Int J CARS

DOI 10.1007/s11548-014-0996-x

ORIGINAL ARTICLE

AneuSearch: a software prototype for intracranial aneurysm

searching and clinical decision support
Jian Wu Harvey Ho Peter Hunter Ping Liu

Received: 9 January 2014 / Accepted: 16 March 2014

CARS 2014

Abstract
Purpose Clinical decisions for treating intracranial aneurysms
(IA) require integrating information in various forms and
from multiple sources. We aim to establish a framework
namely AneuSearch to integrate relevant information in IA
management and also allow for efficient IA searching based
on carefully designed criteria.
Methods The backbone of AneuSearch is an open-source
three-tier DICOM image management system called DCM4Chee, which is a Java implementation for PACS. A supplementary database (AneuSearchDB) was developed to contain
morphological features, hemodynamic and histological data.
The relational tables in AneuSearchDB correspond to the
most fundamental questions raised by neurosurgeons during
IA treatment. The system was developed through collaborations between bioengineers and neurosurgeons.
Results The prototype software has been deployed to computers in a Mianyang Central Hospital in China. Currently, the
system contains the data of 105 IA patients, seven hemodynamic simulation results and nine histological section
images. This system was queried as per given criteria and
can also provide blood flow data after running an external
computational fluid dynamics software.
Conclusions The prototype software provides a novel tool
to IA management. Future works include incorporating IA
treatment criteria in IA rupture risk assessment.
Jian Wu and Harvey Ho have contributed equally to this paper.
J. Wu P. Liu
Department of Neurosurgery, Mianyang Central Hospital,
Mianyang City, Sichuan Province, China
H. Ho (B) P. Hunter P. Liu
Bioengineering Institute, University of Auckland,
Auckland, New Zealand
e-mail: harvey.ho@auckland.ac.nz

Keywords Database Intracranial aneurysm Information

system Modeling

Introduction
Intracranial aneurysms (IAs) are vascular disorders that are
prevalent in 15 % of the population [1]. Most IAs remain
asymptomatic and cause no problem in ones lifetime as
the chance of rupture is low (about 1 in 10,000) [1]. However, IA rupture results in subarachnoid hemorrhage (SAH),
which is associated with a 50 % mortality rate. Nowadays, more unruptured IAs are detected incidentally due to
advanced medical scanning. Decisions must be made concerning whether clinical interventions (e.g., surgical clipping, or endovascular coiling) are required, or a conservative
treatment be taken. Up to this day, clinicians rely on their
experience with the size, location and shape of the aneurysm
to determine the aneurysm rupture risk. However, it has been
suggested that multiple factors such as blood flow stresses,
genetic factors, etc, could also play a role [2,3]. Additional
information are often stored in various documents, or remain
unknown (e.g., blood flow stress) unless processed by biomedical scientists. Hence, an information system that is capable of presenting all relevant data to clinicians for his/her
optimal interventional decisions is desirable.
Previous IA database studies range from single institutional study [4] to nationwide survey [5] and to multinational review of medical records [6]. These databases, however, did not store medical scanning images of patients but
were mainly used to perform metadata analysis. A 4-year
European e16 million project, @neurIST, has developed a
cluster of useful tools for the understanding and modeling
of various aspects of IA [7,8]. In particular, @neuFuse pro-

123

Int J CARS

vided a computational pipeline from aneurysm segmentation

to 3D flow analysis [8]. Similarly, another project namely
AneuRisk provided semiautomatic tools for IA geometry
reconstruction, which can then be used for hemodynamics
simulation [9]. The integrated data (morphology + hemodynamics) were used to correlate predictive landmarks for
rupture [9]. A recent project AngioLab further added in a virtual stenting module to mimic the endovascular intervention
procedures [10]. These tools have greatly advanced the biomedical research for IA and associated translational studies.
However, a few specific clinical requirements remain unaddressed. For example, these systems usually do not contain
histological analysis data in the database and/or may have not
implemented IA feature-based searching/indexing against a
picture archiving and communication system (PACS) [7,8].
The primary goal of this work is to develop a prototype
namely AneuSearch for such an information system. Two
basic yet essential tasks need to be tackled. Firstly, it should
allow an authorized user to search an IA database as per clinically relevant criteria. Secondly, the system should integrate
various available data (image, clinical records, etc), which
exist in various clinical reports, into a single system. Moreover, the system should be capable of utilizing blood flow
analysis carried out by external software. The long-term aim
is to use the software to aid clinical decisions in treating IAs.

Methods
Clinical environment and available data
The software prototype was custom developed by Auckland
Bioengineering Institute (ABI) for the Neurosurgery Department (referred as the Department in the text hereafter) of
Mianyang Central Hospital (MCH), the major health care
provider in a medium sized Chinese city (the Mianyang City)
with a population of 5.3 million. The department manages
1,620 patients each year. About 110 or 5 % of them are
IA patients, approximately, 90 % are admitted due to SAH.
Upon admission into the Department, the patients usually
receive a CT scan (GE Medical Systems/LightSpeed 16). The
scanning images are stored in a PACS system of the Radiology Department and can be accessed from workstations in the
neurosurgery department. Depending on the patients condition, digital subtraction angiography (DSA) may be arranged
to identity the locations of IAs. DSA images are also stored
in that PACS system. Radiology diagnosis reports containing
volume rendering images are also accessible and are used to
identify the size and location of the aneurysm.
Besides the CT images, MRI scans are occasionally performed in the hospital. Furthermore, histology analyses of
dissected IA tissue are carried out in the Pathology Department for selected IA patients. The results provide images of

123

Fig. 1 The infrastructure of AneuSearch: the framework is built upon

DCM4Chee, with a supplementary database (AneuSearchDB) implemented to perform IA feature-based search. The connection between the
image database (PACDB) and AneuSearchDB occurs via the patient ID

arterial tissue, showing endothelial cells and smooth muscle

cells (using H&E staining). The reports are saved in Word
document.
Thus, the clinical data available in the Department for
perioperational IA analysis are in the form of numbers, free
text, pixel images (in DICOM or JPEG) and other documents
(in Word or PDF format).

Infrastructure of AneuSearch
The software infrastructure is based on DCM4Chee, an opensource DICOM image management system [9]. In brief, the
Dcm4Chee is a Java wrapper for the DICOM standard that
can be used for Java-based application development [9]. It
is a three-tier structure based on Java 2 Enterprise Edition,
deployed with the JBoss Application server and the MySQL
database server (Fig. 1).
DCM4Chee has been used in image indexing/searching
in large medical image databases such as Dicoogle [10], the
Cardiac Atlas Project (CAP) [11]. The project CAP contains over 1 million images of 5,000 patients. The number of
patients in AneuSearch (1,000) is at the same scale as that
in CAP.
Note that the PACS system in DCM4Chee is not the same
PACS used in the Radiology Department. Rather, the studies
are exported from the latter into a local disk and then imported
into DCM4Chee.

Int J CARS
Table 1 Classification criteria for IAsthe location
Criteria

Data type

Remarks

Location

Text

Size

Real number

Categories: ACA, MCA, PCA,

ACoA, PCoA, VA, BA
Unit: millimeter

Aspect ratio

Real number

aneurysm size;
aspect ratio;
patient conditions (hypertension, vasospasm, etc);
Treatment (to provide information concerning the treatment outcome)

The ratio of IA height

versus neck width

Incorporating clinical queries into AneuSearchDB

In the Department, the searching criteria of the existing PACS
system simply consist of the patient ID, name and the scan
time. It cannot perform feature-based IA search. AneuSearchDB is designed to address this problem by answering the
most common queries raised by a neurosurgeon to aid his/her
decision-making. these questions commonly include:
(1) Where the aneurysm is located and its size?
(2) The shape of this aneurysm and whether it is a saccular
or a fusiform aneurysm. If it is a saccular aneurysm, then
is it wide or narrow necked?
(3) Whether the aneurysm is ruptured or not. If it is unruptured is an intervention required, and if ruptured, are there
any complications?
(4) Finally, the general condition of the patient must be taken
into account. Thus, one must consider does he/she have
other medical conditions that could restrict clinical interventional decisions?
Questions (1) and (2) seek anatomical and morphological
features of an IA, which in current clinical settings are easily
answered by visual inspection of scan images. Questions (3)
and (4) require more in-depth analysis of the patients condition. The queries may be translated into separate fields in
a database table as follows:
aneurysm location;

The first three morphological parameters are the most basic,

and their specific definitions are given in Table 1 below.
The locations of IAs are classified into seven categories: anterior cerebral artery (ACA), middle cerebral artery
(MCA), posterior cerebral artery (PCA), anterior communicating artery (ACoA), posterior communicating artery
(PCoA), vertebral artery (VA) and basilar artery (BA). In
case, an IA arises from a junctional structure, all arterial categories involved are included in the location definition.
The primary key of the ClinicalData table is the patient
ID imported from the DCM4Chee image database PACSDB
(Fig. 1). However, IA searching is based on the lightweight
AneuSearchDB rather than the much larger PACSDB database.

Relational tables in AneuSearchDB

Currently, three relational tables have been implemented in
AneuSearchDB for the current version. The tables contain
data of morphological features, hemodynamics and histology
analysis results. A schematic of the database is shown in Fig.
2.
Note that every patient should have morphological features of IA, but may not have corresponding hemodynamic
and/or histological data. Histological analysis reports are
only available in a small percentage of patients when such an
analysis is deemed necessary by the neurosurgeon in charge.
Likewise, hemodynamic data are only available when such
an analysis is requested and must be carried out by a bioengineer.

Fig. 2 Schematic of the relational tables in AneuSearchDB

123

Int J CARS

Patient data and security

All IA cases treated in MCH from March 2011 to March
2012, i.e., in a one-year-period were identified by neurosurgeons and exported from the PACS system into a hard disk.
The images were then imported into AneuSearch. The security of the system was maintained by the internal security
policy of DCM4Chee. Since the system was designed to be
deployed in the intranet eventually, no internet access was
permitted.
The project was approved by the hospital authority. Since
the study is a retrospective study of existing IA cases, no
ethical approval was required for the current project.
Hemodynamics analysis
An important factor in the genesis, growth and rupture of an
aneurysm is the hemodynamic forces acting on it, and the
resultant constituent changes on its wall [3]. Biomechanic
analysis yields quantitative data of blood flow forces and
wall deformation to assist the IA rupture risk assessment.
This has been added as an advanced module in conjunction
with a computational pipeline developed for blood flow simulations in large arteries [2,12]. Note that the actual hemodynamics analysis is not carried out within the framework, but
completed by using a third party computational fluid dynamics (CFD) software ANSYS CFX (ANSYS Inc, PA, USA)
which is commercially available.
While we refer the interested reader to the literature for
more detailed description of the modeling pipeline, it should
be stressed that the simulated flow data and/or resulted index
(e.g., WSSwall shear stress, OSIOscillation index) are
largely affected by boundary conditions defined in the flow
simulation. The boundary conditions, such as the inflow flow
rates provided by surgeons, are different at various locations
of an arterial tree. Thus, we consider it more appropriate not
to add WSS or OSI fields into the Hemodynamics table
at this stage, but only provided a WSS distribution map (as an
image) and the link to simulation result file. The clinicians are
then able to check the flow results from a postprocessing tool
(e.g., CFX-post) by himself/herself. This design is deemed as
suitable in a clinical environment because a surgeon usually
does not have the time and/or the expertise to run a CFD
software from scratch, but he/she can observe and analyze
the 3D simulation results worked out by a bioengineer.

tested with different versions of Microsoft Windows (Windows XP, 7, 8), which are the operating systems currently
used in the hospital. Currently, the system is loaded with a
dataset of one hundred and five (105) IA patients containing
26,642 DICOM images of total 12.12 GB disk space (515 K
per file). The system also contains six hemodynamic result
files and histology analysis reports for nine patients. It was
deployed in both MCH and ABI to facilitate the communication between neurosurgeons and computer scientist / bioengineers.
Figure 3 shows the snapshots of the graphic user interface
(GUI) of the system. In Fig. 3a, a basic interface was designed
so that a user can query the IA database based on a single or
multiple criteria, and the results are shown in a table (Fig. 3b).
Another web-based interface (Fig. 3c) was implemented to
allow the administrator of AneuSearchDB to edit/add patient
data in the table.
Exemplary IA search

Results

The system can perform IA searches based on the criteria

defined in AneuSearchDB. For example, a basic search based
on the Location criterion yielded that 32 % (33/105) of
patients had PCoA aneurysms and 24 % (25/105) of them
suffered from ACoA aneurysms. The percentage of anterior, middle and posterior cerebral artery (ACA, MCA, PCA,
respectively) aneurysms were 5.4, 13.5 and 2.7 %, respectively. Aneurysms arising in the VA and BA which supply
the posterior cerebral circulation accounted for 13.5 % of all
IAs, and 9.5 % of the patients have multiple aneurysms.
The data acquired indicate that IAs are more likely to
develop at the ACoA and PCoA of the cerebral vasculature,
where the blood flow patterns are very complex due to multiple bifurcations and also because ACoA and PCoA function
as bridges between (unbalanced) collateral cerebral flows
[12]. This is consistent with that reported in other studies
(e.g., that summarized in [13])
Figure 4 illustrates an example when the search criterion
of aneurysm size (15 mm) was set, the system returned a
list of aneurysms larger than 15 mm. Among these largest
aneurysms, a giant aneurysm (34.2 mm 27.5 mm) was
found in a patient suffering from severe headache and vision
loss due to mass effects from the giant aneurysm. The computational results (WSS, OSI, flow pattern, etc) of the simulation were accessible by the clinical team to aid pathophysiological/surgical understanding of giant aneurysms. A
detailed description of the simulation conducted for this giant
aneurysm was reported recently by the authors in [14].

Software implementation and deployment

Integration of data from multiple sources

The concepts outlined in the Methods section have been

implemented into the prototype AneuSearch. It has been

Four cases have been identified from the 105 patients

whereby multiscale data from the aneurysmal scale to tis-

123

Int J CARS

Fig. 3 Interface design of AneuSearch. a Search criteria panel and button; b search results populated in a table; c web-based interface to edit/add
data in the table

sue/cellular scales were available (Fig. 5). At this stage, no

further quantitative analysis has been made to link these data.
However, the bioengineer co-authors (HH and PH) have suggested a multiscale-framework for aneurysmal growth and
remodeling [2]. The information system presented in this
paper will greatly benefit the framework described in [2].
Evaluation

Q1: Have you previously used an IA specialized software

other than AneuSearch?
Q2: Do you think an IA-focused software is useful in the
process of IA management?
Q3: Do you think the current version AneuSearch is useful
in clinical decision-making?
Q4: What are the features you like in the system?
Q5: Which features need to be further developed?

An internal meeting was arranged to evaluate the software

and to collect feedback from the clinicians in the Department. The attendants included two neurosurgeons, one neuroradiologist, one neurosurgeon resident and two nurses. Five
questions were provided:

For Q1, all clinicians could identify at least one 3D construction software but they agreed that these software are focused
on general use, not specially designed for IAs; for Q2, all
clinicians agreed an IA-focused software could be useful for
daily IA management; for Q3, it was suggested that more

123

Int J CARS

Fig. 4 Example of feature-based IA search and the link to hemodynamic analysis: 1 Search for aneurysms with size larger than 15 mm; 2
A giant aneurysm was found, its DICOM images were retrieved from
PACSDB for postprocessing; 3 3D flow analysis was carried out for the

giant aneurysm; 4 Simulated blood flow data were saved in AneuSearchDB and can be accessed to aid pathophysiological/surgical research
of giant aneurysms

modules needed to be added as its function was still limited in its current prototype version. For Q4, the clinicians
were positive about the concept of simulating blood flow in
aneurysms which could shed lights on rupture risk assessment. For Q5, it was suggested that more clinically related
fields such as the SAH grades, complication and symptom
descriptions could be added to the table.
While the above evaluations were from users perspective, feedbacks were also obtained from the chief engineer of
CAP, which also uses the DICOM4chee architecture. Searching of relevant IAs from the 105 patients was close to real
time performance (0.2 s); therefore, the operating speed was
not an issue. However, comments were made concerning
image data de-identification and web security if the software would be put into multiple medical center trials. On
the other hand, since the prototype software was designed
for the internal use of a single center at this stage, these concerns are not relevant for this prototype version. Indeed, the
neurosurgeons prefer having patient data as important fields
in the database to aid them recall the actual treatment scenarios.

Discussion

123

In this paper, we introduced a prototype software (AneuSearch) which has been installed in the neurosurgery department of MCH. This system supplements the current PACS
system in that it enables feature-based aneurysm search
depending on the most clinically related concerns (e.g.,
aneurysm size, location, aspect ratio, etc) and other relevant
data such as hemodynamics and histological data. Since more
feedbacks are being collected concerning the GUI of the software and its database design, its exact impacts on patient care
and future clinical procedures are yet to be investigated. For
instance, the database searching at the pre-surgery stage is
different from that at the diagnosis stage, thus a different
GUI needs to be presented. After streamlining these software details, the system may be recommended to neurosurgeon peers in different hospitals for a multiple-center clinical
trial.
Extension of the current system will occur in two directions, at the patient and the population levels. At the patient
level, the system will be extended to incorporate more clin-

Int J CARS

Fig. 5 Information from multiple sources were brought together to aid clinical decisions. These data are also important for medical research, e.g.,
for the study of arterial wall remodeling under hemodynamics forces

ical related field, as briefed in the Evaluation section. At

the population level, we hope to gain insights into the
IA occurrence patterns in populations as per their gender,
age and other characteristics (e.g., hypertension) as has
been shown in the Results section. The strategy is to collect a large amount of data (biomechanical, histological
analysis, cellular/molecular measurements, etc) for a few
selected patients while including a small amount of routine measurements (e.g., blood pressure) on a large number
of people.
We emphasize that although the current prototype, according to its current design, is used in relation to just one disease
(cerebral aneurysms), the infrastructures core technologies
could potentially be transferable to serve the needs of many
other areas of (brain) diseases. It is hoped that by augmenting the AneuSearch database (AneuSearchDB), the system
can be customized for neuropathology research. For instance,
for the research on cerebral vasospasm (CV), one of the most
common complications of SAH after IA rupture (represented
as severe constrictions of cerebral arteries) [15], a Vasospasm
table can be implemented with such fields as the constriction
percentage of CV, the medicine administered, the treatment

outcomes, etc, and thus providing a general picture of CV

development and management. The concept is straightforward, but there are still no relevant bioinformatics systems
for CV, to our knowledge.
Another obvious choice would be for brain tumor patients
treated in the same department, including 150 meningioma cases/year and 100 glioma cases/year. This is
another substantial area that needs to be explored from a
bioinformatics perspective. The experiences gained from
the AneuSearch project will certainly contribute to this
vision.

Conclusion
In this paper we introduced the prototype of an information
system AneuSearch to integrate clinical data of IA patients
to support clinical decisions. The core of the system is based
on an open-source three-tier DICOM image management
system (DCM4Chee/JBoss/MySQL), with a supplementary
AneuSearch database developed to incorporate important IA
features and diagnosis data. We showed that the system could

123

Int J CARS

perform meaningful aneurysm search and facilitate further

hemodynamics analysis.
Acknowledgments The project was partially support by an International Mobility Fund of the Royal Society of New Zealand.

References
1. Brisman JL, Song JK, Newell DW (2006) Cerebral aneurysms. N
Engl J Med 355(9):928939
2. Ho H, Suresh V, Kang W, Cooling MT, Watton PN, Hunter PJ
(2011) Multiscale modeling of intracranial aneurysms: cell signaling, hemodynamics, and remodeling. IEEE Trans Biomed Eng
58(10):29742977
3. Watton PN, Raberger NB, Holzapfel GA, Ventikos Y (2009) Coupling the hemodynamic environment to the evolution of cerebral
aneurysms: computational framework and numerical examples. J
Biomech Eng 131(10):101003101014
4. Huang J, McGirt MJ, Gailloud P, Tamargo RJ (2005) Intracranial
aneurysms in the pediatric population: case series and literature
review. Surg Neurol 63(5):424432
5. Andaluz Norberto, Zuccarello Mario (2008) Recent trends in the
treatment of cerebral aneurysms: analysis of a nationwide inpatient
database. J Neurosurg 108:11631169
6. Molyneux AJ, Kerr RS, Birks J, Ramzi N, Yarnold J, Sneade M,
Rischmiller J (2009) Risk of recurrent subarachnoid haemorrhage,
death, or dependence and standardised mortality ratios after clipping or coiling of an intracranial aneurysm in the International Subarachnoid Aneurysm Trial (ISAT): long-term follow-up. Lancet
Neurol 8(5):427433
7. Benkner S, Arbona A, Berti G, Chiarini A, Dunlop R, Engelbrecht G, Frangi AF, Friedrich CM, Hanser S, Hasselmeyer P, Hose
RD, Iavindrasana J, Khler M, Iacono LL, Lonsdale G, Meyer
R, Moore B, Rajasekaran H, Summers PE, Whrer A, Wood S
(2010) @neurIST: Infrastructure for advanced disease management through integration of heterogeneous data, computing, and
complex processing services. IEEE Trans Inf Technol Biomed
14(6):13651377

123

8. Villa-Uriol MC, Berti G, Hose DR, Marzo A, Chiarini A, Penrose

J, Pozo J, Schmidt JG, Singh P, Lycett R, Larrabide I, Frangi AF
(2011) @neurIST complex information processing toolchain for
the integrated management of cerebral aneurysms. Interface Focus
1(3):308319
9. Warnock MJ, Toland C, Evans D, Wallace B, Nagy P (2007) Benefits of using the DCM4CHE DICOM archive. J Digit Imaging
20(S1):125129
10. Costa C, Ferreira C, Bastio L, Ribeiro L, Silva A, Oliveira JL
(2010) Dicoogle-an open source peer-to-peer PACS. J Digit Imaging 24(5):848856
11. Fonseca CG, Backhaus M, Bluemke DA, Britten RD, Chung JD,
Cowan BR, Dinov ID, Finn JP, Hunter PJ, Kadish AH, Lee DC,
Lima JAC, Medrano-Gracia P, Shivkumar K, Suinesiaputra A, Tao
W, Young AA (2011) The Cardiac Atlas Projectan imaging database for computational modeling and statistical atlases of the heart.
Bioinformatics 27(16):22882295
12. Ho H, Sands G, Schmid H, Mithraratne K, Mallinson G, Hunter
P (2009) A hybrid 1D and 3D approach to hemodynamics modelling for a patient-specific cerebral vasculature and aneurysm. In
MICCAI: Medical image computing and computer-assisted intervention, pp 323330
13. Brisman JL, Song JK, Newell DW (2006) Cerebral aneurysms. N
Engl J Med 355(9):928939
14. Ho H, Wu J, Hunter P (2011) Blood flow simulation in a giant
intracranial aneurysm and its validation by digital subtraction
angiography. In Wittek A, Nielsen PMF, Miller K (eds) Computational biomechanics for medicine, Springer, New York, pp 1526
15. Pluta RM (2005) Delayed cerebral vasospasm and nitric oxide:
review, new hypothesis, and proposed treatment. Pharmacol Ther
105(1):2356