1045

Hippocampal Sclerosis in Children with Intractable Temporal Lobe Epilepsy:

Detection with MR Imaging

J. 0. Grattan-Smith1 A. S. Harvey2 P. M. Desmond3 C. W. Chow4

OBJECTIVE. Hippocampal sclerosis is the most common lesion associated with temporal lobe epilepsy. Temporal lobectomy Is effective for the control of medically

refractory

seizures

In these patients.

Before the advent of MR imaging,

hippocampal

sclerosis was rarely diagnosed preoperatively. The purpose of this study was to determine the frequency of hippocampal sclerosis In children with Intractable temporal lobe epilepsy and the accuracy and reliability with which hippocampal sclerosis can be diagnosed on the basis of MR findings in children. MATERIALS AND METHODS. We reviewed the MR images of 53 children (mean age, 10 years) with medically refractory temporal lobe epilepsy. The MR images were reviewed blindly and Independently by two radiologists on two occasions, and were classified as showing hippocampal sclerosis, other lesions, or no abnormality. Hippocampal sclerosis was diagnosed when evidence of hlppocampal atrophy was present or hippocampal signal intensity was abnormal without evidence of a mass lesion. RESULTS. MR images showed hippocampal sclerosis in 30 chIldren (57%), other lesions in 10 (19%) (tumors in eight, cavernous angloma in one, and ectopic gray matter in one), and no abnormality in 1 3 (24%) (lntraobserver agreement: kappa = 0.77 and 0.84, Interobserver agreement: kappa = 0.76). MR lateralizatlon was concordant with ictal EEG in 36 (92%) of 39 children. Hippocampal sclerosis was bilateral in one child and associated with extrahippocampal lesions In nine. Hippocampal sclerosis was detected on MR images of 11 (85%) of 13 children with pathologic confirmation of hippocampal sclerosis. Beneficial results were seen in 26 (90%) of 29 children who had temporal lobectomy. CONCLUSION. Hippocampal sclerosis is the most common lesion in children with intractable temporal lobe epilepsy, and it can be detected reliably and accurately on MR images. We suspect that hippocampal sclerosis is underdiagnosed in children, possIbly leading to postponement of surgery In children with refractory seizures.
AJR

1993;161

:1045-1048

Received May 6, 1993; June 28, 1993.

1

accepted

after

revision

Department
Fiemington

of Radiology,
Rd., Parkville,

Royal

Childrens
3052,

Hospital,

victoria,

Australia. Address correspondence tan-Smith.

2Depantment of Paediatncs, boume, Melboume, Australia.

to J. D. Gnatof Mel-

University

3Department
Hospital, Childrens tralia. Melboume, Hospital,

of Radiology,
Australia.

Royal

Melboume Royal
Aus-

4Department

of Anatomical
Parkville,

Pathology,
victoria, 3052,

0361-803X/93/1615-1045 ci American Roentgen

Ray Society

Temporal lobe epilepsy is the most common form of intractable epilepsy in children [1]. Temporal lobectomy is effective in controlling seizures in approximately 80% of children with intractable temporal lobe epilepsy [2-5]. Although hippocampal sclerosis is seen in 60-70% of autopsy and surgical series in adults with refractory temporal lobe epilepsy, it has been variably reported in studies of temporal lobectomy in children [2-7]. Recent reports [8-14] indicate that hippocampal sclerosis can be reliably detected with MR imaging in 60-100% of patients; however, a number of issues remain unresolved. There is no universal agreement on the ideal method of MA imaging of the hippocampus or criteria for diagnosing hippocampal sclerosis [1 3, 14]. The application to clinical practice of quantitative MR techniques is unclean [11-14]. We reviewed the MR images of children with well-documented intractable temponal lobe epilepsy to determine the frequency of hippocampal sclerosis and the accuracy and reliability with which hippocampal sclerosis could be diagnosed on the basis of MR findings.

Materials
Fifty-three

and Methods
consecutive patients of the Royal Childrens Hospital,

Ethical

considerations

control subjects, use of children

precluded the use of healthy children as and practical problems with blinding precluded the with extratemponal lesions as control subjects.

Melbourne,

with

medically

refractory

seizures

and clinical

and ictal

EEG evidence of temporal lobe epilepsy underwent MR imaging between 1987 and 1992. The 27 boys and 26 girls were 2-1 7 years
old (mean, 10 years) atthe time of imaging. The children were identified from a review of ictal EEG recordings and clinical records of all children with partial seizures who underwent video-EEG monitoring in the department of neurology. Children were included only if sei-

Chance-adjusted intenobserver and intraobserver measured with Cohens kappa statistic.

agreement

was

Results The consensus MR diagnoses for the 53 children were hippocampal sclerosis in 30 (57%), other lesions in 10 (19%), and no abnormality in 13 (24%). Intraobserver and intenobserven agreement was good (intraobserven kappa, 0.77 and 0.84; interobserver kappa, 0.76). Intenobserver disagreement concerned eight children in whom the ultimate diagnosis was hippocampal sclerosis or no abnormality. Six of the eight had MR studies with the 0.3-T unit. Unilateral abnormalities of the temporal cortex were diagnosed on MR images in 39 children, concordant with the EEG and ictal SPECT lateralization in 36 children (92%). Hippocampal sclerosis was unilateral in 29 children and bilateral in one (Figs. 1-3). Twenty-eight children had hippocampal atrophy and increased signal on T2-weighted images; two children had marked hippocampal atrophy without observable signal abnormality. Bilateral hippocampal sclerosis was diagnosed on the basis of bilateral hippocampal atrophy and increased signal intensity (Fig. 3). In nine children, hippocampal sclerosis was associated with lesions outside the temporal cortex: anachnoid cyst in the middle cranial fossa in two (one ipsilateral and one contralatenal to hippocampal sclerosis), peniventniculan high signal intensity in two who had ventniculopenitoneal shunts, ipsilatenal focal occipital atrophy in three, ipsilateral paraventnicular tumor
projecting into the tnigone in one, and generalized cerebral

zune onset was clearly seizure characteristics

localized to the temporal region and clinical were consistent with temporal lobe onset.

Children were excluded if EEG abnormalities were poorly localized or extratemponal or if clinical seizure characteristics were inconsistent with temporal lobe onset. Ictal single-photon emission computed tomography (SPECT) with 99mTchexamethylpropyleneamine

oxime was performed during video-EEG recording in 24 children. At the end of the study period, anterior temporal lobectomy had been performed in 28 children and a lesion had been resected in
one. Seven of these children have been reported previously [2].

Mean age at surgery was 11 years (range, 5-1 8 years), and the mean follow-up period was 22 months (range, 1 month to 4 years). Results of histologic examinations in 29 children were reviewed by a
pediatric neuropathologist who did not know the patients identity or

MR findings.

After surgery,

21 children

were seizure

free on having

auras only, three rarely had seizures, and two had fewer seizures than before. Seizure frequency was unchanged in one child, and one child with Rasmussens syndrome died. MR images were obtained with a 0.3-T iron core resistive magnet (B3000, Fonan, Melville, NY) in 28 children, with a 1 .5-T superconducting magnet (Magnetom, Siemens AG, Enlangen, Germany) in 18 children, and with both units in seven. Fifteen children had multipie studies. With the 0.3-T unit, generally coronal inversion-recovery

(1500/30/500 [TR/Ti/TE]) and axial and coronal spin-echo (2000/ 35,85 [TRITE]) images were obtained. A 5.1-mm slice thickness, 0.5-mm interslice gap, 192 x 256 matrix, and 24.5-cm field of view were used. With the 1 .5-T unit, usually magnetization-prepared
rapid gradient-echo images (MP RAGE,
x

1 0/4/1

0#{176} [TR/TE/flip

angle])
ar field

were obtained
of view, 2.8-mm

by using
slice

a 180
thickness,

256 matrix,
and

25-cm

nectangugap. Axial

no intenslice

and coronal spin-echo images (2750/20,80) were obtained by using a 75#{176} angie, one excitation, flip 4-mm slice thickness, 1-mm interslice gap, 1 92 x 256 matrix, and 23-cm field of view. Coronal images were pulse gated. With both units, axial images were parallel and

atrophy in one. MR diagnoses of the other lesions in the temporal cortex were tumor in eight, cavernous angioma in one, and hemimacrencephaly with ipsilatenal ectopic gray matter in one.

Of the 17 children

who had surgery

and in whom

hippo-

coronal

images

were

perpendicular

to the long axis

of the hippoc-

ampus, defined on a sagittal localizer. Because of rapid advances in MR technology

of the study, images were obtained with two

during the course

machines, and

different

consequently the quality protocols were modified

of the images varied. In addition, imaging as software capabilities changed. Tech-

niques for quantitative analysis were not available at the time of the study. Young children were sedated when necessary, under supenvision of a pediatric anesthetist. MR images were reviewed independently by a pediatric nadiolo-

gist and a neuroradiologist. Each MR study was reviewed twice by each radiologist. Images were reviewed in random order and without knowledge of the patients identity, clinical and EEG data, the nadiologistsfirst interpretation,or the other radiologistsinterpretation.MR images of children who had multiple examinations were reviewed

together.

All sequences

and planes

of imaging

were reviewed,

but

particular attention was directed to the medial temporal structures. Images were classified as showing hippocampal sclerosis, other lesions, or no abnormality. Abnormalities were classified as left sided, right sided, on bilateral. Hippocampal sclerosis was diagnosed when evidence of hippocampal atrophy was present or hippocampal signal intensity was abnormal without evidence of a mass lesion. When different diagnoses were made by one radiologist, the MR

campal sclerosis was diagnosed on the basis of MR findings, histologic examination showed hippocampal sclerosis in 11 and no abnormality in six in whom appropriate hippocampal tissue was not present in the surgical specimen. Of the six children who had surgery and in whom a tumor was diagnosed on the basis of MR findings, histologic examination of the available specimen showed dysembryoplastic neuroepithelial tumor in two, ganglioglioma in one, astrocytoma in one, chronic encephalitis in one, and no abnormality in one. The cavernous angioma and gray matter hetenotopia were histologically confirmed. Of the four children who had surgery and in whom no abnormality was diagnosed on the basis of MR findings (all four imaged with the low-field-strength unit), histologic examination showed hippocampal sclerosis in two and no abnormality in two in whom appropriate hippocampal tissue was not present in the surgical specimen. MR images showed evidence of hippocampal sclerosis in 11 (85%) of 13 children with histologically proved hippocampal sclerosis.

Discussion This study reports the MR findings in a group of children with well-documented and intractable temporal lobe epilepsy. The diagnosis of temporal lobe epilepsy was based on typical clinical seizure characteristics and ictal EEG localization to the temporal lobe. Although intracranial electrodes

images were reviewed once more and a final diagnosis When different diagnoses were made by two radiologists, were reviewed in conjunction and a consensus diagnosis

was made. the images was made.

AJR:161,

November

1993

HIPPOCAMPAL

SCLEROSIS

IN CHILDREN

1047

Fig. 1.-12-year-old boy with right-sided hippocampal sclerosis. A, Angled coronal MP RAGE (10I4l10) MR Image shows marked left bounded superiorly by alveus. B and C, Coronal proton density-weighted (B) and T2-welghted localized to hippocampus on right side.

hippocampal (C) (2500/30,80)

atrophy

(arrow)

on right show

side.

Compare of Increased

with

normal

hippocampus (arrows)

on is

MR Images

the area

signal

Intensity

Fig. 2.-9-year-old girl with left-sided hippocampal sclerosis. A and B, Angled axial (A) and coronal (B) T2weighted (2500/80) MR Images show poorly defined area of increased signal Intensity (arrows) extending along hlppocampus. Hippocampal atrophy was better visualized on MP RAGE coronal sequence (not shown).

Fig. 3.-7-year-old girl with bilateral hippocampal sclerosis. A, Angled coronal MP RAGE (10I4I10) MR Image shows bilateral hippocampal atrophy (arrows), more marked on left side than on right B and C, Proton density-weighted (B) and T2-weighted (C) (2500/30,80) MR images show areas of Increased signal Intensity (arrows) hlppocampus bilaterally. Independent bilateral temporal lobe hyperperfusion was seen on Ictal SPECT Images (not shown).

side.

localized

to

were not used, EEG localization

with scalp electrodes

was

supported by the results of ictal SPECT in 21 children and the beneficial results of temporal lobectomy in 26 children. Abnormalities on MR images were detected in 75% of children, hippocampal sclerosis being the most common lesion.

EEG

confirmation

of latenalization

was

present

in 92%.

Overall agreement between the two observers was good, indicating that hippocampal sclerosis and other lesions can be detected reliably on MR images of children. Hippocampal sclerosis was diagnosed on the basis of MR findings in 57% of children in our study and was confirmed

ity of these images would not be acceptable now. Hippocampal sclerosis was found at pathologic examination in two children, and no abnormality was found in two children in whom appropriate hippocampal tissue was not available for examination. Some children with normal MR findings had minor degrees of hippocampal asymmetry, but in the absence of an appreciable increase in hippocampal signal on T2-weighted images, hippocampal sclerosis was not diagnosed. Quantitative techniques may have led to a diag-

nosis of hippocampal sclerosis in some of these children. It is also possible that some of the children with normal MR
findings who have tional abnormalities not undergone surgery have subtle migraon microdysgenesis of the temporal con-

pathologically

in all children

who

had

surgery

in whom

appropriate hippocampal tissue was available tion. The method of mesial temporal resection

for examinaat our institu-

tex, which is not detectable

with MR.

tion (subpial aspiration meant that adequate

examination was available

as opposed hippocampal
in only

to en bloc resection) tissue for pathologic

who had

11 of 17 children

MR evidence

of hippocampal

sclerosis.

Hippocampal

sclero-

sis is detected on MR images on at surgery in 60-70% of adults with intractable temporal lobe epilepsy [10, 13]. The prevalence of hippocampal sclerosis in children with intractable temporal lobe epilepsy can be determined only from surgical series, where it has been variably reported in 960% [2-6]. The studies showing fewer children with hippocampal sclerosis may reflect the absence of hippocampal

All diagnoses of hippocampal sclerosis based on MR images were confirmed pathologically if appropriate tissue was available for examination. Our results suggest that hippocampal sclerosis is as prevalent in children, and as readily diagnosed on the basis of MR findings, as in adults. We sus-

pect that hippocampal

dren, possibly leading children with intractable ACKNOWLEDGMENTS

sclerosis

is underdiagnosed
of

in chilsurgery in

to postponement seizures.

tissue

in their surgical

specimens,

a selection

bias toward

early surgery when imaging shows a mass lesion, on a different spectrum of temporal lobe abnormalities in children who have intractable temporal lobe epilepsy in early childhood.

We thank Peter Buchanan for administrative assistance; Eric Gilford, Michael Kean, and the MRI staff at St. Vincents Hospital, Melbounne, for imaging our patients; Ian J. Hopkins, Lloyd K. Shield, Kevin J. Collins, and Geoffery L. KIug for contributions to patients
treatment; and Brian M. Tress, Ian J. Hopkins, and Sam F. Berkovic for reviewing the manuscript and providing helpful suggestions.

The
include

reported
hippocampal

MR

features
atrophy,

of

hippocampal
signal

sclerosis
intensity on

increased

T2-weighted

images,

loss of internal

architecture

of the hip-

pocampus, and loss of signal intensity on Ti -weighted images [13, 14]. In our study, only the first two were used as
criteria for a diagnosis of hippocampal sclerosis. Interest-

REFERENCES
1 . Aicardi J. Epilepsy in children. New York: Raven, 1986:140-175 2. Hopkins IJ, KIug GL. Temporal lobectomy for the treatment of intractable complex partial seizures of temporal lobe origin in early childhood. 0ev Med Child Neurol 199133:26-31 and extended in eariy childin early childman-

ingly,

increased

signal

intensity

on T2-weighted

images

was

seen in 28 of 30 children

with hippocampal

sclerosis;

the first

3. Adelson

echo of the double-echo coronal sequence was particularly useful in marginal cases. This is in contrast to some studies in adults, in whom demonstrable signal abnormalities were

WJ, Chugani HT, et al. Temporal temporal resections for the treatment of intractable seizures hood. Pediatr Neurosurg 1992;18:169-178 M, Levin B, Jayakan P, et al. Temporal lobectomy
J, Hwang P, Becker Epilepsia 1992;33:298-303 J, Hoffman HJ, Kobayashi 1987:21:792-797

PD, Peacock

4. Duchowny
hood. 5. Drake

rare [1 5]. A recent study with T2 quantitative measurement has shown increased signal intensity in patients with intractable temporal lobe epilepsy when no signal abnormality was visible (Jackson GD et al., presented at the American Epilepsy Society meeting, December 1992). It is possible
that signal abnormalities associated with hippocampal scle-

LE. Surgical

agement of children with temporal

rosurger, 6. Davidson

lobe epilepsy and mass lesions. Neu-

S, Falconen MA. Outcome of surgery in 40 children with temporal lobe epilepsy. Lancet 1975;1 :1260-1 263 7. Babb TL, Brown WJ. Pathologic findings in epilepsy. in: Engel JJ, ed. Surgical treatmentof AJNR the epilepsies. 1991:12:933-940

rosis are more pronounced

in children

or that loss of signal

8. Bronen
pathology.

RA. Imaging findings

New York: Raven, 1987:511-540 in hippocampal sclerosis: correlation

with

intensity occurs with increasing age. In our study, nine children with hippocampal sclerosis had additional lesions outside the cortex of the temporal lobe, raising doubt about localization of the epileptogenic focus. Arachnoid cysts and peniventnicular white matter gliosis are unlikely to cause seizures by themselves. However, cortical atrophy or tumor in the posterior cortex can manifest electnically and clinically as temporal lobe epilepsy. The children

9. Jackson GD, Berkovic SF, Tress BM, Kalnins RM, Fabinyi GC, Bladin PF.
Hippocampal sclerosis can be reliably detected by magnetic resonance imaging. Neurology 1990;40:1869-1 875 10. Berkovic SF, Anderman F, Olivier A, et al. Hippocampal sclerosis in tempenal lobe epilepsy demonstrated by magnetic resonance imaging. Ann Neurol 1991 :29:17-1 82 11. Jack CRJ, Sharbrough FW, Twomey CK, et aI. Temporal lobe seizures: lateralization with MR volume measurements of the hippocampal formation. Radiology 1990;175:423-429

with discordant MR and EEG abnormalities and the children with bilateral seizure onset have not undergone surgery. No lesion was detectable on MR images of 13 children,
seven
unit.

of whom
In four

were
children

examined
with

with the higher-field-strength

normal

MR findings,

temporal

12. Ashtari M, Barr WB, Schaul N, Bogerts B. Three-dimensional low-angle shot imaging and computerized volume measurement of the hippocampus in patients with chronic epilepsy of the temporal lobe. AJNR 1991: 12:941-947 13. Bronen RA. Epilepsy: the role of MR imaging. AJR 1992;159:1165-1174 14. Jackson GD, Berkovic SF, Duncan JS, Conneily A. Optimizing the diagnosis of hippocampal sclerosis using MR imaging. AJNR 1993:14:753-762

lobectomy was performed because of results of electrophysiologic studies. These four children were examined on the low-field-strength magnet early in the study period. The qual-

15. Cascino GD, Jack CR, Parisi JE, et al. Magnetic resonance imagingbased volume studies in temporal lobe epilepsy: pathological correlations.
Ann Neurol 1991:30:31-36