THE ANATOMY OF THE CAUDA EQUINA ON

CT SCANS AND MRI

MARK S. COHEN, ERIC J. WALL, CHARLES W. KERBER,

JEAN-JACQUES ABITBOL, STEVEN R. GARFIN

From the University ofCaljfornia, San Diego

The nerve roots of the cauda equina may be visualised by contrast-enhanced CT scans and by surface-
coil MRI. We have identified the pattern of anatomy from L2-L3 to L5-Sl in 10 human cadaver specimens
and correlated this with anatomical dissections. Individual roots are slightly more distinct on contrast-
enhanced CT than on surface-coil MRI. There is a crescentic oblique pattern of nerve roots at the lower
lumbar levels which is still apparent in the more crowded proximal sections. In all cases, the axial images
correlated precisely with the dissections. Current imaging modalities can help the clinical understanding and
management of abnormalities in this region of the spine.

The nerve roots within the thecal sac of the cauda equina injection of 2.5% glutaraldehyde in phosphate buffer into
can be demonstrated on contrast-enhanced CT scans and the thecal sac through a small laminotomy site ; 3 ml of
by surface-coil MRI. The roots have been described as lohexolcontrast agent(Winthrop, New York) was added.
forming a crescentic pattern (Naidich et al 1980), which The entire cauda equina was then exposed by removal of
spreads diffusely and occupies the dorsal aspect of the the posterior spinal elements and embedded in situ with
dural sac at lower lumbar levels (Resj#{246}et al 1979; liquid nitrogen. A suture was used to mark each
Monajati et al 1987 ; Ross et al 1987). However, there lumbosacral disc level and the fixed and embedded dural
have been no reports of the position of the individual sac was lifted out of the vertebral canal for imaging.
spinal nerve roots on axial images, nor have the The CT images were obtained on a GE 9800 unit,
radiographic observations been correlated with anatom- performing scans at 80 KVP, 40 mA, with 1.5 mm slices
ical findings. and a 2 5 scan time. The MR images were obtained on a
We have recently identified an organised pattern of GE Signa 1 .5-Tesla superconductive unit with a circular
the intrathecal nerve roots in the cauda equina (Wall et surface-coil. A multi-echo sequence with a repetition
al 1990), using fixation and dissection. This had the time (TR) of 2200 ms and echo times (TE) of 20 ms and
potential for artifact by displacement of the roots. Our 60 ms were used, with a 256 x 256 acquisition matrix
new study was undertaken to validate this previous work and two excitations. Slices were 3 mm thick with a 1 .5 cm
and correlate it with contrast-enhanced CT scans and interslice gap and a field of view of 12 cm.
MRI studies. Each specimen was subsequently sectioned at each
disc level from L2-L3 to L5-Sl using a microtome knife
(Cambridge Instruments, Edison, New Jersey). Sections
MATERIALS AND METHODS
were photographed, drawn and correlated with the
Ten fresh adult human cadavers without evidence of corresponding CT and MR images. The specimens were
spinal surgery were studied, radiographs being taken of then dissected under 3.5 x loupe magnification, tracing
each specimen to exclude congenital abnormality or gross the individual spinal roots in their course through the
spinal pathology. The cauda equina was fixed in situ by thecal sac.

M. S. Cohen, MD, Orthopaedic Surgeon RESULTS

E. J. Wall, MD, Orthopaedic Surgeon
C. W. Kerber, MD, Radiologist
J.-J. Abitbol, MD, FRCS C, Assistant Professor of Surgery The nerve roots were well visualised on both contrast-
S. R. Garfin, MD, Professor ofOrthopaedic Surgery enhanced CT and surface-coil MR images, although
University ofCalifornia, San Diego, San Diego, CA 92103, USA.
individual roots were slightly more distinct on CT than
Correspondence should be sent to Dr S. R. Garfin at UCSD Medical
Center (H-894), 225 Dickinson Street, San Diego, CA 92103, USA. MR. In all cases, the nerve root patterns seen on the axial
images correlated precisely with the individual nerve
© 1991 British Editorial Society ofBone and Joint Surgery
0301-620X/91/3l70 $2.00 roots seen at dissection. We discuss the observed image
JBoneJointSurg[Br] 1991 ; 73-B:38l-4.
pattern in relation to the identified nerve roots at each

VOL. 73-B, No. 3, MAY 1991 381

382 M. S. COHEN, E. J. WALL, C. W. KERBER, J.-J. ABITBOL, S. R. GARFIN

CONTRAST-ENHANCED CT SCAN SURFACE-COIL MRI ANATOMICAL DIAGRAM

Fig. lc

LS-Sl intervertebral level. A crescentic pattern ofnerve roots is seen within the thecal sac, with the SI root lying anterior and lateral.

L4-L.

Fig. 2c

L4-L5 intervertebral level. The LS root is situated anterolaterally, displacing the 51 root, and the lower sacral roots are positioned dorsally.
There is clear separation of the motor (anterior and medial) and sensory (posterior and lateral) roots.

Fig. 3c

L3-L4 intervertebral level. Although the anatomy is more crowded, the motor bundles of the L4, LS and 51 nerve roots are visible anterior
and medial to their corresponding multifascicular sensory bundles.

Fig. 4c

L2-L3 intervertebral level. The roots occupy most of the thecal sac at this level, with the motor bundle anterior to its sensory counterpart for

each root.

THE JOURNAL OF BONE AND JOINT SURGERY

 THE ANATOMY OF THE CAUDA EQUINA ON CT SCANS AND MRI 383

SECtION OF SPECIMEN disc level. The pattern is best appreciated by progression

from caudal to cephalad (Figs 1 to 4).
L5-S1 (Fig. 1). At the L5-Sl disc level, a crescent-shaped
pattern ofnerve roots is seen on both CT and MR images.
The 51 root is seen as a projected large anterolateral
density, with its motor bundle always anteromedial to
the sensory component. The lower sacral roots are
dorsally placed in the thecal sac.
L4-L5 (Fig. 2). At the L4-L5 interspace, the L5 root
assumes an anterolateral position, displacing the 51 root
posteromedially. The lower sacral roots are grouped
dorsally.
L3-L4 (Fig. 3). At the L3-L4 disc level, the imaged nerve
roots become more crowded, although a layered pattern
is still discernible. The L4 root displaces the L5 and 51
Fig. ld
roots dorsally and towards the mid-line. These three root
bundles assume an oblique, layered configuration. Within
each root layer, the motor bundle remains medial and
anterior to its corresponding larger and multifascicular
sensory bundle. The 52-55 roots remain dorsal.
L2-L3 (Fig. 4). The L3 root adds to the layered pattern at
the L2-L3 intervertebral level. The nerve roots occupy
the majority of the thecal sac at this level and are closely
.. 1 apposed. Although the individual roots are less distinct
on both CT and MR images, the layered root pattern can
still be appreciated on some specimens. Dissections
revealed a tightly apposed oblique layering of individual
nerve roots at this level. The motor bundles form the
ventral surface of the layers.

DISCUSSION

Both contrast-enhanced CT (Resj#{246}et al 1979 ; Naidich et

al 1980; Haughton and Williams 1982; Dorwart and
Genant 1983; Takata et al 1988) and surface-coil MRI
(Reicher et al 1986; Monajati et al 1987; Ross et al 1987)
allow the visualisation ofcauda equina nerve roots within
the thecal sac, and several authors have described a
crescent-shaped pattern of nerve roots (Resj#{246}et a! 1979;
Naidich et al 1980 ; Ross et al 1987). At more caudal
lumbar levels the roots are said to separate into “delicate
entities” (Ross et al 1987), becoming more diffuse
(Monajati et al 1987), and more dorsal (Resj#{246}et al 1979)
Fig. 3d
within the thecal sac.
Our study has identified and described the individual
nerve roots seen on both types of image at each disc level
of the cauda equina, showing a crescentic pattern at the
lower lumbar levels. More cephalad, the roots group
together and become less distinct, but dissections have
shown that the oblique layered pattern can be related to
individual roots. The lower sacral roots (S2-S5) remain
in the dorsal aspect of the thecal sac, while the lumbar
and first sacral roots progressively separate before leaving
the anterolateral part of the canal.
Naidich et al (1980) noted that the imaged roots
were arranged in pairs. We have shown that within each
Fig. 4d root layer, the motor bundle is anterior and medial to its

VOL. 73-B. No. 3, MAY 1991

384 M. S. COHEN, E. J. WALL, C. W. KERBER, J.-J. ABITBOL, S. R. GARFIN

Contrast-enhanced CT scan
of a patient at the L4-LS
Fig. S intervertebral level. The lay-
ered pattern of nerve roots is
confirmed in vivo.

Three-dimensional reconstructions from axial CT sections of the cauda equina nerve roots. Figure 6a - Ending at the L4-LS intervertebral
level. Figure 6b - Ending at the L3-L4 level. These reformations allow the individual roots to be traced through the thecal sac.

larger sensory bundle. These separate components can No benefits in any form have been received or will be received
from a commercial party related directly or indirectly to the subject of
be visualised on both CT and MR images for the lumbar this article.
and first sacral roots.
This nerve root arrangement is not an artefact of the REFERENCES
technique : the imaging studies confirm that any dissec-
Dorwart RH, Genant HK. Anatomy of the lumbosacral spine. Radio/
tion artefact is minimal. Contrast-enhanced CT images C/in North Am 1983; 21 :201-20.
performed on living patients for the evaluation of sciatica Haughton VM, Williams AL. Computed tomography of the spine. St
have verified this nerve root arrangement (Fig. 5). Louis, etc : CV Mosby Co. 1982.
Using computer software and our CT images, it has Moniijati A, Wayne WS, Rauschning W, Ekholm SE. MR of the cauda
equina. AJNR 1987; 8:893-900.
been possible to construct three-dimensional models of
NaidiCh TP, King DG, Moran CJ, Sagel SB. Computed tomography of
the nerve roots in the cauda equina (Fig. 6). This enabled
the lumbar thecal sac. J Comput Assist Tomogr 1980; 4:37-41.
us to trace individual spinal roots through the thecal sac, Reicher MA, Gold RH, Halbach VV, Rauschning W, Wilson GH, Lufkin
and confirmed the nerve root anatomy between each RB. MR Imaging of the lumbar spine : anatomic correlations and
the effects oftechnical variations. AJR 1986:147:891-8.
axial section. It is possible to rotate these images and to
Resjo IM, Harwood-Nash DC, Fitz CR, Chuang S. Normal cord in
study sections in any desired plane. As CT and MR infants and children examined with computed tomographic
imaging techniques improve, the identification and metrizamide myelography. Radiology 1979; 130 :691-6.
tracing of individual nerve roots will help the understand- Ross JS, Masaryk TJ, Modic MT, et al. Delamater R, BoWman H,
Wilbur G, Kaufman B. MR imaging oflumbar arachnoiditis. AJR
ing and management ofabnormalities ofthe lumbosacral 1987; 149:1025-32.
spine. Takata K, Inoue S-I, Takahashi K, Ohtsuka Y. Swelling of the cauda
equina in patients who have herniation of a lumbar disc : a possible
The authors wish to express their appreciation to East County
pathogenesis of sciatica. J hone Joint Surg [Am] 1988 ; 70-A:
Radiology, La Mesa, California for the use of its CT facilities, and the
361-8.
University of California, San Diego, Magnetic Resonance Institute for
the use of its scanner. We would also like to thank Dimensional Wall EJ, Cohen MB, Massie JB, Rydevik B, Garfin SR. Cauda equina
Medicine Inc, Minnetonka, Minnesota for the use of its three- anatomy I : intrathecal nerve root organization. Spine 1990;
dimensional computerised software. 15:1244-7.

THE JOURNAL OF BONE AND JOINT SURGERY