Hippocampal Sclerosis in Temporal Lobe Epilepsy: Findings at 7 T 1

NEURORADIOLOGY: Hippocampal Sclerosis in Temporal Lobe EpilepsyHenry et al26 years. The mean age of the remainingeight patients with TLE was 28 years.The clinical characteristics of the patientsare summarized in Table 1 .Qualitative Visual AssessmentIn control subjects, the internal structurewas visible in all sections of the head,body, and tail of the hippocampus ( Fig 3 ).In each patient ( Table 2 ), the abnormalhippocampus had at least one ofthe following signs: atrophy ( n = 7), T2-weighted hyperintense signal ( n = 6),and flattening of the head ( n = 6). Ineach subject, hippocampal striation andadditional detailed anatomic featureswere clearly apparent. In each controlsubject, it was possible to delineate continuoushippocampal striation. This whitematter band appeared to be less distinguishedfrom surrounding gray matterin all sclerotic hippocampi. Thus, the partialloss of hippocampal striation wasobserved in the hippocampus on the sideof ictal onset in all but two patients(patients 1 and 7) and in none of thecontrol subjects. Definite hippocampalmalrotations were observed in three patients;they were seen twice in patientswith ipsilateral onset and once in a patientwith contralateral to ictal onset( Fig 4 ). Malrotations were observed infour control subjects: They were observedbilaterally in subject 1 and on theleft side in subjects 6, 10, and 11. Thedetermination of malrotation showed novariation between the first and secondobservations in 18 of 19 subjects; however,one healthy control subject (subject4) was rated as having bilateralhippocampal malrotation at the secondviewing but not at the first. Ratingsof hippocampal size, signal, shape, andstriation did not vary between the firstand second viewings.Each patient had either no or onehippocampal digitation on the side ofictal onset ( Table 2 ). Three patients alsohad either no or one contralateral hippocampaldigitation. Several patientshad definite hippocampal atrophy. Interestingly,one patient (patient 1) hadbilateral nondigitated hippocampalheads but only mild hippocampal atrophyon the epileptogenic side and contralateralnormal hippocampal volumeFigure 2Figure 2: Manual segmentation of two subregions of body of hippocampus on T2-weighted 7-T MRimages in subject 2 (24-year-old man). (a, b) Most posterior coronal section in (a) head and (b) bodyof hippocampus, as the image on the left of these pairs. Red cross marks the same point on the pairedcoronal and sagittal images, and on sagittal section (as the image on the right of these pairs) indicates levelof coronal section (front is on left side of sagittal image). (c) Limits of CA1–CA3 (dark blue) and CA4 anddentate gyrus (light blue) regions on coronal section in middle of body (with no colored subregions on leftimage and colored subregions on middle image, both showing the same image plane), as indicated on thesagittal section (right image). (d) Three-dimensional rendering of two subregions in body of hippocampus,displayed separately (left, with CA1–CA3 in dark blue, and middle, with CA4 and dentate gyrus in light blue)and together (right).( Fig 4, A ). All healthy subjects had twoor three bilateral digitations of the hippocampalhead; however, one of these22 hippocampal heads had one digitation,consistent with reports of postmortemstud ies in cadavers withoutRadiology: Volume 261: Number 1—October 2011 n radiology.rsna.org 203

NEURORADIOLOGY: Hippocampal Sclerosis in Temporal Lobe EpilepsyHenry et alFigure 3Figure 4Figure 3: Coronal T2-weighted 7-T MR images through the, A , B , head, C ,body, and, D , tail of hippocampus in four control subjects (31-year-old man,24-year-old woman, 19-year-old woman, and 22-year-old man, respectively)show internal structure of hippocampus. Hippocampal striation (arrows) wasvisible in all.Figure 4: Coronal T2-weighted 7-T MR images through, A , head and, B – D ,body of hippocampus in three patients. A , Patient 1 (30-year-old woman).Absence of hippocampal digitations bilaterally and slightly increased signalintensity in head of right hippocampus. Hippocampal striation was barely visiblein head of right hippocampus (arrows). B , Body of left hippocampus in patient1 (arrow) had an unusual malrotated vertical shape. C , Patient 3 (22-year-oldman). Hippocampal atrophy and increased signal intensity predominating inCA4 and dentate gyrus region in body of left hippocampus (arrow). D , Patient 8(29-year-old woman) had hippocampal atrophy and increased signal intensitypredominating in CA1–3 region in body of the left hippocampus (arrow). Thispatient also had globular hippocampal shape.brain disease ( 28 ). Within and betweenthe two observers who each independentlycounted hippocampal digitationstwice, this method was used to generate25 instances of complete agreementon all four readings among 38hippocampi in healthy subjects andpatients with epilepsy. There were fourinstances of intraobserver disagreementand nine instances of interobserverdisagreement (complete data are givenin Appendix E1 [online]). When disagreementsoccurred, the number of digitationsreported never varied by morethan one (usually one digitation vs twodigitations or two vs three digitations).In determining paucity versus multiplicityof digitations in the 38 hippocampalheads, k values were 0.93 for intrarateragreement in both raters and 0.80 forinterrater agreement.Quantitative EvaluationSubregional segmentation of the hippocampalbody on T2-weighted images provedfeasible in all control subjects and204 radiology.rsna.org n Radiology: Volume 261: Number 1—October 2011

NEURORADIOLOGY: Hippocampal Sclerosis in Temporal Lobe EpilepsyHenry et alTable 1Clinical Characteristics of PatientsPatient No./Age (y)/SexEpilepsy Predisposing FactorsAge at SeizureOnset (y)ElectroencephalographicIctal OnsetClinical MR ImagingFindingSeizure Type1/30/FFebrile convulsions of infancy, family 28 R temporal R MTS A, C, Ghistory of epilepsy2/25/M Gestational or perinatal injury 13 R temporal R MTS A, C, G, SE3/22/M Febrile convulsions of infancy 5 L temporal L MTS A, C, G4/20/F None 14 L temporal L MTS A, C5/32/M Central nervous system infection 28 R temporal R MTS A, C6/49/F None 11 L temporal L MTS A, C, G7/20/F None 17 L temporal L MTS A, C, G8/29/FGestational or perinatal injury andfamily history of epilepsy23 L temporal L MTS A, CNote.—A = aura, C = complex partial, G = generalized tonic-clonic, L = left, MTS = mesial temporal sclerosis, R = right, SE = generalized convulsive status epilepticus.Table 2Visual Assessment of Hippocampi in Patients with TLEHeadBodyPatient No. and SideIctal Onset SideShape Signal Intensity Hippocampal Striation No. of Digitations Size Signal Hippocampal StriationMR1Right Y Flat 2 Flat, partial 0 1 0 N ...Left N ... ... Less fl at, partial 0 ... ... ... Y2Right Y Flat 2 Flat, partial 0 3 3 Blurred ...Left N ... ... Less fl at, partial 1 ... ... Blurred but less ...3Right N ... ... ... 1 ... ... ... ...Left Y Flat 2 Less visible, fl at, partial, 0 3 3 Visible, thinner ...blurred4Right N ... ... Reduced visibility 3 ... ... ... ...Left Y Normal 2 Reduced visibility 1 0 1 Less visible ...5Right Y Flat 2 Not visible 0 3 2 Less visible ...Left N ... ... ... 2 ... ... ... ...6Right N ... ... ... 3 ... ... ... ...Left Y Flat 1 Thin, blurred 0 3 1 Blurred Y7Right N ... ... ... 2 ... ... ... ...Left Y Partially 0 ... 1 0 1 Normal ...fl a t8Right N ... ... Visible, interrupted 2 ... ... ... ...Left Y Flat 2 Not visible 0 3 3 Visible, thinner YNote.—Atrophy and T2-weighted signal intensity ratings: 0 = normal, 1 = possibly present (mild), 2 = probably present (moderate), 3 = defi nitely present (severe). MR = hippocampal malrotation,N = no, NA = not applicable, Y = yes.Radiology: Volume 261: Number 1—October 2011 n radiology.rsna.org 205

NEURORADIOLOGY: Hippocampal Sclerosis in Temporal Lobe EpilepsyHenry et alpatients. Examples of resulting structuresin control subjects and patients with malrotationsand those without are shown( Fig 5 ). The mean subregional hippocampalvolumes in control subjects and patientsare displayed in Table 3. As expected,the average volume of the CA1–CA3 regionappeared smaller for sclerotic hippocampi.Surprisingly, CA4 and thedentate gyrus were slightly larger on thecontralateral side in patients. We observedtwo hippocampal subregions, oneconsisting predominantly of the lateralAmmon horn and the other consistingpredominantly of dentate gyrus with CA4,that were slightly smaller on the left sideof healthy subjects than on the right.This permitted us to compare the ratioof lateral Ammon horn (CA1–CA3) todentate CA4 in healthy subjects withthat in patients with TLE.Patient subregional measurementswith respect to the control population areshown in Figure 6 . The values in healthycontrol subjects were, on average, slightlysmaller on the left side than on the rightside for both CA1–CA3 volume and CA4and dentate gyrus volume. Mean AIfor CA1–CA3 of the entire hippocampalbody was 0.12 (range, 2 0.04 to 0.47);mean AI for volume per section of CA1–CA3 was 0.06 (range, 2 0.20 to 0.34);mean AI for CA4 and dentate gyrus ofthe entire body was 0.06 (range, 2 0.06to 0.22); mean AI for volume per sectionof CA4 and dentate gyrus was 0.00(range, 2 0.24 to 0.18); and mean ratioof AIs for CA1–CA3 to CA4 and dentategyrus of the entire body was 0.06(range, 2 0.08 to 0.34). The volume ofthe CA1–CA3 subregion was less thanthe mean volume in control subjects onboth sides in all patients but patient 1.The average volume per section wasmore sensitive than the entire volume.Interestingly, the ratio of CA1–CA3 toCA4 and the dentate gyrus was less thanthe mean ratio in control subjects inall patients on both sides of the hippocampus.Furthermore, in six of eightpatients, the ratio of CA1–CA3 to CA4and dentate gyrus on the side of ictalonset was less than the range in thecontrol group for that side ( Fig 6e ). Volumetricasymmetry of the hippocampalCA1–CA3 subregions was observed inTable 3Volume of Hippocampal Body Subregions on T2-weighted MR ImagesVolume (mm 3 )five patients with TLE and HS. Patient 3had substantial volume reduction comparedwith values in control subjectsfor both CA1–CA3 and CA4 and dentategyrus and a lower asymmetry indexfor the ratio than for both volumesControl GroupPatient GroupRight Left Ictal Onset Side Contralateral SideCA1–CA3Whole body 238 6 88 207 6 50 130 6 61 194 6 33Volume per section 21 6 4 20 6 3 11 6 4 19 6 1CA4 and dentate gyrusWhole body 172 6 46 162 6 45 166 6 47 180 6 17Volume per section 16 6 3 15 6 3 15 6 4 18 6 2Note.—Data are means 6 standard deviations.Figure 5Figure 5: Manually segmented volumes of the two subregions of body of hippocampus (CA1–3[dark blue] and CA4 and dentate gyrus [light blue]) in four control subjects ( h2 , h3 , h1 , and h10 )and four patients with epilepsy ( t3 , t5 , t6 , and t8 ) in the slightly oblique coronal view, with the leftside of the brain on right side of the image. Images in subjects 1 and 10 show unilateral malrotationof the hippocampus. Image in patient 3 shows left HS, image in patient 5 shows right HS, andimages in patients 6 and 8 show left HS with malrotation of hippocampus.separately. The distribution of atrophy infive patients was consistent with Ammonhorn sclerosis, while the distribution ofatrophy in patient 3 was consistent withcombined end-folium and Ammon hornsclerosis. The sclerosis was not detected206 radiology.rsna.org n Radiology: Volume 261: Number 1—October 2011

NEURORADIOLOGY: Hippocampal Sclerosis in Temporal Lobe EpilepsyHenry et alFigure 6Figure 6: (a–e) Graphs of hippocampal subregional volumetricdata. DG = dentate gyrus. Paired graphs for each measurementshow data by side in the left graph, and AI data in the right graph.Healthy subjects’ values are pooled in a single column to the left ofeach graph, with each patient’s data displayed separately. Horizontallines facilitate comparison of values in patients with TLEwith the range of values in the control group. For data by side,the lowest value in the control group is indicated by a color-codeddotted line for side, except in c and d , in which one line appearsbecause right and left sides had the same lowest values in thecontrol group; for AI data, a pair of solid horizontal lines indicateshighest and lowest values in control group.in patients 1 or 7 with subregional volumetryof the hippocampal body. Reproducibilityresults are given in AppendixE2 (online). Briefly, results were betterfor measurements based on the volumeper section than for measurementsbased on the entire body (relative indexvalues averaged for both right- and leftsidedCA1–CA3 and both right- andleft-sided CA4 and dentate gyrus in fivesubjects: 12% for entire body, 5% forvolume per section). This was expectedbecause of the limited number of sections(about 10), which makes an errorin even one section become an errorof about 10% for the complete volume.Volumes for the second segmentationwere larger than those for the first segmentation,with two exceptions (out of40). This can be explained by the factthat most borders are made of partialvolumes between two or more structuresand the fact that the threshold is subjective.Finally, errors were larger forCA4 and the dentate gyrus than for theCA1–CA3 region (relative index valuesfor the volume per section averaged forthe five subjects for both right- and leftsidedCA1–3, 3%; relative index valuesfor the volume per section averagedfor the five subjects for both right- andleft-sided CA4 and dentate gyrus, 6%).Overall, variability between these twosets of segmentations was lower than thatbetween the healthy control subjects foreach type of measurement.In summary, high-spatial-resolutionhigh-contrast 7-T MR imaging revealedthree findings that clinical MR imaging inpatients with mesial TLE did not. First,most of the patients in whom clinicalRadiology: Volume 261: Number 1—October 2011 n radiology.rsna.org 207

NEURORADIOLOGY: Hippocampal Sclerosis in Temporal Lobe EpilepsyHenry et alMR imaging revealed only atrophy overthe entire hippocampal formation hadatrophy specific to the main portion ofthe Ammon horn, with little or no atrophyin the dentate gyrus. Second, absenceor paucity of hippocampal headdigitations was observed on the epileptogenicside in all of the patients withTLE and in only one of the healthy controlsubjects on only one side. Third, theinternal structure of malrotated hippocampiwas clearly visible, unlike inprevious 1.5-T studies ( 17 ).DiscussionOur findings suggest that absence orpaucity of digitations of the hippocampalhead may represent a specific hippocampaldeformity in patients with mesialTLE. This deformity occurred most oftenin atrophic hippocampi on the sideof electroencephalographic ictal onset,but it also occurred on 7-T MR imagesin nonatrophic hippocampi contralateralto ictal onset. Patients with moderate orsevere hippocampal atrophy have beenreported to have loss of these digitationsat clinical MR imaging ( 9 ). Theselower-spatial-resolution and lowercontrast-resolutionstudies may enableus to detect reduced digitations on thebasis of atrophy-related enlargement ofthe adjacent ventricle so that the superiorhippocampal surface has high contrastagainst a widened cerebrospinal fluidspace. At 7 T, we consistently detectedthe thin white matter layer of the alveus,situated between the gray matter massesof the hippocampus and amygdala, to accuratelydefine the hippocampal surface,including its gyral pattern, even in the absenceof moderate or severe hippocampalatrophy with enlarged cerebrospinal fluidspaces. Normal gyral morphology of thehippocampal head may be absent or reducedin many patients with mesial TLEand to some extent may be independentof hippocampal atrophy. Tissue examinationwould be necessary to determineif this hippocampal deformity representsa malformation of hippocampal developmentwith microstructural dysplasia.Others have questioned whether subtledysplasias may be associated withHS. Isolated hippocampal tectonic malformationsare detected microscopicallywithin small portions of sclerotic hippocampiin patients with surgically treatedTLE ( 29 ). Isolated hippocampal dysplasiashave rarely been reported in studies ofmesial TLE with tissue correlation to clinicalMR imaging ( 30,31 ). In syndromesother than TLE, clinical MR imaging candepict hippocampal hypertrophy andother large hippocampal malformationsin association with extrahippocampalmalformations of cortical development( 17,32–34 ). An isolated hippocampalmalformation, malrotation of the hippocampalbody, has been reported atMR imaging not only in patients withmesial TLE but also in healthy subjectswithout epilepsy ( 18,19 ). Here, 7-T MRimages clearly showed major internalstructures of the hippocampus and enableddetailed evaluation of normal andabnormal hippocampal folding and rotation.Ultrahigh-field-strength MR imagingmay prove to be the preferredtechnique with which to study largeaffected and control groups and definebenign and abnormal hippocampalmalformations in relationship to specificneurologic syndromes.The limitations of the current studyare largely related to the relatively smallnumber of subjects imaged and to theincomplete development of 7-T brainimaging techniques. Our method forcounting digitations of the hippocampalhead was associated with concordantinterpretations across independent observers,but it has not been comparedwith an independent measure of theseconvolutions. Our intrahippocampal volumetrictechnique has not yet been validatedfor reproducibility and relationshipswith tissue findings. Larger groups ofhealthy control subjects, patients withmesial TLE, patients with epilepsy butnot TLE, and patients with nonepilepticbrain disease must be studied to determinethe specificity of our findings formesial TLE. Observations regarding apparenthippocampal deformity on 7-Timages in patients with mesial TLE shouldbe confirmed with histopathologic examinationof tissue specimens at subsequentepilepsy surgery. Cadaveric imagingshould be performed to confirmrelationships between submillimetric hippocampalsurface and intrahippocampalfeatures of 7-T images and histologicresults. Avoiding head motion enoughto benefit from the higher spatial resolutionwas a challenge in several subjectsin this study. New strategies willbe needed to permit routine acquisitionof images with submillimetric spatialresolution, particularly in patients withbrain disease. We did not have availablesoftware designed for automated hippocampalsegmentation and surface morphometryat 7-T-generated spatial andcontrast resolution, such as that whichis available for 1.5- and 3-T studies ( 35 ).The greater complexity of intrahippocampalsubstructures in the head and tailof the hippocampus limited our currentsubregional segmentation approach tothe hippocampal body. If 7-T MR imagingis developed to depict subtle hippocampalalterations reliably and thesealterations are shown to accurately defineanatomic abnormalities, considerableadditional effort will be required toclarify their relationships with epileptogenesisand to establish any useful rolein presurgical evaluation of epilepsy. Ourin vivo findings strongly support future effortsin the study of larger groups withimproved acquisition and image analysisprotocols specifically adapted to the informationprovided by 7-T MR imaging.Brain imaging with 7-T MR probablycan be used to fully define a widerange of macroscopically visible findingsin patients with hippocampal sclerosis,including atrophy of hippocampal subregionsand deformities of the hippocampalhead and body. Future applicationsof 7-T MR imaging in presurgical evaluationsmay benefit the numerous individualswith refractory mesial TLE who havenormal or nonspecifically abnormal brainimaging findings at clinical MR imaging( 6–8 ). Increased spatial and contrastresolution at 7 T might enable unilateraldetection of mild HS in patients currentlyconsidered to have MR-negativeTLE, leading to efficacious ablative surgery.Alternatively, improved detectionof bilateral HS in patients who currentlyhave only unilateral hippocampal abnormalitiesdetected with clinical MR imagingmight enable us to avoid intracranial208 radiology.rsna.org n Radiology: Volume 261: Number 1—October 2011