Anda di halaman 1dari 281

Magnetic Resonance Imaging in Stroke

Magnetic resonance imaging (MRI) provides non-invasive

information about the brains blood ow, water movement
and biochemical abnormalities following stroke, and
advances in MRI are transforming the investigation and
treatment of cerebrovascular disease. Echoplanar techniques
with diusion- and perfusion-weighted imaging, together
with developments in magnetic resonance spectroscopy and
angiography, are replacing CT scanning as the diagnostic
modality of choice. In this profusely illustrated book, world
leaders in these technologies review the scientic basis and
clinical applications of MRI in stroke. It will appeal to a broad
readership including stroke physicians, neurologists,
neurosurgeons, rehabilitation specialists, and others with a
clinical or research interest in cerebrovascular disease.

Stephen Davis is Professor of Neurology at the University of

Melbourne. He heads the Stroke Research Group and is
Co-Director of the Brain Imaging Laboratory at the Royal
Melbourne Hospital, where he is Director of Neurology.

Marc Fisher is Professor of Neurology at the University of

Massachusetts, and a leading authority on the use of MRI in
the evaluation of stroke therapies.

Steven Warach is Chief of the Section on Stroke Diagnostics

and Therapeutics in the Stroke Branch at NINDS, National
Institutes of Health, Bethesda, Maryland. He pioneered the
use of diusion and perfusion MRI in the evaluation of stroke
and in clinical trials.
Magnetic Resonance
Imaging in Stroke

Edited by

Stephen Davis
University of Australia, Melbourne

Marc Fisher
University of Massachusetts Memorial Medical Care, USA

Steven Warach
National Institutes of Health, Bethesda, MD, USA

Cambridge, New York, Melbourne, Madrid, Cape Town, Singapore, So Paulo

Cambridge University Press

The Edinburgh Building, Cambridge , United Kingdom
Published in the United States of America by Cambridge University Press, New York
Information on this title:

Cambridge University Press 2003

This book is in copyright. Subject to statutory exception and to the provision of

relevant collective licensing agreements, no reproduction of any part may take place
without the written permission of Cambridge University Press.

First published in print format 2003

isbn-13 978-0-511-06377-0 eBook (NetLibrary)
isbn-10 0-511-06377-6 eBook (NetLibrary)

isbn-13 978-0-521-80683-1 hardback
isbn-10 0-521-80683-6 hardback

Cambridge University Press has no responsibility for the persistence or accuracy of

s for external or third-party internet websites referred to in this book, and does not
guarantee that any content on such websites is, or will remain, accurate or appropriate.

Every effort has been made in preparing this book to provide accurate and up-to-date
information which is in accord with accepted standards and practice at the time of
publication. Nevertheless, the authors, editors and publishers can make no warranties
that the information contained herein is totally free from error, not least because
clinical standards are constantly changing through research and regulation. The authors,
editors and publisher therefore disclaim all liability for direct or consequential
damages resulting from the use of material contained in this book. Readers are strongly
advised to pay careful attention to information provided by the manufacturer of any
drugs or equipment that they plan to use.

List of contributors page vii

Preface xiii

1 The importance of specific diagnosis in 1

stroke patient management
John N. Fink and Louis R. Caplan

2 Limitations of current brain imaging

modalities in stroke 15
P. Alan Barber and Stephen M. Davis

3 Clinical efficacy of CT in acute cerebral

ischemia 31
Rdiger von Kummer

4 Computerized tomographic-based 47
evaluation of cerebral blood flow
Lawrence R. Wechsler, Steven Goldstein
and Howard Yonas

5 Technical introduction to MRI 55

Rohit Sood and Michael Moseley

6 Clinical use of standard MRI 69

Brian M. Tress

7 MR angiography of the head and neck: 85

basic principles and clinical applications
Robert R. Edelman and Joel Meyer

8 Stroke MRI in intracranial hemorrhage 103

Peter D. Schellinger, Olav Jansen and
Werner Hacke

9 Using diffusionperfusion MRI in animal 113

models for drug development
Marc Fisher

vi Contents

10 Localization of stroke syndromes using 121 16 New MR techniques to select patients for 207
diffusion-weighted MR imaging (DWI) thrombolysis in acute stroke
Max Wintermark, Marc Reichhart, Vincent N. Thijs and Gregory W. Albers
Reto Meuli and Julien Bogousslavsky
17 MRI as a tool in stroke drug development 223
11 MRI in transient ischemic attacks: clinical 135 Steven Warach
utility and insights into pathophysiology
18 Magnetic resonance spectroscopy in stroke 233
Jerey L. Saver and Chelsea Kidwell
Dawn E. Saunders and Martin M. Brown
12 Perfusion-weighted MRI in stroke 147
19 Functional MRI and stroke 251
William A. Copen and A. Gregory Sorensen
Amy Brodtmann, Leeanne Carey and
13 Perfusion imaging with arterial spin 161 David G. Darby
David C. Alsop and John A. Detre Index 263
Colour gures between pp. 120 and 121.
14 Clinical role of echoplanar MRI in stroke 175
Stephen Davis and Mark Parsons

15 The ischemic penumbra: the evolution of

a concept 191
Georey A. Donnan, Peter M. Wright,
Romesh Markus, Thanh Phan and
David C. Reutens

Stephen M. Davis,
Department of Neurology,
Royal Melbourne Hospital,
Victoria 3050,

Marc Fisher,
Department of Neurology,
Memorial Health Care,
119 Belmont Street,
MA 01605,

Steven Warach,
National Institutes of Health,
10 Center Drive,
MSC 1063, Room B1D733,
MD 29892-1063,

Gregory W. Albers,
Stanford Stroke Center,
Stanford University Medical Center,
Palo Alto,
CA 94394,

viii List of contributors

David C. Alsop, Leeanne Carey,

Department of Radiology, Department of Neurology,
Beth Israel Deaconess Medical Center and Harvard Royal Melbourne Hospital,
Medical School, Parkville,
USA Victoria,
P.A. Barber,
Department of Neurology, William A. Copen,
Royal Melbourne Hospital, Department of Radiology,
University of Melbourne, Massachusetts General Hospital,
Parkville, PO Box 9657,
Victoria 3050, 55 Fruit Street,
Australia Boston,
MA 02114,
Julien Bogousslavsky, USA
Department of Neurology,
University Hospital (CHUV) BH07, David G. Darby,
1011 Lausanne, Department of Neurology,
Switzerland Royal Melbourne Hospital,
Amy Brodtmann, Victoria 3050,
Department of Neurology, Australia
Royal Melbourne Hospital,
Parkville, Stephen M. Davis,
Victoria 3050, Department of Neurology,
Australia Royal Melbourne Hospital,
University of Melbourne,
Martin M. Brown, Parkville,
Stroke Medicine, Victoria 3050,
Institute of Neurology, Australia
University College London,
The National Hospital for Neurology and John A. Detre,
Neurosurgery, Departments of Neurology and Radiology,
Queen Square, University of Pennsylvania Medical Center,
London, 3 W Gates,
UK 3400 Spruce Street,
Louis R. Caplan, PA 191044283,
Department of Neurology, USA
Beth Israel Deaconess Medical Center,
Boston, Georey A. Donnan,
MA, National Stroke Research Institute,
USA Heidelberg,
List of contributors ix

Robert R. Edelman, Rdiger von Kummer,

Department of Radiology, Room 5106 Department of Neuroradiology,
Evanston Hospital, University of Technology,
2650 Ridge Avenue, Fetscherstr. 74,
Evanston, Dresden,
IL 60201, Saxonia D-01307,
USA Germany

John N. Fink, Romesh Markus,

Department of Neurology, National Stroke Research Institute,
Christchurch School of Medicine, Heidelberg,
New Zealand Victoria,
Marc Fisher,
Department of Neurology, Reto Meuli,
UMACS, Department of Diagnostic and Interventional
Memorial Health Care, Radiology,
119 Belmont Street, University Hospital (CHUV) BH07,
Worcester, 1011 Lausanne,
MA 01605, Switzerland
Joel Meyer,
S. Goldstein, Department of Radiology,
University of Pittsburgh Health System, Evanston Northwestern Healthcare,
Stroke Institute, Northwestern University School of Medicine,
Departments of Neurology and Neurosurgery, 2650 Ridge Avenue,
200 Lothrop Street, Evanston,
PA 15213, IL 60201

Werner Hacke, Michael Moseley,

Department of Neurology, Department of Radiology,
University of Heidelberg, 1201 Welch Road,
Germany Stanford University,
CA 94305-5488,
Olav Jansen, USA
Department of Neuroradiology,
University of Kiel, Mark Parsons,
Germany Department of Neurology,
Royal Melbourne Hospital,
Chelsea Kidwell, University of Melbourne,
UCLA Stroke Center, Parkville,
710 Westwood Plaza, Victoria 3050,
Los Angeles, Australia
CA 90095,
x List of contributors

Thanh Phan, A. Gregory Sorensen,

National Stroke Research Institute, Massachusetts General Hospital,
Heidelberg, NMR Center,
Victoria, 149 13th Street,
Australia Charlestown,
MA 02129,
Marc Reichhart, USA
Department of Neurology,
University Hospital (CHUV) BH07, Vincent N. Thijs,
1011 Lausanne, Department of Neurology,
Switzerland U2 Gasthuisberg,
Herestraat 49,
David C. Reutens, 3000 Leuven,
National Stroke Research Institute, Belgium
Victoria, Brian M. Tress,
Australia The University of Melbourne Department of
Dawn E. Saunders, c/o Post Oce,
Department of Neuroradiology, Parkville,
The National Hospital of Neurology and Victoria 3050,
Neurosurgery, Australia
Queen Square,
London, Steven Warach
UK National Institutes of Health,
Jerey L. Saver, 10 Center Drive,
UCLA Stroke Center, MSC 1063, Room B1D 733
710 Westwood Plaza, Bethesda
Los Angeles, MD 298921063,
CA 90095, USA
Lawrence R. Wechler,
Peter D. Schellinger, University of Pittsburgh Health System,
Department of Neurology, Stroke Institute,
University of Heidelberg, Departments of Neurology and Neurosurgery,
Im Neuenheimer Feld 400, 200 Lothrop Street,
D69120 Heidelberg, PA 15213,
Germany USA

Rohit Sood, Max Wintermark,

Department of Radiology, Department of Diagnostic and Interventional
Stanford University, Radiology,
CA 94305-5488, University Hospital (CHUV) BH07,
USA 1011 Lausanne,
List of contributors xi

Peter M. Wright, H. Yonas,

National Stroke Research Institute, University of Pittsburgh Health System,
Heidelberg, Stroke Institute,
Victoria, Departments of Neurology and Neurosurgery,
Australia 200 Lothrop Street,
PA 15213,

Stroke is a leading cause of death in Western coun-

tries, with a mortality rate higher than most forms
of cancer and now the commonest cause of long-
term adult disability. Stroke diagnosis and manage-
ment were revolutionized by the widespread
introduction of computed tomographic (CT) scan-
ning in the 1970s. CT scanning sensitively excludes
cerebral hemorrhage, but early ischemic changes
can be subtle. In the rst few hours after stroke
onset, when acute therapies such as thrombolysis
are being considered, CT is often normal, although
acute ischemic changes have become better recog-
nized in recent years. Conventional magnetic reso-
nance imaging (MRI) became widely available in
most countries a decade after the advent of CT
scanning, but has had a limited role in stroke diag-
nosis and management. Although MRI provides far
better imaging of posterior fossa structures and
facilitated non-invasive angiography (MRA), its
sensitivity in acute stroke is not much better than
CT. Other functional imaging techniques such as
single photon emission computed tomography
(SPECT) and positron emission tomography (PET)
have been valuable research tools, but have not
been of routine clinical use in the management of
Since the 1990s, the increasingly widespread
availability of echoplanar MRI technology facili-
tated the introduction of diusion-weighted
imaging (DWI), perfusion imaging (PWI) and mag-
netic resonance spectroscopy (MRS). Diusion-
weighted imaging allows the hyperacute evaluation
of the ischemic core within minutes of stroke onset
and the distinction between acute and chronic

xiv Preface

ischemic lesions. It represents an extraordinary being tested in randomized trials. A series of

advance in stroke imaging, specically in the region chapters details the diagnostic advances facili-
of ischemic tissue that is usually destined for infarc- tated by MRA, DWI, PWI and MRS. Following a
tion. PWI provides a measure of the hypoperfused review of the pathophysiology and clinical impor-
tissue at risk, particularly in the ischemic penum- tance of the ischemic penumbra, our contributors
bra, where acute therapies are targeted. Currently, illustrate the role of MRI in drug development and
PWI is dependent on contrast injection, but arterial selection of acute therapies. Recent studies
spin labelling may well supersede this technique. provide insights into the use of MRI in individual-
These new MRI methods also permit topographic ization of the time window, providing a tissue
analysis of acute infarcts and some insights into clock for therapeutic interventions such as
stroke pathophysiology and prognosis. Concurrent thrombolysis. Currently, MR-based studies are
MRA allows analysis of acute arterial occlusion and testing the hypothesis that perfusiondiusion
monitors recanalization. Magnetic resonance spec- mismatch, the postulated MR signature of the
troscopy provides insights into metabolically ischemic penumbra, can suggest the benet of
deranged cerebral tissues and provides information thrombolysis beyond the clinically established 3-
that is complementary to DWI and PWI. These new hour time window. Finally, functional brain
techniques are transforming the diagnosis and imaging using brain activation studies and MRI
management of acute stroke. We believe that CT is are leading to a better understanding of brain pro-
likely to be widely replaced by these new MR tech- cessing and brain recovery after stroke.
niques within the next few years. This has already In this book, we have targeted neurologists, other
occurred in many expert stroke centres. stroke physicians, neuroradiologists and other clin-
In this book we have aimed to provide a com- icians involved in stroke diagnosis, imaging and
prehensive and up-to-date summary of the dra- management. We have aimed to encapsulate the
matic developments that have occurred in this development, current and emerging clinical role of
eld in the last few years and have also tried to MRI in stroke. We are grateful for the contributions
predict likely advances. The scope of the text of our chapter authors, all leaders in the eld of
includes background on the importance of MRI and stroke. A few years ago, experts debated
precise stroke diagnosis, the current uses of CT whether MRI, in acute stroke diagnosis, was ready
including perfusion imaging and an introduction for prime time. After reading this book, we suspect
to standard and echoplanar MRI techniques. you will agree that it is.
Recent advances in MRI permit exclusion of
intracerebral hemorrhage and this is currently Stephen Davis, Marc Fisher and Steven Warach, 2002
The importance of specific diagnosis in stroke patient

John N. Fink1 and Louis R. Caplan2

Department of Neurology, Christchurch School of Medicine, New Zealand
Department of Neurology Beth Israel Deaconess Medical Center, Boston, MA, USA

Introduction: A stroke is not a stroke cians should manage patients according to

methods that have been tested by well-designed
Stroke cannot be considered a diagnosis in itself. randomized controlled trials. Unfortunately, few
Stroke refers to any damage to the brain or spinal therapies for patients with stroke have been tested
cord caused by a vascular abnormality, the term gen- with randomized trials, and even fewer have been
erally being reserved for when symptoms begin thoroughly investigated for patients with specic
abruptly. Stroke is anything but a homogeneous stroke subtypes.
entity, encompassing disorders as dierent as Randomized trials have limitations, including the
rupture of a large blood vessel that causes ooding issue of numbers v. specicity, or lumping v. split-
of the subarachnoid space with blood, the occlusion ting. To provide statistically valid results, random-
of a tiny artery supplying a small but strategic brain ized trials must contain large numbers of patients
site and thrombosis of a venous conduit obstructing with enough end points to analyse within a rela-
outow of blood from the brain. Each stroke subtype tively short period; therefore lumping must pre-
carries with it dierent implications for acute treat- dominate over splitting. But, if the results are to be
ment, prognosis and secondary prevention. Each useful for clinical practice, the data must be speci-
stroke patient has additional variables that inuence cally applicable to individual patients. Too often,
management, including the time from onset to pres- there are signicant obstacles to doing this.
entation, the severity of the lesion, and associated Investigators have continued to design trials as if
comorbidities as well as social and psychological they expect a single treatment to be eective for all
factors. The availability of non-invasive imaging ischemic stroke patients, resulting in inevitable dis-
techniques has revolutionized the diagnostic appointment. Even when treatments have been
process, enabling a much greater understanding of found eective, there is still a great deal of room for
the relevant pathophysiological processes active in improvement. For example, aspirin has been
the individual patient. This chapter provides an proven to be eective for early secondary preven-
overview of how the specic diagnostic information tion of stroke generally, but only prevents 25% of
available from non-invasive investigations can be recurrent strokes within 14 days.1,2 Cost contain-
applied to the management of individual patients. ment and the need to involve a large number of
centres with varying expertise and resources in
trials results in a minimum of patient investigation.
Lumping vs. splitting As a result, accurate subgroup comparisons in trials
become impossible, even when these are reported
The goal of every clinician is to provide the best in a post hoc analysis. Patients who are too ill, old,
care for his or her patients. Where possible, physi- young, or of child-bearing age are often excluded

2 John N. Fink and Louis R. Caplan

from trials. Those unable to give informed consent Table 1.1. Stroke classification
or who have too complex, or multiple, illnesses are
also frequently left out. The type of patients that are (a) Clinical stroke classication systems
excluded from these trials are those that doctors are Traditional
Transient ischemic attack (TIA)
called on to care for every day.
Minor stroke
The term evidence-based must be used cau-
Reversible ischemic neurologic decit (RIND)
tiously when applied to a particular circumstance if
Stroke in progress
that circumstance has not been specically studied. Completed stroke
Information from trials must be weighed according (a) Oxfordshire Community Stroke Project6
to the context of specic treatment decisions for Total anterior cerebral infarction syndrome (TACI)
individual patients. George Thibault said it well:3 Partial anterior cerebral infarction syndrome (PACI)
Lacunar infarction syndrome (LACI)
We then need to decide which approach in our large thera-
Posterior cerebral infarction syndrome (POCI)
peutic armamentarium will be most appropriate in a particu-
(b) Etiologic classication systems
lar patient, with a particular stage of diseases and particular
coexisting conditions, and at a particular age. Even when ran-
Large artery
domized clinical trials have been performed (which is true
for only a small minority of clinical problems), they will often
Small vessel
not answer this question specically for the patient sitting in
Other determined etiology
front of us in the oce or lying in the hospital bed.
Undetermined etiology
The complexity of managing stroke patients is (a) Baltimore-Washington8
increasing. Improvements in diagnostic accuracy Atherosclerotic vasculopathy
have raised new questions about the correct appli- Non-atherosclerotic vasculopathy
cation of existing treatments. There have been Vasculopathy of uncertain cause (lacunar infarct)
Cardiac/transcardiac embolism
many new developments in stroke therapeutics,
including intravenous and intra-arterial thromboly-
Migrainous stroke
sis, catheter-based interventions such as angio-
Oral contraceptive or exogenous estrogen use
plasty and stenting for both extracranial and Other drug related
intracranial stenoses, the development of new anti- Indeterminate
platelet agents with potentially complimentary
mechanisms of action, and hypothermic treatment,
to name a few. The exact place for all of these thera- and how we diagnose stroke. Early stroke classica-
pies is not established, yet it is extremely likely that tions relied on clinical information. Terms such as
many of the new treatments that are currently transient ischemic attack (TIA), minor stroke,
unproven will be able to deliver improved out- reversible ischemic neurologic decit (RIND),
comes for carefully selected patients. Ignoring stroke in progress and completed stroke were
these new diagnostic and therapeutic develop- used to distinguish stroke subtypes.4 These simplis-
ments is not an option, although a conservative tic distinctions now have little clinical usefulness.
approach must be taken when potentially hazard- Even the term TIA is becoming obsolete as smaller
ous therapies have not been rigorously tested. A infarctions have become detectable with magnetic
specic diagnosis is required to optimize treatment resonance imaging (MRI).5
selection. Subsequent classications have increasingly
focused on stroke etiology, because of its impor-
tance in determining treatment strategies for sec-
Advances in imaging and stroke diagnosis ondary prevention of stroke (Table 1.1). This has
required an increasing emphasis on the results of
Advances in imaging have led to dramatic changes imaging investigations, rather than clinical features.
in our understanding of stroke pathophysiology The authors of the Trial of Org 10172 in Acute
The importance of specific diagnosis in stroke patient management 3

Stroke Treatment (TOAST) classied strokes as lar system. Brain MRI examinations for stroke
being due to large artery atherosclerosis, cardioem- should routinely include magnetic resonance
bolism, small vessel occlusion, other determined angiography (MRA) of the intracranial vasculature.
etiology or undetermined etiology.7 This system of Magnetic resonance venography (MRV) and MRA
stroke classication represents an important of the cervical carotid and vertebral arteries can
advance, but still has shortcomings that limit its easily be performed at the same sitting as brain
application to the diagnosis and management of imaging, without the need for contrast. Assessment
individual patients. One major limitation is the of the aortic arch and proximal vessels is possible
oversimplied large artery classication. This cat- with gadolinium-enhanced MRA. Diusion-
egory lumps embolic strokes from sources in the weighted MR imaging (DWI) and perfusion imaging
aorta, large vessel origins in the thorax, cervical (PI) enable determination in real time of the pres-
arterial lesions, and intracranial arterial stenoses ence and severity of an ischemic decit and the
with strokes due to thrombotic occlusion of cervical response of the brain to the insult. These new tech-
or intracranial vessels of either anterior or posterior niques enable the concept of stroke diagnosis to go
circulations. beyond that of simple stroke etiology to establish a
Stroke subtype classications used today, such as comprehensive and dynamic model of stroke
the TOAST system, reect the type of stroke pathophysiology for individual patients.
imaging techniques that were generally available a
decade ago, namely non-contrast computed
tomography (CT) head scan and ultrasound exam- Initial stroke diagnosis
inations of the cervical carotid arteries and of the
heart. A diagnostic strategy that continues to rely
Stroke or stroke-mimic
solely on these modalities will not achieve a more
accurate diagnosis. Not only are important parts of The initial diagnostic step should be to determine if
the vascular system overlooked entirely by such an the event is due to stroke or a non-vascular stroke
approach, but the accuracy of even these simple mimic. Clinical information remains very impor-
classications is often poor.9 A lacunar stroke tant in distinguishing disorders such as migraine,
cannot be reliably diagnosed on the basis of clinical seizure, and factitious and psychogenic disorders
and acute CT ndings.10 Some patients with a from stroke. Sometimes the diagnosis is relatively
lacunar syndrome have multiple acute lesions on clear, but when this is not the case, imaging results
diusion-weighted MRI, consistent with an are critical. A typical appearance on a CT scan will
embolic etiology.11 Moreover, diagnosis using the often conrm the diagnosis of stroke; however,
traditional approach is not made in real time, but false-negative CT ndings are common in the acute
retrospectively. A subacute CT scan is required if phase, particularly if image quality is poor, readers
the diagnosis of lacunar infarction is to be con- are inexperienced or if the patient presents with
rmed and a cortical lesion excluded. Ultrasound lacunar or brainstem stroke.10,1214 Diusion-
tests may be obtained days after the initial presen- weighted MRI is extremely sensitive to acute brain
tation. This is a critical limitation that prevents a ischemia and false-negatives are very rare, with the
specic diagnosis prior to consideration of acute exception of small brainstem lacunes.12,15 DWI is
stroke therapies that can only be overcome if other therefore the diagnostic modality of choice when
protocols for acute imaging and assessment are the diagnosis of stroke is uncertain and positive evi-
used.9 dence of a stroke is required. The importance of an
Newer imaging techniques that allow rapid, non- accurate diagnosis even at this level should not be
invasive assessment of a much greater extent of the underestimated; as many as 20% of initial stroke
vascular system are now widely available. MRI, as diagnoses are erroneous,16 and some patients with
this book demonstrates, is an extremely powerful stroke mimic have been treated with thrombolysis
technique for imaging the brain and cerebrovascu- as a result.17
4 John N. Fink and Louis R. Caplan

management decisions are made. MRI is very sensi-

Arterial occlusion, arterial rupture, or venous
tive for the initial detection of AVM and other vas-
cular abnormalities in the brain, including
The next level of stroke diagnosis is primarily to dis- cavernous angiomata, which are often undetect-
tinguish hemorrhagic from ischemic stroke. able with DSA. MRA has a role in follow-up of any
However, conceptualizing the mechanism and its untreated lesions and screening of high-risk
vascular pathology ensures that stroke due to families.2527
venous thrombosis is not overlooked. The majority
of the remainder of the chapter considers diagnosis
of ischemic stroke; cerebral venous thrombosis and Specific diagnosis and management of
hemorrhagic stroke are considered briey below. ischemic stroke

Cerebral venous thrombosis

Acute stroke
Although cerebral venous thrombosis (CVT) is rare
in comparison with other stroke types, it is treat- The NINDS trial28 established the eectiveness of
able and the diagnosis is frequently missed on CT intravenous tissue plasminogen activator (tPA) for
scan. MRI is very sensitive in the detection of CVT;18 acute ischemic stroke within 3 hours of symptom
however, the diagnosis can be overlooked if it is not onset. While this is a major advance in stroke
considered in the dierential diagnosis, or if sus- treatment, the advancement must not stop there.
ceptibility-weighted (T2*) imaging or MR venogra- Thrombolysis according to the NINDS protocol
phy is not specically requested. Patients adds one favourable outcome for every 13 patients
presenting with what appear to be lobar hemor- treated, while causing harm to one in 17.28 The
rhages on CT (young patients with temporal lobe NINDS trial and other negative multi-centre trials
hemorrhage especially) are particularly at risk of of intravenous thrombolysis2933 relied on a CT
being misdiagnosed and mismanaged before the scan and a clock to characterize their patients
correct diagnosis is made.19 before treatment decisions were made. This was
appropriate at the time as other rapid methods of
Hemorrhagic Stroke more detailed assessment were not generally avail-
CT scanning has generally been considered the able, but inevitably resulted in some patients
investigation of choice for identication of intracra- being exposed to the risk of treatment without
nial blood; however MRI protocols including T2* hope of benet, such as patients whose vessels
imaging are now able to reliably detect acute cere- have spontaneously recanalized,34 or those with
bral hemorrhage, and are far superior to CT in the little salvageable brain tissue within the hypoper-
detection of subacute and chronic hemorrhage.2023 fused region. At the same time, some patients who
The sensitivity of T2* and FLAIR MRI for the detec- might benet beyond 3 hours were denied treat-
tion of acute subarachnoid hemorrhage is compar- ment on the basis of time alone, not individual
able with CT.24 pathophysiological features.35 The PROACT trials
Clinicians are already familiar with the need to have subsequently demonstrated the potential of
make a specic diagnosis of the cause of hemor- intra-arterial thrombolytic agents.36,37 Occlusions
rhage, when it is detected. The development of of the internal carotid artery and proximal middle
catheter-based interventions for treatment of aneu- cerebral artery do not respond as well as more
rysms and arterio-venous malformations (AVM) distal occlusions to intravenous thrombolysis, and
has meant that an even more detailed characteriza- may be better treated via the intra-arterial
tion of the size, morphology and anatomic location route.3739 Determining the appropriate applica-
of these lesions is required to determine the appro- tions for these potentially hazardous therapies
priate therapeutic approach. Digital subtraction requires a specic diagnosis.
angiography (DSA) is generally required before nal More advanced, rapid, non-invasive imaging
The importance of specific diagnosis in stroke patient management 5

techniques for assessment of acute ischemia are bility images, and additional CT scanning is not
increasingly available. Multimodal stroke MRI required before administering acute treatments.2022
protocols that include diusion-weighted imaging, Patients with multifocal small chronic hemorrhagic
perfusion imaging, MRA and susceptibility- lesions due to presumed amyloid angiopathy can
weighted imaging can be performed rapidly, also be identied, who may be at increased risk of
exclude brain hemorrhage, dene areas of hypoper- hemorrhage if thrombolytic agents are given.23,43
fusion and tissue damage and identify occluded The scanning time of an acute imaging protocol is
arteries, enabling decisions about thrombolysis to less than 15 minutes. In the last 4 years at our insti-
be made according to individual pathophysiologi- tution, we have performed perfusion studies in over
cal criteria.34,35 DWI and MRA can enable an 300 acute stroke patients and have treated 29 acute
unequivocal diagnosis of acute stroke to allow stroke patients with t-PA on the basis of MRI results
stroke patients who might be excluded from throm- alone.
bolysis on CT-based criteria to be treated, such as
patients presenting with seizure at stroke onset,
Stroke etiology and secondary prevention
hypoglycemia or hyperglycemia. Denition of the
ischemic penumbra with diusion and perfusion A detailed diagnosis of stroke etiology is required to
MRI may allow expansion of the therapeutic plan management strategies for secondary stroke
window beyond the current 3-hour guideline for prevention. This requires identication of the loca-
selected patients.35 Parameters are being estab- tion and nature of the vascular lesions responsible,
lished to identify those with an unacceptably high identication of systemic stroke risk factors and
risk of hemorrhage due to the severity of the consideration of the likely pathophysiological
ischemic damage present at the infarct core.40 mechanism of stroke. The elements of specic diag-
Continued renements in MR perfusion imaging nosis of ischemic stroke are summarized in Table
techniques promise to allow more accurate predic- 1.2.
tions of the volume of brain tissue that is at risk of
infarction if reperfusion does not occur, based on Diagnosis of vascular lesions
perfusion thresholds.41 All levels of the vascular supply to the brain should
In addition to enabling more specic application be considered when determining stroke etiology,
of thrombolytic therapies, physicians can use that is: the heart, aorta, proximal carotid or verte-
detailed knowledge of their patients pathophysio- bral arteries in the thoracic cavity, cervical carotid
logy to select candidates for other acute stroke ther- and vertebral arteries and intracranial vessels. Not
apies. In particular, patients who are not candidates only must the anatomical location of vascular
for t-PA but who have a persistent vascular occlu- lesions be determined, but knowledge of the nature
sion and a signicant volume of brain at risk of and severity of lesions is required, also. MRA can
infarction due to tenuous collateral supply may provide a comprehensive assessment of the vascu-
benet from hypertensive therapy to improve col- lar tree to determine the location and severity of
lateral circulation.42 vascular lesions. MRI with MRA is the non-invasive
investigation of choice for the diagnosis and follow-
Practical application of acute stroke MRI up of carotid and vertebral artery dissection.44
Multimodal stroke MRI has been in use for several Duplex ultrasound remains more established than
years in institutions in many countries, including MRA for assessment of cervical internal carotid
our own hospital. The hardware and software artery lesions, but promising results are being
required are increasingly available. Stroke fellows shown with contrast-enhanced MRA,45 and verte-
can be trained to perform the studies enabling 24- bral artery assessment is superior with MRA. MRA
hour coverage independent of technician rosters. has great promise in the evaluation of aortic
Acute hemorrhagic stroke can be accurately iden- lesions; 46 it is possible that in the future, MRI of the
tied using an MRI protocol that includes suscepti- heart and great vessels will reduce the need for the
6 John N. Fink and Louis R. Caplan

Table 1.2. Approach to ischemic stroke diagnosis more invasive procedure of transesophageal echo-
cardiography. Transcranial Doppler
1. Initial ischemic stroke diagnosis: Ultrasonography is a useful method of assessing
(a) stroke vs. non-vascular stroke mimic the major intracranial vessels but MRA or CTA oer
(b) ischemic stroke vs. hemorrhagic stroke vs. venous
the convenience of being performed at the same
time as brain imaging. Digital subtraction angiogra-
2. Acute stroke pathophysiology
phy is still required when intravascular interven-
(a) severity and extent of ischemic brain injury
(b) persistence and severity of cerebral hypoperfusion tions are contemplated, on occasion to distinguish
(c) identication of vascular occlusive lesion between critical stenosis and occlusion of the inter-
3. Stroke etiology: vascular lesion nal carotid artery, and to conrm the diagnosis of
(a) location of vascular lesion(s) certain non-atherosclerotic vasculopathies, such as
e.g. cardiac, aorta, vascular origins, cervical bromuscular dysplasia, inammatory and infec-
vessels, intracranial vessels tious arteritides, drug abuse-associated vasculopa-
(b) nature of vascular lesion(s) thy, and radiation-induced stenosis.
e.g. cardiac: thrombus, AF, valvular, PFO,
akinesis, endocarditis, other Specic vascular diagnosis and management
vascular: atherosclerosis severity, ulceration,
other high risk features
Cardiac-origin embolism
other lesions dissection,
A full discussion of the diagnosis and management
vasospasm, bromuscular
dysplasia, arteritis, drug- of cardiac-origin embolism is beyond the scope of
associated vasculopathy this chapter and is available elsewhere.47 Secondary
4. Systemic stroke risk factors prevention strategies can include anticoagulants,
(a) traditional risk factor identication: hypertension, antiplatelet agents or their combination, antibio-
smoking, diabetes, hyperlipidemia. tics, antiarrhythmics and cardioversion, pace-
(b) thrombophilia maker, surgery, or catheter-based interventions.
acquired: antiphospholipid syndrome, Therapeutic decisions depend on a specic diagno-
polycythemia, thrombocytosis, sis of the structural lesions involved and the likely
hyperbrinogenemia, other composition of the embolic particle itself.47,48
inherited: protein C, S, ATIII deciency,
prothrombin mutation
Lesions of the aorta and great vessels
(c) other, e.g. hyperhomocysteinemia
That the aorta is an important source of brain
5. Stroke mechanism
(a) embolic stroke embolism is now well established.49,50 The embolic
(b) in situ thrombosis risk is greatest for thick, complex and mobile
(c) lacunar infarction plaques.51,52 Gadolinium-enhanced MRA can estab-
(d) hemodynamic / watershed stroke lish this diagnosis quickly and accurately.46 The best
(e) vasospasm treatment to prevent embolism from aortic lesions
6. Stroke severity is not yet known. Cases have been reported where
(a) clinical features, e.g. NIH Stroke Scale Score aortic thrombotic masses have disappeared after
(b) lesion volume / location anticoagulant therapy.53,54 Intravenous thrombo-
7. Patient factors lytic treatment 55 and surgical removal of protrud-
(a) premorbid functioning, age
ing atheromas56 have also been reported to be
(b) comorbidities
successful in treating patients with aortic athero-
(c) psychological, social and economic factors
Atheromatous disease of the origins of the verte-
bral arteries is a common, yet often overlooked
source of posterior circulation TIA and stroke.57
Antiplatelet agents or anticoagulants are generally
The importance of specific diagnosis in stroke patient management 7

the rst line of treatment, but angioplasty and ecial for carefully selected patients with poor
stenting of such lesions may sometimes be appro- untreated prognosis refractory to medical therapy.77
Pathophysiological stroke diagnosis
Cervical vascular lesions Vascular imaging studies dene the structural
Carotid endarterectomy is well-established for the lesions important in stroke etiology, but may not
treatment of symptomatic severe (7099%) internal show whether the stroke was due to thrombotic,
carotid artery (ICA) stenosis.59,60 The benet of embolic or hemodynamic mechanism, and do not
endarterectomy for symptomatic moderate inform about the nature of the embolic material
(5069%) stenosis is more modest and decisions itself. Some stroke subtypes, such as migrainous
about treatment must take individual and surgeon stroke, may not be associated with a structural vas-
characteristics into account.60,61 The benetrisk cular lesion. Clinical information must be com-
ratio for carotid endarterectomy for unselected bined with imaging data to achieve a specic
patients with asymptomatic ICA lesions is even diagnosis and tailor management for the individual
lower62 and treatment decisions must be individu- patient.
alized.63,64 A signicant increase in severity of sten-
osis increases stroke risk and favours surgery.65 Thrombosis and embolism
Identication of individual patients with higher Our understanding of stroke pathophysiology has
stroke risk who would benet most from surgical changed dramatically during recent years, empha-
treatment may be possible using TCD microembo- sizing the importance of embolism in stroke patho-
lus detection,66 or possibly platelet scintigraphy67 or genesis.47,48 The majority of non-lacunar ischemic
indicators of cerebral perfusion or vascular strokes are likely to be embolic in origin. Secondary
reserve68,69 including MRI perfusion techniques;70 prevention strategy depends on identication of
however, more studies are still required.71 the donor source, risk factors, and consideration of
Certain cervical carotid artery lesions may be the likely nature of the embolic particle itself.47,48
better treated with intravascular interventions than Thrombosis is likely when complete ICA occlu-
traditional endarterectomy. Careful patient selec- sion is found, although even then embolism from
tion is required; indications might include high cer- the distal ICA thrombus may be the nal stroke
vical lesions with dicult surgical access, mechanism, and embolism from the heart may
radiation-induced stenosis, postsurgical restenosis, have caused the ICA occlusion.7880 Thrombosis
bromuscular dysplasia and patients with high sur- may be an important mechanism when intracranial
gical risk due to severe medical comorbidity. A ran- vascular stenoses are present. It may be dicult to
domized controlled trial of carotid stenting and know in some cases whether intracranial stenoses
endarterectomy is planned.72 detected in the subacute period represent chronic
Dissection of the internal carotid artery generally lesions or partial recanalization of an embolus.
does not require surgical intervention, even when Repeat imaging at a chronic time point with MRA,
aneurysms are associated;73 however, patients may CTA or TCD may be required.
benet from a period of anticoagulation.
Hemodynamic stroke
Intracranial stenoses The importance of hemodynamic factors as sole
The identication of intracranial stenoses can have mechanism in stroke etiology has also been over-
important prognostic implications.74,75 Whether emphasized in the past. Many strokes that may pre-
anticoagulation is more appropriate treatment than viously have been considered hemodynamic,
aspirin for patients with intracranial disease is cur- particularly posterior borderzone infarctions are
rently the subject of a multicentre randomized con- likely to be caused by embolism.8184 However,
trolled trial.76 Intracranial angioplasty and stenting, impaired regional blood ow due to severe vascular
in the hands of experienced operators, may be ben- stenosis or occlusion is likely to contribute to the
8 John N. Fink and Louis R. Caplan

pathogenesis of embolic stroke. A small embolic patient with apparently well-controlled blood-
vascular occlusion is more likely to result in infarc- pressure should prompt consideration of intensi-
tion when insucient collateral circulation is cation of treatment. Ambulatory blood pressure
present, and low ow may impair clearance of recording may be very helpful to optimize the man-
emboli, or washout.85 Possibly the most reliable agement in individual patients. The role of inten-
marker for hemodynamic infarction is the topo- sive lipid-lowering therapy in secondary stroke
graphic pattern of infarction seen on acute multi- prevention is currently the subject of a large ran-
modal MRI, including DWI, MRA and perfusion domized trial.
imaging, along with the appropriate clinical setting. Young patients and those without major vascu-
Multiple small acute lesions are seen in widespread lar risk factors for stroke should also be tested for
distribution within the internal borderzone region, hereditary and acquired thrombophilic states
in the absence of a vascular occlusion.86 Evidence (Table 1.2), the discovery of which can lead to
of a hemodynamic cause for stroke warrants con- modication of treatments prescribed, such as the
sideration of reduction of antihypertensive medica- use of higher intensity anticoagulation for patients
tions or other measures to raise blood pressure, as with the antiphospholipid antibody syndrome,94
well as consideration of revascularization proce- or the introduction of additional treatments such
dures. Magnetic resonance perfusion and other as venesection for polycythemia, folate supple-
methods of brain perfusion imaging such as SPECT, mentation for hyperhomocysteinemia and use of
or TCD assessment of vascular reserve may agents such as eicosapentanoic acid (sh-oil) to
provide helpful information in the management of reduce brinogen levels in hyperbrinogene-
these patients in the future.70,71 mia.95,96

Migraine and vasoconstriction Multiple possible causes of stroke

The diagnosis of migrainous stroke remains pri- Some individuals presenting with stroke may have
marily clinical. Infarction can be caused by pro- more than one potential cause identied. The
longed intense vasoconstriction87,88 either directly major risk factors that predispose to atherosclero-
as a result of impeded blood ow or due to secon- sis, such as hypertension, cigarette smoking, dia-
dary thrombosis.88,89 Treatment for secondary pre- betes and hypertension promote plaque formation
vention of migrainous infarction should include a and occlusive disease in the coronary arteries, aorta
migraine prophylactic agent as well as antiplatelet and peripheral vasculature, as well as the cranio-
therapy. We have most often used verapamil in this cervical arteries.97 Hypertension also predisposes
setting. the penetrating arteries of the brain to lipohyalino-
sis and atheromatous branch disease. The true fre-
Systemic stroke risk factors quency of multiple potential causes for stroke in
Modiable systemic stroke risk factors must be the stroke population is not known. Variation in
incorporated into the patients diagnosis. reported frequencies in the literature have been
Management must be individualized. Risk factors due to denitions used, such as the degree of
are not simply present or absent, but there is a con- carotid artery stenosis required before it is consid-
tinuum of increasing stroke risk with higher blood ered an etiological candidate, and how thoroughly
pressure and cholesterol levels.9093 The actual the patients in each series were investigated.
stroke risk may dier between individuals with the Improvements in diagnostic techniques will inevi-
same blood pressure recordings, depending on how tably result in more such patients being identied.
accurately recordings in the oce reect true daily Understanding of the activity of all of these pro-
levels, the duration of hypertension and the pres- cesses in the individual patient is important. Data
ence of additional risk factors. Evidence of signi- from the Lausanne Stroke Registry indicated that 46
cant end-organ damage such as extensive cerebral (38%) out of 121 recurrent strokes had a dierent
white-matter disease on CT or MRI scanning in a etiology than the initial index stroke.98 In asympto-
The importance of specific diagnosis in stroke patient management 9

matic carotid stenosis, 40% of strokes observed medical and surgical treatments as well as younger
ipsilateral to severe carotid lesions were attribut- patients, nor do they rehabilitate as well from the
able to cardioembolic or lacunar etiologies.99 In eects of a stroke. Secondary prevention studies
addition, the coexistence of coronary artery disease have demonstrated that the absolute benet of
in patients presenting with stroke should not be treatments such as antihypertensive medication
overlooked. Patients who survive ischemic stroke and carotid endarterectomy may be greater for
face a similar risk of death from future myocardial elderly patients.60,107 Socioeconomic and psycho-
infarction to that from recurrent ischemic stroke.100 logical factors may inuence treatment decisions
for some patients and their families.

Stroke severity

Stroke severity is an important diagnostic consider- Specific diagnosis and stroke patient
ation in determining stroke prognosis, which in management
turn inuences management decisions. Clinical
features, which can be quantied using clinical The implications of the enormous heterogeneity of
scales such as the NIH Stroke Scale generally stroke and stroke patients for patient management
provide the most important prognostic informa- should be obvious. Patients should be regarded as
tion.101 Early ischemic lesion volume detected with individuals and modern non-invasive imaging tech-
DWI is also an independent predictor of stroke niques should be used to obtain a specic diagnosis
outcome.102 Imaging studies can be particularly of stroke pathophysiology for each, in order to
important for prognosis in specic cases. The use of ensure optimal management. Acute stroke therapy
diusion and perfusion MR imaging techniques should be oered when possible, preferably on the
and MRA in determining the prognosis of patients basis of pathophysiological, rather than arbitrary,
presenting with acute stroke has already been dis- criteria. All patients deserve assessment of potential
cussed. Detection of a large infarction involving the risk factors, such as hypertension, diabetes,
entire middle cerebral artery territory in a younger smoking, lifestyle, etc, and appropriate modica-
stroke patient is associated with a high risk of tions should be instituted. The mechanism of stroke
malignant cerebral edema.103 Large infarctions or must be considered. Patients with atherosclerosis
hemorrhages in the posterior fossa may also be who have had evidence of cerebral ischemia should
associated with the development of raised intracra- have an evaluation of their heart, coronary arteries,
nial pressure.104 Recognition of these patterns aorta and extracranial and intracranial arteries.
allows early discussion of treatment options that When atherosclerosis is not the cause, a careful
may include hematoma excision, hypothermia105 search for a specic alternative diagnosis must be
and hemicraniectomy.106 made. Therapeutic strategies should then be insti-
tuted for each of the potential risks found and clini-
cians should carefully weigh the riskbenet ratio of
Patient variables
each strategy based on the totality of their knowl-
Individualized stroke management requires consid- edge of that individual patient. Some treatments
eration of the whole individual. Even once a spe- such as antiplatelet medications or standard antico-
cic pathophysiological diagnosis of stroke is agulants might be eective against more than one of
achieved, other variables peculiar to that individual the lesions found, while other treatments such as
patient must also be considered before planning carotid endarterectomy or intracranial angioplasty
management. Pre-existent or coexistent illness may are eective only for the lesions treated. Some treat-
limit or aect treatment. The patients premorbid ments that should benet one lesion (e.g. coronary
function is also an important consideration. Age is artery bypass grafting) might pose a risk for patients
never an absolute contraindication to stroke with other lesions such as severe extracranial and
therapy, however elderly patients do not tolerate intracranial occlusive disease.
10 John N. Fink and Louis R. Caplan

Treatment decisions should be based on infor- 11 Ay H, Oliveira-Filho J, Buonanno FS et al. Diusion-

mation gained about the individual patient, not on weighted imaging identies a subset of lacunar
assumptions or arbitrary criteria. The availability of infarction associated with embolic source. Stroke
powerful non-invasive imaging technology makes 1999; 30: 264450.
thorough evaluation of each patient a realistic 12 Lansberg MG, Albers GW, Beaulieu C, Marks MP.
Comparison of diusion-weighted MRI and CT in
expectation. The increasing use of these techniques
acute stroke. Neurology 2000; 54: 15571561.
in randomized controlled trials will provide
13 Schriger DL, Kalafut M, Starkman S, Krueger M,
answers to important unresolved questions in
Saver JL. Cranial computed tomography interpreta-
stroke therapeutics and continue to improve out- tion in acute stroke: physician accuracy in determin-
comes for our patients. ing eligibility for thrombolytic therapy. J Am Med
Assoc 1998; 279: 12931297.
14 Beauchamp NJ, Barker PB, Wang PY, vanZijl PC.
Imaging of acute cerebral ischemia. Radiology 1999;
REFERENCES 212: 307324.
15 Ay H, Buonanno FS, Rordorf G et al. Normal
1 International Stroke Trial Collaborative Group. The diusion-weighted MRI during stroke-like decits.
International Stroke Trial (IST): a randomised trial of Neurology 1999; 52: 17841792.
aspirin, subcutaneous heparin, both, or neither 16 Libman RB, Wirkowski E, Alvir J, Rao TH. Conditions
among 19435 patients with acute ischaemic stroke. that mimic stroke in the emergency department:
Lancet 1997; 349: 15691581. implications for acute stroke treatment trials. Arch
2 CAST (Chinese Acute Stroke Trial) Collaborative Neurol 1995; 52: 11191122.
Group. CAST: randomized placebo-controlled trial of 17 Katzan IL, Furlan AJ, Lloyd LE et al. Use of tissue-
early aspirin use in 20 000 patients with acute type plasminogen activator for acute ischemic
ischaemic stroke. Lancet 1997; 349: 16411649. stroke: the Cleveland area experience. J Am Med
3 Thibault GE. Clinical problem solving: too old for Assoc 2000; 283: 11511158.
what? N Engl J Med 1993; 328: 946950. 18 Dormont D, Anxionnat R, Evrard S, Louaille C, Chiras
4 Caplan LR. Are terms such as completed stroke or RIND J, Marsault C. MRI in cerebral venous thrombosis.
of continued usefulness? Stroke 1983; 14: 431433. Am J Neuroradiol 1994; 21: 8199.
5 Kidwell CS, Alger JR, Di Salle F et al. Diusion MRI in 19 Fink JN, McAuley DL. Cerebral venous sinus throm-
patients with transient ischemic attacks. Stroke 1999; bosis a diagnostic challenge. Intern Med J 2001; 31:
30: 11741180. 38490.
6 Bamford J, Sandercock P, Dennis M, Burn J, Warlow 20 Linfante I, Llinas RH, Caplan LR, Warach S. MRI fea-
C. Classication and natural history of clinically tures of intracerebral hemorrhage within 2 hours
identiable subtypes of cerebral infarction. Lancet from symptom onset. Stroke 1999; 30: 22632267.
1991; 337: 15211526. 21 Patel MR, Edelman RR, Warach S. Detection of
7 Adams HP, Bendixen BH, Kappelle LJ et al. hyperacute primary intraparenchymal hemorrhage
Classication of subtype of acute ischemic stroke. by magnetic resonance imaging. Stroke 1996; 27:
Denitions for use in a multicenter clinical trial. 23212324.
TOAST. Trial of Org 10172 in Acute Ischemic Stroke. 22 Schellinger PD, Jansen O, Fiebach JB, Hacke W, Sartor
Stroke 1993; 24: 3541. K. A standardized MRI stroke protocol: comparison
8 Johnson CJ, Kittner SJ, McCarter RJ et al. Interrater with CT in hyperacute intracerebral hemorrhage.
reliability of an etiologic classication of ischemic Stroke 1999; 30: 765768.
stroke. Stroke 1995; 26: 4651. 23 Roob G, Schmidt R, Kapeller P, Lechner A, Hartung
9 Lee LJ, Kidwell CS, Alger J, Starkman S, Saver JL. HP, Fazekas F. MRI evidence of past cerebral micro-
Impact on stroke subtype diagnosis of early diusion- bleeds in a healthy elderly population. Neurology
weighted magnetic resonance imaging and magnetic 1999; 52: 991994.
resonance angiography. Stroke 2000; 31: 10811089. 24 Mitchell P, Wilkinson ID, Hoggard N et al. Detection
10 Toni D, Iweins F, von Kummer R et al. Identication of subarachnoid haemorrhage with magnetic reso-
of lacunar infarcts before thrombolysis in the ECASS nance imaging. J Neurol Neurosurg Psychiatry 2001;
I study. Neurology 2000; 54: 684688. 70: 205211.
The importance of specific diagnosis in stroke patient management 11

25 Ross JS, Masaryk TJ, Modic MT, Ruggieri PM, Haacke delivery in acute middle cerebral artery stroke. Stroke
EM, Selman WR. Intracranial aneurysms: evaluation 1998; 29: 411.
by MR angiography. Am J Neuroradiol 1990; 11: 37 Furlan A, Higashida R, Wechsler L et al. for the
449455. PROACT Investigators. Intra-arterial prourokinase
26 Ronkainen A, Hernesniemi J, Puranen M et al. for acute ischemic stroke. The PROACT II study: a
Familial intracranial aneurysms. Lancet 1997; 349: randomized controlled trial. J Am Med Assoc 1999;
380384. 282: 20032011.
27 Crawley F, Clifton A, Brown MM. Should we screen 38 Pessin MS, del Zoppo GJ, Furlan AJ. Thrombolytic
for familial intracranial aneurysm? Stroke 1999; 30: treatment in acute stroke: review and update of
312316. selective topics. In: Moskowitz MA, Caplan LR, eds.
28 The National Institute of Neurological Disorders and Cerebrovascular Diseases: Nineteenth Princeton
Stroke rt-PA Stroke Study Group. Tissue plasminogen Stroke Conference. Boston: Butterworth-Heinemann,
activator for acute ischemic stroke. N Engl J Med 1995: 409418.
1995; 333: 15811587. 39 Linfante I, Llinas RH, Chaves C, Caplan LR, Schlaug
29 The Multicentre Acute Stroke Trial Italy (MAST-I) G. Reperfusion rates and clinical outcome of MCA
Group. Randomised controlled trial of streptokinase, versus ICA occlusion: MRI/CT before and after t-PA
aspirin, and combination of both in treatment of within 3 hours of symptom onset [Abstract].
acute ischemic stroke. Lancet 1995; 346: 15091514. American Academy of Neurology, 53rd Annual
30 Clark WM, Wissman S, Albers GW, Jhamandas JH, Meeting, 2001.
Madden KP, Hamilton S. Recombinant tissue-type 40 Kidwell CS, Saver JL, Mattiello J et al. A diusion-
plasminogen activator (Alteplase) for ischemic stroke perfusion MRI signature predicting hemorrhagic
3 to 5 hours after symptom onset. The ATLANTIS transformation following intra-arterial thrombolysis
Study: a randomized controlled trial. J Am Med Assoc [Abstract]. Stroke 32: 318.
1999; 282: 20192026. 41 Schlaug G, Beneld A, Baird AE et al. The ischemic
31 Hacke W, Kaste M, Fieschi C et al. for the ECASS penumbra: operationally dened by diusion and
Study Group. Intravenous thrombolysis with recom- perfusion MRI. Neurology 1999; 53: 15281537.
binant tissue plasminogen activator for acute hemi- 42 Hillis AE, Barker PB, Beauchamp NJ, Winters BD,
spheric stroke. The European Cooperative Acute Mirski M, Wityk RJ. Restoring blood pressure reper-
Stroke Study (ECASS). J Am Med Assoc 1995; 274: fused Wernickes area and improved language.
10171025. Neurology 2001; 56: 670672.
32 Hacke W, Kaste M, Fieschi C et al. for the Second 43 Wijdicks EF, Jack CR Jr. Intracerebral hemorrhage
European-Australasian Acute Stroke Study after brinolytic therapy for acute myocardial infarc-
Investigators. Randomised double-blind placebo- tion. Stroke 1993; 24: 554557.
controlled trial of thrombolytic therapy with intrave- 44 Kasner SE, Hankins LL, Bratina P, Morganstern LB.
nous alteplase in acute ischaemic stroke (ECASS II). Magnetic resonance angiography demonstrates vas-
Lancet 1998; 352: 12451251. cular healing of carotid and vertebral artery dissec-
33 The Multicentre Acute Stroke Trial Europe Study tions. Stroke 1997; 28: 19931997.
Group. Thrombolytic therapy with streptokinase in 45 Nederkoorn PJ, Kappelle LJ, van der Graaf Y, Hunink
acute ischemic stroke. N Engl J Med 1996; 335: MG, Eikelboom BC, Mali WP. Duplex ultrasound, MR
145150. angiography and conventional angiography in
34 Caplan LR, Mohr JP, Kistler JP, Koroshetz W. Should carotid artery disease: interpretation of separate test
thrombolytic therapy be the rst-line therapy for results and their combinations [Abstract]. Stroke
acute ischemic stroke? Thrombolysis: not a panacea 2001; 32: 335.
for ischemic stroke. N Engl J Med 1997; 337: 46 Fayad ZA, Nahar T, Fallon JT et al. In vivo magnetic
13091313. resonance evaluation of atherosclerotic plaques in
35 Albers GW. Expanding the window for thrombolytic the human thoracic aorta: a comparison with trans-
therapy in acute stroke: the potential role of acute esophageal echocardiography. Circulation 2000; 101:
MRI for patient selection. Stroke 1999; 30: 22302237. 25032509.
36 del Zoppo GJ, Higashida RT, Furlan AJ, Pessin MS, 47 Caplan LR. Brain embolism. In: Caplan LR, Hurst JW,
Rowley HA, Gent M. PROACT: a phase II randomized Chimowitz MI, eds. Clinical Neurocardiology. New
trial of recombinant pro-urokinase by direct arterial York: Marcel Dekker; 1999: 35185.
12 John N. Fink and Louis R. Caplan

48 Caplan LR. Brain embolism, revisited. Neurology North American Symptomatic Carotid
1993; 43: 12811287. Endarterectomy Trial Collaborators. Benet of
49 Amarenco P, Duyckaerts C, Tzourio C, Hennin D, carotid endarterectomy in patients with sympto-
Bousser MG, Hauw JJ. The prevalence of ulcerated matic moderate or severe stenosis. North American
plaques in the aortic arch in patients with stroke. N Symptomatic Carotid Endarterectomy Trial
Engl J Med 1992; 326: 221225. Collaborators. N Engl J Med 1998; 339: 14151425.
50 Amarenco P, Cohen A, Baudrimont M, Bousser M-G. 62 Executive Committee for the Asymptomatic Carotid
Transesophageal echocardiographic detection of Atherosclerosis Study (ACAS). Endarterectomy for
aortic arch disease in patients with cerebral infarc- asymptomatic carotid artery stenosis. J Am Med
tion. Stroke 1992; 23: 10051009. Assoc 1995; 273: 14211428.
51 The French Study of Aortic Plaques in Stroke Group. 63 Caplan LR. A 79 year-old musician with asympto-
Atherosclerotic disease of the aortic arch as a risk matic carotid artery disease. J Am Med Assoc 1995;
factor for recurrent ischemic stroke. N Engl J Med 274: 13831389.
1996; 334: 12161221. 64 Caplan LR. Caplans Stroke: a Clinical Approach. 3rd
52 Mitusch R, Doherty C, Wucherpfennig H et al. edn. Boston: Butterworth-Heinemann; 2000:
Vascular events during follow-up in patients with 175176.
aortic arch atherosclerosis. Stroke 1997; 28: 3639. 65 Chambers BR, Norris JW. Outcome in patients with
53 Blackshear JL, Jahangir A, Oldenberg WA, Saord RE. asymptomatic neck bruits. N Engl J Med 1986; 315:
Digital embolization from plaque-related thrombus 860865.
in the thoracic aorta: identication with transesoph- 66 Seibler M, Nachtman A, Sitzer M et al. Cerebral
ageal echocardiography and resolution with warfarin microembolism and the risk of ischemia in asympto-
therapy. Mayo Clin Proc 1993; 68: 268272. matic high-grade internal carotid artery stenosis.
54 Freedberg RS, Tunick PA, Culliford AT, Tatelbaum RJ, Stroke 1995; 26: 21842186.
Kronzon I. Disappearance of a large intraaortic mass 67 Manca G, Parenti G, Bellina R et al. 111In platelet scin-
in a patient with prior systemic embolization. Am tigraphy for the noninvasive detection of carotid
Heart J 1993; 125: 14451447. plaque thrombosis. Stroke 2001; 32: 719.
55 Hausmann D, Gulba D, Bargheer K, Niedermeyer J, 68 Markus H, Cullinane M. Severely impaired cerebro-
Comess KA, Daniel WG. Successful thrombolysis of vascular reactivity predicts stroke and TIA risk in
an aortic-arch thrombus in a patient after mesen- patients with carotid artery stenosis and occlusion.
teric embolism. N Engl J Med 1992; 327: 500501. Brain 2001; 124: 457467.
56 Belden JR, Caplan LR, Bojar RM, Payne DD, 69 Gur AY, Bova I, Bornstein NM. Is impaired cerebral
Blachman P. Treatment of multiple cerebral emboli vasomotor reactivity a predictive factor of stroke
from an ulcerated, thrombogenic ascending aorta asymptomatic patients? Stroke 1996; 27: 21882190.
with aortectomy and graft replacement. Neurology 70 Kim JH, Lee SJ, Shin T et al. Correlative assessment of
1997; 49: 621622. hemodynamic parameters obtained with T2*-
57 Caplan LR. Posterior Circulation Disease. Clinical weighted perfusion MR imaging and SPECT in symp-
Findings, Diagnosis and Management. Boston: tomatic carotid artery occlusion. Am J Neuroradiol
Blackwell, 1996. 2000; 21: 14501456.
58 Spelle PM, Martin JB, Weill A et al. Percutaneous 71 Derdeyn CP, Grubb RL Jr, Powers WJ. Cerebral hemo-
transluminal angioplasty and stenting of the proxi- dynamic impairment: methods of measurement and
mal vertebral artery from symptomatic stenosis. Am association with stroke risk. Neurology 1999; 53:
J Neuroradiol 2000; 21: 727731. 251259.
59 NASCET Collaborators. Benecial eect of carotid 72 Hobson RW. Carotid angioplasty-stent: clinical expe-
endarterectomy in symptomatic patients with high- rience and role for clinical trials. J Vasc Surg 2001; 33
grade carotid stenosis. N Engl J Med 1991; 325: 445453. (Suppl): S117123.
60 European Carotid Surgery Trialists Collaborative 73 Guillon B, Brunereau L, Biousse V, Djouhri H, Levy C,
Group. Randomised trial of endarterectomy for Bousser MG. Long-term follow-up of aneurysms
recently symptomatic carotid stenosis: nal results developed during extracranial internal carotid artery
of the MRC European Carotid Surgery Trial (ECST). dissection. Neurology 1999; 53: 117122.
Lancet 1998; 351: 13791387. 74 The Warfarin-Aspirin Symptomatic Intracranial
61 Barnett HJM, Taylor DW, Eliasziw M et al. for the Disease (WASID) Study Group. Prognosis of patients
The importance of specific diagnosis in stroke patient management 13

with symptomatic vertebral or basilar artery stenosis. Clinical Findings, Diagnosis, and Management.
Stroke 1998; 29: 13891392. Boston: Blackwell; 1996: 544568.
75 Bogousslavsky J, Barnett HJM, Fox AJ, Hachinski VC, 89 Caplan LR. Migraine and vertebrobasilar ischemia.
Taylor W, EC/IC Bypass Study Group. Atherosclerotic Neurology 1991; 41: 5561.
disease of the middle cerebral artery. Stroke 1986; 17: 90 Voko Z, Bots ML, Hofman A, Koudstaal PJ, Witteman
11121120. JC, Breteler MM. J-shaped relation between blood
76 Benesch CG, Chimowitz MI. Best treatment for intra- pressure and stroke in treated hypertensives.
cranial arterial stenosis? 50 years of uncertainty. Hypertension 1999; 34: 11811185.
Neurology 2000; 55: 465466. 91 Hansson L, Zanchetti A, Carruthers SG et al. for the
77 Gomez CR, Misra VK, Liu MW et al. Elective stenting HOT Study Group. Eects of intensive blood-pres-
of symptomatic basilar artery stenosis. Stroke 2000; sure lowering and low-dose aspirin in patients with
31: 9599. hypertension: principal results of the Hypertension
78 Jorgenson L, Tovrik A. Ischemic cerebrovascular dis- Optimal Treatment (HOT) randomised trial. Lancet
eases in an autopsy series. Part 2: prevalence, loca- 1998; 351: 175562.
tion, pathogenesis and clinical course of cerebral 92 Tell GS, Crouse JR, Furberg CD. Relation between
infarcts. J Neurol Sci 1969; 9: 285320. blood lipids, lipoproteins, and cerebrovascular ath-
79 Castaigne P, Lhermitte F, Gautier JC, Escourolle R, erosclerosis. A review. Stroke 1988; 19: 423430.
Derouesne C. Internal carotid artery occlusion. A 93 Iso H, Jacobs DR, Wentworth D, Neaton JD, Cohen
study of 61 instances in 50 patients with postmortem JD. Serum cholesterol and six-year mortality from
data. Brain 1970; 93: 231258. stroke in 350,977 men screened for the multiple risk
80 Blackwood W, Hallpike JF, Kocen RS, Mair WG. factor intervention trial. N Engl J Med 1989; 320:
Atheromatous disease of the carotid arterial system 904910.
and embolism from the heart in cerebral infarction: a 94 Khamashta MA, Cuadrado MJ, Mujic F, Taub NA,
morbid anatomical study. Brain 1969; 92: 897910. Hunt BJ, Hughes GR. The management of thrombo-
81 Chaves CJ, Silver B, Schlaug G, Dashe J, Caplan LR, sis in the antiphospholipid-antibody syndrome. N
Warach S. Diusion- and perfusion-weighted MRI Engl J Med 1995; 332: 993997.
patterns in borderzone infarcts. Stroke 2000; 31: 95 Radack K, Deck C, Huster G. Dietary supplementa-
10901096. tion with low-dose sh oils lowers brinogen levels: a
82 Belden JR, Caplan LR, Pessin MS, Kwan E. randomized double-blind controlled study. Ann
Mechanisms and clinical features of posterior Intern Med 1989; 111: 757758.
border-zone infarct. Neurology 1999; 53: 13121318. 96 Kobayashi S, Hirai A, Terano T, Hamazaki T, Tamura Y,
83 Angeloni U, Bozzao L, Fantozzi L, Bastianello S, Kumagai A. Reduction in blood viscosity by eicoso-
Kushner M, Fieschi C. Internal borderzone infarction pentanoic acid. Lancet 1981; 2: 197.
following acute middle cerebral artery occlusion. 97 Caplan LR. Lice, eas and strokes. Arch Neurol 2000;
Neurology 1990; 40: 11961198. 57: 11131114.
84 Baird AE, Donnan GA, Saling M. Mechanisms and 98 Yamamoto H, Bogousslavsky J. Mechanism of second
clinical features of internal watershed infarction. and further strokes. J Neurol Neurosurg Psychiatry
Clin Exp Neurol 1991; 28: 5055. 1998; 64: 771776.
85 Caplan LR, Hennerici M. Impaired clearance of 99 Barnett HJM, Meldrum HE. Carotid endarterectomy:
emboli (washout) is an important link between a neurotherapeutic advance. Arch Neurol 2000; 57:
hypoperfusion, embolism, and ischemic stroke. Arch 4045.
Neurol 1998; 55: 14751482. 100 CAPRIE Steering Committee. A randomised, blinded,
86 Darby DG, Barber PA, Gerraty RP et al. trial of clopidogrel versus aspirin in patients at risk of
Pathophysiological topography of acute ischemia by ischemic events (CAPRIE) Lancet 1996; 348:
combined diffusion-weighted and perfusion MRI. 13291339.
Stroke 1999; 30: 20432052. 101 Adams HP Jr, Davis PH, Leira EC et al. Baseline NIH
87 Solomon S, Lipton RB, Harris PY. Arterial stenosis in Stroke Scale score strongly predicts outcome after
migraine: spasm or arteriopathy? Headache 1990; 30: stroke: a report of the Trial of Org 10172 in Acute
5261. Stroke Treatment (TOAST). Neurology 1999; 53:
88 Caplan LR. Migraine and posterior circulation stroke. 126131.
In: ed. Caplan LR, ed. Posterior Circulation Disease: 102 Thijs VN, Lansberg MG, Beaulieu C, Marks MP, Moseley
14 John N. Fink and Louis R. Caplan

ME, Albers GW. Is early ischemic lesion volume on 105 Schwab S, Schwarz S, Spranger M, Keller E, Bertram
diusion-weighted imaging an independent predictor M, Hacke W. Moderate hypothermia in the treatment
of stroke outcome? A multivariable analysis. Stroke of patients with severe middle cerebral artery infarc-
2000; 31: 25972602. tion. Stroke 1998; 29: 24612466.
103 Hacke E, Schwab S, Horn M, Spranger M, De Georgia 106 Schwab S, Steiner T, Schwarz S, Steiner HH, Jansen O,
M, von Klummer R. Malignant middle cerebral Hacke W. Early hemicraniectomy in patients with
artery territory infarction: clinical course and prog- complete middle cerebral artery infarction. Stroke
nostic signs. Arch Neurol 1996; 53: 309315. 1998; 29: 18881893.
104 Koh MG, Phan TG, Atkinson JLD, Wijdicks EFM. 107 Staessen JA, Gasowski J, Wang JG et al. Risks of
Neuroimaging in deteriorating patients with cerebel- untreated and treated isolated systolic hypertension
lar infarcts and mass eect. Stroke 2000; 31: in the elderly: meta-analysis of outcome trials.
20622067. Lancet 2000; 355: 865872.
Limitations of current brain imaging modalities in stroke

P. Alan Barber and Stephen M. Davis

Department of Neurology, Royal Melbourne Hospital, University of Melbourne, Parkville, Victoria, Australia

Introduction immediate hours after symptom onset. Fear, confu-

sion, language disturbance, and physical decits
The successful management of stroke patients may all aect the ability to comprehend or follow
requires the ability to conrm the diagnosis, iden- commands. Patients are often restless, and fre-
tify the site, extent and age of the lesion, and deter- quently become more so with prolonged scanning
mine underlying pathophysiology. Several studies protocols. Many become claustrophobic within the
have made it clear that this cannot be done on the conned spaces of scanners. Head movement is a
basis of clinical ndings alone.13 At the most basic considerable problem and can result in image
level, clinicians are unable to accurately dieren- quality degradation. While this may be minimized
tiate between cerebral infarction and primary intra- by moulded head holders or masks, ill-tting
cerebral hemorrhage.4 Cerebral imaging is therefore devices may be uncomfortable and increase patient
a prerequisite in the management of almost all agitation and movement. The use of sedatives to
stroke patients. reduce patient movement may mask neurological
Techniques such as computed tomography (CT), deterioration, and is therefore relatively contraindi-
magnetic resonance imaging (MRI), positron emis- cated in the acute stroke setting.
sion tomography (PET) and single photon emission
computed tomography (SPECT) have provided a
window onto the structural and functional changes Computed tomography
that occur during stroke can be examined and have
revolutionized our understanding of the pathophy- Computed tomography is more widely available
siology of ischemia. The advent of thrombolytic and is less expensive than other tomographic
therapy has exposed a need for imaging techniques imaging modalities and has been the investigation
that enable the more rational selection of patients of choice in stroke. CT images can be obtained
for potentially hazardous treatments. Each of these rapidly and non-invasively in unwell and frequently
currently used brain imaging techniques has limita- uncooperative patients. CT identies cerebral
tions in such a role and these will form the focus of hemorrhage and is the investigation of choice for
this chapter. subarachnoid hemorrhage. In contrast to MRI, CT
can be performed in patients with ferromagnetic
metallic foreign bodies. CT is less aected by
Patient factors
patient motion than conventional MRI and also
There are a number of logistic and technical prob- allows full access to a patient during scanning,
lems related to the performance of all imaging making it of use in restless or critically ill patients.
studies in stroke patients. These diculties relate to Bony anatomy is also well delineated. This is an
a patients emotional and physical state in the advantage when bony injury is suspected, but a

16 P. Alan Barber and Stephen M. Davis

Table 2.1. Advantages and limitations of cerebral computed tomography (CT) in acute stroke

Advantages of CT in stroke Limitations of CT in stroke

Widely available and relatively inexpensive Exposure to radiation

Non-invasive Low sensitivity for infarction at the hyperacute stage
Sensitive to hemorrhage, investigation of choice for Difculties distinguishing acute from chronic infarction
subarachnoid hemorrhage
Good visualization of bony anatomy Poor imaging of posterior fossa structures
Good access to patients during scanning procedure Low interobserver agreement on the degree of ischemic
Rapid image acquisition, useful in restless patients Inability to stratify patients in acute stroke trials
CT angiography and CT perfusion imaging can be added Inability to delineate hypoperfused but viable cerebral
to conventional protocols tissue

limitation in the posterior cranial fossa where tening of cortical sulci, an asymmetry between the
beam hardening artefacts are produced by bone at sylvian ssures and compression of the ventri-
the base of the skull. A summary of the advantages cles.711
and limitations of CT in stroke can be found in These early ischemic changes are frequently
Table 2.1 and a full discussion of the role of CT in subtle and CT is often normal in the therapeutically
stroke can be found in Chapter 3. critical rst hours after stroke onset. Furthermore,
The principle behind CT is similar to conven- these changes mean that it can be particularly di-
tional X-ray imaging. Radiation passes through cult to visualize small strokes on acute CT studies.
tissue from multiple directions; detectors measure This is a result of the small volume of hypodense
the degree of attenuation of the exiting radiation, ischemic tissue and the minimal compressive
and images are reconstructed in cross-section.5 eects on surrounding brain. Thus, CT is not par-
Thus, CT scanning requires exposure to radiation, ticularly eective in identifying small subcortical or
with the average dose equivalent to the background brainstem lacunar strokes.1214
radiation received in 1 year.6 Repeated studies Over the rst days following stroke onset,
should therefore be avoided unless absolutely nec- infarcted tissue becomes more hypodense and
essary. Pregnancy is also a relative contraindication clearly demarcated.15 Cerebral swelling also
to CT, although the risks to the fetus must be increases reaching a maximum between one and
weighed against the potential benets of the infor- ve days. During the second week, some patients
mation provided. may develop a progressive increase in the density of
the infarct, which may result in fogging of the
infarct. This phenomenon may persist for up to a
Sequence of changes that are seen on CT following
fortnight and as a result strokes may become more
a stroke
dicult to identify and smaller strokes may be
The early CT changes of ischemia include paren- overlooked completely. The density of an infarct
chymal hypodensity and cerebral swelling. then decreases leaving a hypodense and atrophic
Parenchymal hypodensity corresponds to an lesion.16,17
increase in the intracellular water components of Despite this temporal sequence of changes, it is
aected brain tissue and is best seen as a loss of often dicult to dierentiate between an acute or
greywhite matter dierentiation in the cerebral subacute, and chronic infarction on CT. This is a
cortex, insular ribbon or basal ganglia (Figs. 2.1 and particular problem with small subcortical lesions
2.2). Cerebral swelling is the result of the accumula- that occur on a background of periventricular white
tion of extracellular water and can be seen as a at- matter disease. The presence of such a lesion does
Limitations of current brain imaging modalities in stroke 17



Fig. 2.1. Early ischemic changes on computed tomography (CT) may be subtle. (a) CT obtained at 2 hours 45 minutes in a patient
who presented with left hemineglect and hemiparesis, show a hyperdense right middle cerebral artery (MCA) sign (arrow), with a
subtle loss of distinction between the basal ganglia and surrounding white matter (arrowheads). (b) Magnetic resonance angio-
graphy and diusion-weighted imaging in the same patient at 2 hours 15 minutes show an occluded right MCA and a hyperin-
tense region of ischemia in the right subcortical and basal ganglia regions.

not necessarily mean that the lesion is relevant to a tor (t-PA) be avoided in patients with CT evidence
patients clinical presentation. of major ischemia on a screening CT scan.22,23
The subtlety of the early changes of ischemia As these patients may have a greater risk of
results in a low sensitivity for cerebral infarction at hemorrhagic transformation and worse outcome
the hyperacute stage. Indeed the sensitivity of CT following t-PA.11,21,24
for stroke in patients imaged within 5 hours has In practical terms, these limitations mean that
been reported to be as low as 58%.18 The diculty the presence, location and size of an infarct
identifying hyperacute ischemia on CT also results cannot be reliably determined in the rst hours
in a substantial interobserver variability in the after stroke onset. As a consequence, many
detection of acute infarction.19,20 Furthermore, the patients are treated with thrombolytic therapy
reliability and reproducibility of CT in the estima- without denite pre-therapy cerebral CT conr-
tion of the degree of ischemic change is mation of diagnosis of stroke. The major role for
modest.11,21 This is of importance, as some guide- hyperacute CT is to exclude primary intracerebral
lines recommend that tissue plasminogen activa- hemorrhage and other non stroke pathology. CT
18 P. Alan Barber and Stephen M. Davis



Fig. 2.2. Early ischemic changes on computed tomography (CT) may be subtle. (a) CT obtained at 2 hours in a patient who pre-
sented with left hemineglect and hemiparesis, shows a hyperdense right middle cerebral artery (MCA) sign (arrow) and subtle loss
of distinction between cortical grey and white matter in the right middle cerebral artery territory (arrowheads). (b) Magnetic reso-
nance angiography and diusion-weighted imaging in the same patient at 4 hours shows an occluded right MCA with a hyperin-
tense region of ischemia involving the whole of the right MCA territory.

may also lead to an underestimation of the volume Computed tomographic angiography and
of ischemic change so that patients are exposed to perfusion imaging
the risks of thrombolytic therapy, despite guide-
lines to the contrary. Conversely, excessive Computed tomographic angiography (CTA) is a
concern about the degree of ischemic change on rapid, reliable and safe method for imaging the
CT can lead to treatment being inappropriately intracranial vasculature. Data is acquired fol-
withheld.25 In acute stroke trials, CT cannot be lowing the administration of an intravenous
used to reliably stratify participants according to contrast agent. Reconstruction of this data pro-
infarct location or size before therapy is com- duces angiographic images that can be displayed
menced. Patients with similar strokes can only be in three dimensions. CTA is able to delineate
compared retrospectively after treatment has been the site of vessel occlusion and length of
given. occluded segment in acute stroke patients, and
Limitations of current brain imaging modalities in stroke 19

Table 2.2. Advantages and limitations of conventional magnetic resonance imaging (MRI) in acute stroke

Advantages of MRI in stroke Limitations of MRI in stroke

Widely available Expensive

Rapid and non-invasive Contraindicated in patients with metal fragments,
pacemakers and other electrically or magnetically
active implanted devices
High soft tissue resolution Low sensitivity for infarction in the hyperacute stage
Sensitive to brainstem ischemia Difculties estimating the degree of ischemic change
Can weight images to augment tissue contrast Unable to delineate hypoperfused but viable cerebral
Multiple sequences can be obtained providing more Unable to stratify patients in acute stroke trials
information than from any one sequence alone

compares well with other vascular imaging tech- Magnetic resonance imaging
CT can also be used to produce maps of cerebral Magnetic resonance imaging is the imaging modal-
perfusion by following the change in Hounseld ity of choice for most neurologic conditions. MR
number that occurs with the passage of an intrave- images can be obtained rapidly, non-invasively and
nous contrast agent through the cerebral vascular with no exposure to ionizing radiation. Images can
bed. This change in density is proportional to the be generated in any orientation. MRI has high soft
concentration of the contrast agent, and can there- tissue resolution and sensitivity to tissue edema.
fore be used to determine ow information. CT per- MRI has a higher sensitivity than CT to brainstem
fusion-imaging techniques have been used to and cerebellar ischemia, making it of particular use
calculate maps of tracer transit time and, more in the posterior fossa. MR images can be weighted
recently, perfused cerebral blood volume.30,31 A to augment contrast between tissue types.
further CT perfusion imaging technique used in Furthermore, specialized pulse sequences have
stroke is xenon enhanced CT (XeCT).3234 However, been developed to highlight specic tissue proper-
xenon is a mild short acting anesthetic and the use ties. A number of dierent sequences can be
of XeCT is not widespread. obtained at the same imaging session, providing
While the clinical role of CTA and CT perfusion more information than from any single sequence
imaging is still yet to be dened, these techniques alone. For example, information provided by T1-
have enormous potential in acute stroke. As almost and T2-weighted imaging can be augmented by that
all patients have CT scans, clinically useful vascular provided by susceptibility-weighted sequences
and blood ow information can be obtained with (sensitive to hemorrhage), diusion-weighted
little added time or expense. In addition, CTA is less imaging (tissue destined to infarct without prompt
invasive than conventional angiography and both intervention), MR angiography (vascular anatomy),
CTA and perfusion CT can be performed in patients MR perfusion imaging (cerebral perfusion), and MR
who are critically ill or claustrophobic, or who have spectroscopy (concentration of specic cerebral
metallic foreign bodies such as pacemakers. metabolites). The technical aspects and role of MRI
Multimodal CT may therefore directly compete in stroke will be discussed in later chapters.
with the MR imaging techniques discussed in later However, MRI has a number of limitations in
chapters. However, CT is unable to delineate hypo- stroke (Table 2.2). First, MRI is approximately twice
perfused but potentially salvageable tissue of the as expensive as CT. Claustrophobic patients may
ischemic penumbra. have diculties tolerating the studies. MRI is
contraindicated in patients with fragments of
metal, for example within the eyes, and in patients
20 P. Alan Barber and Stephen M. Davis

with intracranial aneurysm clips, cardiac pacemak-

ers or any other electrically or magnetically active
implanted devices that could interact with the mag-
netic eld.

Early ischemic changes are subtle

Ischemic changes may be seen in conventional MR
studies within 1 to 2 hours of stroke onset. The ear-
liest ndings include loss of the normal ow void
within major intracranial vessels and the presence
of arterial enhancement if contrast has been
used.35,36 The rst morphological changes are due
to the development of parenchymal swelling with
eacement of cortical sulci and distortion of the
ventricular system. These changes are rst seen in
T1-weighted sequences and may be present in up to
half of patients within 6 hours.36 This early swelling
occurs without signal change and is most likely
related to the onset of cytotoxic edema, which can
develop within minutes in experimental ischemia.37
Signal changes only appear with the development
of vasogenic edema and are not usually found
before eight hours on T2-weighted sequences or 16
hours on T1-weighted sequences.
As a consequence, the overall sensitivity of MRI
for ischemia is low in the rst few hours following
the onset of symptoms.36,38,39 Conventional MRI is
also subject to the same limitations as cerebral Fig. 2.3. Conventional magnetic resonance imaging
sequences may not dierentiate acute from chronic infarc-
CT when it comes to delineating acute from
tion. Conventional T2-weighted imaging (T2-WI) (A) and
chronic infarction, and stratifying patients
diusion-weighted imaging (DWI) (B) in three patients with
according to infarct presence, location and size
subcortical infarction imaged at 11 hours (a), 10 hours (b)
prior to therapy or randomization in acute stroke and 10.5 hours (c) after the onset of symptoms. Ischemic
trials (Fig. 2.3). changes are seen in the periventricular and subcortical white
matter on T2-WI but there is no way of distinguishing acute
from chronic infarction. Hyperintense regions of acute
MRI in intracerebral hemorrhage
infarction are clearly seen on DWI.
There has been a widespread belief that conven-
tional MRI is less sensitive than CT in the detection
of intracerebral hemorrhage. Much of this pessi-
mism is based on a number of early trials using MR imaging may detect subarachnoid hemorrhage.45
scanners with low eld strengths. However, there is However, until large randomized controlled trials
increasing evidence that echoplanar gradient-echo conrm the ability of MRI to detect intraparenchy-
T2*-weighted imaging, also termed susceptibility mal and subarachnoid hemorrhage, a screening CT
weighted imaging, is reliable in the detection of must be considered in all stroke patients. The
acute intraparenchymal hemorrhage.4044 In addi- expense of any screening CT must be added to the
tion, uid-attenuated inversion recovery (FLAIR) cost of performing MRI.
Limitations of current brain imaging modalities in stroke 21

Table 2.3. Advantages and limitations of single photon emission computed tomography (SPECT) in stroke

Advantages of SPECT in stroke Limitations of SPECT in stroke

Generally well tolerated Technically and logistically complex

Images cerebral perfusion Exposure to ionizing radiation
Can demonstrate collateral ow or remote phenomena Still require CT or MRI imaging
such as diaschisis
Can demonstrate ischemic tissue while CT and No standardized image acquisition or analysis protocols
conventional MRI are normal
With 99Tc-HMPAO, images reect perfusion at the time Unable to delineate the ischemic penumbra
of injection, so that scanning can be delayed
SPECT has no established clinical role in stroke

Echoplanar magnetic resonance imaging within the body. This raises concerns about the
generation of cardiac dysrhythmias or tetanic
Echoplanar MRI (EPI) oers advantages over con-
muscle contractions and some patients have
ventional MRI or CT. Echoplanar images can be
reported mild twitching or pain. However, the
obtained after a single measurement or shot, so that
current frequencies for switched gradient elds are
all data may be collected after a single excita-
usually well below the threshold for neuromuscular
tion.4648 This is in contrast to conventional MR
spin- and gradient-echo sequences in which only a
small portion of the data required to construct an
image is collected after each excitation. As a result,
EPI enables whole brain imaging in seconds. This in Single photon emission computed tomography
turn has facilitated the development of functional
brain imaging. EPI may be sensitized to ow and Single photon emission computed tomography
diusion in the same way as conventional spin- (SPECT) is a rapid, relatively non-invasive perfu-
echo or gradient-echo imaging sequences.47 sion imaging modality that has been used since the
However, because of EPIs very rapid imaging capa- late 1970s to study cerebral blood ow changes.
bilities, whole brain perfusion-weighted imaging SPECT is particularly well suited to the investiga-
(PWI) and diusion-weighted imaging (DWI) can tion of stroke patients (Table 2.3). SPECT can
be obtained within minutes. This rapid imaging provide assessments of the degree and extent of
capability renders EPI relatively insensitive to ischemia, give an indication of likely underlying
movement artefact. pathogenesis, and demonstrate the presence of col-
EPI has several limitations in stroke patients. lateral ow and remote physiological phenomena
Firstly, it is as expensive and subject to the same such as diaschisis (Fig. 2.4, see colour plate
contraindications as conventional MRI. EPI is sen- section). However, SPECT is technically and logisti-
sitive to chemical shift artefacts that require the use cally complex, and at present has no established
of fat suppression corrections. Susceptibility arte- clinical role in stroke. A full discussion of the limita-
facts, which result in image distortions, are a partic- tions of SPECT in stroke requires some understand-
ular problem in the posterior fossa and around the ing of its underlying principles. These will be
paranasal sinuses. EPI also has the potential to gen- summarized in the following section.
erate electric currents, which is related to the rate of
switching of the magnetic eld when rapidly alter-
Acquisition of SPECT studies
nating radiofrequency pulses are applied. A change
in magnetic eld causes a change in electric eld Perfusion SPECT acquires three-dimensional data
and results in the generation of electric currents that represent regional perfusion by localizing
22 P. Alan Barber and Stephen M. Davis

gamma ray-emitting radiotracers. These radiotracers isopropyl-p-iodoamphetamine (IMP, Spectamine)

are distributed in the brain according to the rate of and the inert gas 133xenon have been largely super-
CBF. The emitted photons are then detected using seded by the newer radiotracers.
gamma cameras or one of several dierent ring-
type gamma-ray detector systems. These may be
SPECT analysis
rotating gamma detector arrays, xed detector
systems, or single or multihead camera systems. SPECT studies can provide both semiquantitative
Image acquisition time depends on the desired and qualitative assessments of cerebral perfusion.
quality of the image, as well as the imaging system This is in contrast to PET in which absolute meas-
and radiotracer used. In general, high-resolution ures of blood ow can be determined.
images can be obtained within 2030 minutes.51 Semiquantitative SPECT analysis takes into account
After data acquisition, individual count rates at the number of radioactive counts within a region of
each location are calculated and ltered to avoid interest, expressed as a fraction of the total counts
excessive amplication of noise. Computer gener- within the brain or with respect to some unaected
ated reconstruction then enables data to be pre- region such as the contralateral hemisphere or cere-
sented three-dimensionally or as a series of bellum.59 This requires high-resolution whole brain
two-dimensional slices in the axial, coronal or sag- imaging, so that image acquisition and analysis may
ittal planes. An image consists of a matrix of pixels, take 30 minutes or more. Simpler qualitative acqui-
each reecting the number of counts emitted by sition protocols have been developed that require
the gamma rays.52 In areas of reduced perfusion, fewer slices with lower resolution. These protocols
less radiotracer is emitted and consequently the can be combined with visual analysis so that perfu-
region appears less bright. sion may be reported as simply being normal, low,
absent or high.60 This results in shorter acquisition
and analysis times. Qualitative SPECT measures
Radiolabelled tracers
have been shown to correlate with measured hypo-
The most commonly used tracer for cerebral perfu- perfusion volumes, and both stroke severity and
sion SPECT is 99m-Technetium hexamethylpropyle- outcome.60,61,62
neamine oxime (99Tc-HMPAO, exametazime,
Ceretec, Amersham International). 99Tc-HMPAO is
SPECT in stroke
a commercially available, low molecular weight,
lipophilic macrocyclic amine. It is less expensive With acute cerebral ischemia, up to 90% of SPECT
and oers better spatial resolution than earlier scans will be abnormal within eight hours of
radiotracers. 99Tc-HMPAO is unstable in vitro and symptom onset. SPECT may demonstrate the pres-
needs to be injected soon after preparation, which ence, location and degree of ischemia at a time
assuming technicians are available, takes approxi- when the changes on CT or MRI are subtle or
mately 30 minutes. 99Tc-HMPAO has an in vivo inconclusive.51,63 An exception to this is with
half-life of approximately 6 hours so that imaging lacunar infarction in which SPECT studies are often
may proceed immediately or be delayed for several normal. The degree of hypoperfusion or volume of
hours with images reecting the state of cerebral a hypoperfused region also correlate with clinical
perfusion at the time of the injection.5356 The in state, stroke outcome and nal infarct size.6469 This
vivo retention of 99Tc-HMPAO prevents early repeat prognostic information is over and above that pro-
imaging and results in a greater exposure to radia- vided by clinical scale scores alone.60,7072
tion than other tracers.57,58 99Tc-ethyl cysteinate SPECT studies may facilitate patient targeted
dimer (ECD, Neurolite) is also used in perfusion acute stroke therapy. Ueda et al.73 found that
SPECT and has the advantage over 99Tc-HMPAO of ischemic tissue with CBF less than 35% of cerebel-
being stable in vitro for approximately 6 hours.51 lar ow may be salvaged if intra-arterial thromboly-
Other perfusion SPECT radiotracers such as 123I N- sis is initiated up to 5 hours after stroke onset, but is
Limitations of current brain imaging modalities in stroke 23

at risk of hemorrhagic transformation after this no widespread and standardized SPECT image
time. SPECT may help exclude patients from partic- acquisition or analysis protocol. This has made it
ular acute stroke therapies. For example, patients dicult to compare results obtained by dierent
with increased isolated radioisotope uptake have a centres. As a consequence, the use of SPECT in
favourable outcome. These patients are likely to large multicentre acute stroke therapy trials has
have reperfused at the time of imaging and should been limited.80
avoid thrombolytic therapy. Patients with a focal SPECT is unable to assess cerebral metabolism. It
absence of perfusion are likely to have vessel occlu- is therefore unable to determine the presence of
sion without collateral ow and may be at risk of hypoperfused yet still potentially viable penumbral
hemorrhagic transformation or death following tissue (Fig. 2.6, see colour plate section). This infor-
thromobolytic therapy.7477 The feasibility of using mation can only be inferred in retrospect by deter-
SPECT to help decide acute stroke therapy has been mining whether any reperfusion was maintained at
demonstrated by a number of groups. Serial SPECT outcome (nutritional reperfusion). Thus, SPECT
may also be used to monitor spontaneous changes cannot be used to limit therapy to only those with
or therapy-induced changes in perfusion over time potentially salvageable ischemic tissue.
(Fig. 2.5, see colour plate section).62,75,7881

Positron emission tomography

SPECT limitations

Perfusion SPECT is a generally safe and well- Positron emission tomography (PET) has provided
tolerated procedure. One of the major concerns is major insights into the response of cerebral tissues
exposure to ionizing radiation. The International to reduced cerebral perfusion. In stroke, PET
Commission on Radiological Protection has devel- enables the simultaneous measurement of regional
oped a measure, the eective dose equivalent CBF, the consumption of oxygen with the cerebral
(EDE), which takes into account the relative radio- metabolic rate of oxygen (CMRO2), the consump-
sensitivity of each organ and tissue.82 The EDE for tion of glucose with the cerebral metabolic rate of
Tc-HMPAO of an injected dose of 500 MBq, with glucose (CMRglc), and the oxygen extraction frac-
bladder voiding every 3.5 hours, is 6.9 mSv. This is tion (OEF). These measures permit the delineation
similar to that received from a radionuclide bone of CBF thresholds for electrical and structural
scan and is 43% of the average annual background failure and as such PET has enabled the character-
radiation in the United States.51 SPECT is contrain- ization of the ischemic penumbra. The role of PET
dicated in pregnancy. in imaging the ischemic penumbra will be dis-
A further major limitation of perfusion SPECT is cussed in more detail in Chapter 15. However, while
poor image quality compared to CT or MR imaging. it has been a useful research tool, PET is not well
This is the result of the eects of photon attenua- suited to the acute investigation of stroke patients
tion and scatter, inadequate spatial resolution and (Table 2.4).
partial volume eects. Structural imaging with CT
or MRI must also be performed to exclude intracra-
Acquisition of PET images
nial hemorrhage and other non-stroke pathologies.
Thus two separate imaging sessions, usually in With PET imaging, patients are given radionuclide
dierent hospital departments, are required. This tracers that are radioisotope labelled biological
adds to the time needed for acute investigation and molecules, e.g. 11C-CO2, 15O-O2, 15O-H2O, 18F-
delays the institution of any therapy. uorodeoxyglucose (18F-FDG), or drugs, e.g. 18F-
SPECT is more operator dependent than CT or uoromisonidazole (18F-FMISO). As these tracer
MRI. Thus, the interpretation of results often isotopes decay, a positron is emitted. A positron is a
requires a close collaboration between nuclear subatomic particle with the same mass as an
medicine physicians and clinicians.51There is also electron but with a positive, rather than negative,
24 P. Alan Barber and Stephen M. Davis

Table 2.4. Advantages and limitations of positron emission tomography (PET) in stroke

Advantages of PET in stroke Limitations of PET in stroke

Quantication of cerebral blood ow Expensive

Demonstration of cerebral metabolism Exposure to ionizing radiation
Enables mapping of neuroreceptors Technically and logistically complex
Demonstration of the ischemic penumbra Limited to research or large tertiary institutions
Still require CT or MRI imaging
No standardized image acquisition or analysis protocols
PET has no established clinical role in stroke

charge. The positron then interacts with an electron rapidly converted to15OH2O by red blood cell car-
in a matterantimatter reaction, resulting in the bonic anhydrase. However, 15OH2O based tech-
annihilation of both particles and the release of two niques may underestimate CBF as a result of the
gamma rays (photons). The gamma rays diverge incomplete permeability of the bloodbrain barrier
from the site in opposite directions and are to H2O and incomplete tissue extraction of H2O at
detected by one of a large number of external high ow rates.84
detector pairs congured in one or more rings. The two main techniques for measuring CBF are
By conguring the system so that only the near the steady state and autoradiographic methods. In
simultaneous arrival of these photons is detected the steady state method, tracer is administered
(coincidence detection), only those photons arising until an equilibrium is reached. At this point, the
from between the detectors will be recorded. Not all amount of 15OH2O entering the brain in arterial
photon pairs will reach the detectors because of a blood is equal to the amount lost to radioactive
change in direction (Compton scatter), or absorption decay and venous outow. Simultaneous measure-
by intervening tissues. Accurate quantication of ment of arterial blood radioactivity enables quan-
images requires correction for this absorption and tication of regional CBF from PET imaging. The
scatter of photons. Two- or three-dimensional images autoradiographic method of CBF measurement
can then be generated according to the distribution of utilizes a single bolus of tracer (15OH2O or
regional radioactivity. Mathematical models that 15
OCO2) followed by PET scanning.8587 Tracer
relate radionuclide tissue concentration to the activity in the arterial blood is also measured, with
physiological variable under study are then used. CBF values determined assuming a constant blood:
These models attempt to account for tracer delivery brain partition coecient for the tracer. Each of
to tissues, tracer distribution and metabolism within these techniques have advantages and disadvan-
the tissues, the eects of recirculation of both metab- tages which are beyond the scope of this chapter
olized and non-metabolized tracer, and the amount and will not be discussed further.
of tracer remaining within the bloodstream.

Cerebral blood volume

Cerebral blood flow
Cerebral blood volume (CBV) can be measured
Cerebral blood ow (CBF) is the most useful using 11C or 15O labelled CO. The CO binds irrever-
physiological parameter to be measured by PET in sibly to hemoglobin forming labelled carboxyhe-
stroke (Fig. 2.7, see colour plate section). The meas- moglobin. Once equilibrium is reached, the
urement of CBF is based on the principles of inert, measured peripheral blood radioactivity will be
freely diusible tracers.83 Most PET CBF studies are proportional to the regional carboxyhemoglobin
performed using either intravenously injected concentration, red cell mass and CBV. This enables
OH2O or inhaled 15OCO2; where 15OCO2 is quantication of CBV after adjustments are made
Limitations of current brain imaging modalities in stroke 25

for the hematocrit in peripheral blood and cerebral patients. Protocols typically require the placement
microvessels. CBV can then be used to correct other of an arterial catheter to monitor plasma concentra-
PET measurements such as the CMRO2 and OEF, for tions of radiolabelled tracers. Diculties related to
radiotracer not extracted by the tissues. the use of inhalational tracers (e.g. 15O-labelled O2
or CO2) are common and may result from inability
to comprehend the required task. Other problems
Cerebral metabolic rate of oxygen and oxygen
arise from comorbidities aecting respiratory func-
extraction fraction
tion and unstable blood pressure.
Measures of cerebral oxygen metabolism are PET is expensive and requires specialized equip-
usually measured using a steady-state method after ment and sta. The short half-life of the commonly
the inhalation of 15OO2.88 Tissue radioactivity is used PET radioisotopes (t for 15O 2 min, 11C 20
proportional to the amount of oxygen extracted min, 18F 110 min) mean that they must be gener-
from the blood (OEF). The CMRO2 can then be cal- ated by a cyclotron that is either at or near the PET
culated from the relationship between OEF and scanner. This contributes to the signicant techni-
CBF, described by the equation: cal and logistical complexities associated with
scanning patients at the acute stage. As a result,
CRMO2 CBFOEFarterial O2 content
PET scanners are limited to a small number of large
Tracer that remains bound to hemoglobin can tertiary hospitals and research centres. The acquisi-
result in an overestimation of OEF and CMRO2. This tion and analysis of PET images is also time con-
error is greatest at where CBF is low, such as suming. Additional time is needed with techniques
ischemic or infarcted regions but can be corrected such as 18F-FMISO PET, which requires a delay
by the independent measurement of CBV. Cerebral between tracer injection and imaging.
oxygen metabolism can also be measured using a As a result of these diculties, it is unlikely that
single breath inhalation method,89 and a dynamic PET will have a place in the routine investigation of
method in which dynamic PET scans are obtained acute stroke patients in the foreseeable future. To
after single breath, multiple breath, or continuous date, there have been few studies investigating the
inhalation of labelled oxygen.90 potential role of PET in acute therapeutic interven-
tion. Most of these have examined small numbers of
Other PET techniques used to study cerebral patients with considerable variations in the time
ischemia interval between symptom onset and PET
Additional PET techniques used to study cerebral studies.9597 The only data on the eect of thromb-
ischemia have included the measurement of olysis in humans studied with PET comes from
glucose metabolism (CMRglc) using a modication Heiss et al.,98 who reported on 12 stroke patients
of the autoradiographic method described by studied with PET either before or during the admin-
Sokolo and colleagues.91 Benzodiazepine recep- istration of alteplase. None the less, PET has pro-
tors can be mapped using 11C-umazenil.92 Studies vided an invaluable contribution to our knowledge
using the hypoxia marker 18F-uoromisonidazole of stroke pathophysiology, particularly the charac-
have been reported as a potential method for terization of the ischemic penumbra.
directly identifying the ischemic penumbra.93

Limitations of PET

PET is a generally safe procedure. The radiation Of the imaging modalities described above, only CT
exposure in a typical study is similar to that received and MRI have established roles in the clinical
in many other routine imaging studies.94 A moder- setting. However, both CT and MRI have low sensi-
ate degree of patient cooperation is required, which tivities for acute ischemia in the hyperacute phase
can be a problem in unwell, aphasic or unconscious and neither imaging modality is able to identify
26 P. Alan Barber and Stephen M. Davis

hypoperfused but potentially salvageable ischemic 9 Tomura N, Uemura K, Inugami A, Fujita H, Higano S,
tissue. SPECT and PET have added signicantly to Shishido F. Early CT nding in cerebral infarction:
our knowledge of the pathophysiology of stroke. obscuration of the lentiform nucleus. Radiology 1988;
However for logistic reasons, neither is particularly 168: 463467.
suited to the acute evaluation of stroke patients. 10 von Kummer R, Nolte PN, Schnittger H, Thron A,
Ringelstein EB. Detectability of cerebral hemisphere
PET is the only established method of identifying
ischaemic infarcts by CT within 6 h of stroke.
the penumbra but its cost, technical complexity
Neuroradiology 1994; 38: 3133.
and limited availability mean it is far from suitable
11 von Kummer R, Allen KL, Holle R. Acute stroke: use-
for routine clinical use. While SPECT is certainly fulness of early CT ndings before thrombolytic
less expensive and more widely available than PET, therapy. Radiology 1997; 205: 327333.
there is no accepted standardized image acquisi- 12 Donnan GD, Tress B, Bladin P. A prospective study of
tion and analysis protocol. lacunar infarction using computerised tomography.
Neurology 1982; 32: 4956.
13 Bamford J, Sandercock P, Jones L, Warlow C. The
natural history of lacunar infarction: the Oxfordshire
AC K N OW L E D G E M E N T S Community Stroke Project. Stroke 1987; 18: 545551.
14 Lindgren A, Norrving B, Rudling O, Johansson BB.
We thank Dr Stephen Read PhD, FRACP for his Comparison of clinical and neuroradiological ndings
in rst-ever stroke. A population-based study. Stroke
advice on the preparation of this chapter.
1994; 25: 13711377.
15 Hakim AM, Ryder-Cooke A, Melanson D. Sequential
computerized tomographic appearance of strokes.
Stroke 1983; 14: 893897.
REFERENCES 16 Becker H, Desch H, Hacker H, Pencz A. CT fogging
eect with ischemic cerebral infarcts. Neuroradiology
1 Madden KP, Karanjia PN, Adams HP, Jr., Clarke WR. 1979; 18: 185192.
Accuracy of initial stroke subtype diagnosis in the 17 Skriver EB, Olsen TS. Transient disappearance of cere-
TOAST study. Trial of ORG 10172 in acute stroke treat- bral infarcts on CT scan, the so-called fogging eect.
ment. Neurology 1995; 45: 19751979. Neuroradiology 1981; 22: 6165.
2 Toni D, Fiorelli M, De Michele M et al. Clinical and 18 Horowitz SH, Zito JL, Dinnarumma R, Patel M, Alvir J.
prognostic correlates of stroke subtype misdiagnosis Computed tomographic-angiographic ndings within
within 12 hours from onset. Stroke 1995; 26: the rst ve hours of cerebral infarction. Stroke 1991;
18371840. 22: 12451253.
3 Kidwell CS, Alger JR, Di Salle F et al. Diusion MRI in 19 von Kummer R, Holle R, Gizyska U et al. Interobserver
patients with transient ischemic attacks. Stroke 1999; agreement in assessing early CT signs of middle cere-
30: 11741180. bral artery infarction. Am J Neuroradiol 1996; 17:
4 Warlow CP, Dennis MS, van Gijn J et al. Stroke: A 17431748.
Practical Guide to Management. 2nd edn. Oxford: 20 Schriger DL, Kalafut M, Starkman S, Krueger M, Saver
Blackwell Science Ltd; 2001. JL. Cranial computed tomography interpretation in
5 Gilman S. Imaging of the brain. First of two parts. N acute stroke: physician accuracy in determining eli-
Engl J Med 1998; 338: 812820. gibility for thrombolytic therapy. J Am Med Assoc 1998;
6 Royal College of Radiologists. Making the Best Use of a 279: 12931297.
Department of Clinical Radiology. 3rd edn. London: 21 Marks MP, Holmgren EB, Fox A, Patel S, von Kummer
Royal College of Radiologists; 1995. R, Froehlich J. Evaluation of early computed tomo-
7 Garcia JH. Experimental ischemic stroke: a review. graphic ndings in acute ischemic stroke. Stroke 1999;
Stroke 1984; 15: 514. 30: 389392.
8 Unger E, Littleeld J, Gado M. Water content and 22 Adams HP, Brott TG, Furlan AJ et al. Guidelines for
water structure in CT and MR signal changes: possible thrombolytic therapy for acute stroke: A supplement
inuence in detection of early stroke. Am J to the guidelines for the management of patients with
Neuroradiol 1988; 9: 687691. acute ischemic stroke. A statement for healthcare pro-
Limitations of current brain imaging modalities in stroke 27

fessionals from a special writing group of the stroke 35 Biller J, Yuh WTC, Mitchell GW, Bruno A, Adams HP,
council, American Heart Association. Stroke 1996; 27: Jr. Early diagnosis of basilar artery occlusion using
17111718. magnetic resonance imaging. Stroke 1988; 19:
23 Neurology Report of the Quality Standards Committee 297306.
of the American Academy of Neurology. Practice advi- 36 Yuh WTC, Crain MR, Loes DJ, Greene GM, Ryals TJ,
sory: Thrombolytic therapy for acute ischemic stroke Sato Y. MR imaging of cerebral ischemia: ndings in
summary statement. Neurology 1996; 47: 835839. the rst 24 hours. Am J Neuroradiol 1991; 12: 621629.
24 Mies G, Auer L, Ebhardt G, Traupe H, Heiss W. Flow 37 Schuier FJ, Hossmann K-A. Experimental brain
and neuronal density in tissue surrounding chronic infarcts in cats II: ischemic brain edema. Stroke 1980;
infarction. Stroke 1983; 14: 2227. 11: 593601.
25 Barber PA, Darby DG, Desmond PM et al. 38 Spetzler RF, Zabramksi JM, Kaufman B, Yeung HN.
Identication of major ischemic change: diusion- Acute NMR changes during MCA occlusion: a prelimi-
weighted imaging versus computed tomography. nary study in primates. Stroke 1983; 14: 185191.
Stroke 1999; 30: 20592065. 39 Brant-Zawadzki M, Pereira B, Weinstein P et al. MR
26 Knauth M, von Kummer R, Jansen O, Hahnel S, imaging of acute experimental ischemia in cats. Am J
Dorer A, Sartor K. Potential of CT angiography in Neuroradiol 1986; 158: 701705.
acute ischemic stroke. Am J Neuroradiol. 1997; 18: 40 Patel R, Edelman R, Warach S. Detection of hyper-
10011010. acute primary intraparenchymal hemorrhage by
27 Shrier DA, Tanaka H, Numaguchi Y, Konno S, Patel U, magnetic resonance imaging. Stroke 1996; 27:
Shibata D. CT angiography in the evaluation of acute 23212324.
stroke. Am J Neuroradiol 1997; 18: 10111020. 41 Atlas SW, Thulborn KR. MR detection of hyperacute
28 Na DG, Byun HS, Lee KH et al. Acute occlusion of the parenchymal hemorrhage of the brain. Am J
middle cerebral artery: early evaluation with triphasic Neuroradiol 1998; 19: 14711507.
helical CTpreliminary results. Radiology 1998; 207: 42 Schellinger PD, Jansen O, Fiebach JB, Hacke W, Sartor
113122. K. A standardized MRI stroke protocol: comparison
29 Wildermuth S, Knauth M, Brandt T, Winter R, Sartor K, with CT in hyperacute intracerebral hemorrhage.
Hacke W. Role of CT angiography in patient selection Stroke 1999; 30: 765768.
for thrombolytic therapy in acute hemispheric stroke. 43 Linfante I, Llinas RH, Caplan L, Warach S. MRI fea-
Stroke 1998; 29: 935938. tures of intracerebral hemorrhage within 2 hours from
30 Muzelaar JP, Fatouros PP, Schroder ML. A new method symptom onset. Stroke 1999; 30: 22632267.
for quantitative regional cerebral blood volume meas- 44 Ebisu T, Tanaka C, Umeda M et al. Hemorrhagic and
urements using computed tomography. Stroke 1997; nonhemorrhagic stroke: diagnosis with diusion-
28: 19982005. weighted and T2-weighted echoplanar MR imaging.
31 Hunter GJ, Hamberg LM, Ponzo JA et al. Assessment Radiology 1997; 203: 823828.
of cerebral perfusion and arterial anatomy in hyper- 45 Singer MB, Atlas SW, Drayer BP. Subarachnoid space
acute stroke with three-dimensional functional CT: disease: a diagnosis with uid-attenuated inversion
early clinical results. Am J Neuroradiol 1998; recovery MR imaging and comparison with gadolin-
19: 2937. ium-enhanced spin-echo MR imaging blinded
32 Gur D, Good WF, Wolfson SK, Jr., Yonas H, Shabason L. reader study. Radiology 1998; 208: 417422.
In vivo mapping of local cerebral blood ow by xenon- 46 Manseld P. Multi-planar image formation using NMR
enhanced computed tomography. Science 1982; 215: spin echoes. J Phys C 1977; 10: 349352.
12671268. 47 Turner R, Le Bihan D, Chesnick AS. Echo-planar imaging
33 Yonas H, Darby JM, Marks EC, Durham SR, Maxwell C. of diusion and perfusion. MRM 1991; 19: 247253.
CBF measured by Xe-CT: approach to analysis and 48 Edelman R, Wielopolski P, Schmitt F. Echoplanar MR
normal values. J Cereb Blood Flow Metab 1991; 11: imaging. Radiology 1994; 192: 600612.
716725. 49 Kanal E. An overview of electromagnetic safety con-
34 Firlik AD, Kaufmann AM, Weschler LR, Firlik KS, siderations associated with magnetic resonance
Furkui MB, Yonas H. Quantitative cerebral blood ow imaging. Ann NY Acad Sci 1992; 649: 202224.
determinations in acute ischemic stroke. Relationship 50 Athey T. Current FDA guidance for MR patient expo-
to computed tomography and angiography. Stroke sure and considerations for the future. Ann NY Acad
1997; 28: 22082213. Sci 1992; 649: 242257.
28 P. Alan Barber and Stephen M. Davis

51 Report of the Therapeutics and Technology 64 Rango M, Candelise L, Perani D et al. Cortical
Assessment Subcommittee of the American Academy pathophysiology and clinical neurologic abnormal-
of Neurology. Assessment of brain SPECT. Neurology. ities in acute cerebral ischemia. A serial study with
1996; 46: 278285. single photon emission computed tomography. Arch
52 Eberl S, Zimmerman RD. The computer and its appli- Neurol 1989; 46: 13181322.
cation in nuclear medicine. In: Murray I, Ell P, eds. 65 Raynaud C, Rancurel G, Samson Y et al.
Nuclear Medicine in Clinical Diagnosis and Pathophysiologic study of chronic infarcts with I-123
Treatment. Edinburgh: Churchill Livingstone; 1994: isopropyl iodo-amphetamine (IMP): the importance
12991314. of periinfact area. Stroke 1987; 18: 2129.
53 Costa DC, Ell PJ, Cullum ID, Jarritt PH. The in vivo dis- 66 Bushnell DL, Gupta S, Micoch AG, Barnes WE.
tribution of 99Tc-HM-PAO in normal man. Nucl Med Prediction of language and neurologic recovery after
Commun 1986; 7: 647658. cerebral infarction with SPECT imaging using N-iso-
54 Lassen NA, Andersen AR, Friberg L, Paulson OB. The propyl-p-(I-123) iodoamphetamine. Arch Neurol 1989;
retention of [99m-Tc]-,-HM-PAO in the human brain 46: 665669.
after intracarotid bolus injection: a kinetic analysis. J 67 Launes J, Nikkinen P, Lindroth L, Brownell AL,
Cereb Blood Flow Metab 1988; 8: S13S22. Liewendahl K, Ivanainen M. Brain perfusion defect
55 Neirinck RD, Burke JF, Harrison RC, Forster AM, size in SPECT predicts outcome in cerebral infarction.
Andersen AR, Lassen NA. The retention mechanism Nucl Med Commun 1989; 10: 891900.
of technetium-99m-HMPAO: intracellular reaction 68 Giubilei F, Lenzi GL, Di Piero V et al. Predictive value
with glutathione. J Cereb Blood Flow Metab 1988; 8: of brain perfusion single-photon emission computed
S4S12. tomography in acute ischemic stroke. Stroke 1990; 21:
56 Holman BL, Devous MD. Functional Brain SPECT: the 895900.
emergence of a powerful clinical method. J Nucl Med 69 Davis SM, Chua M, Lichtenstein M, Rossiter SC, Tress
1992; 33: 18881904. BM, Hopper JL. Perfusion changes with recovery after
57 Masdeu JC, Brass LM, Holman BL, Kushner MJ. Brain stroke. Aust NZ J Med 1993; 23: 68.
single-photon emission computed tomography. 70 Laloux P, Richelle F, Jamart J, De Coster P, Laterre C.
Neurology 1994; 44: 278285. Comparative correlations of HMPAO SPECT indices,
58 Fayad PB, Brass LM. Single photon emission com- neurological score, and stroke subtypes with clinical
puted tomography in cerebrovascular disease. Stroke outcome in acute carotid infarcts. Stroke 1997; 26:
1991; 22: 950954. 816821.
59 Infeld B, Binns D, Lichtenstein M, Hopper JL, Davis 71 Bowler JV, Wade DT, Jones BE, Nijran K, Steiner TJ.
SM. Volumetric analysis of cerebral hypoperfusion on Single-photon emission computed tomography using
SPECT: validation and reliability. J Nucl Med 1997; 38: hexamethylpropyleneamine oxime in the prognosis of
14471453. acute cerebral infarction. Stroke 1996; 27: 8286.
60 Alexandrov AV, Black SE, Ehrlich LE et al. Simple visual 72 Lvblad K-O, Jakob PM, Chen Q et al. Turbo spin-echo
analysis of brain perfusion on HMPAO SPECT predicts diusion-weighted MR of ischemic stroke. Am J
early outcome in acute stroke. Stroke 1996; 27: Neuroradiol 1998; 19: 201208.
15371542. 73 Ueda S, Sakaki S, Yuh WTC, Nochide I, Ohta S.
61 Alexandrov AV, Bladin CF, Ehrlich LE, Norris JW. Outcome in acute stroke with successful intra-arterial
Noninvasive assessment of intracranial perfusion thrombolysis and predictive value on initial single-
in acute cerebral ischemia. J Neuroimaging 1995; 5: photon emission-computed tomography. J Cereb
7682. Blood Flow Metab 1999; 19: 99108.
62 Alexandrov AV, Grotta JC, Davis SM, Lassen NA. Brain 74 Alexandrov AV, Black SE, Ehrlich LE, Caldwell CB,
SPECT and thrombolysis in acute ischemic stroke: Norris JW. Predictors of hemorrhagic transformation
time for a clinical trial. J Nucl Med 1996; 37: occurring spontaneously and on anticoagulants in
12591262. patients with acute ischemic stroke. Stroke 1997; 28:
63 Baird AE, Austin MC, McKay WJ, Donnan GA. 11981202.
Sensitivity and specicity of 99Tc-HMPAO SPECT cere- 75 Ueda T, Hatakeyama T, Kumon Y, Sakaki S, Uraoka T.
bral perfusion measurements during the rst 48 hours Evaluation of risk of hemorrhagic transformation in
for the localization of cerebral infarction. Stroke 1997; local intra-arterial thrombolysis in acute ischemic
28: 976980. stroke by initial SPECT. Stroke 1994; 25: 298303.
Limitations of current brain imaging modalities in stroke 29

76 Davis SM, Chua M, Lichtenstein M, Rossiter SC, 88 Frackowiak RS, Lenzi GL, Jones T, Heather JD.
Hopper JL, Tress BM. Cerebral hypoperfusion in Quantitative measurement of regional cerebral blood
stroke prognosis and brain recovery. Stroke 1993; 24: ow and oxygen metabolism in man using 15O and
16911696. positron emission tomography: theory, procedure,
77 Limburg M, van Royen EA, Hijdra A, de Bruine JF, and normal values. J Comput Assist Tomogr 1980; 4:
Verbeeten B, Jr. Single-photon emission computed 727736.
tomography and early death in acute ischemic stroke. 89 Mintun MA, Raichle ME, Martin WR, Herscovitch P.
Stroke 1991; 21: 11501155. Brain oxygen utilization measured with O-15 radio-
78 Hanson SK, Grotta JC, Rhoades H et al. Value of single- tracers and positron emission tomography. J Nucl Med
photon emission-computed tomography in acute 1984; 25: 177187.
stroke therapeutic trials. Stroke 1993; 24: 13221329. 90 Ohta S, Meyer E, Thompson CJ, Gjedde A. Oxygen
79 Overgaard K, Sperling B, Boysen G et al. Thrombolytic consumption of the living human brain measured
therapy in acute ischemic stroke. A Danish pilot study. after a single inhalation of positron emitting oxygen. J
Stroke 1993; 24: 14391446. Cereb Blood Flow Metab 1992; 12: 179192.
80 Alexandrov AV, Masdeu JC, Devous MD, Black SE, 91 Sokolo L, Reivich M, Kennedy C et al. The
Grotta JC. Brain single-photon emission CT with [14C]deoxyglucose method for the measurement of
HMPAO and safety of thrombolytic therapy in acute local cerebral glucose utilization: theory, procedure,
ischemic stroke. Proceedings of the meeting of the and normal values in the conscious and anesthetized
SPECT safe thrombolysis study collaborators and the albino rat. J Neurochem 1977; 28: 897916.
members of the brain imaging council of nuclear 92 Heiss WD, Graf R, Fujita T et al. Early detection of irre-
medicine. Stroke 1997; 28: 18301834. versibly damaged ischemic tissue by umazenil posi-
81 Grotta JC, Alexandrov AV. tPA-associated reperfusion tron emission tomography in cats. Stroke 1997; 28:
after acute stroke demonstrated by SPECT. Stroke 20452051; discussion 20512052.
1998; 29: 429432. 93 Read SJ, Hirano T, Abbott DF et al. The fate of hypoxic
82 Eberl S, Zimmerman RD. Radiation protection and tissue on 18F-uoromisonidazole positron emission
dosimetry in clinical practice. In: Murray I, Ell P, eds. tomography after ischemic stroke. Ann Neurol 2000;
Nuclear Medicine in Clinical Diagnosis and 48: 228235.
Treatment. Edinburgh: Churchill Livingstone; 1994: 94 Assessment: positron emission tomography. Report of
13671388. the Therapeutics and Technology Assessment
83 Kety SS, Schmidt CF. The nitrous oxide method for the Subcommittee of the American Academy of
quantitative determination of cerebral blood ow in Neurology. Neurology 1991; 41: 163167.
man: theory, procedure and normal values. J Clin 95 Hakim AM, Evans AC, Berger L et al. The eect of
Invest 1946; 25: 476482. nimodipine on the evolution of human cerebral
84 Eichling JO, Raichle ME, Grubb RL, Jr., Ter-Pogossian infarction studied by PET. J Cereb Blood Flow Metab
MM. Evidence of the limitations of water as a freely 1989; 9: 523534.
diusible tracer in brain of the rhesus monkey. Circ 96 Yamauchi H, Fukuyama H, Ogawa M, Ouchi Y, Kimura
Res 1974; 35: 358364. J. Hemodilution improves cerebral hemodynamics in
85 Kety SS. Measurement of local blood ow by the internal carotid artery occlusion. Stroke 1993; 24:
exchange of an inert, diusible substance. Med Res 18851890.
1960; 8: 228236. 97 Heiss W-D, Holtho V, Pawlik G, Neveling M. Eect of
86 Herscovitch P, Markham J, Raichle ME. Brain blood nimodipine on regional cerebral glucose metabolism
ow measured with intravenous H2(15)O. I. Theory in patients with acute ischemic stroke as measured by
and error analysis. J Nucl Med 1983; 24: 782789. positron emission tomography. J Cereb Blood Flow
87 Raichle ME, Martin WR, Herscovitch P, Mintun MA, Metab 1990; 10: 127132.
Markham J. Brain blood ow measured with intrave- 98 Heiss WD, Grond M, Thiel A, von Stockhausen HM,
nous H2(15)O. II. Implementation and validation. J Rudolf J. Ischaemic brain tissue salvaged from infarc-
Nucl Med 1983; 24: 790798. tion with alteplase. Lancet 1997; 349: 15991600.
Clinical efficacy of CT in acute cerebral ischemia

Rdiger von Kummer

Department of Neuroradiology, University of Technology, Dresden, Saxonia, Germany

Patients with acute cerebral ischemia represent it seems as if hyperattenuating lesions are easier to
with hemiparesis, hemianopia, speech disturbance, detect than hypoattenuating lesions.
or impairment of consciousness. The dierential
diagnosis is intracranial hemorrhage, cerebral
CT detection of intracranial blood
venous thrombosis, focal encephalitis, demyelina-
tion disorder or tumour. Brain imaging is necessary In patients with acute stroke, blood may be present
to assess the exact diagnosis and the acute patho- in one or more cranial compartments: brain paren-
physiological state of the brain. Both pieces of chyma, ventricles, subarachnoid space, and subdu-
information will guide treatment and will nally ral or epidural space. Clinically, an acute
determine the clinical outcome of the patient. parenchymal hemorrhage cannot reliably be distin-
Kent and Larson proposed ve levels of clinical guished from ischemic stroke. After acute hemor-
ecacy for assessing diagnostic technology: (i) rhage, blood appears as a hyperattenuated, often
technical capacity; (ii) diagnostic accuracy; (iii) space-occupying mass. The degree of hyperattenua-
diagnostic impact; (iv) therapeutic impact; and (v) tion depends on the amount of blood, whether it is
patient outcome.1 In this chapter, I will study the clotted or not, and whether the blood is intermixed
question what unenhanced CT is able to assess in with cerebrospinal uid (CSF) or brain tissue.
patients with acute stroke; how accurate this infor- Hemorrhages related to coagulopathies or treat-
mation is; and whether imaging with CT has any ment with anticoagulants or thrombolytics are often
impact on stroke diagnosis, stroke treatment and, inhomogeneous with uid levels (Fig. 3.1). The sen-
nally, on the clinical outcome of patients. sitivity of CT for detection of parenchymal hemor-
rhage is considered to be almost 100%, but small
hemorrhages into the brain parenchyma or sub-
Level 1 of clinical efficacy: technical capacity of arachnoid space can be missed. The investigators of
CT in acute cerebral ischemia the European Cooperative Acute Stroke Studies
(ECASS I and II) rarely missed a small parenchymal
Technical capacity is the capability of CT to repro- hemorrhage and a subarachnoid hemorrhage (SAH)
ducibly display recognizable images that demon- in 1420 patients (0.1%) that was later detected by
strate pathology with good intra- and interobserver neuroradiologists. Emergency physicians in another
reliability.2 Based on changes in X-ray attenuation, study had an error rate in stroke detection by CT
CT is capable of detecting intracranial hemorrhage, twice that of neurologists and radiologists, and only
thrombo-embolic occlusion of major brain arteries, 17% of emergency physicians, 40% of neurologists
brain tissue swelling without edema, and ischemic and 52% of radiologists achieved a 100% sensitivity
brain edema. Intra- and interobserver reliability for the identication of intracranial hemorrhage.3
was not studied for all of these ndings. Generally, This observation underlines that some hemorrhages

32 Rdiger von Kummer

Fig. 3.1. CT of a 79-year-old man with a stroke after myocar-

dial infarction and thrombolytic therapy shows multiple
hematomas with uid levels (arrows).

are subtle, interobserver reliability does not only

depend on the technical capacity of CT , and that
CT reading requires special training and experience.
Acute hemorrhages usually show hyperattenua-
tion without surrounding edema. If marked edema
is present under those circumstances, underlying
neoplasm or venous obstruction should be sus- Fig. 3.2. Patient with left-sided hemiparesis since less than 5
pected. The location of the hematoma often pro- hours. CT depicts hyperattenuating right middle cerebral
artery (MCA) trunk (arrow). The left MCA trunk (arrowhead)
vides clues about its underlying etiology. Multiple
is not hyperattenuating.
hemorrhagic lesions should make one think about
metastatic disease, coagulopathy or cerebral
amyloid angiopathy. on unenhanced CT (Fig. 3.2). The interobserver
CT is 90% sensitive for the detection of SAH within agreement on such hyperdense artery signs varied
the rst 24 hours of bleeding. When blood later between poor and moderate ( 0.20 and 0.63).57
intermixes with cerebrospinal uid, the density will
be similar to the adjacent brain and dicult to visu-
CT detection of brain tissue swelling
alize. If the hemorrhage is small, it may be missed
entirely by the scan, necessitating lumbar puncture The enlargement of anatomical structures like the
for denitive diagnosis in patients with a history and cerebral cortex and the eacement of cerebral
a syndrome highly suspicious for SAH. The sensitiv- spinal uid spaces suggest brain tissue swelling. CT
ity of CT in detecting subarachnoid blood declines to may detect brain tissue swelling without hypoat-
approximately 50% at 1 week after SAH.4 tenuation for a short period early after arterial or
Calcication of the basal ganglia can occasionally venous obstruction (Fig. 3.3). Compensatory arte-
be mistaken for a deep intraparenchymal hemor- rial dilatation due to low perfusion pressure8 or
rhage. It has a similar degree of attenuation to acute passive arterial dilatation due to high venous pres-
blood, but may be distinguished by its characteris- sure9 cause this type of swelling. Six neuroradiolo-
tic location and tendency to be bilateral. gists agreed on tissue swelling in 45 CT of acute
stroke patients with a 0.56 0.596.

CT detection of arterial obstruction

CT detection of ischemic brain edema
Thrombo-embolic occlusion of major brain arteries
may be represented as a hyperattenuating arterial Brain tissue swelling with hypoattenuation is best
segment in comparison to other arterial segments explained by ischemic edema (Fig. 3.4). Cerebral
Clinical efficacy of CT in acute cerebral ischemia 33

(a) (b)

Fig. 3.3. (a) CT obtained 2 hours after symptom onset in a 58-year-old man with hyperattenuating right MCA trunk (not shown)
shows swelling of the cortical layer within the right MCA territory (arrows). (b) 6 days later the swelling is still present in addition
to a subcortical ischemic infarct. The patient was treated with IV rt-PA. His NIH stroke Score improved from 17 at baseline to 11 at
day 90 after stroke. (Courtesy of Professor Otto Busse, Minden, Germany).

blood ow (CBF) of 812 ml /100 g per min is 2.6 HU in gels20 and by 2.1 HU in experimental cryo-
accepted as the ow threshold for structural integ- genically induced brain edema.19 Co-registration of
rity.10 A sudden decrease in cerebral perfusion CT attenuation and CBF revealed that hypoattenu-
below 812 ml/100 g per min causes the grey ation develops only in areas of critically hypoper-
matter to immediately take up water.1114 (The CBF fused brain tissue.21 The mean CBF in the aected
threshold that triggers a decrease in the ADC is MCA territory of patients with hypoattenuation of
considerably higher: 3040 ml/100 g per min.15) the basal ganglia (7 ml per 100 g and min) was sig-
The amount of water accumulating during ische- nicantly lower than that of patients who did not
mia is signicantly correlated to the duration of have these ndings (17 ml per 100 g and min).22 A
ischemia. The mean ischemic blood ow was 5.8 decline in cerebral blood volume (CBV) may add to
0.4 ml per 100 g and min.16 Signicant resolution the decrease in X-ray attenuation.
of brain edema occurred only if ischemia lasted Using a CT window width of 80 HU, the
less than 15 minutes.13 This early type of ischemic minimal visible contrast is at about 4 HU, which
edema thus indicates the tissue with ow values corresponds with an increase in brain tissue water
below the threshold of structural integrity and may content of about 1.5%. This suggests that the very
herald ischemic necrosis. Imaging of ischemic rst and potentially reversible stage of the devel-
edema could then identify the proportion of oping ischemic edema cannot be seen on CT
ischemic brain tissue that is irreversibly because the decline of attenuation has not
damaged.17 reached the contrast resolution of 4 HU. In other
CT attenuation is linearly proportional to specic words: a normal CT in a patient with stroke
gravity and thus allows monitoring of tissue water excludes hemorrhage and other disease, but
content.18,19 An increase by 1% of tissue water cannot exclude ischemic edema at a very early
content causes a decrease of X-ray attenuation by stage. This insensitivity of CT for the very rst
34 Rdiger von Kummer



Fig. 3.4. CT within 6 hours of the sudden onset of left-sided hemiparesis. (a) The two sections show hypoattenuating tissue
(arrows) within the territory of the pericallosal artery and of an MCA branch and in addition marked eacement of sulci due to
ischemic brain tissue swelling. (b) Three days later the follow-up CT shows a well-marked cerebral infarct in the same territory.
Tissue swelling is still present.
Clinical efficacy of CT in acute cerebral ischemia 35

stage of ischemic edema may have an advantage: with MCA trunk occlusion during the rst 2 hours.
edema at that stage may be potentially reversible, The incidence of positive CT ndings increased to
so that CT depicts ischemic edema only at its irre- 89% in the third hour after symptom onset and to
versible stage. Hypoattenuation of brain paren- 100% thereafter.35 In another series of patients with
chyma on CT after arterial occlusion indicates hemispheric stroke, the incidence of early CT signs
severe ischemia under the critical level of structu- of infarction was 82%.36 In patients selected for
ral integrity for 1 to 3 hours. Under clinical condi- thrombolytic therapy, 12 of 23 patients (52%) had a
tions, hypoattenuation may appear by 22 minutes parenchymal hypodensity on the CT performed
after the onset of symptoms.17 It has not yet been within 3 hours of stroke onset.21 In a series of 100
studied, however, whether hypoattenuation of consecutive patients with MCA infarction, CT
ischemic brain tissue can disappear if the tissue is detected hypodensity of the lentiform nucleus in
reperfused less than 3060 minutes after stroke 48% and of the insular cortex in 59% of the patients
onset and to what extent changes in CBV contrib- within 14 hours of stroke onset.37 In a recent ran-
ute to changes in CT attenuation. domized trial on pro-urokinase in patients with
Because of its subtlety, early ischemic edema is MCA occlusions, the investigators detected infarct
recognized with only moderate interobserver reli- signs in 125 of 171 patients (73%) on CT within the
ability.6,7,23 The National Institute of Neurological rst 6 hours.38
Disorders and Stroke rt-PA stroke trialists and In summary, CT has the capacity to identify
others demonstrated that training in CT reading dierent pathophysiological states that all results in
considerably aects the sensitivity to detect the same clinical picture of an acute stroke syn-
ischemic edema.23,24 Attempts were made to drome: intracranial hemorrhage, ischemic edema
improve the capability of CT in detecting ischemic and ischemia without ischemic edema. In addition,
edema by performing a density-dierence analysis arterial occlusion is detected in some patients.
between both cerebral hemispheres, by varying Because these CT ndings represent dierent
window width and centre level, and by using a pathophysiological states, it is incorrect to address
quantitative score.2527 them all as early infarct signs. Moreover, it makes
The frequency of positive CT ndings in acute no sense, to study and count certain CT signs, like
stroke may also be diminished by the prejudice that obscuration of the lentiform nucleus, loss of the
CT is negative within the rst 24 to 48 hours after insular ribbon, loss of greywhite matter distinc-
stroke that is still discussed in review articles28 and tion, hypodensity, because all these phenomena
books.29 In contrast, many studies have described are caused by hypoattenuating grey matter due to
positive ndings within the rst 36 hours of stroke ischemic edema.
onset: Tomura et al. studied 25 patients with
embolic cerebral infarction between 40 and 340
minutes after the onset of symptoms.30 Twenty- Level 2 of clinical efficacy: diagnostic accuracy
three CT scans (92%) were positive with obscura- of CT in acute cerebral ischemia
tion of the lentiform nucleus caused by
hypodensity.30 Bozzao et al. observed parenchymal Diagnostic accuracy is the accuracy in detecting
hypodensity in 25 of 36 patients (69%).31 A loss of and classifying pathology, measured in true and
the insular ribbon was reported in 23 of 27 (85%) false positive ratios.2 The optimal design for assess-
patients.32 Horowitz et al. reported on hypodensity ing the diagnostic accuracy of CT is a prospective,
and mass eect in 56% of 50 scans.33 When compar- blinded comparison to a reference standard in a
ing MR vs. CT imaging in identical patients within consecutive series of patients from a relevant clini-
3 hours of symptom onset, CT was positive in 19 cal population.39 Because CT was the rst modality
patients (53%) and MRI in 18 (50%) patients with that could image the brain in vivo, a reference stan-
hemispheric stroke.34 We reported 17 positive CT dard for CT ndings in acute cerebral ischemia has
scans (68%) performed in a series of 25 patients seldom been available.
36 Rdiger von Kummer

Table 3.1. Cerebral hemorrhagic events in the placebo groups of sion. We studied the hyperdense middle cerebral
randomized trials on thrombolysis artery sign (HMCAS) in 53 patients with DSA
proven unilateral MCA trunk occlusions.35 The
Cerebral hemorrhagic events HMCAS was positive in 25 patients (sensitivity:
All Symptomatic fatal 47%) and 100% specic (no false-positive ndings).
Trial n (%) (%) (%) Tomsick et al.47 described a sensitivity of 79% and a
specicity of 93% in 25 patients. The HMCAS may
ECASS I 307 37 3 2
be false-positive in patients with calcication of the
NINDS I and II 312 4 1 0.3
MAST E 154 40 3 n.a.
arterial wall or high hematocrit (Fig. 3.5).
MAST I 156 10 1 n.a.
ASK 166 17 3 2
Diagnostic accuracy of CT in detecting ischemic brain
ECASS II 391 40 6 1
tissue swelling
PROACT II 59 57 4 n.a.
For the discussion of false-negative CT in acute
Notes: ECASS, European Cooperative Acute Stroke Study;
stroke, it is important to recognize that no reference
NINDS, National Institute of Neurological Disorder and
Stroke rt-PA Study; MAST-E, Multicentre Acute Stroke
standard exists to determine the accuracy of CT in
TrialEurope; MAST-I, Multicentre Acute Stroke Trial assessing brain tissue swelling with and without
Italy; ASK, Australian Streptokinase Trial; PROACT, Prolyse edema. A normal CT without any hypoattenuation
in Acute Cerebral Thromboembolism Trial. of the brain parenchyma in a patient with acute
stroke may be true negative and represent a distur-
bance of brain perfusion above the level that
Diagnostic accuracy of CT in detecting intracranial
induces edema. Such a CT could be false negative if
the hypoattenuation is too subtle or too small to be
Surgery or autopsy regularly conrms the CT detected by CT. Another issue is whether hypoat-
nding of intracranial hemorrhage. To my knowl- tenuation of tissue on early CT in stroke patients is
edge, the rate of false negative CT in brain hemor- specic for irreversible damage. This was recently
rhage has not been extensively studied. prospectively studied in 786 patients.17 In this
Presumably, CT does not detect subtle and small study, CT predicted ischemic necrosis in 449
hemorrhages into the brain parenchyma or CSF patients that was conrmed by follow-up CT in 433
space. The frequency of post-ischemic hemorrhagic patients (positive predictive value96%, specicity
transformations is higher in pathological series 85%). The false-positive ndings in 16 patients
than with serial CT40 and varies among the placebo were explained in retrospect by image artefacts due
groups of major stroke trials (Table 3.1). The latter to poor technique. A true normalization of prior
observations support the impression that the fre- hypoattenuating tissue was not observed.
quency of post-ischemic transformation is aected
by the CT technique, the time and frequency of CT
studies, the denition of a hemorrhage on CT, the Level 3 of clinical efficacy: diagnostic impact of
experience in CT assessment and the patient popu- CT in acute cerebral ischemia
lation under study.
It seems as if advances in imaging are being made
faster than the ability to fully evaluate their diagnos-
Diagnostic accuracy of CT in detecting arterial
tic and therapeutic impact. Diagnostic impact is the
accuracy and clinical value compared with existing
The determination of the specicity of a hyperat- alternatives.2 MRI was introduced into clinical prac-
tenuating arterial segment requires the comparison tice as the second modality that can image brain
with digital subtraction angiography (DSA) as a parenchyma, and is now considered as an evolving
gold-standard for the assessment of arterial occlu- standard of care in acute stroke.41 It is an unresolved
Clinical efficacy of CT in acute cerebral ischemia 37

Fig. 3.5. CT of a 19-year-old man with Down syndrome, heart failure, and a hematocrit of 71% shows a hyperattenuating left MCA
(arrow) and hyperattenuating other vessels, although contrast was not given.

issue, however, whether CT or MRI has a higher a gelatin model, there was a linear relationship
clinical ecacy in patients with acute stroke. between water content, CT attenuation, T1, and T2
relaxation time. An increase in water content of 6%,
however, resulted in only a 19% increase in signal
Diagnostic impact on intracranial hemorrhage
intensity of gel on the T2-weighted image, but in a
CT was the rst imaging modality that could reliably 25% change in CT attenuation in the same speci-
dierentiate between hemorrhagic and ischemic mens.20 In an experiment with boiled eggs, Unger et
stroke. It seems as if susceptibility weighted MR al. demonstrated that T2 relaxation is not only
sequences have the same sensitivity to detect brain aected by water content. Egg hardening caused
hemorrhage in acute stroke.42 Again, no randomized dramatic T2 shortening and modest T1 shortening,
comparison has been done so far, to my knowledge, but had no eect on CT attenuation,20 suggesting
to decide whether CT can be completely replaced by that the proportion of bound or structured water
MRI in acute stroke. It is another controversy, aects T2 relaxation. It is still unclear, however,
whether special MR sequences like FLAIR, have the whether the eects of water structure play a role in
same accuracy as CT in detecting SAH.4345 the acute ischemic brain. Typical aws of studies
comparing CT with MRI were no reference stan-
dard, or a poorly dened one,4649 and MRI per-
Diagnostic impact on ischemic edema
formed considerably later than CT.47,49 To our
The diagnostic accuracy and clinical value of both knowledge, no study compared CT and MRI in a
techniques have rarely been properly compared. In randomized manner. Not surprisingly, some studies
38 Rdiger von Kummer

(a) Fig. 3.6. (a) CT performed 90 minutes after onset of left-sided

hemiparesis shows a wedge shaped area of hypoattenuating
brain tissue (arrows). (b) The MRI was performed 2 hours
after symptom onset. The T2W image does not show any
pathology in this area. The DWI shows disturbed diusion
within the area depicted by CT 30 minutes earlier.

Clinical efficacy of CT in acute cerebral ischemia 39

Fig. 3.7. Two CT scans within 4 days of a 73-year-old man with occlusion of the left internal carotid artery and progressive right-
sided hemiparesis and aphasia. The ischemic lesions are mainly located in the end-supply area of the left MCA (arrows).

found that MRI including diusion-weighted edema detected by CT. We showed that the extent of
imaging (DWI) is more sensitive than CT4649, hypoattenuating brain tissue due to ischemic
whereas other studies did not show that MRI is edema is associated with the neurological score
superior to CT in detecting ischemic changes of and with clinical outcome.17,35,5456 Fiorelli et al.
cerebral parenchyma.34,50 concluded from their experience with emergency
CT was used to dierentiate among types of cere- CT in acute ischemic stroke that it adds signi-
bral ischemia like territorial infarcts (Fig. 3.6) cantly to the prediction of clinical outcome made
caused by emboli, infarcts in end-supply areas or on clinical grounds.57
watershed areas (Fig. 3.7) often associated with
major artery occlusion or stenosis, and dissemi-
nated small infarcts (Fig. 3.8) caused by small Level 4 of clinical efficacy: therapeutic impact
vessel disease.51 Others have challenged the of CT in acute cerebral ischemia
concept that stroke mechanisms can be inferred
from interpretation of lesion patterns on brain Therapeutic impact is the eect of the diagnostic
scans.52,53 test on patient care, on prognostic information, on
Another approach to the diagnostic impact of CT treatment selection, and on the costs and risks of
is the interpretation of the extent of ischemic the diagnostic process.2
40 Rdiger von Kummer

has an increased risk for brain hemorrhage.55,59 This

hypothesis was conrmed by ECASS II.60,61
Moreover, the ECASS II data showed, that acute
stroke patients with a normal CT have a less severe
neurological decit and a better clinical course
compared to patients with hypoattenuated brain
tissue. This is illustrated in Fig. 3.9 showing the
NIHSS over time in the ECASS I and II populations
in relationship with the ndings on their baseline

Level 5 of clinical efficacy: impact on patient

outcome of CT in acute cerebral ischemia

Impact on patient outcome is the most important

criterion for the usefulness of a diagnostic test
showing that it leads to a change in patient man-
agement that improves patient outcome.2 ECASS I
and II are the only studies with prospectively
dened categories for the extent of ischemic edema
on baseline CT. We will use here the combined
Fig. 3.8. 59-year-old man with acute mild stroke since 3
dataset of both trials to study whether the response
hours. CT shows multiple small subcortical ischemic lesions
to rt-PA treatment was dierent in patients with a
(arrows) typical for dissiminated small vessel disease.
normal CT compared to patients with a small
region of edema ( 33% of MCA territory) (Fig. 3.10)
For the rst time in history, CT enabled the iden- and compared to patients with more extensive
tication of patients suering from cerebral ische- edema (> 33% of MCA territory) (Fig. 3.11). This
mia. As a consequence, specic treatment like dataset includes 1350 patients. According to the
thrombolysis could be tested. CT has thus an enor- common assessment of 3 neuroradiologists, 688
mous therapeutic impact just by dierentiating patients had a normal CT at baseline, 618 patients
between ischemic and hemorrhagic stroke. had a small ischemic edema, and only 89 patients
In acute cerebral ischemia, the only treatment had extensive ischemic edema. (Such large edema
that is proved to improve clinical outcome is intra- was an exclusion criterion in ECASS I and II.) Figure
venous thrombolysis with rt-PA applied within 3 3.9 shows that only patients with small ischemic
hours of symptom onset or intra-arterial infusion of edema had benet from rt-PA treatment. The odds
pro-urokinase in patients with MCA occlusion if ratio for a benecial outcome without a functional
applied within 6 hours of symptom onset.38,58 decit (Rankin 0 and 1) at 3 months was 1.20 (95%
Whereas the NINDS rt-PA Study Group used CT to CI, 0.89 1.63) for patients with normal CT, 1.47
exclude patients with intracranial hemorrhage, the (95% CI, 1.03 2.09) for patients with small
PROACT investigators excluded patients from the ischemic edema, and 0.71 (95% CI, 0.21 2.41) for
study who had a hypoattenuated area on CT patients with large edema. This suggests that rt-PA
exceeding one third of the MCA territory. The rst treatment may enhance the chance for stroke
ECASS study supported the hypothesis that patients by almost 50% to not be disabled after 3
patients with such a region of extensive ischemic months. In contrast, rt-PA treatment did not
edema did not benet from rt-PA treatment and demonstrate a benecial effect in patients with
Clinical efficacy of CT in acute cerebral ischemia 41

> 33%, placebo
30 > 33%, rt-PA
33%, placebo
27.5 33% rt-PA
no, placebo
no, rt-PA








NIH-B NIH-24 NIH-90d

Fig. 3.9. Baseline CT ndings in ECASS I and II and the response to rt-PA. NIHSS, National Institute of Health Stroke Scale; NIH-B,
NIHSS at baseline; NIH-24, NIHSS at 24 hours after stroke onset; NIH-90d, NIHSS at 3 months after stroke onset.

normal CT, and may be deleterious in patients with research methods, small sample sizes, or incom-
a large region of ischemic edema. The last group plete reporting; (d) studies with multiple aws in
was, however, too small to really prove the impact research methods or reports of opinion unsubstan-
of this CT nding on patient outcome. tiated by data.
Kent and Larson proposed four grades of method- ECASS II was a double-blinded randomized
ologic quality for studies of diagnostic tests:1 multicentre-controlled trial with a prospective
(a) studies with broad generalizability to a variety of protocol of CT assessment. This trial provides high-
patients and no signicant aws in research quality evidence (grade B evidence) that the CT
methods; (b) studies with a narrower spectrum of nding of hypoattenuating brain tissue in less than
generalizability and with only a few aws that are one-third of the MCA territory is associated with a
well described so that their impact on conclusions benecial eect of rt-PA treatment in ischemic
can be assessed; (c) studies with several aws in stroke.
42 Rdiger von Kummer



Fig. 3.10. Two CT scans in a 42-year-old man. (a) The rst scan was obtained 2 hours after stroke onset and shows a hyperattenu-
ated left MCA trunk (long arrow) and branches and a hypoattenuating left insular cortex and putamen (short arrows). (b) The
second scan was obtained 24 hours later. The hypoattenuating brain tissue is still restricted to the left insular cortex and putamen
(short arrows) despite a hyperattenuated left MCA (long arrow).
Clinical efficacy of CT in acute cerebral ischemia 43

(a) (b)

Fig. 3.11. Two CT scans in a 68-year-old man. (a) The rst scan obtained 140 minutes after stroke onset shows an extended area of
hypoattenuated brain tissue (arrows) covering the entire right MCA territory in this section. (b) Seven days later and after placebo
treatment, the follow-up scan shows a complete right MCA infarction with slight hemorrhagic transformation.

REFERENCES 7 Marks M, Holmgren E, Fox A, Patel S, von Kummer R,

Froehlich J. Evaluation of early computed tomo-
1 Kent D, Larson E. Disease, level of impact, and quality graphic ndings in acute ischemic stroke. Stroke 1999;
of research methods; three dimensions of clinical e- 30: 389392.
cacy assessment applied to magnetic resonance 8 Gibbs J, Wise R, Leenders K, Jones T. Evaluation of
imaging. Invest Radiol 1992; 27: 245254. cerebral perfusion reserve in patients with carotid-
2 Powers W. Testing a test. A report card for DWI in acute artery occlusion. Lancet 1984; 8372: 310314.
stroke. Neurology 2000; 54: 15491551. 9 Yuh W, Simonson T, Wang A et al. Venous sinus occlu-
3 Schriger D, Kalafut M, Starkman S, Krueger M, Daver J. sive disease: MR ndings. Am J Neuroradiol 1994; 15:
Cranial computed tomography interpretation in acute 309316.
stroke. Physician accuracy in determining eligibility 10 Hossmann KA. Viability thresholds and the penumbra
for thrombolytic therapy. J Am Med Assoc 1998; 279: of focal ischemia. Ann Neurol 1994; 36: 557565.
12931297. 11 Watanabe O, West C, Bremer A. Experimental regional
4 Schievink W. Intracranial aneurysms. N Engl J Med cerebral ischemia in the middle cerebral artery terri-
1997; 336: 2840. tory in primates. Part 2: Eects on brain water and
5 Tomsick T, Brott T, Chambers A, et al. Hyperdense electrolytes in the early phase of MCA stroke. Stroke
middle artery sign on CT: ecacy in detecting middle 1977; 8: 7176.
cerebral artery thrombosis. Am J Neuroradiol 1990; 11: 12 Schuier FJ, Hossmann KA. Experimental brain infarcts
473477. in cats. II. Ischemic brain edema. Stroke 1980; 11:
6 von Kummer R, Holle R, Grzyska U et al. Interobserver 593601.
agreement in assessing early CT signs of middle cere- 13 Todd N, Picozzi P, Crockard A, Ross Russel R. Duration
bral artery infarction. Am J Neuroradiol 1996; 17: of ischemia inuences the development and resolution
17431748. of ischemic brain edema. Stroke 1986; 17: 466471.
44 Rdiger von Kummer

14 Gotoh O, Asano T, Koide T, Takakura K. Ischemic brain and center level settings. Radiology 1999; 213:
edema following occlusion of the middle cerebral 150155.
artery in the rat. I: The time courses of the brain water, 27 Barber P, Demchuk A, Zhang J, Buchan A. Validity and
sodium, and potassium contents and bloodbrain- reliability of a quantitative computed tomography
barrier permeability to 125I-Albumin. Stroke 1985; 16: score in predicting outcome of hyperacute stroke
101109. before thrombolytic therapy. Lancet 2000; 355:
15 Wang Y, Hu W, Perez-Trepichio A et al. Brain tissue 16701674.
sodium is a ticking clock telling time after arterial 28. Gilman S. Imaging of the brain. N Engl J Med 1998;
occlusion in rat focal cerebral ischemia. Stroke 2000; 338: 812820.
31: 13861392. 29 Sorensen A, Reimer P. Cerebral MR Perfusion Imaging.
16 Todd N, Picozzi P, Crockard A, Ross Russel R. Stuttgart, New York: Thieme; 2000.
Reperfusion after cerebral ischemia: Inuence of 30 Tomura N, Uemura K, Inugami A, Fujita H, Higano S,
duration of ischemia. Stroke 1986; 17: 460465. Shishido F. Early CT nding in cerebral infarction.
17 von Kummer R, Bourquain H, Bastianello S et al. Early Radiology 1988; 168: 463467.
prediction of irreversible brain damage after ischemic 31 Bozzao L, Bastianello S, Fantozzi LM, Angeloni U,
stroke by computed tomography. Radiology 2001; 219: Argentino C, Fieschi C. Correlation of angiographic
95100. and sequential CT ndings in patients with evolving
18 Phelps M, Gado M, Homan E. Correlation of eective cerebral infarction. Am J Neuroradiol 1989; 10:
anatomic number and electron density with attenua- 12151222.
tion coecients measured with polychromatic X-rays. 32 Truwit C, Barkovich A, Gean-Marton A, Hibri N,
Radiology 1975; 117: 585588. Norman D. Loss of the insular ribbon: Another early
19 Rieth KG, Fujiwara K, Di Chiro G et al. Serial measure- CT sign of acute middle cerebral artery infarction.
ments of CT attenuation and specic gravity in Radiology 1990; 176: 801806.
experimental cerebral edema. Radiology 1980; 135: 33 Horowitz SH, Zito JL, Donnarumma R, Patel M, Alvir J.
343348. Computed tomographic angiographic ndings
20 Unger E, Littleeld J, Gado M. Water content and within the rst ve hours of cerebral infarction. Stroke
water structure in CT and MR signal changes: Possible 1991; 22: 12451253.
inuence in detection of early stroke. Am J 34 Mohr J, Biller J, Hilal S et al. Magnetic resonance
Neuroradiol 1988; 9: 687691. versus computed tomographic imaging in acute
21 Grond M, von Kummer R, Sobesky J, Schmlling S, stroke. Stroke 1995; 26: 807812.
Heiss W-D. Early computed-tomography abnormal- 35 von Kummer R, Meyding-Lamad U, Forsting M et al.
ities in acute stroke. Lancet 1997; 350: 15951596. Sensitivity and prognostic value of early computed
22 Firlik A, Kaufmann A, Wechsler L, Firlik K, Fukui M, tomography in middle cerebral artery trunk occlu-
Yonas H. Quantitative cerebral blood ow determina- sion. Am J Neuroradiol 1994; 15: 915.
tions in acute ischemic stroke. Relationship to com- 36 von Kummer R, Nolte PN, Schnittger H, Thron A,
puted tomography and angiography. Stroke 1997; 28: Ringelstein EB. Detectability of hemispheric ischemic
22082213. infarction by computed tomography within 6 hours
23. Grotta J, Chiu D, Lu M et al. Agreement and variability after stroke. Neuroradiology 1996; 38: 3133.
in the interpretation of early CT changes in stroke 37 Moulin T, Cattin F, Crpin-Leblond T et al. Early CT
patients qualifying for intravenous rtPA therapy. signs in acute middle cerebral artery infarction:
Stroke 1999; 30: 15281533. Predictive value for subsequent infarct locations and
24 von Kummer R. Eect of training in reading CT scans outcome. Neurology 1996; 47: 355375.
on patient selection for ECASS II. Neurology 1998; 51 38 Furlan A, Higashida R, Wechsler L et al. Intra-arterial
(Suppl 3): S50S52. prourokinase for acute ischemic stroke. J Am Med
25 Bendszus M, Urbach H, Meyer B, Schultheiss R, Assoc 1999; 282: 20032011.
Solymosi L. Improved CT diagnosis of acute middle 39 Lijmer J, Mol B, Heiserkamp S et al. Empirical evi-
cerebral artery territory infarcts with density- dence of design-related bias in studies of diagnostic
dierence analysis. Neuroradiology 1997; 39: 127131. tests. J Am Med Assoc 1999; 282: 10611066.
26 Lev M, Farkas J, Gemmete J et al. Acute stroke: 40 Hornig C, Dorndorf W, Agnoli A. Hemorrhagic cere-
improved nonenhanced CT detectionbenets of soft- bral infarction: a prospective study. Stroke 1986;
copy interpretation by using variable window width 17: 179185.
Clinical efficacy of CT in acute cerebral ischemia 45

41 Hacke E, Warach S. Diusion-weighted MRI as an 52 Caplan L, Hennerici M. Impaired clearance of emboli

evolving standard of care in acute stroke. Neurology (washout) is an important link between hypoperfu-
2000; 54: 15481549. sion, embolism, and ischemic stroke. Arch Neurol
42 Schellinger P, Jansen O, Fiebach J, Hacke W, Sartor K. A 1998; 55: 14751482.
standardized MRI stroke protocol: comparison with 53 Hennerici M, Daertshofer M, Jakobs L. Failure to
CT in hyperacute intracerebral hemorrhage. Stroke identify cerebral infarct mechanisms from topography
1999; 30: 19741975. of vascular territory lesions. Am J Neuroradiol 1998;
43 Ogawa T, Inugami A, Shimosegawa E et al. 19: 10671074.
Subarachnoid hemorrhage: Evaluation with MR 54 von Kummer R, Bastianello S, Bozzao L, Manelfe C,
imaging. Radiology 1993; 186: 345351. Hacke W. Early prediction of fatal ischemic brain
44 Atlas S. MR imaging is highly sensitive for acute sub- edema. Stroke 1996; 27: 181.
arachnoid hemorrhage . . . Not! Radiology 1993; 186: 55 von Kummer R, Allen K, Holle R et al. Acute stroke:
319322. usefulness of early CT ndings before thrombolytic
45 Noguchi K, Ogawa T, Inugami A et al. Acute subarach- therapy. Radiology 1997; 205: 327333.
noid hemorrhage: MR Imaging with uid-attenuated 56 von Kummer R, Bourquain H, Manelfe C, Bastianello
inversion recovery pulse sequences. Radiology 1995; S, Bozzao L, Meier D. Predictive value of early CT in
196: 773777. acute ischemic stroke. Stroke 1999; 30: 250.
46 Kertesz A, Black S, Nicholson L, Carr T. The sensitivity 57 Fiorelli M, Toni D, Bastianello S et al. Computed
and specicity of MRI in stroke. Neurology 1987; 37: tomography ndings in the rst few hours of ischemic
15801585. stroke: implications for the clinician. J Neurol Sci.
47 Bryan N, Levy L, Whitlow W, Killian J, Preziosi T, 2000; 173: 1017.
Rosario J. Diagnosis of acute cerebral infarction: 58 NINDS Stroke Study Group. Tissue plasminogen acti-
Comparison of CT and MR Imaging. Am J Neuroradiol vator for acute ischemic stroke. N Engl J Med 1995;
1991; 12: 611620. 333: 15811587.
48 Barber P, Darby D, Desmond P et al. Identication of 59 Larrue V, von Kummer R, del Zoppo G, Bluhmki E.
major ischemic change : diusion-weighted imaging Hemorrhagic transformation in acute ischemic stroke.
versus computed tomography. Stroke 1999; 30: Potential contributing factors in the European
20592065. Cooperative Acute Stroke Study. Stroke 1997; 28:
49 Lansberg M, Albers G, Beaulieu C, Marks M. 957960.
Comparison of diusion-weighted MRI and CT in 60 Hacke W, Kaste M, Fieschi C et al. Randomised
acute stroke. Neurology 2000; 54: 15571561. double-blind placebo-controlled trial of thrombolytic
50 Barber P, Demchuk A, Hill M et al. A comparison of CT therapy with intravenous alteplase in acute ischaemic
versus MR imaging in acute stroke using ASPECTS: stroke (ECASS II). Lancet 1998; 352: 12451251.
Will the new replace the old as the preferred 61 Larrue V, von Kummer R, Mller A, Bluhmki E. Risk
imaging modality? Stroke 2001; 32: 325 (abstract). factors for severe hemorrhagic transformation in
51 Ringelstein E, Biniek R, Weiller C, Ammeling B, Nolte P, ischemic stroke patients treated with rt-PA. Stroke
Thron A. Type and extent of hemispheric brain infarc- 2001; 32: 438441.
tions and clinical outcome in early and delayed
middle cerebral artery recanalization. Neurology 1992;
42: 289298.
Computerized tomographic-based evaluation of cerebral
blood flow

Lawrence R. Wechsler, Steven Goldstein and Howard Yonas

University of Pittsburgh Health System, Stroke Institute, Departments of Neurology and Neurosurgery, PA, USA

Introducton eective in attenuating X-rays and can therefore be

employed as a contrast agent. Unlike iodine, Xe is
Functional neuroimaging in the form of cerebral not reactive and is freely diusible throughout most
blood ow (CBF) measurement continues to be a tissues in our body including the bloodbrain
rapidly expanding tool in the care of patients with barrier. These properties which are not dependent
cerebrovascular disease, head trauma, seizure disor- on radioactivity have led to a commercially avail-
ders and many other disease states involving the able product that can interface with most modern
central nervous system. Computerized tomographic CT scanners and is capable of measuring blood
(CT)-based assessment of cerebral blood ow (CBF) ow quantitatively. 1,2
oers many advantages in the care of patients with After inhalation, Xe diuses into the circulation
disorders of the central nervous system. CT-based through the pulmonary capillary bed then rapidly
technology capable of evaluating CBF can be readily crosses the bloodbrain barrier into brain tissue.
combined with routine CT scanning equipment thus The CT scanner is able to detect small changes in
increasing the availability and decreasing the costs of attenuation due to the presence of Xe and the
this technology. Monitoring of patients with respira- change in density is directly proportional to the
tory and hemodynamic instability is also more easily concentration of Xe in the tissues. Since the Xe con-
done using CT based technology. In addition, centration in the blood can be measured indirectly
patients with mechanical heart valves, permanent by monitoring the end expiratory Xe concentration,
cardiac pacemakers and other ferromagnetic devices and the change in density is measured directly by
can be safely studied. Two primary CT-based imaging the sequential CT images, computation of cerebral
techniques are clinically available to evaluate CBF; blood ow on a pixel by pixel basis can be calcu-
stable xenon enhanced CT (XeCT) and dynamic CT lated.3 The actual arterial concentration can be esti-
perfusion imaging (CTP). These techniques are based mated based on the end expiratory concentration,
upon two entirely dierent mathematical models. and this allows the calculation of quantitative CBF
XeCT is based upon the well-established diusable values. CBF for all pixels is displayed in a colour-
tracer model, while CTP is based upon a non- coded ow map and areas of interest can be drawn
diusable tracer kinetic model that can be applied to to outline vascular territories.
both CTP and magnetic resonance perfusion (MRP). A series of images are obtained throughout the
brain over a 6-minute interval. This includes two
baseline images at each slice location followed by
Xenon CT cerebral blood flow six additional images at each level during
41 2 minutes of Xe inhalation. Only 28% Xe concen-
Xenon (Xe) is a naturally occurring element that is trations are required currently, which is well below
an inert gas at room temperatures. Like iodine, Xe is the 80% dose usually associated with anesthesia.

48 Lawrence R. Wechsler, Steven Goldstein and Howard Yonas

This method has the advantage in stroke patients of tions are easily overcome and cerebral blood ow
allowing CBF measurements without moving from data can be obtained expeditiously.
the CT table. Since most stroke patients undergo CT
scanning to detect hemorrhage, CT based CBF
measurements eliminate the need to move the Clinical applications of XeCT
patient and reduce overall time to acquire physio-
logical data, a point that is particularly important in The primary goal of thrombolytic therapy in acute
treatment of patients with acute stroke. Set-up time stroke is to re-establish tissue perfusion before the
for XeCT CBF studies requires only a few minutes onset of irreversible ischemia. Two major determi-
and the entire study is completed in 510 minutes. nants of irreversible cerebral ischemia leading to
Processing data and construction of CBF maps tissue infarction include the duration and severity
adds another 5 minutes so that the time from com- of reduced CBF.4 In a review of thresholds of ische-
pletion of the standard head CT to obtaining clini- mia in dierent animal models, Hossmann demon-
cally useful CBF information is only 1520 minutes. strated a distinct rank order of susceptibilities to
Some technical pitfalls exist. The major problem eventual infarction.5 The sodiumpotassium ratio
encountered is related to patient motion. Since the of brain tissue, an important marker of anoxic
technique requires a series of sections to be depolarization, increases at CBF values below
obtained at precisely the same locations, any move- 1015 ml/100 g per min6 and extracellular ion
ment between slice acquisitions can be a source of changes occur at CBF values between 6 and 15
error. That is, if a patient moves between images ml/100 g per min.79 In the baboon, Yonas et al.
then the pixels are misaligned and the accuracy of demonstrated that selective occlusion of the lentic-
the calculations is decreased. With attention to ulostriate arteries caused reduction of CBF to < 8
gaining the trust of the patient, careful positioning ml/100 g/min. At this level of CBF tissue infarction
and securing the head, and occasional intravenous was consistently produced after 60 minutes.10,11
sedation, greater than 90% of awake studies yield This suggests that the early extent of severe ische-
useful information. If a patient is paralysed and mia denes the nal extent of infarction.
intubated, the yield of excellent studies approaches
100%. Gas leaks in the administration set or in the
Acute stroke
mouth piece can also occur, though the monitoring
systems usually detect these quickly. Finally, in In patients with acute stroke, direct measurement of
theory, pulmonary disease sucient to cause poor CBF is potentially helpful to maximize the benet of
diusion of the Xe gas across the capillary mem- acute stroke therapy and reduce the complications.
branes from the alveoli to the bloodstream could To accomplish these goals, an imaging method
impair the uptake and alter the apparent equilib- must distinguish brain already infarcted or destined
rium, although in over 9000 studies at our institu- to infarct from reversibly ischemic brain capable of
tion, this has rarely been a problem. The one issue returning to normal functioning with reperfusion.
related to pulmonary disease that occasionally has In addition, some patients recanalize in the rst few
been problematic occurs in patients who require hours after stroke despite persistent neurological
high percentages of oxygen in a ventilator in order decits. In such cases, normal CBF values suggest
to maintain oxygenation. Then, the addition of Xe that reperfusion therapy is not necessary, possibly
gas may require decreasing the percentage of avoiding angiography or thrombolytic therapy.
oxygen during the blood ow examination and, in Xenon CT provides quantitative CBF values that
some clinical situations, this may be intolerable to can potentially be used to triage acute stroke
the patient. Agitation or vomiting occurs rarely, patients. CBF values below 6 cm3/100 g per min
particularly with the use of lower Xe concentrations measured in the setting of acute stroke correlated
associated with current CT scanning technology. In with subsequent infarction on follow-up CT
the vast majority of patients, the technical limita- whether or not reperfusion had occurred following
Computerized tomographic-based evaluation of cerebral blood flow 49

thrombolytic therapy.12 Patients with occlusion of technique in acute stroke will be able to consis-
the M1 segment of the middle cerebral artery (MCA) tently identify areas of brain potentially recoverable
had signicantly lower mean CBF values in the by reperfusion or other acute stroke therapies. CBF
MCA territory than those without MCA occlusion.13 values less than 20 ml/100 g per min indicating
All patients with M1 occlusions had mean MCA CBF ischemia but greater than 6 ml/100 g per min, the
values less than 20 cm3/100 g per min. Thus Xe CT level at which irreversible ischemia occurs even
may identify within the rst few hours after stroke within the rst few hours after stroke, hold the
areas of brain already infarcted as well as predict promise of identifying such reversibly ischemic
the presence of proximal arterial occlusion. areas. This must be conrmed by prospective
Reperfusion of infracted brain does little to studies showing that brain regions with this level of
improve neurological function but may cause CBF go on to infarction if untreated, but recover
neurological worsening due to hemorrhage or cere- after acute stroke intervention.
bral edema with herniation. Mean CBF values in A possible problem with using CBF measurement
the MCA territory less than 15 ml3/100 g per min in acute stroke is that measurements represent only
were associated with a greater risk of hemorrhage one point in a time continuum from the moment of
or herniation in patients treated with thrombolytic stroke onset. The CBF value obtained may vary with
therapy after acute stroke.14 CBF values were also time and may not be representative of CBF values
shown to more accurately predict complications of over the entire time since onset.19 Studies per-
symptomatic hemorrhagic conversion or hernia- formed in a primate model of middle cerebral
tion due to edema than initial NIHSS (Figs. 4.1, 4.2, artery stroke suggests, however, that without reper-
see colour plate section).15 fusion there is little uctuation in CBF within the
Initial CBF values may also predict long-term rst few hours. For each individual animal, the
outcome in patients with acute stroke. Mean CBF probability of an MCA qCBF value remaining
measured within 6 hours of stroke onset was sig- within the 010 ml/ 100 g per min range at any
nicantly lower in patients with severe decits or point in time over 6 hours range was 95% (see Fig.
death 1 year or more after stroke compared to those 4.3). Clinically, this would suggest that qCBF deter-
with mild decits.16 CBF also predicted long-term mination in ischemic stroke, demonstrating the
outcome better than time to treatment which cor- presence of an ischemic core measured within the
related poorly with long-term decits. Similarly, rst 6 hours after symptom onset, is unlikely to be
CBF measured by XeCT in the rst few hours after signicantly dierent if measured before, or after,
stroke was more predictive of subsequent infarction the clinical study upon which treatment decisions
on CT than time from onset of symptoms to CBF may potentially be based.
In some patients, despite signicant neurological
Chronic cerebral ischemia
decits, CBF in the aected vascular territory is
normal. This suggests reperfusion has already The role of hemodynamic factors in the risk of
occurred and that there is little to gain by attempt- recurrent stroke in patients with extracranial
ing thrombolysis. In many such patients, neurologi- carotid occlusive disease remains controversial.
cal decits resolve over the ensuing 24 hours From 1977 to 1985, The International Cooperative
without acute stroke interventions.18 The ability to Study of ExtracranialIntracranial Arterial
avoid therapy with the potential to cause intracere- Anastomosis compared the ecacy of the
bral hemorrhage in patients destined to improve STAMCA bypass procedure to medical manage-
may be a valuable contribution of quantitative CBF ment with aspirin therapy in a group of patients
measurements in the setting of acute stroke. with minor strokes or TIAs within 3 months prior to
Intermediate levels of CBF measured by XeCT surgery.20 Despite its negative results some clini-
may predict brain regions that can be salvaged with cians believe that patients with compromised
acute stroke therapy. Ultimately, the optimal CBF hemodynamics may benet from revascularization
50 Lawrence R. Wechsler, Steven Goldstein and Howard Yonas

80 LMCA territory (4-level average at each time): all primates

PRIMATE 1-30 -1998
60 PRIMATE 3-6-1998
CBF (ml /100 g/min)

PRIMATE 7-17-1998
PRIMATE 9-18-1998
40 PRIMATE 7-15-2000
PRIMATE 11-3-2000



BASE1 OCCL 120 min 240 min 360 min

Fig. 4.3. Left MCA qCBF over time after MCA occlusion in a primate model of MCA
infarction. Note that when qCBF approaches 10 cm3/100 g per min there is little varia-
tion of ow over time.

procedures. Recent data suggests that patients

Intracerebral hemorrhage (ICH)
demonstrated to have compromised cerebrovascu-
lar reserve may dene a subgroup at high risk for XeCT may provide valuable insights into the
recurrent stroke. Increased oxygen extraction frac- optimal blood pressure management of patients
tion (OEF) as measured by positron emission with ICH. Parenchymal ICH, commonly associated
tomography (PET) has been reported to be asso- with hypertension, remains the most common
ciated with a statistically higher incidence of prior form of hemorrhagic stroke and is an important
ischemic events (42%) than those patients with cause of morbidity and mortality.25 Substantial
normal OEF (16%) in patients with carotid occlu- growth in the overall volume of ICH occurs in 26%
sive disease.21,22 Using XeCT before and after an of patients from baseline exam to a 1 hour CT and
intravenous acetazolamide challenge both Yonas23 an additional 12% of all patients experience further
and Webster24 identied a group of patients with growth in the volume of ICH between a 1 and 20
carotid occlusive disease and evidence of hemody- hour CT scan.26 Volume expansion is a potentially
namic compromise or steal dened as a 5% important contributor to overall patient morbidity
decrease in qCBF measured after acetazolamide and mortality. Blood pressure reduction may
challenge. These patients subsequently had a stroke reduce this rate of early volume expansion;27
rate of 36% within 6 months. This demonstrates however, regional or global cerebral ischemia may
that measurement of cerebrovascular reserve after be precipitated in brain areas surrounding the
vasodilatory challenge as well as increased OEF as hemorrhage by decreasing CBF.28 In an attempt to
measured by PET may identify patients at high risk address these issues, investigators have begun to
of recurrent cerebral ischemic after symptomatic monitor global and regional CBF during attempts
carotid occlusive disease. Planning of prospective at aggressive blood pressure reduction in ICH.
trials are under way to reassess the utility of intra- Powers and colleagues studied CBF changes after
cranial bypass in patients with carotid occlusion blood pressure reduction in nine patients with
and compromised hemodynamic prole based on hypertensive ICH with PET. Intravenous labetolol
increased OEF and cerebrovascular reserve. or nicardipine were used to achieve a 15% reduc-
Computerized tomographic-based evaluation of cerebral blood flow 51

tion in mean arterial pressure (MAP). These nine with SAH. When TCD velocities become elevated,
patients did not demonstrate a signicant change CBF studies exclude hyperemia and identify
in mean hemispheric or regional CBF at the lower regions of impending ischemia even before symp-
blood pressure.29 Gebel et al. used XeCT to toms appear. In patients with neurological symp-
examine the eect of aggressive blood pressure toms, measurement of CBF helps determine
reduction on CBF in a retrospective series of six whether vasospasm is responsible for the symp-
hypertensive patients with ICH. Five of the six toms.
patients tolerated reduction to normotensive blood
pressure levels without a signicant decline in
global or regional quantitative CBF.30 In one CT perfusion
patient, blood pressure reduction caused CBF
values in the surround of the hemorrhage to fall With the development of helical CT scanning tech-
into the ischemic range suggesting that aggressive nology it is now possible to track a bolus of intrave-
blood pressure reduction may not be tolerated in nous contrast as it passes through the brain.
all patients. Further eorts are under way, designed Assuming there is a linear relationship between CT
to demonstrate the utility of CBF monitoring enhancement and the concentration of contrast
during blood pressure reduction in ICH, and the material within the brain tissues and arteries, both
ecacy of this strategy to improve outcome in a tissue enhancement curve and an arterial
these critically ill patients. enhancement curve can be measured during the
rst pass of contrast following intravenous injec-
tion. From these curves, CBF, cerebral blood
Subarachnoid hemorrhage (SAH) and vasospasm
volume (CBV) and mean transit time (MTT) can be
Symptomatic vasospasm results in decreased cere- calculated. In addition, the ability to measure an
bral perfusion at the parenchymal level leading to arterial input function allows calculation of quanti-
tissue ischemia and in some cases irreversible tative CBF.33 However, current CT technology
necrosis. Although transcranial Dopper (TCD) permits imaging of only a single slice compared
velocities measured in large conductance vessels with four slices imaged by XeCT. This is a consider-
correlate well with angiographic evidence of vaso- able disadvantage in patients with stroke since the
spasm, they may accurately reect decreased CBF. area of ischemia may extend across a much greater
Hyperemia also occurs in the setting of SAH and region of brain or a single slice may miss a small
may also cause elevated TCD velocities.31 TCD area of ischemia entirely. Quantitation of CBF using
monitors blood ow velocity in the large arteries of this method also has limitations. Values may
the Circle of Willis. When leptomeningeal collateral change depending upon the artery chosen to be
is adequate, the brain may not be ischemic despite used as the arterial input function.34 In one patient
thre presence of vasospasm. The combination of with recent stroke studied at our institution by CT
TCD and a measure of tissue perfusion such as perfusion and XeCT values correlated in areas of
Xe CT may help avoid the pitfalls inherent in using low ow, but in regions of normal CBF by XeCT
TCD alone to monitor patients with SAH.32 TCD is there was poor correlation with CT perfusion
also insensitive to second- and third-order vessel values.
spasm that can occur without signicant involve-
ment of the more proximal vessel examined by
Clinical applications of CT perfusion
TCD. Although TCD generally correlates well with
angiographic vasospasm it is likely that measure- CT perfusion has only recently been introduced
ment of CBF yields the most important evidence of and only a few clinical studies assessing the utility
clinically signicant vasospasm, that is, signicant of this technique have been published. Due to the
tissue hypoperfusion. XeCT provides a rapid and problems with quantitation of CBF using this
readily available technique for monitoring patients method, most studies used ratios comparing values
52 Lawrence R. Wechsler, Steven Goldstein and Howard Yonas

in the aected regions with the contralateral hemi- minimal delay in time to treatment. In addition,
sphere. Such comparisons also may introduce error they must be widely available in order to have a
since infarction in one hemisphere has been shown true clinical impact. CT- based systems now oer
to be associated with decreased CBF values in the the ability to provide exquisite tissue and vascular
contralateral unaected hemisphere.35 Koenig et anatomy using conventional CT and CT angiogra-
al.36 found that CT perfusion studies in patients phy. In addition, CT-based perfusion techniques
with stroke within 6 hours of onset predicted the also provide physiologic information that is becom-
development of infarction on follow-up CT scans. ing increasingly important in clinical decision
Since only a single slice was obtained, only infarc- making. All this information (CT, CTA, XeCT and/or
tion at this brain level was predicted. Other studies CTP) may be obtained in under 2030 minutes
demonstrated a correlation between volume of using current technology.
hypoperfusion and angiographic data37 as well as
the clinical application of this technique in acute
stroke patients.38 Much more work is needed to
establish the utility of CT perfusion in stroke
patients. Quantitation also must be improved
before it can rival XeCT for the measurement of
1 Good WF, Gur D. Xenon-enhanced CT of the brain:
ischemic levels of CBF. As CT technology improves,
eect of ow activation on desired cerebral blood ow
it is likely that multiple slice imaging will be pos- measurements. Am J Neuroradiol 1991; 12: 8385.
sible, making this method more applicable to eval- 2 Gur D,Yonas H, Wolfson SK Jr. et al. Xenon and iodine
uation of stroke. enhanced cerebral CT: a closer look. Stroke 1981;
An alternative to bolus tracking CT perfusion is 12(5): 573578.
CT enhancement imaging of the brain parenchyma 3 Kety SS, Schmidt CF. The nitrous oxide method for the
following a bolus of contrast for studies such as CT quantitative determination of cerebral blood ow in
angiography.39 In areas of ischemia there is less man: theory, procedure, and normal values. J Clin
enhancement most likely reecting reduced blood Invest 1948; 27: 476483.
volume. This assessment can be easily accom- 4 Jones TH, Morawetz RB, Crowell RM et al. Thresholds
of focal cerebral ischemia in awake monkeys. J
plished in the acute stroke setting following CT
Neurosurg 1981; 54: 773782.
angiography without the need for any additional
5 Hossmann K-A. Viability thresholds and the penum-
contrast or injections. Although not quantitative it
bra of focal ischemia. Ann Neurol 1994; 36(4):
gives a rapid assessment of the localization and size 557564.
of brain regions aected by ischemia and may 6 Hossman K-A, Schuier FJ. Experimental brain infarcts
predict areas of brain that are likely to go on to in cats. Pathophysiological observations. Stroke 1980;
infarction.40 11: 583592.
7 Astrup J, Symon L, Branston NM. Cortical evoked
potential and extracellular K and H at critical levels
Conclusions of brain ischemia. Stroke 1977; 8: 5157.
8 Branston NM, Strong AJ, Symon L. Extracellular potas-
The need for quantitative physiologic information sium activity, evoked potential and tissue blood ow.
Relationships during progessive ischaemia in baboon
continues to expand as more emphasis is placed on
cerebral coretex. J Neurol Sci 1977; 32: 305321.
acute therapeutic intervention in a host of disease
9 Morawetz RB, Crowell RH, DeGirolami U. Regional
states aecting the central nervous system. These
cerebral blood ow thresholds during cerebral ische-
techniques oer the possibility of improved patient mia. Fed Proc 1979; 38: 24932494.
selection and decreased complications from 10 Yonas H, Gur D, Claassen D, Wolfson SKJ, Moosy J.
aggressive interventional techniques. However, in Stable xenon enhanced computed tomography in the
order to be of clinical usefulness these techniques study of clinical and pathologic correlates of focal
must be capable of being rapidly performed with ischemia in baboons. Stroke 1988; 19: 228238.
Computerized tomographic-based evaluation of cerebral blood flow 53

11 Yonas H, Gur D, Claassen D, Wolfson SKJ, Moosy J. 21 Derdeyn CP, Yundt KD, Videen TO, Carpenter DA,
Stable xenon-enhanced CT measurement of cerebral Grubb RL, Powers WL. Increased oxygen extraction
blood ow in reversible focal ischemia in baboons. J fraction is associated with prior ischemic events in
Neurosurg 1990; 73: 266273. patients with carotid occlusion. Stroke 1998; 29:
12 Kaufaman AM, Firlik AD, Fukui MB, Wechsler LR, 754758.
Jungreis CA, Yonas H. Ischemic core and penumbra in 22 Grubb RL, Derdeyn CP, Fritsch SM et al. Importance of
human stroke. Stroke 1999; 30(1): 9399. hemodynamic factors in the prognosis of sympto-
13 Firlik AD, Kaufmann AM, Wechsler LR, Firlik KS, Fukui matic carotid occlusion. JAMA 1998; 280: 10551060.
MB, Yonas H. Quantitative cerebral blood ow deter- 23 Yonas H, Smith H, Durham S, Pentheny SL, Johnson
minations in acute ischemic stroke: relationship to DW. Increased stroke risk predicted by compromised
computed tomography and angiography. Stroke 1997; cerebral blood ow reactivity. J. Neurosurg 1993; 79:
28: 22082213. 483489.
14 Firlik AD, Yonas H, Kaufmann AM et al. Relationship 24 Webster MW, Makaroun MS, Steed DL, Smith HA,
between cerebral blood ow and the development of Johnson DW, Yonas H. Compromised cerebral blood
swelling and life-threatening herniation in acute ow reactivity is a predictor of stroke in patients with
ischemic stroke. J Neurosurg 1998; 89: 6773. symptomatic carotid artery occlusive disease. J Vasc
15 Goldstein S, Yonas H, Gebel JM et al. Acute cerebral Surg 1995; 21: 338345.
blood ow as a predictive physiologic marker for 25 Broderick JP, Brott T, Tomsick BT et al. Intracerebral
symptomatic hemorrhagic conversion and clinical hemorrhage is more than twice as common as sub-
herniation after thrombolytic therapy. Stroke 2000; 31: arachnoid hemorrhage. J Neurosurg 1993; 78: 188191.
275. Accepted as oral presentation to the 25th AHA 26 Brott T, Broderick J, Kothari R et al. Early hemorrhage
International Stroke Conference 2000. growth in patients with intracerebral hemorrhage.
16 Goldstein S, Jungreis CA, Wechsler LR et al. Time to Stroke 1997; 28: 15.
treatment and cerebral blood ow as predictive meas- 27 Dandapani BK, Suzuki S, Kelley RE, Reyes-Iglesias Y,
ures of clinical outcome in patients treated with intra- Duncan RC. Relation between blood pressure and
arterial thrombolytic therapy. Accepted as an oral outcome in intracerebral hemorrhage. Stroke 1995; 26:
presentation at the American Society of 2124.
Neuroradiology (ASNR) 39th Annual Meeting April 26, 28 Powers WJ. Acute hypertension after stroke: the scien-
2001. tic basis for treatment decisions. Neurology 1993; 43:
17 Kilpatrick MM, Goldstein S, Yonas H et al. Sensitivity 461467.
and specicity of quantitative cerebral blood ow vs. 29 Powers WJ, Adams RE, Yundt KD et al. Acute
time from symptom onset as a predictor of cerebral Pharmacological hypotension after intracerebral
infarction. Abstract accepted as a poster presentation hemorrhage does not change cerebral blood ow.
to the 26th American Heart Association International Stroke 1999; 30: 242; abstract.
Stroke Conference, February 1416, 2001. Stroke 2001; 30 Gebel JM, Kassam AB, Snyder JV et al. Eects of
32 (1); 348. aggressive blood pressure reduction on cerebral blood
18 Firlik AD, Rubin G, Yoans H, Wechsler LR. Relation ow in patients with acute intracerebral hemorrhage.
between cerebral blood ow and neurologic decit Stroke 2000; 31: 283.
resolution in acute ischemic stroke. Neurology 1998; 31 Iacopino DG, Todaro C, Alafaci C et al. Transorbital
51: 177182. Doppler: an approach that increases transcranial
19 Jovin TG, Goldstein S, Gebel J, Wechsler LR, Ott, M-B, Doppler sensitivity in the detection of SAH
Yonas H. Patterns of core and penumbra in acute M1 Vasospasm. Stroke 1993; 24: 519
occlusion and their clinical correlates. Abstract 32 Clyde BL, Resnick DK, Yonas H, Smith HA, Kaufmann
accepted as a poster presentation to the 26th AM. The relationship of blood velocity as measured by
American Heart Association International Stroke transcranial doppler ultrasonography to cerebral
Conference, February 1416, 2001. Stroke 2001; 32 (1); blood ow as determined by stable xenon computed
348. tomographic studies after aneurismal subarachnoid
20 The EC/IC Bypass Study Group. Failure of extracra- hemorrhage. Neurosurgery 2001; 38: 896.
nialintracranial arterial bypass to reduce the risk of 33 Cenic A, Nabavi DG, Craen RA, Gelb AW, Lee T-Y.
ischemic stroke. Results of an international random- Dynamic CT measurement of cerebral blood ow: A
ized trial. N Engl J Med 1985; 313: 11911200. validation study. Am J Neuroradiol 1999; 20: 6373
54 Lawrence R. Wechsler, Steven Goldstein and Howard Yonas

34 Wintermark M, Thiran J, Maeder P, Schnyder P, Meuli 38 Hunter GJ, Hamberg LM, Ponzo JA et al. Assessment
R. Simultaneous measurement of regional cerebral of cerebral perfusion and arterial anatomy in hyper-
blood ow by perfusion CT and stable xenon CT: a acute stroke with three-dimensional functional
validation study. Am J Neuroradiol 2001; 22: 905914. CT: early clinical results. Am J Neuroradiol 1998; 19:
35 Firlik AD, Kaufmann AM, Wechsler LR et al. Quantitative 2937.
cerebral blood ow determinations in acute ischemic 39 Hunter GJ, Hamberg LM, Ponzo JA et al. Assessment
stroke: relationship to computed tomography and of cerebral perfusion and arterial anatomy in hyper-
angiography. Stroke 1997; 28: 22082213. acute stroke with three-dimensional functional
36 Koenig M, Kraus M, Theek C, Klotz E, Gehlen W, CT: early clinical results. Am J Neuroradiol 1998; 19:
Heuser L. Quantitative assessment of the ischemic 2937.
brain by means of perfusion-related parameters 40 Lev MH, Segal AZ, Farkas J et al. CT Blood volume
derived from perfusion CT. Stroke 2001; 32: 431437. imaging of hyperacute MCA stroke: prediction of nal
37 Lee KH, Cho S, Byun HS et al. Triphasic perfusion infarct volume and clinical outcome in response to
computed tomography in acute middle cerebral intra-arterial thrombolysis. Stroke 2001; 32:
artery stroke. Arch Neurol 2000; 57: 990999. 20212028.
Technical introduction to MRI

Rohit Sood and Michael Moseley

Department of Radiology, Stanford University, CA, USA

Introduction The integrated MR stroke protocol

Since the rst image of the human wrist was Because MRI can be made sensitive to many dier-
obtained by Paul Lauterbur in 1972, magnetic res- ent tissue contrasts ranging from T1 relaxation
onance imaging (MRI) has developed signicantly times to proton diusion, and since many MRI
to a stage where it has found applications to most sequences can be done in a few seconds to minutes
of the speciality branches in medicine. In the last scan time, stroke protocols using MRI have begun
decade there has been a rapid development in including many dierent sequences in order to
gradient and RF coil technology and this has maximize the amounts of information within a 20
resulted in the implementation of fast imaging to 30 minutes examination time. The following MRI
techniques such as echo planar imaging (EPI), on methods that can be found in a typical MRI stroke
clinical imaging systems. It is now possible to protocol are described more fully below. The
acquire a single image of the brain in less than parameters that may be used in a typical stroke
50 ms using EPI.1 During this period, MR has been protocol are outlined in Table 5.1.
used for obtaining functional information about
the dierent physiological processes in the brain,
such as diusion and perfusion. This has resulted The T1-weighted image
in the development of newer techniques, with the
MR systems being equipped with faster gradients. The time taken by the net nuclear magnetization
MRI has been used in the acute stroke setting over vector to reach the thermal equilibrium with the
the past few years, due to the unique sensitivity of surrounding environment (lattice) is called the
several new MRI techniques to rapidly detect spin-lattice relaxation time or T1. The T1 relaxation
cerebral ischemia. These new techniques include time depends on the environment of the relaxing
diusion-weighted imaging (DWI), perfusion- spin nuclei. For example, proton spins in the cere-
weighted imaging (PWI), magnetic resonance brospinal uid (CSF) have a T1 of 1900 ms (at 1.4 T
spectroscopy (MRS) and high-speed MR- static magnetic eld) while the T1 of protons in the
angiography (MRA). In this chapter, we will briey thalamus is 700 ms.2 Thus the thalamus has a
outline the basic physical principles of MRI, shorter T1 compared to the CSF due to the inherent
which will lay a foundation for the following chap- property of the tissue. In general, proton spins that
ters. A brief overview of the MR techniques that are not bound to molecules and exist in free form
have been developed for investigating stroke (as in the CSF, water) have longer T1 relaxation
patients is provided. It is assumed, however, that times. The image generated from a signal with a
the reader is familiar with the basic physics of greater T1 dependence is called a T1-weighted
MRI. image (Figure 5.1 upper left). It is possible on the

56 Rohit Sood and Michael Moseley

Table 5.1. Typical stroke protocola

1. Sagittal scout Scout

Patient position: Supine; Head First; Head Coil
Imaging parameters: Sagittal; 2D; GRE
Scan timing: 1 echo; TE 20 ms; TR 500 ms
Acquisition timing: 256 3 128; 1/2 NEX; Phase FOV 1; Autoshim; S/I
Scanning range: FOV 24; slice thickness 3 skip 1; 20 slices Time: 0:36

2. FSE FSE First and Second Echo

Imaging parameters: Oblique; 2D; Spin-echo (Fast)
Scan timing: 1 echo; TE 17/100, TR 9000
User CV: Min Acquisitions 1
Acquisition timing: 256 3 192; 1 NEX; A/P; Autoshim; Phase correct
Scanning range: FOV 24; 5 skip 2.5; 20 slices Time: 3:00

3. Circle of Willis MRA

Imaging parameters: Oblique; 3D; Vasc TOF SPGR; Flow Comp; MT; EDR
Vascular options: Collapse; 19 Projections; Ramp pulse I -> S
Scan timing: 25; MinTE; TR 5 44
Scan setup: Autoshim; Bandwidth 15.63
Scanning range: FOV 20 cm; 2; 32 locations; overlap 4
Saturation: Explicit sat (S)
Acquisition time: 256 3 128; A/P; Phase 0.75; 1 NEX; Time: 2:17

4. Diffusion DWI 1 DTI

Imaging parameters: Oblique; 2D; psd: DW-EPI
User CV: FLAIR 5 0, 1
Diff options: b-values 5 014,000 (28 NEX)
Diff grad options: 0, x, y, z, then: xy, xz, yz, -xy, -xz, y-z
Scan timing: 1 Shot; MinTE; TR 5 4000
Acquisition timing: 128 3 128; Phase FOV 1; R/L;
Scanning range: FOV 24; 5 skip 2.5; 20 slices; GRx Time 6:36

5. Bleed screen GRE EPI

Imaging parameters: Oblique; 2D; GE-EPI; Multiphase
Scan timing: 1 Shot; TE 5 60; TR 5 2000; Flip 60
Acquisition timing: 128x128; 1 NEX; Phase FOV 1; R/L;
Autoshim; Phase correct
Scanning range: FOV 24; 5 skip 2.5; 20 slices Time: 0:06

6. Perfusion PWI
Imaging parameters: Oblique; 2D; GE-EPI; Multiphase
Multiphase: 40 phases; Interleaved; Minimum delay
Scan timing: 1 Shot; TE 5 60; TR 5 2000; Flip 60
Acquisition timing: 128x128; 1 NEX; Phase FOV 1; R/L;
No Autoshim; Phase correct
Scanning range: FOV 24; 5 skip 2.5; 12 slices Time: 1:21

Note: a These parameters are listed for the GE Horizon Series and are meant for illustrative
purposes only
Technical introduction to MRI 57

(a) (b) (c ) (d ) (e) (f )

(g) (h) (i ) (j )

Fig. 5.1.
(a) T1-weighted image of the brain (TR/TE 500/18.0 ms) used for scout.
(b) T2-weighted axial image of the brain is commonly acquired using fast spin echo (FSE) sequence (TR/TE 4 s/100 ms).
This can also be acquired as the b0 image from the SE EPI series for DWI.
(c) Image acquired using FLAIR sequence from the FSE or SE EPI image.
(d) Diusion-weighted image acquired at high b value. Note the right-sided lesion.
(e) ADC map obtained from the b0 and the high b value images. Note the low ADC value in the lesion seen in the
DW image.
( f ) Upper right: A GRE EPI bleed screen depicting possible bleed.
(g) Lower row left: Time of Flight (3D TOF) MRA angiogram.
(h) Maps acquired from the PWI exam. rCBV, rMTT, and rCBF
(i ) Arterial spin label (ASL) image (FAIR method) of normally perfused slice.
( j ) CE-MRA image acquired during arterial phase of contrast bolus injection.

MR scanner to vary parameters such as TR (repeti- the long T1 of CSF) while the white matter appears
tion time), TE (echo time) and TI (inversion time, if bright (high signal due to a relatively shorter T1
using an inversion recovery sequence) in order to than CSF). The T1 property of tissue can also be
obtain T1 weighting. T1-weighted images are typi- modied due to the presence of a T1 shortening
cally obtained using spin echo sequences with agent (e.g. paramagnetic contrast agent such as
short TR and short TE. On a T1-weighted image of gadolinium or the iron in methemoglobin).
the brain, the CSF typically appears dark (no signal MR sequences that generate a T1-weighted image
due to signal saturation because of the short TR and by virtue of adding an inversion recovery pulse to
58 Rohit Sood and Michael Moseley

invert protons with a particular T1 are STIR (short early subacute phase of a parenchymal hematoma
tau inversion recovery) and FLAIR (fLuid attenu- the conversion of hemoglobin to methemoglobin
ated inversion recovery, Fig. 5.1(c)) FLAIR imaging has a T2 shortening eect on the vascular lesion.
has been particularly useful in imaging of patients This is due to the fact that methemoglobin, formed
with onset of cerebral ischemia3 even though it is by the oxidative denaturation of hemoglobin, has
more properly an inversion recovery T2-weighted ve unpaired electrons on the heme iron making it
spin echo sequence. paramagnetic and thus resulting in T2 shortening
The T1-weighted image is typically added to the (along with strong T1 shortening).5 As a result the
stroke protocol usually as a pilot or scout image lesion will appear dark on a T2-weighted image due
series for prescribing slices for subsequent image to spin related dephasing eects (or loss of phase
series, although T1-weighted image series could coherence). However, T2-weighted imaging tech-
also be acquired prior to and after gadolinium con- niques such as conventional and fast spin echo
trast injection to look for breakdowns in the have not been very useful in the neurological
bloodbrain barrier or even as a quick depiction of assessment of acute ischemia due to subtle changes
evolving mass eects postischemia. occurring in the parenchymal tissue on these
images and the hyperintense signal from CSF
masking the subtle pathological ndings.
The T2-weighted image The T2-weighted image is acquired, together with
a proton density-weighted image in one integrated
Due to mutual interaction between the spins, com- image series employing a long TR together with a
ponents of the magnetization vector get out of short TE (the proton density weighted image) and
phase and result in decay of the net magnetization. long TE (the T2-weighted image). The utility of the
The time for the components to decay is known as T2-weighted image is to depict increases in water
the T2 or transverse relaxation time. Smaller mole- content due primarily to edema or inammation.
cules like water have longer T2 relaxation times The use of the T2 weighted image to detect hyper-
compared to large macromolecules and in general acute and acute ischemia, however, is hampered by
T2 as well as T1 are functions of water content, that the slow evolution of T2 over the rst 24 hours fol-
is the higher the water content in the tissue, the lowing an ischemic insult. The T2 weighted image is
longer the T2 and T1. CSF for example has a T2 of largely considered to be the MRI gold standard for
250 ms compared to thalamus whose T2 is 75 ms (at lesion volume at later time points beyond the rst
1.4T).2 Hence on a T2-weighted image of the brain 24 hours.
(Fig. 5.1(b)) the CSF appears bright (due to longer
T2) as compared to cortical grey matter. Examples
of MR sequences that are used for acquiring T2- Magnetic resonance angiography
weighted images include conventional spin echo
(CSE), rapid acquisition with relaxation enhance- Magnetic resonance angiography (MRA) is a major
ment (RARE) or fast spin echo/turbo spin echo and advance in application of MRI to imaging of patent
EPI-spin echo. blood ow in vessels. It is now used routinely with
Paramagnetic contrast agents such as gadolinium conventional MR studies of the brain for investigat-
(and its chelates) in high concentration have a T2 ing ischemic lesions and arteriovenous malforma-
(and T2*; discussed later under perfusion-weighted tions. The MRA techniques can be broadly classied
imaging) shortening eect (in addition to T1 short- as (i) the inow of moving protons orthogonal to the
ening).4 On a T2-weighted image this may result in slice that have not seen prior RF pulses producing a
loss of phase coherence and signal loss causing the hyperintense time of ight eect within the vascu-
tissue to appear dark. Compounds with paramag- lar walls and (ii) spins moving during the applica-
netic properties found in the body also demon- tion of and in the direction of an imaging gradient
strate T2 shortening eects. For example, during the that produce a signal phase shift (dependent on the
Technical introduction to MRI 59

type and direction of blood ow), so-called the sion-weighted imaging) relaxation process of any
phase contrast technique. ensemble of the spins. Another important source of
Both two-dimensional6 and three-dimensional7 contrast in MRI is the signal loss caused by proton
gradient echo TOF techniques have been popular dephasing in the presence of coherent and incoher-
for stroke-related MRA clinical utility. The two- ent ow. Random diusion (Brownian motion of
dimensional TOF technique produces images with water molecules) of protons into areas of varying
decreased spatial resolution in the slice select magnetic eld strengths leads to random phase
direction and has a longer TE compared to the shifts (not seen in stationary tissue). There is a
three-dimensional technique. In the three- resultant loss of signal from these protons as a result
dimensional TOF MRA technique, a thick slab or of these phase shifts. In pure water, protons diuse
volume of tissue is excited and extra phase encod- greater distances (2025 microns) as compared to
ing steps are applied along the slice select direction tissue water (715 microns) within an imaging time
(in addition to those in the phase encoding direc- of 40 ms. Thus the process of diusion, including
tion) to divide the volume into multiple partitions rate and direction of diusion reect the hindered
or sections along the slice coverage. A three-dimen- motion of water molecules in a given medium.10 A
sional Fourier transform is then performed to parameter that quantitates the property of diusion
reconstruct the image. The three-dimensional data and proton displacement is called the apparent
sets have voxel dimensions of less than 0.80.8 diusion coecient (ADC). A typical ADC value in
0.8 mm, which is typically smaller than that can be human brain is 1 103 mm2/s. The ADC value of
achieved using two-dimensional techniques.7 water in tissue changes if the motion of protons is
Other advantages of the three-dimensional over the hindered along any one direction, resulting in the
two-dimensional technique are higher signal to property of a non-random, anisotropic diusion
noise ratio (SNR), improved slice prole and reduc- pathway, resulting in not a single average ADC,
tion in T2* eects8 at the expense of the loss of more appropriately in an ADC tensor. This property
inow through deeper volumes. of anisotropic diusion tensor imaging (DTI) is seen
A new MRA technique that acquires imaging data in myelin bres in white matter, where the protons
before, during and after administration of a single or move faster along the direction of least resistance
double dose gadolinium contrast agent bolus injec- causing the ADC to be directionally dependent
tion is known as contrast-enhanced MRA (CE- (anisotropic).11 In such conditions the direction of
MRA).9 This technique has become very popular the applied diusion sensitizing gradient in the ST
and is commonly used in stroke protocols today for (Stejskal and Tanner) pulse sequence (see below)
investigating intracranial vascular diseases over can be chosen by using one of the x, y or z axis eld
large peripheral elds of view from typically the gradients or more. This allows assessment of the
aortic arch to the circle of Willis. By injecting gado- diusion anisotropy tensor12 and can be used to
linium, the blood T1 can be shortened to a few tens synthesize a diusion tensor map (DTI map) the
of milliseconds and three-dimensional data sets can practice of which is becoming more popular in
be longitudinally acquired with a very short exam- routine stroke MRI.
ination time. This technique oers several advan- Building upon spin echo imaging techniques, it is
tages over non-contrast enhanced conventional possible to encode the ADC information by modify-
MRA and oers useful information to the stroke ing pulse sequences to include extra diusion sen-
protocol at the expense of a contrast bolus injection. sitizing gradients which are applied along typically
six varying directions. This technique is known as
the StejskalTanner, ST technique or pulsed mag-
Diffusion-weighted imaging netic eld gradient technique.13 The diusion
encoding gradients (noted in Fig. 5.2) are a set of
The signal intensity in MRI depends on proton matching gradient pulses played along the direc-
density, T1, T2 and T2* (explained later under perfu- tion of diusion of protons. The diusion of
60 Rohit Sood and Michael Moseley

Fig. 5.2. Diusion-weighted imaging using spin echo SE EPI sequence.The three main steps involved in obtaining a diusion map
of the brain are acquiring MR images with various b-values (typically b0 and b8001000 s/mm2), plot SI vs. b-values to obtain
the ADC map.

protons during the application of pulsed gradients (45 ms), b values of up to 2500 s/mm2 can be
( in Fig. 5.2) causes them to acquire a phase achieved depending upon the diusion sensitizing
change (or dephase). As a result, the diusion of gradient strength. The b value signies the diusion
protons causes incomplete spin rephasing and of protons under the inuence of pulsed diusion
attenuation of spin echo signal during data acquisi- gradients. The dierences between the slow and
tion. On the acquired DW (diusion-weighted) fast diusion rates are best visualized at higher b
images, relatively slow diusion appears hyperin- values, however, at the cost of SNR. In order to max-
tense while rapid diusion appears hypointense. imize the b values, larger gradients than those com-
The signal intensity of the acquired signal (echo) monly used are necessary, which has prompted
depends on the following parameters which are improved design and commercial availability of
known and can be varied: stronger gradients.14
(a) : the time interval between the leading edges A typical diusion examination consists of rapid
of the diusion sensitizing gradients (DG), acquisition (as shown in Fig. 5.2) of a DW image
(b) , G: the duration and amplitude of the diu- with b0 (no diusion weighting) followed by
sion sensitizing gradient pulses (DG) rapid acquisition of a DW image at a higher b value
(c) b value is related to the above parameters and (typically 1000 s/mm2). All the acquired images are
is given by bexp[-22G2( /3) ADC].13 then used to obtain the ADC value and to synthe-
Given the typical parameters of (53 ms), size an ADC map (Fig. 5.1(e), 5.2) from the slope of
Technical introduction to MRI 61

the signal decrease with increasing the value. The such as multiple sclerosis,20 which may provide
calculation of b value is inuenced somewhat by better understanding about the underlying disease
the interaction of diusion encoding gradient with process. However, in routine stroke DWI, the trace
slice selective and frequency encoding gradients.15 ADC (the average of diagonal elements of the
These gradients themselves act in a lesser way as tensor matrix) images acquired from the DWI series
dephaserephase gradients and thus may contrib- is used. The trace averaging tends to remove the
ute to the observed diusion weighting.16 eects of anisotropic diusion thus simplifying the
Patient motion and CSF pulsation during detection of ischemic from normally perfused
imaging using conventional spin echo imaging brain.
techniques results in motion degraded MR images.
This is due to motion artefacts, also known as
ghosting artefacts, arising from motion during the Perfusion-weighted imaging
long acquisition times seen with the conventional
spin echo technique. With the development of fast Perfusion is the steady-state delivery of blood
imaging techniques such as single shot EPI and half (nutrients and oxygen) to tissue parenchyma repre-
Fourier RARE, it has been possible to suppress senting coherent motion of water and cellular
motion artefacts even with large diusion gradients material.21 Since both perfusion-weighted imaging
applied. Routine diusion-weighted imaging is per- (PWI) and magnetic resonance angiography (MRA)
formed using DWSEEPI (diusion weightedspin map the ow of blood in vessels, there may be some
echoecho planar imaging) in regular clinical prac- confusion between the applications of these two
tice due to widespread availability of these tech- techniques. MRA is useful in demonstrating the
niques on commercial clinical MR systems (Fig. macroscopic vasculature (in both arteries and
5.2). veins) and detecting morphological changes in
The role of diusion-weighted imaging in clinical blood vessels themselves such as stenosis, aneu-
practice became clear when it was noted that the rysms, dissections and malformations. PWI detects
ADC of water was signicantly slower in regions of microscopic ow at the capillary level and therefore
ischemia compared with normally perfused regions is useful to investigate changes taking place at the
of brain with ADC decreasing by 30% to 60% after cellular level.
the onset of stroke.17 The ability of a non-invasive Perfusion is typically measured in ml/100 g of
MR method such as DWI, to quickly and accurately tissue/min. The normal gray matter is perfused at
detect and characterize cerebral ischemia occurred the rate of 5060 ml/100 g/min. There are three
at a critical time in neuroimaging, when thrombo- important parameters that are used to quantify and
lytic and neuroprotective agents were entering the assess brain tissue perfusion: cerebral blood ow
clinical arena. The common use of DWI in stroke (CBF), cerebral blood volume (CBV) and mean
protocols is to provide the rst look at the extent of transit time (MTT). CBV is dened simply as the
the ischemic lesion, although more often than not, amount of blood in a given amount of tissue at any
the DWI sequences used to rule out stroke in most time (ml/g) as compared to CBF, which represents
CNS exams. the amount of blood moving through a certain
The concept of diusion-weighted imaging was amount of tissue per unit time (ml/g per min). The
modied to account for these orientational varia- ratio of CBV and CBF is dened as the mean transit
tions in the ADC measurements and hence the time (units minutes).
concept of diusion tensor18,19 involves the appli- Perfusion imaging techniques can be divided into
cation of diusion gradients in dierent directions two broad groups, depending upon the type of con-
and the data obtained is used to extract and charac- trast mechanism:
terize tissue microstructure such as the white (i) signal monitoring using exogenous (injectable)
matter tracts. The concept of DTI has been success- contrast agents such as gadolinium DTPA (Gd-
fully applied to the study of white matter diseases DTPA; T1 shortening agent), oxygen 17 in H2O17
62 Rohit Sood and Michael Moseley

(T2 shortening agent) or superparamagnetic ture by synthesizing perfusion maps. PWI is per-
agents (T2* shortening). The technique that formed by obtaining images at regular intervals
uses Gd-DTPA as the exogenous contrast agent before, during and after injection of a magnetic sus-
will be discussed in greater detail, since it is the ceptibility contrast agent. For example, after inject-
most commonly used technique in the clinical ing 0.12 mmol/kg of a gadolinium chelate at the
setup; and rate of 5 ml/second, rapid imaging is performed
(ii) Endogenous or inherent contrast agents such using gradient echo EPI sequence and data sets
as arterial spin labelling technique (ASL). (typically consisting of 1215 slices) are acquired
The basic underlying concept of perfusion imaging serially at 2-second (usually the TR) interval (Fig.
involves visualizing (and quantifying) signal 5.3 upper left). Once the data set is acquired, a plot
changes during the vascular transit of an injected of signal intensity change vs. time (Fig. 5.3 upper
contrast agent such as gadolinium (Gd), dyspro- right) is obtained from each of these images using
sium (Dy) or iron (Fe) based compounds using postprocessing techniques. Local change in relaxiv-
rapid T2* sensitive imaging techniques such as EPI. ity (R2*1/T2*(t) 1/T2*(0)) is then obtained by
These substances produce stronger magnetic eld creating a R2* map using the signal intensity
heterogeneity eects than other substances due to values from the images. The change in R2* with
the inherent property of magnetic susceptibility time is tted to a gamma variate curve 24 (so called
(dened as the ratio of induced magnetic eld in a because the curve resembles the plot of a mathe-
substance to the applied static magnetic eld). As a matical function called gamma variate function).
result, these substances cause perturbations in the From this curve (Fig. 5.3 lower left) a number of
local magnetic eld, which inuence water protons parameters are then extracted such as relative CBF,
within a certain range. Water protons undergo relative CBV (Fig. 5.1(h)), relative MTT (Fig. 5.3,
dephasing as they diuse through the sphere of lower right), arrival time (AT), peak height, time to
inuence of these contrast agent particles. The loss peak (TTP) and FWHM (full width at half
of phase coherence and signal loss occurs due to maximum).25 Thus it is the relation between the
T2* relaxation and it follows an exponential decay signal intensity vs. time plot and concentration of
similar to T2. This eect can be reversed with spin the tracer in the voxel over time that provides the
echo based techniques and hence makes these necessary information about cerebral hemodynam-
techniques relatively insensitive to T2* eects. ics and perfusion.
However, gradient echo based imaging techniques A number of physiological factors inuence the
are very sensitive to small changes in magnetic sus- parameters that can be extracted from the signal vs.
ceptibilities, and as a result have proven to be time or concentration vs. time plots. For example,
useful in perfusion imaging.22 The spheres of inu- hematocrit (Hct), which is dened as the ratio of
ence of these particles exists beyond immediate cerebral red blood cell volume and cerebral blood
vicinity and have more signicant long-range volume and has units of percentage, which must be
eects than do T1 shortening eects (which exist considered with MR techniques when using plasma
around unpaired electron clouds only and hence markers.26 Some other parameters that may indi-
are short-range eects). The long-range eects of rectly inuence perfusion imaging are heart rate
these T2* sensitive agents are advantageous since a and mean arterial blood pressure.
small concentration can greatly inuence freely
diusing protons throughout the brain paren-
chyma.23 Perfusion imaging using endogenous contrast
Regional blood ow in brain that is compromised agents
due to a pathological process (such as stroke) can
now be assessed by studying the dynamics of Another popular technique used for perfusion MR
transit of T2* sensitive agents through the vascula- imaging is the arterial spin labelling (ASL)
Technical introduction to MRI 63

Fig. 5.3. Upper left: Series of GRE EPI images acquired from 12 slices every 2 seconds during a bolus injection of 0.1mmol/kg Gd.
Upper right: Signal intensity vs. time curve for a given region of interest. (lower left) Relaxivity vs. time curve obtained from the
perfusion-weighted images and the t. Lower right: Synthesized maps of the hemodynamically weighted images from the R2 data
tting. Note the decrease in signal with time as the contrast agent reaches the region of interest.

technique. An advantage of this technique is that it tions to study and investigate various pathological
is possible non-invasively, to obtain perfusion conditions such as stroke,28 traumatic brain inju-
maps of the brain. ries,29 degenerative brain diseases30 and brain
In this technique, protons in a slice proximal to tumours.31
the imaged slice are tagged with an inversion 180
RF pulse. These saturated blood protons tagged in
the proximal slice move in to the imaged slice due Magnetic resonance spectroscopy
to arterial blood ow during the inversion time
where the signal intensity is lowered relative to the Magnetic resonance spectroscopy (MRS) of the
control image acquired simultaneously. These brain also known as neurospectroscopy is a non-
images are then subtracted to obtain a perfusion invasive MR technique that gives the relative con-
map with hyperintense regions corresponding to centration of certain chemical compounds within
the inowing blood protons. This technique has 2 to 3 cm3 of tissue.
been combined with EPI to enable fast imaging and MRS is based on principles similar to MRI. Each
is called EPISTAR.27 proton is shielded from the static magnetic eld by
The ASL technique has found numerous applica- electrons in the orbital (orbits around the proton in
64 Rohit Sood and Michael Moseley

which the electrons are revolving) around it. This Conclusion

causes the precessional frequency of each proton to
be dierent from its neighbours. The interaction of Magnetic resonance imaging is now an established
the static eld and electrons causes the eld at the clinical imaging modality with applications extend-
nucleus to be slightly altered and is termed as ing from investigative radiology to functional
chemical shift.32 The chemical shift is measured as imaging in neurosciences. The development of
parts per million (ppm) and is proportional to the newer MR techniques such as diusion- and perfu-
magnitude of the external static eld. It is this prop- sion-weighted imaging combined with fast imaging
erty of chemical shift that allows the spectroscopist capabilities on commercial systems has revolution-
to detect a wide variety of individual protons linked ized our understanding of the pathophysiological
to larger molecules such as proteins and carbohy- mechanisms involved in the many clinical condi-
drates. In a typical MRS spectrum obtained from tions especially cerebral ischemia and stroke.
brain tissue (Fig. 5.4), a number of peaks corre- Today, MRI is being increasingly used for investigat-
sponding to frequencies of chemical nuclei will be ing patients with stroke and in the evaluation of
seen, with the area under the peak representing the newer drugs, which may prove to be useful in the
amount of chemical in the tissue. treatement and management of cerebral ischemic
H spectroscopy is the most commonly used disorders.
technique in clinical practice. Localized proton
spectroscopy has recently emerged as a useful tool
for investigating brain diseases such as Alzheimers REFERENCES
disease,33 stroke,34 epilepsy,35 multiple sclerosis,36
and brain tumours37 and is now available on com- 1 Edelman R, Wielopolski R, Schmitt F. Echo Planar
mercial MR systems. Imaging. Radiology 1994; 192: 600.
A typical MRS spectrum of adult brain (spectra 2 Wehrli F, MacFall J, Prost J. Impact of the choice of the
vary for newborns, children under 8 years and operating parameter on MR images. In: Partain C,
elderly) is shown in gure (Fig. 5.4). The spectrum Price R, Patton J, eds. Magnetic Resonance Imaging.
Philadelphia: WB Saunders, 1988.
is read from right to left with the rst and tallest
3 Rother J, De Crespigny A, DArceuil H et al. Recovery
peak assigned to neuronal marker NA (N-acetylas-
of apparent diusion coecient after ischemic
partate) resonating at 2 ppm. The second tallest
induced spreading relates to cerebral perfusion gradi-
resonance is Cr (creatine including phosphocrea- ent. Stroke 1996; 27: 980.
tine) resonating at ~3 ppm. Adjacent to the Cr peak 4 Gore J. Contrast agents and relaxation eects. In: Atlas
is the smaller but important Ch (choline) peak. The S, ed. Magnetic Resonance Imaging of the Brain and
ratio Ch/Cr of about 0.5 is characteristic of brain Spine. New York: Raven Press, 1991.
gray matter. Peaks corresponding to glucose and 5 Gomori J, Grossman R, Goldberg H. Intracranial
myoinositol (ml) are also seen in the spectrum of hematomas: imaging by high eld MR. Radiology
the normal adult brain. 1985; 157: 87.
MRS has been suggested to be more sensitive 6 Gullberg G, Wehrli F, Shimakawa A et al. MR vascular
than MRI in detecting hypoxic damage.38 It has imaging with fast gradient refocusing pulse sequence
and reformatted images from transaxial sections.
been proposed that MRS may be able to detect
Radiology 1987; 165: 241.
cerebral ischemia within seconds of its onset as
7 Masaryk T, Modic M, Ross J et al. Three dimensional
compared to DWI that provides warning as early as
(volume) gradient echo imaging of the carotid bifur-
1 h after the onset. Research has shown that the cation: preliminary clinical experience. Radiology
metabolite prole obtained in patients with stroke 1989; 171: 801.
is characteristic and correlates well with biochemi- 8 Haacke E, Tkach J, Parrish T. Reduction of T2* dephas-
cal changes taking place during the disease ing in gradient eld echo imaging. Radiology 1989;
process.34 170: 457.
Technical introduction to MRI 65

Signa 1.5T SYS#LCMR0C0

Ex:4238 Signa 1.5T SYS#LCMR0C0
probe-p 2000/35 64 avg Probe-test
Se: 5/4 Ex:4238 probe-p 2000/35 64 avg Probe-test
36 M Probe-test 36 M Probe-test
Im:1/1 Se: 6/5
Mach .2 Ratio 03/04/98 Mach .2 Ratio 03/04/98
NA 126 1.66 Im:1/2
Cr 76 REF
13:24 NA 76 1.95 13:33
NA Ch 46 0.61 NA Cr 39 REF
Ch 44 1.13
mI 48 0.65
mI 27 0.69
H2O 120729 1596.89
H2O 87664 2241.00

RMS Noise Cr SNR

RMS Noise Cr SNR
2.24 33.71 Ch 1.62 24.10
Voxel Loc. & Dims.
Voxel Loc. & Dims.
mI Ch R 5.0 P 53.9 S 6.3 Cr
19.2 18.3 20.0 mI L 21.0 P
34.2 S 6.3
18.3 20.0

Fig. 5.4. Normal MR spectra of the adult male human brain. The magnetic resonance spectra of a 36-year-old normal male adult
human brain have been produced by PROBE (GEs spectroscopy sequence) (TE 35 ms). The metabolite, NA is a neuronal marker,
Creatine (Cr) is found in cell energy stores and Choline (Ch) is found in cell membranes. Normal variations in the spectral peaks
are in the order of 15 %. (Image courtesy: Daniel Spielman, Lucas MRIS, Stanford Radiology, Stanford University.)
66 Rohit Sood and Michael Moseley

9 Jung H, Chang K, Choi D et al. Contrast enhanced MR 22 Villringer A, Rosen B, Believeau J. Dynamic imaging
Angiography for the diagnosis of intracranial vascular with lanthanide chelates on normal brain: contrast
disease. Am J Roentgenol 1995; 165: 1251. due to magnetic susceptibility eects. Magn Reson
10 Moseley M, Cohen Y, Kucharczyk J. Anisotropy in Med 1988; 6: 164.
diusion-weighted MRI. Magn Reson Med 1990; 14: 23 Kucharczyk J, Roberts T, Moseley M et al. Applications
330. of contrast enhanced perfusion sensitive MR imaging
11 Moseley M, Kucharczyk J, Mintorovitch J et al. in the diagnosis of cerebrovascular disorders. J Magn
Diusion-weighted MRI of anisotropic water diusion Reson Imaging 1993; 3: 1.
in cat CNS. Radiology 1990; 176: 439. 24 Ostergaard L, Weissko R, Chesler D et al. High resolu-
12 Basser P, Mattiello J, LeBihan D. MR imaging of ber tion measurement of cerebral blood ow using intra-
tract direction and diusion in anisotropic tissues vascular tracer bolus passages. Part I: Mathematical
Twelfth annual meeting of the Society of Magnetic approach and statistical analysis. Magn Reson Med
Resonance Imaging in Medicine. New York: Society of 1996; 36: 15.
Magnetic Resonance in Medicine, 1993. 25 Sorensen G, Reimer P. Perfusion MRI with dynamic
13 Stejskal E, Tanner J. Use of spin echo in pulsed mag- susceptibility contrast imaging. In: Sorensen G, ed.
netic eld gradient to study anisotropic, restricted Cerebral MR Perfusion Imaging: Principles and
diusion and ow. J Chem Phys 1965; 42: 288. Current Applications. Theime Medical Publishers,
14 Beneld A, Prasad P, Edelman R et al. On the optimal 2001.
b-value for measurement of lesion volumes in acute 26 Barbier E, Lamalle L, Decorps M. Methodology of
human stroke by diusion-weighted imaging. Fourth brain perfusion imaging. J Magn Reson Imaging 2001;
annual scientic meeting of the International Society 13: 496.
of Magnetic Resonance Imaging in Medicine. New 27 Edelman R, Siewert B, Darby D et al. Qualitative
York: International Society of Magnetic Resonance in mapping of cerebral blood ow and functional
Medicine, 1996. localization with EPI and signal targeting with alter-
15 Neeman M, Freyer J, Sillerud L. Pulsed-gradient SE nating radio frequency (EPISTAR). Radiology 1994;
studies in NMR imaging: eects of imaging gradients 192: 513.
on the determination of diusion gradients. J Magn 28 Siewert B, Schlaug G, Edelman R, Warach S.
Reson 1990; 90: 303. Comparison of EPISTAR and T2* weighted gadolinium
16 LeBihan D, Breton E, Lallemand D. MR imaging of enhanced perfusion imaging in patients with acute
intravoxel incoherent motion: application to diusion cerebral ischemia. Neurology 1997; 48: 673.
and perfusion in neurologic disorders. Radiology 29 Forbes M, Hendrich K, Kochanek P et al. Assessment
1988; 161: 401. of cerebral blood ow and CO2 reactivity after con-
17 Moseley M, Kucharczyk J, Mintorovitch J. Diusion- trolled cortical impact by perfusion magnetic reso-
weighted MR imaging of acute stroke: correlation nance imaging using arterial spin labeling in rats. J
with T2 weighted and magnetic susceptibility- Cereb Blood Flow Metab 1997; 17: 865.
enhanced MR imaging in cats. Am J Neuroradiol 1990; 30 Alsop D, Detre J, Grossman M. Assessment of cere-
11: 423. bral blood ow in Alzheimers disease by spin
18 Basser P, Mattiello J, Le Bihan D. Estimation of the labeled magnetic resonance imaging. Ann Neurol
eective self-diusion tensor from NMR spin echo. J 2000; 47: 93.
Magn Reson 1994; 103: 247. 31 Gaa J, Warach S, Wen P, Thangaraj V, Weilopolski E.
19 Le Bihan D, Mangin J, Poupon C et al. Diusion tensor Noninvasive perfusion imaging of human brain
imaging: concepts and applications. J Magn Reson tumors with EPISTAR. Eur Radiol 1996; 6: 518.
Imaging 2001; 13: 534. 32 Brown T, Kincaid B, Ugurbil K. NMR chemical shift
20 Teivsky A, Ptak T, Farkas J. Investigation of apparent imaging in three dimensions. Proc Natl Acad Sci USA
diusion coecient and diusion tensor anisotropy 1982; 79: 3523.
in acute and chronic multiple sclerosis. AJNR 1999; 20: 33 Meyerho D, MacKay S, Constans J et al. Axonal injury
1491. and membrane alternations in Alzheimers disease
21 Sorensen G, Reimer P. Cerebral hemodynamics-what suggested by in vivo proton magnetic resonance
are they. In: Sorensen G, ed. Cerebral Mr Perfusion spectroscopy. Ann Neurol 1994; 36: 40.
Imaging : Principles and Current Applications. Theime 34 Howe F, Maxwell R, Saunders D et al. Proton spectros-
Medical Publishers, 2001. copy in vivo. Magn Reson Q 1993; 9: 31.
Technical introduction to MRI 67

35 Cendes F, Andermann F, Dubeau F et al. Proton mag- 37 Barker P, Glickson J, Bryan R. In vivo magnetic reso-
netic resonance spectroscopic images and MRI volu- nance spectroscopy of human brain tumors. J Magn
metric studies for lateralization of temporal lobe Reson Imaging 1993; 5: 32.
epilepsy. Magn Reson Imaging 1995; 13: 1187. 38 Matson G, Weiner M. Spectroscopy. In: Stark D,
36 Arnold D, Mathews P, Franscis G et al. Proton mag- Bradley W, eds. Magnetic Resonance Imaging. New
netic resonance spectroscopy of human brain in vivo York: Mosby, 1999: 182.
in the evaluation of multiple sclerosis: assessment of
the load of disease. Magn Reson Med 1990; 14: 154.
Clinical use of standard MRI

Brian M. Tress
The University of Melbourne Department of Radiology, and Department of Radiology The Royal Melbourne Hospital, Victoria, Australia

Introduction Vasogenic edema spreads easily through the white

matter due to its relatively less dense cellular
MRI is supremely sensitive to the abnormal accu- density and more capacious extravascular space.
mulation of water, much more so than is CT, so it The process of cytotoxic edema predominates in
might be expected that MRI is superior to CT in the the rst 68 hours. Subsequently, vasogenic edema
detection of ischemic infarction. Although this becomes progressively more dominant and is
principle is applicable to subacute infarction, it is largely responsible for the brain swelling in the rst
not the case in the rst 6 hours after the onset of few days after the onset of infarction.
the ischemic insult. A basic understanding of the
pathological processes which precede infarction is
necessary in order to understand their MRI mani- MRI technique
The basic standard MRI sequences which should be
applied in screening for stroke are T1 and T2-
Pathology of ischemic infarction weighted axial or sagittal scans through the whole
brain. The type of T2-weighted scans used is impor-
Deprivation of oxygen supply to neurones, whether tant. Conventional T2-weighted spin echo
as a result of embolus, thrombosis or prolonged sequences allow the acquisition of minimally
hypotension, leads initially to malfunction in as T2-weighted images or proton-weighted images as
little as a few seconds, e.g. a StokesAdams attack, the rst echo of a double echo sequence. Images
in which brainstem ischemia induces unconscious- acquired from the second echo of the spin echo
ness within seconds of cardiac asystole. If the sequence will be both heavily T2-weighted and sen-
ischemic insult is prolonged the highly energy sitive to the magnetic susceptibility eects of ferro-
dependent sodium pump mechanism, which is magnetic blood products such as
responsible for maintaining a tenfold dierence in deoxyhemoglobin, intracellular methemoglobin
extracellular to intracellular sodium concentration, and hemosiderin. The T2-weighted multiple spin
begins to fail and sodium, water and calcium ions echo sequences used most commonly include fast
pass from the extracellular to the intracellular spin echo, turbo spin echo and rapid spin echo.
space. The cell swells, producing cytotoxic edema Although they provide exquisite T2 weighting and
and the extracellular space is simultaneously spatial resolution, they suer in that they are rela-
reduced. If the diminished oxygen supply is main- tively insensitive to magnetic susceptibility eects
tained, the less energy-dependent capillary endo- and, hence, to hemorrhage. In the stroke context
thelial cells start to lose their function and the that is unacceptable. Therefore if one of these tech-
normally tight junctions between them begin to niques is used, a further T2-weighted magnetic sus-
lose their integrity. Intravascular uid leaks into the ceptibility sensitive technique should be added.
extravascular space, producing vasogenic edema. Appropriate sequences include echoplanar imaging

70 Brian M. Tress


Fig. 6.1. Right middle cerebral artery territory infarct at 6 hours. T1 and T2-weighted scans through vertex (a), (b) show swollen
right precentral and postcentral gyri, but no intrinsic signal abnormality. T1-weighted post contrast scan (c) shows stasis in pre-
central and central arteries and pial enhancement from leptomeningeal collateral feeders.

spin echo (EPI SE) or T2-weighted gradient echo more detailed time of ight (TOF) study of the
(GE). intracranial circulation can be obtained in more
A very appropriate substitute for proton- cooperative patients.
weighted sequences is uid attenuated inversion Diusion-weighted imaging (DWI), when avail-
recovery (FLAIR). While retaining heavy T2- able, should be a mandatory sequence in any
weighting the cerebrospinal uid is nulled, making protocol for stroke evaluation.
any parenchymal abnormality more conspicuous.
Furthermore, FLAIR sequences have been reported
Plain MRI appearances of ischemic infarction
to be the only MRI sequence with high sensitivity
and specicity in the detection of acute subarach-
Conventional T1- and T2-weighted sequences
noid hemorrhage1 and to be able to detect vessel
occlusion and perfusion decits more sensitively Appearances vary dramatically, according to the
than conventional spin echo sequences2,3. FLAIRs time at which imaging takes place after the
drawback is its relative insensitivity to infratentorial ischemic insult. If imaging is initiated within the
intraaxial lesions. rst 6 hours, during which time cytotoxic edema
In hyperacute stroke MRA sequences provide val- predominates, standard T2- and T1-weighted
uable and accurate information in respect to the images are completely normal in the majority of
patency of the major intracranial vessels. A rapid cases. In a minority of cases careful scrutiny of the
phase contrast MRA sequence can be obtained in T1-weighted images may show subtle evidence of
restless patients in less than 2 minutes, while a mass eect (Fig. 6.1). This is in contrast to CT, the
Clinical use of standard MRI 71


(c )

Fig. 6.1. cont.

72 Brian M. Tress

(a) (b)

Fig. 6.2. Twelve-month-old infarct in left occipital lobe and 1-week-old infarct in right occipital lobe. (a) T1-weighted and (b)
T2-weighted images show uid lled gliotic area in left occipital lobe with compensatory dilatation of left occipital horn and
slightly hemorrhagic and swollen right occipital lobe infarct.

sensitivity of which is variably reported from 30 to acute time frame. Sharply demarcated zones of
80% within the rst 12 hours. hypointensity on T1-weighted scans are particularly
By 12 hours, at which time a signicant portion likely to represent older infarcts (Fig. 6.2).
of the induced edema is of the vasogenic type, T2- Occasionally slowed ow within supplying arter-
weighted scans show evidence of prolonged T2 ies will be seen as hyperintensity within the vessels,
relaxation, manifested as increased signal in the particularly in T1-weighted and FLAIR scans (Fig.
aected area. In the case of classical peripheral 6.3). More commonly, absence of the normal ow
embolic infarction the abnormal signal is charac- void signal is seen in either or both T1- and
teristically within the cerebral cortex, as well as the T2-weighted scans. This sign is more easily iden-
underlying white matter, leading to a relative loss of tied in the larger vessels of the Circle of Willis and
grey/white dierentiation. The peripheral portions the M1 and M2 segments of the middle cerebral
of the abnormal signal are in some cases sharply artery. When present in the middle cerebral artery,
dened, often clearly demarcating the border zones this sign can be regarded as the MRI equivalent of
between classical vascular territories (Fig. 6.2). the dense middle cerebral artery sign, as seen with
Occasionally peripheral vasogenic edema predomi- CT. Absence of ow void or even hyperintense
nates. It is seen as nger like projections within signal is usually easily identied within the cavern-
white matter. Prolonged T1 relaxation is often not as ous segment of the internal carotid artery, the inter-
prominent in the rst 24 hours, so that reduced nal carotid in the carotid canal and within the
signal intensity on T1-weighted scans implies a less basilar artery. It is important to recognize that these
Clinical use of standard MRI 73

(a) (b)

Fig. 6.3. Right-sided medullary infarct at 24 hours. (a) T2-weighted scan shows hyperintense signal in right side of medulla, sharply
delineated medially at the midline. Note markedly reduced signal void in right vertebral artery. FLAIR image (b) also shows right
medullary infarct and right vertebral artery hyperintensity, indicating right vertebral occlusion or severe stenosis or markedly
slowed ow.

signal changes can be the result of occlusion or

Contrast-enhanced T1-weighted sequences
severe stenosis and slow ow.
A second blind period for conventional The injection of an intravenous contrast medium
sequences has been described in the second week, containing gadolinium signicantly increases the
when both T1- and T2-weighted sequences can ability of conventional sequences to diagnose
revert almost to normality, even in the presence of a infarction. Within the rst 24 hours the signal inten-
sizeable completed infarct (Fig. 6.4)4. This is attrib- sity in arteries supplying the infarcted region is
uted to revascularization and petechial hemorrhage increased due to slow ow and T1 shortening by the
eectively neutralizing the eect of vasogenic and gadolinium in approximately 80% of cases (Fig.
any persistent cytotoxic edema. Even diusion- 6.1).5 From day 2 to day 5 cortical surface contrast
weighted sequences may be negative if it happens enhancement may result from leptomeningeal col-
that imaging is taking place at the point that diu- laterals (Fig. 6.1).5
sion is about to evolve from restriction to increase. From approximately 4 days parenchymal
At this stage contrast enhanced sequences are par- enhancement commences as a result of revascular-
ticularly valuable because the brain parenchyma ization by leaky vessels. The phenomenon is seen
will usually show vivid parenchymal contrast in 100% of cases by day 7.5,6 The contrast enhance-
enhancement (Fig. 6.4). Sequences particularly sen- ment pattern is variable. A gyral pattern of
sitive to blood breakdown products, such as gradi- enhancement is common (Fig. 6.4), but streaky,
ent echo T2* sequences, will usually detect even the patchy and even ring patterns of enhancement may
small amount of blood associated with petechial be seen. Some form of parenchymal enhancement
hemorrhage (Fig. 6.5). is seen in almost 100% of infarcts during the second
74 Brian M. Tress

(a) (b)

Fig. 6.4. Infarct involving right insular cortex.

(a) At 24 hours a T2-weighted scan shows
clearcut increased signal intensity within
the right insular cortex.
(c ) (b) At 10 days the T2-weighted scan shows
only very subtle hyperintensity in the
right insular cortex.
(c) Post-contrast T1-weighted scan shows
vivid parenchymal contrast enhancement
in the insular cortex and adjacent right
frontal cortex.
Clinical use of standard MRI 75

(a) (b)

(c ) (d )

Fig. 6.5. Left frontal infarct at 12 days.

(a) T1-weighted image shows serpentine hyperintense areas in infarct periphery.
(b) T2-weighted fast spin echo image shows extensive areas of signal void within central portions of infarct.
(c) T2-weighted EPI gradient echo image from diusion weighted sequence without added gradients (Bo image), demonstrates
far more extensive signal voids than T2-weighted FSE image.
(d ) Phase contrast MRA shows left internal carotid occlusion, with left middle cerebral artery patent via ow from the right inter-
nal carotid and anterior cerebral arteries.
76 Brian M. Tress

and third weeks, but has largely disappeared by the order middle cerebral branches, but takes in
second month. excess of 4 minutes for data acquisition. This may
All forms of contrast enhancement are seen in result in degraded images due to movement
only a minority of small deep (lacunar) infarcts. induced artefact when scanning restless acute
stroke patients.
MRA of the carotid arteries in the neck has been
Fluid attenuated inversion recovery (FLAIR)
signicantly improved by the development of con-
FLAIR is a routinely available technique which pro- trast enhanced MRA7 (see Chapter 7), which has
duces heavily T2-weighted images, at the same time increased the accuracy of the evaluation of the
nulling or completely subtracting the normally carotid bifurcation, particularly in regards to spec-
bright cerebrospinal uid signal. The technique is icity. The addition of this technique to the stan-
particularly suited to the identication of periven- dard evaluation means that information regarding
tricular lesions such as demyelination plaques, and the condition of brain parenchyma and the major
is often used as a substitute for minimally T2- arteries of supply from arch to circle of Willis can
weighted (proton weighted) sequences because of be evaluated in one examination, provided the
the relative conspicuity of any lesion with a pro- patient is able to co-operate. Even non-contrast-
longed T2. Its sensitivity in the detection of supra- enhanced MRA is at least equal to duplex Doppler
tentorial lesions is arguably greater than that of ultrasound in the evaluation of the carotid bifurca-
conventional T2-weighted sequences. The major tion.8 Its major advantage is in the demonstration
disadvantage of the sequence is its relative insensi- of atherosclerotic disease in Asian and black
tivity to brainstem and cerebellar lesions. ethnic groups, in whom the intracranial arteries
Like conventional T2-weighted sequences FLAIR are more frequently involved than in caucasian
shows no parenchymal signal abnormality in the populations.
rst 68 hours after the ischemic insult. However, it
has been observed that vessels in the ischemic
Magnetization Transfer Contrast (MTC)
region will show a hyperintense signal instead of
the usual signal void in 1096 % of cases (Figs. This technique exploits the cross-relaxation
6.3,6.6).2,3 The hyperintense signal is associated between mobile water protons and the restricted
particularly with large vessel occlusion or severe proton pool, which includes protons in or around
stenosis, slow ow and decreased perfusion. It cor- macromolecules such as myelin. The application of
responds to intravascular enhancement seen in a broad frequency, o resonance pulse reduces the
contrast enhanced sequences. magnetization of the restricted proton pool to zero,
while only partly reducing the free proton magnet-
ization. Contrast enhancement after intravenous
Magnetic resonance angiography (MRA) contrast medium injection can be maximized by
this technique.9 The ratio of signal intensity before
MRA is a very useful adjunct to conventional and after application of the o resonance pulse
sequences. Middle cerebral or internal carotid (magnetization transfer ratio, or MTR) provides an
occlusion can be conrmed in acute stroke indicator of demyelination, which has lead to its
patients, with appropriate impact on management. use in multiple sclerosis. In subacute and chronic
The various forms of MRA have been described in stroke MTR has been found to be closely related to
Chapter 7. Phase contrast MRA can be performed axonal damage and to correlate well with motor
in less than 2 minutes to provide acceptable decit.10
images of the terminal internal carotid arteries The principal use for MTC in the stroke context is
and the vertebrobasilar system and the Circle of in TOF MRA, where it suppresses background
Willis (Fig. 6.7). Time of Flight MRA provides more parenchymal signal, increasing the conspicuity of
detailed anatomical images, often showing third small vessels.
Clinical use of standard MRI 77

(a) (b )

(c ) (d )

Fig. 6.6. FLAIR images of 4-hours-old infarct, showing high signal intensity within stem of left middle cerebral artery (a) and
Sylvian branches (b) but no parenchymal signal change. Diusion-weighted image (c) demonstrates restricted diusion in left
middle cerebral artery territory, indicating early infarct. Phase contrast MRA (d ) shows left middle cerebral artery occlusion at its

Hemorrhagic infarction time. They are normal sequelae of bland infarction

and should not be considered part of the spectrum
Approximately 60% of ischemic or bland infarcts of true hemorrhagic infarction.
develop evidence of petechial hemorrhage in the True hemorrhagic infarction is most often secon-
second and subsequent weeks. This is seen as dary to initially bland infarcts in which the feeding
patchy small, multifocal areas of signal loss within arteries have recanalized, either spontaneously, or
the infarcted cortex on T2-weighted sequences (Fig. as a result of therapeutic thrombolysis, allowing
6.5). These hemorrhages are not suciently large to high pressure blood ow into irreversibly damaged
be seen in most CT scans performed at the same capillaries. It is also a frequent sequel to venous
78 Brian M. Tress

(a) (b)

(c ) (d )

Fig. 6.7. Right middle cerebral artery territory infarct at 6 hours.

(a) T2-weighted scan shows no abnormality
(b) Phase contrast MRA shows right middle cerebral artery occlusion.
(c) T2-weighted scan shows clearcut infarct in right temporal lobe at 3 days.
(d) Phase contrast MRA demonstrates restoration of right middle cerebral artery patency.

thrombosis, usually after thrombus has retrogradely T2-weighted GE.11 The rst images of diusion-
propagated into cortical veins from the venous weighted sequences before the application of the
sinuses. The MRI appearance varies accordingly. superimposed gradients are appropriately mag-
Hyperacute parenchymal hemorrhage can be netic susceptibility sensitive, as the diusion
detected by MRI by utilizing one of the T2-weighted sequences are usually EPI T2-weighed GE or SE
magnetic susceptibility sensitive sequences such as (Fig. 6.5(c)).
Clinical use of standard MRI 79

Arterial origin hemorrhagic infarcts usually have Differential diagnosis

a background of characteristic bland infarct
appearances. They are largely conned to an arte-
Normal anatomical variants
rial vascular territory and involve the cortex as well
as the deeper structures. Superimposed on this VirchowRobin spaces are perivascular spaces sur-
background are heterogeneous zones of soft tissue rounding perforating arteries. They represent
with the typical appearance of hematoma. The extensions of the subarachnoid space into the brain
signal characteristics of the hematoma component and are particularly prominent in the region of the
are dependent on its age. anterior perforated substance, where the lenticulo-
Venous infarcts, on the other hand, transgress striate arteries enter the base of the brain in the
arterial territorial boundaries (Fig. 6.8). Those middle cranial fossae. The only consistent associa-
venous infarcts secondary to superior sagittal sinus tion is with increasing age.14 They are virtually
thrombosis are often bilateral in the superior con- always visible in T1- and T2-weighted spin echo
vexities of the hemispheres. The hemorrhagic images of the lentiform nuclei and internal capsule,
element is heterogeneous and contrast enhance- where they occasionally reach sizes of more than
ment is often minimal or absent. If the hemorrhage one centimetre (Fig. 6.9). They are characteristically
is more than 3 days old the aected venous sinuses sited in close relation to the anterior commissure.
may be hyperintense on T1-weighted scans (Fig. They can be dierentiated from lacunar infarcts in
6.8). Occasionally the involved cortical veins may that they are round or linear and follow the signal
be similarly hyperintense. If the thrombotic process intensities of cerebrospinal uid on all sequences.15
has propagated into the medullary veins, subepen- It is important to perform proton-weighted
dymal veins such as the internal cerebral veins and sequences or FLAIR, as heavily T1- and T2-weighted
the great vein of Galen may also be hyperintense, or sequences alone cannot distinguish between the
contain no ow void. Extension to the deep venous two entities.
system produces an almost pathognomonic pattern
of ischemic edema, involving both thalami and the
Brain tumours
upper brainstem.
Magnetic resonance venography (MRV), which is In the vast majority of cases mass eect, enhance-
basically MRA with arterial ow suppressed, will ment pattern, surrounding vasogenic edema and
often provide the denitive diagnosis of sinus transgression of more than one arterial territory
thrombosis, but there are pitfalls. There may be easily dierentiates tumour from infarct, even in
gaps in up to 31% of non-dominant transverse the absence of a characteristic history.
sinuses in normal subjects.12 Acute thrombus may Occasionally, an infarct can be truly simulated by
produce signal voids on T2-weighted sequences, tumour in an isolated examination. Gliomas of
due to the presence of deoxyhemoglobin or intra- low or intermediate grade are most often respon-
cellular methemoglobin. The voids can simulate sible. Protoplasmic, non-enhancing astrocytomas
patency. This artefact can be overcome by carefully can be conned to the cortex and underlying
perusing the sinuses in all sequences.13 Subacute medulla in an apparent single vascular territory
thrombus may appear hyperintense on T1-weighted and have sharp margins. Alternatively, an infarct
images. The maximal pixel intensity algorithm used can sometimes induce abundant vasogenic
to construct two and three dimensional images edema and vivid contrast enhancement, occa-
from time of ight MRA source images cannot dis- sionally with a ring pattern to simulate a malig-
tinguish between the hyperintensity of owing nant tumour. If MRS and diusion-weighted
blood and subacute thrombus. The MRV images imaging are not immediately available the sim-
may then appear normal. Phase contrast MRV is plest method of dierentiating between tumour
more reliable as only moving protons are included and infarct is simply to rescan in 1 to 2 weeks
in the nal image (Fig. 6.8).13 time. An infarct will have signicantly changed in
80 Brian M. Tress

(a) (b)

(c) (d)

Fig. 6.8. Axial T2-weighted image (a) through posterior fossa structures shows mottled signal within the right sigmoid sinus. Note
inammatory changes in the adjacent mastoid air cells. Axial T2-weighted image through occipital lobes (b) shows irregular lobu-
lated zones of abnormal signal in lateral temporal and occipital lobes.
Sagittal and lateral sinus thrombosis. Sagittal T1-weighted images show hyperintensity in superior saggittal sinus (c) and lateral
sinus (d ). Note hemorrhagic infact in adjacent occipital lobe. MRV studies conrm lack of ow in superior sagittal sinus (e) and
right lateral sinus (f ).
Clinical use of standard MRI 81

(e ) (f )

Fig. 6.8. (cont.)

appearance, both in mass eect and contrast

enhancement pattern, while a tumour will be
unchanged or slightly bigger.


The clinical presentation will usually be signi-

cantly dierent, but the MRI appearances can be
similar. The major dierence is that inammatory
processes do not respect normal arterial territories.
Herpes simplex encephalitis characteristically
involves one or both temporal lobes (Fig. 6.10). The
involvement is predominantly in the posterior cere-
bral artery supplied portion of the medial temporal
lobe initially, but almost invariably has already
involved the middle cerebral artery supplied lateral
aspect by the time of rst presentation. As the
inammation extends superiorly through the exter-
nal capsule, the lentiform nucleus is almost always
preserved, unlike middle cerebral artery territory
Fig. 6.9. VirchowRobin (perivascular) spaces. T1-weighted The early cerebritis stage which precedes the
axial image shows symmetrical, large uid lled ovoid spaces development of cerebral abscess may manifest as
in the anterior perforated substance in close proximity to the an area of nonspecic non-enhancing edema,
anterior commissure. usually at the cortico-medullary junction. It is more
82 Brian M. Tress



Fig. 6.10. Herpes simplex encephalitis. T1-weighted (a) and T2-weighted axial images, demonstrate abnormal signal and swelling
in the anterior portion of the left temporal lobe and less marked changes in the medial portion of the right temporal lobe.
Note that the left temporal lobe abnormality transgresses the left middle and posterior cerebral artery territory.
Clinical use of standard MRI 83

(a) (b)

Fig. 6.11. Left frontal cerebritis. T2-weighted scan (a) shows abnormal signal in left frontal lobe anteriorly with pattern of vaso-
genic edema in the underlying white matter. Note lesion crosses the left anterior and middle cerebral artery territories.
Post-contrast T1-weighted scan (b) shows patchy left frontal enhancement.

often associated with a vasogenic pattern of edema REFERENCES

than are infarcts, but may have a patchy enhance-
ment pattern before the development of a capsule 1 Noguchi K, Ogawa T, Inugami, A et al. Acute subarach-
and ring enhancement at about 2 weeks (Fig. 6.11). noid haemorrhage: MR imaging with uid-attenuated
inversion recovery pulse sequences. Radiology 1995;
Once developed, the capsule will usually show evi-
196: 773777.
dence of signal voids on T2-weighted scans, thought
2 Kamran S, Bates V, Bakshi R, Wright P, Kinkel W,
to be due to a combination of microhaemorrhages
Milatich R. Signicance of hyperintense vessels on
and free radicals. FLAIR MRI in acute stroke. Neurology 2000; 55:
3 Toyoda K, Ida M, Fukuda K. Fluid-attenuated inver-
Conclusion sion recovery intraarterial signal: an early sign of
hyperacute cerebral ischemia. Am J Neuroradiol 2001;
Standard MRI sequences are extremely sensitive to 22: 10151016.
the detection of acute, subacute and chronic 4 Pereira AC, Doyle VL, Clifton A, Howe FA, Griths JR,
infarcts. However, standard sequences are negative Brown MM. Case reports. The transient disappearance
in the majority of infarcts (within the rst 68 of cerebral infarction on T2 weighted MRI. Clin Radiol
2000; 55: 725727.
hours). Sensitivity is increased by the use of con-
5 Elster AD, Moody DM. Early cerebral infarction;
trast enhancement and FLAIR, but accurate diag-
gadopentate dimeglumine enhancement. Radiology
nosis of hyperacute infarction depends on the use
1990; 177: 627632.
of diusion-weighted imaging. Standard MRA 6 Karonen JO, Partanen PL, Vanninen RL, Vainio PA,
sequences add further valuable information regard- Aronen HJ. Evolution of MR contrast enhancement
ing patency of major intracranial vessels. patterns during the rst week after acute ischemic
stroke. Am J Neuroradiol 2001; 22: 103111.
84 Brian M. Tress

7 Fellner FA, Felher C, Wufke, R et al. Fluoroscopically 11 Patel MR, Edelman RR, Warach S. Detection of hyper-
triggered contrast-enhanced 3D MR DSA and 3D time acute primary intraparenchymal hemorrhage by mag-
of ight turbo MRA of the carotid arteries: rst clinical netic resonance imaging. Stroke 1996; 27: 23212324.
experiences in correlation with ultrasound, X-ray 12 Ayanzen RH, Bird CR, Keller PJ, McCully FJ, Theobald
angiography and endarterectomy ndings. Magn MR, Heiserman JE. Cerebral MR venography: normal
Reson Imaging 2000; 18: 575585. anatomy and potential diagnostic pitfalls.
8 Patel MR, Kurtz KM, Klufas RA, et al. Preoperative assess- Am J Neuroradiol 2000; 21: 7478.
ment of the carotid bifurcation. Can magnetic resonance 13 Provenzale JM, Joseph GJ, Barboriak DP. Dural sinus
angiography and duplex ultrasonography replace con- thrombosis: ndings on CT and MR imaging and
trast angiography? Stroke 1995; 26: 17531758. diagnostic pitfalls. Am J Radiol 1998; 170: 777783.
9 Mathews VP, King JC, Elster AD, Hamilton CA. 14 Heier LA, Bauer CJ, Schwartz L, Zimmerman RD,
Cerebral infarction: eects of dose and magnetization Morgello S, Deck MD. Large Virchow-Robin spaces:
transfer saturation at gadolinium-enhanced MR MR-clinical correlation. Am J Neuroradiol 1989; 10:
imaging. Radiology 1994; 190: 547552. 929936.
10 Pendlebury ST, Lee MA, Blamire AM, Styles P, 15 Bokura H, Kobayashi S, Yamaguchi S. Distinguishing
Matthews PM. Correlating magnetic resonance silent lacunar infarction from enlarged Virchow-Robin
imaging markers of axonal injury and demyelination spaces: a magnetic resonance imaging and pathologi-
in motor impairment secondary to stroke and multi- cal study. J Neurol 1998; 245: 116122.
ple sclerosis. Magn Reson Imaging 2000; 18: 369378.
MR angiography of the head and neck: basic principles and
clinical applications

Robert R. Edelman and Joel Meyer

Department of Radiology, Evanston Northwestern Healthcare
Northwestern University School of Medicine, 2650 Ridge Avenue, Evanston, IL 60201, USA

Since the rst publication of a clinical magnetic res- properties that can be manifested in an MR image
onance angiogram (MRA) in 1985,1 there has been can be a great advantage for distinguishing arteries
extensive growth in the vascular applications of and veins from stationary background tissue, and
magnetic resonance.2 Although for many years a permits the measurement of ow (not possible by
purely investigational tool, the technology and vali- CT), but it also introduces ample sources for arte-
dation studies have progressed to the point that facts.
MRA has largely supplanted X-ray angiography Blood has a long T1 relaxation time (1.2 s). The
(XRA) for evaluation of the extracranial carotid T2 relaxation time is substantially longer for arterial
arteries, and is often an alternative to XRA for eval- blood (high oxyhemoglobin content), as compared
uation of the vertebral arteries and circle of Willis. with venous blood (high deoxyhemoglobin
In this chapter, we will rst review the basic prin- content). Without the confounding inuence of
ciples of MRA, including time of ight and phase ow, blood would appear dark (e.g. similar to
contrast techniques, and introduce the use of para- muscle on T1-weighted images) and bright on T2-
magnetic contrast agents for MRA. Advantages and weighted images. However, ow has a profound
pitfalls of MRA as compared with duplex sonogra- eect on the signal intensity of blood so that the
phy (DUS) and XRA will be addressed, and we will intrinsic relaxation properties of arterial and
consider future directions for this rapidly advanc- venous blood are not usually apparent, the excep-
ing technology. tion being where ow is absent (i.e. stasis in a

Basic principles of MRA

Time of flight

Unlike computed tomography (CT), which relies One class of MRA techniques is called time of
solely on the attenuation of X-ray photons to gener- ight.3 The term refers to the fact that the transit of
ate an image, MR uses a combination of magnetic blood from one position to another will alter the
elds and radiofrequency energy in order to signal intensity from the spins, and enable its
produce images. The appearance of blood in an MR dierentiation from stationary tissues. On spin-
image depends on the intrinsic magnetic relaxation echo and fast spin-echo images, the time-of-ight
properties (T1 and T2), the oxygenation status and eect causes rapidly owing spins to appear dark
physical state of the blood (e.g. venous vs. arterial (hence the terminology dark blood technique).
vs. hematoma), the direction, rate, and pulsatility of The reason is that the owing spins wash out of the
ow, as well as the presence of exogenously admin- plane of section before the sequence of 90 and 180
istered contrast agents. The variety of physical degree pulses has been played out, so that the MR

86 Robert R. Edelman and Joel Meyer

signal is lost. Additionally, extra radiofrequency ties are lower, as for venography of the intracranial
pulses called saturation pulses may be applied to venous sinuses. Alternatively, one can use a multi-
the inowing spins so as to further saturate and ple overlapping thin-slab 3D acquisition (MOTSA)
suppress their signal.4 Alternatively, one can more (Fig. 7.1), which is less sensitive to signal loss from
eectively eliminate the signal from blood by using saturation and turbulence than 3D or 2D, respec-
the dual inversion method.5,6 Dark blood tech- tively.10
niques are helpful for routine anatomic imaging Both 2D and 3D time-of-ight MRA are helpful for
and for direct imaging of the vessel wall and associ- assessing the extracranial carotid bifurcation. The
ated plaque or thrombus.7,8 2D technique generally yields good image quality
For MRA, on the other hand, one is usually trying for non-stenotic vessels and in slow ow situations
to render the owing spins brighter than other (e.g. string sign). However, it is not as helpful for the
tissues, so that one can apply a maximum intensity circle of Willis. The 3D technique provides more
projection (MIP) algorithm to produce an accurate measurements of stenoses because of its
angiogram-like picture.9 In order to achieve this lesser sensitivity to signal loss from turbulent ow
goal, a gradient-echo pulse sequence is applied. and smaller voxels. In both cases, images are
Unlike the spin-echo case, only a single radiofre- acquired over a small eld of view in the axial orien-
quency pulse is used during each repetition of the tation, so that only a relatively small portion of the
sequence, so that the washout eects can be carotids in the vicinity of the bifurcation is imaged.
neglected. Typically, one uses a short repetition Thus, a tandem stenosis would be easily missed, as
time (TR) so that scan times are minimized (of the is also the case with duplex sonography.
order of a few seconds for 2D acquisitions and less
than 30 seconds for 3D acquisitions). The combina-
Phase contrast
tion of the short TR and a large excitation ip angle
depresses the signal intensity of stationary tissues, The pulse sequences used for MR imaging typically
as these spins do not have adequate time to recover apply a pair of gradients of opposite polarity
their longitudal magnetization between sequence (bipolar) along the frequency and slice-select
repetitions. directions. The net phase shift for stationary spins
Conversely, rapidly owing spins move into and at the echo time (TE) is zero. However, moving
out of the plane of section in such a short time spins accumulate a non-zero phase shift from the
period that they only experience one or a few bipolar gradients; this velocity-dependent phase
radiofrequency pulses before they are replaced by shift is the basis for phase contrast MRA. The phase
fresh, fully magnetized spins. Thus, the owing shift can be used to quantify velocity on phase-
spins produce an intense signal that is readily contrast images.11,12 Alternatively, the phase shift
dierentiated from stationary tissue, particularly if allows owing blood to be displayed without any
the orientation of the blood vessel is nearly orthog- signal from the stationary tissues. This feature
onal to the plane of section. The data may be comes in handy when there are paramagnetic
directly acquired as thin sections (2D) or as a thick hemoglobin breakdown products or contrast-
slab that is subsequently reconstructed into thin enhancing tissues in the vicinity of a blood vessel,
sections (3D). In the case of 3D, the ow must be as these structures would appear bright in a time of
more rapid than in the 2D scenario since a thick ight MRA. Two-dimensional (thick slice) phase-
slab, rather than thin slice, is imaged and the contrast MRA can be obtained in a matter of
owing spins have a greater distance to traverse seconds, whereas 3D phase-contrast MRA more
before exiting. Thus, 3D time-of-ight MRA is typically is acquired over several minutes.13
reserved for situations where the expected ow Phase-contrast MRA is sensitive to velocities over
velocities are high, such as the extracranial carotid a range that is determined by the ratio of velocity to
arteries and Circle of Willis, whereas 2D time of the user-selected velocity-encoding sensitivity
ight would be greatly preferred where the veloci- (VENC). Ideally, a VENC is chosen that is slightly
MR angiography of the head and neck: basic principles and clinical applications 87

(a) (b)

(c) (d)

Fig. 7.1. Two-dimensional TOF v. MOTSA. (a), (b) Two-dimensional TOF MRA: targeted MIP projection of the left carotid and ver-
tebral artery shows loss of normal signal of the left internal carotid artery several centimetres distal to the bifurcation. Similar loss
of signal in the adjacent left external carotid artery favours swallowing artefact rather stenoses. (c), (d ) MOTSA: targeted MIP pro-
jection of the left carotid bifurcation and verterbral artery conrms the presence of a swallowing artefact with normal appearance
of both internal and external carotid arteries.

larger than the peak velocity. For instance, a VENC MRA will alias blood owing at high velocities will
of 1 m/s might be chosen for an MRA of the falsely appear to ow at lower velocities, or the
abdominal aorta, whereas a VENC of 10 cm/s might signals may be lost due to dephasing. Conversely, if
be chosen for imaging of the intracerebral veins. If the VENC is too large, then the signal-to-noise ratio
the VENC is too small, then the phase-contrast is diminished for the blood vessels.
88 Robert R. Edelman and Joel Meyer

There are several helpful clinical applications of in order to measure the diameter accurately. With
phase-contrast MRA in the head and neck. Two- current gradient technology, typical in-plane reso-
dimensional phase contrast can be used as a scout lution for CE 3DMRA is of the order of 1 mm, which
sequence for the carotid arteries, for rapid assess- could not be adequate to measure a 2 mm stenosis
ment of dural sinus patency and for the evaluation of the internal carotid artery. Enhancement of
of collateral ow patterns. Three-dimensional background tissues can make it dicult to process
phase-contrast MRA provides a lengthier but more the images into a MIP. This is particularly a problem
complete depiction of the intracranial veins14. for imaging of the carotid siphon due to rapid
Phase-contrast MRA is eective even after the enhancement of the cavernous sinus, and as a
administration of a paramagnetic contrast agent; in result time of ight remains the preferred tech-
fact, the vascular signal-to-noise ratio will benet. nique for MRA of the Circle of Willis. Another draw-
back is the cost of the contrast agent. However, the
combination of shorter scan time allowing for
Contrast-enhanced MRA
higher patient throughput, the improved reliability
Time-of-ight images can suer from artefacts in and eld of view as compared with time-of-ight,
the presence of slow ow or turbulence. The desire and better image quality may balance concerns
to eliminate such artefacts prompted the develop- about the additional cost.
ment of contrast-enhanced 3D techniques, which
lessen the sensitivity to turbulence and slow Contrast agent dosage and timing
ow.15,16 The typical dose of gadolinium chelate for extracra-
Paramagnetic contrast agents boost the signal nial carotid and vertebral studies is on the order of
intensity from blood by greatly shortening its T1 0.1 mmol/kg delivered intravenously at a rate of
relaxation time. For instance, during the rst pass approximately 2 cc/s, followed by 10 cc of saline
of a typical dose of gadolinium chelate (e.g. administered at the same rate. The matrix and slice
0.1 mmol/kg or around 1020 cc for an average- thickness should be selected such that the voxel
sized adult), the T1 relaxation time of blood may be dimensions are of the order of 1 mm or less. It is
shortened by a factor of 10 or more. As a conse- critical to time the start of the 3D data acquisition
quence, one can use a very short repetition time so that it is coincident with the arrival and lling of
(TR, e.g. 5 ms) in order to acquire entire 3D data the target vessel with contrast medium. Several
sets in much shorter times (e.g. 1020 seconds) approaches have been used for this purpose. One
than with time-of-ight or phase contrast. can administer a 12 cc test bolus and image with a
The benets of using a contrast agent can be sub- fast 2D sequence in order to time the arrival of con-
stantial. For instance, contrast-enhanced 3D MRA trast medium.17 One can place a tracker region
(CE 3DMRA) never suers from saturation artefact, within the vessel of interest and automatically
since the vessels appear bright by virtue of the detect the arrival of the contrast bolus.18 A foolproof
eects of the contrast agent alone. Because the TE timing approach is to image with a rapid,
is so short (e.g. 2ms), CE 3DMRA is insensitive to uoroscopic 2D acquisition and then trigger the 3D
signal loss from turbulent ow, an effect that arte- acquisition when the contrast medium is seen to
factually enlarges stenoses in time-of-ight and arrive in the target vessel.
phase-contrast MRA. The scan time is typically an
order of magnitude shorter than with 3D time-of- Technical considerations
ight and phase-contrast. The optimal technique used for CE 3DMRA
There are also drawbacks to the use of contrast depends on the particular hardware and software
agents. The need for a very short TR and TE may congurations of the MR system. For instance, MR
limit spatial resolution, so that the accuracy of systems with the newest gradient hardware can
measuring stenoses may be degraded. For instance, achieve TRs of the order of 4 ms or less, so that
one typically needs at least 3 pixels to span a vessel rst-pass, high resolution studies of the extracra-
MR angiography of the head and neck: basic principles and clinical applications 89

nial carotid arteries can be acquired in the 15 s or Comparison of time-of-ight and CE 3DMRA
less needed to avoid jugular venous enhancement. Although the benets of contrast enhancement are
With further shortening of TR and a modest reduc- well recognized, published data do not yet fully
tion in the number of phase-encodes, one can support the replacement of time-of-ight tech-
acquire multiphase MRA with a time resolution of niques for carotid MRA. There are certain circum-
10 s or better, thereby eliminating the need to get stances where CE 3DMRA is clearly advantageous
the timing of the contrast bolus exactly right.19 (Fig. 7.3). For instance, because data can be
However, such rapid imaging may come at the acquired in a coronal or sagittal orientation span-
expense of spatial resolution, which is needed for ning a large eld of view, the method is preferred
accurate measurement of carotid stenoses.20 for studies that need to detect tandem stenoses
Moreover, older MR systems cannot provide such involving the origin or siphon of the carotid artery.
short TRs, in which case a dierent approach This is advantageously done in conjunction with a
called elliptico-centric phase-encoding21 is head and neck phased array coil, which spans the
helpful. region from the aortic arch through the circle of
The concept underlying elliptico-centric phase- Willis. The CE 3DMRA method is also preferred for
encoding is that the order in which the phase- patients with tortuous vessels, as well as for
encode gradients are played out will alter the patients who might not hold still for the 510
appearance of the image in fundamental ways. minutes required for 3D time-of-ight MRA.
Data acquired during the play-out of the weakest Turbulence causes exaggeration of stenosis with
phase-encoding gradients (centre of k-space) deter- time of ight or phase contrast methods, but is
mine the overall image contrast, whereas data much less an issue with CE 3DMRA. On the other
acquired during the play-out of the stronger phase- hand, the nominal spatial resolution for time-of-
encoding gradients (outer portions of k-space) ight MRA is usually better than with CE 3DMRA,
determine spatial resolution. In a 3D sequence, two which in some circumstances improves the
orthogonal phase-encoding gradients are played specicity of the stenosis measurement.
out (vs. one phase-encode gradient for a 2D
sequence). If the two phase-encode gradients are
played out in the usual linear order, then the image Clinical applications
contrast is similar throughout the period of data
acquisition. In the elliptico-centric scheme, the
Extracranial carotid and vertebral arteries
weakest phase-encodes are played out for both
phase-encode directions at the start of the data Stenotic disease
acquisition, so that this early data determines the In 1991 the NASCET study showed that carotid
eventual tissue contrast of the MRA. For instance, if endarterectomy gave superior outcomes to medical
these early data are acquired during the rst pass of therapy for symptomatic patients with signicant
contrast agent through the extracranial carotid extracranial carotid stenosis.22 A more recent study
arteries prior to enhancement of the jugular veins, indicated that even asymptomatic patients may
then the jugular veins will not be visible in the MRA benet from surgical therapy.23 Although X-ray
(Fig. 7.2). This is true even if the data acquisition is angiography was the gold standard for the NASCET
extended over a minute or longer. Thus, the benet study, it involves a small stroke risk of the order of
of elliptico-centric phase-encoding is that one can 0.54%.24 Consequently, there is a strong need for
dierentiate the arteries from veins even though an accurate non-invasive test for extracranial
one has a slower gradient system and must use a carotid disease.
lengthier data acquisition. The longer data acquisi- In a blinded study involving 176 carotid arteries,
tion provides the opportunity to collect more data 3D TOF MRA had a sensitivity of 94%, a specicity
and therefore to obtain higher spatial resolution of 85% and an accuracy of 88% for the identication
than would be possible during a rst-pass study. of 70% to 99% stenosis; two-dimensional TOF MRA
90 Robert R. Edelman and Joel Meyer

(a) (b)

Fig. 7.2. Gadolinum-enhanced MRA with elliptic centric encod-

(c) ing: carotid stenosis. (a), (b), (c) AP view and multiple oblique
projections show bilateral stenoses of the carotid arteries. On
the left, there is high grade stenosis of the common and left
internal carotid arteries. On the right, there is moderate steno-
sis of the common carotid artery bifurcation extending to
involve the right internal carotid artery. Note the excellent
visualization of the origins of the great vessels and the vertebral

had a sensitivity and specicity that were approxi- and DUS can replace X-ray angiography in the
mately 10% lower than those of three-dimensional majority of patients (Figs. 7.4, 7.5). The use of CE
TOF MRA.25 Duplex sonography had a sensitivity of 3DMRA can eliminate the artefacts associated with
94%, a specicity of 83%, and an accuracy of 86%. time of ight MRA and also permit a much larger
Combining data from three-dimensional TOF MRA eld of view that encompasses the carotid origins
and DU, allowing for CA only for disparate results, and siphons.26 A recent study comparing CE
yielded a sensitivity of 100%, a specicity of 91%, 3DMRA, CTA, and digital subtraction angiography
and accuracy of 94% among concordant non-inva- found that both MRA and CTA accurately depicted
sive tests, with CA required in 16% of arteries. The extracranial carotid stenoses, although the superior
authors concluded that the combination of MRA spatial resolution aorded by CTA made it easier to
MR angiography of the head and neck: basic principles and clinical applications 91

(a) (b)

Fig. 7.3. Two-dimensional TOF vs. enhanced MRA: string sign.

(c) (a) Two-dimensional TOF demonstrates diminished ow
within the left internal carotid artery (arrows). Note the
decreased signal intensity of the left ICA vs. the right ICA. (b),
(c) Enhanced MRA better demonstrates the abnormal left
internal carotid through the level of the skull base (arrows).
Note the irregular areas of narrowing involving the left
pre-cavernous and cavernous internal carotid arteries.

detect ulcerations.27 With ongoing improvements in used in patients with extracranial carotid steno-
spatial resolution, the performance of CE 3DMRA sis.29,30 as well as lesions of the Circle of Willis.
appears to match that of DSA.28
Flow measurement Magnetic resonance can be used not only to image
By using 2D phase contrast angiography, particu- the lumen of the vessel but the wall as well using
larly with cardiac-gated cine to depict ow over the dark blood imaging methods. MRI is highly accu-
entire cardiac cycle, one can measure ow veloci- rate at measuring the carotid wall area31 and T2-
ties and ow rates in the extracranial carotid arter- weighted images can discriminate between the
ies as well as in other vessels. This method can be brous cap and lipid core of a plaque.32
92 Robert R. Edelman and Joel Meyer

(a) (b)

Fig. 7.4. Intracranial stenosis: right internal carotid artery. (a)

(c) Axial T2-weighted MR shows absence of the normal ow void
within the right internal carotid artery (curved arrow). (b) MIP
projection of the anterior circulation demonstrates decreased
signal intensity of the right internal carotid relative to the left
internal carotid artery. (c) Targeted MIP of the right internal
carotid artery conrms a high grade stenosis of the cavernous
right internal carotid artery.

Dissections of the extracranial portions of the or intimal ap, mural hematoma, and luminal
carotid and vertebral arteries can be traumatic or irregularity similar to that demonstrated by X-ray
spontaneous in origin. The combination of MRI angiography. One can also use MRA to serially
and MRA is an accurate non-invasive means for the follow the response to antiplatelet or anticoagulant
detection of carotid and vertebral dissections (Fig. therapy.38 Most studies of MRA for carotid and ver-
7.6).3337 Typically, MRI reveals a crescentic intra- tebral dissection have relied on time of ight or
mural hyperintensity on both T1- and T2-weighted phase contrast techniques. Consequently, patient
images. MRA may reveal a smoothly tapered vessel motion, turbulence, or slow ow can cause an
MR angiography of the head and neck: basic principles and clinical applications 93

(a) (b)

Fig. 7.5. Vertebral artery origin stenosis. (a), (b), (c) Enhanced
(c) MRA of the aortic arch and great vessels shows excellent visual-
isation of the vertebral artery origins with stenoses bilaterally.
Volume rendered oblique projection better demonstrates the
proximal vertebral lesions.

incomplete or erroneous interpretation. Given their aicted more with extracranial carotid stenosis,
smaller size and greater tortuousity, the accuracy is whereas the intracranial circulation is aected
less for vertebral than carotid dissections. It is likely more commonly in blacks, Hispanics, Japanese,
that CE 3DMRA will give more reliable results, but and Chinese.39 In setting of acute stroke, informa-
this is not yet veried in the published literature. tion about the status of the intracranial arteries is
critical to determination of stroke subtype and
Circle of Willis associated prognosis, and helps to determine the
Stenotic disease choice to employ thrombolytic therapy.40
The incidence of intracranial arterial stenosis Transcranial doppler, MRA,4143 and CTA are non-
varies among ethnic groups. Whites tend to be invasive approaches for evaluating intracranial
94 Robert R. Edelman and Joel Meyer

(a) (b)

Fig. 7.6. Gadolinium-enhanced MRA with elliptic centric

(c) encoding: carotid dissection bilateral pseudoaneurysms. (a),
(b), (c) Oblique MIP projections demonstrate dilation of the
cervical internal carotid arteries bilaterally consistent with
pseudoaneurysms associated with bilateral carotid dissections
(arrows). Volume rendered projections more clearly document
the bilateral carotid pseudoaneurysms.

disease. Transcranial doppler is inexpensive and In another study, imaging with DWI and MRA
available at the bedside, but in some cases there is within 24 hours of hospitalization substantially
not an adequate temporal window. In one study, a improved the accuracy of early ischemic stroke
total of 50 middle cerebral arteries in 25 patients subtype diagnosis.45 While useful for evaluation of
with a history of ischemic stroke were studied with the Circle of Willis,46 time of ight MRA also has
MRA and CTA.44 Inter- and intraobserver variabil- limitations. Whereas spatial resolution is adequate
ity was good and comparable to that achieved with for the primary branches of the Circle of Willis, it is
duplex sonography for extracranial carotid steno- less so in the more distal, smaller branches and
sis, although MRA proved more reliable than CTA. the level of diagnostic condence diminishes.
MR angiography of the head and neck: basic principles and clinical applications 95

(a) (b)

Fig. 7.7. Unruptured anterior communicating artery aneurysm. (a), (b) 73-year-old male with headache. MIP of the TOF MRA
anterior circulation shows bilobed aneurysm of the anterior communicating complex. Note the hypoplastic right A1 segment.
Volume rendering provides better appreciation of the relationship of the aneurysm to the anterior cerebral artery origins.

Collateral ow and function

Collateral ow patterns are readily shown by selec- 90%, the reported sensitivities have ranged from
tive application of presaturation pulses over 0.67 to nearly 1.5761 A prospective study involving
feeding vessels47 or by phase contrast MRA. Phase 142 patients demonstrated an accuracy per patient
contrast MRA has been applied to demonstrate col- of 0.87 at CTA and 0.85 at MRA.62 The accuracy per
lateral ow patterns in patients with internal aneurysm for the best observer was 0.73 at CTA
carotid artery obstructions.48,49 In patients with and 0.67 at MRA. Interobserver agreement was
moyamoya, a progressive cerebrovascular occlusive good; however, the sensitivity for detection of
disorder of unknown etiology, the diagnostic accu- aneurysms smaller than 5 mm was 0.57 for CTA
racy of MRA has been reported to be excellent, and 0.35 for MRA compared with 0.94 and 0.86,
close to that of X-ray angiography.50,51 Moreover, respectively, for detection of aneurysms 5 mm or
MRA can be used to assess vascular anatomy and larger. The accuracy of both CTA and MRA was
hemodynamic response after surgical lower for detection of internal carotid artery
intervention.52 MRA has also been used in conjunc- aneurysms compared with that at other sites. Thus,
tion with acetazolamide challenge to assess the MRA is not a complete replacement for X-ray
vasoreactivity of the intracranial vessels for normal angiography. The accuracy of MRA might be
and stenotic vessels53 and to show age-related sucient to screen patients who are at risk for
decreases in cerebral blood ow.54 intracranial aneurysms.63 The costbenet of
screening MRA is controversial. One study of
Aneurysm screening MRA found aneurysms in 25 of 626 rst-
Both MRA55,56 and CTA have been reported to be degree relatives of patients with history of sub-
accurate tests for intracranial aneurysms (Fig. 7.7), arachnoid bleed from ruptured intracranial
although the sensitivity diminishes greatly for aneurysms.64 The authors state that
smaller lesions. Despite an accuracy of the order of Implementation of a screening program for the
96 Robert R. Edelman and Joel Meyer

(a) (b)

Fig. 7.8. MR venography 2D TOF: venous thrombosis. (a), (b),

(c) (c), (d ), (e ) Axial T2-MR (a) imaging of the posterior fossa
demonstrates low signal intensity involving the left transverse
sinus mimicking a normal ow void. Axial T1-weighted MR (b)
of the transverse sinus show subtle increased signal that sug-
gests recent sinus thrombosis. T1-images with gadolinium (c),
(d ) document a large lling defect involving the left transverse
sinus as well as collateral circulation along the left tentorium
cerebellum. Two-dimensional TOF MR venography (e) shows
no ow within the transverse sinuses bilaterally with collateral
circulation noted on the right.

rst-degree relatives of patients with sporadic sub- sucient to obviate the need for selective X-ray
arachnoid hemorrhage does not seem warranted at angiography prior to intervention. Contrast-
this time, since the resulting slight increase in life enhanced 3D MRA should provide more complete
expectancy does not oset the risk of postoperative evaluation. One study using real-time monitored
sequelae. Other studies favour the use of MRA.65 auto-triggered elliptic centric-ordered MRA found
it to be superior to TOF MRA for evaluation of
Vascular malformation intracranial AVMs, resulting in improved depiction
MRI depicts the nidus of a vascular malformation, of the components of an AVM, in particular the
whereas MRA depicts the feeding arteries and nidus and draining veins.69 When compared with
draining veins.6668 However, image quality is not DSA, the authors reported that CE 3DMRA consis-
MR angiography of the head and neck: basic principles and clinical applications 97

(d) (e)

tently depicted AVM components and their orienta- particularly dicult to achieve for neck applica-
tion. tions.

Balanced steady-state free precession

Intracranial veins
Although available for many years, balanced
Patency of the dural sinuses and cerebral veins can steady-state free precession (SSFP) techniques
be well assessed by any of several techniques, have only come into their own with the availability
including 2D time of ight,70 3D phase contrast,7174 of faster gradients and shorter TR/TE. Examples of
as well as CE 3D MRA (Fig. 7.8). The combination of SSFP include true FISP76 and FIESTA. SSFP images
MRI and MRA is ideal as it provides information show a contrast dependence on the T2/T1 ratio.
about vessel patency as well as the signal character- Because of its long T2 relaxation time, blood natu-
istics and thus age of thrombus.75 rally appears bright with SSFP imaging. The tech-
nique enables the use of very short TR so that
imaging is extremely rapid. Although SSFP acqui-
Works in progress sitions are relatively insensitive to the T1 shorten-
ing eect of a contrast agent, one can sensitize the
Spiral imaging acquisition by means of an inversion prepulse for
Although spoiled gradient-echo pulse sequences CE MRA. Drawbacks of SSFP include artefacts
such as FLASH (fast low angle shot) and MPSPGR from o-resonance eects (improved through the
(multiplanar spoiled gradient-echo) predominate use of very short TR) and the fact that other
for time of ight and CE 3D MRA, other pulse tissues, such as cerebrospinal uid and fat, also
sequences may be helpful. For instance, spiral appear bright.
imaging is a method whereby two orthogonal gradi-
ents are oscillated to collect a series of echoes. The Variants in K-space sampling
method is insensitive to ow-related dephasing so Since not all portions of k-space are equal with
that vessels appear bright. However, excellent shim- respect to tissue contrast, there are potential
ming of the main magnetic eld is required to avoid benets to sampling the various portions of k-space
blurring from o-resonance eects, which may be at diering rates. This approach enables more rapid
98 Robert R. Edelman and Joel Meyer

imaging during the rst pass of a contrast agent. toolbox for neurovascular disorders. The technol-
For instance, one might collect the outer portions ogy is not entirely mature insofar as the pulse
of k-space less frequently than the central portions sequences and gradient hardware continue to
since the amplitudes of the latter are relatively improve. For instance, steady-state free precession
insensitive to the eects of the contrast bolus and and parallel imaging methods have not yet been
remain fairly constant. Examples of this approach extensively evaluated for neurovascular applica-
include keyhole77 and TRICKS.78 By sampling the tions, but no doubt will improve the accuracy and
outer portions of k-space less frequently than speed of diagnosis. It is now generally accepted
usual, the scan time is substantially shortened pro- that MRA can replace X-ray angiography for the
viding the opportunity for rapid, multiphase presurgical evaluation of the carotid endarectomy
imaging while maintaining spatial resolution. candidate. There remain circumstances where X-
ray angiography is still needed, such as in a patient
Parallel imaging who has a pacemaker or a potentially ferromag-
The time and spatial resolution of neurovascular netic intracranial aneurysm clip (MRI contraindi-
MRA studies can be greatly improved by the use of cated), as well as in those patients who are unable
parallel imaging techniques.7981 Sample commer- to remain still long enough to acquire an artefact-
cial nomenclature for parallel imaging methods free MRA study. X-ray angiography may be indi-
includes SMASH, SENSE, and ASSET. The basic cated as the arbitrating examination for the patient
idea of parallel imaging is that the individual ele- in whom the duplex sonogram and MRA yield
ments of the phased array coil used to detect the strikingly divergent results. In addition to X-ray
MR signal have a sensitivity for the signal that angiography, contrast-enhanced CTA using fast,
depends on distance and orientation from the coil multislice scanners can provide accurate depiction
element. With the use of appropriate mathemati- of the extra- and intracranial carotid arteries, and
cal algorithms, this property can be used to extract is advocated by some as a more reliable test than
phase-encode information in parallel to the MRA. However, CT is relatively contraindicated in
phase-encode information acquired through the someone with signicantly impaired renal function
application of the magnetic eld gradient. The net or allergy to iodinated contrast medium.
eect is that the scan time can be reduced by a For the intracranial circulation, MRA is well
factor up to the number of coil elements, or the accepted for imaging of the dural sinuses and larger
spatial resolution can be improved by this factor. intracranial veins, e.g. for the diagnosis of dural
In practice, accelerations of 24 have been shown, sinus thrombosis. Using phase contrast MRA one
and in principle much higher acceleration factors can easily depict patterns of collateral ow.
are feasible by the use of appropriately designed Stenoses involving the proximal branches of the
phased array coils. With the availability of parallel circle of Willis are well shown, although the accu-
imaging, which is just beginning to come into clin- racy declines as one attempts to image smaller,
ical practice, the only real limit on speed and more distal vessels. MRA is reliable for detecting
spatial resolution for MRA is the available signal- stenoses and occlusions involving the vertebro-
to-noise ratio, which will depend on magnetic basilar system, although the specicity will depend
eld strength, contrast agent dosage and rate of on the level of spatial resolution (as is also the case
infusion, as well as radiofrequency coil and for the extracranial carotid arteries). Although MRI
patient-related factors. is invaluable for detecting vascular malformations
and MRA can depict major feeding arteries and
draining veins, it has not eliminated the need for X-
Conclusions ray angiography prior to embolization or surgical
In a relatively short period of time, MRA has come We see that MRA has already had a major clinical
to occupy a valuable place in the diagnostic impact despite the relative immaturity of the tech-
MR angiography of the head and neck: basic principles and clinical applications 99

nology. No doubt the breadth and depth of clinical gradient-echo MR imaging. Am J Neuroradiol 1991; 12:
applicability will continue to expand as the tech- 481488.
nology moves forward. 15 Prince MR. Gadolinium-enhanced MR aortography.
Radiology 1994; 191: 155164.
16 Cloft HJ, Murphy KJ, Prince MR, Brunberg JA. 3D
gadolinium-enhanced MR angiography of the carotid
arteries. Magn Reson Imaging 1996; 14: 593600.
17 Kim JK, Farb RI, Wright GA. Test bolus examination in
the carotid artery at dynamic gadolinium-enhanced
1 Wedeen VJ, Meuli RA, Edelman RR et al. Projective
MR angiography. Radiology 1998; 206: 283289.
imaging of pulsatile ow with magnetic resonance.
18 Foo TKF, Saranathan M, Prince MR, Chenevert TL.
Science 1985; 230: 946948.
Automated detection of bolus arrival and initiation of
2 Yucel EK, Anderson CM, Edelman RR et al.
data acquisition in fast, three dimensional,
Magnetic resonance angiography: update on applica-
gadolinium-enhanced MR angiography. Radiology
tions for extracranial arteries. Circulation 1999; 100:
1997; 203: 275280.
19 Levy RA, Makin JH. Three-dimensional contrast-
3 Keller P. Time of ight magnetic resonance angiogra-
enhanced MR angiography of the extracranial carotid
phy. Neuroimaging Clin N Am 1992; 4: 639656.
arteries: two techniques. Am J Neuroradiol 1998; 19:
4 Felmlee JP, Ehman RL. Spatial presaturation: a method
for suppressing ow artefacts and improving depic-
20 Serfaty JM, Chirossel P, Chevallier JM. Accuracy of
tion of vascular anatomy in MR imaging. Radiology
three-dimensional gadolinium-enhanced MR angio-
1987; 164: 559564.
graphy in the assessment of extracranial carotid artery
5 Edelman RR, Chien D, Kim D. Fast selective black
disease. Am J Radiol 2000; 175: 455463.
blood MR imaging. Radiology 1991; 181: 655660.
21 Wilman AH, Riederer SJ. Performance of an elliptical
6 Simonetti OP, Finn JP, White RD, Laub G, Henry DA.
spiral centric view order for signal enhancement and
Black blood T2-weighted inversion-recovery MR
motion artefact suppression in breathhold three
imaging of the heart. Radiology 1996; 199: 4957.
dimensional gradient echo imaging. Magn Reson Med
7 Demarco JK, Rutt BK, Clarke SE. Carotid plaque char-
1997; 38: 793802.
acterization by magnetic resonance imaging: review
22 North American Symptomatic Carotid
of the literature. Top Magn Reson Imaging 2001; 12:
Endarterectomy Trial Collaborators. Benecial eect
of carotid endarterectomy in symptomatic patients
8 Edelman RR, Mattle HP, Wallner B et al. Extracranial
with high-grade stenosis. N Engl J Med 1991; 325:
carotid arteries: evaluation with black blood MR
angiography. Radiology 1990; 177: 4550.
23 Executive Committee for the Asymptomatic Carotid
9 Laub G. Displays for MR angiography. Magn Reson
Atherosclerosis Study. Endarterectomy for asympto-
Med 1990; 14: 222229.
matic carotid artery stenosis. J Am Med Assoc 1995;
10 Blatter DD, Parker DL, Ahn SS et al. Cerebral MR
273: 14211428.
angiography with multiple overlapping thin slab
24 Hankey GJ, Warlow CP, Sellar RJ. Cerebral angio-
acquisition. Radiology 1992; 183: 379389.
graphic risk in mild cerebrovascular disease. Stroke
11 Dumoulin CL. Phase contrast MR angiography tech-
1990; 21: 209222.
niques. Magn Reson Imaging Clin N Am 1995; 3:
25 Patel MR, Klufas RA, Kim D et al. Preoperative assess-
ment of the carotid bifurcation: can magnetic reso-
12 Walker MF, Souza SP, Dumoulin CL. Quantitative ow
nance angiography and duplex ultrasonography
measurement in phase contrast MR angiography. J
replace contrast arteriography? Stroke 1995; 26:
Comput Assist Tomogr 1988; 12: 304313.
13 Pernicone JR, Siebert JE, Potchen EJ, Pera A,
26 Remonda L, Heid O, Schroth G. Carotid artery steno-
Dumoulin CL, Souza SP. Three-dimensional phase-
sis, occlusion, and pseudo-occlusion: rst-pass,
contrast MR angiography in the head and neck. Am J
gadolinium-enhanced, three-dimensional MR angio-
Neuroradiol 1990; 155: 457466.
graphy preliminary study. Radiology 1998; 209:
14 Tsuruda JS, Shimakawa A, Pelc NJ, Saloner D. Dural
sinus occlusion: evaluation with phase-sensitive
27 Randoux B, Marro B, Koskas F et al. Carotid artery
100 Robert R. Edelman and Joel Meyer

stenosis: prospective comparison of CT, three- Thrombolytic therapy with streptokinase in acute
dimensional gadolinium-enhanced MR, and conven- ischemic stroke. N Engl J Med 1996; 335: 145150.
tional angiography. Radiology 2001; 220: 179185. 41 Rother J, Schwartz A, Wentz KU, Rautenberg W,
28 Huston J, Fain SB, Wald JT. Carotid artery: elliptic Hennerici M. Middle cerebral artery stenosis: assess-
centric contrast-enhanced MR angiography com- ment by magnetic resonance angiography and tran-
pared with conventional angiography. Radiology 2001; scranial Doppler ultrasound. Cerebrovasc Dis 1994; 4:
218: 138143. 273279.
29 van Everdingen KJ, Klijn CJM, Kappelle LJ, Mali 42 Fujita N, Hirabuki N, Fujii K et al. MR imaging of
WPTM, van der Grond J. MRA ow quantication in middle cerebral artery stenosis and occlusion: value of
patients with a symptomatic internal carotid artery MR angiography. Am J Neuroradiol 1994; 15: 335341.
occlusion. Stroke 1997; 28: 15951600. 43 Korogi Y, Takahashi M, Mabuchi N et al. Intracranial
30 Vanninen R, Koivisto K, Tulla H, Manninen H, vascular stenosis and occlusion: diagnostic accuracy
Partanen K. Hemodynamic eects of carotid of three-dimensional, Fourier transform, time-of-
endarterectomy by magnetic resonance ow ight MR angiography. Radiology 1994; 193: 187193.
quantication. Stroke 1995; 26: 8489. 44 Wong KS, Lam WWM, Liang E, Huang YN, Chan YL, Kay
31 Yuan C, Beach KW, Hillyer Smith L, Jr, Hatsukami TS. R. Variability of magnetic resonance angiography and
Measurement of atherosclerotic carotid plaque size in computed tomography angiography in grading middle
vivo using high resolution magnetic resonance cerebral artery stenosis. Stroke 1996; 27: 10841087.
imaging. Circulation 1998; 98: 26662671. 45 Lee LJ, Kidwell CS, Alger J, Starkman S, Saver JL.
32 Toussaint JF, LaMuraglia GM, Southern JF, Fuster V, Impact on stroke subtype diagnosis of early diusion-
Kantor HL. Magnetic resonance images lipid, brous, weighted magnetic resonance imaging and magnetic
calcied, hemorrhagic, and thrombotic components resonance angiography. Stroke 2000; 31: 1081.
of human atherosclerosis in vivo. Circulation 1996; 94: 46 Masaryk TJ, Modic MT, Ross JS et al. Intracranial cir-
932938. culation: preliminary clinical experience with three-
33 Klufas RA, Hsu L, Barnes PD, Patel MR, Schwartz RB. dimensional volume MR angiography. Radiology 1989;
Dissection of the carotid and vertebral arteries: 171: 793799.
imaging with MR angiography. Am J Roentgenol 1995; 47 Edelman RR, Mattle HP, OReilly GV, Wentz KU, Cheng
164: 673677. L, Zhao B. Magnetic resonance imaging of ow
34 Levy C, Laissy JP, Raveau V et al. Carotid and vertebral dynamics in the circle of Willis. Stroke 1990; 21: 5665.
artery dissections: three-dimensional time-of-ight 48 Miralles M, Dolz JL, Cotillas J et al. The role of the
MR angiography and MR imaging versus conventional circle of Willis in carotid occlusion: assessment with
angiography. Radiology 1994; 190: 97103. phase contrast MR angiography and transcranial
35 Provenzale JM, Morgenlander JC, Gress D. duplex. Eur J Vasc Endovasc Surg 1995; 10: 424430.
Spontaneous vertebral dissection: clinical, conven- 49 Hartkamp MJ, van der Grond J, van Everdingen KJ,
tional angiographic, CT, and MR ndings. J Comput Hillen B, Mali WPTM. Circle of Willis collateral ow
Assist Tomogr 1996; 20: 185193. investigated by magnetic resonance angiography.
36 Homan M, Sacco RL, Chan S, Mohr JP. Noninvasive Stroke 1999; 30: 2671.
detection of vertebral artery dissection. Stroke 1993; 50 Hasuo K, Mihara F, Matsushima T. MRI and MR
24: 815819. angiography in moyamoya disease. J Magn Reson
37 Kitanaka C, Tanaka J, Kuwahara M, Teraoka A. Imaging 1998; 8: 762766.
Magnetic resonance imaging study of intracranial ver- 51 Houkin K, Aoki T, Takahashi A, Abe H. Diagnosis of
tebrobasilar artery dissections. Stroke 1994; 25: moyamoya disease with magnetic resonance angio-
571575. graphy. Stroke 1994; 25: 21592164.
38 Kasner SE, Hankins KK, Bratina P, Morgenstern LB. 52 Yoon H-K, Shin H-J, Lee M, Byun HS, Na DG, Han BK.
Magnetic resonance angiography demonstrates vas- MR Angiography of moyamoya disease before and
cular healing of carotid and vertebral artery dissec- after encephaloduroarteriosynangiosis. Am J Radiol
tions. Stroke 1997; 28: 19931997. 2000; 174: 195200.
39 Caplan LR, Gorelick PB, Hier DB. Race, sex and occlu- 53 Mandai K, Sueyoshi K, Fukunaga R et al. Evaluation of
sive cerebrovascular diseases: a review. Stroke 1986; cerebral vasoreactivity by three-dimensional time-of-
17: 648655. ight magnetic resonance angiography. Stroke 1994;
40 Multicenter Acute Stroke Trial-Europe Study Group. 25: 18071811.
MR angiography of the head and neck: basic principles and clinical applications 101

54 Buijs PC, Krabbe-Hartkamp MJ, Bakker CJ et al. Eect evaluation of three-dimensional time-of-ight MR
of age on cerebral blood ow: measurement with angiography. Radiology 1990; 175: 443448.
ungated two-dimensional phase-contrast MR angiog- 68 Kauczor HU, Engenhart R, Layer G et al. 3D TOF MR
raphy in 250 adults. Radiology 1998; 209: 667674. angiography of cerebral arteriovenous malformations
55 Ross JS, Masaryk TJ, Modic MT, Ruggieri PM, Haacke after radiosurgery. J Comput Assist Tomogr 1993; 17:
EM, Selman WR. Intracranial aneurysms: evaluation by 184190.
MR angiography. Am J Neuroradiol 1990; 11: 449455. 69 Farb RL, McGregor C, Kim JK et al. Intracranial arteri-
56 Schuierer G, Huk WJ, Laub G. Magnetic resonance ovenous malformations: real-time auto-triggered
angiography of intracranial aneurysms: comparison elliptic centric-ordered 3D gadolinium-enhanced MR
with intra-arterial digital subtraction angiography. angiography initial assessment. Radiology 2001; 220:
Neuroradiology 1992; 35: 5054. 244251.
57 Ogawa T, Okudera T, Noguchi K et al. Cerebral 70 Mattle HP, Wentz KU, Edelman RR et al. Cerebral
aneurysms: evaluation with three-dimensional CT venography with MR. Radiology 1991; 178:453458.
angiography. Am J Neuroradiol 1996; 17: 447454. 71 Liauw L, van Buchem MA, Spilt A et al. MR angiogra-
58 Preda L, Gaetani P, Rodriguez et al. Spiral CT angiogra- phy of the intracranial venous system. Radiology 2000;
phy and surgical correlations in the evaluation of 214: 678682.
intracranial aneurysms. Eur Radiol 1998; 8: 739745. 72 Tsuruda J, Saloner D, Normal D. Artefacts associated
59 Huston J, III, Nichols D, Luetmer P et al. Blinded with MR neuroangiography. Am J Neuroradiol 1992;
prospective evaluation of sensitivity of MR angiogra- 13: 14111422.
phy to known intracranial aneurysms: importance of 73 Medlock MR, Olivero WC, Hanigan WC, Wright RM,
aneurysm size. Am J Neuroradiol 1994; 15: 16071614. Winek SJ. Children with cerebral venous thrombosis
60 Aprile I. Evaluation of cerebral aneurysms with MR- diagnosed with magnetic resonance imaging and
angiography. Riv Neuroradiol 1996; 9: 541550. magnetic resonance angiography. Neurosurgery 1991;
61 White PM, Wardlaw JM, Easton V. Can noninvasive 31: 870876.
imaging accurately depict intracranial aneurysms? A 74 Vogl TJ, Bergman C, Villringer A, Einhaupl K, Lissner J,
systematic review. Radiology 2000; 217: 361370. Felix R. Dural sinus thrombosis: value of venous MR
62 White PM, Teasdale EM, Wardlaw JM, Easton V. angiography for the diagnosis and follow-up.
Intracranial aneurysms: CT angiography and MR Radiology 1994; 162: 11911198.
angiography for detection prospective blinded com- 75 Isensee C, Reul J, Thron A. Magnetic resonance
parison in a large patient cohort. Radiology 2001; 219: imaging of thrombosed dural sinuses. Stroke 1994; 25:
739749. 2934.
63 Raaymakers TW, Buys PC, Verbeeten Jr B et al. MR 76 Carr JC, Simonetti O, Bundy J et al. Angiography of the
angiography as a screening tool for intracranial heart with segmented true fast imaging with steady-
aneurysms: feasibility, test characteristics, and inter- state precision. Radiology 2001; 219: 828834.
observer agreement. Am J Radiol 1999; 173: 14691475. 77 van Vaals JJ, Brummer ME, Dixon WT et al. Keyhole
64 The Magnetic Resonance Angiography in Relatives of method for accelerating imaging of contrast agent
Patients with Subarachnoid Hemorrhage Study uptake. J Magn Reson Imaging 1993; 3: 671675.
Group. Risks and benets of screening for intracranial 78 Korosec FR, Frayne R, Grist TM, Minstretta CA. Time-
aneurysms in rst-degree relatives of patients with resolved contrast-enhanced 3D MR angiography.
sporadic subarachnoid hemorrhage. [see comments]. Magn Reson Med 1996; 36: 588595.
N Engl J Med 1999; 341: 13441350. 79 Sodickson DK, Manning WJ. Simultaneous acquisition
65 Brown BM, Soldevilla F. MR angiography and surgery of spatial harmonics (SMASH): fast imaging with
for unruptured familial intracranial aneurysms in radiofrequency coil arrays. Magn Reson Med 1997; 38:
persons with a family history of cerebral aneurysms. 591603.
Am J Radiol 1999; 173: 133138. 80 Sodickson DK, McKenzie CA, Li W, Wol S, Manning
66 Edelman RR, Wentz KU, Mattle HP et al. Intracerebral WJ, Edelman RR. Contrast-enhanced 3D MR angio-
arteriovenous malformations: evaluation with selec- graphy with simultaneous acquisition of spatial har-
tive MR angiography and venography. Radiology 1989; monics: A pilot study. Radiology 2000; 217: 284289.
173: 831837. 81 Weiger M, Pruessmann KP, Kassner A et al. Contrast-
67 Marchal G, Bosmans H, Van Fraeyenhoven L et al. enhanced 3D MRA using SENSE. J Magn Reson
Intracranial vascular lesions: optimization and clinical Imaging 2000; 12: 671677.
Stroke MRI in intracranial hemorrhage

Peter D Schellinger,1 Olav Jansen2 and Werner Hacke1

Department of Neurology, University of Heidelberg, Germany
Department of Neuroradiology, University of Kiel, Germany

Introduction of residues from ICH were only demonstrated by

MRI. This chapter will cover the MRI signatures of
Non-traumatic intracranial hemorrhage (ICH) ICH in all stages, but with a focus on the dieren-
accounts for 1015% of all strokes, but up to 25% tial diagnosis in hyperacute stroke patients, where
of more severe strokes.1,2 The etiology of ICH in stroke MRI becomes more and more important
the vast majority of patients is arterial hyperten- for guiding therapy in ischemic stroke patients. It
sion (63.5%), followed by coagulopathies (15%) will briey deal with subarachnoid hemorrhage
and vessel malformations (8.5%), and less fre- (SAH), and discuss future prospects such as the
quently amyloid angiopathy, vasculitis, intoxica- detection of perihemorrhagic pathological pro-
tions, cavernoma, or cerebral venous thrombosis.3 cesses, which may contribute to the morbidity of
In the hyperacute emergency assessment ICH.
(< 612 h) computed tomography (CT) is the diag-
nostic standard and modality of choice to dier-
entiate between hyperacute ICH and ischemic MRI signatures of acute, subacute and chronic
stroke.46 In general, MRI at this stage is consid- ICH
ered to be of little value for the diagnosis of intra-
cerebral or subarachnoidal hemorrhage, and Gomori and Grossman had already described in
many authors claim that the sensitivity of MRI for the mid-1980s the stage-dependent signal inten-
detecting hyperacute ICH is poor.710 Throughout sity of ICH in dierent MRI sequences in experi-
the chapter we arbitrarily dened hyperacute as mental13 as well as clincial situations.1417 The
(< 12 h), acute (12 h to 7 d), subacute (7 d to 3 mo) ability to detect acute hemorrhage by MRI is
and chronic (> 3 mo) stages. While hyperacute related to the oxygen saturation of hemoglobin
ICH is hyperdense on acute CT scans with pro- and its degradation. As a hematoma ages, the
gressing time and hematoma degradation, there is hemoglobin passes through several forms (oxyhe-
a loss of density and the ICH may appear isodense moglobin, deoxyhemoglobin, and methemoglo-
or hypodense. MRI is far superior to CT in the bin) prior to red cell lysis and breakdown into
subacute and chronic stages especially with ferritin and hemosiderin. The T2 relaxation rate
regard to concomitant or underlying pathology.11 varies quadratically with the concentration of
In a study of 129 patients with ICH Steinbrich et deoxyhemoglobin.13 Acute hematomas are charac-
al. found sensitivities of 46% (MRI) and 93% (CT) terized by central hypointensity on T2-weighted
in the hyperacute and acute stage but 97% (MRI) images (T2WI) whereas they appear isointense on
and 58% (CT) in the subacute and 93% (MRI) and conventional T1WI. Methemoglobin formation
17% (CT) in the chronic stage.12 Also, petechial leads to T1-shortening and therefore to a centri-
bleeding, discrete foci of contusion and evidence petal hyperintensity on T1WI, which also changes

104 Peter D. Schellinger, Olav Jansen and Werner Hacke

Table 8.1. Signal characteristics of ICH

Sequence Hyperacute (< 12 h) Acute (12 h7 d) Subacute (7 d3 m) Chronic (> 3 m)

T1WI Spin echo Isointense Isointense to Hyperintense Cystic defect

T2WI Spin echo Hyperintense Hypointense Hypointense Hypointense scar
FLAIR Hyperintense rim, Signal loss Signal loss Hypointense scar
Hypointense core
DWI Hyperintense rim, Hypointense/ Signal loss Hypointense scar
Hypointense core Signal loss
T2*WI (PWI) Signal loss Signal loss Signal loss Hypointense scar

T2WI causing a hyperintense signal. This may MRI in hyperacute ICH

allow approximate classication of the stage of the
ICH. In late subacute and chronic ICH, a hypoin- Hyperacute stroke demands the dierentiation
tense rim caused by the paramagnetic eect of between ischemic stroke and intracranial hemor-
hemosiderin demarcates the border zone of the rhage from the radiologist, which is impossible by
hematoma, which is better seen on T2WI than on clinical means.6,23 It also demands a generalized
T1WI. Hyperacute ICH presents as a non-specic diagnostic workup, where the important pathophy-
hyperintensity on T2WI and cannot be seen on siologic aspects of hyperacute ischemic stroke can
T1WI unless there is a substantial mass eect. An be investigated: (i) where and how large is the
excellent overview about this can be found in actual area of irreversible ischemic brain damage?
Osborn, Diagnostic Neuroradiology.18 New (ii) how old is the infarction? (iii) is there tissue at
sequences, which are more susceptible to para- risk of infarction that is potentially salvageable and
magnetic eects of deoxyhemoglobin, however, how much tissue is at risk? (iv) is there a vessel
may detect hyperacute ICH (see below). Due to the occlusion and where is it? (5) is an ICH or another
complexity of ICH appearance in dierent underlying, nonischemic disease present? The
sequences at dierent stages (see Table 8.1) as advent of new MRI techniques such as perfusion-
opposed to CT, there is a considerable amount of (PWI) and diusion- (DWI) weighted imaging has
discomfort among clinicians and radiologists with revolutionized diagnostic imaging in ischemic
regard to the role of MRI especially in the evalua- stroke.24,25 DWI may delineate infarcted brain tissue
tion of hyperacute stroke. Besides dierentiation in less than 1 h after symptom onset, probably
between ischemia and hemorrhage, MRI may within minutes,26 although there is cumulating evi-
provide information with regard to the etiology of dence that in the very early stage of stroke there
ICH. Basal ganglia hemorrhage in combination may be reversible DWI changes.27,28 while PWI
with older microbleeds or vascular leukencepha- denes the area of cerebral hypoperfusion. Several
lopathy implies a hypertensive etiology of ICH as studies have reported early ndings of stroke MRI
do lobar hematomas in the elderly.19 Multiple cor- within the rst 6 to 12 hours, demonstrating the
tical ICH in the elderly may suggest an amyloid feasibility and practicality of this method in the
angiopathy.20 In young patients, additional setting of acute stroke and thrombolytic
information from MRA may detect aneurysms, therapy.25,2935 In essence, the presence of a vessel
arteriovenous malformations, vasculitis like occlusion by MRA is associated with a PWI/DWI
changes, such as multiple intracranial stenoses, or mismatch, the stroke MRI setting that denes the
cerebral venous thrombosis and thus may make ideal candidate for thrombolysis.33,36 As already
an invasive conventional angiography unneces- stated, the diagnosis of ICH is still the domain of
sary.21,22 CT rather MRI. The need to perform both, CT for
Stroke MRI in intracranial hemorrhage 105

exclusion of ICH and stroke MRI to guide therapeu- window.42 ICH was unambiguously identied on
tic eorts, is time-consuming (time is brain), med- the basis of all stroke MRI sequences. Volumetric
icoeconomically questionable, and also analysis showed a good raw correlation of hemat-
inefcient.37,38 Therefore, characteristic MRI fea- oma volumes in all MRI images compared to CT. As
tures of hyperacute ICH especially on newer already reported by Rosen et al., images obtained
sequences should be described and utilized. with sequences with a high sensitivity for suscepti-
Many authors claim that the sensitivity of MRI for bility eects (T2*-WI, FLAIR) generally overestimate
detecting hyperacute ICH is poor.7,10 The appear- the actual hematoma size in comparison to the
ance of intracranial hemorrhage on magnetic reso- lesion volume assessed on CT.44 Hematoma
nance (MR) imaging depends primarily on the age volumes on DWI (median and mean dierence
of the hematoma and the type of MR contrast (i.e. 3.97%/4.36%, SD 37.42%) followed by FLAIR
T1- or T2-weighted). Immediately after ICH there (median and mean dierence 2.91%/6.25%, SD
are intact red blood cells with intracellular oxyhe- 28.39%) corresponded best with lesion size on
moglobin within a protein and uid-rich serum.12 CT. Conventional T2WI substantially underesti-
Oxyhemoglobin is diamagnetic and does not cause mated (median and mean dierence 17.24% /
any susceptibility eects.18 The key substrate for 12.98%, SD 34.46%) and T2*-WI substantially
early MRI visualization of hemorrhage is deoxyhe- overestimated (median and mean dierence
moglobin, a blood degradation product with para- 17.94% / 18.86%, SD 24.45%) the hemat-
magnetic properties due to unpaired electrons. As oma size. The typical appearance of ICH on the
long as the deoxyhemoglobin is intracellular within MRI images was a heterogeneous focus of high and
the erythrocytes during the very rst hour(s) after low signal intensities. With increasing susceptibility
ICH onset, it does not cause rapid dephasing of weight, the central area of hypointensity became
proton spins in T2WI or in T2*WI thus not allow- more pronounced. The T2*-WI showed no or few
ing for a signal loss within the rst 12 hours.12,39 areas of hyperintensity, or merely a faint ring
However, the paramagnetic deoxyhemoglobin around a central core of signal loss. Figures 8.1 and
changes the local magnetic eld gradient and this 8.2 illustrate the appearance of a larger and a
subsequently may lead to a dephasing of station- smaller hyperacute ICH on CT and MRI in two rep-
ary or passing protons inside and in the vicinity of resentative patients; these ndings were common
the ICH.11 The short T2 observed when deoxyhe- to all nine patients.
moglobin forms is enhanced on higher-eld- Linfante et al. successfully investigated 5 hyper-
strength systems and on gradient-echo images acute ICH-patients with MRI between 23 minutes
and is reduced with fast spin-echo MR tech- and 2 hours from symptom onset.43 They report a
niques.10 characteristic signal- and evolution pattern of
There are only few publications which report a hyperacute ICH, dividing the lesion into three
high diagnostic accuracy of MRI for ICH within a parts. In the very rst minutes after symptom onset,
time window of 6 hours or less after symptom the centre of the lesion appears heterogeneous on
onset.4043 Patel et al. were the rst to demonstrate all images (T2WI, T2*WI, T1WI), which is due to a
early MRI changes in susceptibility-weighted local predominance of oxyhemoglobin. The periph-
T2*WI within 5 hours in six patients with acute ery of the hematoma is hypointense on susceptibility-
ICH.41 These authors, however, did not use a stan- weighted images more than on T2WI and shows a
dardized MRI protocol and did not perform a volu- progressive enlargement over the next minutes and
metric comparative analysis of CT and MRI. rst hours. This is due to a progressive centripetal
Our group compared CT and stroke MRI ndings increase in concentration of deoxyhemoglobin.
in nine patients with acute ICH in a randomized There is a surrounding rim, which appears hyperin-
stroke imaging protocol (CT, DWI, PWI with T2*W tense on T2WI and T2*WI but hypointense on
source images, T2WI, FLAIR images, MRA); all T1WI and represents perifocal vasogenic edema.
patients received CT and MRI within a 3 to 6 h time While both authors42,43 suggest that stroke MRI is
(a) (b) (c) (d )

Fig. 8.1. 65-year-old woman with hypertension and obesity presenting 4 hours after symptom onset with somnolence, left-sided sensorimotor hemiparesis, dysarthro-
phonia, and hemineglect to the left. CT 11 2 h and MRI 33 4 h after symptom onset: (a) initial CT shows an ICH located in the right external capsule, (b) T2-WI shows the
same hyperintense lesion. (c) FLAIR and (d ) DWI revealed an irregular hyperintense right putaminal lesion with a hypointense core more prominent on the strongly
susceptibility-weighted and strongest on T2*-weighted images.

(e) (f ) (g) (h)

Fig. 8.1 cont. 40-year-old man with untreated arterial hypertension suering from severe headache, dense right-sided paresis, severe non-uent aphasia, and forced
gaze deviation to the left. CT 2 h and MRI 3 h after symptom onset: (e) initial CT shows a left putaminal ICH and ( fh) T2WI an irregular hyperintense putaminal lesion
on the left DWI, FLAIR, and T2*-weighted PWI source images show an increasing area of signal loss within the core of the left-sided hematoma with a heterogeneous
appearance caused by small amounts of deoxyhemoglobin.
Stroke MRI in intracranial hemorrhage 107


Fig. 8.2. 65-year-old patient with left-sided hemiplegia. Baseline CT shows an external capsule ICH 4 h after symptom onset (left).
Perfusion-weighted images (PWI, MTT-maps) show a large perfusion decit (hyperintensity) of the complete right hemisphere
(middle and right).


41-year-old patient with right-sided hemiplegia and aphasia. Baseline CT shows a large basal ganglia ICH 3 h after symptom onset
(left). Perfusion-weighted images (PWI, MTT-maps) show a large perfusion decit (hyperintensity) of the complete left hemi-
sphere (middle and right).


Same patient as middle row 24 h after symptom onset. Surgical removal of the hematoma as seen on the postsurgical CT (left).
Normalization of perfusion disturbance on the left hemisphere according to PWI (middle and right) with small artefacts in the
former ICH cavity due to susceptibility (T2*) eects.
108 Peter D. Schellinger, Olav Jansen and Werner Hacke

diagnostic in the evaluation of hyperacute ICH, SAH.5054 Conventional T1WI and T2WI were shown
they conclude consistently that the randomized to be ineective in the diagnosis of acute SAH with
and prospective acquisition of a larger group with sensitivities of 3550%.55 This may, in part, be due to
blinded evaluation is needed to conrm these pre- the fact that the rules, which apply for intracerebral
liminary observations. degradation of ICH may not be valid for SAH due to
These results support the hypothesis that small a higher oxygen partial pressure in cerebrospinal
amounts of deoxyhemoglobin are present within the uid (CSF)17 and that the formation of blood clots
very rst minutes after onset of ICH and detectable which also causes T2-shortening is impaired.9 The
by susceptibility-weighted MRI sequences. Figure paramagnetic susceptibility eect, which may be
8.1 suggests that T2*WI are suited best for the diag- utilized to detect intracerebral hemorrhage is not
nosis of ICH, which holds true for the qualitative suitable for the detection of SAH as the latter fre-
detection also of relatively small hematomas without quently is localized in the vicinity of the cranial vault
mass eect. Further information such as the pres- or skull base, thus areas which are prone to artifacts.
ence of space occupying edema, midline shift, and A rise in CSF-protein concentration leads to a T1-
ventricular hemorrhage is best derived from conven- shortening and therefore signal intensity increase,
tional T2WI. In addition to the standardized stroke which is seen better on PDWI than T1WI,50,56 an
protocol, postcontrast T1WI scans may be obtained observation contradicted by other authors.57,58 The
after PWI if another primary disease is suspected detection of SAH may be improved by using PDWI
(e.g. apoplectic glioma or metastases). Multicentre with shorter repetition times (TR) to visualize the
trials to acquire larger patient numbers are currently T1-eect.59 FLAIR sequences have been reported to
under way. Patients should be randomized for the be suitable for the diagnosis of not only subacute
sequence of CT and MRI to avoid a negative selec- but also acute SAH.48,49,52,54,60 The latter studies have
tion bias for CT, if both modalities are to be com- all been performed with low eld scanners (0.5
pared with regard to sensitivity and specity. tesla), however, with increasing eld strengths and
faster FLAIR sequences (usually 10 to 15 min,
TURBOFLAIR 3 to 6 min) there also are more pul-
MRI in subarachnoid hemorrhage sation artefacts, which may impair their sensitivity
for SAH. Wiesmann et al. recently reported a very
CT is also the (imaging) standard of care for the diag- high sensitivity (100%) for the combined use of
nosis of acute subarachnoid hemorrhage45,46 with a PDWI and FLAIR images in 19 patients.59,60
sensitivity of 85100%. The CT-based diagnosis of Artefacts, which were seen in one-third of their
SAH may be dicult in the posterior fossa due to patients could be determined as such by the com-
bone artefacts as well as with small amounts of bined use of PDWI and FLAIR images, all PD and
blood. In the subacute stage the sensitivity of CT for FLAIR studies were negative in ten control patients.
SAH diminishes substantially with sensitivities of Also many newer MRI sequences such as MRA, DWI
50% after 1 week, 30% after 2 weeks and 0% at 3 and PWI be useful to detect early complications of
weeks45 necessitating the diagnostic gold standard in SAH such as vasospasm and spreading depression.61
SAH, i.e. lumbar puncture. There currently is only Further studies aimed at the diagnostic reach of
scarce information about the sensitivity of stroke stroke MRI in acute SAH are worthwhile.
MRI for SAH. Initial assumptions that MRI may also
be suitable in the diagnosis of acute SAH were
derived from in vitro studies.47,48 In subacute SAH (5 Future prospects of stroke MRI in ICH
days to 2 weeks) MRI with uid attenuated inversion
recovery (FLAIR) sequences and proton density (PD) Even in the new millenium there is still controversy
weighted images is clearly superior to CT (sensitivity with regard to which patients should receive best
100% vs. 45%).49 Some investigators reported that medical treatment and whom should surgery per-
MRI may also be eective in the diagnosis of acute formed on.62,63 In general, patients with mild
Stroke MRI in intracranial hemorrhage 109

symptoms and small ICH, severe symptoms and as possible, especially with the time constraints
large ICH as well as deep grey matter ICH are not imposed by acute therapies for ischemic stroke such
operated, because surgery does not aect the as thrombolytic therapy. MRI stroke protocols
natural disease course and may even lead to com- including DWI, PWI and susceptibility weighted
plications and deterioration in patients with mild images are very promising with regard to the charac-
symptoms or basal ganglia hemorrhage.64 terization of acute stroke patients and the identica-
Conversely, young patients with a low comorbidity, tion of patients suitable for specic therapy.
moderately sized right hemispheric ICH, moderate Surprisingly, though, MRI is still not generally con-
clinical symptoms at presentation with a rapid sidered to be the primary and only diagnostic tool in
deterioration are likely to prot from hematoma hyperacute stroke patients, as there is doubt regard-
evacuation, especially when microsurgical tech- ing the ability to detect hyperacute ICH. While a
niques are applied.65 There are inconsistent data larger number of patients would be useful to conrm
with regard to the pathological processes in the these ndings, preliminary results suggest that stroke
vicinity of the ICH and their contribution to acute MRI with T2*WI is as sensitive as CT in the diagnosis
and chronic clinical decit. Some authors postulate of ICH and that there are characteristic features of
the presence of perihemorrhagic ischemic lesions hyperacute intracerebral hemorrhage on MRI. Also,
(animal experimental data) whether caused by the combination of PDWI and FLAIR images may
toxic eects, apoptosis or reduction of cerebral reliably dierentiate acute and subacute SAH from
blood ow,66,67 others deny this.68 Also in clinical ICH and ischemic strokes. In addition to that, SAH
studies evidence has been provided for69,70 and usually presents with a dierent clinical picture than
against71 the presence of ischemic perihemorrhagic do ICH and ischemic stroke, unless there is a sub-
areas. Results of the studies by Videen et al., stantial intracerebral hemorrhage accompanying
however, suggest that oligemia may be reactive to primary SAH, which then is apparent on T2*WI. The
ICH and represent a reduced metabolic oxygen rate initial and exclusive use of stroke MRI therefore is
with a constant oxygen extraction fraction rather feasible, cost-eective and time saving. In conclu-
than being pathologic oligemia or ischemia.70 A sion, stroke MRI may be the diagnostic tool of choice
small pilot study utilized a stroke MRI protocol with not only for patients with subacute and chronic ICH
T2WI, FLAIR, DWI, PWI and MRA to evaluate and SAH but also in the initial assessment of patients
ICH.42 There were perihemorrhagic perfusion de- with hyperacute ischemic or hemorrhagic stroke as
cits according to PWI in all nine patients, and large well as hyperacute SAH.
decits in three of nine patients, while there were
no manifest ischemic infarcts according to DWI
(Fig. 8.2).72 However, whether this was due to raised
intracranial pressure, local compression of the REFERENCES
middle cerebral artery, diaschisis, toxic eects of
blood products or other mechanisms cannot be 1 WHO Task Force. Stroke 1989. Recommendations on
answered at this point. Further animal studies and stroke prevention, diagnosis, and therapy. Report of
a prospective clinical study, which may shed more the WHO Task Force on Stroke and other
Cerebrovascular Disorders. Stroke 1989; 20:
light on the pathophysiological, clinical and prog-
nostic relevance of a perihemorrhagic zone of cere-
2 Jorgensen HS, Nakayama H, Raaschou HO, Olsen TS.
bral impairment are currently under way.
Intracerebral hemorrhage versus infarction: stroke
severity, risk factors, and prognosis. Ann Neurol 1995;
38: 4550.
Conclusion 3 Rosenow F, Hojer C, Meyer-Lohmann C et al.
Spontaneous intracerebral hemorrhage. Prognostic
In hyperacute stroke, it is of utmost importance that factors in 896 cases. Acta Neurol Scand 1997; 96:
diagnostic eorts are as specic and time ecacious 174182.
110 Peter D. Schellinger, Olav Jansen and Werner Hacke

4 Higer HP, Pedrosa P, Schaeben W, Bielke G, Meindl S. 18 Osborn AG. Intracranial hemorrhage. In: Osborn AG,
Intracranial hemorrhage in MRT. Radiology 1989; 29: ed Diagnostic Neuroradiology. Mosby: Year Book Inc;
297302. 1994: 154198.
5 Jansen O, Heiland S, Schellinger P. Neuroradiological 19 Fazekas F, Kleinert R, Roob G, Kleinert G et al.
diagnosis in acute ischemic stroke. Value of modern Histopathologic analysis of foci of signal loss on gradi-
techniques. Nervenarzt 1998; 69: 465471. ent-echo T2*-weighted MR images in patients with
6 Hacke W, Stingele R, Steiner T, Schuchardt V, Schwab spontaneous intracerebral hemorrhage: evidence of
S. Critical care of acute ischemic stroke. Intens Care microangiopathy-related microbleeds. Am J
Med 1995; 21: 856862. Neuroradiol 1999; 20: 637642.
7 Hayman LA, Taber KH, Ford JJ, Bryan RN. 20 Hagen T. Intracerebral hemorrhage in the context
Mechanisms of MR signal alteration by acute intracer- of amyloid angiopathy. Radiology 1999; 39:
ebral blood: old concepts and new theories. Am J 847854.
Neuroradiol. 1991; 12: 899907. 21 Ruiz-Sandoval JL, Cantu C, Barinagarrementeria F.
8 Hayman LA, Pagani JJ, Kirkpatrick JB, Hinck VC. Intracerebral hemorrhage in young people: analysis of
Pathophysiology of acute intracerebral and subarach- risk factors, location, causes, and prognosis. Stroke
noid hemorrhage: applications to MR imaging. Am J 1999; 30: 537541.
Radiol 1989; 153: 135139. 22 Jansen O, Knauth M, Sartor K. Advances in clinical
9 Weingarten K, Zimmerman RD, Cahill PT, Deck MD. neuroradiology. Akt Neurologie 1999; 26: 17.
Detection of acute intracerebral hemorrhage on MR 23 Hacke W, Steiner T, Schwab S. Critical management of
imaging: ineectiveness of prolonged interecho inter- the acute stroke: medical and surgical therapy. In:
val pulse sequences. Am J Neuroradiol 1991; 12: Batjer HH, Caplan LR, Freiberg L, Greenlee RG, Jr.,
475479. Kopitnik TH, Jr. and Young WL, eds Cerebrovascular
10 Bradley WG, Jr. MR appearance of hemorrhage in the Disease. Hagerstown: Lippincott-Raven; 1996:
brain. Radiology 1993; 189: 1526. 523533.
11 Felber S, Auer A, Wolf C et al. MRI characteristics of 24 Warach S, Boska M, Welch KM. Pitfalls and potential
spontaneous intracerebral hemorrhage. Radiology of clinical diusion-weighted MR imaging in acute
1999; 39: 838846. stroke. Stroke 1997; 28: 481482.
12 Steinbrich W, Gross-Fengels W, Krestin GP, Heindel W, 25 Barber PA, Darby DG, Desmond PM et al. Prediction
Schreier G. Intracranial hemorrhages in the magnetic of stroke outcome with echoplanar perfusion- and
resonance tomogram. Studies on sensitivity, on the diusion-weighted MRI. Neurology 1998; 51:
development of hematomas and on the determina- 418426.
tion of the cause of the hemorrhage. Rofo Fortschr Geb 26 Conturo TE, McKinstry RC, Aronovitz JA, Neil JJ.
Rontgenstr Neuen Bildgeb Verfahr 1990; 152: Diusion MRI: precision, accuracy and ow eects.
534543. NMR Biomed 1995; 8: 307332.
13 Gomori JM, Grossman RI, Yu-Ip C, Asakura T. NMR 27 Kidwell CS, Alger JR, Di Salle F et al. Diusion MRI in
relaxation times of blood: dependence on eld patients with transient ischemic attacks. Stroke 1999;
strength, oxidation state, and cell integrity. J Comput 30: 11741180.
Assist Tomogr 1987; 11: 684690. 28 Kidwell CS, Saver JL, Mattiello J et al. Thrombolytic
14 Gomori JM, Grossman RI, Goldberg HI, Zimmerman reversal of acute human cerebral ischemic injury
RA, Bilaniuk LT. Intracranial hematomas: imaging by shown by diusion/perfusion magnetic resonance
high-eld MR. Radiology 1985; 157: 8793. imaging. Ann Neurol 2000; 47: 462469.
15 Gomori JM, Grossman RI. Mechanisms responsible 29 Tong DC, Yenari MA, Albers GW, OBrien M, Marks MP,
for the MR appearance and evolution of intracranial Moseley ME. Correlation of perfusion- and diusion-
hemorrhage. Radiographics 1988; 8: 427440. weighted MRI with NIHSS score in acute (<6.5 hour)
16 Grossman RI, Gomori JM, Goldberg HI et al. MR ischemic stroke. Neurology 1998; 50: 864870.
imaging of hemorrhagic conditions of the head and 30 Jansen O, Schellinger PD, Fiebach JB, Hacke W, Sartor
neck. Radiographics 1988; 8: 441454. K. Early recanalization in acute ischemic stroke saves
17 Grossman RI, Kemp SS, Ip CY et al. Importance of oxy- tissue at risk dened by MRI. Lancet 1999; 353:
genation in the appearance of acute subarachnoid 20362037.
hemorrhage on high eld magnetic resonance 31 Baird AE, Warach S. Imaging developing brain infarc-
imaging. Acta Radiol Suppl. 1986; 369: 5658. tion. Curr Opin Neurol 1999; 12: 6571.
Stroke MRI in intracranial hemorrhage 111

32 Beaulieu C, de Crespigny A, Tong DC, Moseley ME, 45 van Gijn J, van Dongen KJ. The time course of aneu-
Albers GW, Marks MP. Longitudinal magnetic reso- rysmal haemorrhage on computed tomograms.
nance imaging study of perfusion and diusion in Neuroradiology 1982; 23: 153156.
stroke: evolution of lesion volume and correlation with 46 Scotti G, Ethier R, Melancon D, Terbrugge K, Tchang S.
clinical outcome. Ann Neurol 1999; 46: 568578. Computed tomography in the evaluation of intracra-
33 Schellinger PD, Jansen O, Fiebach JB et al. Monitoring nial aneurysms and subarachnoid hemorrhage.
intravenous recombinant tissue plasminogen activa- Radiology 1977; 123: 8590.
tor thrombolysis for acute ischemic stroke with diu- 47 Chakeres DW, Bryan RN. Acute subarachnoid hemor-
sion and perfusion MRI. Stroke 2000; 31: rhage: in vitro comparison of magnetic resonance and
13181328. computed tomography. Am J Neuroradiol 1986; 7:
34 Schellinger PD, Jansen O, Fiebach JB et al. Feasibility 223228.
and practicality of MR imaging of stroke in the man- 48 Melhem ER, Jara H, Eustace S. Fluid-attenuated inver-
agement of hyperacute cerebral ischemia. Am J sion recovery MR imaging: identication of protein
Neuroradiol 2000; 21: 11841189. concentration thresholds for CSF hyperintensity. Am J
35 Schellinger PD, Fiebach JB, Jansen O et al. Stroke mag- Roentgenol 1997; 169: 859862.
netic resonance imaging within 6 hours after onset of 49 Noguchi K, Ogawa T, Seto H et al. Subacute and
hyperacute cerebral ischemia. Ann Neurol 2001; 49: chronic subarachnoid hemorrhage: diagnosis with
460469. uid-attenuated inversion-recovery MR imaging.
36 Rordorf G, Koroshetz WJ, Copen WA et al. Regional Radiology 1997; 203: 257262.
ischemia and ischemic injury in patients with acute 50 Jenkins A, Hadley DM, Teasdale GM, Condon B,
middle cerebral artery stroke as dened by early Macpherson P, Patterson J. Magnetic resonance
diusion-weighted and perfusion-weighted MRI. imaging of acute subarachnoid hemorrhage. J
Stroke 1998; 29: 939943. Neurosurg 1988; 68: 731736.
37 Powers WJ, Zivin J. Magnetic resonance imaging in 51 Matsumura K, Matsuda M, Handa J, Todo G. Magnetic
acute stroke: not ready for prime time. Neurology resonance imaging with aneurysmal subarachnoid
1998; 50: 842843. hemorrhage: comparison with computed tomography
38 Powers WJ. Testing a test: a report card for DWI in scan. Surg Neurol. 1990; 34: 7178.
acute stroke. Neurology 2000; 54: 15491551. 52 Noguchi K, Ogawa T, Inugami A et al. Acute subarach-
39 Mattle HP, Edelman RR, Schroth G, OReilly GV. In: noid hemorrhage: MR imaging with uid-
Spontaneous and Traumatic Hemorrhage in Clinical attenuated inversion recovery pulse sequences.
and Magnetic Resonance Imaging, Vol. 1. 2nd ed. Radiology 1995; 196: 773777.
Philadelphia, PA: W B Saunders Co; 1996: 53 Chrysikopoulos H, Papanikolaou N, Pappas J et al.
652702. Acute subarachnoid haemorrhage: detection with mag-
40 Ebisu T, Tanaka C, Umeda M, et al. Hemorrhagic and netic resonance imaging. Br J Radiol 1996; 69: 601609.
nonhemorrhagic stroke: diagnosis with diusion- 54 Noguchi K, Seto H, Kamisaki Y, Tomizawa G,
weighted and T2-weighted echo-planar MR imaging. Toyoshima S, Watanabe N. Comparison of uid-
Radiology 1997; 203: 823828. attenuated inversion-recovery MR imaging with CT in
41 Patel MR, Edelman RR, Warach S. Detection of hyper- a simulated model of acute subarachnoid hemor-
acute primary intraparenchymal hemorrhage by mag- rhage. Am J Neuroradiol 2000; 21: 923927.
netic resonance imaging. Stroke 1996; 27: 55 Ogawa T, Inugami A, Shimosegawa E et al.
23212324. Subarachnoid hemorrhage: evaluation with MR
42 Schellinger PD, Jansen O, Fiebach JB, Hacke W, Sartor imaging. Radiology 1993; 186(2): 345351.
K. A standardized MRI stroke protocol: comparison 56 Satoh S, Kadoya S. Magnetic resonance imaging of
with CT in hyperacute intracerebral hemorrhage. subarachnoid hemorrhage. Neuroradiology 1988; 30:
Stroke 1999; 30: 765768. 361366.
43 Linfante I, Llinas RH, Caplan LR, Warach S. MRI fea- 57 Atlas SW. MR imaging is highly sensitive for acute sub-
tures of intracerebral hemorrhage within 2 hours from arachnoid hemorrhage . . . not! Radiology 1993;
symptom onset. Stroke 1999; 30: 22632267. 186(2): 319322; discussion 323.
44 Rosen BR, Belliveau JW, Chien D. Perfusion imaging 58 Atlas SW, Thulborn KR. MR Detection of hyperacute
by nuclear magnetic resonance. Magn Reson Q 1989; parenchymal hemorrhage of the brain. Am J
5: 263281. Neuroradiol 1998; 19: 14711477.
112 Peter D. Schellinger, Olav Jansen and Werner Hacke

59 Wiesmann M, Mayer TE, Medele R, Bruckmann H. 67 Deinsberger W, Vogel J, Fuchs C, Auer LM, Kuschinsky
Diagnosis of acute subarachnoid hemorrhage at W, Boker DK. Fibrinolysis and aspiration of experi-
1.5 Tesla using proton-density weighted FSE and MRI mental intracerebral hematoma reduces the volume
sequences. Radiology 1999; 39: 860865. of ischemic brain in rats. Neurol Res 1999; 21:
60 Noguchi K, Ogawa T, Inugami A, Toyoshima H, 517523.
Okudera T, Uemura K. MR of acute subarachnoid 68 Mun Bryce S, Kroh FO, White J, Rosenberg GA. Brain
hemorrhage: a preliminary report of uid-attenuated lactate and pH dissociation in edema: 1H- and 31P-
inversion-recovery pulse sequences. Am J Neuroradiol NMR in collagenase-induced hemorrhage in rats. Am
1994; 15: 19401943. J Physiol 1993; 265: R697R702.
61 Busch E, Beaulieu C, de Crespigny A, Moseley ME. 69 Dethy S, Goldman S, Blecic S, Luxen A, Levivier M,
Diusion MR imaging during acute subarachnoid Hildebrand J. Carbon-11-methionine and uorine-
hemorrhage in rats. Stroke 1998; 29: 21552161. 18FDG PET study in brain hematoma. J Nucl Med
62 Kanno T, Nagata J, Nonomura K et al. New approaches 1994; 35: 11621166.
in the treatment of hypertensive intracerebral hemor- 70 Videen TO, Dunford-Shore JE, Diringer MN, Powers
rhage. Stroke 1993; 24: 196100. WJ. Correction for partial volume eects in regional
63 Morgenstern LB, Frankowski RF, Shedden P, Pasteur W, blood ow measurements adjacent to hematomas in
Grotta JC. Surgical treatment for intracerebral hemor- humans with intracerebral hemorrhage: implementa-
rhage (STICH). Stroke 1998; 51: 13591363. tion and validation. J Comput Assist Tomogr1999; 23:
64 Fernandes HM, Gregson B, Siddique S, Mendelow AD. 248256.
Surgery in intracerebral hemorrhage : the uncertainty 71 Hirano T, Read SJ, Abbott DF et al. No evidence of
continues. Stroke 2000; 31: 25112516. hypoxic tissue on 18F-uoromisonidazole PET after
65 Auer LM, Deinsberger W, Niederkorn K et al. intracerebral hemorrhage. Neurology 1999; 53:
Endoscopic surgery versus medical treatment for 21792182.
spontaneous intracerebral hematoma: a randomized 72 Schellinger PD, Fiebach JB, Jansen O, Hacke W.
study. J Neurosurg 1989; 70: 530535. Penumbra Nachweis bei Hirnblutungen mit multi-
66 Bullock R, Brock Utne J, van Dellen J, Blake G. modaler MRT. Akt Neurol 1998; 25: 150.
Intracerebral hemorrhage in a primate model: eect
on regional cerebral blood ow. Surg Neurol 1988; 29:
Using diffusionperfusion MRI in animal models for drug

Marc Fisher
Department of Neurology, University of Massachusetts, Medical School, Worcester, MA, USA

Introduction engendered a careful reassessment of the devel-

opment plan, both preclinical and clinical, used
The development of therapies for acute ischemic to bring the drugs in question to advanced evalua-
stroke is a dicult venture with many potential tion.
pitfalls. Currently, the only approved therapy for Preclinical evaluation of purported acute stroke
acute ischemic stroke is intravenous tissue plas- therapies plays a vital role in the drug development
minogen activator (t-PA) given within 3 hours of process. A cogent argument would be that a drug
stroke onset. The approval of rt-PA stemmed from not demonstrating ecacy in animal stroke models
the National Institutes of Neurological Disorders is very unlikely to show ecacy in humans.
and Stroke (NINDS) trials that demonstrated a Conversely, treatment eects in animal stroke
signicant treatment eect with rt-PA on several models certainly do not guarantee success in clini-
dierent types of 90-day outcome measures and cal trials. For both thrombolytic and neuroprotec-
global assessment of these measures.1 Two other tive drugs, the gold standard for detecting
acute stroke therapy trials yielded statistically sig- treatment ecacy in animal models remains eects
nicant positive eects on the prespecied on ischemic lesion volume evaluated postmortem
primary outcome measures. In the Stroke Therapy by histological methods.5 The goal of acute stroke
Ancrod Trial (STAT), ancrod, a defrinogenating therapy is to salvage ischemic brain tissue and
agent, initiated within 3 hours of stroke onset prevent progression to infarction. The salvage of
improved outcome, but not as favorably as rt-PA ischemic tissue is then presumed to be associated
given within the same time window.2 Another with improved neurological and functional
ancrod trial with a 6-hour window to the initiation outcome. Animal stroke models and their use in
of treatment was stopped and the drug has not evaluating treatment eects of potential acute
been approved by regulatory agencies. The other stroke therapies are prone to many diculties.6
positive acute stroke trial was the PROACT 2 trial Experiments must be performed in a blinded
that evaluated the thrombolytic agent, randomized manner with careful attention to the
proUrokinase, within 6 hours of stroke onset in a measurement and maintenance of physiological
group of relatively severe stroke patients with variables. A doseresponse relationship between
angiographically conrmed middle cerebral drug treatment eects and lesion volume should be
artery occlusion.3 The trial only included 180 dened, as well as the relationship of dose to side
patients and did not lead to regulatory approval. eects. The survival time of animals after dosing
There have been a large number of other acute should be extended for days or even weeks after
stroke therapy trials, both phase 2 and phase 3 initial success with shorter survival to ensure that
trials that have not achieved a favourable the treatment eect is not transient.6 Lastly, drugs
outcome.4 These non-successful trials have are typically initially evaluated in rodent stroke

114 Marc Fisher

models, but consideration of further testing in Diffusionperfusion MRI in animal stroke

gyrencephalic species closer to humans should be models
entertained for novel, rst in class drugs. In addi-
tion to determining eects on ischemic lesion Diusion MRI (DWI) is an imaging technique that
volume, it has been recommended that eects on allows for the rapid and easy identication of
behavioural and functional outcome be evaluated regions of focal brain ischemia in animal stroke
to more closely reproduce the likely outcomes to be models (Fig. 9.1, see colour plate section). With
used in clinical trials.7 The important role of animal DWI, ischemic lesions can be detected within
stroke models for screening of potentially eective minutes of stroke onset and the evolution of the
drugs for acute ischemic stroke is clear and the ischemic lesion can be followed both temporally
main treatment eect to be assessed in animals will and spatially.12 With serial DWI imaging, the pattern
likely remain reduction of acute and/or subacute of lesion growth can be assessed, e.g. the spread of
infarct volume. the ischemic lesion from the basal ganglia to the
Proceeding from animal evaluation of novel cortex. Information about the severity of the
acute stroke therapies to clinical development is a ischemic lesion can also be derived because abso-
major step that requires a substantial investment of lute values of the apparent diusion coecient
time, money and other resources by the drugs (ADC) can be determined. The precise derivation of
developer, e.g. a pharmaceutical or biotechnology low ADC values in focal ischemic brain injury
company. Clinical trials require a large sample size remains uncertain, but the most likely explanation
to be adequately powered to detect treatment is the development of cytotoxic edema secondary
eects on the traditional neurological, functional to high-energy metabolism failure.13 This energy
and handicap measures used to determine treat- failure leads to loss of ion homeostasis and an
ment eects.8 The sample size employed in most inux of extracellular water into the cell, i.e. cyto-
phase 2 trials is insucient to detect even substan- toxic edema. It appears that lower ADC values are
tial trends of treatment ecacy and decisions associated with a greater degree of ischemic injury
about proceeding to phase 3 trials are largely based and less chance for reversibility with an appropriate
on safety data and intuition. Recently, it has been and timely intervention. Rats have been the species
proposed that diusionperfusion MRI might help most widely studied in DWI animal stroke studies
to bridge the gap between preclinical evaluation and these experiments have provided much valu-
and advanced clinical trials.9 For a neuroprotective able information about stroke pathophysiology and
drug, an eect on ischemic lesion growth on a temporal characteristics.
baseline diusion MRI to a day 3090 T2 scan might DWI animal stroke studies are providing informa-
conrm an eect on ischemic lesion evolution in tion about the development of secondary lesions
vivo and potentially be predictive of salutary eects after successful reperfusion, so-called reperfusion
on clinical outcome in a phase 3 trial.10 For a injury. With early reperfusion, either with a throm-
thrombolytic drug, demonstrating reperfusion bolytic drug or by mechanical reperfusion, several
eects on perfusion imaging in a phase 2 trial laboratories have demonstrated that initially
would reassure participants and sponsors that the detected ADC abnormalities can be reversed.14,15
agent restores blood ow and the appropriate dose However, a few hours later secondary declines in
to be used in the phase 3 trial.11 Diusion ADC values are observed, occurring initially in
perfusion MRI can be used in animal models to regions where the initial ADC declines were most
evaluate the same parameters as those to be severe (Fig. 9.2, see colour plate section). By 12
assessed in phase 2 and 3 clinical trials, providing a hours after reperfusion, the secondary DWI lesion
link between preclinical and clinical drug develop- volumes are approximately 50% of the volume
ment. The two MRI techniques have been used observed prior to reperfusion.16 This lesion volume
extensively in animal stroke models and more continues to increase, becoming almost the same
recently in clinical stroke trials. as the volume observed during occlusion by 24
Using diffusionperfusion MRI in animal models for drug development 115

hours after reperfusion and this late secondary reduced perfusion in the ischemic region, the
ischemic lesion is highly correlated with the histo- severity of the perfusion abnormality and the
logically conrmed infarct volume. The secondary return of adequate perfusion with mechanical or
ADC decline observed after reperfusion might in thrombolytic intervention in an animal model or
part represent reperfusion injury, but the situation stroke patient.
may be more complex. A recent histology study Combining information from PWI with DWI
demonstrated that ischemic regions where ADC provides a powerful tool for the study of acute
values had returned to normal still have neuronal ischemic stroke in animal models. PWI changes
and glial cell abnormalities that tend to worsen can be related to ADC declines to study the rela-
over time.17 This observation implies that a normal tionship between the severity of perfusion
DWI study after reperfusion does not necessarily changes to the development and severity of tissue
indicate tissue normalization, although late secon- injury. Using the two MRI parameters together
dary deterioration appears to occur even without can provide insights about the existence or lack of
complete reperfusion related ADC recovery. The an ischemic penumbra. An estimate of the
mechanisms for the secondary ADC declines and ischemic penumbra would be hypoperfused
tissue injury are not clearly established, but likely tissue on PWI that is not yet abnormal on DWI.21
contributors are secondary energy failure, mito- This region is the so-called diusionperfusion
chondrial respiratory function deterioration and mismatch and the presence of a DWIPWI mis-
apoptosis. Secondary ADC declines is not limited to match is being used currently in clinical stroke
animal studies. The UCLA group has seen such sec- trials to enrich patient selection for enrolment.
ondary ADC declines in approximately half of their The hypothesis being that mismatch patients are
patients after successful intra-arterial thromboly- those likely to still have an ischemic penumbra
sis.18 The clinical signicance of these secondary and a better chance to respond to therapeutic
ADC declines remains to be rmly established, but intervention. In a preliminary animal study of the
a worse clinical outcome is likely when this occurs. DWIPWI mismatch, it was observed that with the
Perfusion MRI (PWI) is the other MRI modality permanent suture occlusion model the mismatch
that has been widely applied in animal stroke was no longer signicantly dierent by 90
models. The two techniques used to acquire PWI minutes after occlusion and that it had essentially
are the bolus-tracking method and the arterial disappeared by 120 minutes.22 There was a signi-
spin-labelling technique.19 Bolus-tracking PWI is cant mismatch at 30 and 60 minutes after occlu-
much more commonly used and requires the sion. In another group of animals who were
injection of a contrast agent that has susceptibil- mechanically reperfused after 60 minutes of tem-
ity eects, impairing the acquisition of T2*-signal. porary occlusion, the ischemic lesion volume on
Multiple, repetitive images are obtained over a DWI did not increase after reperfusion. This
30-second time interval to generate a signal observation supports the concept that the hypo-
intensity curve. This signal intensity curve can perfused region with normal ADC values can be
then be used to obtain a number of perfusion- salvaged with timely reperfusion. The DWIPWI
related parameters including the mean transit mismatch is likely only a rough approximation of
time, cerebral blood volume and an estimate of the ischemic penumbra because some of the
the cerebral blood ow.20 With accurate determi- abnormal region on DWI can be reversed and
nation of the arterial input function and decon- infarction prevented. A more precise denition of
volution techniques, more precise measurements the ischemic penumbra on DWIPWI will likely
of cerebral blood ow can be obtained. In clinical require determination of absolute ADC, cerebral
practice or animal studies, perfusion maps based blood ow and perhaps other MRI parameters.
upon the mean transit time or time to peak of the These determinations will come from successful
signal intensity curve are commonly employed. treatment experiments initially in animal stroke
These PWI maps can be used to document models and later in stroke patients.
116 Marc Fisher

Preclinical evaluation of new stroke therapies species such as cats and/or primates. Determining
eectiveness in higher species with a time window
Animal stroke models are used to assess in vivo to initiation of treatment 12 hours or longer after
therapeutic responses of novel pharmaceuticals to stroke onset will likely provide more reassurance of
determine if the agents have activity in focal brain drug activity than relying solely on rodent data.
ischemia. A large variety of stroke models are used Experiments employing cats and especially pri-
to evaluate therapy in several dierent animal mates are dicult, time consuming and expensive
species. None of the animal stroke models precisely to perform. The data derived are very likely to be
reproduce the clinical conditions associated with worth the eort because they will provide impor-
ischemic stroke. Therefore, relating experiences tant clues as to how best to proceed with clinical
with these animal stroke models to human stroke development of the drug. However, the leap from
requires an extrapolation and extension of the animal studies that focus on infarct size and to
results. The typical paradigm employed for the some extent behavioural measures to clinical trials
development of neuroprotective stroke drugs is to that use neurological and/or functional handicap
develop a molecule targeted at one or more aspects measures remains problematic. Using diu-
of the cascade of ischemic injury. The drug is then sionperfusion MRI in animal stroke models and
screened for obvious toxicology problems such as clinical drug development may help to bridge the
blood pressure lowering eects, eects on gap.
bodybrain temperature, cancer inducing eects
and other major side eects. Non-toxic agents are
then applied to rodent models of focal brain ische- Using diffusionperfusion MRI in animals to
mia and the recommended initial experiments assess therapy
usually are performed in a rat permanent ischemia
model with a short period of poststroke survival The ability of diusionperfusion MRI to rapidly
(2448 hours). Currently, the most commonly used evaluate stroke in vivo provides a powerful tool to
rat stroke model is the suture occlusion model that assess the eects of neuroprotective and thrombo-
can be used for both permanent and temporary lytic therapies in animal stroke models. Much more
occlusion experiments.23 Dose ranging is then pre- experience with diusion MRI to assess eects of
formed around the initially eective dose to dene neuroprotectants is currently available. The rst
the minimally eective and maximally tolerated report using diusion MRI to evaluate the
dose. The endpoint of these investigations is infarct sodiumcalcium channel modulator, RS-87476,
size as determined by histological evaluation. appeared in 1991, demonstrating that an in vivo
Functional and behavioural endpoints should also treatment eect could be detected.24 Subsequently,
be evaluated, especially as the length of survival is reduction of ischemic lesion size in vivo on diu-
extended to determine if initial treatment eects sion MRI with the NMDA antagonist, Cerestat, was
are maintained over an extended time period. The observed in both temporary and permanent rat
time to initiation of treatment is another important focal ischemia models.25 The treatment eect
variable that should be assessed in the animal observed during the 3 hours of imaging and drug
models. Drugs that are only eective when given infusion was somewhat larger than the reduction of
before or shortly after stroke onset are not attractive infarct size observed histologically. This observa-
candidates for clinical development for practical tion was then used to model what the likely optimal
and nancial reasons. A short time window will infusion time of Cerestat should have been to
make clinical trial design dicult and a short time maintain maximal neuroprotective benets. Other
window for initiation of therapy will only have a neuroprotective drugs have also been evaluated
small market, if proven to be successful. Drugs that with diusion MRI with the detection of signicant
are shown to be eective with a reasonable time in vivo treatment eects that were conrmed by
window in rats should then be evaluated in higher delayed histological conrmation of infarct size
Using diffusionperfusion MRI in animal models for drug development 117

reductions (Fig. 9.3, see colour plate section).

Another important piece of information that can be
obtained with the use of diusion MRI to evaluate
neuroprotective therapeutic eects is the time
course of activity. Two groups evaluated the glycine
antagonist, ZD9379, and an interesting and unex-
pected delayed treatment eect was observed.
When therapy with ZD9379 was initiated 30
minutes after permanent middle cerebral artery
occlusion with the suture model, Takano et al. did
not observe any obvious dierence in lesion
volume between the treated and placebo groups
until 3.5 hours after occlusion when a trend
towards a smaller ischemic lesion volume was seen
in the ZD9379 group (Fig. 9.4).26 The diusion
imaging ceased at that time point, but at 24 hours Fig. 9.4. Ischemic lesion volumes are similar in the placebo
after stroke onset the treated group had a 43% and ZD9379 groups for the rst 3 hours after stroke onset. At
3.5 hours the volume in the treated group (closed circles)
smaller ischemic lesion volume by histological
begins to diverge from the volume in the placebo group
analysis. This experiment suggested a delayed treat-
(open squares)(reproduced with permission, Stroke 1997; 28:
ment eect with ZD9379 that would not have been
appreciated with analysis of delayed, histological
infarct size alone. Interestingly, the delayed tissue
salvage occurred primarily in the border zone quately reperfused. Combining PWI with DWI
between the middle and anterior cerebral arteries, a allows for assessment of the eects of reperfusion
region of modest blood ow decline thought to rep- on ischemic lesion size. Several groups have
resent a portion of the ischemic penumbra in the reported that mechanical reperfusion by with-
suture occlusion model. In a second reported treat- drawal of the suture occluder 3060 minutes after
ment experiment with ZD9379, Qiu et al. initiated stroke onset is associated with a marked reduction
treatment immediately after stroke onset and in the hypoperfused zone and shrinkage of the
imaged for 6 hours with diusion MRI.27 No signi- ischemic region on DWI.15,28 Withdrawal of the
cant dierence of ischemic lesion size was detected suture occluder 90 minutes or later after stroke
at 2.5 hours after stroke onset, but at 6 hours onset is typically associated with adequate reperfu-
ischemic lesion size was 41% smaller in the ZD9379 sion on PWI, but little or no reduction of the
group. This experiment with a longer imaging time ischemic lesion on DWI.
window supports the implications of the rst Several reports have appeared concerning the use
experiment that ZD9379 had a delayed treatment of PWI and DWI to assess thrombolytic therapy in
eect and was able to reverse initially abnormal animal stroke models. Jiang et al. evaluated rt-PA in
ADC regions. a rat embolic stroke model.14 Initiating treatment
PWI has been used to evaluate reperfusion eects with rt-PA 1 hour after embolization was associated
of both mechanical reperfusion and thrombolytic with signicant reduction in the hypoperfused
therapy. The suture occlusion model is amenable to region on PWI and an improvement in the extent of
reperfusion experiments by withdrawal of the the ischemic lesion determined by a combination of
suture and this can be accomplished in the MRI DWI and T2 ndings. Treatment eects of
unit to allow for rapid and continuous assessment. ProUrokinase were evaluated by another group in a
PWI can be used to conrm successful suture with- rat embolic model.29 Pro-urokinase was given either
drawal and to quantify the percentage of initially intravenously or intra-arterially beginning 30
hypoperfused tissue that is subsequently ade- minutes after embolization. Frequent PWI and DWI
118 Marc Fisher

Control group I.A. group I.V. group

% hemisphere hypoperfused

Minutes Minutes Minutes

Fig. 9.5. The percentage of the ischemic hemisphere that is hypoperfused declined in most animals treated with intra-arterial
(I.A.) or intravenous (I.V.) pro-urokinase (reproduced with permission, Neurology 1998; 50: 870875).

studies were performed to evaluate the eects of bination of PWI and DWI, the true ecacy of the
treatment on brain perfusion and ischemic lesion drug would not have been realized in the temporary
growth. In comparison to the vehicle group, both occlusion experiment. The power of using both PWI
pro-urokinase treatment approaches were observed and DWI in animal treatment is thus apparent.
to signicantly improve perfusion and to maintain
the ischemic lesion at the same size as before treat-
ment began (Fig. 9.5). In the vehicle group, a signi- Conclusions
cant increase of ischemic lesion size on DWI
occurred over time, and the postmortem ischemic The limited success so far in the development of
lesion size was signicantly larger in this group at 24 acute stroke therapies strongly suggests that new
hours than in the two pro-urokinase groups. approaches to the drug development process will
Combining PWI and DWI in temporary occlusion be necessary to enhance the likelihood of nding
models evaluating neuroprotective drugs may also eective therapies in the future. Diusion
be useful. A treatment experiment with an endothe- perfusion MRI in animal stroke models should be
lin antagonist in a 90-minute temporary occlusion valuable in this process for a number of reasons.
experiment demonstrated an interesting result. The These MRI techniques provide important informa-
24-hour infarct volume demonstrated by histology tion about in vivo drug activity that can help to
was signicantly smaller in the treated group as conrm the site of action, time window of treatment
compared to the placebo group.30 Interestingly, PWI eect and appropriate duration of therapy. The
disclosed that only 60% of the animals adequately experience in animals can then be directly trans-
reperfused after withdrawal of the suture occluder ferred to early clinical trials where the same end-
and it was these successfully reperfused animals that points can be evaluated, conrming or refuting the
demonstrated a neuroprotective eect. This result preclinical results in stroke patients. The use of
suggested that the endothelin antagonist would not diusionperfusion MRI both in preclinical testing
be eective in permanently occluded animals and, and in clinical development of acute stroke thera-
indeed, a subsequent experiment with a permanent pies will likely continue to expand, especially once
occlusion model did not demonstrate a signicant the MRI modalities provide support for the approval
reduction of ischemic lesion size. Without the com- of a new acute stroke therapy. Additionally, their
Using diffusionperfusion MRI in animal models for drug development 119

utility will also increase as we move into the multi- 12 Reith W, Hasegowa Y, Latour LL et al. Multislice
therapy era of acute stroke therapy when combina- diusion mapping for 3D evolution of cerebral ische-
tions of neuroprotective and thrombolytic therapy mia in a rat stroke model. Neurology 1995; 45:
are employed together to maximize outcome. 172177.
13 Latour LL, Svoboda K, Mitra PP, Sotak CH. Time-
dependent diusion of water in biological model
system. Proc Natl Acad Sci USA 1994; 91: 12291233.
14 Jeans Q, Zhang RL, Zhang ZB et al. Diusion, T2,
REFERENCES and perfusion-weighted magnetic resonance
imaging of middle cerebral artery embolic stroke
1 The National Institute of Neurological Disorders and and recombinant tissue plasminogen activator
Stroke rt-PA Stroke Study Group. Tissue plasminogen intervention in the rat. J Cereb Blood Flow Metab
activator for acute ischemic stroke. N Engl J Med 1995; 1998; 18: 758767.
333: 15811587. 15 Li F, Haw S, Tatlisumak T et al. Reversal of acute
2 Sherman DG, Atkinson RP, Chippendale T et al. for the apparent diusion coecient abnormalities and
STAT participants. Intravenous ancrod treatment for delayed neuronal death following transient focal cere-
acute ischemic stroke. J Am Med Assoc 2000; 283: bral ischemia in rats. Ann Neurol 1999; 46: 333342.
23952403. 16 Li F, Liu KF, Silva MD et al. Secondary decline in
3 Furlan AJ, Higashida R, Wechsler L et al. for the apparent diusion coecient and neurological
PROACT Investigators. Intra-arterial prourokinase for outcome after a short period of focal brain ischemia
acute ischemic stroke. The PROACT-II Study. J Am Med in rats. Ann Neurol 2000; 48: 236244.
Assoc 1999; 282: 20032011. 17 Li F, Liu KF, Silva MD et al. Acute normalization of the
4 Fisher M, Schaebitz W. An overview of acute stroke apparent diusion coecient of water is not asso-
therapy. Arch Neurol 2000; 160: 31964006. ciated with reversal of astrocytic swelling and neuro-
5 Li F, Irie K, Anwer U, Fisher M. Delayed triphenyltetra- nal shrinkage. Am J Neuroradiol 2002; 23: 180188.
zolium chloride staining remains useful for evaluating 18 Kidwell CS, Saver JL, Mattiello J et al. Thrombolytic
cerebral infarct volume in a rat stroke model. J Cereb reversal of acute human cerebral ischemic injury
Blood Flow Metab 1997; 12: 11321135. shown by diusion/ perfusion magnetic resonance
6 Stroke Therapy Academic Industry Roundtable. imaging. Ann Neurol 2000; 47: 462469.
Recommendations for standards regarding preclinical 19 Rosen BR, Belliveau JW, Vevea JM, Brady TJ. Perfusion
neuroprotective and restorative therapies. Stroke imaging with NMR contrast agents. Magn Reson Med
1999; 30: 27522758. 1990; 14: 249265.
7 Marshall JWB, Dun KJ, Green AR, Ridley RM. NXY- 20 Sorensen AG, Copen WA, Ostergaard L et al.
059, a free radical-trapping agent substantially lessens Hyperacute stroke: simultaneous measurement of rel-
the functional disability resulting from cerebral ische- ative cerebral blood volume, relative cerebral blood
mia in a primate species. Stroke 2001; 32: 190198. ow, and mean tissue transit time. Radiology 1999;
8 Davis SM. Endpoints and statistical concerns. In 212: 519527.
Fisher M, ed., Stroke Therapy, Butterworth- 21 Neuman-Hacin T, Wittsack H-J, Wenerski F et al.
Heinemann, Boston, 2000. Diusion and perfusion-weighted MRI: The DWI/PWI
9 Neumann-Haefelin T, Moseley ME, Albers GW et al. mismatch region in acute stroke. Stroke 1999; 30:
New magnetic resonance imaging methods for cereb- 15911597.
rovascular disease: Emerging clinical applications. 22 Omae T, Silva M, Mayzel O et al. Temporal evolution of
Ann Neurol 2000; 31: 559570. diusion-perfusion mismatch in a rat stroke model.
10 Warach S, Pettigrew LC, Dashe JF et al. Eect of citico- Stroke 2001; 32: 352.
line on ischemic lesions as measured by diusion- 23 Laing RJ, Jablonski J, Laing RW. Middle cerebral artery
weighted MRI. Ann Neurol 2000; 48: 713772. occlusion in the rat. Which method works best? Stroke
11 Schellinger FD, Jansen O, Fiebach JB et al. 1993; 294298.
Monitoring intravenous recombinant tissue plasmi- 24 Kucharczyk KJ, Mintorovitch J, Moseley ME et al.
nogen activator thrombolysis for acute ischemic Ischemic brain damage: reduction by
stroke with diusion and perfusion MRI. Stroke 2000; sodiumcalcium ion channel modulator RS-87476.
31: 13181328. Radiology 1991; 179: 221227.
120 Marc Fisher

25 Minematsu K, Fisher M, Li L et al. Eects of a novel 28 Muller TB, Haraldseth D, Jones RA et al. Combined
NMDA antagonist on experimental stroke rapidly and perfusion and diusion-weighted magnetic reso-
quantitatively assessed by diusion-weighted MRI. nance imaging in a rat model of reversible middle
Neurology 1993; 42: 397403. cerebral artery occlusion. Stroke 1995; 26: 451458.
26 Takano K, Tatlisumak T, Formato J et al. A glycine site 29 Takano K, Carano RAD, Tatlisumak T et al. The ecacy
antagonist attenuates infarct size in experimental of intraarterial and intravenous Prourokinase in an
focal ischemia: postmortem and diusion mapping embolic stroke model evaluated by diusion
studies. Stroke 1997; 28: 12551263. perfusion magnetic resonance imaging. Neurology
27 Qiu H, Hedlund CW, Gewalt SL, Benveniste H, Bare 1998; 50: 870875.
TM, Johnson GA. Progression of a focal ischemic 30 Tatlisumak T, Carano RAD, Takano K, Opgenorth TJ,
lesion in a rat brain during treatment with a Sotak CH, Fisher M. A novel endothelin antagonist,
novel glycine NMDA antagonist. JMRI 1997; 7: A-127722, attenuates ischemic lesion size in rats with
793744. temporary middle cerebral artery occlusion. Stroke
1998; 29: 850858.
Localization of stroke syndromes using diffusion-weighted MR
imaging (DWI)

Max Wintermark,1 Marc Reichhart,2 Reto Meuli1 and Julien Bogousslavsky2

Department of Diagnostic and Interventional Radiology and
Department of Neurology, University Hospital (CHUV), 1011 Lausanne, Switzerland

Introduction induces dephasing of the transverse magnetization

of the static and mobile molecules. Both the 180
DWI is an MR imaging technique in which micro- pulse and the second gradient pulse, which is the
scopic water motion is responsible for the contrast same size as the rst but opposite in sign, rephase
within the image.15 DWI has assumed the role of a the stationary spins, but not the mobile molecules,
valuable imaging technique because it provides thus resulting in substantial signal loss.25
information that is not available on standard T1- Clinical practice uses dierent representations of
and T2-weighted MR images. By showing hyper- the results of DWI data processing: diusion-
acute brain ischemia within minutes after stroke weighted images, DWI trace and ADC maps, which
onset, diusion-weighted imaging has gained are all equivalent.
importance in the assessment of stroke, whereas Diusion of bipolar gradient pulses used to
CT or T2-weighted MR images become positive only outline diusion properties of matter is typically
after several, usually 5 or 6 hours after stroke onset. applied in one direction at a time. Because of aniso-
In a rodent model, sensitivity of diusion-weighted tropic eects, data obtained in using a one-
imaging in the detection of acute infarction has direction diusion gradient only can give
amounted to 60% within 50 minutes and 100% misleading information. Application of a diusion
within 2 hours after symptomatology onset.612 gradient perpendicular (rather than parallel) to the
orientation of a white matter tract results in a lower
ADC measurement. If this particularity can be used
Clinical representations of DWI results in myelin bre tractography, it may also be confus-
ing and thus result in diagnostic pitfalls in the
Diusion of water molecules alters conventional detection and delineation of cerebral ischemia.
T1- and T2-weighted MR imaging, because it This is the reason why bipolar gradient pulses are
induces a signal dephasing and a signal loss. On the usually applied along at least three orthogonal sam-
other hand, through adequate MR sequences, this pling directions, and the diusion coecients thus
signal loss can be turned into a specic informa- sampled allow for calculation of a DWI-trace, which
tion, which constitutes the basis of DWI. represents the average of diusion-weighted
Enhancement of the DWI signal is aorded by images in all three directions:
adding a bipolar gradient, which consists of two

b D
sequential pulses superimposed to the classical S S0 e T2
90180 spin echo sequence. The rst gradient
pulse is applied between the 90 and the 180 in which the rst term in brackets represents the T2
pulses. Motion during and after this gradient pulse shine-through eect, D the diusion coecient

122 Max Wintermark, Marc Reichhart, Reto Meuli and Julien Bogousslavsky

and b a factor describing the strength of bipolar

gradient pulses. On a DWI trace, the anisotropy and
orientation of myelin bres are completely
removed. Such images show small contrast
between white and grey matter and, as discussed
later, the diusion coecient of white matter is
decreased in acute stroke to approximately the
same extent as that of grey matter.1315
The DWI trace is not solely the result of diusion
characteristics but also of the T2 decay: some inves-
tigators have called this phenomenon the T2 shine-
through eect.16 One can get rid of this T2
shine-through eect through computing isotropic
or directionally averaged diusion coecient
(ADC) (Fig. 10.3). Admitting a single water compart-
ment due to fast exchange across cell membranes,
and thus a monoexponential decay in voxels, the
ADC map can be calculated from at least two b
values, typically 500 and 1000 [s/mm2]. The ADC
map shows relative signal intensities only due to
dierences in tissue diusion.14,16
On ADC maps, tissues have an appearance oppo-
site to that seen on the DWI trace, reecting the
negative exponent in the equation relating signal
intensity to the ADC. Thus, CSF is hyperintense on
ADC maps, whereas on a DWI trace, CSF is hypoin-
tense. On diusion-weighted images, lesions with
restricted water diusion are hyperintense while
they are hypointense on ADC maps.14
Information carried by a DWI trace and ADC Fig. 10.1. 65-year-old male patient showing a right armleg
maps are complementary. The DWI trace aords hemisyndrome with a left face palsy, allowing for the suspi-
sharp delineation of pathologic processes, through cion of a left pontine ischemic lesion. The MR examination
both the removal of anisotropy of myelin bres and obtained 6 hours after onset relates the symptomatology to
an acute left paramedian pontine ischemic stroke (arrow),
the absence of contrast between grey and white
featuring hyperintensity on T2-weighted image and DWI-
matter while ADC maps allow one to get rid of the
trace, and hypointensity on the ADC map. Even retrospec-
T2 shine-through.
tively, this stroke could not be recognized on the admission
non-contrast cerebral CT obtained 4 hours after stroke onset.

DWI sensitivity to strokes according to their

location percentage increases to 94%. Interobserver agree-
ment is better in DWI than in conventional
DWI is more accurate than CT in localizing MRI.1719
ischemic lesions shortly after stroke onset. DWI can show small lesions adjacent to the cere-
Admission CT does localize lesions correctly in only brospinal uid. The diagnosis of a small cortical or
53% of patients. On the other hand, conventional brainstem infarct may be dicult on T2WI images
MRI correctly identies acute cerebral ischemia in while DWI (Figs. 10.110.3) easily depicts such
71% to 80% of cases. With the addition of DWI, this lesions. Fluid attenuated inversion recovery (FLAIR)
Localization of stroke syndromes using diffusion-weighted MR imaging (DWI) 123

Fig. 10.2. 58-year-old female patient admitted for headaches, nausea, vomiting, vertigo and dysarthria. Physical examination let a
left cerebellar stroke be suspected. The latter was demonstrated in the territory of the left superior cerebellar artery by T2-
weighted, DWI-trace and ADC obtained 24 hours after admission. On the other hand, only a slight hypodensity could be retro-
spectively identied in the corresponding area on the admission cerebral CT, obtained 6 hours after onset.

demonstrates similar sensitivity to ischemic lesions which have been reported in 19% to 31% of patients
than DWI, but does not allow demonstration of with stroke in the posterior circulation. The occur-
early lesions and dierentiation between new and rence of false-negative DWI ndings was signicantly
old lesions.20 higher for small lesions and during the rst 24 hours
In the vertebro-basilar territory, there has been after stroke onset. In the anterior circulation, false-
some concern about false-negative DWI studies, negative initial DWI studies are much rarer (2%).21
124 Max Wintermark, Marc Reichhart, Reto Meuli and Julien Bogousslavsky

Fig. 10.3. 48-year-old male patient admitted for thrombosis of the vertebro-basilar artery responsible for a locked-in syndrome.
Intra-arterial thrombolysis was attempted without success. Post-thrombolysis CT and MR examinations demonstrate hemor-
rhagic transformation of a complete vertebro-basilar stroke. DWI aords precise delineation of ischemic areas in both cerebellar
hemispheres as well as in the brainstem. Hemorrhagic transformation is responsible for susceptibility eects, resulting in pontine
hyposignal on T2-weighted images and DWI-trace due to the T2-shine through.
Localization of stroke syndromes using diffusion-weighted MR imaging (DWI) 125

Fig. 10.4. 75-year-old male patient admitted at our institution

after sudden onset of a left facearmleg sensitivomotor
hemisyndrome. Physical examination revealed a left hemi-
anopsia, a rightward gaze deviation, a dysarthria and a left
hemineglect. The patient underwent cerebral CT and MR
examinations, respectively, 2 and 3 hours after admission.
The DWI-trace clearly depicts an acute stroke extending to
the supercial right middle cerebral artery territory (arrow),
whereas the latter is much more subtle on the T2-weighted
image and especially on the non-contrast cerebral CT, where
it features a cortical ribbon loss sign.

DWI and stroke extent

The NINDS and ECASS studies have demonstrated

an increased risk of hemorrhagic transformation in
stroke patients with a large area of hypodensity on
admission CT when treated with thrombolytic
therapy.22,23 The risk of hemorrhagic transformation
was signicantly increased for stroke involving
more than one-third of the middle cerebral artery
(MCA) territory.24 However, even if hypo-
attenuation of more than one-third of the MCA ter-
ritory was an exclusion criterion in ECASS, it often
remained unidentied at the time of enrolment and
11% of the total study population with large stroke
were inappropriately included in this study.25 This
observation correlated with a moderate interob-
server agreement in the identication of early
infarct signs involving more than one-third of the
MCA territory on the admission CT.25 During ECASS
II, a special eort was made to train participating
physicians to identify early infarct signs. This
resulted in a reduction in the number of inappro-
priately randomized patients to 5%.26
DWI has proved to be more sensitive than CT in
the assessment of the involved extent of the MCA
territory (Fig. 10.4). DWI and CT sensitivity ranges
between 57% and 86% and 14% and 43%, respec-
tively. Specicity has been reported excellent for
both imaging techniques and interobserver agree-
ment has shown only moderately good in both
imaging techniques (kappa: 0.6 for DWI and 0.5 for
CT).25,27 Furthermore, DWI has shown a more e-
cient early predictor of the volume of tissue ulti-
mately infarcted than CT.27
126 Max Wintermark, Marc Reichhart, Reto Meuli and Julien Bogousslavsky

examination at the time of their rst symptomatic

event. In our registry, among 1000 consecutive
patients with rst-ever stroke, 3% show infarcts in
multiple territories supplied by carotid arteries (Fig.
10.5), 2% display multiple infarcts in the vertebro-
basilar territory (Fig. 10.6) whereas 2% show multi-
ple infarcts in both territories. Embolism arising
from aortic branch vessel atheromatosis and
cardiac origin is the main cause of multiple cerebral
infarcts. Whether the latter is the result of the frag-
mentation of a single embolism, or whether it
causes multiple territory infarction by proximal
occlusion of major vessels with distal hemody-
namic infarction still remains unclear.28
Among patients with multiple cerebral ischemic
lesions, 30% show multiple acute infarcts, 56%
single acute infarct and multiple old infarcts, and
14% multiple acute and multiple old infarcts.29
One limitation of conventional MR imaging is
that acute and old infarcts appear very similar. T2-
weighted images do not allow for dierentiation of
acute vs. chronic infarcts, whereas T1-weighted
images with gadolinium enhancement are helpful
in only 12% of patients.29 This can cause diagnostic
confusion regarding which lesion is acute and
symptomatic, even with the help of neurological
ndings, especially in elderly patients whose con-
ventional MR examination demonstrates a large
number of high-signal foci in the corona radiata,
basal ganglia and brainstem on T2-weighted
Fig. 10.5. 52-year-old male patient found unconscious at
images. Moreover, in such patients, it may be di-
home. Admission cerebral CT was unremarkable. The patient
cult to determine whether new neurological symp-
was transferred to our institution and a cerebral MR was per-
formed. The latter demonstrated an extensive stroke involv- toms represent a new ischemic event, or just the
ing both anterior cerebral artey and left sylvian artery unmasking of a prior decit due to an intercurrent
territories. The precise stroke extent was better delineated on illness (Fig. 10.7).
DWI-trace and ADC than on T2-weighted image. Follow-up DWI overcomes conventional MR imaging limita-
cerebral CT, obtained 24 hours after cerebral MR displayed tions. Indeed, DWI can easily distinguish between
development of a malignant cerebral edema. The patient new and old ischemic brain injuries, as they respec-
died from intracranial hypertension. tively show hypo- and hypersignal on ADC maps
(Fig. 10.8). On the other hand, both are hyperin-
tense on the DWI trace, which, however, aords
Multiple cerebral infarcts: DWI demonstrates sharp delineation of pathologic processes through
the one responsible for the acute the removal of anisotropy of myelin bres and the
symptomatology absence of contrast between grey and white matter.
In 77% to 100% of patients with multiple cerebral
Patients with cerebrovascular diseases often have infarcts, DWI not only delineates early ischemic
multiple ischemic lesions seen on conventional MR brain injury better than conventional MR, but also
Localization of stroke syndromes using diffusion-weighted MR imaging (DWI) 127

Fig. 10.6. 72-year-old male patient admitted for violent headaches. Physical examination demonstrated severely altered cerebel-
lous tests, as well as dysarthria and bilateral corticospinal signs. MR examination obtained 10 hours after symptomatology onset
related these clinical symptoms and signs with bilateral acute strokes in the territory of superior cerebellar arteries.

successfully identies the acute lesion responsible lesions and thus often provides relevant informa-
for the clinical decit.20,29 Moreover, in 18% of acute tion allowing precise clinicaltopographic correla-
stroke patients, DWI has been reported to localize tions.
the symptomatic lesion in a dierent vascular terri- Identifying the origin of stroke, for instance a
tory from the one suspected, which relied upon the proximal source of embolism in the case of multiple
combination of the initial clinical evaluation and acute lesions or a hemodynamic cause for water-
the conventional MR examination. Finally, in 20% shed infractions, has considerable repercussions on
of acute stroke patients, DWI allowed to make clear stroke patient management, on investigations as
that lesions thought to be acute on the conven- well as on therapy.
tional MR sequences were actually old (Fig. 10.9).30 In patients with multiple ischemic events in the
anterior circulation (Fig. 10.10), imaging of the
carotid bifurcation should denitely be performed
Therapeutic impact of DWI because of the well-established benets of carotid
endarterectomy in surgically suitable candidates
DWI aords accurate localization of strokes. As it with symptomatic or signicant carotid stenosis.
successfully discriminates acute from old infarcts, it Moreover, the benet of carotid endarterectomy
represents a remarkable tool in the study of the has shown considerably higher in patients with
pathophysiology of multiple cerebral ischemic more recent ischemic symptoms.31 For patients
128 Max Wintermark, Marc Reichhart, Reto Meuli and Julien Bogousslavsky

Fig. 10.7. 76-year-old male patient admitted for aphasia and visual eld defect. Cerebral MR examination was obtained 4 hours
after symptomatology onset. T2-weighted images show multiple focal hyperintensities (arrowheads) in both basal nuclei, as well
as in external capsulae. These multiple punctual lesions have high-diusion properties, featuring hyposignal on DWI-trace and
hypersignal on the ADC maps. They relate to lacunas and dilated VirchowRobin spaces. On the other hand, more extensive acute
ischemic lesions (arrows) can be identied in the left temporal pole and in the left parieto-occipital region. They are also hyperin-
tense on T2-weighted images, but demonstrate low diusion properties, with hyper- and hyposignal on DWI-trace and ADC map,

with anterior circulation strokes who do not have symptomatic vertebro-basilar stenosis, anticoagu-
marked carotid stenoses, some propose MR angio- lation may be more eective than antiplatelet
graphy imaging of the distal carotid or of the proxi- agents.32 New therapies, such as vertebral angio-
mal anterior or middle cerebral arteries, plasty in patients with marked posterior circulation
conventional angiography or transcranial Doppler. occlusive disease, are being evaluated.33
If substantial intracranial lesions are identied, Finally, DWI can more frequently dierentiate a
recent evidence suggests that anticoagulation small deep subcortical infarct from a cortical (Fig.
therapy may be more benecial than antiplatelet 10.12) or a combined cortical/subcortical lesion
agents, and ongoing trials are attempting to deter- than conventional MR imaging can. In patients
mine the optimal therapeutic approach for these with small deep infarcts, many clinicians limit the
patients.32 diagnostic evaluation because of the relatively low
If the acute symptomatic lesion is found in the yield of detailed vascular and cardiac imaging in
posterior circulation (Fig. 10.11), diagnostic studies these patients, particularly if they have predispos-
may focus on the vertebro-basilar system. For a ing risk factors.34
Localization of stroke syndromes using diffusion-weighted MR imaging (DWI) 129

Fig. 10.8. 70-year-old male patient showing a left facearmleg Fig. 10.9. 70-year-old female patient admitted for a right
hemisyndrome with complex oculomotricity alteration and sensitivomotor facearmleg hemisyndrome associated with
internuclear ophthalmoplegia. Cerebral CT and MR examina- a motor aphasia. These ndings were related to a left anterior
tions were obtained 7 and 9 hours after symptomatology onset, choroidial stroke both by CT and MR examinations. Non-
respectively. Cerebral MR demonstrates three ischemic lesions contrast cerebral CT and T2-weighted images, as well as DWI-
in the brainstem, one of them being retrospectively identied trace, demonstrated a signal abnormality involving the
on the cerebral CT. Two of these ischemic lesions are acute. posterior arm of the left internal capsula, as well as portions
They feature hypersignal on T2- and DWI-trace, and hyposignal of the left thalamus and lentiform nucleus, adjacent to the
on the ADC maps. They are located in the right paramedian left lateral ventricle. ADC, displaying increased diusion
pons and mesencephalon, and in the antero-lateral portion of coecients in this area, revealed this stroke to be actually
the left pons (arrows). On the other hand, an old lesion, also evolving from at least 1 or 2 weeks.
hyperintense on T2-weighted images and ADC maps, is
identied in the left central paramedian pons (arrowhead). This
combination explained the patients intricate clinical picture.
130 Max Wintermark, Marc Reichhart, Reto Meuli and Julien Bogousslavsky

Fig. 10.10. 50-year-old male patient showing a complete right hemisyndrome. Cerebral MR examination performed 9 hours after
onset attributed the symptomatology to multiple acute ischemic lesions, featuring hypersignals on DWI-trace and hyposignals on
ADC maps. These lesions are distributed in the left anterior and middle cerebral artery territories. This distribution allowed for
pathology of the left internal carotid artery to be suspected. Cervical angio-MR and Doppler sonography demonstrated severe
atherosclerotic changes.

Conclusion 4 Stejskal EO, Tanner JE. Spin diusion measurements:

spin echoes in the presence of a time-dependent eld
In conclusion, DWI can easily be added to conven- gradient. J Chem Phys 1965; 42: 288292.
5 Basser PJ, Mattiello J, Le Bihan D. Estimation of the
tional MR imaging. In acute stroke patients, it pro-
eective self-diusion tensor from the NMR spin
vides potentially relevant ndings that do not show
echo. J Magn Reson 1994; 103: 247254.
on conventional MR sequences. It improves the
6 Moseley M, Kucharczyk J, Mintorovirch J et al.
diagnosis of multiple brain infarcts, distinguishing Diusion-weighted MR imaging of acute stroke: cor-
between acute and old lesions. It also aords a better relation with T2-weighted and magnetic susceptibility
correlation between clinical symptoms and the site enhanced MR imaging in cats. Am J Neuroradiol 1990;
of ischemic lesions. Thus, it considerably improves 11: 423429.
stroke patient care, leading to investigations focused 7 Moseley M, Cohen Y, Mintorovitch J et al. Early detec-
on the etiology of the cerebral infarct. It may some- tion of regional cerebral ischemia in cats: comparison
times result in a change of the therapeutic strategy, of diusion- and T2-weighted MRI and spectroscopy.
hopefully preventing further stroke recurrence. Magn Reson Med 1990; 14: 330346.
8 Jones SC, Perez-Trepichio AD, Xue M, Furlan AJ, Awad
IA. Magnetic resonance diusion-weighted imaging:
sensitivity and apparent diusion constant in stroke.
REFERENCES Acta Neurochir 1994; 60: 207210.
9 Warach S, Chien D, Li W, Ronthal M, Edelman RR. Fast
1 Carr HY, Purcell EM. Eects of diusion on free pre- magnetic resonance diusion-weighted imaging of
cession in nuclear magnetic resonance experiments. acute human stroke. Neurology 1992; 42: 17171723.
Phys Rev 1954; 94: 630638. 10 Baird A, Warach S. Magnetic resonance imaging of acute
2 Woessner DE. Eects of diusion in nuclear magnetic stroke. J Cereb Blood Flow Metab 1998; 18: 583609.
resonance spin-echo experiments. J Chem Phys 1961; 11 Beaulieu C, de Crespigny A, Tong DC, Moseley DC,
34: 20572061. Albers GW, Marks MP. Longitudinal magnetic reso-
3 Stejskal EO. Use of spin echoes in a pulsed magnetic- nance imaging study of perfusion and diusion in
eld gradient to study anisotropic restricted diusion stroke: evolution of lesion volume and correlation
and ow. J Chem Phys 1965; 43: 35973603. with clinical outcome. Ann Neurol 1999; 46: 568578.
Localization of stroke syndromes using diffusion-weighted MR imaging (DWI) 131

Fig. 10.11. 77-year-old male patient with sudden onset of cortical blindess and decreased level of consciousness. Cerebral CT and
MR examinations were obtained 6 and 7 hours after admission. They display acute strokes in both occipital lobes, as well as in the
posterior arm of the right internal capsula. MR aords better demonstration of bilateral involvement of thalami, which explained
the decreased level of consciousness. Stroke occurrence in both posterior cerebral artery territories raised suspicion about a
vetebro-basilar pathology in this patient. Both angio-MR and doppler-sonography demonstrated a stenosis of the distal portion of
the vertebro-basilar artery.
132 Max Wintermark, Marc Reichhart, Reto Meuli and Julien Bogousslavsky

Fig. 10.12. 75-year-old male patient admitted for a left sensitivomotor hemisyndrome with aphasia. Rapid resolution of symptoms
was observed, already in the emergency room. This clinical pattern was explained by a right parieto-occipital cortical infarct
(arrows). The latter was demonstrated only by DWI-trace and ADC, whereas it could not be seen on cerebral CT and T2-weighted
MR examination.
Localization of stroke syndromes using diffusion-weighted MR imaging (DWI) 133

12 Gonzalez RG, Schaefer PW, Buonanno F et al. 24 von Kummer R, Allen KL, Holle R et al. Acute stroke:
Diusion-weighted MR imaging: diagnostic accuracy usefulness of early CT ndings before thrombolytic
in patients imaged within 6 hours of stroke symptom therapy. Radiology 1997; 205: 327333.
onset. Radiology 1999; 210: 155162. 25 von Kummer R, Holle R, Gizyska U et al. Interobserver
13 van Gelderen P, de Vleeschouwer MHM, DesPres D, agreement in assessing early CT signs of middle cere-
Pekar J, van Zijl PCM, Moonen CTW. Water diusion bral artery infarction. Am J Neuroradiol 1996; 17:
and acute stroke. Magn Reson Med 1994; 31: 154163. 17431748.
14 Ulug AM, Beauchamp N, Bryan RN, van Zijl PCM. 26 Hacke W, Kaste M, Fieschi C et al. Randomised
Absolute quantitation of diusion constants in double-blind trial placebo-controlled trial of
human stroke. Stroke 1997; 28: 483490. thrombolytic therapy with intravenous alteplase in
15 Neil JJ, Shiran SI. McKinstry RC et al. Normal brain in acute ischaemic stroke (ECASS II). Second
human newborns: apparent diusion coecient and EuropeanAustralasian Acute Stroke Study
diusion anisotropy measured by using diusion investigators. Lancet 1998; 352: 12451251.
tensor MR imaging. Radiology 1998; 209: 5768. 27 Lansberg MG, Albers GW, Beaulieu C, Marks MP.
16 Provenzale JR, Engelter ST, Petrella JR, Smith JS, Comparison of diusion-weighted MRI and CT in
MacFall JR. Use of MR exponential diusion-weighted acute stroke. Neurology 2000; 54: 15571561.
images to eradicate T2 shine-through eect. Am J 28 Bogousslavsky J, Van Melle G, Regli F. The Lausanne
Neuroradiol 1999; 172: 537539. Stroke Registry: analysis of 1,000 consecutive patients
17 Mohr JP, Biller J, Hilal SK et al. Magnetic resonance with rst ever stroke. Stroke 1988; 19: 10831092.
versus computed tomographic imaging in acute 29 Altieri M, Metz RJ, Mller C, Maeder P, Meuli R,
stroke. Stroke 1995; 26: 807812. Bogousslavsky J. Multiple brain infarcts: clinical and
18 Lutsep HL, Albers GW, DeCrespigny A, Kamat GN, neuroimaging patterns using diusion-weighted mag-
Marks MP, Moseley ME. Clinical utility of diusion- netic resonance. Eur Neurol 1999; 42: 7682.
weighted magnetic resonance imaging in the assess- 30 Albers GW, Lansberg MG, Norbash AM et al. Yield of
ment of ischemic stroke. Ann Neurol 1997; 41: 574580. diusion-weighted MRI for detection of potentially
19 Lansberg MG, Norbash AM, Marks MP et al. relevant ndings in stroke patients. Neurology 2000;
Advantages of adding diusion-weighted magnetic 54: 15621567.
resonance imaging to conventional magnetic reso- 31 The European Carotid Surgery Trial Group.
nance imaging for evaluation of acute stroke. Arch Randomised trial of endarterectomy for recently
Neurol 2000; 57: 13111316. symptomatic carotid stenosis: nal results of the MRC
20 Kumon Y, Zenke K, Kusunoki K et al. Diagnostic use of European Carotid Surgery Trial (ECST). Lancet 1998;
isotropic diusion-weighted MRI in patients with 351: 13791387.
ischaemic stroke: detection of the lesion responsible for 32 The WarfarinAspirin Symptomatic Intracranial
the clinical decit. Neuroradiology 1999; 41: 777784. Disease Study Group. The warfarinaspirin sympto-
21 Oppenheim C, Stanescu R, Dormont D et al. False- matic intracranial disease study. Neurology 1995; 45:
negative diusion-weighted MR ndings in acute 14881493.
ischemic stroke. Am J Neuroradiol 2000; 21: 33 Marks MP, Marcellus M, Norbash AM et al. Outcome
14341440. of angioplasty for atherosclerotic intracranial stenosis.
22 The NINDS t-PA Stroke Study Group. Intracerebral Stroke 1999; 30: 10651069.
hemorrhage after intravenous t-PA therapy for 34 Mead GE, Wardlaw JM, Lewis SC, McDowall M,
ischemic stroke. Stroke 1997; 28: 21092118. Dennis MS. Can simple clinical features be used to
23 Hacke W, Kaste M, Fieschi C et al. Intravenous throm- identify patients with severe carotid stenosis on
bolysis with recombinant tissue plasminogen activa- Doppler ultrasound? J Neurol Neurosurg Psychiatry
tor for acute hemispheric stroke. The European 1999; 66: 1619.
Cooperative Acute Stroke Study (ECASS). J Am Med
Assoc 1995; 274: 10171025.
MRI in transient ischemic attacks: clinical utility and
insights into pathophysiology

Jerey L. Saver and Chelsea Kidwell

UCLA Stroke Center, Los Angeles, CA, USA

The brain responds dynamically to transient epi- reecting three somewhat dissimilar ischemic
sodes of ischemic insult. Standard brain imaging episodes that can underlie clinically similar TIAs.
techniques, computed tomography (CT) and con- A very brief or low-intensity period of focal ische-
ventional magnetic resonance imaging (MRI), are mia may disrupt synaptic transmission and
insensitive to dynamic and regionally varying produce transient neurologic decits without
neural parenchymal responses to tissue ischemia. causing early cytotoxic edema or permanent
In contrast, the novel MRI techniques of perfusion tissue injury. In these cases, perfusion MRI may
and diusion imaging permit visualization of these show focally reduced cerebral blood ow, but
critical tissue processes, and have aorded new both acute diusion MRI, sensitive to early cyto-
insights into the physiopathology of human cere- toxic edema, and late T2 imaging, sensitive to
bral ischemia. In addition, clinical studies have increased water content, a marker of permanent
demonstrated that magnetic resonance imaging is parenchymal injury, will be unrevealing. A some-
of substantial clinical utility in patients with tran- what more severe transient ischemic insult may
sient ischemic attacks (TIAs). suciently disrupt cellular energetic state to
impair maintenance of ionic gradients across cell
membranes, producing cytotoxic edema, but not
Overview cause advanced bioenergetic failure. Early resto-
ration of blood ow may permit cellular
The current conventional denition of transient re-energization and restoration of ionic gradients,
ischemic attack is neurologic symptoms due to with edema resolution. In this setting, acute per-
focal cerebral ischemia that resolve completely fusion MRI during the episode and diusion MRI
within 24 hours (Special Report from the National close to the time of the episode will be abnormal,
Institute of Neurological Disorders and Stroke, but late T2 imaging will be unrevealing. A more
1990). Defects of this denition, already apparent in profound ischemic insult may produce loss of cell
the CT era, have been demonstrated even more membrane integrity, in addition to failure of syn-
pointedly by MRI investigations. We will conse- aptic transmission and cytotoxic edema, with
quently propose a new denition of TIA, informed resulting permanent parencyhmal injury.
by MRI ndings, at the conclusion of this chapter. However, early recruitment of alternative neural
Nonetheless, all MRI studies to be reviewed in this circuitry and synaptic outgrowth, neuroplasticity
chapter have employed this conventional deni- and neurorepair, may allow rapid resolution of
tion. clinical decits. In this setting, patients with
Recent studies in animal models and in human rapidly transient neurologic signs may exhibit
patients with conventionally dened TIAs have early perfusion, early diusion, and late T2 abnor-
identied three distinct tissue patterns on MRI, malities on MRI imaging.

136 Jeffrey L. Saver and Chelsea Kidwell

Standard MR imaging studies in patients with identied as acute (68%) were < 1.5 cm in diameter
transient ischemic attacks and 58% were purely cortical. Thirty-seven per cent
of the acute infarcts were multiple in nature.
Patients with clinical TIAs frequently have had prior Contrast enhancement occurred in 5 of the 45
cerebral ischemic events, both clinically manifest patients studied, but in only 2 patients contributed
and silent. Conventional MRI is more sensitive than to the delineation of the acute lesion. Evidence of
standard CT in identifying these pre-existing infarction in these patients was associated with
lesions, with MR in dierent studies showing evi- higher frequency of signicant vascular or cardiac
dence of at least one infarct somewhere in the cere- disorders.
brum in 81% of TIA patients,1 vs. 068% in CT Bhadelia and colleagues studied 100 TIA patients
series. Some of these infarcts are in locations that from the Cardiovascular Health Study imaged with
could have accounted for the decits observed standard MRI sequences.9 Brain infarcts were dem-
during the TIA. Among patients meeting clinical onstrated in 46% of the TIA patients compared to
criteria for TIA, 3139% demonstrate neuroanatom- 28% of patients without a history of TIA. In stepwise
ically relevant infarcts on conventional magnetic logistic regression analysis, diastolic blood pressure
resonance imaging1,2 and 248% on standard com- and internal carotid intima-media thickness were
puted tomography.38 It is dicult with conven- predictive of infarction on MRI. These authors also
tional MRI and CT to determine what proportion of found an increased frequency of cortical infarcts
these appropriately localized infarcts occurred at and multiple infarcts in the TIA patients.
the time of the index TIA, and what proportion In their study of 64 patients with carotid territory
existed prior to the presenting event. TIAs studied with MRI, Kimura and colleagues
The earliest report of MRI ndings in TIA patients found 16 of 41 patients demonstrated contrast
came from Awad and colleagues.7 This group enhancement.2 The majority of contrast-enhancing
studied 22 patients with both MRI and CT. They lesions were cortical (81%). Aphasia or confusional
found 77% of patients had focal ischemic changes state, hypertension and presence of an emboligenic
on MRI compared to 32% on CT. However, the cardiac or arterial source were more frequently
majority of lesions did not correlate with symptom- observed in patients with enhancement. The
atology. increased rate of contrast enhancement in this
Fazekas and colleagues reported the results of study compared to the Fazekas report may be
routine MR imaging in 62 patients with hemi- related to dierences in patient characteristics and
spheric TIAs.1 Forty-ve of these patients also had timing of the MRI studies.
contrast-enhanced studies. They found that 81% of Walters and colleagues performed serial MRI
their cohort had MRI evidence of focal ischemic studies over a 2-year period in 125 TIA patients and
injury, and reported that 31% demonstrated evi- compared the results to 75 controls. They found
dence of an acute TIA-associated infarct. The de- that 47% of the TIA patients demonstrated evidence
nition Fazekas and colleagues employed for acute of new asymptomatic lesions on follow-up imaging
TIA-associated infarcts was: lesions (i) located in a compared to 12% of controls. Thirteen TIA patients
vascular territory potentially corresponding to the had new cerebral transient symptoms, of which two
patients TIA symptoms, (ii) having signal charac- had new relevant MRI lesions. In addition, four TIA
teristics of appearing hyperintense on T2-weighted patients had experienced a stroke during the
scans and isointense or only minimally hypoin- follow-up period (three ischemic events, one intra-
tense on T1-weighted scans, and (iii) showing cerebral hemorrhage). Factors that correlated with
regional swelling or contrast enhancement. This an increased risk of a new ischemic lesion included
denition is problematic, as many lesions remote diastolic blood pressure, male sex, age, and initial
in time may demonstrate this T2/T1 pattern and the severity of MRI ischemic lesions. Also of note, these
accuracy of rater discrimination of swelling for authors found that TIA patients had an accelerated
small lesions is uncertain. The majority of infarcts rate of brain atrophy compared to controls.
MRI in transient ischemic attacks: clinical utility and insights into pathophysiology 137

MR spectroscopy in transient ischemic attacks


% with DWI lesion

MR spectroscopy is an interesting new application 60
of MRI for the study of patients with TIA. A prelimi- 6
nary report by Giroud and colleagues of 5 TIA 40

patients found no dierences in NAA/creatine ratio,

but did nd an increase in lactate/creatine ratio.
Further studies are needed to conrm and clarify 0
1030 0.51 13 36 612 1224
the signicance of these ndings.
min h h h h h

s d
Diffusion MRI in transient ischemic attacks
Fig. 11.1. Relation of TIA symptom duration to presence or
absence of DWI abnormality among 42 consecutive patients
Studies from several groups have demonstrated
in UCLA series. (From UCLA Stroke Center.)
that diusion MR imaging provides a more sensi-
tive and specic evaluation of ischemic insult in
TIA patients as compared to standard CT and MRI
studies.1013 hospital with symptoms suggestive of a cerebral
TIA. Forty-two underwent MRI scanning and 19 did
not, due to metal implant (5), refused/claustropho-
UCLA study
bia (6), MRI technical diculty (3), and non-stroke
We studied consecutive TIA patients using diusion attending preference (5).
MR imaging and compared them to a group of con- Among the 42 clinical TIA patients, 20 (48%)
temporaneous completed stroke patients to deter- demonstrated a focal abnormality on diusion-
mine: (i) the incidence of diusion-weighted weighted imaging, consistent with acute neural
imaging (DWI) and apparent diusion coecient bioenergetic compromise. In ve (25%) of these 20
(ADC) abnormalities in TIA patients compared to patients, there was no lesion correlate on initial T2-
standard T2-weighted MRI sequences; (ii) whether weighted (T2W) sequences (example of a more
the presence of a diusion MR imaging abnormal- recent similar patient is shown in Fig. 11.1). The
ity correlates with the duration, location or mecha- remaining 15 patients did exhibit T2-weighted
nism of symptoms; (iii) whether the diusion MR lesion abnormalities in the same regions as DWI
imaging signature in TIA diers from that in com- alterations (example of a more recent similar
pleted stroke; and (iv) the impact of diusion patient is shown in Fig. 11.2). Two patients with
imaging data on clinical diagnosis of TIA localiza- visible but very small DWI abnormalities did not
tion and mechanism.10 have a measurable ADC volume. There was a strong
Clinical, conventional MRI, and diusion MRI correlation between ADC and DWI volumes
data were collected on consecutive patients pre- (r0.77). There were no signicant dierences
senting to the UCLA Medical Center over a 6-month between patients who demonstrated DWI abnor-
period with symptoms of a transient ischemic malities and those who did not, in age, sex, and
attack. TIAs were dened as symptoms of pre- presence of hypertension, diabetes, tobacco use,
sumed ischemic cerebrovascular etiology lasting hypercholesterolemia, or history of prior stroke or
less than 24 hours. Patients with brainstem and/or TIA.
hemispheric symptoms were included while A precise estimation of TIA duration (all were
patients with isolated amaurosis fugax were unequivocally less than 24 hours) was available for
excluded. All MRI scans were obtained within 3 15 of the 20 patients with DWI abnormalities and 17
days of symptom onset. During the enrolment of the 22 without DWI abnormalities. Duration of
period, 61 patients were admitted to our university TIA symptoms for patients without a DWI
138 Jeffrey L. Saver and Chelsea Kidwell

Fig. 11.2. A 63-year-old male presented with right arm weakness lasting 30 minutes. Top row shows consecutive axial DWI
sequences demonstrating left corona radiata lesion (thick arrows) not obviously apparent on FLAIR sequences (bottom row, thin
arrows). (From UCLA Stroke Center.)

abnormality was mean 3.2 hours ( 4.7 hours stan- MR imaging, while two in the DWI positive category
dard deviation), median 0.5 hours vs. mean 7.3 were still symptomatic at the time of MR imaging.
hours ( 6 hours standard deviation), median 4.0 Interval from time of resolution of TIA symptoms to
hours for patients with a DWI abnormality (t test for time of MRI for patients with a DWI abnormality
dierence in means, P0.03). The percentage of was mean 12.7 hours, median 8.8 hours vs. mean
patients with a DWI abnormality within various 12.9 hours, median 5.1 hours in patients without a
symptom duration intervals increased as the total DWI abnormality (rank sums test for dierence in
duration of symptoms increased (Fig. 11.3). medians, P0.7).
The mean time from symptom onset to MRI In the 20 patients with diusion MRI abnormal-
study for all TIA patients was 17 hours (range ities, DWI signal changes were localized to the
1.2573 hours), and did not signicantly dier brainstem in four patients, the cerebellum in two
between the two groups (mean 15.8 hours for DWI- patients, subcortical hemispheric structures in
negative patients, mean 19.5 hours for DWI- seven patients, and cortical regions in seven
positive patients). One patient in the group without patients. Vascular territories aected were super-
DWI lesions was still symptomatic at the time of the cial middle cerebral artery in six patients, deep
MRI in transient ischemic attacks: clinical utility and insights into pathophysiology 139

Fig. 11.3. A 79-year-old patient presented with dysarthria and bilateral arm incoordination lasting one hour. DWI (left image)
sequence showing right cerebellar lesions, also subtly evident on T2-weighted sequence (right image). (From UCLA Stroke Center.)

middle cerebral artery in six patients, brainstem botic mechanisms, one changed post-DWI to likely
perforators in four patients, and posterior cerebral cardioembolic and one changed to likely small
arteries in two patients. In these 20 patients, nal vessel; of seven initial cardioembolic diagnoses,
etiologic mechanism was felt to be small vessel one changed to likely large vessel atherothrombotic
lacunar in nine patients, large vessel atherothrom- and one changed to likely small vessel; and of nine
botic in four patients, and cardioembolic in seven initial small vessel diagnoses, one changed to likely
patients. large vessel atherothrombotic and one changed to
DWI results altered the attending physicians likely cardioembolic.
opinion regarding vascular localization in 7/20 There were signicant dierences in the DWI
patients, anatomical localization in 8/20 patients and ADC MR signatures between the TIA
and probable TIA mechanism in 6/20 patients. The patients with DWI abnormalities and the patients
types of alterations in diagnosis were quite varied with completed stroke (Table 11.1). Completed
and no single pattern predominated. For example, stroke patients had larger volumes and greater
among etiological diagnoses, of four patients ini- intensities of ADC and DWI alteration than TIA
tially suspected to have large artery atherothrom- patients.
140 Jeffrey L. Saver and Chelsea Kidwell

Table 11.1. Comparison of DWI signature between TIA patients with diffusion MR abnormality and patients with completed stroke

TIA patients (n 20) Completed stroke patients (n 23) P value

mean (SEM) mean (SEM)

DWI lesion volume (cm3) 2.9 (8.4) 22.2 (7.8) 0.0002a

ADC volume (cm3) 0.7 (3.4) 10.5 (3.2) 0.0001a
DWI intensity 35% (5%) 62% (5%) 0.001b
Mean ADC value (m2/s) 442 (6.7) 409 (6.1) 0.009b

Rank Sums Test
T test. Modied, with permission, from reference 10.

Table 11.2. Time intervals and yield of diffusion MRI in transient ischemia attack patients: three series

Time from TIA onset Frequency of positive

Series TIA duration (mean) (h) to MRI (mean) (h) DWI ndings on MRI (%)

UCLA (n42) 3.2a 17 48

Duke (n40) 4.8 37 35
MGH (n57) 1.9 39 46

Median duration was 2.0 hours.

All 20 TIA patients demonstrating DWI abnor- ings.12,13 Though diering somewhat in cohort
malities were contacted for a follow-up MRI, and characteristics and timing of MRIs, the four series
nine of these patients agreed to return for repeat show convergent results regarding the frequency of
neuroimaging. Three patients were studied with DWI positivity among TIA patients, ranging from
head CT and six with brain MRI 27 months post 35% in the Duke cohort, 37% in the Takayama
event. Of these nine patients, ve (three MRI, two cohort, 46% in the MGH cohort, and 48% in the
CT) demonstrated a subsequent infarct in the UCLA cohort (Table 11.2). Aggregating the four
region corresponding to the original DWI abnor- series, among 158 TIA patients studied, 42% exhib-
mality while four (three MRI, 1 CT) did not. Five of ited diusion MRI abnormalities. The Duke investi-
the 22 patients without a DWI abnormality under- gators found that in their cohort, as in ours, TIAs of
went follow-up imaging (three MRI, two CT), 2 longer clinical duration were more likely to be DWI
weeks to 15 months post event. None demonstrated positive. Among DWI positive patients, mean TIA
a subsequent relevant infarct. duration was 7.1 hours in the Duke cohort and
7.3 hours in the UCLA cohort; in DWI negative
patients mean TIA duration was 3.2 hours in both
Duke, Massachusetts General Hospital and
cohorts. In contrast, TIA duration was not a predic-
additional studies
tor of DWI positivity in the MGH series. In part, this
Studies of diusion MRI in TIA patients performed discrepancy may be due to the briefer average
by investigators at Duke University, Massachusetts duration of TIAs and the longer interval from TIA
General Hospital (MGH), and by Takayama and col- oset to MR imaging in the MGH study. The MGH
leagues have conrmed and extended our nd- investigators did nd that prior, non-stereotyped
MRI in transient ischemic attacks: clinical utility and insights into pathophysiology 141

TIAs, identied stroke etiology, and cortical symp- patients. These dierences support the concept
toms were independent clinical predictors of DWI that the cerebral ischemia experienced by patients
positivity. These clinical factors seem to index with TIAs is lesser in volume and severity than that
larger, more severe ischemic episodes, as does experienced by patients with clinically completed
longer duration of clinical decits, and this under- stroke syndromes.
lying physiologic factor is likely to be the most criti- Both the UCLA and Duke series found a strong
cal for the appearance of DWI abnormality. statistical correlation between duration of TIA
Takayma and colleagues studied 19 TIA patients symptoms and presence of a lesion on DWI. This
with diusion imaging and found that seven (37%) correlation, however, was not absolute. DWI lesions
demonstrated DWI positivity. On follow-up appeared in patients with clinical episodes as brief
imaging, all of the DWI lesions had evolved to per- as 10 minutes, while some patients in the DWI neg-
sistent T2 lesions. ative group had symptoms lasting more than
In addition to the above series, there are two case 12 hours. DWI abnormalities do appear to be
reports in the literature of DWI lesions associated uncommon, if present at all, in patients with clini-
with TIAs, both reporting reversibility of the DWI cal symptoms lasting less than 5 minutes.
abnormalities on follow-up imaging.14,15 In one of In addition to improving our understanding of
these patients, reversal of an initial perfusion the underlying pathophysiological processes that
abnormality was also demonstrated.15 occur with TIAs, these data add to a growing body
of evidence demonstrating the clinical utility of
DWI.17,18 A variety of studies have demonstrated
Discussion of diffusion MR findings
that the diagnosis of TIA is often dicult, especially
These studies of diusion MRI in transient for the non-neurologist.19,20 Kraaijeveld and col-
ischemic attack provide important new insights leagues found kappa measures of inter-rater agree-
into the pathophysiology of TIAs and the clinical ment of only 0.65 among eight experienced
utility of new MRI sequences in TIA patients. Across neurologists diagnosing 56 TIA patients and of only
all series, more than two of every ve cerebral TIA 0.31 for determination of the vascular territory
patients demonstrated diusion MRI evidence of involved.21 The size, appearance and location of
acute bioenergetic compromise. In the UCLA study, DWI lesion(s) in TIA may help guide physicians in
among TIA patients with early DWI abnormalities determining the underlying etiologic mechanism
who had follow-up imaging, approximately one- and in choosing the optimal therapeutic regimen to
half exhibited late CT or MRI evidence of estab- reduce the probability of recurrent TIAs or com-
lished infarction. Together, these data suggest that pleted stroke in the future.
approximately one quarter of cerebral TIA patients In the UCLA study, information obtained from
actually have cerebral infarction with transient the DWI study led to a change in the suspected ana-
signs. We, and others, also identied a distinct tomical localization, vascular localization and TIA
subset of TIA patients, representing about one fth mechanism in over one third of patients. In addi-
of TIA cases, who had early DWI abnormalities but tion to clarifying the site and source of ischemia in
no late evidence of established infarction. This patients with clinically denite TIAs, diusion
nding in TIA patients suggests that DWI abnor- imaging also can be quite helpful in patients with
malities may be reversible in humans if early resto- atypical transient neurologic symptoms, when it is
ration of blood ow is obtained. This observation unclear whether the event was a TIA vs. migraine,
has been conrmed, with important additional hyperventilation, brief seizure, or other TIA mimic.
complexities, by MRI studies in patients undergo- Although DWI abnormalities have rarely been
ing reperfusion after thrombolytic stroke therapy.16 reported in TIA mimics, a visualized diusion
In TIA patients, the ADC volume, mean ADC abnormality in these cases generally provides sup-
value, DWI volume and DWI signal intensity were portive evidence of the diagnosis of TIA.
all signicantly less abnormal than in acute stroke The observation that DWI alone was positive in
142 Jeffrey L. Saver and Chelsea Kidwell

25% of patients, while 75% had correlative lesions are experiencing ongoing ischemia will need to be
identied retrospectively on T2-weighted imaging claried by large series of concurrent perfusion
underestimates the diagnostic impact of DWI. Even studies. The severity and size of the perfusion
in the patients with T2 visible lesions, the diusion decit might also be an indicator of the reversibility
imaging provided added clinical utility. Many of the of the diusion abnormality. Finally, the pathologic
T2 positive patients had multiple foci of increased correlates of DWI changes in TIA require investiga-
T2 signal, and determining which, if any, T2 foci tion, including how often signal abnormalities
were new and related to the recent TIA may not reect, at the histopathologic level, absence of
have been possible without the DWI sequences. infarction, incomplete infarction, or complete
Standard T2-weighted sequences alone are gener- infarction.27
ally incapable of reliably dierentiating acute from
chronic events.
Identifying which patients have a new infarct on A new, tissue-based definition of transient
imaging may have important prognostic value.22 In ischemic attacks
the Dutch TIA trial, evidence of any cerebral infarct
on CT was an independent risk factor for subse- Time-based denitions for TIA rst arose in the
quent stroke, myocardial infarction, or vascular 1950s as an imprecise means to distinguish
death.23,24 Evans and colleagues reported that in TIA between those cerebral ischemic episodes that
patients, CT veried infarction increased the risk of caused brain injury and those that did not, in the
death by 109% over a 10-year period following the absence of imaging or other laboratory measures
TIA.5 However, this study did not correlate the risk that could directly determine tissue parenchymal
with evidence of a new, appropriately located TIA status. Proposed time cutos varied widely. A 1958
related infarct. Eliasziw and colleagues did not nd NIH committee on classication of cerebrovascular
an increased risk of ipsilateral stroke in a group of disease suggested that TIAs could last as long as 1
TIA patients with severe carotid stenosis treated hour.28 Acheson and Hutchinson, in 1964, also
medically as part of the North American employed a 1-hour threshold to distinguish TIA
Symptomatic Carotid Endarterectomy Trial; from stroke.29 However, also in 1964, Marshall
however, these results cannot be generalized to all employed a 24-hour limit in dening TIA, although
TIA patients.25 Only larger series with long-term his data showed that symptoms lasted less than
follow-up will be able to distinguish if there is a 1-hour in three-fourths of his patients.30 In the 1975
dierence in prognosis in TIA patients without revision of the NIH classication document, a
diusion abnormalities, TIA patients with transient 24-hour limit for TIAs was adopted.31
diusion abnormalities but no eventual T2 lesion, Accumulating evidence suggests that any time
and patients with diusion abnormalities and a cuto for TIA is inaccurate in reecting end organ
subsequent T2 lesion. injury. The current 24-hour operational denition is
Several issues require further study. The clinical especially misleading. Large-scale studies have
prognostic signicance of nding an associated claried our understanding of the typical duration
DWI abnormality in a patient with a TIA remains of TIAs, showing that most TIAs resolve within
uncertain. We concur with the general view 1060 minutes rather than lasting several hours.32,33
advanced by Caplan that all TIA patients are at sig- Moreover, diusion MR ndings of diusion
nicant risk of subsequent vascular events and it is change in patients with spells as brief as 10 minutes
the underlying mechanism rather than the duration challenge the simplistic assumption that, because
of symptoms that is most critical to determine.26 clinical TIA symptoms rapidly resolve, signicant
However, it may be that within each mechanism ischemic tissue injury must not occur.
category, longer duration of a TIA or presence of a MR imaging studies have demonstrated the
DWI abnormality identies a subgroup at increased untenability of any denition of TIA based solely on
risk. How often patients with DWI abnormalities clinical manifestations and an arbitrarily assigned
MRI in transient ischemic attacks: clinical utility and insights into pathophysiology 143

time window, rather than tissue changes and physio- This research denition recognizes three levels of
logic processes. While the likelihood of DWI altera- strength of evidence for a spell being a TIA. The
tions is directly related to the duration of symptoms, lowest level is clinical criteria alone. The next level
some patients with spells as brief as 10 minutes will higher is clinical criteria plus a supportive test
show parenchymal changes on diusion imaging showing that no parenchymal tissue injury
and some with spells exceeding 12 hours will show occurred during the spell. The supportive test
no diusion alteration. There is not likely to be a should be highly sensitive to subtle brain paren-
fundamental biologic dierence between a patient chymal injury, such as diusion MR. Conventional
whose symptoms last 59 minutes and a patient CT and MR are insucient. Serum biomarkers of
whose symptoms last 61 minutes, or between 23 brain parenchymal injury, such as the S-100
hour 59 minute spells and 24 hour 1 minute spells. protein and neuron specic enolase,34,35 may be
Accordingly, we propose the following new de- useful alternative laboratory measures for this
nition of transient ischemic attacks, based on fun- denition. The highest level of evidence addition-
damental physiologic processes indexed by ally includes imaging evidence of focal hypoperfu-
imaging or other laboratory measures, rather than a sion during the spell. This nding helps to exclude
strict time limit: A TIA is a brief episode of neuro- seizures, compressive neuropathies, and other TIA
logic dysfunction due to focal cerebral ischemia, that mimics and rules in focal hypoperfusion as the eti-
is not associated with permanent brain injury. ology of the episode. Any of the wide variety of
Although most TIAs last 1 minute to 2 hours, a perfusion imaging modalities available could
minority last up to 24 or more hours. provide the data required for this level of evidence,
Under this denition, the diagnosis of TIAs may including CT perfusion imaging, xenon CT, perfu-
be rendered on clinical grounds alone. However, for sion MR, single positron emission tomography
research purposes, it is useful to have a more (SPECT), positron emission tomography (PET),
detailed, strictly operationalized denition, again transcranial Doppler ultrasound, and cerebral
based on the presence or absence of tissue injury angiography.
rather than time interval. We propose the following:
A TIA is a brief episode of neurological dysfunc-
tion, presumed to be due to focal cerebral Conclusion
ischemia, that is not associated with perma-
nent brain injury. Although most TIAs last 1 Magnetic resonance imaging has fundamentally
minute to 2 hours, a minority last up to 24 or altered our understanding of the pathophysiology
more hours. of transient ischemic attack. The spectrum of
Highly probable TIA: transient neurological dys- ischemic tissue alterations underlying transient
function presumed to be due to focal cerebral clinical symptoms is now understood to variably
ischemia, imaging evidence of focal hypoper- include synaptic transmission failure, cytotoxic
fusion during episode, no laboratory/imaging edema and permanent tissue injury, and these pro-
evidence of tissue injury. cesses are easily delineated in individual patients
Probable TIA: transient neurological dysfunction on MR imaging. In routine clinical practice, MR
presumed to be due to focal cerebral ischemia, permits conrmation of focal ischemia rather than
perfusion imaging not performed during another process as the cause of a patients decit,
episode, no laboratory/imaging evidence of improves accuracy of diagnosis of the vascular
tissue injury. localization and etiology of TIA, and assesses the
Possible TIA: transient neurological dysfunction extent of pre-existing cerebrovascular injury.
presumed to be due to focal cerebral ischemia, Accordingly, MRI, including diusion sequences,
perfusion imaging not performed during should now be considered a preferred diagnostic
episode, no sensitive laboratory/imaging test test in the investigation of the patient with poten-
of tissue injury performed. tial transient ischemic attacks.
144 Jeffrey L. Saver and Chelsea Kidwell

AC K N OW L E D G E M E N T S 10 Kidwell CS, Alger JR, Di Salle F et al. Diusion MRI in

patients with transient ischemic attacks. Stroke 1999;
We are grateful to our collaborators in the UCLA 30: 11741180.
Magnetic Resonance Imaging in Human Cerebral 11 Engelter ST, Provenzale JM, Petrella JR, Alberts MJ.
Reperfusion Study Group: Jerey R. Alger, PhD, Diusion MR imaging and transient ischemic
Gary Duckwiler, MD, Y. Pierre Gobin, MD, Reza attacks [letter; comment]. Stroke 1999; 30:
Jahan, MD, James Mattiello, PhD, Sidney Starkman,
12 Ay H, Buonanno FS, Schaefer PW et al. Clinical and
MD, Pablo Villablanca, and Fernando Vinuela, MD.
diusion-weighted imaging characteristics of an iden-
tiable subset of TIA patients with acute infarction
(Abstract). Stroke 1999; 30: 235A.
13 Takayama H, Mihara B, Kobayashi M, Hozumi A,
REFERENCES Sadanaga H, Gomi S. [Usefulness of diusion-
weighted MRI in the diagnosis of transient ischemic
1 Fazekas F, Fazekas G, Schmidt R, Kapeller P, attacks]. No To Shinkei 2000; 52: 919923.
Oenbacher H. Magnetic resonance imaging corre- 14 Lecouvet FE, Duprez TP, Raymackers JM, Peeters A,
lates of transient cerebral ischemic attacks. Stroke Cosnard G. Resolution of early diusion-weighted and
1996; 27: 607611. FLAIR MRI abnormalities in a patient with TIA.
2 Kimura K, Minematsu K, Wada K, Yonemura K, Yasaka Neurology 1999; 52: 10851087.
M, Yamaguchi T. Lesions visualized by contrast- 15 Neumann-Haefelin T, Wittsack HJ, Wenserski F et al.
enhanced magnetic resonance imaging in transient Diusion- and perfusion-weighted MRI in a patient
ischemic attacks. J Neurol Sci 2000; 173: 103108. with a prolonged reversible ischaemic neurological
3 Dennis M, Bamford J, Sandercock P, Warlow C. decit. Neuroradiology 2000; 42: 444447.
Prognosis of transient ischemic attacks in the 16 Kidwell CS, Saver JL, Mattiello J et al. Thrombolytic
Oxfordshire Community Stroke Project. Stroke 1990; reversal of acute human cerebral ischemic injury
21: 848853. shown by diusion/perfusion magnetic resonance
4 Koudstaal PJ, van Gijn J, Frenken CW et al. TIA, RIND, imaging. Ann Neurol 2000; 47: 462469.
minor stroke: a continuum, or dierent subgroups? 17 Lee LJ, Kidwell CS, Alger J, Starkman S, Saver JL.
Dutch TIA Study Group. J Neurol Neurosurg Psychiatry Impact on stroke subtype diagnosis of early
1992; 55: 9597. diusion-weighted magnetic resonance imaging and
5 Evans GW, Howard G, Murros KE, Rose LA, Toole JF. magnetic resonance angiography. Stroke 2000; 31:
Cerebral infarction veried by cranial computed 10811089.
tomography and prognosis for survival following tran- 18 Lutsep HL, Albers GW, DeCrespigny A, Kamat GN,
sient ischemic attack. Stroke 1991; 22: 431436. Marks MP, Moseley ME. Clinical utility of diusion-
6 Bogousslavsky J, Regli F. Cerebral infarct in apparent weighted magnetic resonance imaging in the assess-
transient ischemic attack. Neurology 1985; 35: ment of ischemic stroke. Ann Neurol 1997; 41:
15011503. 574580.
7 Awad I, Modic M, Little JR, Furlan AJ, Weinstein M. 19 Ferro JM, Falcao I, Rodrigues G et al. Diagnosis of
Focal parenchymal lesions in transient ischemic transient ischemic attack by the nonneurologist. A
attacks: correlation of computed tomography and validation study. Stroke 1996; 27: 22252229.
magnetic resonance imaging. Stroke 1986; 17: 20 Calanchini PR, Swanson PD, Gotshall RA et al.
399403. Cooperative study of hospital frequency and
8 Davalos A, Matias-Guiu J, Torrent O, Vilaseca J, Codina character of transient ischemic attacks. IV. The reli-
A. Computed tomography in reversible ischaemic ability of diagnosis. J Am Med Assoc 1977; 238:
attacks: clinical and prognostic correlations in a pros- 20292033.
pective study. J Neurol 1988; 235: 155158. 21 Kraaijeveld CL, van Gijn J, Schouten HJ, Staal A.
9 Bhadelia RA, Anderson M, Polak JF et al. Prevalence Interobserver agreement for the diagnosis of transient
and associations of MRI-demonstrated brain infarcts ischemic attacks. Stroke 1984; 15: 723725.
in elderly subjects with a history of transient ischemic 22 Toole JF. The Willis lecture: transient ischemic attacks,
attack. The Cardiovascular Health Study. Stroke 1999; scientic method, and new realities. Stroke 1991; 22:
30: 383388. 99104.
MRI in transient ischemic attacks: clinical utility and insights into pathophysiology 145

23 The Dutch TIA Trial Study Group. Predictors of 28 Ad Hoc Committee on Cerebrovascular Disease of the
major vascular events in patients with a tran- Advisory Council of the National Institute of
sient ischemic attack or nondisabling stroke. Neurological Disease and Blindness. A classication of
The Dutch TIA Trial Study Group. Stroke 1993; 24: and outline of cerebrovascular diseases. Neurology
527531. 1958; 8: 395434.
24 van Swieten JC, Kappelle LJ, Algra A, van Latum JC, 29 Acheson J, Hutchinson EC. Observations on the
Koudstaal PJ, van Gijn J. Hypodensity of the cerebral natural history of transient cerebral ischaemia. Lancet
white matter in patients with transient ischemic 1964; 2: 871874.
attack or minor stroke: inuence on the rate of subse- 30 Marshall J. The natural history of transient ischaemic
quent stroke. Dutch TIA Trial Study Group. Ann Neurol cerebrovascular attacks. Q J Med 1964; 33: 309324.
1992; 32: 177183. 31 NIH. A classication and outline of cerebrovascular
25 Eliasziw M, Streier JY, Spence JD, Fox AJ, Hachinski diseases. II. Stroke 1975; 6: 564616.
VC, Barnett HJ. Prognosis for patients following a 32 Levy DE. How transient are transient ischemic
transient ischemic attack with and without a cerebral attacks? Neurology 1988; 38: 674677.
infarction on brain CT. North American Symptomatic 33 Dyken ML, Conneally M, Haerer AF et al. Cooperative
Carotid Endarterectomy Trial (NASCET) Group. study of hospital frequency and character of transient
Neurology 1995; 45: 428431. ischemic attacks. I. Background, organization, and
26 Caplan LR. Are terms such as completed stroke or clinical survey. J Am Med Assoc 1977; 237: 882886.
RIND of continued usefulness? Stroke 1983; 14: 34 Elting JW, de Jager AE, Teelken AW et al. Comparison
431433. of serum S-100 protein levels following stroke and
27 Li F, Liu KF, Silva MD et al. Transient and permanent traumatic brain injury. J Neurol Sci 2000; 181: 104110.
resolution of ischemic lesions on diusion-weighted 35 Persson L, Hardemark HG, Gustafsson J et al. S-100
imaging after brief periods of focal ischemia in rats : protein and neuron-specic enolase in cerebrospinal
correlation with histopathology. Stroke 2000; 31: uid and serum: markers of cell damage in human
946954. central nervous system. Stroke 1987; 18: 911918.
Perfusion-weighted MRI in stroke

William A. Copen1 and A. Gregory Sorensen2

Massachusetts General Hospital Department of Radiology, Boston, USA
Massachusetts General Hospital NMR Center, Charlestown, USA

Introduction sented. Finally, the clinical interpretation of perfu-

sion-weighted images is discussed.
Perfusion is the circulation of blood through living
tissue via a capillary bed that permits transport of
oxygen, nutrients and other substances to and from Contrast agents in PWI
the bloodstream. Correspondingly, perfusion-
weighted magnetic resonance imaging (PWI) Any cross-sectional imaging technique that studies
encompasses a set of techniques that create images brain perfusion must detect the presence of blood
depicting hemodynamics at the microvascular in vessels much smaller than the tissue voxels that
level. This is in contrast to vascular imaging studies correspond to individual image pixels. In PWI, as
such as magnetic resonance angiography that show with non-MR-based techniques relying on X-ray
changes occurring in larger arteries and veins. The computed tomography or nuclear imaging, this is
pathological event initiating ischemic stroke often accomplished by measuring local concentrations of
originates in such larger vessels. However, infarc- a contrast agent, or tracer, that is dissolved in the
tion is directly caused by an impairment of perfu- blood. Performing PWI with MRI rather than some
sion, that is, an inadequacy of circulation through other imaging modality oers the theoretical
the capillary bed of aected brain tissue. Therefore, advantage that the contrast agent can be either an
PWI oers the opportunity to study the pathophy- endogenous contrast agent that is naturally present
siological events that lead most directly to ischemic in the blood, or an exogenous one that is injected
damage. In some cases, these events are largely or for the purpose of obtaining images.
completely undetected by techniques that study Two approaches to using endogenous contrast
only larger vessels. Furthermore, when brain tissue agents have been successful experimentally, though
is threatened but not irreversibly damaged by they are not yet optimized for clinical imaging in
impaired perfusion, the cerebral vasculature exhib- the setting of stroke. In the rst, the contrast agent
its characteristic and identiable responses that is hemoglobin. Deoxyhemoglobin is a paramag-
can be identied by PWI. For this reason, PWI may netic compound, and therefore has eects on MR
play an important role in guiding therapeutic images similar to those of other paramagnetic con-
rescue of tissue that is threatened by ischemia. trast agents such as gadolinium. This observation is
This chapter rst reviews the techniques the basis of techniques such as BOLD (blood
employed in PWI, beginning with the role of con- oxygen level dependent) imaging,13 which has
trast agents, and the MR pulse sequences that are been used successfully in numerous experimental
usually chosen. The postprocessing algorithms that studies of brain activation. These changes do reect
are used to convert raw data into clinically inter- brain perfusion to some degree. Unfortunately, the
pretable perfusion-weighted images are then pre- signal changes produced by deoxyhemoglobin are

148 William A. Copen and A. Gregory Sorensen

small, in the order of 1% at usual eld strengths, always administered in one of several available che-
and they are dicult to produce uniformly in the lated forms. After intravenous injection, any of the
brain, and may depend not just on brain perfusion gadolinium chelates rapidly equilibrate throughout
but brain utilization of oxygen; this is an area of the extracellular space in most organs of the body.
active investigation. Consequently, techniques However, the chelate is too large a complex to pass
using hemoglobin as an intrinsic contrast agent are through the intact bloodbrain barrier. Therefore,
not usually used clinically. the contrast agent normally enters the brain paren-
A second approach utilizes the protons within chyma via small arteries several seconds after injec-
owing blood as an endogenous contrast agent. In tion, passes through the capillary bed, and then
this method, called arterial spin labelling (ASL), an exits through small veins within a matter of
additional magnetic pulse, sometimes from a seconds. Very little gadolinium remains within
second radiofrequency coil, is performed in the parenchymal regions with an intact bloodbrain
region of an artery supplying the brain, such as an barrier.
internal carotid artery. This additional pulse is used Conventional contrast-enhanced brain MRI
to excite or tag in some way the protons (spins) in exploits this fact in order to highlight regions in
the blood as they ow by. These excited, or which the bloodbrain barrier is absent or
labelled spins then serve as a contrast agent as damaged. This is accomplished by acquiring T1-
they enter the brains capillaries, enabling genera- weighted images several minutes after intravenous
tion of perfusion-weighted images.46 ASL shows injection of a gadolinium-based contrast agent, at a
great promise as a PWI technique of the future, and time when the agent has been washed out of
oers several potential advantages. These include normal brain tissue. The gadolinium persists in the
lesser expense, lack of side eects from exposure to interstitial spaces of regions without an intact
exogenous contrast agents, the potential for abso- bloodbrain barrier. When water molecules whose
lute measurement of cerebral blood ow, and the protons have been excited during an MRI pulse
ability to label easily only those spins passing sequence pass very close to the gadolinium ion,
through a particular artery, so that the perfusion they are able to transfer some of their excess energy
territory of a particular artery can be demarcated. to the gadolinium ion, in a process alternatively
Unfortunately, with currently available technology, known as spin-lattice relaxation, longitudinal relax-
the signal-to-noise resolution of ASL techniques is ation, or T1 relaxation. The relaxation process
much lower than that of methods using exogenous occurs spontaneously in any substance whose
contrast agents. This means that ASL requires protons have been excited, whether or not an exo-
image acquisition times that are unacceptably long genous contrast agent is present. The process is
in the clinical setting in order to produce perfusion- accelerated, however, in the presence of gadolin-
weighted images of comparable quality. ium, because transfer of energy from excited spins
Furthermore, these techniques may have artefacts is more ecient when the energy acceptor is one
and shortcomings particularly in patients with dis- with a large magnetic moment such as gadolinium.
turbed cerebral perfusion that have not yet been This relaxivity eect of gadolinium results in short-
fully characterized. ened T1, and increased signal intensity on T1-
Because of the above limitations, PWI techniques weighted images.
now used for imaging stroke patients generally rely Though PWI uses the same contrast agents as
on exogenously administered contrast agents. All conventional contrast-enhanced MRI techniques,
currently approved MRI contrast agents contain the PWI relies on two fundamentally dierent princi-
gadolinium ion, which has a particularly high mag- ples. First, in PWI, the contrast agent is injected as a
netic moment because of its seven unpaired 4f sub- rapid bolus, and sequential images are acquired
orbital electrons. It is this property that makes during the intravascular transit of that bolus, rather
gadolinium a useful MRI contrast agent. than some time later. In this way, local concentra-
The free gadolinium ion is toxic, and is therefore tions of the agent can be measured as they rise and
Perfusion-weighted MRI in stroke 149

fall rapidly within regions of brain tissue whose Gadoliniums magnetic susceptibility eect,
bloodbrain barrier is intact. Second, PWI tech- however, operates over much longer ranges. If present
niques generally rely upon gadoliniums suscepti- in large concentration, such as during its rst pass
bility eect on MR images, rather than the through the cerebral vasculature, gadolinium within a
relaxivity eect discussed above. small blood vessel accelerates spin dephasing within
The susceptibility eect occurs because gadolin- a radius approximately equal to the diameter of the
ium ions and other paramagnetic substances are vessel. This means that, when PWI is accomplished
more magnetically susceptible than diamagnetic using T2- or T2*-weighted pulse sequences that are
substances, meaning that the former behave as sensitive to gadoliniums susceptibility eect, all of
small magnets in the presence of an externally the spins in each tissue voxel are potentially aected
applied magnetic eld, causing non-uniform by the presence of the contrast agent. Consequently,
microscopic variations in that eld. These varia- larger changes in signal intensity are produced, more
tions alter the precession frequencies of nearby precise measurements of gadolinium concentration
excited magnetic spins, so that they lose their phase are possible, and more accurate perfusion-weighted
coherence more quickly. This is manifested as loss images can be generated.
of signal intensity in T2- or T2*-weighted images.
The same susceptibility eect is responsible for loss
of signal intensity in the vicinity of hemorrhagic Pulse sequences
blood breakdown products, metallic hardware such
as aneurysm clips, and other highly susceptible For the reasons discussed above, PWI is accom-
substances within the brain. In PWI, loss of signal plished most eectively using T2 or T2*-weighted
in T2 or T2*-weighted images due to magnetic sus- pulse sequences, which are sensitive to the spin
ceptibility is exploited in order to measure the local dephasing caused by gadoliniums magnetic sus-
concentration of gadolinium dynamically during ceptibility. Elsewhere in clinical practice, T2*-
intravascular transit. weighted gradient echo pulse sequences are often
PWI techniques are usually designed to rely on employed when susceptibility eects are of para-
gadoliniums susceptibility eect rather than its mount interest, such as when evaluating for the
relaxivity eect, chiey because these eects are presence of blood breakdown products remaining
exhibited over dierent ranges. The relaxivity eect from previous hemorrhage. Gradient echo pulse
requires that an excited hydrogen nucleus sequences can be used in PWI, and demonstrate
approach a gadolinium ion very closely, within excellent sensitivity to gadoliniums susceptibility
several angstroms, in order for transfer of energy to eect. However, clinical experience has shown that
occur. The magnitude of the relaxivity eect falls o spin echo sequences may be more useful, despite
with the sixth power of the distance between the such sequences smaller sensitivity to magnetic
two species. Although signal in MR images comes susceptibility. This is because signal changes in spin
mostly from hydrogen nuclei within water mole- echo images are more dependent on gadolinium
cules, which are distributed among all tissue com- concentrations within the capillary bed and other
partments, the intact bloodbrain barrier limits microscopic vessels, whereas gradient echo images
gadolinium to the intravascular compartment, assign relatively more weighting to larger vessels.
which occupies only about 4% of tissue volume in Tissue metabolism is most directly dependent on
grey matter, and 2% in white matter. Therefore, if capillary-level hemodynamics, so spin echo images
PWI were to use T1-weighted images, and measure are more suited to depicting the conditions that
gadolinium concentration based on relaxivity- determine tissue viability. Furthermore, perfusion-
related increases in signal intensity, at most about weighted images based on gradient echo pulse
4% of excited spins would be aected, and the con- sequences are so dominated by large vessels that
trast agent would produce only small changes in subtle but important dierences in parenchymal
signal intensity. perfusion can be dicult to appreciate.
150 William A. Copen and A. Gregory Sorensen

Table 12.1. Sample imaging parameters for perfusion-weighted based contrast agent. Note that the image darkens
imaging. as the contrast agent enters the cerebral circulation,
then lightens again as the agent is washed out of
TR: 1500 ms.
the brain. Qualitatively speaking, the degree to
TE: 70 ms
which each part of the brain darkens reects the
Field of view: 20 cm20 cm
Matrix size: 128128
regional perfusion of that part of the brain. The
Slice thickness: 5 or 6 mm next section describes the computational methods
Interslice gap: 1 mm or interleaved that are used to transform raw images such as those
Number of sequential images for each slice: 45 in Fig. 12.1 into perfusion-weighted images that
Number of slices: 11 convey quantitative information about brain perfu-
Total imaging time: 68 seconds (45 images 1.5 seconds) sion.
Timing of bolus injection: 10 seconds after beginning of
Gadolinium dose: 0.2 mmol/kg Image postprocessing
Rate of injection: 5 ml/s
Saline ush: 25 ml
In conventional MR images, each pixels brightness
Rate of injection of saline ush: 5 ml/s
reects the magnitude of the radiofrequency signal
Size of intravenous catheter: ideally 18 gauge or larger.
that was emitted by a corresponding voxel of tissue.
Notes: In PWI, these signal-intensity-based images must
These parameters have been used extensively with a undergo an additional processing step, in which
1.5 tesla General Electric scanner, with an echo planar one of several computational methods is employed
imaging modication by Advanced NMR Systems. in order to produce images that depict one of a
variety of perfusion-related parameters.
The simplest of these parameters to measure is
regional cerebral blood volume, or rCBV.
Table 12.1 lists sample image acquisition param- Consider, for example, the images marked 0 s
eters that have been used successfully in spin echo and 9 s in Fig. 12.1. The 0 s image was acquired
pulse sequences used for PWI. Note that, because just as the bolus of contrast agent reached the
spin echo sequences are less sensitive than gradient brain. Because of the pulse sequence used, it is a
echo sequences in detecting the magnetic eld predominantly T2*-weighted image. The 9 s
inhomogeneities that result from magnetic suscep- image, however, was acquired while the contrast
tibility, the usual dose of gadolinium required is agent was present in the cerebral vasculature. It is,
0.2 mmol per kg of body weight, double that used in therefore, both a T2-weighted and rCBV-weighted
conventional contrast-enhanced imaging. image. This is because, relative to the original 0 s
Successful implementation of PWI requires that image, each voxels signal intensity in the 9 s
sequential images be acquired rapidly, with a tem- image has fallen by an amount related to the
poral interval of less than 2 seconds, in order to quantity of gadolinium within that voxel, due to
sample adequately the eects of the contrast agent accelerated T2 relaxation. The change in the rate of
as it passes through the vasculature. This can be relaxation, R2, is linearly proportional to the
accomplished by almost any medium-eld scanner quantity of gadolinium within the voxel, and
in clinical use today. However, multislice PWI therefore to the volume of blood in that voxel. So,
requires the use of fast imaging techniques such as for each voxel,
echo planar imaging, which can be implemented
rCBV kR2
only on scanners with high performance gradient
systems. where k is a constant, and R2 is the change in the
Figure 12.1 shows a sequence of images acquired rate of T2 relaxation. The change in signal intensity
during the intravascular transit of a gadolinium- that results from T2 relaxation is an exponential
Perfusion-weighted MRI in stroke 151

Fig. 12.1. Sequential axial T2*-weighted images acquired during intravascular transit of a bolus of gadolinium. The smaller images
were acquired every 1.5 seconds, and were used to generate the larger map of rCBF at lower left. The graph at lower right shows
how mean pixel signal intensity (relative to baseline pre-gadolinium signal intensity) varied with time in regions of interest placed
in the left (L) and right (R) hemispheres. The locations of these regions of interest are shown on the rCBF map at lower left.

process, so the above equation can be converted intensity at time t and S0 is the signal intensity that
into one incorporating signal intensities as follows: was present at time zero, before the arrival of the
contrast agent. Note that only relative rCBV can be
rCBVt k ln

calculated, because the value of the constant k
depends on numerous factors that are dicult to
where rCBVt is the volume of blood within a given monitor, including, for example, the delivery of gad-
voxel at time t, k is a new constant, St is the signal olinium to the brain and its subsequent clearance.
152 William A. Copen and A. Gregory Sorensen

The above equation allows calculation of rCBV in high serum concentrations are dicult to achieve
each voxel, provided that at least two images are for a prolonged length of time, because gadolin-
acquired: a baseline image acquired before the ium-based contrast agents are rapidly cleared from
arrival of the contrast agent, and a second image the blood in organs other than the brain, where
acquired while the contrast agent is in the cerebral they quickly exit into the interstitial space. About
vasculature. Nevertheless, rCBV is usually calcu- 20% of an injected dose of contrast agent can be
lated using far more than two images. One reason lost in its rst pass through the lungs alone. Slow
for this is that acquiring multiple images allows infusion would preclude attainment of adequate
more precise estimation of rCBV, simply because gadolinium concentrations within the cerebral vas-
more precise estimations of any quantity can be culature.
obtained by averaging repeated measurements. Rapid bolus injection of gadolinium is also pref-
Another reason is that the contrast agent is usually erable because this allows computation of perfu-
injected as a rapid bolus, rather than a steady-state sion parameters other than rCBV. Two such
infusion (see below). This means that the agent is parameters can be intuitively appreciated by visual
not homogeneously dissolved in the blood pool. inspection of the two signal-intensity-vs.-time
Depending on dierences in regional hemodynam- curves in the lower right portion of Fig. 12.1. It is
ics, the bolus may arrive in dierent parts of the clear that signal intensity falls more gradually
brain at slightly dierent times. If only one post- (meaning that gadolinium concentration rises
contrast image were acquired, erroneously higher more gradually) in the right hemisphere, which is
rCBV would be assigned to regions through which undergoing infarction, than in the left hemisphere,
the contrast bolus happened to be arriving at the which is normal. Physiologically, this is because
time of image acquisition. Using multiple images blood ow to the right hemisphere is abnormally
helps to minimize this eect by averaging values low. It is also clear that gadolinium remains in the
obtained during the entire passage of the bolus right hemisphere for a longer time than in the left
through every voxel. hemisphere. Physiologically, this is because the
Customarily, approximately ten to twenty pre- mean transit time of any particular small volume of
contrast images are acquired before the arrival of blood through the brains vasculature is longer in
the contrast agent in the brain, and averaged to the right hemisphere. Regional cerebral blood ow
obtain each pixels baseline S0 value. Subsequently, (rCBF) and mean transit time (rMTT) are perfusion
approximately 20 to 40 more images are acquired parameters that, along with rCBV, may be useful in
while the contrast agent passes through the brain. guiding stroke therapy. Like rCBV, they can be com-
For each pixel, rCBV is calculated in each image, puted from concentration-vs.-time curves like
and the average rCBV value over time is assigned those in Fig. 12.1. Unlike rCBV, however, their com-
to each pixel to create a calculated map of rCBV putation is mathematically dependent upon bolus
in each part of the brain (see Figs. 12.3, 12.4, and rather than steady-state injection of a contrast
12.5). agent. Computation of rCBF and rMTT also is more
As mentioned above, the contrast agent in PWI is mathematically complex than computation of
usually injected as a single rapid bolus, rather than rCBV.
as a constant infusion. While it may seem that Before proceeding to the calculations necessary
either injection method would be acceptable, there to compute rCBF and rMTT, it will be worthwhile to
are two reasons why rapid bolus injection is prefer- consider the qualitative examples in Fig. 12.2,
able. First, in order for gadoliniums T2 relaxivity which illustrates conceptually some of the dicul-
eects to extend outside of the blood vessels in ties of determining rCBF using PWI. The top row of
which it is contained, it must be present in fairly Fig. 12.2 (part (a)) of Fig. 12.2 depicts an idealized
high concentration. At low concentrations, gadolin- PWI examination, in which the entire quantity of
iums T1 eects predominate instead. Suciently gadolinium that enters a voxel of brain tissue
Perfusion-weighted MRI in stroke 153



Fig. 12.2. Ideal and actual arterial input function and tissue concentration curves. In these four curves, time is measured on the x-
axis, and gadolinium concentration on the y-axis. The two curves on the left are arterial input functions, depicting the delivery of
a bolus of gadolinium to a hypothetical tissue voxel as a function of time. The two curves on the right depict the concentration of
gadolinium in that voxel as a function of time. Ideally, as in (a) the entire gadolinium bolus is delivered during a single instant in
time. In this case, tissue concentration immediately reaches its maximum, and rCBF in each voxel is proportional to the initial
maximum value attained in that voxel. (b) depicts a more realistic situation, in which the bolus of gadolinium is delivered over a
brief but not instantaneous period of time. In this case, tissue concentration rises and falls more gradually, and rCBF must be cal-
culated by deconvolution (see text).

arrives at the same instant in time, as shown by the In reality, it is impossible to achieve instantane-
arterial input function, AIF(t), on the left. In this ous arrival of the entire gadolinium bolus. Even
idealized case, the concentration function, C(t), mechanical power injectors are limited in the rate
which describes the concentration of gadolinium in at which they can deliver contrast agents into the
the voxel as a function of time, reaches its bloodstream. Furthermore, for practical reasons,
maximum immediately, and then falls gradually as the gadolinium must be injected into a vein, rather
the gadolinium bolus leaves the voxel via the than directly into an artery supplying the brain. As
venous system. In this case, computation of rCBF is the bolus travels through the venous system, the
straightforward and maximally accurate; it is pro- right heart, the pulmonary vasculature, and nally
portional to the gadolinium concentration at the the arteries supplying the brain, it is dispersed in
instant when the bolus arrives. the blood, so that the arrival of gadolinium in a
154 William A. Copen and A. Gregory Sorensen

brain voxel is dispersed in time. This dispersion artery rather than microvascular ow, i.e. early
is amplied by slower injection of the contrast arrival of the tracer, large signal drop during tracer
agent bolus, or in patients whose cardiac output is passage, and rapid return to baseline signal inten-
low. sity. After several such pixels have been chosen,
The bottom row of Fig. 12.2 (part (b)) depicts this their signal intensities at each time point are aver-
noninstantaneous arrival of the tracer bolus that aged, and the resulting signal-intensity-vs.-time
occurs in reality, and the eect that it has on C(t). In curve is converted to a concentration-vs.-time
this case, computation of rCBF in this voxel is no curve, which is used as AIF(t). The automated
longer straightforward. This is because C(t) is now deconvolution process then takes several more
mathematically dependent not only on rCBF, but minutes, depending on available computing
also on two other functions: AIF(t), and R(t), a resources, to generate maps of rCBF. After rCBF
residue function that describes the proportion of has been calculated for each image pixel, calcula-
tracer remaining in the voxel as a function of time. tion of rMTT is straightforward, because rCBV,
The residue function is not shown explicitly in Fig. rCBF, and rMTT are related by the central volume
12.2, but has exactly the same shape as the concen- theorem:
tration function in part A. Mathematically, it can be
said that C(t) is the product of rCBF and the convo- rMTT
lution of AIF(t) and R(t). In order to calculate rCBF,
R(t) must be determined, using a somewhat Because deconvolution is computationally inten-
complex process called deconvolution. A number sive and requires calculation of an arterial input
of algorithms for deconvolution exist. We have used function, some centres have used more mathemati-
singular value decomposition, which is relatively cally simple approximations of rCBF and rMTT. For
robust, in that it works well in a variety of situations example, the slope of the increase in gadolinium
in which perfusion is altered.7,8 concentration after bolus arrival can be used as an
All deconvolution methods are similar in that approximation of rCBF. The width of the concentra-
they require that AIF(t) be supplied along with tion curve when it reaches half of its maximum
C(t). Generally, one AIF to be used for all image value can be used as an approximation of rMTT.
pixels is obtained by selecting manually several While such approximations lack the mathematical
pixels that lie adjacent to a major cerebral artery, rigor of quantities determined by deconvolution,
and therefore can be used to sample the concen- their calculation is more straightforward.
tration of tracer within the artery. A middle cere-
bral artery is typically used, because much of its
course is usually visualized within a single axial Interpretation of perfusion-weighted images
image section, and this makes selection of adja-
cent pixels easier. Note that pixels that lie within When rCBV, rCBF, and rMTT are evaluated together,
the artery cannot be used, because, for technical considerable information regarding regional cere-
reasons, the relationship between signal intensity bral perfusion is available. This is because these
and gadolinium concentration is more complex quantities illuminate not only the degree and phys-
for those pixels than for pixels just outside the ical location of ischemic insult, but also the nature
artery. Pixel selection can be performed by a phy- of the cerebrovascular response.
sician, technologist, or other trained personnel in The fundamental determinant of cerebral perfu-
a few minutes, or can be done by an automated or sion is cerebral perfusion pressure, or CPP. The
semiautomated computer program. blood vessels of the brain adapt to changes in CPP
Selection of pixels to sample arterial ow is by adjusting vascular resistance so that, within a
based both on location and on signal characteris- wide range of local CPP values, rCBF is maintained
tics. The ideal pixels to choose are those that dem- close to its normal value of about 50 cm3 of blood
onstrate the characteristics of ow in a major per 100 g of brain tissue per minute in grey matter,
Perfusion-weighted MRI in stroke 155

and somewhat less in white matter. When decreases from an informed estimation of how much brain
in CPP are so great that this autoregulatory process tissue will be aected, and how severe neurological
begins to fail, rCBF falls. However, the vasculature is decits are likely to be, if thrombolytic therapy is
still able to compensate with further reductions in not initiated.
vascular recruitment and vasodilation. This tends PWI can help to identify tissue that is at risk of
to preserve or even increase rCBV. Importantly, inclusion into growing infarcts. This tissue may be
because the eective cross-sectional area of the represented by regions of diusion-perfusion mis-
microvasculature is increased, the velocity of blood match, i.e. that portion of the brain in which abnor-
ow is reduced. Consequently, MTT of blood within malities are seen on acute-stage PWI, but not on
each capillary is increased. When red blood cells DWI. Infarcts are more likely to grow substantially
spend a greater amount of time in each capillary, when the area of diusionperfusion mismatch is
brain tissue is able to extract a larger amount of large.13,1820
oxygen, possibly enough to survive for some period This observation raises the question of how infor-
of time despite impaired rCBF. In light of these mation from the various kinds of perfusion-
eects, elevations in rMTT can be seen as indica- weighted images should be integrated in order to
tions of the brains attempts to compensate for identify tissue at risk and predict lesion enlarge-
ischemic insult. Accordingly, a drop in rCBV could ment. One study found that volumes of abnormal-
signify exhaustion of the vasculatures autoregula- ities on initial rCBV images provide the best
tory reserve. estimate of nal lesion volumes, though slightly
Understanding of cerebrovascular hemodynam- underestimating those volumes. That study found
ics suggests several potential roles for PWI in caring that abnormalities on initial rCBF or rMTT images
for stroke patients. PWI is of greatest clinical inter- are often much larger, and may overestimate the
est in acute stroke, when timely therapeutic reper- amount of tissue that will be involved.18 Another
fusion may result in the preservation of study also found that rCBF abnormalities tended to
ischemically threatened tissue that otherwise be larger than nal infarct volumes, but found that,
would undergo infarction. In this setting, PWI is among patients with a large region of diusion
most often performed along with diusion- perfusion mismatch, abnormalities on rCBF maps
weighted magnetic resonance imaging, or DWI. provided the best single estimate of nal infarct
DWI is much more accurate than other imaging progression.20
modalities in identifying damaged tissue very early These studies are consistent with the view that
after stroke onset.911 However, infarcts often reduced rCBF represents ischemic threat to brain
enlarge over time to encompass more tissue than tissue, increased rMTT represents the brains active
initially appears abnormal on DWI.1218 response to that threat, and reduced rCBV may rep-
Now that thrombolytic therapy is commonly resent failure of that response to meet metabolic
employed in acute stroke management, predicting needs. In this view, regions of reduced rCBV repre-
the extent to which acute infarcts will grow in the sent tissue that is very likely destined for infarction,
days following presentation is increasingly impor- but regions of abnormal rCBF or rMTT may or may
tant. Intravenous or intra-arterial thrombolysis is not proceed to infarction (see Fig. 12.3 and Fig.
capable of saving substantial quantities of brain 12.4). There have been eorts to develop models
tissue that otherwise would have undergone infarc- that combine all three of these parameters, along
tion, resulting in marked reduction of permanent with other imaging measurements, in order to
neurologic decits. However, thrombolysis also predict the probability of infarction of any given
carries considerable risk of catastrophic intracra- tissue voxel.21
nial hemorrhage. The decision of whether or not to Besides informing the decision of whether or
accept this risk and proceed with thrombolysis is a not to proceed to thrombolysis by identifying
highly individualized one. In making that decision, tissue that is at risk, PWI has also been used to
clinicians, patients and their families would benet determine whether reperfusion has occurred,
156 William A. Copen and A. Gregory Sorensen

Fig. 12.3. Extension of an acute infarct into an area of abnormal perfusion. In this patient, acute-stage diusion-weighted imaging
(DWI) showed a small, subtle ischemic lesion in the left centrum semiovale. A map of rCBV, acquired during the same examina-
tion, showed a matching small, subtle region of decreased rCBV. However, the other two PWI maps showed a much larger region
of decreased rCBF and increased rMTT. Abnormal regions in each image are marked by arrows. A follow-up DWI image obtained
one day later shows that the infarct had grown to encompass much of the ischemically threatened tissue identied in the rCBF
and rMTT maps.

either spontaneously or as a result of thrombolytic that already is contained in currently available per-
therapy.22,23 Figure 12.5 provides an example of an fusion-weighted images.
acute stroke patient in whom spontaneous reper-
fusion occurred prior to imaging. In this patient,
thrombolytic therapy would confer no potential Conclusions
benet, while conferring a signicant risk of hem-
orrhage. PWI describes a set of techniques in which sequential
As clinical experience with PWI in acute stroke magnetic resonance images of the brain are acquired
accrues, it is likely that more precisely formulated during intravascular transit of a bolus of a gadolin-
guidelines will emerge for interpreting information ium-based exogenous contrast agent, in order to
Perfusion-weighted MRI in stroke 157

Fig. 12.4. Failure of an acute infarct to extend into an area of abnormal perfusion. In this patient, acute-stage DWI showed a small
ischemic lesion involving the left corona radiata. PWI showed larger regions of abnormal rCBV, rCBF, and rMTT. Abnormal regions
in each image are marked by arrows. A follow-up CT image obtained 2 days later shows that the infarct had not grown signicantly
into the region of acutely impaired perfusion.

produce images that depict hemodynamics at the ing thrombolytic therapy. Tissue that demonstrates
microvascular level. Although PWI can be accom- abnormal perfusion, without abnormal appearance
plished with most scanners in clinical use today, on acute-stage diusion-weighted imaging, may be
multislice PWI requires fast imaging pulse sequences tissue that is at risk of inclusion into a growing infarct,
such as echo planar imaging. The raw images and therefore tissue that may be rescued by timely
obtained during intravascular transit of the contrast therapeutic reperfusion. PWI can also identify cases in
agent are used to generate images depicting regional which spontaneous or therapeutic reperfusion of
variations in any of several perfusion-related parame- ischemic areas has occurred, so that further thrombo-
ters, which may include cerebral blood volume, cere- lytic therapy would be dangerous and without poten-
bral blood ow and mean transit time. These are most tial benet. Additional clinical experience with PWI
useful in acute stroke, when information about will enable more precise and detailed denition of the
regional perfusion may help to guide decisions regard- role of PWI in stroke management.
158 William A. Copen and A. Gregory Sorensen

5 Williams DS, Detre JA, Leigh JS, Koretsky AP. Magnetic

resonance imaging of perfusion using spin inversion
DWI rCBV of arterial water. [erratum appears in Proc Natl Acad
Sci USA 1992 May 1; 89(9): 4220]. Proc Natl Acad Sci
USA 1992; 89: 212216.
6 Edelman RR, Siewert B, Darby DG et al. Qualitative
mapping of cerebral blood ow and functional local-
ization with echo-planar MR imaging and signal tar-
geting with alternating radio frequency. Radiology
1994; 192: 513520.
7 stergaard L, Weissko RM, Chesler DA, Gyldensted
C, Rosen BR. High resolution measurement of cere-
bral blood ow using intravascular tracer bolus pas-
sages. Part I: Mathematical approach and statistical
rCBF rMTT analysis. Magn Reson Med 1996; 36: 715725.
8 stergaard L, Sorensen AG, Kwong KK, Weissko RM,
Gyldensted C, Rosen BR. High resolution measurement
of cerebral blood ow using intravascular tracer bolus
passages. Part II: Experimental comparison and prelim-
inary results. Magn Reson Med 1996; 36: 726736.
9 Lvblad KO, Laubach HJ, Baird AE et al. Clinical expe-
rience with diusion-weighted MR in patients with
Fig. 12.5. Spontaneous reperfusion. In this patient, acute- acute stroke. Am J Neuroradiol 1998; 19: 10611066.
stage DWI showed an ischemic lesion in the left occipital lobe 10 Singer MB, Chong J, Lu D, Schonewille WJ, Tuhrim S,
(arrow). PWI showed that rCBV and rCBF were not decreased, Atlas SW. Diusion-weighted MRI in acute subcortical
but rather elevated (arrows) in this region, signifying the infarction. Stroke 1998; 29: 133136.
characteristic hyperemia or luxury perfusion that often 11 Gonzalez RG, Schaefer PW, Buonanno FS et al.
follows spontaneous reperfusion of ischemic brain tissue. Diusion-weighted MR imaging: diagnostic accuracy
There was no denite rMTT abnormality. In this patient, in patients imaged within 6 hours of stroke symptom
thrombolytic therapy would pose a signicant risk of hemor- onset. Radiology 1999; 210: 155162.
rhage, probably without potential gain. 12 Sorensen AG, Buonanno FS, Gonzalez RG et al.
Hyperacute stroke: evaluation with combined multi-
section diusion-weighted and hemodynamically
REFERENCES weighted echo-planar MR imaging. Radiology 1996;
199: 391401.
1 Ogawa S, Lee TM. Magnetic resonance imaging of 13 Baird AE, Beneld A, Schlaug G et al. Enlargement of
blood vessels at high elds: in vivo and in vitro meas- human cerebral ischemic lesion volumes measured by
urements and image simulation. Magn Reson Med diusion-weighted magnetic resonance imaging. Ann
1990; 16: 918. Neurol 1997; 41: 581589.
2 Ogawa S, Lee TM, Kay AR, Tank DW. Brain magnetic 14 Barber PA, Darby DG, Desmond PM et al. Prediction
resonance imaging with contrast dependent on blood of stroke outcome with echoplanar perfusion- and
oxygenation. Proc Natl Acad Sci USA 1990; 87: diusion-weighted MRI. Neurology 1998; 51: 418426.
98689872. 15 Schwamm LH, Koroshetz WJ, Sorensen AG et al. Time
3 Ogawa S, Lee TM, Nayak AS, Glynn P. Oxygenation- course of lesion development in patients with acute
sensitive contrast in magnetic resonance image of stroke: serial diusion- and hemodynamic-weighted
rodent brain at high magnetic elds. Magn Reson Med magnetic resonance imaging. Stroke 1998; 29:
1990; 14: 6878. 22682276.
4 Detre JA, Subramanian VH, Mitchell MD et al. 16 van Everdingen KJ, van der Grond J, Kappelle LJ,
Measurement of regional cerebral blood ow in cat Ramos LM, Mali WP. Diusion-weighted magnetic
brain using intracarotid 2H2O and 2H NMR imaging. resonance imaging in acute stroke. Stroke 1998; 29:
Magn Reson Med 1990; 14: 389395. 17831790.
Perfusion-weighted MRI in stroke 159

17 Rordorf G, Koroshetz WJ, Copen WA et al. Regional relative cerebral blood ow most accurately identies
ischemia and ischemic injury in patients with acute tissue destined to infarct. Stroke 2001; 32: 15811587.
middle cerebral artery stroke as dened by early 21 Wu O, Koroshetz WJ, Ostergaard L et al. Predicting
diusion-weighted and perfusion-weighted MRI. tissue outcome in acute human cerebral ischemia
Stroke 1998; 29: 939943. using combined diusion- and perfusion-weighted
18 Sorensen AG, Copen WA, stergaard L et al. MR imaging. Stroke 2001; 32: 933942.
Hyperacute stroke: simultaneous measurement of rel- 22 Marks MP, Tong DC, Beaulieu C, Albers GW, de
ative cerebral blood volume, relative cerebral blood Crespigny A, Moseley ME. Evaluation of early reperfu-
ow, and mean tissue transit time. Radiology. 1999; sion and i.v. tPA therapy using diusion- and perfu-
210: 519527. sion-weighted MRI [see comments]. Neurology 1999;
19 Neumann-Haefelin T, Wittsack HJ, Wenserski F et al. 52: 17921798.
Diusion- and perfusion-weighted MRI. The 23 Kidwell CS, Saver JL, Mattiello J et al. Thrombolytic
DWI/PWI mismatch region in acute stroke. Stroke reversal of acute human cerebral ischemic injury
1999; 30: 15911597. shown by diusion/perfusion magnetic resonance
20 Parsons MW, Yang Q, Barber PA et al. Perfusion mag- imaging. Ann Neurol 2000; 47: 462469.
netic resonance imaging maps in hyperacute stroke:
Perfusion imaging with arterial spin labelling

David C. Alsop1 and John A. Detre2

Department of Radiology,
Beth Israel Deaconess Medical Center and Harvard Medical School, USA
Departments of Neurology and Radiology
University of Pennsylvania Medical Center, USA

Introduction tions such as delayed ow or collateral ow through

multiple paths.7 Third, measurements of absolute
Unlike diusion imaging, which is unique to MRI, blood ow and volume are not readily possible.
imaging of cerebral blood ow is possible with a Finally, motion during the bolus is a frequent
number of tomographic techniques including PET,1 problem because of poor patient cooperation and
SPECT,2 and X-ray CT. 3,4 Relative to these other discomfort from the injection. Motion is an inher-
approaches, blood ow imaging with MRI can be ent problem with most imaging techniques, but
advantageous because of its speed, spatial resolu- those that depend upon changes in a larger back-
tion, lack of ionizing radiation, and MRIs ability to ground signal intensity are most prone to error.
obtain other, spatially registered images such as An alternative technique for the measurement of
diusion, MR angiography, or even MR spectros- perfusion MRI known as arterial spin labelling
copy within the same study. (ASL) shows promise for addressing many of the
Imaging of hemodynamic parameters such as limitations of bolus contrast for perfusion assess-
time to peak, mean transit time, relative blood ow, ment.8,9 This approach employs electromagnetic
and relative blood volume is possible with the bolus elds to non-invasively label the nuclear spins of
injection of a contrast agent.5,6 This dynamic sus- water in the inowing arterial blood. These labelled
ceptibility contrast (DSC) approach yields a large spins then enter the tissue where the labelling
signal change, which permits the sensitive charac- causes an attenuation of signal proportional to per-
terization of the passage of the bolus through the fusion. While the signal change resulting from
vasculature. It is particularly good at delineating labelling is much smaller than with bolus contrast
very delayed or extended transit of the bolus injection, ASL oers a exibility of acquisition and
through a region of tissue and has thus been widely quantication strategies that can be advantageous
applied to the study of stroke in animals and in numerous applications.
While clearly a powerful technique, hemody-
namic MRI with contrast injection is not without Arterial spin labelling methods
weaknesses. First, it requires a well-timed i.v. injec-
tion of a contrast agent. The injection increases the The spatially selective radiofrequency (RF) excita-
cost of the study signicantly, requires good venous tion and inversion capabilities of magnetic reso-
access, preferably a power injector, and limits the nance make possible a totally non-invasive method
short term repeatability of the study. Second, for perfusion imaging, which is known as arterial
because it is an intravascular tracer, quantication spin labelling. The basic principle of ASL perfusion
of perfusion can be confused by pathologic condi- imaging is simple, (see Fig. 13.1). If blood ows

162 David C. Alsop and John A. Detre

Fig. 13.1. The arterial spin labelling technique. Normally positive nuclear spins ow from feeding arteries into the slice of tissue
being imaged, left. Spatially selective inversion can be applied inferior to the slice to be imaged such that the sign of the spins in
the feeding arteries is changed. This negative signal ows into the imaged slice and decreases the signal intensity.

through a region of tissue, then it must be supplied The small size of the ASL eect means that it is
by an artery which comes from outside of the virtually impossible to obtain an ASL measurement
region. Any spatially selective alteration of signal without rst subtracting out the unwanted, perfu-
outside of the region will change the spins in the sion unrelated background signal. This is typically
feeding artery. If time is allowed for these changed achieved by acquiring one image with some pertur-
spins to enter the tissue, then there will be a change bation of the inowing arterial spins, the labelled
in the spins within the tissue. image, and a second image with no perturbation of
ASL is a competition between T1 decay and the arterial spins, the control image. Subtraction of
inow of new spins. With f (the ow or perfusion) the two images yields a perfusion sensitive image
being the rate of inow and 1/T1 being the rate of (Fig. 13.2) As with DSC imaging, there would be no
decay, the fractional change in the image intensity signal in this subtraction image without perfusion
due to ASL tends to be approximately f T1. For a but quantication of perfusion requires further
T1 of 1 second and a ow of 60 ml/100 g/min, equal analysis
to 1 ml/100 g per s(0.01/s), the ratio is 1%.
Typically, ASL signal changes in human grey matter
Labelling strategies for ASL
are of this order and white matter ow is consider-
ably smaller. Compared to DSC imaging, the signal The simplicity of the ASL approach has been some-
change in ASL is quite small and one might initially what muddled by the introduction of a number of
consider the signal unmeasurable. However, with dierent approaches to labelling, each with its own
modern scanners equipped with fast imaging hard- acronym. The clearest distinction between labelling
ware to minimize motion related noise, images of approaches is between continuous ASL and pulsed
the perfusion-related signal can readily be meas- ASL. In continuous ASL,9 spins are continuously
ured, albeit at lower spatial resolution than typi- inverted at a specic location before they enter the
cally used for clinical MRI. tissue. Thus spins which enter the tissue rst are
Perfusion imaging with arterial spin labelling 163

Fig. 13.2. The subtraction method used for most ASL imaging. An image acquired with labelling is subtracted from an image
without labelling, the control. The resulting very small dierence, typically less than 1% of the image intensity, reects perfusion.

labelled rst and spins which enter later are Edelman et al.10 were the rst to apply pulsed
labelled later. In contrast, the pulsed labelling tech- techniques in humans. They employed a slab selec-
niques10,11 apply a single perturbing pulse to the tive inversion pulse inferior to the imaged brain
spins outside the slice and then wait as the spins region to label arterial spins and a superior inver-
enter the tissue. Because spins which enter the sion slab during the control image preparation. The
tissue later are perturbed earlier than in the contin- most distinct alternative strategy employs a non-
uous technique, there is additional T1 decay of the selective inversion pulse for the labelling and a slab
label and generally the signal change with pulsed selective inversion containing the entire imaged
techniques is smaller than with continuous tech- region for a control. This inversion technique was
niques.12,13 Nevertheless, pulsed techniques employ rst proposed by Kwong et al.11 Because any slight
more traditional RF and gradient strategies and can dierence in the ip angle of the two pulses will
produce excellent images. For simplicity, the pulsed have a big eect upon the measured signal, special
techniques will be discussed rst. RF pulses with low sensitivity to variations in RF
In pulsed techniques, an RF pulse is applied at a amplitude and very sharp slice proles must be
time approximately 1 second before imaging. employed.17
Almost always an inversion RF pulse is used Continuous ASL usually employs a special RF
because it creates the largest change in the arterial labelling scheme known as ow driven adiabatic
magnetization and, by analogy to inversion recov- inversion.18 One can approximate continuous satu-
ery, the time before imaging is referred to as TI. Two ration8 of spins as they ow past a certain plane by
images are obtained, one with the region contain- applying a series of thin saturation pulses repeat-
ing the inowing artery inverted and one where it is edly, but trying the same strategy with inversion
not inverted. The only dierences between all of the pulses is problematic; some spins may remain
pulsed labelling strategies is what inversions are within the saturation slab for two or more inver-
applied to other tissue, including the imaged sions and get doubly inverted. Because inversion
region, and what is done to insure that the dier- provides a stronger signal change in ASL, it is con-
ences between the two RF preparations do not sidered highly desirable to use a continuous inver-
cause direct eects on the image which are inde- sion technique. Flow-driven adiabatic inversion is a
pendent of perfusion. Such direct eects can result simple technique because it requires only turning
from imperfect inversion slice proles14 and dier- on a constant gradient and R; however, the concept
ences in the magnetization transfer eects of the of ow driven adiabatic inversion is foreign to most
two preparation sequences.15,16 MRI practitioners who typically use only pulsed RF.
164 David C. Alsop and John A. Detre

For the purposes of perfusion imaging, however, it can be taken from literature values21 or an attempt
is only necessary to understand that the technique can be made to measure it in vivo. Likewise, the
denes a plane such that all spins owing through tissue T1 can be measured, since it may be dierent
the plane get inverted at the time they ow than the T1 of blood. Including the eects of the
through. dierent T1s complicates the discussion,13 so for
The greatest attraction of continuous ASL is that now, assume T1 of blood and tissue are the same.
it can achieve a signal approximately two to three For pulsed experiments, the eect of the inver-
times larger than with pulsed ASL.12 To achieve this sion is to create a bolus of labelled spins that is of
large a signal change, however, the labelling must duration tb. If we wait long enough for the bolus to
be left on for many seconds. The decreased number enter the tissue, a time TI, before imaging, then the
of averages achievable per unit time with this large labelled signal will be just the perfusion times the
signal change partially osets the signal advan- duration of the bolus times a decay factor for T1
tage.19 Still, continuous ASL is generally more sensi- decay over TI.
tive than pulsed techniques and has some If we had not waited long enough, then some of
advantages in the details of implementation. The the labelled blood destined for the tissue would still
greatest disadvantages of the continuous labelling be in the feeding vessels and the perfusion would
approach are the unusual RF requirements of the be underestimated. One of the challenges of pulsed
labelling. First, many scanners are designed speci- ASL quantication is that tb is not known because
cally for brief, widely spaced pulses of RF at high the inversion pulse inverts a thickness of tissue
power. Flow-driven adiabatic inversion requires containing arteries, not a duration. Wong and col-
weaker RF for an extended period of time, so imple- laborators22 applied a saturation pulse at a certain
mentation on some scanners may be dicult. The time after the inversion to eliminate any labelled
long duration of the labelling can also lead to high blood still remaining in the inverted slab. This
RF power deposition in the subject that can makes the tb always be equal to TITsat.
approach the FDA limits. Nevertheless, continuous For continuous ASL experiments the perfusion
labelling can be performed on many mainstream quantication is slightly dierent. Since the spins
clinical scanners at power levels well within the are not all labelled at the same time we have to
FDA limits.13 average the T1 decay of the duration of the bolus.
Often the term TI is not used for continuous experi-
ments. Instead w, a postlabelling delay is dened
Quantification of perfusion with ASL
such that TItb w.13 As with the pulsed experi-
ASL perfusion measurement diers from DSC ment, ow will be underestimated unless one waits
imaging in several important ways. First, the tracer, long enough for all of the bolus to enter the tissue.
water spins, can readily diuse into tissue.20 This If the T1 of tissue and blood are signicantly
makes it impossible to measure venous blood dierent, then there will be added uncertainty in
volume with ASL but it also makes ASL insensitive the quantication of perfusion. Generally, if one
to the vessel permeability issues that aect DSC uses the T1 of blood in the above equations and T1
perfusion imaging when the bloodbrain barrier is of tissue is shorter, the perfusion will be underesti-
damaged. Another important dierence is that the mated, and if the T1 of tissue is longer, then perfu-
tracer is rapidly decaying. Finally, the tracer is sion will be overestimated.13
injected into the arteries right near the tissue, so
deconvolution of an intravenous bolus is unneces-
Motion artefacts and ASL
sary. Typically, the arterial input function is
assumed based upon the labelling applied whether While the small signal level of ASL reduces the
pulsed or continuous. signal-to-noise ratio of ASL perfusion images, sur-
Because the tracer is rapidly decaying, it is prisingly good images can be obtained if motion
important to know the decay rate, T1. T1 of blood artefacts can be reduced and eliminated. Though
Perfusion imaging with arterial spin labelling 165

Fig. 13.3. Background suppressed ASL. The background signal has been attenuated to less than the grey matter perfusion signal
such that the negative ASL contribution changes the sign of the labelled image. The dierence image, right, is much less sensitive
to motion artefact than in traditional ASL.

snapshot imaging is typically used for ASL studies, Further reduction of motion artefacts is possible
the subtraction of images acquired at dierent times by suppression of the background signal, which is
still can be sensitive to motion. Experience with normally much larger than the perfusion related
patient studies has shown, however, that good image signal change. A very successful strategy for sup-
quality can be obtained with simple strategies to pression was rst proposed by Dixon et al.25 First
reduce motion and recent developments suggest apply a highly selective 90 degree pulse to the tissue
even greater robustness to motion is possible. to be imaged. Next perform the labelling pulse.
Typically imaging of perfusion with ASL requires Later, apply one or more non-selective inversion
multiple averages to achieve an acceptable signal-to- pulses timed to null the tissue signal at the imaging
noise ratio. If the label and control images are inter- time. In the absence of inow, there will be zero
leaved, i.e. rst a control, then a label, then a control, signal from the tissue but the inowing blood signal
then a label, etc., the time dierence between the two will be the same magnitude.
sets of images will be only one TR, even if the averag- Such a strategy was employed by Ye et al.26 to
ing continues for many minutes. This rapid alterna- reduce noise from the motion of the background
tion between label and control tends to give signal in an ASL experiment by a factor of 10. We
improved robustness to motion and other sources of have found that optimization of inversion timing
noise in the scanner, such as drifts in sensitivity. In for suppression over a broad range of T1s can
normal subjects and reasonably cooperative volun- produce suppression of the background signal by a
teers, this subtraction strategy is sucient to achieve factor of greater than 100. This makes the signal
good quality images.23 Robustness to motion can be from ASL larger than the background signal and
further improved by using postprocessing methods greatly reduces concerns of motion artefact (Fig.
to reduce motion. Using one particular strategy24 we 13.3). Background suppression changes the relative
have been able to drastically reduce residual motion motion sensitivity of the two MR perfusion imaging
in ASL perfusion images of acute stroke patients and techniques. Background suppressed ASL is much
demented patients, even when subjects were speak- less sensitive to motion than DSC imaging.
ing during the scan. Background suppression can be adapted to most
166 David C. Alsop and John A. Detre

Fig. 13.4. ASL dierence images acquired with background suppression, continuous labelling, and a 3D fast spin echo imaging

ASL imaging sequences. For continuous labelling the imaging of the whole brain an unacceptably
sequences, some of the inversion pulses must be long process. Just as with traditional imaging, inter-
spatially selective with their boundary located leaving of slice acquisition is the obvious solution.
exactly at the labelling plane. Inversion of the spins This approach is challenging for several reasons.
past the labelling plane will change the sign of the First, the ASL label decays with T1. Acquiring
accumulating label so the continuous labelling images over a period of time comparable to T1 will
must be switched between label and control after cause loss of signal, and signal-to-noise ratio, in the
each inversion.27 Example continuous ASL images later images, relative to single slice acquisition.
obtained with background suppression are shown Second, it may take longer for blood to enter a thick
in Fig. 13.4. slab than to enter a thin one. This longer transit
time will necessitate longer waits before imaging
and consequently some signal decay due to T1.
Multislice ASL imaging
Finally, it is harder to control for systematic dier-
In principle, acquiring more than one slice with ences in the direct eects of the labelling RF on the
ASL is straightforward, but in practice there are a tissue signal across a thick slab.
number of complications. Since ASL benets from The rst problem requires that the dierent slices
averaging, acquiring one slice at a time would make be acquired as rapidly as possible so that T1 decay
Perfusion imaging with arterial spin labelling 167

of the label is not a concern. A reasonable number popular for multislice pulsed ASL experiments
of slices can be acquired with very fast imaging because of their sharp slice proles as well as insen-
systems, but the 20 or more slices required for sitivity to B1 variations. The FOCI pulse in particu-
whole brain coverage probably cannot be satisfac- lar can be made to have very sharp boundaries and
torily achieved this way. Imaging time per prepara- has helped a great deal with pulsed labelling
tion sequence can be reduced by using multishot studies.
versions of the single-shot sequences currently Magnetization transfer is a physical process that
being used for ASL. Alternatively, three- can cause attenuation of tissue magnetization by RF
dimensional acquisition can be used so that all of at a considerable oset in frequency from the slice.
the slices are acquired at exactly the same time after Even if the slice prole is perfect, the magnetization
labelling. Multishot and three-dimensional scans transfer eects of two dierent inversion pulses
could increase the noise in a perfusion study need not be perfectly matched. Magnetization
because motion-related phase errors can cause transfer is an especially important issue for adia-
ghosts and other artefacts. The use of background batic labelling since the RF is on for a long time. RF
suppression can help a great deal with this must be applied during the preparation period of
approach.26 the control image in order to avoid a large system-
The longer transit into a thick slab of tissue is an atic error. Single-slice studies simply moved the
unavoidable result of imaging a thicker region of labelling plane above the slice instead of below the
tissue. If the ow into the thick slab is so slow that a slice.9 This serves only as a good control for one
very long wait has to be used, then the signal-to- slice. A second, small RF coil can be used for the
noise ratio may suer dramatically. In such a case, a labelling that produces only a weak RF eld in the
series of sequential single slice scans may give imaged region and thus minimizes magnetization
comparable or superior results. Usually, however, a transfer eects28,29 but this does require a favour-
large volume of tissue is perfused by a relatively able labelling geometry and special hardware. We
large and fast feeding artery, so ow is not markedly have shown30 that sinusoidal modulation of the
delayed in transit. If there is a stenotic vessel, then amplitude of the RF pulse during the control period
delay in the large vessel could be more of a serious can mimic the magnetization transfer eects of the
issue. labelling while allowing most of the blood spins to
Finally direct labelling eects on the measured pass uninverted. Sinusoidal modulation can be
signal must be eliminated for accurate results. A thought of as splitting the inversion plane into two
number of strategies to control for direct eects of separate inversion planes that doubly invert the
RF only work perfectly when there is one, thin inowing blood. The eciency of the scheme is not
slice.9 Generally, RF can aect tissue directly in two perfect, however, and an improved control for
ways, either through imperfect slice proles14 or multislice continuous ASL would be desirable.
magnetization transfer.15 Talagala et al.31 have proposed a strategy that con-
All of the discussion above has assumed that one trols for magnetization transfer near the centre slice
could perfectly invert or saturate a square slab of better than the original superior slice strategy,9
tissue with a very fast boundary. In practice, there is making possible the imaging of a thin slab near the
a transition region between the inversion and no centre slice. We are currently evaluating a promis-
inversion area. Depending on the RF pulse used, ing alternative strategy, which involves modulation
the transition may be very broad. For a slice near of both the gradient and RF amplitudes.32
the end of an imaged slab, the transition region will MT can also be a signicant but smaller problem
pass directly through the slice and systematic errors for pulsed labelling, especially when adiabatic RF
can occur. To overcome these errors, very carefully pulses are used. Several groups have suggested
engineered inversion pulses have been used. Two strategies involving double pulses with altered
adiabatic pulses, the hyperbolic secant pulse and, properties that can help to eliminate MT systematic
more recently, the FOCI pulse17 have become errors.15,16
168 David C. Alsop and John A. Detre

tion with diusion MRI, is becoming an important

Validation of ASL perfusion MRI
part of the evaluation of acute stroke. These data
ASL methods have been validated against existing are used to conrm the diagnosis of stroke, to
methods for quantifying cerebral perfusion. CBF establish a baseline against which stroke therapies
values obtained using continuous ASL have been can be assessed, and to contribute to prognosis.
validated against microspheres in a rat model of Perfusion imaging may be particularly valuable in
middle cerebral artery occlusion33 and against the triage of patients for thrombolytic therapy,
15O-PET scanning in humans.34 where patients with normal perfusion might be
spared the risks of thrombolysis, while patients
with perfusion decits in the absence of signi-
Applications to cerebrovascular disease cant structural change might be aorded the ben-
ets of thrombolysis even outside the accepted
temporal window.
Perfusion mapping in animal models
ASL perfusion MRI provides a means of non-
One important role for ASL is in the study of animal invasively quantifying cerebral perfusion in the
models of acute stroke.3537 A large body of investi- acute setting without contrast administration and
gation has dened ischemic thresholds in animal is ideal in instances where intravenous access is
models using diusible tracers,38 and these results dicult to obtain or where serial perfusion meas-
are directly comparable to values obtained using urements are required. The use of contrast-
ASL methods. In addition, because no injection is enhancement in magnetic resonance angiography
required, ASL can be repeatedly performed to map can greatly improve the visualization of luminal
blood ow over the course of infarction and reperfu- proles, but requires no prior contrast administra-
sion treatments. Perfusion maps can easily be tion during the scanning session. This represents
acquired every few minutes for many hours of lesion another setting in which an ability to obtain perfu-
evolution. Alternatively, ASL perfusion can be inter- sion images without contrast administration is
leaved with other measures of stroke physiology advantageous.
including diusion and spectroscopic MRI so that We have reliably obtained interpretable data
comparisons of perfusion decits and energetics from patients with acute stroke using continuous
can be made. These methods can be used to evalu- ASL perfusion MRI39 (Fig. 13.5). Perfusion decits
ate the mechanisms of MRI measures in humans corresponding to the vascular distribution of the
and to further the understanding of stroke physiol- patients symptoms were observable, and ranged
ogy. Animal MRI scanners, such as for rats and mice, from small focal decits of a similar size to diu-
are relatively inexpensive and already widely avail- sion lesions, to large decits extending well beyond
able in the imaging research community. the diusion lesion. In some instances, bright
linear features were present in ASL perfusion MRI
scans, suggesting eects of delayed arterial transit.
Clinical applications in stroke
While ASL arterial transit eects are probably most
Patients presenting with neurological symptoms problematic in the acute setting, it remains possible
on an ischemic basis should manifest regional to identify areas of hemodynamic compromise and
hypoperfusion, and the distribution of observed transit eects themselves may provide diagnosti-
hypoperfusion with respect to known vascular dis- cally relevant information. We have begun to
tributions should be diagnostically informative. In compare the characteristics of DSC and ASL perfu-
principle, perfusion imaging should provide the sion images to assess the strengths and limitations
greatest sensitivity for stroke detection because of both techniques (Fig. 13.6).
hypoperfusion occurs in advance of metabolic and While embolism rather than primary hypoperfu-
subsequently structural changes. The measure- sion has been considered to be the cause of most
ment of cerebral perfusion using MRI, in combina- cerebrovascular symptoms, patients presenting
Perfusion imaging with arterial spin labelling 169

Fig. 13.5. Quantitative perfusion images in two patients with acute stroke. Unilateral hypoperfusion is apparent in both cases.
Residual bright signal in vessels, most apparent in the bottom case, indicates labelled blood has not reached the tissue. ASL
underestimates perfusion in such cases but signal in vessels can be used as an indicator of collateral ow.

with stroke, transient ischemic attack (TIA), or predicting cerebrovascular ischemia, particularly in
severe carotid stenosis may have clinical features situations of increased embolization such as cardi-
suggesting hypoperfusion. Our studies using con- ovascular surgery.
tinuous ASL perfusion MRI in such a cohort suggest CBF is maintained over a broad range of perfu-
that resting perfusion abnormalities are indeed sion pressures by a property of the cerebrovascular
prevalent,40 particularly in patients with high grade system termed autoregulation. Because autoregu-
stenotic lesions of the cerebral vasculature, (Fig. latory mechanisms in the cerebral vasculature can
13.7). These ndings are consistent with other maintain CBF through vasodilatation, it has been
reports correlating cerebral perfusion or perfusion suggested that CBF alone is an inadequate measure
reserve with the presence of extracranial stenoses of hemodynamic compromise. Thus, while resting
of the carotid arteries. Although most studies have reductions in perfusion are clearly abnormal, alter-
failed to clearly implicate primary hypoperfusion as ations in hemodynamic reserve are also signicant
a cause of large vessel stroke, hypoperfusion has because they suggest that the autoregulatory
been found to be predictive of recurrent stroke. capacity of the cerebral vasculature may be
This discrepancy has begun to be reconciled exhausted. Cerebrovascular reserve is tested by
through the hypothesis that hypoperfusion may measuring the increase in CBF induced by carbon
inuence the outcome from cerebral emboliza- dioxide inhalation or acetazolamide administration
tion.41 This hypothesis, if true, suggests an increas- (cerebrovascular reactivity). A number of studies
ingly important role for perfusion imaging in have demonstrated that cerebrovascular reserve
170 David C. Alsop and John A. Detre

Fig. 13.6. Comparison of ASL and DSC perfusion imaging in acute stroke. The pattern of hypoperfusion is similar in both

Fig. 13.7. Resting perfusion imaging in a patient with left carotid stenosis. Borderzone hypoperfusion as indicated by arrows, is
apparent in a subset of patients with symptomatic cerebrovascular disease.

impairment is particularly signicant in patients tively and non-invasively measuring the eects of
with borderzone ischemia, and that abnormalities pharmacological augmentation throughout the
in augmentation are predictive of stroke. Perfusion brain.42 An example of fractional augmentation
MRI provides a convenient method for quantita- maps made with ASL perfusion is shown in Fig.
Perfusion imaging with arterial spin labelling 171

Fig. 13.8. Characterization of cerebrovascular reserve with ASL and acetazolamide challenge. Resting ASL images, top, show little
abnormality but maps of fractional augmentation of ow, bottom, show impaired reserve in the right MCA distribution.

13.8, and demonstrates impaired cerebrovascular ment strategies. A recent hypothesis posits that
reserve in the vascular distribution of a right middle hypoperfusion may reduce clearance of microem-
cerebral artery stenosis. boli and exacerbate their ischemic consequences,41
Measurement of regional CBF is likely to have in which case perfusion imaging may be used to
several other applications in cerebrovascular predict ischemic complications during cardiovas-
disease and stroke. Management of hypertension cular surgery.
has been widely recognized as the most important
intervention in stroke prevention.43 While clear
benets have been demonstrated for even modest Discussion
reductions in systolic or diastolic blood pressure,
the lower bounds for such reductions have not ASL perfusion MRI can be a useful tool for the
been established, and many patients with chronic assessment of stroke both in model systems and in
hypertension may have alterations in CBF autoreg- the clinical setting. Relative to other perfusion
ulation. Non-invasive and repeatable perfusion measurement techniques, the greatest weakness of
imaging could allow antihypertensive therapy to be ASL for clinical applications is the rapid decay of
titrated based on resulting CBF values. Perfusion the tracer. If blood takes a long time to reach the
imaging can also help to distinguish between tissue, then a loss of signal is unavoidable. While
embolic and hemodynamic mechanisms of the use of a postlabelling delay can be used to mini-
ischemic symptoms, since this may be dicult to mize the systematic eect of the transit time on the
determine clinically yet may suggest diering treat- measured perfusion, the signal-to-noise ratio will
172 David C. Alsop and John A. Detre

be badly compromised. This eect was observed in make it preferable for perfusion measurement in
both our study of acute stroke39 and in cases with some studies of stroke and cerebrovascular disease.
severe intracranial stenosis.40 ASL studies at higher Three great advantages of ASL are non-
elds strengths should improve perfusion quanti- invasiveness, repeatability and absolute quantica-
cation in these situations due to longer T1 values at tion. Traditional ASL implementations may be more
higher elds. It can be argued that the ASL perfu- motion sensitive than DSC, but the use of new
sion images provide much the same information as background suppression techniques can make ASL
qualitative time-to-peak DSC images: the identi- less sensitive than DSC. Preliminary clinical studies
cation of regions with abnormal vascular supply. It indicate that traditional ASL is already useful in the
is also interesting to note that the transit delays assessment of acute stroke. Further technical devel-
observed in some of our acute stroke studies opments promise to address most of the challenges
exceeded the 1.5 second postlabelling delay we of acute stroke evaluation with ASL.
employed. These delays are comparable to the
normal mean transit times of intravascular tracers.
Attempts to quantify perfusion with DSC using
deconvolution or tting techniques5,44 will also be
inaccurate in the presence of such long delays.
Because of the acquisition exibility of ASL, it is
1 Matthew E, Andreason P, Carlson RE et al.
possible to alter the labelling to overcome transit Reproducibility of resting cerebral blood ow measure-
time limitations. If labelling is performed within ments with H215O positron emission tomography in
one cm or less of the imaged region, then inow of humans. J Cereb Blood Flow Metab 1993; 13: 748754.
blood to the tissue occurs through small nearby 2 Waldemar G. Functional brain imaging with SPECT in
vessels. Such a labelling strategy would eliminate normal aging and dementia. Cereb Brain Metab Rev
the eects of large vessel stenosis or tortuous col- 1995; 7: 89130.
lateral ow on the observed images. This labelling 3 Axel L. Cerebral blood ow determination by rapid-
strategy can be approximated by single slice studies sequence computed tomography. Radiology 1980;
where all the blood inowing from very nearby the 137: 676686.
4 Yonas H, Darby JM, Marks EC, Durham SR, Maxwell C.
slice is labelled. While single slice acquisitions are
CBF measured by XeCtapproach to analysis and
less ecient than multislice when the transit time
normal values. J Cereb Blood Flow Metab 1991; 11:
through the brain is short, they can become com-
petitive when the transit time is long. For example, 5 Ostergaard L, Sorensen AG, Kwong KK, Weissko RM,
a continuous labelling experiment with a postlabel- Gyldensted C, Rosen BR. High resolution measure-
ling delay of 1.5 seconds suers more than a factor ment of cerebral blood ow using intravascular tracer
of 3 loss of signal due to the decay of the label bolus passages .2. Experimental comparison and pre-
during that time relative to an acquisition with no liminary results. Magn Reson Med 1996; 36: 726736.
delay and zero transit time. Ten slices can be 6 Rosen BR, Belliveau JW, Buchbinder BR et al. Contrast
acquired sequentially with the same signal-to-noise agents and cerebral hemodynamics. Magn Reson Med
ratio and in the same time-period as the multislice 1991; 19: 285292.
acquisition if moving the label very close to the 7 Calamante F, Gadian DG, Connely A. Delay and dis-
persion eects in dynamic susceptibility contrast
slice eliminates the long transit time into the tissue.
MRI: simulations using singular value decomposition.
Future developments such as selective labelling of
Magn Reson Med 2000; 44: 466473.
blood based on velocity45 rather than spatial posi-
8 Detre JA, Leigh JS, Williams DS, Koretsky AP. Perfusion
tion may also overcome these transit limitations. imaging. Magn Reson Med 1992; 23: 3745.
While DSC MRI is certainly a powerful method 9 Williams DS, Detre JA, Leigh JS, Koretsky AP. Magnetic
for assessing hemodynamic abnormalities, ASL resonance imaging of perfusion using spin inversion
oers a number of advantages, which, with of arterial water. Proc Natl Acad Sci USA 1992; 89:
improved implementation and optimization, may 212216.
Perfusion imaging with arterial spin labelling 173

10 Edelman RR, Siewert B, Darby DG et al. Qualitative 24 Alsop DC, Detre JA. Reduction of excess noise in fMRI
mapping of cerebral blood ow and functional local- using noise image templates (abstr). Proc Int Soc
ization with echo-planar MR imaging and signal tar- Magn Reson Med Fifth Scientic Meeting and
geting with alternating radio frequency. Radiology Exhibition. Berkeley, CA: International Society for
1994; 192: 513520. Magnetic Resonance in Medicine, 1997: 1687.
11 Kwong KK, Chesler DA, Weissko RM et al. MR 25 Dixon WT, Sardashti M, Castillo M, Stomp GP.
Perfusion studies with T1-weighted echo planar Multiple inversion recovery reduces static tissue
imaging. Magn Reson Med 1995; 34: 878887. signal in angiograms. Magn Reson Med 1991; 18:
12 Buxton RB, Frank LR, Wong EC, Siewert B, Warach S, 257268.
Edelman RR. A general kinetic model for quantitative 26 Ye FQ, Frank JA, Weinberger DR, McLaughlin AC.
perfusion imaging with arterial spin labeling. Magn Noise reduction in 3D perfusion imaging by attenuat-
Reson Med 1998; 40: 383396. ing the static signal in arterial spin tagging (ASSIST).
13 Alsop DC, Detre JA. Reduced transit-time sensitivity in Magn Reson Med 2000; 44: 92100.
non-invasive magnetic resonance imaging of human 27 Alsop DC, Detre JA. Background suppressed 3D RARE
cerebral blood ow. J Cereb Blood Flow Metab 1996; ASL perfusion imaging. Proc Int Soc Magn Reson Med.
16: 12361249. Seventh Scientic Meeting, Philadelphia, 1999.
14 Frank LR, Wong EC, Buxton RB. Slice prole eects in 28 Silva AC, Zhang WG, Williams DS, Koretsky AP.
adiabatic inversion: application to multislice perfu- Multislice MRI of rat-brain perfusion during amphet-
sion imaging. Magn Reson Med 1997; 38: 558564. amine stimulation using arterial spin-labeling. Magn
15 Golay X, Stuber M, Pruessmann KP, Meier D, Boesiger Reson Med 1995; 33: 209214.
P. Transfer insensitive labeling technique (TILT): 29 Zaharchuk G, Ledden PJ, Kwong KK, Reese TG, Rosen
application to multislice functional perfusion BR, Wald LL. Multislice perfusion and perfusion terri-
imaging. J Magn Reson Imaging 1999; 9: 454461. tory imaging in humans with separate label and image
16 Edelman RR, Chen G. EPISTAR MRI: multislice coils. Magn Reson Med 1999; 41: 10931098.
mapping of cerebral blood ow. Magn Reson Med 30 Alsop DC, Detre JA. Multisection cerebral blood ow
1998; 40: 800805. MR imaging with continuous arterial spin labeling.
17 Yongbi MN, Branch CA, Helpern JA. Perfusion imaging Radiology 1998; 208: 410416.
using FOCI RF pulses. Magn Reson Med 1998; 40: 31 Talagala SL. Multi-slice perfusion MRI using continu-
938943. ous arterial spin labeling: controlling for MT eects
18 Dixon WT, Du LN, Faul DD, Gado M, Rossnick S. with simultaneous proximal and distal RF. Proc Int Soc
Projection angiograms of blood labeled by adiabatic Magn Reson Med. Sixth Scientic Meeting and
fast passage. Magn Reson Med 1986; 3: 454462. Exhibition. Berkeley, CA: International Society for
19 Wong EC, Buxton RB, Frank LR. A theoretical and Magnetic Resonance in Medicine, 1998: 381.
experimental comparison of continuous and pulsed 32 Alsop DC. Improved eciency for multi-slice continu-
arterial spin labeling techniques for quantitative ous arterial spin labeling using time varying gradients.
perfusion imaging. Magn Reson Med 1998; 40: Proc Int Soc Magn Reson Med. Ninth Scientic
348355. Meeting and Exhibition. Berkeley, CA: International
20 Raichle ME, Eichling JO, Straatmann MG, Welch MJ, Society for Magnetic Resonance in Medicine, 2001:
Larson KB, Ter-Pogossian MM. Bloodbrain barrier 1562.
permeability of 11C-labeled alcohols and 15O-labeled 33 Walsh EG, Minematsu K, Leppo J, Moore SC.
water. Am J Physiol 1976; 230: 543552. Radioactive microsphere validation of a volume local-
21 Bryant RG, Marill K, Blackmore C, Francis C. Magnetic ized continuous saturation perfusion measurement.
relaxation in blood and blood clots. Magn Reson Med Magn Reson Med 1994; 31: 147153.
1990; 13: 133144. 34 Ye FQ, Berman KF, Ellmore T et al. H215O PET valida-
22 Wong EC, Buxton RB, Frank LR. Quantitative imaging tion of steady-state arterial spin tagging cerebral
of perfusion using a single subtraction (QUIPSS and blood ow measurements in humans. Magn Reson
QUIPSS II). Magn Reson Med 1998; 39: 702708. Med 2000; 44: 450456.
23 Siewert B, Bly BM, Schlaug G et al. Comparison of the 35 Wang Y, Hu WX, Perez-Trepichio AD et al. Brain tissue
BOLD- and EPISTAR-technique for functional brain sodium is a ticking clock telling time after arterial
imaging by using signal detection theory. Magn Reson occlusion in rat focal cerebral ischemia. Stroke 2000;
Med 1996; 36: 249255. 31: 13861391.
174 David C. Alsop and John A. Detre

36 Lythgoe MF, Thomas DL, Calamante F et al. Acute 41 Caplan LR, Hennerici M. Impaired clearance of
changes in MRI diusion, perfusion, T-1, and T-2, in a emboli (washout) is an important link between hypo-
rat model of oligemia produced by partial occlusion of perfusion, embolism, and ischemic stroke. Arch
the middle cerebral artery. Magn Reson Med 2000; 44: Neurol 1998; 55: 14751482.
706712. 42 Detre JA, Samuels OB, Alsop DC, Gonzalez-At JB,
37 Jiang Q, Zhang RL, Zhang ZG, Ewing JR, Divine GW, Kasner SE, Raps EC. Noninvasive magnetic reso-
Chopp M. Diusion-, T-2, and perfusion-weighted nance imaging evaluation of cerebral blood ow with
nuclear magnetic resonance imaging of middle cere- acetazolamide challenge in patients with cerebrovas-
bral artery embolic stroke and recombinant tissue cular stenosis. J Magn Reson Imaging 1999; 10:
plasminogen activator intervention in the rat. J Cereb 870875.
Blood Flow Metab 1998; 18: 758767. 43 Bronner LL, Kanter DS, Manson JE. Primary pre-
38 Hossman KA. Viability thresholds and the penumbra vention of stroke. N Engl J Med 1995; 333:
of focal ischemia. Ann Neurol 1994; 36: 557565. 13921400.
39 Chalela JA, Alsop DC, Gonzalez-Atavales JB, Maldjian 44 Rempp KA, Brix G, Wenz F, Becker CR, Guckel F,
JA, Kasner SE, Detre JA. Magnetic resonance perfusion Lorenz WJ. Quantication of regional cerebral blood-
imaging in acute ischemic stroke using continuous ow and volume with dynamic susceptibility con-
arterial spin labeling. Stroke 2000; 31: 680687. trast-enhanced MR-imaging. Radiology 1994; 193:
40 Detre JA, Alsop DC, Vives LR, Maccotta L, Teener JW, 637641.
Raps EC. Noninvasive MRI evaluation of cerebral 45 Norris DG, Schwarzbauer C. Velocity selective radio-
blood ow in cerebrovascular disease. Neurology frequency pulse trains. J Magn Reson 1999; 137:
1998; 50: 633641. 231236.
Clinical role of echoplanar MRI in stroke

Stephen Davis and Mark Parsons

Department of Neurology, Royal Melbourne Hospital and University of Melbourne, Victoria, Australia

Introduction decisions to be made based on assessment of

pathophysiology in the individual patient, rather
Stroke is a leading cause of death in Western coun- than relying on the arbitrary time windows used
tries, with a case mortality rate of 24% within the in clinical trials.11,12
rst month, higher than most forms of cancer.1 It is Echoplanar magnetic resonance imaging (EPI)
the commonest cause of long-term adult disability.2 enables rapid, non-invasive imaging and analysis
Recent important advances in acute stroke therapy of cerebral pathophysiology in acute stroke.
include the established benets of stroke units, and Echoplanar techniques include diusion-
the introduction of the rst proven interventional weighted imaging (DWI), a measure of tissue water
therapies. These include intravenous t-PA within 3 diusion, perfusion-weighted imaging (PWI),
hours,3 acute use of aspirin4,5 and administration of which measures various indices of cerebral perfu-
intra-arterial thrombolysis within 6 hours of stroke sion, and magnetic resonance spectroscopy
onset.6 A range of new strategies are being investi- (MRS), allowing analysis of in vivo biochemistry.
gated, which generally involve either rapid reperfu- Because DWI can rapidly detect ischemic cerebral
sion with thrombolytic agents, or neuroprotection tissue, which is usually destined to infarct, and
from the damaging neurotoxic cascade that follows distinguish acute from chronic infarction, it repre-
ischemia.7 sents an important clinical advance over com-
Both strategies focus on treatment of the puted tomography (CT) and conventional
ischemic penumbra.810 An understanding of the magnetic resonance (MR) scanning. Both of these
pathophysiologic basis of brain ischemia and the modalities are relatively insensitive to early brain
penumbra is important in interpreting the mag- ischemia.
netic resonance (MR) changes imaged in acute In addition, combined use of DWI and PWI can
stroke. The penumbra is a variable region of criti- allow assessment of at-risk or threatened brain
cally underperfused but still viable tissue sur- tissue in the ischemic penumbra. The presence of
rounding the infarcted core. Despite penumbral tissue is estimated as the dierence, or
symptomatic hypoperfusion, essential energy- mismatch, between the perfusion rim on PWI and
requiring processes such as the NaK ATPase the ischemic core, dened by DWI.13 Rapid evalua-
system, are still able to maintain ionic gradients, tion of the presence and extent of the penumbra is
and hence membrane and neuronal integrity.9,10 likely to be valuable in the selection of patients for
The penumbra represents a therapeutic window acute therapy. For example, thrombolysis can
for intervention. However, the duration of this reperfuse and hence rescue threatened penumbral
interval is uncertain and is likely to vary consider- tissue, with reduction in eventual infarct size and
ably from person to person. Rapid identication improved outcome.14 Although tissue plasminogen
of the penumbra can allow rational therapeutic activator (t-PA) improves outcome in ischemic

176 Stephen Davis and Mark Parsons

Table 14.1. Comparison of CT, SPECT, PET, conventional MRI, EPI modalities in acute stroke


Current use in Standard Restricted role, Research technique Alternative Increasing use
stroke diagnostic chiey in research technique to CT, (particularly DWI)
technique centres but less used as
Advantages Widely available. Perfusion Measures perfusion, Lack of bone Highly sensitive
Sensitively information. metabolism, at artefacts; better diagnosis early
excludes Complementary risk penumbral imaging brain- ischemia. Potential
hemorrhage to CT, MRI tissue stem to image
penumbral tissue
Disadvantages Insensitive for Cannot replace Expensive, Cost. Relatively Availability. PWI
early ischemia, structural imaging logistically difcult. insensitive in requires post-
at risk tissue techniques. Does Very limited rst 6 hours. processing
not allow availability Previous concerns
measurement of about exclusion
metabolism, of hemorrhage
at risk tissue
Future clinical May be replaced Diminishing Research only EPI applications Might become
role in stroke by EPI investigative
modality of choice

stroke within 3 hours of onset, not all patients Imaging techniques in brain ischemia (Table
treated benet, as there is a signicant risk of 14.1)
hemorrhage. Furthermore, the time window for
benet may well exceed 3 hours in some patients. Computed tomographic scanning (CT) has been
Because of the heterogeneity of human stroke and the investigation of choice for acute stroke evalua-
the cerebral circulation, it is recognized that there is tion for the past three decades. It sensitively
no universal time window for acute treatments. excludes hemorrhage, but ischemic changes can be
Combined DWI/PWI, with rapid diagnosis of cere- subtle. In the rst 6 hours of stroke onset, CT is
bral pathophysiology, is likely to facilitate acute often normal. Major ischemic changes can be iden-
therapeutic decisions. tied in some cases within about 2 hours of
The high sensitivity and specicity of EPI in onset,1618 but CT remains a relatively blunt instru-
stroke together with its potential to guide therapy ment in the evaluation of hyperacute brain ische-
have led to the widespread view that DWI/PWI, mia (Table 14.1). Various functional imaging
combined with magnetic resonance angiography methods have been used in the evaluation of acute
(MRA), is likely to replace CT as the primary modal- stroke. Positron emission tomography (PET)
ity in stroke investigation. Others are more cau- enables imaging and measurement of both perfu-
tious,15 emphasizing that more clinical correlations sion and metabolism. While important information
are needed, particularly in conrming its accuracy about cerebral pathophysiology in acute brain
compared with CT scanning. ischemia has been obtained from PET studies,
Clinical role of echoplanar MRI in stroke 177

Table 14.2. Echoplanar MR techniques in stroke

Techniques DWI PWI MRS

Measures Translational movement of Cerebral perfusion indices NAA, lactate Chemical shift
water (ADC values) (mean transit time, cerebral imaging
blood volume, cerebral
blood ow)

Acute stroke Light bulb signal due to Reduced blood ow Reduced NAA, increased lactate
attenuated water diffusion

Clinical role Highly sensitive diagnosis of Hypoperfusion in acute Early chemical derangements may
acute ischemia Differentiation stroke. DWI/PWI mismatch complement DWI/PWI information
from chronic ischemia allows estimation of at risk
penumbral tissue

acute stroke studies are logistically dicult, the cations that have been dened by EPI research and
technique is expensive and not widely available. clinical experience include the sensitive diagnosis
Single photon emission computed tomography of acute ischemia, the dierentiation of acute from
(SPECT) is more suited to acute studies, is widely chronic infarcts, the prognostic value of acute DWI
available and can give information about cerebral and PWI lesions, insights into infarct topography
blood ow abnormalities. Acute hypoperfusion in and pathophysiology, use as a surrogate endpoint
stroke correlates with outcome and repeat imaging in drug trials and the potential of combined
allows assessment of reperfusion, also shown to DWI/PWI as a powerful tool in providing a template
correlate with functional recovery.12,1922 Although for acute therapy (Table 14.2).
SPECT can be used to retrospectively evaluate the There is a view in experienced centres that
eects of acute therapies,2124 it cannot delineate multimodality MRI can be used as the sole diagnos-
metabolism or identify the irreversibly damaged tic modality in acute stroke.28 In contrast, others29
and hence non-viable infarct core. Therefore, the have argued for more studies to demonstrate the
inability of SPECT to dene the critical penumbral evidence for the replacement of CT. There has
region has limited its role in stroke management. nonetheless been a debate in the recent literature
Conventional magnetic resonance imaging (MRI) as to whether EPI is ready for prime time in acute
reveals the structure of the brain with excellent stroke evaluation.30,31 We and others certainly
spatial resolution. However, standard MRI with T2- believe that it is.12,28,32,33
weighted imaging does not reliably show changes
in the brain within 6 hours of stroke onset. This is
likely to be the time window for most acute stroke Clinical applications of EPI in acute stroke
In contrast, echoplanar imaging (EPI) of the
Diagnosis of acute and chronic ischemia
brains diusion of water (DWI), perfusion (PWI) or
metabolite concentrations on magnetic resonance Diusion-weighted imaging (DWI) involves meas-
spectroscopy (MRS) can reveal abnormalities urement of alterations in the diusion of water
within 12 minutes of the onset of ischemia.13,25,26 It molecules. At present, DWI is the most used of the
is an increasingly available technology and far more EPI parameters in clinical practice. With acute
useful than T2-weighted MRI within the rst 48 ischemia, there is a marked reduction in the appar-
hours of acute stroke.26,27 The major clinical appli- ent diusion coecient of water (ADC), reecting
178 Stephen Davis and Mark Parsons

Fig. 14.1. Patient scanned at 1 hour after symptom onset. There is an extensive diusion lesion in the right middle cerebral artery

early disruption of high-energy metabolism, with about 710 days after stroke onset, although earlier
failure of ion pumps and cytotoxic edema. Ischemic in some cases.38 This phenomenon is clinically
lesions appear hyperintense on DWI, termed the useful in distinguishing acute from chronic infarcts,
lightbulb sign and are apparent within minutes of often dicult with CT or conventional MR scan-
onset (Fig. 14.1). These brain regions represent the ning.32,33,38
irreversibly damaged infarct core, and the minimal DWI is both highly sensitive and specic for
extent of the nal infarct size in most cases. Very acute stroke. In 194 cases studied within 24 hours of
early reperfusion has been shown to reverse small stroke onset, Lovblad et al. reported 88% sensitivity
DWI lesions in animal models and occasional and 95% specicity. In the rst 6 hours, the sensitiv-
human cases,34,35 as will be seen below. ity rose to 94% with 100% specicity. We found
Within the rst few days after stroke onset, there 100% sensitivity and 100% specicity in patients
is typically expansion of the non-viable infarct core, with a clinical diagnosis of acute stroke, imaged
demonstrated by serial DWI scans.12,36,37 Eventually, within 6 hours of onset.39 In one patient with a
the ADC values in the infarcted tissue increase, normal DWI scan, further investigations led to the
related to a breakdown in cell membranes with diagnosis of an acute brachial plexopathy. However,
resultant increase in extracellular space and DWI-negative stroke has been reported by many
increased diusional movement of water mol- groups, particularly with small, deep lesions in the
ecules. Hence, older infarcts appear hypointense on brainstem.33,40
DWI scans compared with normal brain. This The high-intensity lightbulb sign is of great clin-
change is called pseudonormalization and occurs ical value in dierentiating acute from chronic
Clinical role of echoplanar MRI in stroke 179

less used in routine practice than DWI, but provides

invaluable complementary information in acute
stroke, particularly in measurement of acute
pathophysiology and at-risk tissue in the ischemic
penumbra, as will be seen below. PWI involves the
acquisition of whole brain T2*-weighted perfusion
scans every 2 seconds following paramagnetic con-
trast injection (Gadolinium-DTPA) and makes use
of the signal loss that occurs during the dynamic
tracking of the rst pass of the contrast agent.47,48 A
signal intensity-time curve allows measurements of
relative cerebral blood volume (CBV), mean transit
Fig. 14.2. Patient imaged at 12 hours after onset of right hemi-
time (MTT) and an estimation of regional cerebral
paresis. There is marked white matter ischemic change on blood ow (CBF), where CBFCBV/MTT (Fig. 14.3,
T2-weighted imaging, but the culprit lesion for acute symp- see colour plate section). Because the MTT is
toms is not apparent. DWI shows an acute lesion in the left hyperintense in regions of impaired perfusion,
internal capsule. most groups have used the MTT as an index of CBF.
However, work in our laboratory has shown that
CBF is likely to be a better correlate of neurological
infarcts, particularly in patients with multiple impairment and functional outcome.14
ischemic lesions of dierent ages (Fig. 14.2). A Magnetic resonance angiography complements
number of studies have compared the clinical PWI and DWI by allowing determination of middle
utility of DWI with conventional MRI. The Albers cerebral artery (MCA) stem and major branch
group41,42 showed that DWI could sometimes iden- patency 37,49. The table time for the performance of
tify the stroke lesion at a site other than clinically DWI, PWI and MRA is approximately 20 minutes, a
thought, show multiple lesions and demonstrated practical duration for most stroke patients. In
that lesions considered acute on conventional MRI summary, combined DWI and PWI allow hyper-
were actually old. They also conrmed the super- acute, sensitive and specic diagnosis of ischemia,
iority of DWI over CT within the rst few hours after distinction of older strokes, measurement of at-risk
stroke onset, as shown by others.43 DWI has also tissue in the ischemic penumbra and concurrent
provided insight into the pathophysiology of visualization of arterial pathology such as stenosis
patients with transient ischemic attacks. DWI or occlusion.
abnormalities are present in nearly 50% of patients,
with full reversibility in about half of these.44
Stroke prognosis
Persisting DWI lesions were particularly common in
patients with symptoms lasting greater than 1 hour. Lovblad et al.50 pointed out the acute prognostic
Echoplanar imaging has also allowed rapid gradi- potential of DWI lesions in ischemic stroke. Large
ent echo susceptibility-weighted sequences that acute DWI lesions usually predict poor outcome.
can sensitively exclude acute hemorrhage.45 In a We and others have emphasized the prognostic
small series of patients with hemorrhage within value of combined DWI and PWI lesion
2 hours of symptom onset, all patients were iden- volumes.11,12,51 Both DWI and PWI lesion volumes
tied with MRI.46 Because of the slight uncertainty obtained within 6 hours of stroke onset have been
about the accuracy of MRI in acute cerebral hemor- correlated with the neurological impairment scores
rhage, a large trial, involving randomization of at 24 hours and infarct size at 7 days on T2-weighted
acute stroke patients to either CT or MRI, is being MRI.51 Our group reported that the volume of both
mounted to address this question. the DWI and PWI lesions correlated with acute clin-
Perfusion-weighted imaging (PWI) is currently ical impairment scores, the eventual infarct size at
180 Stephen Davis and Mark Parsons

fused brain. However, some patients have sponta-

neous reperfusion of penumbral tissue between
acute and subacute studies, associated with signi-
cantly improved clinical outcome. The nal infarct
size is usually intermediate between the DWI core
and perfusion boundary in patients with a demon-
strable penumbra, and the patients neurological
course typically correlates with this evolution.
It should be emphasised that PWI > DWI mis-
match patterns are, at best, an approximation of
the true ischemic penumbra. PWI may overesti-
mate the mismatch region, as current bolus-
Fig. 14.4. PWI DWI mismatch pattern. Patient imaged at tracking techniques may not reliably distinguish
3 hours after onset of aphasia and right hemiparesis. The PWI between critically hypoperfused tissue at risk and
lesion is extensive and involves the entire MCA territory. The oligemic regions with lesser degrees of hypoperfu-
DWI lesion is much smaller, in the striatocapsular region. sion, not destined to infarct.5355 The recent use of
thresholded PWI maps, which identify diering
the chronic 90-day stage, as well as the outcome severities of contrast bolus delay, might circumvent
functional status of the patient.12 However, these this problem.5659 In addition, DWI lesions may
relationships dynamically evolve over the rst few overestimate irreversibly damaged tissue, as rapid
days after stroke onset, which helps explain the reperfusion with thrombolysis has been shown to
dierences in the volume of the acute parameters lead to permanent reversal of some DWI lesions.35
measured within hours of stroke onset, compared In the future, sophisticated statistical algorithms
with the nal infarct size.12,36 may be able to predict tissue fate on a voxel-by-
voxel basis using DWI and PWI parameters rather
than the currently used volumetric techniques.60
Relationship between DWI and PWI : imaging tissue
Nonetheless, the presence of a PWI > DWI mis-
at-risk in the ischemic penumbra
match pattern is at present, the most practical de-
Baird et al. rst explored the dierent patterns of nition of at-risk tissue for the clinician, and this
match and mismatch between DWI and PWI in group of patients are likely to have the greatest
acute stroke.36 Our group12 reported that 6 patterns potential to benet from thrombolytic
could be identied in acute ischemia, which we therapy.35,53,61,62
found predicted the evolution of ischemic decits, We have also identied patients with hypoperfu-
and outcome. The most common of these is the sion yet normal DWI, typical of transient ischemic
putative penumbral pattern, dened by a PWI lesion attacks, where there is tissue at risk, but without
of greater volume than a DWI lesion (Fig. 14.4). frank infarction. However, as previously noted, a
Warach et al. rst proposed that PWI > DWI mis- signicant proportion of TIA patients have DWI
match represented the ischemic penumbra, brain abnormalities.44 We and others have explored the
tissue at risk of infarction that might be potentially relationships between MRA ndings and DWI/PWI
rescued by acute therapy such as thrombolysis.11,13 patterns. In acute stroke, MCA stem occlusion on
We found that a penumbral pattern was present MRA studies is usually associated with a penumbral
in 75% of stroke patients imaged within 6 hours and pattern (Fig. 14.5).37,49 Untreated, these patients
that the probability of a penumbral pattern show typically greater expansion of the ischemic
decreased with time over the rst 24 hours after core, larger nal infarct size and worse clinical
stroke.12,52 In untreated patients, an acute penum- outcome than penumbral patients with a patent
bral pattern generally predicts expansion of the artery.37 However, the penumbral pattern can also
infarcted DWI core into the surrounding hypoper- be seen with normal MRA. These patients are likely
Clinical role of echoplanar MRI in stroke 181

Fig. 14.5. Typical picture of right MCA occlusion at 4 hours (arrow). There is associated PWI>DWI mismatch, with subsequent
expansion of the ischemic core at day 90 T2-weighted imaging (right).

Fig. 14.6. Non-mismatch (non-penumbral) pattern. At 15 hours after symptom onset this patient had no visible vessel occlusion
on MRA, and the acute DWI lesion is equal in size to the acute PWI lesion. There is no expansion of the ischemic core at 3 month
T2-weighted imaging.

to have occlusion of more distal second and third the ischemic core between acute, subacute and
order arterial branches, beyond the spatial resolu- outcome studies,37 and nal infarct size is similar to
tion of MRA.37 the size of the acute DWI lesion (Fig. 14.6).
Non-mismatch patterns are those in which PWI Therefore, it is generally considered that patients
lesions are smaller or absent compared with DWI. with such non-mismatch patterns are unlikely to
In these patients, there is usually no expansion of have penumbral tissue present.
182 Stephen Davis and Mark Parsons

Fig. 14.8. Patient scanned 13 hours after onset of right senso-

rimotor hemiparesis. DWI highlights acute infarction in the
posterior limb of the left internal capsule that is indistin-
guishable from chronic ischemic change on concomitant
T2-weighted imaging.

Fig. 14.7. Cardio-embolic infarction. DWI at 23 hours after

with MRI or CT is often of limited clinical utility in
presentation with a right cortical hand, shows the culprit
the acute setting. In contrast, Singer et al.65
lesion, but also an occult acute diusion lesion (arrow).
reported a 94.6% accuracy with DWI in acute sub-
cortical infarction. Lacunar infarcts are often multi-
ple. The high intensity lightbulb signature on DWI
Infarct topography and pathogenesis
allows delineation of acute lacunar infarction from
Our studies of acute infarct topography have shown chronic white matter lesions (Fig. 14.8). Singer et al.
that the proximal MCA territory in the insular noted that in 10% of cases, the acute infarction was
region is particularly susceptible to infarction.52 not detected on conventional MRI. Delineation of
However, the range of MCA subcortical and cortical the acute lesion with DWI was possible in all cases
infarct types are readily identied including lacunar of multiple lacunes, corresponding to the acute
lesions, striatocapsular and branch arterial infarcts. clinical symptomatology. Similarly, Noguchi et al.66
An increasing number of studies have indicated used DWI in patients with lacunar infarcts and were
that the combined use of DWI and PWI can help also able to identify acute small infarcts and separ-
determine the likely pathophysiology of acute ate these from chronic lesions. In 3 of 35 patients,
stroke. The Los Angeles group showed that multi- small hyperacute brain stem infarcts, within
modal MRI (DWI and MRA) within 24 hours sub- 6 hours of onset were seen only with DWI and in 5
stantially improved the diagnosis of stroke subtype, patients, fresh small infarcts adjacent to multiple
using the TOAST and Oxfordshire classications.63 old infarcts could only be distinguished with DWI.
A pattern of a proximal DWI lesion with separate, Other reports have shown the superiority of DWI
distal PWI lesions in the MCA territory suggests an over conventional MRI for the diagnosis of clini-
embolic pathogenesis.52 Cardiac emboli are sug- cally relevant, penetrating artery infarcts.67 A recent
gested by a cluster of acute DWI lesions in dierent study used DWI to determine whether clinical
arterial territories (Fig. 14.7). One group reported lacunar syndromes predicted lacunar infarcts.68
that acute, multiple brain infarcts were detected in Twenty-three patients with clinical lacunar syn-
29% of a large series studied with DWI, due to either dromes had DWI within 3 days of onset. In most
large-vessel disease or cardiogenic embolism.64 cases, acute lacunar syndromes correlated with
In subcortical infarction, conventional imaging small subcortical infarcts compatible with single-
Clinical role of echoplanar MRI in stroke 183

location could give rise to a variety of clinical

lacunar syndromes.71 Conversely, Lindgren et al.68
found a good correlation between lacunar syn-
dromes and subcortical lesions using DWI.

Use of DWI/PWI as surrogate endpoints in

investigational drug trials

With the lengthy list of failed neuroprotective strate-

gies in Phase III trials, it has been suggested that
DWI/PWI studies could be used to test proof of
Fig. 14.9. Patient presenting with left pure motor hemiparesis. concept in early human studies.37,72 Because DWI
DWI at 9 hours reveals right striatocapsular infarction, and a lesions correlate with acute and outcome clinical
small region of infarction in the right parietal cortex (arrow). scores and infarct size, they can be used as valid sur-
Despite having a classical clinical syndrome, a proximal
rogate measures of stroke outcome in acute investi-
embolic source must be the pathogenic mechanism for the
gational drug trials.12,50 Attenuation of the typical
expansion of acute DWI lesions has been used as a
surrogate endpoint of drug ecacy. Fishers group
used DWI in a rat stroke model to evaluate a puta-
penetrator occlusion. However, a small number had tive neuroprotective NMDA antagonist.73 They cor-
cortical involvement. related drug ecacy with the imaging results and
These newer MRI techniques may contribute to a reduced histological infarct size. The neuroprotec-
modication of the lacunar hypothesis, namely that tive agent citicholine was evaluated in an MR sub-
they are small vessel infarcts due to in situ lipohya- study, measuring lesion growth between the acute
linosis and microatheroma, rather than embolic stage and 12 weeks.72 Although no dierence was
pathology. There is accumulating evidence that found in the primary endpoint, lesion volume
acute DWI identies a subset of patients presenting reduction correlated with measured neurological
with typical lacunar syndromes with a pattern of improvement and lesion volume attenuation was
ischemic lesions suggesting an embolic pathogene- correlated with the drug between weeks 1 and 12.72
sis6870 (Fig. 14.9). In a study of 62 patients with clas- Similar proof of concept studies with PWI and DWI
sical lacunar syndromes studied with DWI within have been performed or are currently under way
3 days, 16% had multiple regions of increased signal with other neuroprotective agents.74 Such studies
intensity. The index lesion could be identied in the can be performed with 50100 patients per treat-
hemisphere or brainstem, while these subsidiary ment arm and are likely to indicate the probability
infarcts usually occurred in leptomeningeal artery of success in large and expensive phase III studies.72
territories and were more often associated with an
embolic stroke source.69 Our group recently found
Selection of patients for acute therapies
that in 16 patients with clinical presentations of
subcortical infarcts, 10 had some evidence to As discussed earlier, patients with an acute penum-
suggest an embolic etiology, with a pattern of bral (PWI > DWI mismatch) pattern should be ideal
multiple, acute lesions in more than one single candidates for thrombolysis.11,12 The pivotal NINDS
small-vessel territory.70 Schonewille et al. used DWI trial established the ecacy of t-PA in the treatment
to study 43 patients presenting acutely with classi- of acute ischemic stroke, and led to its licensing as
cal lacunar syndromes. They conrmed that the rst acute stroke therapy in the USA, for selected
lacunar syndromes could be caused by deep patients within 3 hours of stroke onset.3,75 Trials
ischemic infarcts, but also by supercial infarction with a 6-hour window have been negative and only
and hemorrhage. Furthermore, lesions in the same indicated a trend towards clinical ecacy.76
184 Stephen Davis and Mark Parsons

outcome.77 However, these acute CT changes are

subtle, and dierent interpretations can be made by
skilled observers.78 We compared the size of the DWI
lesions with the extent of early ischemia on acute CT
scan (Fig. 14.10). Ischemic lesions greater than one
third of the MCA territory were identied by DWI in
all cases, while DWI showed more extensive lesions
than CT in some cases.39 Large acute DWI lesions
may predict an unacceptably high rate of hemor-
rhage with thrombolytic therapy.
We, and collaborating centres in Australia, have
(b )
used PWI and DWI techniques to study 20 patients
before and after treatment with t-PA within 6 hours
of stroke onset14 (Fig. 14.11). Most patients had a
penumbra. Reperfusion occurred in most patients,
with attenuation of infarct expansion, in contrast
with our matched, historical controls who did not
receive thrombolysis. These results support the
hypothesis that t-PA facilitates reperfusion and
attenuates infarct expansion in penumbral
patients. These results are consistent with data
Fig. 14.10. (a) CT at 2.5 hours after onset of left hemiparesis. from other groups.61,62,79 Marks et al.62 studied a
There is perhaps some subtle reduced density in the deep small group of patients, although only after t-PA,
white matter. (b) DWI performed prior to CT highlights the and found early reperfusion compared with
greater sensitivity of this technique for acute ischemia. untreated subjects. The UCLA group studied
patients before and after intra-arterial thrombolysis
and found resolution of perfusion decits and
It is clear that not all patients treated within 3 indeed some shrinkage of the initial DWI lesion
hours of ischemic stroke onset benet from throm- volume.35 The Heidelberg group61 concluded that
bolytic therapy. In addition, a therapeutic window MRI was an eective and comprehensive tool in
of 3 hours signicantly limits the number of stroke evaluating stroke patients for thrombolysis. They
patients who can be treated. Ideally, stroke physi- also indicated that PWI/DWI represented the target
cians would like to be able to identify patients with group for thrombolysis and that reperfusion in the
the greatest potential to benet from, and the penumbra followed recanalization.
lowest risk of induced hemorrhage as a result of Trials are now under way in Australia and North
thrombolysis and to extend the current therapeutic America to prospectively evaluate the hypothesis
window beyond three hours in selected patients. that the presence and extent of the ischemic
This would involve a strategy to individualize penumbra using EPI will predict patients most
the therapeutic time window in acute stroke likely to respond to thrombolytic therapy. In
patients. Australia, the EPITHET trial involves a double-
There is also speculation that use of combined blind, randomized, multicentre controlled trial of
DWI/PWI might improve patient selection in terms t-PA vs. placebo in 100 eligible patients with hemi-
of safety. Early major ischemic changes on CT scan, spheric infarction 36 hours after stroke onset.80 In
dened by the ECASS Investigators as change involv- the USA, the DEFUSE trial involves patients openly
ing more than one-third of the middle cerebral treated with t-PA up to 6 hours, to determine which
artery (MCA) territory, appear to predict a higher risk DWI/PWI parameters best predict response to t-
of hemorrhagic transformation and worse PA.79 In Germany, centres are already selecting
Clinical role of echoplanar MRI in stroke 185

Fig. 14.11. Patient with acute right MCA occlusion at 2.5 hours and associated PWI DWI mismatch. Following intravenous
thrombolysis, recanalization of the occluded vessel has occurred, with complete reperfusion (lower PWI map) and no infarct

patients for thrombolysis beyond 3 hours, using highlighting the complementary nature of MRS and
evidence of PWI/DWI mismatch on MRI.61 MRI. NAA begins to decline within several hours of
acute infarction,81,8385 and is thought to indicate a
loss of neuronal viability.89 Several groups have
MR spectroscopy in clinical practice
found a correlation between NAA levels and stroke
Magnetic resonance spectroscopy (MRS) allows the outcome that is independent of infarct size,
non-invasive assessment of metabolic changes emphasizing the potential clinical utility of this
during stroke.81 This information can be obtained technique.90,91
in conjunction with other MRI sequences so that A number of studies have suggested that MRS
correlations between anatomical and physiological may herald the presence of hypoperfused but still
changes can be made with changes in the biochem- viable neural cells within or surrounding infarcted
ical processes occurring in cerebral ischemia.82 tissue, as these cells continue to produce
Lactate and N-acetylaspartate (NAA) are two lactate.86,92,93 Therefore, lactate levels measured
metabolites that can be measured by 1H MRS, within the penumbra may be a marker of tissue that
which are of major relevance in stroke (Fig. ultimately progresses to infarction. This clearly has
14.12).8385 Lactate, which is not detectable in non- implications for acute stroke management.
ischemic brain, is elevated almost immediately Furthermore, our group has recently demonstrated
after the onset of cerebral ischemia,86,87 and indi- that acute hyperglycemia leads to elevated lactate
cates a switch from oxidative metabolism to ane- production, which in turn leads to increased con-
robic glycolysis.88 We have found that lactate levels version of penumbral tissue to infarction.94 The
in the acute ischemic region provide independent more widespread availability, along with shorter
predictive information when combined with DWI, imaging times of multivoxel MRS, or chemical shift
186 Stephen Davis and Mark Parsons

Fig. 14.12. Acute left MCA branch diusion lesion imaged at 14 hours after symptom onset in which proton magnetic resonance
spectroscopy (MRS) with a single voxel in the core of the infarct shows reduced NAA to choline ratio and a large lactate peak. A
voxel placed in the homologous region on the left for comparison shows a normal spectrum, there is no lactate peak.

imaging (CSI), may enable assessment of metabolic REFERENCES

changes occurring within the full extent of the
penumbra (Fig. 14.13, see colour plate 1 Anderson CS, Jamrozik KD, Burvill PW, Chakera TM,
section).85,93,95 The addition of CSI to a multimodal Johnson GA, Stewart-Wynne EG. Ascertaining the true
incidence of stroke: experience from the Perth
acute stroke MRI examination may confer addi-
Community Stroke Study, 19891990. Med J Aust 1993;
tional accuracy to the identication of at-risk
158: 8084.
ischemic tissue.
2 Bennett SA, Magnus P. Trends in cardiovascular risk
factors in Australia. Results from the National Heart
Foundations Risk Factor Prevalence Study, 19801989.
Conclusions and future directions Med J Aust 1994; 161: 519527.
3 Tissue plasminogen activator for acute ischemic
We conclude that EPI with DWI, PWI and MRS oers stroke. The National Institute of Neurological
unique insights into acute brain ischemia. These Disorders and Stroke rt-PA Stroke Study Group. N Engl
techniques have now moved from the research arena J Med 1995; 333: 15811587.
into routine clinical use. Given the highly sensitive 4 The International Stroke Trial (IST): a randomised trial
detection of ischemia and the potential for rational of aspirin, subcutaneous heparin, both, or neither
among 19435 patients with acute ischaemic stroke.
selection of acute therapies, we predict that EPI will
International Stroke Trial Collaborative Group. Lancet
replace CT as the method of choice for acute stroke
1997; 349: 15691581.
evaluation, particularly in patients being considered
5 CAST: randomised placebo-controlled trial of early
for acute interventional stroke therapy. aspirin use in 20,000 patients with acute ischaemic
stroke. CAST (Chinese Acute Stroke Trial)
Collaborative Group. Lancet 1997; 349: 16411649.
Clinical role of echoplanar MRI in stroke 187

6 Furlan A, Higashida R, Wechsler L et al. Intra-arterial 19 Davis SM, Chua MG, Lichtenstein M, Rossiter SC,
prourokinase for acute ischemic stroke. The PROACT Binns D, Hopper JL. Cerebral hypoperfusion in stroke
II study: a randomized controlled trial. Prolyse in prognosis and brain recovery. Stroke 1993; 24:
acute cerebral thromboembolism. J Am Med Assoc 16911696.
1999; 282: 20032011. 20 Baird AE, Donnan GA, Austin MC, Fitt GJ, Davis SM,
7 Davis SM. Tissue rescue therapy for ischaemic stroke. McKay WJ. Reperfusion after thrombolytic therapy in
J Clin Neurosci 1995; 2: 715. ischemic stroke measured by single-photon emission
8 Astrup J, Siesjo BK, Symon L. Thresholds in cerebral computed tomography. Stroke 1994; 25: 7985.
ischemia the ischemic penumbra. Stroke 1981; 12: 21 Infeld B, Davis SM, Donnan GA et al. Streptokinase
723725. increases luxury perfusion after stroke. Stroke 1996;
9 Ginsberg MD, Pulsinelli WA. The ischemic penumbra, 27: 15241529.
injury thresholds, and the therapeutic window for 22 Yasaka M, OKeefe GJ, Chambers BR et al.
acute stroke [editorial; comment]. Ann Neurol 1994; Streptokinase in acute stroke: eect on reperfusion
36: 553554. and recanalization. Australian Streptokinase Trial
10 Hossmann KA. Viability thresholds and the penumbra Study Group. Neurology 1998; 50: 626632.
of focal ischemia [see comments]. Ann Neurol 1994; 23 Barber PA, Davis SM, Infeld B et al. Spontaneous
36: 557565. reperfusion after ischemic stroke is associated with
11 Warach S, Dashe JF, Edelman RR. Clinical outcome in improved outcome. Stroke 1998; 29: 25222528.
ischemic stroke predicted by early diusion-weighted 24 Infeld B, Davis SM, Donnan GA et al. Nimodipine and
and perfusion magnetic resonance imaging: a prelimi- perfusion changes after stroke. Stroke 1999; 30:
nary analysis. J Cereb Blood Flow Metab 1996; 16: 5359. 14171423.
12 Barber PA, Darby DG, Desmond PM et al. Prediction 25 Edelman RR, Wielopolski P, Schmitt F. Echo-planar
of stroke outcome with echoplanar perfusion- MR imaging. Radiology 1994; 192: 600612.
and diusion-weighted MRI. Neurology 1998; 51: 26 Sorensen AG, Buonanno FS, Gonzalez RG et al.
418426. Hyperacute stroke: evaluation with combined multi-
13 Warach S, Gaa J, Siewert B, Wielopolski P, Edelman RR. section diusion-weighted and hemodynamically
Acute human stroke studied by whole brain echo weighted echo-planar MR imaging. Radiology 1996;
planar diusion-weighted magnetic resonance 199: 391401.
imaging. Ann Neurol 1995; 37: 231241. 27 Lutsep HL, Albers GW, DeCrespigny A, Kamat GN,
14 Parsons MW, Barber PA, Darby DG, Tress BM, Donnan Marks MP, Moseley ME. Clinical utility of diusion-
GA, Davis SM. Hyperacute diusion- and perfusion- weighted magnetic resonance imaging in the assess-
weighted MRI identied t-PA responders in stroke. ment of ischemic stroke. Ann Neurol 1997; 41: 574580.
Ann Neurol 2002; 51: 2837. 28 Baird AE, Warach S. Magnetic resonance imaging of
15 Wardlaw JM, Sandercock PA, Warlow CP, Lindley RI. acute stroke. J Cereb Blood Flow Metab 1998; 18:
Trials of thrombolysis in acute ischemic stroke: does 583609.
the choice of primary outcome measure really matter? 29 Keir SL, Wardlaw JM. Systematic review of diusion
Stroke 2000; 31: 11331135. and perfusion imaging in acute ischemic stroke.
16 Barber PA, Demchuk AM, Zhang J, Buchan AM. Stroke 2000; 31: 27232731.
Validity and reliability of a quantitative computed 30 Powers WJ, Zivin J. Magnetic resonance imaging in
tomography score in predicting outcome of hyper- acute stroke: not ready for prime time [editorial;
acute stroke before thrombolytic therapy. ASPECTS comment]. Neurology 1998; 50: 842843.
Study Group. Alberta Stroke Programme Early CT 31 Bryan RN. Diusion-weighted imaging: to treat or not
Score. Lancet 2000; 355: 16701674. to treat? That is the question. Am J Neuroradiol 1998;
17 Moulin T, Cattin F, Crepin-Leblond T et al. Early CT 19: 396397.
signs in acute middle cerebral artery infarction: pre- 32 Warach S, Boska M, Welch KM. Pitfalls and potential
dictive value for subsequent infarct locations and of clinical diusion-weighted MR imaging in acute
outcome. Neurology 1996; 47: 366375. stroke. Stroke 1997; 28: 481482.
18 Wardlaw JM, Lewis SC, Dennis MS, Counsell C, 33 Lovblad KO, Laubach HJ, Baird AE et al. Clinical expe-
McDowall M. Is visible infarction on computed rience with diusion-weighted MR in patients with
tomography associated with an adverse prognosis in acute stroke [see comments]. Am J Neuroradiol 1998;
acute ischemic stroke? Stroke 1998; 29: 13151319. 19: 10611066.
188 Stephen Davis and Mark Parsons

34 Minematsu K, Li L, Sotak CH, Davis MA, Fisher M. 48 Edelman RR, Siewert B, Darby DG et al. Qualitative
Reversible focal ischemic injury demonstrated by mapping of cerebral blood ow and functional local-
diusion-weighted magnetic resonance imaging in ization with echo-planar MR imaging and signal tar-
rats. Stroke 1992; 23: 13041310; discussion geting with alternating radio frequency. Radiology
13101311. 1994; 192: 513520.
35 Kidwell CS, Saver JL, Mattiello J et al. Thrombolytic 49 Rordorf G, Koroshetz WJ, Copen WA et al. Regional
reversal of acute human cerebral ischemic injury ischemia and ischemic injury in patients with acute
shown by diusion/perfusion magnetic resonance middle cerebral artery stroke as dened by early
imaging. Ann Neurol 2000; 47: 462469. diusion-weighted and perfusion-weighted MRI.
36 Baird AE, Beneld A, Schlaug G et al. Enlargement of Stroke 1998; 29: 939943.
human cerebral ischemic lesion volumes measured by 50 Lovblad KO, Baird AE, Schlaug G et al. Ischemic lesion
diusion-weighted magnetic resonance imaging [see volumes in acute stroke by diusion-weighted mag-
comments]. Ann Neurol 1997; 41: 581589. netic resonance imaging correlate with clinical
37 Barber PA, Davis SM, Darby DG et al. Absent middle outcome. Ann Neurol 1997; 42: 164170.
cerebral artery ow predicts the presence and evolu- 51 Tong DC, Yenari MA, Albers GW, OBrien M, Marks MP,
tion of the ischemic penumbra. Neurology 1999; 52: Moseley ME. Correlation of perfusion- and diusion-
11251132. weighted MRI with NIHSS score in acute (6.5 hour)
38 Warach S, Mosley M, Sorensen AG, Koroshetz W. Time ischemic stroke [see comments]. Neurology 1998; 50:
course of diusion imaging abnormalities in human 864870.
stroke [letter; comment]. Stroke 1996; 27: 12541256. 52 Darby DG, Barber PA, Gerraty RP et al.
39 Barber PA, Davis SM, Darby DG et al. Screening for Pathophysiological topography of acute ischemia by
thrombolytic therapy in acute stroke: diusion- combined diusion-weighted and perfusion MRI.
weighted imaging vs. computed tomography. Stroke 1999; 30: 20432052.
(Abstract). Cerebrovasc Dis 1999; 9: 73. 53 Fisher M. Diusion and perfusion imaging for acute
40 Wang PY, Barker PB, Wityk RJ, Ulug AM, van Zijl PC, stroke. Surg Neurol 1995; 43: 606609.
Beauchamp NJ. Diusion-negative stroke: a report of 54 Beaulieu C, de Crespigny A, Tong DC, Moseley ME,
two cases. Am J Neuroradiol 1999; 20: 18761880. Albers GW, Marks MP. Longitudinal magnetic reso-
41 Albers GW, Lansberg MG, Norbash AM et al. Yield of nance imaging study of perfusion and diusion in
diusion-weighted MRI for detection of potentially stroke: evolution of lesion volume and correlation
relevant ndings in stroke patients. Neurology 2000; with clinical outcome. Ann Neurol 1999; 46: 568578.
54: 15621567. 55 Barber PA, Consolo HK, Yang Q et al. Comparison of
42 Lansberg MG, Albers GW, Beaulieu C, Marks MP. MRI perfusion imaging and single photon emission
Comparison of diusion-weighted MRI and CT in computed tomography in chronic stroke. Cerebrovasc
acute stroke. Neurology 2000; 54: 15571561. Dis 2001; 11: 128136.
43 Lansberg MG, Norbash AM, Marks MP, Tong DC, 56 Neumann-Haefelin T, Wittsack HJ, Wenserski F et al.
Moseley ME, Albers GW. Advantages of adding Diusion- and perfusion-weighted MRI. The
diusion-weighted magnetic resonance imaging to DWI/PWI mismatch region in acute stroke. Stroke
conventional magnetic resonance imaging for evalu- 1999; 30: 15911597.
ating acute stroke. Arch Neurol 2000; 57: 13111316. 57 Ostergaard L, Sorensen AG, Kwong KK, Weissko RM,
44 Kidwell CS, Alger JR, Di Salle F et al. Diusion MRI in Gyldensted C, Rosen BR. High resolution measurement
patients with transient ischemic attacks. Stroke 1999; of cerebral blood ow using intravascular tracer bolus
30: 11741180. passages. Part II: Experimental comparison and pre-
45 Patel MR, Edelman RR, Warach S. Detection of hyper- liminary results. Magn Reson Med 1996; 36: 726736.
acute primary intraparenchymal hemorrhage by mag- 58 Neumann-Haefelin T, Wittsack HJ, Fink GR et al.
netic resonance imaging. Stroke 1996; 27: 23212324. Diusion- and perfusion-weighted MRI: inuence of
46 Linfante I, Llinas RH, Caplan LR, Warach S. MRI fea- severe carotid artery stenosis on the DWI/PWI mis-
tures of intracerebral hemorrhage within 2 hours from match in acute stroke. Stroke 2000; 31: 13111317.
symptom onset. Stroke 1999; 30: 22632267. 59 Parsons MW, Yang Q, Barber PA et al. Perfusion mag-
47 Kucharczyk J, Vexler ZS, Roberts TP et al. Echo-planar netic resonance imaging maps in hyperacute stroke:
perfusion-sensitive MR imaging of acute cerebral relative cerebral blood ow most accurately identies
ischemia. Radiology 1993; 188: 711717. tissue destined to infarct. Stroke 2001; 32: 15811587.
Clinical role of echoplanar MRI in stroke 189

60 Wu O, Koroshetz WJ, Ostergaard L et al. Predicting 72 Warach S, Pettigrew LC, Dashe JF et al. Eect of citico-
tissue outcome in acute human cerebral ischemia line on ischemic lesions as measured by diusion-
using combined diusion- and perfusion-weighted weighted magnetic resonance imaging. Citicoline 010
MR imaging. Stroke 2001; 32: 933942. Investigators. Ann Neurol 2000; 48: 713722.
61 Schellinger PD, Jansen O, Fiebach JB et al. Monitoring 73 Meadows ME, Fisher M, Minematsu K. Delayed treat-
intravenous recombinant tissue plasminogen acti- ment with a non-competitive NMDA antagonist, CNS-
vator thrombolysis for acute ischemic stroke with 1102, reduces infarct size in rats. Cerebrovasc Dis 1994;
diusion and perfusion MRI. Stroke 2000; 31: 4: 2631.
13181328. 74 Fisher M, Albers GW. Applications of diusion-perfu-
62 Marks MP, Tong DC, Beaulieu C, Albers GW, de sion magnetic resonance imaging in acute ischemic
Crespigny A, Moseley ME. Evaluation of early reperfu- stroke. Neurology 1999; 52: 17501756.
sion and i.v. tPA therapy using diusion- and perfu- 75 Adams HP, Jr., Brott TG, Furlan AJ et al. Guidelines for
sion-weighted MRI [see comments]. Neurology 1999; Thrombolytic Therapy for Acute Stroke: a Supplement
52: 17921798. to the Guidelines for the Management of Patients with
63 Lee LJ, Kidwell CS, Alger J, Starkman S, Saver JL. Acute Ischemic Stroke. A statement for healthcare
Impact on stroke subtype diagnosis of early professionals from a Special Writing Group of the
diusion-weighted magnetic resonance imaging and Stroke Council, American Heart Association. Stroke
magnetic resonance angiography. Stroke 2000; 31: 1996; 27: 17111718.
10811089. 76 Hacke W, Kaste M, Fieschi C et al. Randomised
64 Roh JK, Kang DW, Lee SH, Yoon BW, Chang KH. double-blind placebo-controlled trial of thrombolytic
Signicance of acute multiple brain infarction on therapy with intravenous alteplase in acute ischaemic
diusion-weighted imaging. Stroke 2000; 31: 688694. stroke (ECASS II). Second EuropeanAustralasian
65 Singer MB, Chong J, Lu D, Schonewille WJ, Tuhrim S, Acute Stroke Study Investigators [see comments].
Atlas SW. Diusion-weighted MRI in acute subcorti- Lancet 1998; 352: 12451251.
cal infarction [published erratum appears in 77 Hacke W, Kaste M, Fieschi C et al. Intravenous throm-
Stroke 1998 Mar; 29(3): 731]. Stroke 1998; 29: bolysis with recombinant tissue plasminogen activa-
133136. tor for acute hemispheric stroke. The European
66 Noguchi K, Nagayoshi T, Watanabe N et al. Diusion- Cooperative Acute Stroke Study (ECASS) [see com-
weighted echo-planar MRI of lacunar infarcts. ments]. J Am Med Assoc 1995; 274: 10171025.
Neuroradiology 1998; 40: 448451. 78 Grotta JC, Chiu D, Lu M et al. Agreement and variabil-
67 Oliveira-Filho J, Ay H, Schaefer PW et al. Diusion- ity in the interpretation of early CT changes in stroke
weighted magnetic resonance imaging identies the patients qualifying for intravenous rtPA therapy.
clinically relevant small-penetrator infarcts. Arch Stroke 1999; 30: 15281533.
Neurol 2000; 57: 10091014. 79 Albers GW. Expanding the window for thrombo-
68 Lindgren A, Staaf G, Geijer B et al. Clinical lacunar lytic therapy in acute stroke. The potential role of
syndromes as predictors of lacunar infarcts. A com- acute MRI for patient selection. Stroke 1999; 30:
parison of acute clinical lacunar syndromes and nd- 22302237.
ings on diusion-weighted MRI. Acta Neurol Scand 80 Davis SM, Tress B, Barber PA et al. Echoplanar mag-
2000; 2000: 2. netic resonance imaging in acute stroke. J Clin
69 Ay H, Oliveira-Filho J, Buonanno FS et al. Diusion- Neurosci 2000; 7: 38.
weighted imaging identies a subset of lacunar infarc- 81 Graham GD, Kalvach P, Blamire AM, Brass LM, Fayad
tion associated with embolic source. Stroke 1999; 30: PB, Prichard JW. Clinical correlates of proton mag-
26442650. netic resonance spectroscopy ndings after acute
70 Gerraty RP, Parsons M, Barber PA et al. Acute echopla- cerebral infarction. Stroke 1995; 26: 225229.
nar diusion and perfusion MRI improves diagnostic 82 Moonen CT, van Zijl PC, Frank JA, Le Bihan D, Becker
accuracy in subcortical cerebral infarction (Abstract). ED. Functional magnetic resonance imaging in medi-
Stroke 2000; Accepted for Publication. cine and physiology. Science 1990; 250: 5361.
71 Schonewille WJ, Tuhrim S, Singer MB, Atlas SW. 83 Graham GD, Blamire AM, Howseman AM et al. Proton
Diusion-weighted MRI in acute lacunar syndromes. magnetic resonance spectroscopy of cerebral lactate
A clinical-radiological correlation study. Stroke 1999; and other metabolites in stroke patients. Stroke 1992;
30: 20662069. 23: 333340.
190 Stephen Davis and Mark Parsons

84 Gideon P, Henriksen O, Sperling B et al. Early time netic resonance spectroscopy and initial infarct
course of N-acetylaspartate, creatine and phospho- volume by MRI predicts outcome in patients with
creatine, and compounds containing choline in middle cerebral artery territory infarction. Stroke
the brain after acute stroke. A proton magnetic 1999; 30: 15771582.
resonance spectroscopy study. Stroke 1992; 23: 91 Wardlaw JM, Marshall I, Wild J, Dennis MS, Cannon J,
15661572. Lewis SC. Studies of acute ischemic stroke with proton
85 Barker PB, Gillard JH, van Zijl PC et al. Acute stroke: magnetic resonance spectroscopy: relation between
evaluation with serial proton MR spectroscopic time from onset, neurological decit, metabolite
imaging. Radiology 1994; 192: 723732. abnormalities in the infarct, blood ow, and clinical
86 Higuchi T, Fernandez EJ, Maudsley AA, Shimizu H, outcome. Stroke 1998; 29: 16181624.
Weiner MW, Weinstein PR. Mapping of lactate and N- 92 Petro OA, Graham GD, Blamire AM et al.
acetyl--aspartate predicts infarction during acute Spectroscopic imaging of stroke in humans: histopa-
focal ischemia: in vivo 1H magnetic resonance thology correlates of spectral changes. Neurology
spectroscopy in rats. Neurosurgery 1996; 38: 121129; 1992; 42: 13491354.
discussion 129130. 93 Gillard JH, Barker PB, van Zijl PC, Bryan RN,
87 Petro OA, Ogino T, Alger JR. High-resolution proton Oppenheimer SM. Proton MR spectroscopy in acute
magnetic resonance spectroscopy of rabbit brain: middle cerebral artery stroke. Am J Neuroradiol 1996;
regional metabolite levels and postmortem changes. J 17: 873886.
Neurochem 1988; 51: 163171. 94 Parsons MW, Barber PA, Yang Q et al. Acute hyper-
88 Petro OA, Prichard JW, Ogino T, Shulman RG. Proton glycemia adversely affects stroke outcome: an MRI
magnetic resonance spectroscopic studies of agonal imaging and spectroscopy study. Ann Neurol 2002; 52:
carbohydrate metabolism in rabbit brain. Neurology 2028.
1988; 38: 15691574. 95 Wild JM, Wardlaw JM, Marshall I, Warlow CP. N-
89 Prichard JW. The ischemic penumbra in stroke: pros- acetylaspartate distribution in proton spectroscopic
pects for analysis by nuclear magnetic resonance images of ischemic stroke: relationship to infarct
spectroscopy. Res Publ Assoc Res Nerv Ment Dis 1993; appearance on T2-weighted magnetic resonance
71: 153174. imaging. Stroke 2000; 31: 30083014.
90 Pereira AC, Saunders DE, Doyle VL et al. Measurement
of initial N-acetyl aspartate concentration by mag-
The ischemic penumbra: the evolution of a concept

Georey A. Donnan, Peter M. Wright, Romesh Markus, Thanh Phan and David C. Reutens
National Stroke Research Institute, Heidelberg, Victoria, Australia

Definition Importance

In human ischemic stroke the aected brain tissue The fundamental importance of the concept of the
has metabolic requirements that can no longer be ischemic penumbra is the recognition that ischemic
supported by a sudden interruption, or at least processes may be reversible. In theory, this zone of
reduction, in nutrient supply and waste disposal functionally inactive but structurally intact tissue
resulting from disease in a proximal artery. The may, at any time point, be restored to functional nor-
underlying pathophysiological changes immediately mality. The best clinical validation of this principal is
following arterial occlusion include an initial loss of the observed correlation between the volume of sur-
neuronal electrical activity followed by depletion of viving penumbra and clinical outcome.4,5 The recog-
cellular energy stores and the loss of the transmem- nition of this relationship has stimulated research
brane ion pumps which maintain water, sodium, eorts to develop therapeutic approaches to increase
chloride and potassium balance. Cellular necrosis the proportion of salvageable tissue. Since clinicians
will follow these changes if the cells minimum can now detect penumbral tissue very early after
energy requirements are not met.13 The tissue sup- stroke onset with a variety of neuroimaging modali-
plied by the artery is not aected homogeneously by ties, particularly magnetic resonance imaging, there
this process, and critical blood ow thresholds asso- is great clinical potential for individualized patient
ciated with irreversible tissue infarction, functional selection for therapy. The ability to identify critical
tissue impairment, or benign oligemia can be thresholds for cerebral blood ow rates oers impor-
identied. Several methods of threshold detection tant potential advantages: reperfusion into penum-
using modern imaging techniques lead to diering bral tissue following the use of thrombolysis in stroke
interpretations of this in vivo, and the extensive mol- is associated with improved clinical outcome, com-
ecular and biochemical mechanisms in the penum- pared with a failure to reperfuse;69 those thresholds
bra can only partially be evaluated in human stroke which predict hemorrhagic risk, or lack of thrombol-
patients. The region of incomplete ischemia in ysis may be identied;8 and the lack of penumbral
stroke usually adjacent to the area of profound tissue due to early reperfusion may obviate the need
ischemia has been termed the ischemic penumbra. for therapy. Hence, this proven treatment may poten-
A useful functional denition of the penumbra is tially be enhanced using this model. The ischemic
that region of under-perfused brain tissue that is penumbra may also be useful as a surrogate target
metabolically impaired, classically showing electri- for clinical research into new therapies. This issue
cal inactivity, but with cellular morphology intact. In has become more important since many large phase
humans, this region is closely associated with estab- III trials of neuroprotection in acute ischemic stroke
lished critical thresholds for blood ow rates using have been negative, in spite of good evidence for
positron emission tomography (PET). ecacy in animal models.

192 Geoffrey A. Donnan, Peter M. Wright, Romesh Markus, Thanh Phan and David C. Reutens

Table 15.1. Flow thresholds in ml/100 g per min (% of reference) for irreversible damage and
penumbra from functional neuroimaging studies in acute stroke12

Modality Authors Damage Penumbra

PET Baron et al. (1984)13 12

Powers et al. (1985)14 12 22
Hakim et al. (1989)15 12 18
Furlan et al. (1996)16 7
Marchal et al. (1996)17 22
Heiss et al. (2000)12 4.8 (22) 14.1 (54)
MRI Schlaug et al (1999)18 6 (12) 18.5 (37)
SPECT Giubilei et al. (1990)19 (60) (70)
Shimosegawa et al. (1994)20 (48) (75)
Ezura et al. (1996)8 (35) (70)
Berrouschot et al. (1998)21 (0) (70)
Hatazawa et al. (1999)22 (39) (69)
Ueda et al. (1999)23 (35) (55)
Xe-CT Firlik et al. (1998)24 17.3 35.4
Kaufmann et al. (1999)25 6 20

PET, positron emission tomography; MRI, magnetic resonance imaging; SPECT,
single photo emission computed tomography; Xe-CT, Xenon-enhanced X-ray
computed tomography.

Flow thresholds for the ischemic penumbra region of benign oligemia (mildly reduced blood
ow) was not usually at risk of infarction. The accu-
The concept of the ischemic penumbra was pio- mulated experience from functional imaging
neered in baboon models of focal ischemia using studies in human stroke over the subsequent three
occluded middle cerebral arteries by Astrup, Siesjo decades has identied estimates of the upper and
and Symons.10,11 They identied three distinct lower critical ow thresholds to dene penumbral
regions of blood ow disruption. Regions of regions. These are shown in Table 15.1.
severely hypoperfused tissue (the ischemic core)
with cerebral blood ow less than 610 ml/100 g
per min were usually irreversibly damaged and The dynamic and time-related nature of the
exhibited features of cellular necrosis, such as an penumbra
eux of intracellular potassium. This region was
usually within the basal ganglia and anterior tem- Ischemia will inevitably occur in tissues dependent
poral region. Less profoundly hypoperfused on an acutely occluded large artery. Heiss and col-
penumbral tissue, usually more cortically located, leagues demonstrated in a primate model that a
had a CBF 20 ml/100 g per min, and demon- CBF of up to 14 ml/100 g/min for more than 45
strated functional electrical impairment with an minutes was associated with necrosis, but that even
isoelectric electroencephalograph and absent profound ow disturbances could be associated
evoked potentials but could potentially be salvaged with neuronal survival if reperfused within 20
if the middle cerebral artery occlusion was tran- minutes.26 In another model Morawetz and col-
sient. This region was not associated with early evi- leagues have demonstrated that blood ow below
dence of cellular necrosis.2 A third, peripheral 1012 ml/100 g/min for more than 2 hours was
The ischemic penumbra 193

3 hours. However, accumulating evidence that

No deficit some patients benet from intra-arterial, and pos-
Flow (ml /100 g per min)

sibly intravenous reperfusion treatments for up to 6

hours have limited the acceptance of this view.3133
In addition, Baron and colleagues studied eight
20 Paralysis patients between 717 hours from stroke onset
using positron emission tomography (PET), who
had both acute misery perfusion (see section on
10 Infarction PET), and a nal infarction of more than 16 mm
diameter. Eventually infarcted tissue with preserved
early CMRO2 values above the threshold for irre-
versible infarction were found in each of these eight
0 cases. This pattern constituted 10% to 52% of the
0 1 2 3 // Permanent
nal infarct volume. Several patients still had
Time (h)
penumbra at 1316 hours after stroke, and this late
Fig. 15.1. Developed from Jones et al.28 A monkey model of penumbral tissue made up on average 35% of the
the relationship between the cerebral blood ow in ml/100
nal infarct volume.34 Read and colleagues studied
g/min (y-axis), and the duration of ischemia in causing signs,
24 patients up to 51 hours after ischemic stroke
and tissue infarction. Reversible paralysis develops if the CBF
using 18F-uoromisonidazole (FMISO) as a ligand
falls below 23 ml/100 g/min. Even profound ischemia is toler-
that labels hypoxic but viable tissue in positron
able for very brief periods, and if CBF is below 10 ml/100 g
per min for 2 hours, or below 18 ml/100 g/min in the emission tomography. Penumbral tissue, labelled
setting of permanent occlusion, then irreversible tissue with FMISO, was found in 15 patients and the pro-
infarction occurs. portion of patients with penumbra, and penumbral
volume, decreased with time. On average, 45% of
the tissue that trapped FMISO survived. Up to 68%
consistently associated with tissue infarction,27 and (mean, 17.5%) of the infarct volume was initially
tissue with a CBF of 15 ml/100 g per min could be hypoxic. Most of the tissue initially aected pro-
supported for about 3 hours without infarction, ceeded to infarction.4 Darby and colleagues have
whilst tissue with a CBF of 5 ml/100 g/min only also shown the penumbra, as dened by a mis-
remained viable for 2 hours. The area of irreversibly match between diusion-weighted, and perfusion-
damaged tissue expanded at the expense of the imaging lesion volumes using MRI, gradually
penumbral region over time, leaving only a small declines with time up to about 18 hours following
region of penumbral tissue at 3 hours.28 Flow stroke onset. This is shown in Fig. 15.2.
thresholds appear to dier between animal species, In addition to the severity and duration of arterial
with rats needing over 25 ml/100 g per min,29 and occlusion, the availability of an alternate blood
cats 15 ml/100 g per min for tissue viability.30 Figure supply from collateral vessels,36 and environmental
15.1 demonstrates the relationship between the stressors such as hyperglycemia, arterial hypoxia
severity and duration of ow reduction and the ulti- and hyperthermia may play important roles in
mate fate of the tissue aected. determining the duration of tissue viability. Misery
Heiss and colleagues studied ten human patients perfusion, recurrent spreading depression, excito-
with positron emission tomography within 3 hours toxin release and acidosis all contribute to this
of stroke. In these, penumbral tissue consisted of dynamic state in the penumbra.37,38 Microvessel
only about 18% of the total tissue with reduced CBF, plugging by platelets, brin or neutrophils has been
while 12% was oligemic, and the remaining 70% found in baboon models and will aect the
was critically hypoperfused tissue at these early dynamic process of tissue reperfusion. This may
time points.31 This would imply that there is a low also occur even in the presence of early reopening
likelihood of signicant clinical recovery after of the diseased artery.12,34,3941
194 Geoffrey A. Donnan, Peter M. Wright, Romesh Markus, Thanh Phan and David C. Reutens

100 Spreading Selective

depression cell death

50 core

Non-penumbral HIF
Denatured c-fos, IEGs Hypoxia
0 6 12 18 24 Fig. 15.3. Schematic of multiple molecular penumbras after a
stroke. A zone of selective neuronal cell death borders the
Hours after stroke onset
infarct. The zone of protein denaturation extends outside of
Fig. 15.2. Plot of penumbral or non-penumbral patterns as a per- this and is demarcated by heat shock protein 70 (HSP70)
centage of total number of patients scanned within each period expression in injured neurons. Hypoxia inducible factor (HIF)
(03 hours, 06 hours, 612 hours, 1218 hours, 18 hours).35 is exposed in areas where blood ow is persistently decreased
and oxygen delivery is impaired. This may be coincident with
HSP 70 or extend over more widespread regions depending
on collaterals. Ischemia-induced spreading depression
induces c-fos and many other immediate early genes (IEGs)
The ischemic penumbra: a molecular view at some distances from the infarct, including the ipsilateral
rat occipital and frontal lobes, contralateral cortex and many
Although founded on the concept of critical subcortical structures. Derived from Sharp et al.42
changes of blood ow, the ischemic penumbra can
also be described in molecular terms. While this
concept is based on animal models of cerebral tumour necrosis factor (TNF) alpha and TNF recep-
ischemia, and no means of imaging molecular tors,46,47 transcription regulatory nuclear factor kB
changes are yet available in vivo, there are impor- (NFkB), and the targets for NFkB which include
tant implications for therapy in humans. Sharp and manganese (Mn2), superoxide dismutase,
colleagues described no less than ve distinct inducible nitric oxide synthetase (iNOS), matrix
regions associated with focal cerebral ischemia metalloprotease-9 (MMP-9) and cyclo-oxygenase-2
based on the molecular changes being exhibited.42 (COX-2).42 In addition, the expression of the anti-
apoptotic genes Bcl-2 and Bcl-xL is decreased
among neurons lethally injured, whereas they are
Defining the ischemic core in molecular terms
increased in mildly compromised neurons within
Protein synthesis is inhibited by the release of glu- the penumbral region.
tamate, and energy depletion in ischemia, decreas-
ing by about 50% at a CBF of 55 ml/100 g per min
Defining the ischemic penumbra in molecular terms
and ceasing almost completely below 35 ml/100
g/min.42,43 A molecular delineation of the ischemic Outside the core of the infarct, a series of four
core employs analyses of appropriate proteins, further regions can be dened in molecular terms.
many of which have a short half-life, and hence Given that there is potential reversibility of neu-
rapid reductions in concentration. These include ronal impairment in each zone, the term penum-
intercellular adhesion molecule-1 (ICAM-1 or bra is applicable for each. As can be seen in Fig.
CD54),44 matrix metalloproteases (MMPs),45 15.3, the rst topographical area is immediately
The ischemic penumbra 195

adjacent to the core of the infarct, and extends for tional ischemic penumbra for every patient. An
no more than one centimetre. Histologically, selec- operational denition requires information which
tive neuronal death has been documented.48 will allow identication of the upper and lower
The next zone is an area of protein denaturation, tissue CBF thresholds for penumbra. These must
and is manifest by the continued expression of the correctly identify the ischemic core, the potentially
heat shock protein 70 (HSP70). Induction of the viable tissue, and exclude the benign oligemic
HSP70 gene, together with other heat shock chap- tissue, by ensuring the tissue involved is clinically
erones, acts to renature the aected proteins. While functionally impaired. In humans, this denition is
these changes may be expressed on blood vessels, it dicult to achieve and alternative means of valida-
is their neuronal expression which is important in tion of penumbral imaging techniques are needed.
this potentially viable area surrounding an infarct.42 A pragmatic approach adopted by a number of
The next zone is demarcated by the presence of the investigators is to establish a positive correlation
transcription factor, hypoxia inducible factor 1 between the amount of putative penumbral tissue
(HIF-I). This is induced by changes in molecular that survives and an improved clinical outcome.4,5
oxygen levels in tissue.49 It has been suggested that Another approach is to regard the PET analysis of
this region may represent an area of chronically cerebral blood ow (CBF), regional cerebral meta-
depleted blood ow and, hence, hypoxia surround- bolic rate of oxygen (CMRO2), oxygen extraction
ing infarcts.42 Although it has been shown that HIF- fractions (OEF), as the current gold standard, and
I and related transcription factors may increase performing comparative studies against newer
blood ow, glucose delivery and energy mainte- techniques. However, this approach is logistically
nance under hypoxic conditions, their role is still dicult and is yet to be implemented satisfactorily.
unclear.42 The nal zone (not penumbral by stan-
dard blood ow denitions) is far more widespread,
Positron emission tomography (PET)
and represents areas aected by spreading depres-
sion. This phenomenon induces a number of inter- O PET
mediate early genes (IEGs), particularly c-fos, over Multitracer PET imaging with 15O allows generation
the entire hemisphere, and some may play a role in of quantitative brain maps for CBF, CMRO2, OEF,
protecting the brain against ischemic insults.50 The cerebral blood volume (CBV), and regional cerebral
delineation of these ve zones (one core, and four metabolic rate of glucose. Analysis of data from
penumbral) is extremely useful, since many of several small human stroke studies performed at
these molecular changes represent potential targets 548 h after onset of symptoms allows estimation of
for therapy. It should be said that many of these thresholds of ow and metabolic activity that can
multiple penumbras are still poorly understood in reasonably dierentiate the tissue subtypes of
molecular terms and further research may reveal ischemia, penumbra, oligemia and unaected
further avenues for penumbral salvage, or neuro- tissue at the time of the study. The penumbra is
protection. described by reduced CBF and increased OEF but
with a relatively preserved rCMRO2, indicating
stressed but viable tissue, a state described by
Ischemic penumbra: operational definitions in Baron as miserly perfusion.51 While regions with
humans CMRO2 of less than 65 mol/100 g/min predict
infarction,13 those with an increased OEF 70%,
As outlined earlier, there are many concurrent vas- with CMRO2 relatively preserved despite reduction
cular, environmental and molecular processes in CBF are penumbral. Where the CBF is between
occurring in early stroke. Of the several imaging 12 and 22 ml/100 g per min, the tissue outcome
modalities that oer potential for the assessment of depends on the duration of ischemia, and other
the penumbra in human patients, each assesses undetermined factors,15 although the outcome is
only a portion of these events. None of the available often infarction unless the CMRO2 derangements
modalities oers an exact denition of the func- are particularly mild.16,52 Heiss and colleagues
196 Geoffrey A. Donnan, Peter M. Wright, Romesh Markus, Thanh Phan and David C. Reutens

found that 95% of regions with CBF 4.8 ml/100 g venous contrast to generate angiographic images,
per min became infarcted, but 95% of regions with then in the tissue phase of contrast passage to deter-
CBF 14.1 ml/100 g/min escaped infarction.12 The mine semiquantitative cerebral blood volume analy-
most profound drops in CMRO2 occur in the deep sis.62 The quantitated CBF rates generated do not
MCA territory, consistent with relative preservation correlate well with CBF values determined by PET
of cortical tissue at follow-up.5356 Baron reported and thresholds for dening the ischemic penumbra
for the rst time in humans a well validated opera- have not been identied. Koenig and colleagues
tional model of the ischemic penumbra. Using 30 analysed the relative comparison to the contralateral
patients imaged 518 hours after stroke onset, they hemisphere, semiquantitate CBF, CBV, and time to
identied the ischemic core as tissue with CMRO2 peak perfusion maps in stroke. They found CBF
of metabolic thresholds with multitracer PET values of around 34% compared with 62%, and CBV
studies or MRI have not been established. Limited values of 43% and 62% respectively, for areas of infarc-
validation has been reported correlating surviving tion and areas suering from reversible ischemia.
penumbral tissue with outcome.4 Reversible ischemia was not operationally dened to
be consistent with ischemic penumbra in this study.
While perfusion CT is still restricted to a single slice,
Single photon emission computed tomography
multislice technology is currently being developed.63

Areas of relatively reduced CBF can be determined

Xenon-enhanced CT
using 99mTechnetium-hexamethylpropylene-amino-
oxime (HMPAO), or 99mTechnetium-ethyl-cysteinate This uses stable xenon gas and has not been taken
dimer (ECD). The latter also contribute limited meta- up widely, probably due to diculties managing
bolic information regarding the tissue. Volumes of the gas itself, and to its side eects when used.
reduced, and profoundly reduced CBF in acute Quantitative CBF analysis has been possible in a
stroke can be demonstrated using SPECT. It can be baboon stroke model, essentially validating it as a
compared with a normal population, or a region of technique, but nding lower CBF values than are
interest from the contralateral hemisphere, but an usually documented with PET studies.64 Studies in
accurate description of penumbra is dicult, as the acute human stroke within 6 hours of onset have
regions of ischemic core, and of oligemia are not correlated permanent tissue infarction with CBF
identied well. A number of investigators have values below 20 ml/100 g per min. The most pro-
demonstrated that severe, large regions of reduction found middle cerebral artery strokes had a mean
in uptake of these radioactive labels is associated CBF of 8.6 ml/100 g per min in contrast to a mean
with poor clinical outcome,19,5759 and normal or of 18 ml/100 g per min in those with milder strokes.
increased uptake is predictive of good outcome.57,21,60 Profound reductions in CBF to 6 ml/100 g/min
The relative CBF reductions that predicted the region closely predicted tissue infarction, and were associ-
of infarction at follow-up ranged from no ow up to ated with a small rim of tissue with CBF values of
60% ow as shown in Table 15.2. The range of ow 720 ml/100 g/min.25 Stroke patients had a mean
reduction, as represented by reduced tissue label CBF in areas of ischemia of 17.3 ml/100 g/min
uptake, found to represent tissue that, although compared with 35.4 ml/100 g/min for those with
ischemic, has a moderate chance of escaping infarc- reversible ischemic events.24
tion is between 40 and 70%.8,9,19,21,57,60,61

Magnetic resonance diffusion-weighted and

Computerized tomography (CT) perfusion
perfusion imaging of the penumbra

This new modality has become practicable with Evolving MRI techniques are useful for assessment
helical CT scanners, which are able to track intra- of penumbral tissue in acute stroke. The blood ow
The ischemic penumbra 197

Table 15.2. Operational definitions of the ischemic penumbra in humans: clinically validated definitions of tissue at risk of infarction
but not in the ischemic core. For definitions of the penumbra, evidence is required that the tissue is functionally impaired due to
hypoperfusion, but has not already been recruited into the irreversibly damaged ischemic core

Imaging Denition penumbra Denition core Advantages Disadvantages

SPECT Relative CBF 4070% Relative CBF 060% Cheap Poorly denes core
Available Poor agreement about
Time delays
PET FMZ 15-O PET CBF/FMZ <3.5 times Not time dependent Not clinically validated
mismatch or 2 minute contralateral FMZ No FMZ penumbra
FMZ binding binding denition
No subcortical data
PET F-MISO 18-F MISO trapping Central void uptake Biochemical denition Poor validation
Direct, simple measure Not clinically applicable
PET (15-O) CBF 722 mla CBF <12 mla Reproducible Limited availability
CMRO2 >54 picoMa CMRO2 <65 picoMa Clinically validated High cost
Grey CMRO2 >1.4 mla Grey CMRO2 <1.4 mla Slow technique
OEF > 70% Lower OEF
CT perfusion Relative CBF 3462% Relative CBF <34% Cheap, available, fast Poor PET correlation
Not quantied
Not operationally valid
Xe-CT 720mla CBF <6 mla Increasingly available Poor validation
Quantitation Technically problematic
Anesthetic side effects
MRI DWI/PI PW/DWI mismatch with:
Relative CBV to 137% Relative CBV to 76% Increasingly available Validation retrospective
Relative CBF to 37% Relative CBF to 11% Fast No external validation
Relative ADC to 56% Relative ADC to 91% Clinically feasible Not quantitated
Relative MTT to 165% Oligemia not excluded
Reversible DWI include

SPECTsingle photon emission computed tomography; CBFcerebral blood ow; PETpositron emission tomography;
FMZ 11C umazenil ligand; F-MISO 18Fluoromisonidazole ligand; CMRO2 cerebral metabolic rate of oxygen;
OEFoxygen extraction fraction; CTcomputed tomography; Xe-CTstable xenon CT; MRImagnetic resonance
imaging; DWIdiffusion-weighted imaging; PIperfusion imaging.
ml/100 g tissue per minute

and metabolic imaging methods have the advan- marker of the ischemic penumbra.12 The diusion-
tages of higher spatial resolution and speed, when weighted lesion loosely represents the ischemic
compared to the other imaging modalities that can core, and the various potential measures of relative
assess the ischemic penumbra, and can easily be or semiquantitative cerebral blood ow mark the
combined with high resolution anatomic, and arter- disturbance in perfusion of the tissue. Commonly,
ial imaging in a single imaging session. The rapid in the early stroke setting the ow disturbance is
and clinically available assessment of mismatch contiguous to the diusion lesion, and the area of
between diusion-weighted imaging lesion volume, mismatch between these two tissue volumes is
and the often larger perfusion abnormality is a taken to represent the ischemic penumbra.
198 Geoffrey A. Donnan, Peter M. Wright, Romesh Markus, Thanh Phan and David C. Reutens

vasogenic edema from as early as 6 hours from

Diffusion-weighted MRI (DWI)
stroke onset.70 There is also a correlation between
DWI is increasingly available in the setting of acute the ADC value and cerebral blood ow during
stroke, and for rapid acquisition is performed aterial occlusion.71 The mean acute relative ADC
using echoplanar magnetic resonance imaging reduction in a visible DWI lesion is 58% when com-
methods. This technique is able to detect changes pared to the contralateral hemisphere. In animal
in tissue water motion shortly after a focal models this should correspond to a CBF of 1015
ischemic insult as mentioned in an earlier chapter. ml/100 g per min together with energy failure.18 It
These changes are thought to reect intracellular has also been found that ADC abnormalities may be
water accumulation (cytotoxic edema) related to seen at the periphery of a lesion even before ATP
high-energy metabolism failure, hence accumula- energy stores have been fully depleted, which sug-
tion of osmotically signicant byproducts, and loss gests that some of the ADC-detected lesion may be
of ion homeostatic mechanisms.65 This allows very penumbral rather than representing the infarct
early and sensitive imaging of the damaged tissue core.65 In addition, the reduction of ADC values cor-
in stroke. There are, however, regions where these relate with the severity of ischemia,65 and are more
DW images are contaminated by signicant sus- profoundly decreased toward the core of the ADC
ceptibility artefacts causing anatomic distortions lesion than in the periphery.72 They become more
to the image that are most severe in the posterior severe with longer duration of ischemia, are less
fossa, the anterior and inferior temporal lobes, and intense in reversible than in irreversible areas,73 and
the paranasal sinuses. In addition, these regions after transient arterial occlusion diusion changes
often appear bright due to artefact, making inter- only predict infarction if severely decreased.74
pretation of similarly bright adjacent areas of Kidwell and colleagues also found that the MRI of
tissue damage dicult. The apparent diusion patients with transient ischemic attacks typically
coecient (ADC) map is a quantitated measure of shows less severely reduced ADC values than
water molecule diusion, and is able to take into patients with strokes, even though all ADC values
account errors in detecting ischemic damage, par- identied were moderateseverely reduced (less
ticularly eliminating any unwanted T2 signal which than 545 m2/s).75 Hasegawa and colleagues found
may be visible in the DW images (T2 shine-through that potential ADC thresholds exist that may sepa-
eect). rate reversible from irreversible abnormalities.76 A
A number of investigators using animal models of schematic representation of this penumbral compo-
cerebral ischemia, have shown close relationships nent of the diusion lesion is shown in Fig. 15.4. If
between the tissue with reduced apparent diusion the ischemic insult is short lived, even severe ADC
coecient, tissue with abnormal pH, ATP, and reductions are known to be potentially reversible in
glucose66 and tissue that later is found to have humans. Doege and colleagues reported six patients
histopathological damage.66,67 There is also an evo- with small subcortical grey matter ischemia in
lution in the ADC during prolonged tissue ischemia whom a large portion of the ADC lesion was
that plays a role in tissue outcome.68 A reduction in reversible, as judged from the lesion size on nal T2
ADC values has been demonstrated within minutes image.77 Recovery of tissue showing DWI changes
of arterial occlusion in animal models. The severity has also been reported by others,78 and was found in
of reduction in ADC tends to increase over the next nearly half of patients with TIA who initially pre-
2448 hours, during which time tissue water sented with DWI changes.75 Ueda and colleagues
content begins to increase and is detected on T2- also found that in the rst 72 hours of ischemic
weighted images.69 The ADC value will then increase stroke the ADC map overestimated the nal infarc-
through a phase of pseudo-normalization to the tion size by 182%, although this subacute assess-
increased levels seen in later cellular necrosis. A ment has little bearing on the acute management of
pattern of increasing size of the ADC lesion over 24 stroke.79 ADC values are continuous rather than
hours may be due to both cytotoxic edema and to simply normal or abnormal, and further research is
The ischemic penumbra 199

required to identify more precise thresholds of ADC have identied at-risk tissue, and a reduction to less
abnormality that predict the ischemic core and than 12% of rCBF, or 19% of rCBV identied the
perhaps the penumbra, and how these values inter- core of tissue infarction.18 The severity of clinical
act with the duration of regional ischemia. presentation84 and later outcome of stroke has been
found to correlate with the volume of tissue with
abnormal perfusion acutely. In Fig. 15.5 it can be
Perfusion MRI
seen how relatively more straightforward, and sen-
The use of MRI to detect cerebral perfusion at the sitive it is to measure volumes based on the MTT
capillary level in acute stroke is an evolving technol- and time-to-peak (TTP) maps compared with the
ogy, and at present has the best spatial resolution of CBV and CBF maps when assessing the penumbra.
any imaging technique for this purpose. An An important issue to consider when using per-
injection-free arterial spin labelling technique is in fusion MRI is that the exact critical penumbral
early stages of development and has been discussed thresholds of hypoperfusion are not reliably
in an earlier chapter. The most commonly used identied in any one patient.85 Contributing to
technique employs the injection of a paramagnetic errors in assessment is the diculty with identify-
contrast agent intravenously that will cause T2 ing and deconvolving an arterial input function
shortening, and dynamically tracking the signal from the tissue signal, delay and dispersion of the
changes as the contrast passes through the brain. A tissue impulse function, limitations to the correla-
T2-weighted echoplanar imaging sequence is best tion between gadolinium concentration and the
suited as it is fast, and allows serial scans every 12 perfusion-weighted signal intensity, anatomic
seconds.80,81 This technique allows generation of warping and susceptibility artefact using the echo-
time-contrast concentration curves based on the planar MRI techniques, and abnormalities in perfu-
assumption that contrast concentration is propor- sion in the patients contralateral hemisphere due
tional to the amount of signal loss in the tissue over to prior strokes or diaschisis. Attempts to correlate
the range of concentrations being observed. the MRI determined blood ow indices with the
Using this approximation of tissue gadolinium more established PET methods are ongoing.
concentration in brain tissue, a pixel by pixel analy- Absolute quantitation of CBF for MRI, independent
sis of blood ow, blood volume and mean transit of the PET gold standard, and validating the results
time can be determined wherever the signal to by operationally dening the tissue compartments
noise characteristics are satisfactory. It has been have not yet been possible. It is known that
estimated that in an acute stroke, tissue that has a vasoregulatory mechanisms are important in com-
73% increase in the tissue mean transit time (MTT), pensation for acute reductions in perfusion pres-
and a 29% increase in CBV as compared to the con- sures, and that ow heterogeneity is reduced in
tralateral hemisphere is at risk of infarction.82 It has ischemic areas, thus increasing the mean capillary
also been shown that MTT and CBF brain maps are concentrations of vital energy and nutrient sub-
larger than cerebral blood volume (CBV) maps typi- stances. In humans, there has been some prelimi-
cally, and that these markers of abnormal blood nary work to show that areas of prolonged MTT can
ow are usually more extensive than the diusion be further analysed to determine areas where the
image abnormality.83 The pathophysiological heterogeneity of MTTs is reduced. This implies a
responses to obstructed blood supply to a tissue are maximal capillary dilatation response to ischemic
complex, and the ability to use MRI markers of stress and may identify areas at greater risk of
blood ow disturbance are limited by a lack of infarction.86 Similarly, the CBV is initially increased
absolute quantitation, and hence the severity of in ischemic stroke.18 These ndings are in agree-
abnormality is measured as a relative disturbance ment with PET evidence of an early compensatory
as compared to the patients contralateral (pre- vasodilatory response to reduced perfusion pres-
sumed normal) hemisphere. On this basis, a reduc- sures which results in a relatively maintained CBF
tion of relative CBF (rCBF) to 37%, and rCBV to 47% despite prolongation of MTT.14,87
200 Geoffrey A. Donnan, Peter M. Wright, Romesh Markus, Thanh Phan and David C. Reutens

Benign oligemia have recently reported in abstract form their eval-

often seen in Irreversible, severe uation of perfusion and diusion MRI for predict-
MTT and TTP maps ADC changes
ing hemorrhagic transformation after
intra-arterial rt-PA thrombolysis.90 They found
that mean ADC values of 583 m2/s (in the typical
Thresholds range for irreversible infarction), and MTT values
uncertain of 38 were statistically correlated with hemor-
rhagic transformation, and 746 m2/s and 25 pre-
dicted infarction, whereas 874 m2/s and 17 were
Mismatch Milder, reversible correlated with salvage. Therefore, this describes
penumbra ADC changes a model wherein the most sluggish blood supply,
Fig. 15.4. Proposed schematic of the layers that are present and most damaged tissue is that most likely to
within the region of focal cerebral ischemia, as identied hemorrhage following thrombolysis, and the least
using a diusion/perfusion mismatch in MR imaging in sluggish ow, and least impaired tissue was
humans. If standard imaging techniques are used these layers potentially salvageable. It is of interest to note
are not yet readily identiable in patients with acute ischemic that ADC values of 746 m2/s are only mildly
stroke. abnormal, largely invisible to the eye, and only a
little below the 874 m2/s mean value for tissue
that survived. Using serial MRI around intra-arte-
The most sensitive MRI measure of abnormal rial thrombolysis they found that of the tissue
blood ow, is the time-to-peak (TTP) contrast con- showing initial DWI changes 41% remained
centration. This is known to routinely identify injured at 4 hours after thrombolytic treatment
tissue that is not at risk of infarction, because it (mean ADC was 624 m2/s), 18% showed early
includes tissue that is merely oligemic, rather than reversal of DWI changes but recurrence at 7 days
truly penumbral in the functional sense. This is rep- (mean ADC was 641 m2/s), and 33% showed
resented as benign oligemia in Fig. 15.4, and it can sustained reversal of the DWI lesion (mean ADC
be seen in Fig. 15.5 that the TTP and MTT lesion was 677 m2/s).
size is greater than the CBF and CBV lesions.
However, even using this sensitive marker the
appearance of a mismatch between the usually irre- Magnetic resonance imaging: proton
versibly damaged DWI lesion core and a larger sur- spectroscopy
rounding time-to-peak perfusion abnormality has
been found to correlate with enlargement of the
Proton magnetic resonance spectroscopy
infarction over time.88 In addition, spontaneous
early reperfusion of the hypoperfused area of mis- This evolving technique allows the measurement of
match is a clinically favourable nding.89 It is not several brain metabolites, either with single or mul-
yet clear what combinations of CBV, CBF, MTT and tiple voxel studies (chemical shift imaging).91,92 N-
ADC abnormalities should be used for each individ- acetyl aspartate (NAA) is a putative marker of
ual patient to correctly dene the ischemic penum- neuronal viability and reduces with neuronal loss.93
bra using MRI. Hyperacute changes in NAA concentration may
Imaging the ischemic penumbra is hoped to even occur within 3060 minutes of experimental
rene the patient selection for the use of throm- infarction.94 Lactate, a metabolite that is not usually
bolysis in acute stroke. In theory, the physician detectable in brain tissue, is a byproduct of anaero-
might look for patterns where signicant bic metabolism, particularly in ischemia.95 Data
amounts of penumbral tissue exist, where the from animal studies are consistent with a CBF
ADC values are normal, yet perfusion is threshold of 20 ml/100 g per minute required for
signicantly reduced. Kidwell and colleagues the production of lactate, and that this occurs
The ischemic penumbra 201

Fig. 15.5. Example from a 36-year-old female with complete absence of ow signal in the right middle cerebral artery (MCA) and
internal carotid artery (ICA) occlusion. The mean transit time (MTT), and time-to-peak (TTP) maps show perfusion abnormality
that is far more extensive than the apparent diusion coecient of water map (ADC) lesion and the nal infarct size. Cerebral
blood volume (CBV) and cerebral blood ow (CBF), whilst harder to visualize, are intermediate in size. Acute T2-weighted,
diusion weighted (DWI), and ADC images, with the follow-up T2 imaging are shown in the second row. Acute and follow-up mag-
netic resonance angiography (MRA) show absent ow in the right ICA and MCA. White arrows indicate absent middle cerebral
arteries. Black arrows indicate relevant ischemic changes.

within minutes of such a compromised state.96 It Future directions

has been postulated that lactate elevation may be a
marker of the ischemic penumbra in stroke, and in The combination of diusion-weighted imaging
animal studies high levels of lactate have been cor- and perfusion imaging represent the future of
related with worse outcome, and with greater penumbral imaging as it is readily available in most
reductions in NAA.92 centres and can be applied along with other
202 Geoffrey A. Donnan, Peter M. Wright, Romesh Markus, Thanh Phan and David C. Reutens

hemorrhage sensitive sequences as a rst line tool 3 Branston N, Strong A, Symon L. Extracellular potas-
in the evaluation of patients with sudden neurolog- sium activity, evoked potential and tissue blood ow.
ical decits. The addition of proton/phosphorus/ Relationships during progressive ischemia in baboon
sodium multivoxel MR spectroscopy can help with cerebral cortex. J Neurol Sci 1977; 32: 305321.
further delineation of the ischemic core and 4 Read S, Hirano T, Abbott D et al. The fate of hypoxic
tissue on 18F-uoromisonidazole positron emission
penumbra. However, the prolonged imaging time
tomography after ischemic stroke. Ann Neurol 2000;
for multivoxel MR spectroscopy even for a single
48: 228235.
slice acquisition, substantial postprocessing, and
5 Marchal G, Benali K, Iglesias S, Viader F, Derlon JM,
lack of availability makes this tool likely to be Baron JC. Voxel-based mapping of irreversible
limited to research units and not yet applicable for ischemic damage with PET in acute stroke. Brain
phase II proof of concept studies unless advances 1999; 122: 23872400.
in time eciency occur. Emerging MR imaging 6 Herderschee D, Limburg M, van Royen EA, Hijdra A,
techniques in ischemic stroke evaluation is the use Buller HR, Koster PA. Thrombolysis with recombi-
of blood oxygen level dependent (BOLD) contrast, nant tissue plasminogen activator in acute ischemic
which up until now has been used mainly for stroke: evaluation with rCBF-SPECT. Acta Neurol
anatomical localization, neurocognitive research Scand 1991; 83: 317322.
and mechanism of recovery from ischemic stroke.97 7 Baird AE, Donnan GA, Austin MC, Fitt GJ, Davis SM,
McKay WJ. Reperfusion after thrombolytic therapy in
This technique depends on generation of contrast
ischemic stroke measured by single-photon emis-
by deoxyhemoglobin, which possesses paramag-
sion computed tomography. Stroke 1994; 25: 7985.
netic properties compared to oxyhemoglobin
8 Ezura M, Takahashi A, Yoshimoto T. Evaluation of
which is diamagnetic. The BOLD technique has regional cerebral blood ow using single photon
been used to dierentiate perfused and non-per- emission tomography for the selection of patients for
fused tissues (decreased BOLD signal) during local brinolytic therapy of acute cerebral embolism.
experimental ischemia in cats.98 Additionally, the Neurosurg Rev 1996; 19: 231236.
use of acetazolamide challenge or breath holding 9 Ryu YH, Chung TS, Yoon PH et al. Evaluation of
with BOLD techniques can be used to assess vascu- reperfusion and recovery of brain function before
lar reactivity to determine autoregulatory reserve and after intracarotid arterial urokinase therapy in
capacity such as in patients with occluded carotid acute cerebral infarction with brain SPECT. Clin Nucl
arteries.90,100 The recent description of quantitative Med 1999; 24: 566571.
10 Astrup J, Siesjo BK, Symon L. Thresholds in cerebral
assessment of cerebral blood ow, cerebral blood
ischemia the ischemic penumbra. Stroke 1981; 12:
volume and tissue oxygen extraction suggests some
potential for this technique in evaluating regions,
11 Symon L, Branston NM, Strong AJ, Hope TD. The
together along with diusion-weighted and perfu- concepts of thresholds of ischemia in relation to
sion imaging.101 brain structure and function. J Clin Pathol Suppl (R
Coll Pathol) 1977; 11: 149154.
12 Heiss WD. Ischemic penumbra: evidence from func-
tional imaging in man. J Cereb Blood Flow Metab
REFERENCES 2000; 20: 12761293.
13 Baron JC, Rougemont D, Soussaline F et al. Local
1 Astrup J, Blennov G, Nilsson B. Eects of reduced interrelationships of cerebral oxygen consumption
cerebral blood ow upon EEG pattern, cerebral and glucose utilization in normal subjects and in
extracellular potassium, and energy metabolism in ischemic stroke patients: a positron tomography
the rat cortex during bicuculline-induced seizures. study. J Cereb Blood Flow Metab 1984; 4: 140149.
Brain Res 1979; 177: 115126. 14 Powers WJ, Grubb RL, Jr., Darriet D, Raichle ME.
2 Astrup J, Symon L, Branston NM, Lassen NA. Cerebral blood ow and cerebral metabolic rate of
Cortical evoked potential and extracellular K and oxygen requirements for cerebral function and via-
H at critical levels of brain ischemia. Stroke 1977; 8: bility in humans. J Cereb Blood Flow Metab 1985; 5:
5157. 600608.
The ischemic penumbra 203

15 Hakim AM, Evans AC, Berger L et al. The eect of 27 Morawetz RB, DeGirolami U, Ojemann RG, Marcoux
nimodipine on the evolution of human cerebral FW, Crowell RM. Cerebral blood ow determined by
infarction studied by PET. J Cereb Blood Flow Metab hydrogen clearance during middle cerebral artery
1989; 9: 523534. occlusion in unanesthetized monkeys. Stroke 1978; 9:
16 Furlan M, Marchal G, Viader F, Derlon JM, Baron JC. 143149.
Spontaneous neurological recovery after stroke and 28 Jones TH, Morawetz RB, Crowell RM et al. Thresholds
the fate of the ischemic penumbra. Ann Neurol 1996; of focal cerebral ischemia in awake monkeys. J
40: 216226. Neurosurg 1981; 54: 773782.
17 Marchal G, Beaudouin V, Rioux P et al. Prolonged 29 Tyson GW, Teasdale GM, Graham DI, McCulloch J.
persistence of substantial volumes of potentially Focal cerebral ischemia in the rat: topography of
viable brain tissue after stroke: a correlative PETCT hemodynamic and histopathological changes. Ann
study with voxel-based data analysis. Stroke 1996; 27: Neurol 1984; 15: 559567.
599606. 30 Strong AJ, Venables GS, Gibson G. The cortical
18 Schlaug G,Beneld A, Baird AE et al. The ischemic ischemic penumbra associated with occlusion of
penumbra: operationally dened by diusion and the middle cerebral artery in the cat: 1. Topo-
perfusion MRI. Neurology 1999; 53: 15281537. graphy of changes in blood ow, potassium ion
19 Giubilei F, Lenzi GL, Di Piero V et al. Predictive value activity, and EEG. J Cereb Blood Flow Metab 1983; 3:
of brain perfusion single-photon emission computed 8696.
tomography in acute ischemic stroke. Stroke 1990; 31 Heiss WD, Thiel A, Grond M, Graf R. Which targets
21: 895900. are relevant for therapy of acute ischemic stroke?
20 Shimosegawa E, Hatazawa J, Inugami A et al. Stroke 1999; 30: 14861489.
Cerebral infarction within six hours of onset: predic- 32 Furlan A, Higashida R, Wechsler L et al. Intra-arterial
tion of completed infarction with technetium-99m- prourokinase for acute ischemic stroke. The PROACT
HMPAO SPECT. J Nucl Med 1994; 35: 10971103. II study: a randomized controlled trial. Prolyse in
21 Berrouschot J, Barthel H, Hesse S, Koster J, Knapp Acute Cerebral Thromboembolism. J Am Med Assoc
WH, Schneider D. Dierentiation between transient 1999; 282: 20032011.
ischemic attack and ischemic stroke within the rst 33 Fisher M, Baron JC. Which targets are relevant for
six hours after onset of symptoms by using 99mTc- therapy of acute ischemic stroke? Stroke 2000; 31:
ECD-SPECT. J Cereb Blood Flow Metab 1998; 18: 984986.
921929. 34 Baron J. Mapping the ischemic penumbra with PET:
22 Hatazawa J, Shimosegawa E, Toyoshima H et al. implications for acute stroke treatment. Cerebrovasc
Cerebral blood volume in acute brain infarction: a Dis 1999; 9: 193201.
combined study with dynamic susceptibility contrast 35 Darby DG, Barber PA, Gerraty RP et al.
MRI and 99mTc-HMPAO-SPECT. Stroke 1999; 30: Pathophysiological topography of acute ischemia by
800806. combined diusion-weighted and perfusion MRI.
23 Ueda T, Sakaki S, Yuh WT, Nochide I, Ohta S. Stroke 1999; 30: 20432052.
Outcome in acute stroke with successful intra- 36 Jacewicz M, Tanabe J, Pulsinelli WA. The CBF thresh-
arterial thrombolysis and predictive value of old and dynamics for focal cerebral infarction in
initial single-photon emission-computed spontaneously hypertensive rats. J Cereb Blood Flow
tomography. J Cereb Blood Flow Metab 1999; 19: Metab 1992; 12: 359370.
99108. 37 Fisher M, Garcia JH. Evolving stroke and the
24 Firlik AD, Rubin G, Yonas H, Wechsler LR. Relation ischemic penumbra. Neurology 1996; 47:
between cerebral blood ow and neurologic decit 884888.
resolution in acute ischemic stroke. Neurology 1998; 38 Takano K, Latour LL, Formato JE et al. The role
51: 177182. of spreading depression in focal ischemia
25 Kaufmann AM, Firlik AD, Fukui MB, Wechsler LR, evaluated by diusion mapping. Ann Neurol 1996;
Jungries CA, Yonas H. Ischemic core and penumbra 39: 308318.
in human stroke. Stroke 1999; 30: 9399. 39 Garcia JH, Liu KF, Yoshida Y, Chen S, Lian J. Brain
26 Heiss WD, Rosner G. Functional recovery of cortical microvessels: factors altering their patency after the
neurons as related to degree and duration of occlusion of a middle cerebral artery (Wistar rat). Am
ischemia. Ann Neurol 1983; 14: 294301. J Pathol 1994; 145: 728740.
204 Geoffrey A. Donnan, Peter M. Wright, Romesh Markus, Thanh Phan and David C. Reutens

40 Abumiya T, Fitridge R, Mazur C et al. Integrin 52 Heiss WD, Fink GR, Herolz K, Pietrzyk U, Wagner R,
alpha(IIb)beta(3) inhibitor preserves microvascular Wienhard K. Progressive derangement of periinfarct
patency in experimental acute focal cerebral viable tissue in ischemic stroke. J Cereb Blood Flow
ischemia. Stroke 2000; 31: 14021409; discussion Metab 1992; 12: 193203.
14091410. 53 Touzani O, Yound AR, Derlon JM et al. Sequential
41 Mori E, del Zoppo GJ, Chambers JD, Copeland BR, studies of severely hypometabolic tissue volumes
Arfors KE. Inhibition of polymorphonuclear leuko- after permanent middle cerebral artery occlusion. A
cyte adherence suppresses no-reow after focal cere- positron emission tomographic investigation in
bral ischemia in baboons. Stroke 1992; 23: 712718. anesthetized baboons. Stroke 1995; 26: 21122119.
42 Sharp FR, Lu A, Tang Y, Millhorn DE. Multiple molec- 54 Heiss W, Graf R, Fujita T et al. Early detection of irre-
ular penumbras after focal cerebral ischemia. J Cereb versible damaged ischemic tissue by umazenil
Blood Flow Metab 2000; 20: 10111032. positron emission tomography in cats. Stroke 1997;
43 Mies G, Ishimaru S, Xie Y, Seo K, Hossmann KA. 28: 20452051; discussion 20512052.
Ischemic thresholds of cerebral protein synthesis 55 Heiss W, Grond M, Thiel A et al. Permanent cortical
and energy state following middle cerebral artery damage detected by umazenil positron emission
occlusion in rat. J Cereb Blood Flow Metab 1991; 11: tomography in acute stroke. Stroke 1998; 29:
753761. 454461.
44 Mabuchi T, Kitagawa K, Ohtsuki T et al. Contribution 56 Heiss WD, Kracht LW, Thiel A, Grond M, Pawlik G.
of microglia/macrophages to expansion of infarction Penumbral probability thresholds of cortical
and response of oligodendrocytes after focal cerebral umazenil binding and blood ow predicting tissue
ischemia in rats. Stroke 2000; 31: 17351743. outcome in patients with cerebral ischemia. Brain
45 Heo JH, Lucero J, Abumiya T, Koziol JA, Copeland BR, 2001; 124: 2029.
del Zoppo GJ. Matrix metalloproteinases increase 57 Hanson S, Grotta J, Rhoades H et al. Value of single-
very early during experimental focal cerebral photon emission-computed tomography in acute
ischemia. J Cereb Blood Flow Metab 1999; 19: 624633. stroke therapeutic trials. Stroke 1993; 24: 13221329.
46 Zhai QH, Futrell N, Chen FJ. Gene expression of IL- 58 Laloux P, Richelle F, Jamart J, De Coster P, Laterre C.
10 in relationship to TNF-alpha, IL-beta and IL-2 in Comparative correlations of HMPAO SPECT indices,
the rat brain following middle cerebral artery occlu- neurological score, and stroke subtypes with clinical
sion. J Neurol Sci 1997; 152: 119124. outcome in acute carotid infarcts. Stroke 1995; 26:
47 Feuerstein GZ, Wang X, Barone FC. Inammatory 816821.
gene expression in cerebral ischemia and trauma. 59 Berrouschot J, Barthel H, von Kummer R, Knapp WH,
Potential new therapeutic targets. Ann NY Acad Sci Hesse S, Schneider D. 99m technetium-ethyl-
1997; 825: 179193. cysteinate-dimer single-photon emission CT can
48 Nedergaard M. Neuronal injury in the infarct border: predict fatal ischemic brain edema. Stroke 1998; 29:
a neuropathological study in the rat. Acta 25562562.
Neuropathol 1987; 73: 267274. 60 Marchal G, Bouvard G, Iglesias S et al. Predictive
49 Wang GL, Jiang BH, Rue EA, Semenza GL. Hypoxia- value of (99m) Tc-HMPAO-SPECT for neurological
inducible factor 1 is a basic-helixloophelix-PAS outcome/recovery at the acute stage of stroke.
heterodimer regulated by cellular O2 tension. Proc Cerebrovasc Dis 2000; 10: 817.
Natl Acad Sci USA 1995; 92: 55105514. 61 Hartmann A. Prolonged disturbances of regional
50 Matsushima K, Schmidt-Kastner R, Hogan MJ, cerebral blood ow in transient ischemic attacks.
Hakim AM. Cortical spreading depression activates Stroke 1985; 16: 932939.
trophic factor expression in neurons and astrocytes 62 Hunter GJ, Hamberg LM, Ponzo JA et al. Assessment
and protects against subsequent focal brain of cerebral perfusion and arterial anatomy in hyper-
ischemia. Brain Res 1998; 807: 4760. acute stroke with three-dimensional functional CT:
51 Baron JC, Bousser MG, Rey A, Guillard A, Comar D, early clinical results. Am J Neuroradiol 1998; 19:
Castaigne P. Reversal of focal misery-perfusion syn- 2937.
drome by extra-intracranial arterial bypass in hemo- 63 Koenig M, Kraus M, Theek C, Klotz E, Gehlen W,
dynamic cerebral ischemia. A case study with 15O Heuser L. Quantitative assessment of the ischemic
positron emission tomography. Stroke 1981; 12: brain by means of perfusion-related parameters
454459. derived from perfusion CT. Stroke 2001; 32: 431437.
The ischemic penumbra 205

64 Yonas H, Gur D, Claassen D, Wolfson SK, Jr., Moossy 76 Hasegawa Y, Fisher M, Latour LL, Dardzinski BJ,
J. Stable xenon-enhanced CT measurement of cere- Sotak CH. MRI diusion mapping of reversible and
bral blood ow in reversible focal ischemia in irreversible ischemic injury in focal brain ischemia.
baboons. J Neurosurg 1990; 73: 266273. Neurology 1994; 44: 14841490.
65 Hoehn-Berlage M, Eis M, Back T, Kohno K, Yamashita 77 Doege C, Kerskens C, Romero B et al. MRI of small
K. Changes of relaxation times (T1, T2) and apparent human stroke shows reversible diusion changes in
diusion coecient after permanent middle cere- subcortical gray matter. Neuroreport 2000; 11:
bral artery occlusion in the rat: temporal evolution 20212024.
regional extent, and comparison with histology. 78 Marks MP, de Crespigny A, Lentz D, Enzmann DR,
Magn Reson Med 1995; 34: 824834. Albers GW, Moseley ME. Acute and chronic stroke:
66 Kucharczyk J, Mintorovitch J, Asgari H, Moseley M. navigated spin-echo diusion-weighted MR
Diusion/perfusion MR imaging of acute cerebral imaging. Radiology 1996; 199: 403408.
ischemia. Magn Reson Med 1991; 19: 311315. 79 Ueda T, Yuh W, Maley J, Quets J, Hahn P, Magnotta V.
67 Back T, Hoehn-Berlage M, Kohno K, Hossmann K. Outcome of acute ischemic lesions evaluated by
Diusion nuclear magnetic resonance imaging in diusion and perfusion MR imaging. Am J
experimental stroke. Correlation with cerebral Neuroradiol 1999; 20: 983989.
metabolites. Stroke 1994; 25: 494500. 80 Sorensen AG, Buonanno FS, Gonzalez RG et al.
68 Jiang Q, Zhang Z, Chopp M et al. Temporal evolution Hyperacute stroke: evaluation with combined multi-
and spatial distribution of the diusion constant of section diusion-weighted and hemodynamically
water in rat brain after transient middle cerebral weighted echo-planar MR imaging. Radiology 1996;
artery occlusion. Neurol Sci 1993; 120: 123130. 199: 391402.
69 Moseley ME, Cohen Y, Mintorovitch J et al. Early 81 Warach S, Chien D, Li W, Ronthal M, Edelman RR.
detection of regional cerebral ischemia in cats: com- Fast magnetic resonance diusion-weighted imaging
parison of diusion- and T2-weighted MRI and spec- of acute human stroke. Neurology 1992; 42:
troscopy. Magn Reson Med 1990; 14: 330346. 17171723.
70 Loubinoux I, Volk A, Borredon J et al. Spreading of 82 Sorensen A, Gonzalez R, Copen W et al. Quantitation
vasogenic edema and cytotoxic edema assessed by of diusion/perfusion MRI mismatch in acute
quantitative diusion and T2 magnetic resonance human cerebral infarction. Stroke 1997; 28: 252.
imaging. Stroke 1997; 28: 419426; discussion 426427. 83 Sorensen AG, Copen WA, Ostergaard L et al.
71 Miyabe M, Mori S, van Zijl PC et al. Correlation of the Hyperacute stroke: simultaneous measurement of
average water diusion constant with cerebral blood relative cerebral blood volume, relative cerebral
ow and ischemic damage after transient middle blood ow, and mean tissue transit time. Radiology
cerebral artery occlusion in cats. J Cereb Blood Flow 1999; 210: 519527.
Metab 1996; 16: 881891. 84 Tong D, Yenari M, Albers G, OBrien M, Marks M,
72 Rother J, de Crespigny AJ, DArceuil H, Iwai K, Moseley M. Correlation of perfusion- and diusion-
Moseley ME. Recovery of apparent diusion weighted MRI with NIHSS score in acute (6.5 hour)
coecient after ischemia-induced spreading depres- ischemic stroke. Neurology 1998; 50: 864870.
sion relates to cerebral perfusion gradient. Stroke 85 Calamante F, Thomas DL, Pell GS, Wiersma J, Turner
1996; 27: 980986; discussion 986987. R. Measuring cerebral blood ow using magnetic res-
73 Minematsu K, Li L, Sotak C, Davis M, Fisher M. onance imaging techniques. J Cereb Blood Flow
Reversible focal ischemic injury demonstrated by Metab 1999; 19: 701735.
diusion-weighted magnetic resonance imaging in 86 Ostergaard L, Sorensen AG, Chesler DA et al.
rats. Stroke 1992; 23: 13041310. Combined diusion-weighted and perfusion-
74 Mintorovitch J, Moseley ME, Chileuitt L, Shimizu H, weighted ow heterogeneity magnetic resonance
Cohen Y, Weinstein PR. Comparison of diusion- and imaging in acute stroke. Stroke 2000; 31:
T2-weighted MRI for the early detection of cerebral 10971103.
ischemia and reperfusion in rats. Magn Reson Med 87 Baron JC, Frackowiack RS, Herholz K et al. Use of
1991; 18: 3950. PET methods for measurement of cerebral
75 Kidwell CS, Alger JR, Di Salle F et al. Diusion MRI in energy metabolism and hemodynamics in cere-
patients with transient ischemic attacks. Stroke 1999; brovascular disease. J Cereb Blood Flow Metab 1989;
30: 11741180. 9: 723742.
206 Geoffrey A. Donnan, Peter M. Wright, Romesh Markus, Thanh Phan and David C. Reutens

88 Barber P, Darby D, Desmond P et al. Prediction of 96 Allen K, Busza AL, Crockard HA et al. Acute cerebral
stroke outcome with echoplanar perfusion and ischemia: concurrent changes in cerebral blood ow,
diusion weighted MRI. Neurology 1998; 51: 418426. energy metabolites, pH, and lactate measured with
89 Barber P, Davis S, Infeld B et al. Spontaneous hydrogen clearance and 31P and 1H nuclear magnetic
reperfusion after ischemic stroke is associated with resonance spectroscopy. III. Changes following
improved outcome. Stroke 1998; 29: 25222528. ischemia. J Cereb Blood Flow Metab 1988;
90 Saver J, Kidwell C, Liebeskind D, Starkman S. (Abstr) 8: 816821.
Acute ischemic stroke trials. Stroke 2001; 32: 275278. 97 Neumann-Haefelin T, Moseley ME, Albers GW. New
91 Beauchamp NJ, Jr., Barker PB, Wang PY, van Zijl PC. magnetic resonance imaging methods for cere-
Imaging of acute cerebral ischemia. Radiology 1999; brovascular disease: emerging clinical applications.
212: 307324. Ann Neurol 2000; 47: 559570.
92 Gillard JH, Barker PB, van Zijl PC, Bryan RN, 98 de Crespigny AJ, Wendland MF, Derugin N, Vexler ZS,
Oppenheimer SM. Proton MR spectroscopy in acute Moseley ME. Rapid MR imaging of a vascular chal-
middle cerebral artery stroke. Am J Neuroradiol 1996; lenge to focal ischemia in cat brain. J Magn Reson
17: 873886. Imaging 1993; 3: 475481.
93 Saunders DE, Howe FA, van den Boogaart A, McLean 99 Kleinschmidt A, Steinmetz H, Sitzer M, Merboldt KD,
MA, Griths JR, Brown MM. Continuing ischemic Frahm J. Magnetic resonance imaging of regional
damage after acute middle cerebral artery infarction cerebral blood oxygenation changes under acetazo-
in humans demonstrated by short-echo proton spec- lamide in carotid occlusive disease. Stroke 1995; 26:
troscopy. Stroke 1995; 26: 10071013. 106110.
94 Penrice J, Cady EB, Lorek A et al. Proton magnetic res- 100 Kastrup A, Li TQ, Glover GH, Moseley ME. Cerebral
onance spectroscopy of the brain in normal preterm blood ow-related signal changes during breath-
and term infants, and early changes after perinatal holding. Am J Neuroradiol 1999; 20: 12331238.
hypoxia-ischemia. Pediatr Res 1996; 40: 614. 101 van Zijl PC, Ele SM, Ulatowski JA et al. Quantitative
95 Barker PB, Gillard JH, van Zijl PC et al. Acute stroke: assessment of blood ow, blood volume and blood
evaluation with serial proton MR spectroscopic oxygenation eects in functional magnetic reso-
imaging. Radiology 1994; 192: 723732. nance imaging. Nat Med 1998; 4: 159167.
New MR techniques to select patients for thrombolysis in
acute stroke

Vincent N. Thijs1, and Gregory W. Albers2

Department of Neurology, UZ Gasthuisberg, Leuven, Belgium
Stanford Stroke Center, Stanford University Medical Center, Palo Alto, CA 94304, USA

Introduction initial publication of the trial results, the authors

reported that the benet of tPA was not dierent
The landmark NINDS IV tPA trial showed that treat- between the patients treated within 090 minutes
ment with tPA within 3 hours of symptom onset and the patients treated between 90 and 180
improved neurologic impairment and functional minutes,1 although no adjustment for dierences in
outcome.1Despite an increase in symptomatic baseline stroke severity between the early and late
hemorrhages from 0.6% in the placebo treated treated patients was performed. A subsequent
group to 6.4% in the active treatment group, mor- report by the same investigators found that, after
tality was not increased and functional outcome adjustment for baseline stroke severity, favourable
improved, even among patients with the most outcomes were more frequently observed in the
severe strokes.2 patients who were enrolled earlier and that treat-
The NINDS trial was subsequently criticized ment during the last portion of the 3-hour time
because it subjected some patient groups who were window was associated with a condence interval
assumed to be unlikely to benet from thrombo- that crossed,14 suggesting that the benet might be
lytic agents to a potentially harmful treatment.3,4 reduced during this time period.
For instance, patients with proximal ICA-occlusions In North America, thrombolytic therapy with IV-
are unlikely to benet from IV-tPA.57 Similarly, it tPA is currently used in only about 16% of all
has been suggested that patients with small, deep ischemic stroke patients, despite major endorse-
infarcts should not receive thrombolytics because ments from the American Academy of Neurology,
of the generally favourable outcome and the pre- the American Heart Association and the Canadian
sumed dierent pathological processes underlying Stroke Consortium.1517 The low treatment rate is
small vessel occlusions.8 related to both the perceived risks associated with
In the NINDS trial, no patient subgroups could treatment and the small fraction of eligible patients
be identied that appeared not to benet from tPA. presenting to the hospital within the 3-hour time
However, in this trial stroke subtype was estab- window.1821 A greater number of patients could be
lished based on clinical impression. Many studies treated with tPA if the time window could be
suggest that the identication of stroke subtypes expanded to 6 hours after symptom onset. In recent
cannot be made reliably on clinical criteria and surveys, about 60% of stroke patients arrived in the
early CT alone.913 emergency department within 6 hours of symptom
In the NINDS trial, the only factor that appeared onset.22,23
to improve the favourable response to tPA was the Trials that have attempted to expand the time
time between symptom onset and treatment. In the window of treatment with IV tPA to ve or 6 hours

208 Vincent N. Thijs and Gregory W. Albers

have not been successful.2426 Both the PROACT II tissue that is at risk of infarction but can still be sal-
trial, that compared intra-arterial pro-urokinase vaged by an eective therapy. Other goals are to
with placebo and a recent meta-analysis of all IV identify the etiology and location of the ischemic
thrombolytic stroke trials provided proof of the lesions and to provide an estimate of the prognosis
principle that stroke can be treated up to 6 hours of these lesions if left untreated. Ideally, imaging
after symptom onset.2729 Several explanations have may also be able to detect patients in whom treat-
been proposed to account for the failure of some of ment is harmful or unnecessary.
the individual IV thrombolytic trials beyond 3 The penumbra can be dened as hypoperfused
hours. ECASS I suered from protocol violations brain tissue that is not yet irreversibly damaged but
due to the erroneous inclusion of patients with at risk of infarction. The penumbra can be dened
large regions of early infarction evident on the in various ways depending on the modality used to
baseline CT scan, and the trial showed a positive identify it.3240
eect if these patients were excluded. If ECASS II Biochemically the penumbra can be dened as
had chosen the endpoint of independence (mod- the area in which protein synthesis is suppressed as
ied Rankin Score 02) instead of recovery (mod- a result of hypoperfusion, but ATP depletion, and
ied Rankin 01), possibly a more realistic target therefore energy failure, has not yet occurred.4146
given the longer time window for inclusion, the trial The penumbra can be identied as existing between
would have been positive.25,30 However, despite the two levels of metabolic dysfunction. A lower level
promising results of posthoc analyses of ECASS and exists beyond which ATP depletion and irreversible
ECASS II, ATLANTIS did not nd a benet for IV tPA damage occurs. The higher level distinguishes
thrombolysis, even when dierent endpoints were between hypoperfused tissue without a suppression
considered.26 in protein synthesis which is therefore not at risk of
There are additional theoretical reasons to infarction, from tissue that demonstrates suppres-
explain why these trials failed. The presumed sion of protein synthesis. Animal experiments indi-
target of acute stroke therapy is the ischemic cate that with persisting vascular occlusion brain
penumbra. Yet, none of the clinical trials of new regions beneath the threshold of ATP-depletion
therapeutic agents has attempted to use the exis- gradually expand and within hours equal the size of
tence or size of the penumbra as a selection criter- the regions at the threshold of reduced protein syn-
ion for entry. Only CT technology was used to thesis.34,42,47 This suggests that tissue salvage can
exclude patients with intracerebral hemorrhage or only occur early before this process is complete.
to exclude patients with early extensive cerebral Unfortunately, thresholds of ATP depletion and
edema. The inclusion of patients without substan- reduced protein synthesis can not be evaluated in
tial regions of reversible brain injury in trials of humans using currently available technology.
new therapeutic agents for stroke may have Estimations of these regions using PET, SPECT,
allowed entry of patients in whom treatment was Xenon CT and diusion and perfusion MRI, and
likely to be futile or even harmful. We believe that recently CT-perfusion have been proposed.37,4860
the ability of trials to distinguish eective from PET with 15O-labelled oxygen and water is the
ineective therapeutic agents in stroke can be only established method for identifying the
maximized by using techniques to image the ischemic penumbra in humans.6165 Using this
ischemic penumbra.31 approach the penumbra can be dened as tissue
with a critical ow decrease but preserved oxygen
consumption as reected in a high oxygen extrac-
Imaging the ischemic penumbra: the gold tion fraction or a preserved metabolic rate of
standard(s) oxygen.66 The studies of Baron and coworkers indi-
cate that tissue fullling these requirements is
A theoretical benet of imaging in acute stroke is to present up to 24 hours after occlusion in the border
non-invasively and rapidly identify the presence of areas of the ischemic region and that survival of
New MR techniques to select patients for thrombolysis in acute stroke 209

these areas can occur spontaneously or following show that diusion of water protons is restricted
administration of tPA.6770 These quantitative PET after ischemic stroke.7982 This reduction of water
measurements require arterial sampling and are movement is probably due to cytotoxic edema
cumbersome to obtain. associated with energy failure caused by reductions
Recently novel, semiquantitative PET measure- in cerebral blood ow. The apparent diusion
ments have been proposed to study the penumbra coecient (ADC) is a quantitative parameter that
in humans that partially overcome the complexity of reects the degree of mobility of water within the
quantitative PET measurements. brain parenchyma.
Fluoromisonidazole is a marker of hypoxic yet viable PWI is an evolving MR technology to study cere-
tissue. Read et al. found large areas of increased bral hemodynamics.82,83 Hemodynamic maps of
tracer on FMISO-PET studies up to 43 hours after cerebral blood ow (CBF), cerebral blood volume
symptom onset.71,72 This technique is not suitable as (CBV) and mean transit time (MTT) or time to
a clinical tool since it requires a delay between tracer bolus peak (TTP) can be created by mathematical
injection and imaging of at least 2 hours. No distinc- analysis of the evolution of the intensity of the T2*-
tion can be made between irreversible and reversible weighted gradient or spin echo images after a
injury. Flumazenil, a tracer that binds to GABA- Gadolinium bolus (Fig. 16.1). The advantages of
receptors in grey matter can be used to identify these perfusion-weighted imaging techniques
morphologically intact tissue.73,74 The penumbra can (PWI) are their high resolution and non-invasive
be dened with this technique as the area with a nature compared to positron emission tomography.
critical ow decrease but with preserved radioac- PWI can easily be combined with other MR tech-
tively labelled umazenil binding.75 This technique niques like magnetic resonance angiography (MRA)
permits imaging of the penumbra within a few hours to assess vessel patency and diusion weighted
after symptom onset and does not require arterial imaging (DWI) to assess injury to brain paren-
sampling which is contraindicated if thrombolysis is chyma.79,84,85 Current controversies regarding PWI
given. Heiss et al. estimated the amount of salvage- include the determination of the most accurate and
able tissue within the entire volume of hypoperfused cost-eective MR method to acquire perfusion
tissue using this technique. In patients studied images (gradient-echo or spin-echo), the most
within 12 hours after symptom onset about 50% of precise postprocessing method (single value
the initial ischemic lesion is composed of already decomposition, maximum likelihood approach,
irreversibly damaged tissue and therefore represents gamma-variate) to determine hemodynamics
the ischemic core. A substantial variability was quantitatively and the most predictive parameters
noted: in some patients the core regions measured to identify tissue at risk (MTT, CBF or CBV).8694
only 15.5% of the initial ischemic lesion while in It has been proposed that a combination of per-
others 94% of the ischemic lesion was beyond rever- fusion- and diusion-weighted MR might be able
sibility. Overall, the penumbra measured about 20% to identify the ischemic penumbra.95,96 In this initial
of the hypoperfused lesion volume, ranging from hypothesis the lesion identied on the diusion-
0.3% to 55.9%.76 The complex logistics involved in weighted imaging was thought to represent the core
maintaining a PET facility prohibit widespread use of the infarct and the component of lesion on the
of PET imaging in clinical practice. perfusion-weighted image that extends beyond the
DWI lesion represents the penumbra.97 This mis-
match hypothesis has been challenged by both
Diffusion- and perfusion-weighted imaging to animal and human experiments.
study the ischemic penumbra.
DWI and ADC lesions are reversible
Diusion-weighted imaging can assess the mobility
of water protons within the brain parenchyma.77,78 Animal ischemia experiments have demonstrated
Extensive animal and human experimental data that the lesion identied on diusion-weighted
210 Vincent N. Thijs and Gregory W. Albers

Fig. 16.1. DWI and PWI images from a patient imaged 6.8 hours after symptom onset with a baseline NIHSS of 13. The baseline
DWI lesion expands dramatically to almost equal the size of the baseline perfusion decit.

imaging is larger than the area where ATP depletion lesions is not associated with neurologic improve-
has occurred, indicating that only part of the DWI ment followed by worsening.104
lesion represents the ischemic core.98100 Studies Permanent reversal of early DWI lesions has also
using ADC mapping to provide a quantitative esti- been demonstrated in humans following early
mate of the degree of cytotoxic edema show that reperfusion conrming that early DWI lesions do
lesions with mild ADC usually normalize after not represent the ischemic core in stroke
reperfusion while lesions with more severe ADC patients.105109 Kidwell et al. showed the resolution
reduction go on to infarction.101 Interestingly, ADC of diusion lesions among patients with TIAs.110
lesions can disappear entirely after early successful The same group demonstrated that after recanal-
reperfusion but reappear later.102,103 This delayed ization with intra-arterial (i.a.) t-PA therapy or com-
re-emergence of ADC lesions together with T2 and bined i.v./i.a. t-PA treatment, DWI lesion volumes
DWI hyperintensities occurs following longer dura- regressed from baseline to follow-up.111 In 50% of
tions of hypoperfusion and is associated with the these patients, this reversal of DWI lesions was tem-
presence of tissue necrosis on pathological exam- porary with the development of recurrent lesions
ination. This resolution and re-emergence of DWI within one week. This pattern of disappearance and
New MR techniques to select patients for thrombolysis in acute stroke 211

re-emergence of diusion lesions appears very patients studied within 7 hours after symptom
similar to the animal observations discussed above. onset (Fig. 16.2, see colour plate section).118
Again, this pattern appears not to be associated Therefore, it now appears that the ischemic
with early clinical improvement and subsequent penumbra can be estimated by using a graded,
neurological deterioration.112,113 quantitative analysis of the perfusion lesion and the
Although these data suggest that early diusion ADC map. Lesions with early ADC values below a
lesions do not represent the ischemic core, accurate yet to be determined threshold represent the
determination of the core might be achieved by a ischemic core. Mild, non-critical hypoperfusion can
quantitative analysis of the ADC map together with probably be distinguished from tissue at risk of
parameters derived from PWI. Preliminary multi- infarction using a semiquantitative analysis of the
parametric analyses suggest that very low ADC hypoperfused tissue volume. Tissue with ADC
values represent irreversibly damaged tissue and values above the threshold for the ischemic core
are also at higher risk of hemorrhagic transforma- and PWI evidence of critical hypoperfusion will
tion.113115 correspond to the penumbra (Fig. 16.3).

Visual analysis of the PWI lesion does not Factors influencing DWI and PWI lesion
accurately identify tissue at risk of infarction evolution

The goal of hemodynamic imaging is to rapidly and In untreated stroke patients, DWI lesions typically
accurately identify the area of hypoperfusion. enlarge during the rst week after symptom
Within this lesion, tissue at risk of infarction should onset.95,96,119122 In a randomized placebo-
be distinguished from tissue that is mildly hypoper- controlled trial of the putative neuroprotective
fused, but not at risk of infarction. This is especially agent citicholine, an MR substudy showed that in
important in the design of clinical trials of new untreated patients studied within 24 hours after
thrombolytic agents as the inclusion of patients symptom onset the follow-up lesion volume was
who have large areas of hypoperfused tissue that is 180% the size of the baseline lesion volume,
not at risk of infarction could lead to bias if these although there was substantial variability.123 Factors
patients are overrepresented in either the active that were associated with more impressive lesion
treatment or the control group. growth were the presence of a large hypoperfusion
Recent evidence indicates that visual analysis of lesion on PWI, evidence of occlusion on MRA and
the perfusion lesion volume may substantially over- larger size of the baseline diusion lesion. Patients
estimate the nal infarction volume and critically studied within 12 hours also had more substantial
hypoperfused tissue can be identied by grading lesion growth. These ndings conrm previous
the degree of hypoperfusion. Neumann-Haefelin et observations using DWI and PWI in acute stroke
al. reported that hypoperfused lesions with a time patients.92,95,120
to peak delay (a measure of the degree of hypoper- Darby et al. studied the presence of a PWI/DWI
fusion) of more than 6 seconds were at risk of sub- mismatch over time and found that a PWI/DWI mis-
sequent lesion enlargement in stroke patients match was present in 75% of stroke patients
studied within 24 hours.116 Schlaug et al. studied 15 scanned within 6 hours of symptom onset, 70% of
patients within 6 hours after symptom onset and patients studied at 12 hours and 44% of the patients
found that tissue that was at risk of infarction had scanned at 18 hours after symptom onset.84 The
rCBF values < than 37% of the contralateral side same group studied the evolution of patients who
and mean transit time (MTT) increases of more had isolated hypoperfusion lesions (normal DWI) at
than 165% of the contralateral side.117 We recently baseline. Subsequent infarction occurred in 6/11
reported that mean MTT increases of >4 or patients within the PI abnormality. The remaining
>6 seconds identied tissue at risk of infarction in ve showed no later infarction. The volume of the
212 Vincent N. Thijs and Gregory W. Albers

Ischemic core

Potentially reversible DWI lesion +

Hypoperfused, at risk of infarction

= Penumbra

Non-critical hypoperfusion

Fig. 16.3. Current concept of the ischemic penumbra as imaged using DWI and PWI. The baseline DWI lesion consists of tissue
that is salvageable and tissue that is irreversibly damaged. These compartments can probably be distinguished by quantifying the
ADC. Regions with ADC values below 300350 106 mm2/s have a low likelihood of recovery. The hypoperfusion lesion contains
both tissue at risk of infarction and mild, non-critical hypoperfusion. These compartments can probably be distinguished by
quantifying the PWI lesion. For instance, regions with MTT increase of less than 4 seconds appear to be at minimal risk of infarc-
tion. The ischemic penumbra consists of both the potentially reversible area of the early diusion lesion and tissue with critical
levels of hypoperfusion identied within the entire baseline hypoperfusion lesion (pale area).

PWI lesion did not predict recovery. The degree of patients with a PWI/DWI mismatch might be
the hypoperfusion decit at baseline was not related to the insensitivity of MRA to identify corti-
studied. It is unclear if these patients had mild, non- cal branch vessel occlusion or to the existence of a
critical hypoperfusion or if spontaneous early reper- no reow phenomenon following recanalization or
fusion prevented the development of infarction.124 to the latency for PWI lesions to disappear follow-
Karonen performed serial studies in 49 stroke ing spontaneous or treatment induced reperfu-
patients. Imaging was performed during the rst sion.127 Diusion lesions have been shown to
day after symptom onset, at day 2 and at 1 week. exhibit substantially more growth in patients with a
Eighty per cent of the patients had a PWI/DWI mis- major vessel occlusion compared to patients
match at baseline. The mismatch volumes corre- without a vessel occlusion. These observations
lated signicantly with the presence of lesion support the hypothesis that patients with occluded
growth between day 1 and day 2 and day 1 and vessels still have more tissue at risk of infarction
1 week on diusion-weighted imaging.122 compared to patients who have likely experienced
The presence of a PWI/DWI mismatch is asso- spontaneous reperfusion.125,126
ciated with the presence of a major vessel occlu-
sion, although exceptions occur. Rordorf et al.
studied 17 patients within 12 hours after symptom Monitoring thrombolysis using DWI and PWI
onset and found a PWI/DWI mismatch in 9/10
patients with M1-occlusion. Six out of seven The rationale of thrombolysis in acute stroke is to
patients without M1-occlusion did not have a open occluded vessels and therefore provide bene-
PWI/DWI mismatch.125 Barber et al. found MRA- cial reperfusion. Both the NINDS trial and the
identied vessel occlusion in 9 of 14 patients PROACT trials suggest that reperfusion is the most
studied within 24 hours after symptom onset with a promising strategy for acute stroke therapy to date,
PWI/DWI mismatch. None of the patients without a as none of the putative neuroprotective agents have
mismatch had major vessel occlusion.126 The yet proven benecial in acute stroke patients.128130
absence of an MRA identied vessel occlusion in The frequency of nding occluded vessels within
New MR techniques to select patients for thrombolysis in acute stroke 213

6 hours of stroke onset is high. When examined by match. Early resolution of PWI abnormalities
angiography within 6 hours of acute onset, approxi- occurred frequently in patients treated with IV t-PA
mately 1020% of patients do not show a major and may be a predictor of a favourable clinical
vessel occlusion.5,27,131133 Most patients with mod- response and a reduction in the expansion of the
erate to severe stroke who are enrolled in acute early DWI lesion. The initial MRI scan (typically
intervention trials have either an M1 occlusion performed about 2 hours after the initiation of t-PA
(30% of patients), or an M2 occlusion (25% of therapy) revealed a PWI lesion that was smaller
patients). The remaining patients have either a than the initial DWI lesion in 5/6 t-PA-treated
carotid T occlusion (10%), a proximal occlusion or patients. In contrast, this pattern was only seen in
severe stenosis of the internal carotid artery or a one of the six non-t-PA-treated patients. This sug-
vertebrobasilar occlusion (510%).7,132 gests that most of the t-PA-treated patients had
Only a few serial DWI and PWI studies have com- already begun to experience reperfusion before the
pared the lesion evolution of patients who had evi- rst MRI scan was obtained. Furthermore, 83% of
dence of early reperfusion (either spontaneous or the t-PA-treated patients showed a substantial
with thrombolytic therapy) with patients without decrease in PWI volume between baseline and the
reperfusion. Compared to patients with other rst repeat MR scan whereas none of the non-t-PA
PWI/DWI proles, patients with a PWI/DWI mis- patients showed this degree of improve-
match are thought to be the ideal candidates for ment.120,135,136
thrombolysis as they are the most likely to have sal- Schellinger and Jansen studied 24 patients
vageable brain tissue. treated with t-PA within 6 hours after onset. Most
Barber et al. studied 28 untreated patients with (66%) were scanned prior to t-PA treatment and
MCA territory stroke within 12 hours of stroke others were scanned during or immediately after
onset. Major spontaneous reperfusion dened as a tPA infusion. Vessel occlusion was present in 20/24
>90% reduction in PWI lesion by day 3 occurred in patients and in 11/20 recanalization occurred.
45% of patients and was associated with better clin- Nineteen of 20 patients with vessel occlusions had a
ical outcome and smaller nal infarct size. In PWI/DWI mismatch. In patients with early recanal-
patients with major reperfusion the expansion of ization, functional and radiological outcomes were
the baseline DWI lesion was reduced by 67%.134 signicantly better than in patients who did not
We recently studied 21 patients within 7 hours have recanalization. Patients with a PWI/DWI mis-
after symptom onset with serial diusion and per- match who had early reperfusion had signicantly
fusion-weighted MR. All 21 patients had a measur- greater improvements in NIH Stroke Scale scores
able DWI lesion on the initial scan compared to and smaller nal lesion volumes than patients who
only 10% who had a visible lesion on the initial had a PWI/DWI mismatch but in whom no reperfu-
FLAIR image. Abnormal regions of marked delay in sion occurred.137,138
contrast arrival times (PWI lesions) were seen on Parsons et al. recently described 14 patients in
time-to-peak maps in 16/18 patients. DWI lesion whom DWI and PWI were performed within
volumes grew substantially during the rst week, 3 hours after symptom onset. Eighty per cent of
reaching a mean of 320% at 57 days. Lesion patients with a PWI/DWI mismatch who were
volumes typically decreased between the 57-day treated with t-PA had no signicant expansion of
scan and the 30-day scan, however, the nal infarct the early DWI lesion, whereas all untreated patients
size was nearly twice as large as the early DWI with a mismatch had nal infarcts that were con-
lesion. Patients with DWI/PWI mismatch experi- siderably larger than the early DWI lesions.139
enced larger increases in early DWI lesion growth These small, non-randomized studies suggest
than patients without this pattern. In patients that DWI and PWI could be used to monitor throm-
without a mismatch, the average lesion growth bolytic therapy and indicate that specic MR pat-
between baseline and 30-day follow-up was 120%, terns can be identied among patients who are
compared to 250% in patients with a PWI/DWI mis- likely to respond favorably to thrombolytic therapy.
214 Vincent N. Thijs and Gregory W. Albers

Use of DWI and PWI as surrogate markers in lesion volumes early after stroke onset is not
acute stroke trials aected by factors that can inuence functional
outcome, such as social circumstances or the
In animal stroke models, primary evidence of the quality of rehabilitative treatment. These factors are
ecacy of therapeutic intervention is established very dicult to control in small samples and can
by a reduction in ischemic lesion volume compared bias the results of a small trial. An objective meas-
to non-treated animals. A surrogate end point in a urement, such as a reduction in the early growth of
clinical trial is a laboratory measurement that is the ischemic lesion size, could demonstrate a proof
used as a substitute for a clinically meaningful end- of the principle on which the experimental treat-
point.140 Changes induced by a therapy on a surro- ment was based.128,130,148
gate endpoint are expected to reect changes in the
clinically meaningful endpoint. To be used as an
endpoint, the surrogate marker should be tightly Future prospects
linked to the outcome characteristic. Several
studies using DWI and PWI in humans have We have designed a clinical trial that will test the
reported moderate correlations between the hypothesis that stroke patients with a PWI/DWI
volume of the early DWI and PWI lesion and func- mismatch and early successful recanalization have
tional outcome scales like the Barthel a more favourable clinical and radiological
index.120,141145 We reported that, along with the response to tPA than patients without these fea-
NIHSS and patient age, DWI lesion volume was an tures in patients treated with i.v. t-PA within 3 to 6
independent prognostic marker.146 The modest cor- hours after symptom onset. This non-randomized,
relations between lesion volume on DWI and func- prospective multicentre study of 80 patients will
tional outcome scales can probably be attributed to provide the necessary pilot data to plan a placebo
both the numerous additional factors that inu- controlled randomized trial using MRI selection
ence functional outcome and the relative inade- criteria to include patients most likely to respond
quacy of stroke outcome scales. In addition, lesions favourably to tPA.
of similar volume in dierent brain regions are If the primary hypothesis of this study is correct,
likely to have variable inuences on functional a future placebo-controlled ecacy trial will be
outcome. Most of the available data support the designed to include only patients who have evi-
idea that DWI lesion volumes could be used as a dence of an ischemic penumbra of signicant size
surrogate marker for functional outcome. based on PWI/DWI parameters. The pilot trial will
Although we do not advocate to rely solely on provide accurate sample size calculations and
measurements of lesion volumes in phase III trials, address safety issues (e.g. exclusion of patients with
we believe there is a role for DWI and PWI in the DWI lesions above a certain size to exclude hemor-
selection and development of new therapeutic rhage or herniation). It is possible that MRI proles,
agents in phase II trials. This technology can dra- other than a PWI/DWI mismatch, will predict a
matically reduce the sample size necessary to show favourable clinical response to t-PA. In addition,
meaningful eects by using the patient as his own certain MRI or MRA proles (perhaps an extremely
control. We calculated that to demonstrate the large baseline PWI lesion or proximal or distal
eect of a drug that is able to reduce lesion expan- carotid occlusions) might be associated with a very
sion by 25% between day 1 and day 3, a sample size low rate of early reperfusion and an unfavourable
of only 80 patients per treatment group would be clinical response