James
A. Brunberg,
Robert
C. Dauser,
J.
Constance
MD #{149}Michael A. DiPietro, MD #{149}Joan MD #{149}Karin M. Muraszko, MD #{149}Gregory D’Amato, BS #{149}Jonathon M. Rubin, MD, PhD
L. Venes, MD S. Berkey, MD
Intramedullary Lesions ofthe Pediatric Spinal Cord: Correlation of Findings from MR Imaging, Intraoperative Sonography, Surgery, and Histologic
M
Findings of preoperative magnetic resonance (MR) imaging and radiologist-directed intraoperative sonography (lOS) were correlated with surgical and pathologic findings in 11 pediatric patients with intramedullary spinal cord lesions. There were seven gliomas and one each of primitive neuroectodermal tumor, venous vascular malformation, neurenteric cyst, and active schistosomiasis. MR imaging provided discrete preoperative anatomic localization and cxcluded multicentric lesions but did not reliably distinguish between solid and cystic lesions. lOS helped (a) define the limits of intramedullary mass before the dura mater was opened and (b) differentiate cystic from solid components. The internal architecture of intramedullary lesions, as shown with MR imaging and sonography, was strikingly similar, allowing discrete correlative tocalization for biopsy or tissue resection. Gadolinium-enhanced MR imaging and lOS are complementary imaging techniques that should be used in concert for the evaluation and management of intramedullary lesions of the pediatric spinal cord. Both techniques display regions of cord abnormality, but neither definitively characterizes underlying tissue histology.
Ultrasound
(US),
intraoperative,
I
From
1991;
the
K.M.M.), Neurology Michigan Hospital, revision requested JAB. RSNA, 1991
(JAB.,
1.5 T in nine patients patients. Spin-echo
infection,
vascular
is helpful
and
or regions
inflamma-
neoptastic
pro-
of surrounding
for
guiding
selective
MAD.,
J.M.R.),
B2B311,
Surgery
(J.A.B.,J.L.V.,
as seen
biopsy
R.C.D.,
with
identified
MR
with
imaging,
Eleven
AND
patients
and
be
METHODS
aged
2-15 years who spinal cord mass lepreoperative MR
had intramedullary sions and underwent
imaging
could
lOS.
MATERIALS
radiologist-directed
LOS were
studied. No patient had previously undergone spinal cord surgery or irradiation treatment. MR imaging was performed at
with
or tumor resection. Intraoperative sonography (lOS) provides this capability. Although the use of 105 in many centers has become routine for the localization and characterization of an intraspinal mass (11-14), a direct comparison with MR imaging has not been published previously. The purpose of this study was to correlate preoperative MR imaging of intramedullary spinal cord lesions with lOS and surgical and histologic findings. Our goals were to determine whether regions of altered signal intensity or contrast enhancement on
341.1298,
of Radiology
intramedullary
edema or gliosis (1-4,8,9). Despite the detailed anatomy displayed on preoperative MR images, the surgeon may see only a widened spinal cord after opening the dura mater. Although there also may be superficial discoloration or hyperemia suggesting the tumor site, the sungeon is otherwise blinded with respect to internal cord anatomy. The ability to directly correlate and localize MR findings to the operative field
181 :573-579
Departments
MR images correlated with regions of echogenicity at lOS and whether anatomic patterns or margins of lesions,
AGNETIC
cesses,
351.1298 Radiology
resonance (MR) imaging is increasingly used as the sole preoperative diagnostic procedure in suspected intramedullary lesions of the spinal cord (i-9). For patients with congenital structural alterations, the discrete anatomic detail and soft-tissue resolution of MR imaging often allow a specific preoperative diagnosis (10). For more complex lesions and for acquired intramedullary lesions, MR imaging, though anatomically precise, does not consistently help distinguish between tion,
Index terms: Gadolinium #{149} Magnetic resonance (MR), contrast enhancement, 351.12988 Schistosomiasis, 351.2084 #{149} Spinal cord, cysts, 351.2084, 351.3611 #{149} Spinal cord, MR studies, 341.1214, 351.1214 #{149} Spinal cord, neoplasms, 341.3632, 341.3636, 351.3632, 351.3635, 351.365
Study’
Ti
and at 0.35 T in two images were obtained
weighting
(500-700/20-30
[repeti-
tion time msec/echo time msec]) and T2 weighting (2,000-3,000/80-100) in sagittal and axial planes. Six patients underwent Ti-weighted imaging before and after receiving 0.i mmol/kg gadopentetate dimeglumine
Berlex
Imaging,
Wayne, NJ). MR imaging features tramedullary lesions and adjacent
of incord
were
(Magnevist;
characterized
by
their
location,
con-
tour, signal intensity, and pattern of enhancement with gadopentetate dimeglumine. Localization of the site of the lesion for the correlation of findings from MR imaging, lOS, and surgery was accomplished with
radiographic
verification
of anatomic
reference points included in a scout-view MR image with a large field of view on with
the
fiducial
location
of the
markers
markers
placed
consisted
lesion
relative
on the
skin.
of vitamin
A capsules
taped adjacent to the spinous that one on two markers were of view. Before their removal, tion
was
recorded
on
the
to
The
processes in each their with
skin
so field
posian in-
delibte felt-tip marker. By using the MR image, the distance from the position of the fiducial marker to the site of surgical interest
could
be calculated.
tance
was then
mapped
eration, utilizing the as a point of reference. Radiologist-guided
performed
after
The
same
dis-
on the skin at opmarking real-time
laminectomy
on
the 105
skin was
and before
U.A.B.), and Pathology (C.J.D.), and School of Medicine (G.S.B.), University of 1500 E Medical Center Dr, Ann Arbor, MI 48109-0030. Received March 28, 1991; May 15; revision received June 7; accepted June 10. Address reprint requests to Abbreviations: los = intraoperative
CSF
= cerebrospinal sonography.
fluid,
573
a.
b.
Figure
1. S. mansoni fluent enhancement T2-weighted images
d.
C.
myelitis. (a-c) MR images (600/20 and 3,000/90) demonstrate widening of the caudal spinal cord after the administration of gadopentetate dimeglumine (b), and diffuse increased signal intensity (c). (d) LOS demonstrates markedly increased echogenicity (arrows) in the anterior portion of the
b.
a. Figure
2.
because
Spinal
cord
of movement
venous
angioma,
venous between
with
linear
angioma. regions
c.
(a, b) MR images
sequences.
(a), nodular and conof cord parenchyma on lower spinal cord.
(600/20
(c) LOS performed
of echogenicity
and
before
within
2,500/60) the
the cord
at 0.35 T. Sagittal
dura
mater
at the rostral
is opened
T2-weighted demonstrates
extent
of the angioma
spinal
cord
image increased
possibly
(b) is off axis relative echogenicity
representing
to a
of the
venous
path-
ways.
opening
the
dura
mater
in alt patients.
LOS was performed through solution-filled laminectomy 7.5-MHz acoustic
transducer that gel and enclosed
tic sheath Technology
574
(NeuroSect Laboratories,
Radiology
#{149}
Neither
the saline site with a
was coated in a sterile
with plas-
OR; Advanced Bothell,
touched
ous
serial
in sagittal
tenpretation gist
Wash).
the
was
dura
mater
real-time
images
and
planes.
axial
was provided
to the neurosurgeon,
recordings
nor
the
by the transducer.
were
made
Contiguwere
obtained
Lmmediate
in-
by the radioloand
cassette
of LOS. Images
particular with were
tape
of
interest a Polaroid correlated
aging reached MAD.) K.M.M.).
were
photographed
camera. Surgical with preoperative
and 105, and a consensus
findings MR
im-
was
between radiologists (JAB., and neurosurgeons (J.L.V., R.C.D., MR images and videocassette
November
1991
Figure images
3. Grade II astrocytoma. (600/20 and 3,000/90)
after (b) the administration dimeglumine spinal canal
extending
upper
(a-c) before
MR (a) and
of gadopentetate
demonstrate widening of the by an intramedullary cord lesion
from
the foramen
thoracic
magnum
spine. There
to the
is enhancement
of a mixed solid and cystic component that extends from C-5 to T-3. Increased signal intensity
on
T2-weighted
images
is seen
from the foramen magnum to T-3 (c). (d, e) LOS demonstrates the enhanced portion of the mass to be only slightly more echogenic
region tral
a.
than
the
of gliosis
to this
noncystic
and
level.
but
edema
indicates
*
enlarged
in the cord
ros-
C-5.
C.
D.
dura mater was opened, lOS demonstrated fusiform cord enlargement with diffusely increased echogenicity at the site of enlargement. There were linear regions of increased echogenicity at the cord surface and within the cord rostral to the site of greatest enlargement. When the dura mater was opened, enlarged veins were evident on the cord surface. A venous angioma was partially resected. Color flow Doppler sonography failed to reveal increased flow within the cord parenchyma. Patient
3
A 3-year-old
d.
e.
recordings
viewed
made
at LOS
by the same
complete
surgical
pathologic
were
findings
results
later
investigators were
re-
when
and
tissue
available.
RESULTS Preoperative were completed
MR studies and lOS in ii patients aged
2-15
intramedullary
years
with
spinal
cord lesions. The lesions included seven gliomas (four astrocytomas, two ependymomas, and one other glioma),
and
one
each
are
Patient
1
discussed
below.
MR images of a 6-year-old with spinal cord widening strated
multinodular
and
child demonconfluent
gadolinium enhancement of the anterior two-thirds of the cord at T-iO to
Volume
181
Number
#{149}
with
increased
intensity
on
T2-
weighted images from T-8 through the conus (Fig 1). lOS showed increased echogenicity of the anterior half of the cord at T-iO to L-1. At surgery, regions of MR enhancement correlated with the site of echogenicity
and
strating crosis
the location of tissue Schistosoma mansoni and inflammation.
demon-
ova
ne-
2
cystic
regions
was C-5.
identical T2-weighted
Patient
clear
spinal mass
than
2
At MR
imaging,
an
u-year-old
child had an intramedullary mass extending from T-i to T-4 (Fig 2). Except for curvilinear regions of low signal intensity, the mass was isointense with normal cord parenchyma on Tiweighted images. The mass and the rostral cord demonstrated increased intensity on T2-weighted images, except for curvilinear regions of low signal intensity similar to those seen on Ti-weighted images. Before the
demonstrated
on
Ti-weighted
from
cord was
the
signal intensity medulla. It was
MR
rostrat cystic
it as solid
normal
images
to that of the cord images demon-
strated increased the cord and the
strated
of primitive
neuroectodermal tumor, venous vascular malformation, neurentenc cyst, and active schistosomiasis. Selected patients
T-12,
child
spinal cord widening and decreased intensity of the cord from the foramen magnum to T-6 (Fig 3). The spinal cord from C-5 to T-3 vertebral body levels demonstrated marked enhancement, except for central wellmarginated regions of low signal intensity, which were later demonstrated at sonography and surgery to be cystic. Signal intensity within these
cord,
images
above
of un-
whether
to the enhanced or solid. and
and
105 more
demonechogenic
histologic
ings from a biopsy specimen gtiosis rostral to an enhanced
findshowed grade
II
astrocytoma. The margin between enhanced and nonenhanced cord parenchyma that was seen with MR imaging could not be accurately defined with sonography, although the enhanced portion of the mass was somewhat more echogenic. Both lOS and MR imaging demonstrated remodeling of the posterior aspects of vertebral bodies and widening of the spinat canal from C-S to T-3.
Radiology
575
#{149}
Figure 4. Grade II oligoastrocytoma. (a, b) MR images (600/20) at the middle thoracic level without contrast enhancement (a) and after the administration of gadopentetate dimeglumine (b) demonstrate an expanded spinal canal with a complex mass of mixed signal intensity. There is inhomogeneous enhancement of both the central portion of the mass and of the margins of the apparently cystic components of the lesion. (c) T2-weighted image (3,000/90) centered
lower
than
intensity (arrow). signal tion
a and b demonstrates
at a site of suspected Sharply marginated intensity
of
the
are mass.
seen
(d)
in
lOS
low signal hemorrhage regions of high
the
central
cystic
component at the caudal mass and the echogenic wall of row). (e) The central portion of a small cyst that correlates with
on the gadolinium-enhanced less cystic than that suggested weighted
por-
demonstrates
the
extent of the cyst the mass the one
the (arhas seen
images but is with the T2-
image.
Patient
a.
b.
C.
4
MR imaging performed on a i3year-old patient showed a multiplecompartment lower thoracic and lumbar
spinal
cord
mass,
portions
which demonstrated with gadopentetate 4). Before
the
of
enhancement dimeglumine
dura
mater
was
(Fig opened,
lOS demonstrated septa and cystic portions of the tumor to be morphologically identical to those seen with MR imaging, which distinguished rostral
and
caudal
syrinx
formation
from regions of cystic alteration within the mass on the basis of mural enhancement of the tumor cysts. On the basis of cyst fluid echogenicity, sonography did not help distinguish between intratumoral cyst and adjacent syrinx, although the eccentric location, irregular contour of the cyst wall, and echogenic surrounding tissue distinguished regions of tumoral cyst
formation,
which
MR findings. trast
correlated
All regions
enhancement
sonographic
were images.
Central
on
the
T2-weighted
e.
with
weighted
of MR conechogenic
on
nonen-
hanced portions of the mixed grade oligodendroglioma astrocytoma, which had tow signal intensity on Ti-weighted images and high signal intensity on T2-weighted images, were found at sonography (Fig 4e) and at visual inspection to be markedly less cystic and to have a greater component of solid tissue than that suggested
d.
images.
II
images
intensity 5). It was
and
5
MR images of this 7-year-old child demonstrated enlargement of the cord caudal to C-7, with central decreased signal intensity on Ti-
576
Radiology
#{149}
images the
signal
(CSF),
(Fig find-
nostral
and signal intensity and caudal to the lOS demonstrated
normal.
ings represented a mass or a syrinx. 105 helped localize the changing cord contour at C-7 and demonstrated the lesion to be solid and homogeneously
nor location canal, and
echogenic, with a distinct rostral interface with normal cord. The pathologic diagnosis was ependymoma.
the
tral
otogic
6
MR
demonstrated
of a 6-year-old a sharply
extramedutlary and lesion at T-2, which cord and 6). Signal
displaced intensity
identical
to that
child
marginated
intramedultary compressed
diagnosis
it posteriorly of the lesion
of cerebrospinal
fluid
mass onstrated
in the showed
spinal a ros-
cyst margin that echogenicity of normal. The pathneurenteric
cyst.
imaging
of a 2-year-old
child
a welt-marginated
of low-signal images
inhomogeneous below
of the cord lesion was the ante-
7
weighted
(Fig was
The was was
demonstrated gion
the
of the mass MR imaging
extramedullary not resolved. adjacent cord
was
Patient Patient
MR imaging Patient
increased
on T2-weighted not clear whether
this
intensity rostral low
level.
extending gadolinium
An
from
re-
on Tito C-4, with
signal
intensity
intramedullary C-S
to T-3
dem-
enhancement
November
1991
In the present series of patients, MR imaging, lOS, and surgical and pathologic findings were correlated with regard to the localization and extent of the lesion, the internal architecture of the lesion, and the tissue-imaging characteristics of the lesion. MR imaging allowed definitive preoperative localization of intramedullary lesions of the spinal cord, allowed for accurate characterization of the extent of the lesion, and excluded multiple sites of involvement. Relationships of the lesion to nerve roots and to the dura mater were evident with MR imaging and lOS. Surgical localization from the MR images was accurate in all cases
with
above. quickly
a.
b.
Figure 5. cord caudal strates the mogeneously
Ependymoma. (a) MR image (600/20) demonstrates an abrupt enlargement of the to C-7 (*). Cord signal caudal to this point is isointense with CSF. (b) LOS demonsite of transition (*) from normal cord to solid mass lesion. The mass consists of hoincreased echogenicity, with no regions of cystic alteration and no syrinx forma-
hon.
with intermixed regions of low signal intensity. T2-weighted images demonstrated homogeneously increased signal intensity throughout the cord, and it was not possible to distinguish between nonenhanced tumor and syrinx formation rostral to the region of enhancement. 105 demonstrated syrinx formation above and below the echogenic mass, with the mass having small
regions
of central
cyst
forma-
tion. The central rostral curvilinear contour of the mass seen with MR imaging was easily correlated with sonographic findings. Regions of MR contrast
enhancement
echogenic.
The
was
II astrocytoma.
grade
were
pathologic
markedly
diagnosis
MR imaging techniques allow localization of an intramedullary spinal cord mass lesion to its specific compartment and segmental level within the spinal canal but are not specific for tissue type, do not distinguish between nonenhanced tumor and surrounding edema, and do not reliably distinguish between cystic and solid components (i-9). To supplement preoperative MR imaging during the
181
Number
#{149}
105
approach to spinal cord has been increasingly
lized,
both
2
leuti-
continuous charand abnormal cord anatomy through the opened or unopened dura mater (ui-i4) and to allow correlation of operative findings with preoperative images. 105 demonstrates the normal posterior dura mater as a thin echogenic membrane with the anterior dura mater obscured by adjacent echogenic bone and ligaments. The posterior surface of the vertebral bodies is more echogenic than are disk spaces, allowing identification of segmental intervals. An alteration in the dimension or anterior contour of the spinal canal acterization
to allow
of normal
is easily recognized. ments are echogenic, hesions, vessels, and
DISCUSSION
Volume
surgical sions,
The dentate as are dural nerve roots
versing the subarachnoid
subarachnoid space. space around the
is normally
anechoic,
and
with
ligaadtra-
The cord real-
time sonography, pulsation can be demonstrated in the anterior spinal artery. Although the surface of the spinal cord is echogenic, the normal cord parenchyma is relatively hypoechoic, except for a central linear echogenic complex that may represent the central canal or anterior median fissure.
the
With
techniques
lOS,
identified
the
described
mass
and
the
was surgeon
was assured that the extent of the laminectomy was appropriate before opening the dura mater. Sonography was again used after the dura mater was opened, if further image characterization of the mass was necessary before myelotomy or biopsy was performed. This approach minimized damage to normal cord parenchyma by limiting the myelotomy and allowing the greatest potential access to the central portion of the mass. In all cases, the similarity between the surface and internal architecture of the lesions, as shown at preoperative MR imaging and at lOS, allowed immediate direct correlation of the operative field with preoperative MR imaging. A site selected for biopsy on the basis of MR imaging characteristics
could
be quickly
In the patient inflammation miasis
(Fig
identified
at lOS.
with granulomatous secondary to schistoso1), focal
gadolinium
en-
hancement and hyperechogenicity were limited to the ventral aspect of the cord and corresponded exactly to the focal necrosis found at surgery. Four of the seven gliomas and none of the other four lesions demonstrated large intramedullary fluid cotlections (syringohydromyelia) rostral or caudal to the neoplasm. Identification of the interface between the tumor and syrinx was quickly accomplished with lOS, allowing the surgeon to establish a plane of dissection between the tumor and surrounding cord (Figs 4, 6). Although internal architecture and cord contour were well demonstrated with both modalities, neither MR imaging nor lOS provided specific solidtissue series,
characterization. tumors, regions
tion, gliosis, and edema echogenic than normal
In the current of inflamma-
were more cord paren-
Radiology
577
#{149}
a.
b.
C.
Figure 6. Neurenteric cyst. (a, b) MR images (600/20 and 3,000/100) demonstrate a well-marginated that distorts the contour of the spinal cord and displaces it posteriorly. Signal intensity parallels mater is opened demonstrates posterior displacement of the cord and intramedullary extension rostral
margin
of the
neurenteric
cyst
is well
defined
chyma, as previously reported (iii4). Intramedullary tumors were associated with expansion of cord contour, alteration of intramedullary anatomy, and loss of the central echo complex. In some cases, edema and gliosis could not be differentiated from tumor margins. After the resection of a spinal lesion, 105 allowed the surgeon to assess the completeness
of tumor
evacuation,
the
reex-
pansion of subarachnoid spaces, the effectiveness of cyst or syrinx drainage, or the position of an implanted shunt device. Completeness of resection was better assessed in cases with sonographically welt-marginated lesions
(Figs
i, 5, 6) than
for
glioma
sur-
rounded by biopsy-proved echogenic nontumorat gliosis and edema (Fig 3). Although it is important in operative planning for the neurosurgeon to know whether an intramedullary lesion has a substantial cystic component, solid lesions of the spinal cord and those that contain fluid cannot be routinely distinguished with MR imaging (iS,i6). Cysts generally are nonenhanced areas of low signal intensity on Ti-weighted images and high signat intensity on T2-weighted images. Low signal intensity on T2-weighted images
relating
to flow,
turbulence,
or
phase shift induced by fluid pulsation may occur within a large syrinx but is uncommon within a tumor cyst (i6). None of our patients with either tu578
Radiology
#{149}
(open
intramedullany and extramedullary of CSF. (c) LOS performed before
of the anechoic
cystic
lesion
(solid
the
mass dura
arrow).
A
arrow).
mor cyst or syrinx formation demonstrated low signal intensity on T2weighted images. On Ti-weighted images, cystic cavities can alternatively be isointense or hyperintense, relative to surrounding cord parenchyma if they contain fluid of high protein content or methemoglobin secondary to recent hemorrhage (17). Benign cystic lesions in our series had sharply defined margins, uniform signat intensity, and isointensity of cyst contents relative to CSF (i8). Regions of cord gliosis, edema, or demyelination can be identical in intensity at MR imaging to nonflowing fluid within a cyst or syrinx. Ultrasonography demonstrates cysts as anechoic structures with increased through transmission, whereas regions of edema and gliosis are hyperechoic relative to normal cord parenchyma. Only one patient has reportedly had a homogeneously echogenic intramedutlary cord lesion that was found to be cystic at 105; the lesion contained viscous squamous debris as a component of a teratoma (i9). Cystic
that
structures
rostral
or caudal
to
the mass generally expand the central canal, have smooth walls, cause symmetric expansion of the cord, and represent syrinx formation rather than tumor cyst or necrosis (ii). Tumor cysts are generally smaller, may have irregular walls, and are eccentric in
position series,
within the
the
walls
cord
were
(ii).
more
In our echogenic
than those of normal cord. Cystic alterations occurred both in ependymomas and astrocytomas in our series. Sonography without Doppler analysis may cysts
not distinguish and enlarged
between vascular
small channels
within a mass lesion. Although color flow Doppler imaging has correctly demonstrated arteriovenous malformations in the brain and spinal cord (i4), flow in the venous malformation encountered in the current series could not be demonstrated. Presumably,
blood
flow
was
slow
and
the
equipment available was insensitive to it. Components of the intramedullary mass
that
demonstrated
hancement consistently Not
contrast
en-
with MR imaging were echogenic at sonography.
all echogenic
regions,
however,
demonstrated contrast enhancement. All neoplasms were echogenic, as were regions of edema, gliosis, inflammation, hemorrhage, and venous vascular malformation. Regions of contrast enhancement within a cord mass lesion may be the most productive sites for tissue biopsy, and on the basis
of our
are echogenic raphy
without
experience,
these
at sonography.
regions
Sonog-
contrast-enhanced
MR
imaging for correlation may not distinguish between the tumor and other areas of echogenicity, as deNovember
1991
scribed above. Tissue echogenicity relates to the number of reflecting interfaces encountered by the sound wave. Because gadopentetate dimeglumine
does
not
enhance
all
tumor tissue, it is not always possible with either sonography or MR imaging to distinguish between regions of tumor, gliosis, vascular malformation, edema, and inflammation. Ependymomas have been reported at 105 to be central, symmetric, better marginated within the spinal cord, and more echogenic than astrocytomas (ii). In our limited series, this differentiation could not be confirmed, though it was characteristic of the ependymoma of patient 5. No correlation could be identified between characteristics noted at sonography and those at histologic examination. Homogeneous versus inhomogeneous and diffuse versus focal increased echogenicity patterns noted in the neoplasms of this series did not correlate with the type of tumor. A limitation to image-pathologic conelation in this series stemmed from the method of tissue resection, which incorporated
a piece-by-piece
dissection
of the tumor from the cord with an ultrasonic surgical aspirator (Cavitron Lasersonics, Stanford, Conn) under the operating microscope. Also, the classification of gliomas was not always distinct. Microscopically, ependymomas typically appear as a dense uniform “carpet” of cells, whereas astrocytomas have a looser fibrillary background matrix. The ependymoma of patient S had typical cells, but they were within a loose background matrix that was more typical of an astrocytoma. The astrocytoma of patient 4 had mixed components that resulted in its being classified as oligoastrocytoma, whereas patient 3 had similar oligodendroglial elements only to a slight degree. Tissue characterization as to tumor type or to the presence of a malignant versus a nonmalignant process was not possible with either MR imaging or with lOS. Although MR imaging does not specifically demonstrate calcification,
the presence of scalloped posterior margins of vertebral bodies resulting from chronic mass effect was readily apparent. Similar scalloping and enlargement of the spinal canal was noted at 105 (Figs 3, 6). MR imaging and lOS were found to be complementary techniques for the imaging of intramedullary cord lesions. Alterations in cord contour and patterns of structural intramedullary alteration seen with MR imaging could be quickly correlated with sonographic images, allowing rapid intraoperative correlation
181
Number
#{149}
2
localization findings
7.
Donovan Post MJ, Quencer RM, Green BA, et al. Intramedullary spinal cord metastases, mainly of nonneurogenic origin. AJR 1987; 148:1015-1022. Goy AMC, Pinto RS, Raghavendra BN, Epstein FJ, Kricheff II. Intramedullary spinal cord tumors: MR imaging, with emphasis on associated cysts. Radiology 1986; 161: 381-386.
8.
Sze G, Krol G, Zimmerman
9.
before
2.
3.
4.
Sze G, Stimac
GK, Bartlett
C, et al.
Multi-
center study of gadopentetate dimeglumine as an MR contrast agent: evaluation in patients with spinal tumors. AJNR 1990; 11:967-974. Dillon WP, Norman D, Newton TH, Bolla K, Mark A. Intradural spinal cord lesions: Gd-DTPA-enhanced MR imaging. Radiology 1989; 170:229-237.
Sze G, Bravo S. Krol G.
Spinal
lesions:
quantitative and qualitative temporal evolution of gadopentetate dimeglumine enhancement in MR imaging. Radiology 1989; 170:849-856. Panizel PM, Baleniaux D, Rodesch G, et al. Gd-DTPA-enhanced MR imaging of spinal tumors. AJNR 1989; 10:249-258. Di Chiro G, DoppmanJL, Dwyer AJ, et al. Tumors and anteriovenous malformations of the spinal cord: assessment using MR. Radiology 1985; 156:689-697.
investigation
and comparison
with
10.
11.
12.
13.
14.
15.
16.
References 1.
RD, Deck MDF.
Intramedullary disease of the spine: diagnosis using gadolinium-DTPA-enhanced MR imaging. AJR 1988; 151:1193-1204. Stimac GK, Porter BA, Olson DO, Gerlach R, Genton M. Gadolinium-DTPA-enhanced MR imaging of spinal neoplasms:
preliminary
and
opening the dura mater. lOS allowed the surgeon to define margins of an intramedullary lesion that was not visible at inspection of the cord surface. Combined lOS and preoperative MR imaging provided precise localization for determining the extent of the laminectomy and for defining the location and extent of myelotomy. It identified the depth of tumor resection relative to the location of the anterior surface of the spinal cord or to the conus, allowed assessment of completeness of tumor resection, and defined the presence or absence of syrinx or cyst adjacent to or within the mass. In patients with intramedullary spinal cord lesions, a more specific surgical approach, a more complete resection, and diminished longterm morbidity may be possible when preoperative contrast-enhanced MR imaging and 105 techniques are used in concert. U
5.
Volume
lesion with MR
6.
17.
18.
unenhanced spin-echo and STIR sequences. AJR 1988; 151:1185-1192. BrunbergJA, Latchaw RE, Kanal E, Bunk L Jr. Albnight L. Magnetic resonance imaging of spinal dysraphism. Radiol Clin North Am 1988; 26:181-205. Epstein FJ. The utilization of ultrasonography as a surgical adjunct in the radical excision of one hundred and eighty-six consecutive intramedullary spinal cord neoplasms. In: Marlin AE, ed. Concepts in pediatric neurosurgery. Vol 10. Basel, Switzerland: Karger, 1990; 54-66. Rubin JM, Dohrmann GJ. The spine and spinal cord during neurosurgical operations: real-time ultrasonography. Radiology 1985; 155:197-200. Quencer RM, Morse BMM, Green BA, Eismont FJ, Brost P. Intraoperative spinal sonography: adjunct to metrizamide CT in the assessment and surgical decompression of posttraumatic spinal cord cysts. AJR 1984; 142:593-601.
Rubin J.
in spinal
cord
Kjos BO, Brant-Zawadzki
M, Kucharczyk
W, Kelly WM, Norman D, Newton TH. Cystic intracranial lesions: magnetic resonance imaging. Radiology 1985; 155:363369. Williams AL, Haughton VM, Pojunas KW, Daniels DL, Kilgore DP. Differentiation of intramedullary
19.
Ultrasonography
surgery. In: Rubin JM, Chandler WF, eds. Ultrasound in neurosurgery. New York: Raven, 1990; 107-182. Rubin JM, Aisen AM, DiPietro MA. Ambiguities in MR imaging of tumoral cysts in the spinal cord. J Comput Assist Tomogr 1986; 10:395-398. Enzmann DR. O’Donohue J, Rubin JB, Shuer L, Cogen P. Silverberg G. CSF pulsations within nonneoplastic spinal cord cysts. AJR 1987; 149:149-157.
neoplasms
and
cysts
by
MR. AJR 1987; 149:159-164. PlattJF, Rubin JM, Bowerman RA, DiPietro MA, Chandler WF. Intraoperative sonographic characterization of a cystic intramedullary spinal cord lesion appearing as solid. AJNR 1988; 9:614.
Radiology
579
#{149}
A. Brunberg,
Robert
C. Dauser,
J.
Constance
MD #{149}Michael A. DiPietro, MD #{149}Joan MD #{149}Karin M. Muraszko, MD #{149}Gregory D’Amato, BS #{149}Jonathon M. Rubin, MD, PhD
L. Venes, MD S. Berkey, MD
Intramedullary Lesions ofthe Pediatric Spinal Cord: Correlation of Findings from MR Imaging, Intraoperative Sonography, Surgery, and Histologic
M
Findings of preoperative magnetic resonance (MR) imaging and radiologist-directed intraoperative sonography (lOS) were correlated with surgical and pathologic findings in 11 pediatric patients with intramedullary spinal cord lesions. There were seven gliomas and one each of primitive neuroectodermal tumor, venous vascular malformation, neurenteric cyst, and active schistosomiasis. MR imaging provided discrete preoperative anatomic localization and cxcluded multicentric lesions but did not reliably distinguish between solid and cystic lesions. lOS helped (a) define the limits of intramedullary mass before the dura mater was opened and (b) differentiate cystic from solid components. The internal architecture of intramedullary lesions, as shown with MR imaging and sonography, was strikingly similar, allowing discrete correlative tocalization for biopsy or tissue resection. Gadolinium-enhanced MR imaging and lOS are complementary imaging techniques that should be used in concert for the evaluation and management of intramedullary lesions of the pediatric spinal cord. Both techniques display regions of cord abnormality, but neither definitively characterizes underlying tissue histology.
Ultrasound
(US),
intraoperative,
I
From
1991;
the
K.M.M.), Neurology Michigan Hospital, revision requested JAB. RSNA, 1991
(JAB.,
1.5 T in nine patients patients. Spin-echo
infection,
vascular
is helpful
and
or regions
inflamma-
neoptastic
pro-
of surrounding
for
guiding
selective
MAD.,
J.M.R.),
B2B311,
Surgery
(J.A.B.,J.L.V.,
as seen
biopsy
R.C.D.,
with
identified
MR
with
imaging,
Eleven
AND
patients
and
be
METHODS
aged
2-15 years who spinal cord mass lepreoperative MR
had intramedullary sions and underwent
imaging
could
lOS.
MATERIALS
radiologist-directed
LOS were
studied. No patient had previously undergone spinal cord surgery or irradiation treatment. MR imaging was performed at
with
or tumor resection. Intraoperative sonography (lOS) provides this capability. Although the use of 105 in many centers has become routine for the localization and characterization of an intraspinal mass (11-14), a direct comparison with MR imaging has not been published previously. The purpose of this study was to correlate preoperative MR imaging of intramedullary spinal cord lesions with lOS and surgical and histologic findings. Our goals were to determine whether regions of altered signal intensity or contrast enhancement on
341.1298,
of Radiology
intramedullary
edema or gliosis (1-4,8,9). Despite the detailed anatomy displayed on preoperative MR images, the surgeon may see only a widened spinal cord after opening the dura mater. Although there also may be superficial discoloration or hyperemia suggesting the tumor site, the sungeon is otherwise blinded with respect to internal cord anatomy. The ability to directly correlate and localize MR findings to the operative field
181 :573-579
Departments
MR images correlated with regions of echogenicity at lOS and whether anatomic patterns or margins of lesions,
AGNETIC
cesses,
351.1298 Radiology
resonance (MR) imaging is increasingly used as the sole preoperative diagnostic procedure in suspected intramedullary lesions of the spinal cord (i-9). For patients with congenital structural alterations, the discrete anatomic detail and soft-tissue resolution of MR imaging often allow a specific preoperative diagnosis (10). For more complex lesions and for acquired intramedullary lesions, MR imaging, though anatomically precise, does not consistently help distinguish between tion,
Index terms: Gadolinium #{149} Magnetic resonance (MR), contrast enhancement, 351.12988 Schistosomiasis, 351.2084 #{149} Spinal cord, cysts, 351.2084, 351.3611 #{149} Spinal cord, MR studies, 341.1214, 351.1214 #{149} Spinal cord, neoplasms, 341.3632, 341.3636, 351.3632, 351.3635, 351.365
Study’
Ti
and at 0.35 T in two images were obtained
weighting
(500-700/20-30
[repeti-
tion time msec/echo time msec]) and T2 weighting (2,000-3,000/80-100) in sagittal and axial planes. Six patients underwent Ti-weighted imaging before and after receiving 0.i mmol/kg gadopentetate dimeglumine
Berlex
Imaging,
Wayne, NJ). MR imaging features tramedullary lesions and adjacent
of incord
were
(Magnevist;
characterized
by
their
location,
con-
tour, signal intensity, and pattern of enhancement with gadopentetate dimeglumine. Localization of the site of the lesion for the correlation of findings from MR imaging, lOS, and surgery was accomplished with
radiographic
verification
of anatomic
reference points included in a scout-view MR image with a large field of view on with
the
fiducial
location
of the
markers
markers
placed
consisted
lesion
relative
on the
skin.
of vitamin
A capsules
taped adjacent to the spinous that one on two markers were of view. Before their removal, tion
was
recorded
on
the
to
The
processes in each their with
skin
so field
posian in-
delibte felt-tip marker. By using the MR image, the distance from the position of the fiducial marker to the site of surgical interest
could
be calculated.
tance
was then
mapped
eration, utilizing the as a point of reference. Radiologist-guided
performed
after
The
same
dis-
on the skin at opmarking real-time
laminectomy
on
the 105
skin was
and before
U.A.B.), and Pathology (C.J.D.), and School of Medicine (G.S.B.), University of 1500 E Medical Center Dr, Ann Arbor, MI 48109-0030. Received March 28, 1991; May 15; revision received June 7; accepted June 10. Address reprint requests to Abbreviations: los = intraoperative
CSF
= cerebrospinal sonography.
fluid,
573
a.
b.
Figure
1. S. mansoni fluent enhancement T2-weighted images
d.
C.
myelitis. (a-c) MR images (600/20 and 3,000/90) demonstrate widening of the caudal spinal cord after the administration of gadopentetate dimeglumine (b), and diffuse increased signal intensity (c). (d) LOS demonstrates markedly increased echogenicity (arrows) in the anterior portion of the
b.
a. Figure
2.
because
Spinal
cord
of movement
venous
angioma,
venous between
with
linear
angioma. regions
c.
(a, b) MR images
sequences.
(a), nodular and conof cord parenchyma on lower spinal cord.
(600/20
(c) LOS performed
of echogenicity
and
before
within
2,500/60) the
the cord
at 0.35 T. Sagittal
dura
mater
at the rostral
is opened
T2-weighted demonstrates
extent
of the angioma
spinal
cord
image increased
possibly
(b) is off axis relative echogenicity
representing
to a
of the
venous
path-
ways.
opening
the
dura
mater
in alt patients.
LOS was performed through solution-filled laminectomy 7.5-MHz acoustic
transducer that gel and enclosed
tic sheath Technology
574
(NeuroSect Laboratories,
Radiology
#{149}
Neither
the saline site with a
was coated in a sterile
with plas-
OR; Advanced Bothell,
touched
ous
serial
in sagittal
tenpretation gist
Wash).
the
was
dura
mater
real-time
images
and
planes.
axial
was provided
to the neurosurgeon,
recordings
nor
the
by the transducer.
were
made
Contiguwere
obtained
Lmmediate
in-
by the radioloand
cassette
of LOS. Images
particular with were
tape
of
interest a Polaroid correlated
aging reached MAD.) K.M.M.).
were
photographed
camera. Surgical with preoperative
and 105, and a consensus
findings MR
im-
was
between radiologists (JAB., and neurosurgeons (J.L.V., R.C.D., MR images and videocassette
November
1991
Figure images
3. Grade II astrocytoma. (600/20 and 3,000/90)
after (b) the administration dimeglumine spinal canal
extending
upper
(a-c) before
MR (a) and
of gadopentetate
demonstrate widening of the by an intramedullary cord lesion
from
the foramen
thoracic
magnum
spine. There
to the
is enhancement
of a mixed solid and cystic component that extends from C-5 to T-3. Increased signal intensity
on
T2-weighted
images
is seen
from the foramen magnum to T-3 (c). (d, e) LOS demonstrates the enhanced portion of the mass to be only slightly more echogenic
region tral
a.
than
the
of gliosis
to this
noncystic
and
level.
but
edema
indicates
*
enlarged
in the cord
ros-
C-5.
C.
D.
dura mater was opened, lOS demonstrated fusiform cord enlargement with diffusely increased echogenicity at the site of enlargement. There were linear regions of increased echogenicity at the cord surface and within the cord rostral to the site of greatest enlargement. When the dura mater was opened, enlarged veins were evident on the cord surface. A venous angioma was partially resected. Color flow Doppler sonography failed to reveal increased flow within the cord parenchyma. Patient
3
A 3-year-old
d.
e.
recordings
viewed
made
at LOS
by the same
complete
surgical
pathologic
were
findings
results
later
investigators were
re-
when
and
tissue
available.
RESULTS Preoperative were completed
MR studies and lOS in ii patients aged
2-15
intramedullary
years
with
spinal
cord lesions. The lesions included seven gliomas (four astrocytomas, two ependymomas, and one other glioma),
and
one
each
are
Patient
1
discussed
below.
MR images of a 6-year-old with spinal cord widening strated
multinodular
and
child demonconfluent
gadolinium enhancement of the anterior two-thirds of the cord at T-iO to
Volume
181
Number
#{149}
with
increased
intensity
on
T2-
weighted images from T-8 through the conus (Fig 1). lOS showed increased echogenicity of the anterior half of the cord at T-iO to L-1. At surgery, regions of MR enhancement correlated with the site of echogenicity
and
strating crosis
the location of tissue Schistosoma mansoni and inflammation.
demon-
ova
ne-
2
cystic
regions
was C-5.
identical T2-weighted
Patient
clear
spinal mass
than
2
At MR
imaging,
an
u-year-old
child had an intramedullary mass extending from T-i to T-4 (Fig 2). Except for curvilinear regions of low signal intensity, the mass was isointense with normal cord parenchyma on Tiweighted images. The mass and the rostral cord demonstrated increased intensity on T2-weighted images, except for curvilinear regions of low signal intensity similar to those seen on Ti-weighted images. Before the
demonstrated
on
Ti-weighted
from
cord was
the
signal intensity medulla. It was
MR
rostrat cystic
it as solid
normal
images
to that of the cord images demon-
strated increased the cord and the
strated
of primitive
neuroectodermal tumor, venous vascular malformation, neurentenc cyst, and active schistosomiasis. Selected patients
T-12,
child
spinal cord widening and decreased intensity of the cord from the foramen magnum to T-6 (Fig 3). The spinal cord from C-5 to T-3 vertebral body levels demonstrated marked enhancement, except for central wellmarginated regions of low signal intensity, which were later demonstrated at sonography and surgery to be cystic. Signal intensity within these
cord,
images
above
of un-
whether
to the enhanced or solid. and
and
105 more
demonechogenic
histologic
ings from a biopsy specimen gtiosis rostral to an enhanced
findshowed grade
II
astrocytoma. The margin between enhanced and nonenhanced cord parenchyma that was seen with MR imaging could not be accurately defined with sonography, although the enhanced portion of the mass was somewhat more echogenic. Both lOS and MR imaging demonstrated remodeling of the posterior aspects of vertebral bodies and widening of the spinat canal from C-S to T-3.
Radiology
575
#{149}
Figure 4. Grade II oligoastrocytoma. (a, b) MR images (600/20) at the middle thoracic level without contrast enhancement (a) and after the administration of gadopentetate dimeglumine (b) demonstrate an expanded spinal canal with a complex mass of mixed signal intensity. There is inhomogeneous enhancement of both the central portion of the mass and of the margins of the apparently cystic components of the lesion. (c) T2-weighted image (3,000/90) centered
lower
than
intensity (arrow). signal tion
a and b demonstrates
at a site of suspected Sharply marginated intensity
of
the
are mass.
seen
(d)
in
lOS
low signal hemorrhage regions of high
the
central
cystic
component at the caudal mass and the echogenic wall of row). (e) The central portion of a small cyst that correlates with
on the gadolinium-enhanced less cystic than that suggested weighted
por-
demonstrates
the
extent of the cyst the mass the one
the (arhas seen
images but is with the T2-
image.
Patient
a.
b.
C.
4
MR imaging performed on a i3year-old patient showed a multiplecompartment lower thoracic and lumbar
spinal
cord
mass,
portions
which demonstrated with gadopentetate 4). Before
the
of
enhancement dimeglumine
dura
mater
was
(Fig opened,
lOS demonstrated septa and cystic portions of the tumor to be morphologically identical to those seen with MR imaging, which distinguished rostral
and
caudal
syrinx
formation
from regions of cystic alteration within the mass on the basis of mural enhancement of the tumor cysts. On the basis of cyst fluid echogenicity, sonography did not help distinguish between intratumoral cyst and adjacent syrinx, although the eccentric location, irregular contour of the cyst wall, and echogenic surrounding tissue distinguished regions of tumoral cyst
formation,
which
MR findings. trast
correlated
All regions
enhancement
sonographic
were images.
Central
on
the
T2-weighted
e.
with
weighted
of MR conechogenic
on
nonen-
hanced portions of the mixed grade oligodendroglioma astrocytoma, which had tow signal intensity on Ti-weighted images and high signal intensity on T2-weighted images, were found at sonography (Fig 4e) and at visual inspection to be markedly less cystic and to have a greater component of solid tissue than that suggested
d.
images.
II
images
intensity 5). It was
and
5
MR images of this 7-year-old child demonstrated enlargement of the cord caudal to C-7, with central decreased signal intensity on Ti-
576
Radiology
#{149}
images the
signal
(CSF),
(Fig find-
nostral
and signal intensity and caudal to the lOS demonstrated
normal.
ings represented a mass or a syrinx. 105 helped localize the changing cord contour at C-7 and demonstrated the lesion to be solid and homogeneously
nor location canal, and
echogenic, with a distinct rostral interface with normal cord. The pathologic diagnosis was ependymoma.
the
tral
otogic
6
MR
demonstrated
of a 6-year-old a sharply
extramedutlary and lesion at T-2, which cord and 6). Signal
displaced intensity
identical
to that
child
marginated
intramedultary compressed
diagnosis
it posteriorly of the lesion
of cerebrospinal
fluid
mass onstrated
in the showed
spinal a ros-
cyst margin that echogenicity of normal. The pathneurenteric
cyst.
imaging
of a 2-year-old
child
a welt-marginated
of low-signal images
inhomogeneous below
of the cord lesion was the ante-
7
weighted
(Fig was
The was was
demonstrated gion
the
of the mass MR imaging
extramedullary not resolved. adjacent cord
was
Patient Patient
MR imaging Patient
increased
on T2-weighted not clear whether
this
intensity rostral low
level.
extending gadolinium
An
from
re-
on Tito C-4, with
signal
intensity
intramedullary C-S
to T-3
dem-
enhancement
November
1991
In the present series of patients, MR imaging, lOS, and surgical and pathologic findings were correlated with regard to the localization and extent of the lesion, the internal architecture of the lesion, and the tissue-imaging characteristics of the lesion. MR imaging allowed definitive preoperative localization of intramedullary lesions of the spinal cord, allowed for accurate characterization of the extent of the lesion, and excluded multiple sites of involvement. Relationships of the lesion to nerve roots and to the dura mater were evident with MR imaging and lOS. Surgical localization from the MR images was accurate in all cases
with
above. quickly
a.
b.
Figure 5. cord caudal strates the mogeneously
Ependymoma. (a) MR image (600/20) demonstrates an abrupt enlargement of the to C-7 (*). Cord signal caudal to this point is isointense with CSF. (b) LOS demonsite of transition (*) from normal cord to solid mass lesion. The mass consists of hoincreased echogenicity, with no regions of cystic alteration and no syrinx forma-
hon.
with intermixed regions of low signal intensity. T2-weighted images demonstrated homogeneously increased signal intensity throughout the cord, and it was not possible to distinguish between nonenhanced tumor and syrinx formation rostral to the region of enhancement. 105 demonstrated syrinx formation above and below the echogenic mass, with the mass having small
regions
of central
cyst
forma-
tion. The central rostral curvilinear contour of the mass seen with MR imaging was easily correlated with sonographic findings. Regions of MR contrast
enhancement
echogenic.
The
was
II astrocytoma.
grade
were
pathologic
markedly
diagnosis
MR imaging techniques allow localization of an intramedullary spinal cord mass lesion to its specific compartment and segmental level within the spinal canal but are not specific for tissue type, do not distinguish between nonenhanced tumor and surrounding edema, and do not reliably distinguish between cystic and solid components (i-9). To supplement preoperative MR imaging during the
181
Number
#{149}
105
approach to spinal cord has been increasingly
lized,
both
2
leuti-
continuous charand abnormal cord anatomy through the opened or unopened dura mater (ui-i4) and to allow correlation of operative findings with preoperative images. 105 demonstrates the normal posterior dura mater as a thin echogenic membrane with the anterior dura mater obscured by adjacent echogenic bone and ligaments. The posterior surface of the vertebral bodies is more echogenic than are disk spaces, allowing identification of segmental intervals. An alteration in the dimension or anterior contour of the spinal canal acterization
to allow
of normal
is easily recognized. ments are echogenic, hesions, vessels, and
DISCUSSION
Volume
surgical sions,
The dentate as are dural nerve roots
versing the subarachnoid
subarachnoid space. space around the
is normally
anechoic,
and
with
ligaadtra-
The cord real-
time sonography, pulsation can be demonstrated in the anterior spinal artery. Although the surface of the spinal cord is echogenic, the normal cord parenchyma is relatively hypoechoic, except for a central linear echogenic complex that may represent the central canal or anterior median fissure.
the
With
techniques
lOS,
identified
the
described
mass
and
the
was surgeon
was assured that the extent of the laminectomy was appropriate before opening the dura mater. Sonography was again used after the dura mater was opened, if further image characterization of the mass was necessary before myelotomy or biopsy was performed. This approach minimized damage to normal cord parenchyma by limiting the myelotomy and allowing the greatest potential access to the central portion of the mass. In all cases, the similarity between the surface and internal architecture of the lesions, as shown at preoperative MR imaging and at lOS, allowed immediate direct correlation of the operative field with preoperative MR imaging. A site selected for biopsy on the basis of MR imaging characteristics
could
be quickly
In the patient inflammation miasis
(Fig
identified
at lOS.
with granulomatous secondary to schistoso1), focal
gadolinium
en-
hancement and hyperechogenicity were limited to the ventral aspect of the cord and corresponded exactly to the focal necrosis found at surgery. Four of the seven gliomas and none of the other four lesions demonstrated large intramedullary fluid cotlections (syringohydromyelia) rostral or caudal to the neoplasm. Identification of the interface between the tumor and syrinx was quickly accomplished with lOS, allowing the surgeon to establish a plane of dissection between the tumor and surrounding cord (Figs 4, 6). Although internal architecture and cord contour were well demonstrated with both modalities, neither MR imaging nor lOS provided specific solidtissue series,
characterization. tumors, regions
tion, gliosis, and edema echogenic than normal
In the current of inflamma-
were more cord paren-
Radiology
577
#{149}
a.
b.
C.
Figure 6. Neurenteric cyst. (a, b) MR images (600/20 and 3,000/100) demonstrate a well-marginated that distorts the contour of the spinal cord and displaces it posteriorly. Signal intensity parallels mater is opened demonstrates posterior displacement of the cord and intramedullary extension rostral
margin
of the
neurenteric
cyst
is well
defined
chyma, as previously reported (iii4). Intramedullary tumors were associated with expansion of cord contour, alteration of intramedullary anatomy, and loss of the central echo complex. In some cases, edema and gliosis could not be differentiated from tumor margins. After the resection of a spinal lesion, 105 allowed the surgeon to assess the completeness
of tumor
evacuation,
the
reex-
pansion of subarachnoid spaces, the effectiveness of cyst or syrinx drainage, or the position of an implanted shunt device. Completeness of resection was better assessed in cases with sonographically welt-marginated lesions
(Figs
i, 5, 6) than
for
glioma
sur-
rounded by biopsy-proved echogenic nontumorat gliosis and edema (Fig 3). Although it is important in operative planning for the neurosurgeon to know whether an intramedullary lesion has a substantial cystic component, solid lesions of the spinal cord and those that contain fluid cannot be routinely distinguished with MR imaging (iS,i6). Cysts generally are nonenhanced areas of low signal intensity on Ti-weighted images and high signat intensity on T2-weighted images. Low signal intensity on T2-weighted images
relating
to flow,
turbulence,
or
phase shift induced by fluid pulsation may occur within a large syrinx but is uncommon within a tumor cyst (i6). None of our patients with either tu578
Radiology
#{149}
(open
intramedullany and extramedullary of CSF. (c) LOS performed before
of the anechoic
cystic
lesion
(solid
the
mass dura
arrow).
A
arrow).
mor cyst or syrinx formation demonstrated low signal intensity on T2weighted images. On Ti-weighted images, cystic cavities can alternatively be isointense or hyperintense, relative to surrounding cord parenchyma if they contain fluid of high protein content or methemoglobin secondary to recent hemorrhage (17). Benign cystic lesions in our series had sharply defined margins, uniform signat intensity, and isointensity of cyst contents relative to CSF (i8). Regions of cord gliosis, edema, or demyelination can be identical in intensity at MR imaging to nonflowing fluid within a cyst or syrinx. Ultrasonography demonstrates cysts as anechoic structures with increased through transmission, whereas regions of edema and gliosis are hyperechoic relative to normal cord parenchyma. Only one patient has reportedly had a homogeneously echogenic intramedutlary cord lesion that was found to be cystic at 105; the lesion contained viscous squamous debris as a component of a teratoma (i9). Cystic
that
structures
rostral
or caudal
to
the mass generally expand the central canal, have smooth walls, cause symmetric expansion of the cord, and represent syrinx formation rather than tumor cyst or necrosis (ii). Tumor cysts are generally smaller, may have irregular walls, and are eccentric in
position series,
within the
the
walls
cord
were
(ii).
more
In our echogenic
than those of normal cord. Cystic alterations occurred both in ependymomas and astrocytomas in our series. Sonography without Doppler analysis may cysts
not distinguish and enlarged
between vascular
small channels
within a mass lesion. Although color flow Doppler imaging has correctly demonstrated arteriovenous malformations in the brain and spinal cord (i4), flow in the venous malformation encountered in the current series could not be demonstrated. Presumably,
blood
flow
was
slow
and
the
equipment available was insensitive to it. Components of the intramedullary mass
that
demonstrated
hancement consistently Not
contrast
en-
with MR imaging were echogenic at sonography.
all echogenic
regions,
however,
demonstrated contrast enhancement. All neoplasms were echogenic, as were regions of edema, gliosis, inflammation, hemorrhage, and venous vascular malformation. Regions of contrast enhancement within a cord mass lesion may be the most productive sites for tissue biopsy, and on the basis
of our
are echogenic raphy
without
experience,
these
at sonography.
regions
Sonog-
contrast-enhanced
MR
imaging for correlation may not distinguish between the tumor and other areas of echogenicity, as deNovember
1991
scribed above. Tissue echogenicity relates to the number of reflecting interfaces encountered by the sound wave. Because gadopentetate dimeglumine
does
not
enhance
all
tumor tissue, it is not always possible with either sonography or MR imaging to distinguish between regions of tumor, gliosis, vascular malformation, edema, and inflammation. Ependymomas have been reported at 105 to be central, symmetric, better marginated within the spinal cord, and more echogenic than astrocytomas (ii). In our limited series, this differentiation could not be confirmed, though it was characteristic of the ependymoma of patient 5. No correlation could be identified between characteristics noted at sonography and those at histologic examination. Homogeneous versus inhomogeneous and diffuse versus focal increased echogenicity patterns noted in the neoplasms of this series did not correlate with the type of tumor. A limitation to image-pathologic conelation in this series stemmed from the method of tissue resection, which incorporated
a piece-by-piece
dissection
of the tumor from the cord with an ultrasonic surgical aspirator (Cavitron Lasersonics, Stanford, Conn) under the operating microscope. Also, the classification of gliomas was not always distinct. Microscopically, ependymomas typically appear as a dense uniform “carpet” of cells, whereas astrocytomas have a looser fibrillary background matrix. The ependymoma of patient S had typical cells, but they were within a loose background matrix that was more typical of an astrocytoma. The astrocytoma of patient 4 had mixed components that resulted in its being classified as oligoastrocytoma, whereas patient 3 had similar oligodendroglial elements only to a slight degree. Tissue characterization as to tumor type or to the presence of a malignant versus a nonmalignant process was not possible with either MR imaging or with lOS. Although MR imaging does not specifically demonstrate calcification,
the presence of scalloped posterior margins of vertebral bodies resulting from chronic mass effect was readily apparent. Similar scalloping and enlargement of the spinal canal was noted at 105 (Figs 3, 6). MR imaging and lOS were found to be complementary techniques for the imaging of intramedullary cord lesions. Alterations in cord contour and patterns of structural intramedullary alteration seen with MR imaging could be quickly correlated with sonographic images, allowing rapid intraoperative correlation
181
Number
#{149}
2
localization findings
7.
Donovan Post MJ, Quencer RM, Green BA, et al. Intramedullary spinal cord metastases, mainly of nonneurogenic origin. AJR 1987; 148:1015-1022. Goy AMC, Pinto RS, Raghavendra BN, Epstein FJ, Kricheff II. Intramedullary spinal cord tumors: MR imaging, with emphasis on associated cysts. Radiology 1986; 161: 381-386.
8.
Sze G, Krol G, Zimmerman
9.
before
2.
3.
4.
Sze G, Stimac
GK, Bartlett
C, et al.
Multi-
center study of gadopentetate dimeglumine as an MR contrast agent: evaluation in patients with spinal tumors. AJNR 1990; 11:967-974. Dillon WP, Norman D, Newton TH, Bolla K, Mark A. Intradural spinal cord lesions: Gd-DTPA-enhanced MR imaging. Radiology 1989; 170:229-237.
Sze G, Bravo S. Krol G.
Spinal
lesions:
quantitative and qualitative temporal evolution of gadopentetate dimeglumine enhancement in MR imaging. Radiology 1989; 170:849-856. Panizel PM, Baleniaux D, Rodesch G, et al. Gd-DTPA-enhanced MR imaging of spinal tumors. AJNR 1989; 10:249-258. Di Chiro G, DoppmanJL, Dwyer AJ, et al. Tumors and anteriovenous malformations of the spinal cord: assessment using MR. Radiology 1985; 156:689-697.
investigation
and comparison
with
10.
11.
12.
13.
14.
15.
16.
References 1.
RD, Deck MDF.
Intramedullary disease of the spine: diagnosis using gadolinium-DTPA-enhanced MR imaging. AJR 1988; 151:1193-1204. Stimac GK, Porter BA, Olson DO, Gerlach R, Genton M. Gadolinium-DTPA-enhanced MR imaging of spinal neoplasms:
preliminary
and
opening the dura mater. lOS allowed the surgeon to define margins of an intramedullary lesion that was not visible at inspection of the cord surface. Combined lOS and preoperative MR imaging provided precise localization for determining the extent of the laminectomy and for defining the location and extent of myelotomy. It identified the depth of tumor resection relative to the location of the anterior surface of the spinal cord or to the conus, allowed assessment of completeness of tumor resection, and defined the presence or absence of syrinx or cyst adjacent to or within the mass. In patients with intramedullary spinal cord lesions, a more specific surgical approach, a more complete resection, and diminished longterm morbidity may be possible when preoperative contrast-enhanced MR imaging and 105 techniques are used in concert. U
5.
Volume
lesion with MR
6.
17.
18.
unenhanced spin-echo and STIR sequences. AJR 1988; 151:1185-1192. BrunbergJA, Latchaw RE, Kanal E, Bunk L Jr. Albnight L. Magnetic resonance imaging of spinal dysraphism. Radiol Clin North Am 1988; 26:181-205. Epstein FJ. The utilization of ultrasonography as a surgical adjunct in the radical excision of one hundred and eighty-six consecutive intramedullary spinal cord neoplasms. In: Marlin AE, ed. Concepts in pediatric neurosurgery. Vol 10. Basel, Switzerland: Karger, 1990; 54-66. Rubin JM, Dohrmann GJ. The spine and spinal cord during neurosurgical operations: real-time ultrasonography. Radiology 1985; 155:197-200. Quencer RM, Morse BMM, Green BA, Eismont FJ, Brost P. Intraoperative spinal sonography: adjunct to metrizamide CT in the assessment and surgical decompression of posttraumatic spinal cord cysts. AJR 1984; 142:593-601.
Rubin J.
in spinal
cord
Kjos BO, Brant-Zawadzki
M, Kucharczyk
W, Kelly WM, Norman D, Newton TH. Cystic intracranial lesions: magnetic resonance imaging. Radiology 1985; 155:363369. Williams AL, Haughton VM, Pojunas KW, Daniels DL, Kilgore DP. Differentiation of intramedullary
19.
Ultrasonography
surgery. In: Rubin JM, Chandler WF, eds. Ultrasound in neurosurgery. New York: Raven, 1990; 107-182. Rubin JM, Aisen AM, DiPietro MA. Ambiguities in MR imaging of tumoral cysts in the spinal cord. J Comput Assist Tomogr 1986; 10:395-398. Enzmann DR. O’Donohue J, Rubin JB, Shuer L, Cogen P. Silverberg G. CSF pulsations within nonneoplastic spinal cord cysts. AJR 1987; 149:149-157.
neoplasms
and
cysts
by
MR. AJR 1987; 149:159-164. PlattJF, Rubin JM, Bowerman RA, DiPietro MA, Chandler WF. Intraoperative sonographic characterization of a cystic intramedullary spinal cord lesion appearing as solid. AJNR 1988; 9:614.
Radiology
579
#{149}
