ORIGINAL CONTRIBUTION. Voxel-Based Morphometry Reveals Gray Matter Network Atrophy in Refractory Medial Temporal Lobe Epilepsy

Transcription

1 ORIGINAL CONTRIBUTION -Based Morphometry Reveals Gray Matter Network Atrophy in Refractory Medial Temporal Lobe Epilepsy Leonardo Bonilha, MD, PhD; Chris Rorden, PhD; Gabriela Castellano, PhD; Fabrício Pereira, BSc; Pablo A. Rio, BSc; Fernando Cendes, MD, PhD; Li M. Li, MD, PhD Background: Conventional volumetric studies have shown that brain structures functionally and anatomically related to the hippocampus are smaller in patients with drug-refractory medial temporal lobe epilepsy (MTLE). Objectives: To determine the extent of gray matter atrophy in the brains of patients with MTLE and to examine the pattern of atrophy. Design: We performed a voxel-based morphometric study of 43 consecutive patients with unilateral drugrefractory MTLE (21 patients with right-sided MTLE and 22 patients with left-sided MTLE) whose magnetic resonance images showed signs of unilateral hippocampal atrophy. The data from the patients with MTLE were compared with the data from 49 healthy control subjects to identify differences between groups in gray matter concentration (GMC). Setting: Academic hospital s epilepsy clinic. Results: We observed that patients with left- and rightsided MTLE exhibited GMC reduction in the hippocampus ipsilateral to the seizure origin. In addition, we found GMC reduction in the ipsilateral parahippocampal and isocortical temporal regions. Patients with MTLE also showed GMC reduction in subcortical nuclei such as the thalamus and caudate, in the cerebellum, in the midbrain, and in parieto-occipital regions. Conclusions: Patients with MTLE exhibit a reduction in GMC in regions outside the temporal lobe, specifically in areas that are connected to the hippocampus and parahippocampal region, suggesting an anatomical route for atrophy. Arch Neurol. 2004;61: Author Affiliations: Laboratory of Neuroimaging, State University of Campinas, Campinas, Brazil (Drs Bonilha, Castellano, Cendes, and Li and Messrs Pereira and Rio); and School of Psychology, University of Nottingham, Nottingham, England (Dr Rorden). HIPPOCAMPAL SCLEROSIS (HS) is the most common pathologic abnormality associated with medial temporal lobe epilepsy (MTLE). 1 This abnormality is defined by neuronal loss and gliosis involving the CA1, CA3, and CA4 hippocampal subregions and the granule cell layer of the dentate gyrus, with relative sparing of the CA2 subregion. 2 The use of hippocampal volumetric quantification through magnetic resonance imaging (MRI) 3,4 has been successfully applied to detect HS in vivo, and its reliability is based on this measurement s correlation with neuronal counts of surgically resected hippocampal tissue. 5 Morphometric studies of the medial and anterior portion of the temporal lobe in patients with MTLE have demonstrated a reduction in the volume of structures functionally and anatomically related to the hippocampus, indicating that neuronal damage in the MTLE extends beyond the hippocampus. The entorhinal cortex and perirhinal cortices are the areas that convey most of the incoming information to the hippocampus, 6 and their volume is significantly reduced in patients with MTLE Neurophysiologic studies performed in patients with MTLE have shown that a large network is involved in the generation and maintenance of seizures and that in some cases the onset of seizures is outside of the hippocampus. 14,15 Because the neural network of the MTLE involves areas connected to the hippocampus and limbic system, there is reason to predict neuronal loss in the areas that participate in the network, both inside and outside the temporal lobe. We analyzed whole-brain gray matter concentration (GMC) in patients with MTLE using voxel-based morphometry (VBM). Our aim was to identify gray matter abnormalities in these patients, with special attention to areas functionally connected to the hippocampus and limbic system. 1379

2 METHODS SUBJECTS We recruited 49 healthy subjects who were State University of Campinas (Campinas, Brazil) employees and their acquaintances (17 men) with a mean ± SD age of 34±11 years, ranging from 19 to 60 years. We also studied 43 consecutive patients with chronic refractory MTLE. Twenty-two patients had leftsided MTLE (8 men) with a mean±sd age of 38±8 years, ranging from 18 to 54 years; 21 patients had right-sided MTLE (7 men) with a mean±sd age of 32±8 years, ranging from 17 to 55 years. There was no significant difference in age (F 2,89 =1.6; P=.20) or sex (Pearson 2 =0.02; P=.90) distribution between controls and patients with MTLE. All patients were referred from the outpatient epilepsy clinic of the State University of Campinas Hospital, where they were diagnosed based on their clinical history and physical examination results. Complementary investigation involved interictal electroencephalography (EEG), computed tomography, and MRI. All patients were diagnosed as having epileptic syndrome based on criteria from the International League Against Epilepsy. 16 Seizures were lateralized according to the medical history, a comprehensive neurological examination, interictal EEG, and prolonged video-eeg monitoring seizure onset. Visual analysis of the MRI studies revealed unilateral hippocampal atrophy in all included patients, a finding associated with HS. 17,18 All patients were considered to have drug-refractory MTLE with unilateral seizure onset and unilateral hippocampal atrophy. MRI ACQUISITION We acquired diagnostic MRI studies using a standardized protocol, 17 collecting T1-weighted images with either 1-mm isotropic voxels or mm voxels. All images were acquired with the same 2-T scanner (Prestige; Elscint, Haifa, Israel) using a spoiled gradient-echo sequence (time to repeat=22 ms; echo time=9 ms; flip angle=35 ; matrix= pixels). PREPROCESSING OF DATA The DICOM format images were transformed into ANALYZE format using MRIcro software (C.R.; 19 The VBM analysis was performed using SPM2 software (Wellcome Department of Imaging Neuroscience, London, England; 20 The SPM2 software s normalization stage was used to match the overall size and shape of each individual s MRI study to the T1.img template image supplied by SPM2. Default SPM2 settings were used for normalization parameter estimation (using 12 linear parameters and nonlinear basis functions as well as a brain mask to ensure that the fit was based on the shape of the brain rather than the surrounding scalp). Spatially normalized images were resliced to an isotropic 1.5 mm. Subsequent to normalization, the images from all participants were in a common stereotaxic space, allowing analysis between individuals and close correspondence to other neuroimaging studies. Next the images underwent segmentation of gray matter using SPM2 s built-in routines, which estimate the GMC for each voxel. In conventional VBM, the normalization process can enlarge atrophied regions when matching images from patients to stereotaxic space (the template image used in normalization is based on images from a neurologically healthy sample). Good et al 21 have proposed a technique for modulating the estimated concentration of tissue in segmented images based on the spatial deformations selected during normalization. This technique compensates for the deformation of the brain tissue during the normalization process, preserving the quantity of tissue (eg, gray matter) while ensuring a good spatial alignment between patients and controls. Therefore, we used the code developed by Good and colleagues 21 to modulate the GMC in the brains of participants. This modulation step was applied after the segmentation step. Finally, the images were convolved with a 10-mm isotropic gaussian kernel to minimize gyral interindividual variability. This smoothing creates images that are more normally distributed and permits voxelwise analysis. The resulting images were compared using the t test to search for differences in GMC between control subjects and patients with MTLE. Contrasts were defined to estimate the probability of each voxel being gray matter. STATISTICAL ANALYSIS Group differences for age were assessed using 1-way analysis of variance, and the sex distribution was evaluated with the 2 test. The normalized, segmented, modulated, and smoothed data were analyzed using SPM2 software. We conducted 2 analyses of GMC: one compared patients who had left-sided MTLE with the control participants, and the second compared those who had right-sided MTLE with controls. The parameters for these 2 comparisons were identical and were performed using 2-sample t tests. This analysis included grand mean scaling, proportional threshold masking (set to 0.8), and brain masking. The results from the analysis appear in a parametric map of the t statistic (SPM (t) ), and the SPM (t) is corrected for normal distribution (SPM (z) ). Our statistical threshold was a false discovery rate of 1% 22 with an extent threshold looking for clusters with at least 32 contiguous voxels. This technique attempts to control the rate of statistical false alarms during multiple statistical comparisons. In situations in which a real signal is present in the data, this technique typically offers increased statistical power compared with traditional techniques for controlling for familywise error (eg, the Bonferroni correction). In the context of VBM, a false discovery rate of 0.01 implies that approximately 1 of 100 voxels identified as statistically significant is actually a false alarm. RESULTS The SPM (t) and SPM (z) values are displayed in Table 1 and Table 2, revealing the regions of reduced GMC in patients with left-sided MTLE and right-sided MTLE as compared with control subjects. No regions of GMC excess were observed in patients with MTLE compared with controls, even with a liberal analysis performed without an extent threshold and an uncorrected threshold of P.05. The SPM (t) displayed in a normal T1-weighted MRI template is shown in the Figure. There was a significant reduction of GMC in patients with refractory MTLE that affected the hippocampus ipsilateral to the seizure focus and other regions beyond the medial temporal lobe. We observed a symmetrical pattern of GMC reduction in patients with left- and rightsided MTLE. Although the atrophied regions were similar in the 2 groups, the extent of statistically significant atrophy was greater in patients with left-sided MTLE. In patients with right-sided MTLE, we observed that the re- 1380

3 Table 1. -Based Morphometric Results: GMC Reduction in Patients With Right-Sided MTLE* Location (cor) k (equiv) (unc) (FWE cor) (FDR cor) t Statistic z Score (equiv) Coordinates (x, y, z) Right cerebellum ,-69,-36 Left cerebellum ,-83,-42 Left middle frontal gyrus ,-21, 44 Right temporal lateral isocortex ,-30,-20 Right caudate nucleus , 14, 15 Right thalamus ,-14, 12 Right cingulate gyrus , 39, 29 Posterior portion of right hippocampus ,-41, -2 Right temporo-occipital region ,-80, 3 Anterior portion of right hippocampus and right parahippocampal gyrus , 2,-32 Left parieto-occipital region ,-48, 36 Left thalamus ,-17, 9 Right middle frontal gyrus ,-18, 41 Right dorsal midbrain ,-36,-11 Abbreviations: cor, corrected; equiv, equivalent; FDR, false discovery rate; FWE, familywise error rate; GMC, gray matter concentration; MTLE, medial temporal lobe epilepsy; unc, uncorrected. * P values are corrected for entire volume with significance at P.01. The uncorrected P value was P.001 for all locations. size is 32. Table 2. -Based Morphometric Results: GMC Reduction in Patients With Left-Sided MTLE* Location (cor) k (equiv) P value (unc) (FWE cor) (FDR cor) t Statistic z Score (equiv) (unc) Coordinates (x, y, z) Right cerebellum ,-69,-35 Left cerebellum ,-84,-41 Posterior portion of left hippocampus ,-39, 3 Left frontal operculum , 32, 11 Left cingulate gyrus , -8, 47 Left insulae , 3, -5 Right parieto-occipital area ,-69, 41 Left thalamus ,-11, 20 Anterior portion of left hippocampus ,-15,-18 Left fornix ,-36, 11 Left parieto-occipital area ,-65, 30 Left inferior frontal gyrus , 38, 3 Left caudate , 5, 11 Right caudate , 11, 17 Dorsal midbrain ,-19,-14 Right thalamus ,-15, 17 Right cingulate gyrus ,-41, 54 Right occipital area ,-68, 3 Left superior frontal sulcus , 41, 32 Right frontal operculum , 27,-18 Abbreviations are given in the footnote to Table 1. * P values are corrected for entire volume with significance at P.01. size is 32. duction of GMC affected the right hippocampus and right parahippocampal gyrus, left and right hemispheres of the cerebellum, bilateral thalamus, right caudate nucleus, right temporal isocortex, right temporo-occipital region, left and right middle frontal gyri, dorsal midbrain, and left parieto-occipital region (Figure). In patients with leftsided MTLE, we observed a reduction of GMC in the left hippocampus and left parahippocampal gyrus, left insulae, left and right cingulate gyri, left and right frontal opercula, left and right thalami, left and right parietooccipital regions, left and right cerebellum, midbrain, and bilateral caudate nuclei. COMMENT We performed a VBM analysis of patients with chronic refractory MTLE and found a reduction in GMC in brain areas within and beyond the hippocampus and temporal lobe. We observed a GMC reduction in patients with 1381

4 7 Z 3 7 Z Figure. The parametric map of the t statistic depicts the location and the statistical significance of voxels, with a decreased density of gray matter in patients with right-sided (top row) and left-sided (bottom row) medial temporal lobe epilepsy compared with controls. The map illustrates a multislice display of coronal images of a T1 template of a normal brain. A color bar indicating the z score value is shown at the right. A parasagittal slice of the T1 template is also displayed at the right, showing the location of the slices. Left side of images corresponds to the right brain hemisphere, and right side corresponds to the left brain hemisphere. 4 right- or left-sided MTLE involving areas that are widely distributed in the brain, located not only within the temporal lobes but also in brain regions such as the diencephalon, insulae, midbrain, and other isocortical areas. The VBM analysis suggests that areas functionally and anatomically related to the hippocampus undergo volume reduction. We observed that the patients with right-sided MTLE had a symmetrical pattern of volume reduction compared with those who had left-sided MTLE. This suggests similar pathologic consequences in both hemispheres. However, we found that the extent of statistically significant GMC reduction was more widespread in patients with left-sided MTLE, affecting clusters with a larger number of voxels. This may reflect a different behavior of left-sided MTLE, not in terms of the structures involved but rather in the intensity of damage. In recent studies, Keller et al 23,24 observed that patients with MTLE showed significant alterations in GMC, either an increase or decrease of GMC in specific regions. They found that the hippocampus had a significant decrease in GMC in patients with MTLE compared with controls 23,24 and that the reduction of GMC in the hippocampus was not dependent on the time of seizures. The GMC reduction was also observed in regions such as the dorsal prefrontal cortex of the right brain hemisphere 24 as well as the bilateral thalamic, prefrontal, and cerebellar regions. 23 In turn, GMC reduction in these regions was associated with the duration of epilepsy. The authors observed that patients with MTLE showed a significant increase in GMC in the parahippocampal, pericallosal, and cerebellar regions, 24 and this was interpreted as a reflection of white matter atrophy or structural displacement due to cerebrospinal fluid expansion. The findings of Keller et al 23,24 were not consistent with a previous VBM study by Woermann et al, 25 who studied 10 patients with left-sided MTLE and HS and 10 patients who had MTLE without hippocampal atrophy. Woermann and colleagues found a decrease of GMC in only a few patients when comparing each individual with the control group. They did not observe a reduction of GMC when comparing the group of patients who had MTLE with the group of control subjects, but in patients with MTLE who had normal MRI results, they found an excess of GMC in the inner interface of the temporal posterior region. However, the small number of patients examined in their study 25 may mean that the null results observed in the group comparison were due to limited statistical power. The use of different techniques for image normalization can probably explain the different results between our study and those by Keller et al. 23,24 In their studies, Keller and colleagues observed a widespread GMC reduction in the brains of patients with MTLE but did not observe GMC reduction in areas that are known from conventional morphometric studies to be atrophied in these patients, such as the parahippocampal gyrus, 7-13 thalamus, 26 and caudate nuclei. 27 The detection of GMC changes in the hippocampal region is hampered by the lack of sharp macroscopic boundaries between gray and white matter in the temporal lobe. 25,28 Recently described techniques have aimed to improve normalization and segmentation procedures and have increased the sensitivity of VBM analysis. 28 These technical improvements together with a larger sample size have made it possible to identify GMC differences in the medial temporal lobe region in patients with Alzheimer disease 29 and in healthy subjects who perform repetitive tasks. 30 This may partly explain the discrepancies between the VBM studies of MTLE described previously Normalization is a necessary but potentially treacherous stage of VBM analysis. Normalization is required to ensure that the same brain regions can be compared between individuals, but it can also reduce the structural abnormalities that will be investigated with VBM, transforming the shape of the brain image during the process and as a consequence artificially inflating atrophied areas (ie, ensuring a better spatial match between individuals but consequently yielding less difference in GMC between patients and controls). All prior VBM studies of MTLE have included nonlinear functions during normalization. Nonlinear functions greatly improve the fit of normalization but can also cause dramatic distortions to abnormal tissue. 21 We used a modulation of gray matter volume based on the warping applied during normalization. 21 This technique allows us to use normalization while minimizing the danger of distorting the structural abnormalities in the brains of the patients. Similarly, we used the false discovery rate to increase the sensitivity of multiple voxelwise comparisons, controlling 1382

5 for familywise error while providing a reasonable level of statistical power. 22 These differences in the methods used may well explain the different results we obtained in comparison with previous studies. The regions of GMC reduction beyond the hippocampus in patients with MTLE are not completely defined, specifically outside the temporal lobe and in the diencephalon. One possible explanation is that the areas of GMC reduction and therefore neuronal damage are those contained within the path observed in neurophysiologic studies as the mesolimbic route of seizure propagation. Spencer 14 revisited the issue of neural networks as a basis for the generation and maintenance of seizures, 15 describing possible routes for seizure spread formed by anatomically connected cortical and subcortical structures. 14,15 According to this theory, the hyperexcitability associated with seizures reverberates in the whole network. 14 If MTLE is indeed an expression of dysfunction in the neural networks, one would expect to find atrophy throughout the structures of the network. Animal and functional studies have shown that the hippocampus is densely connected to the parahippocampal region 6,31,32 and to thalamocortical circuits These regions are expected to be part of the network involving the hippocampus, which is affected in MTLE. In this study, we have observed a pattern of GMC atrophy in patients with MTLE that involves the hippocampus, parahippocampal region, and thalamocortical structures. We have found GMC reduction in regions that are functionally and anatomically connected to the medial temporal limbic system. The course of treatment following MRI varied for our patients: only a subset required surgical resections, and only a portion of these had preoperative assessment using single-photon emission computed tomography (a technique that can identify the regions most heavily activated during a seizure). Therefore, we did not have access to histological or blood flow analysis for these patients. However, because they were homogeneous in terms of having drugrefractory MTLE with unilateral MRI signs of HS and without additional lesions, they could be comparable with patients with MTLE who were evaluated in other studies. 1,4,14 We have observed that different areas within the brains of patients with MTLE exhibit GMC reduction. This group of structures has an underlying anatomical order; most of them are heavily connected and are associated with the hippocampus or pertain to the limbic system. The data presented in this article are unable to prove whether these structures are actually interconnected and part of a network. Resolving the relationship between these regions will require evaluating the functional connectivity of the structures and examining the white matter tracts between these areas. We have shown that patients with drug-refractory MTLE exhibit a GMC that is not restricted to the hippocampus but rather affects the brain in a widespread and organized fashion. However, this is an observation based on patients with a long history of poor seizure control and long-term use of antiepileptic drugs. The influence of long-term medication use or recurrent seizures on the pattern of GMC reduction is not yet understood. Moreover, it is not clear whether the pattern of GMC reduction we have observed applies to patients with other forms of refractory epilepsy. It is still unclear why some regions that are synchronously activated by the proposed network exhibit GMC reduction, 14 whereas other regions do not. This may reflect different susceptibility of brain regions to the excitotoxic effects of seizures. In addition, the presence of GMC reduction in regions that are not directly linked to the hippocampus and limbic system 14 may indicate that multiple networks are involved in maintaining and spreading seizures or that an undiscovered network lies beneath the pattern of seizure propagation in MTLE. However, the observation that GMC reduction primarily affects areas with connections to the medial portion of the temporal lobe points to a possible route of neuronal damage located within the networks involved in seizure generation and the maintenance of seizures in patients with MTLE. Accepted for publication: March 25, Correspondence: Li M. Li, MD, PhD, Department of Neurology, Faculty of Medicine, UNICAMP, , Campinas, SP, Brazil (limin@fcm.unicamp.br). Author Contributions: Study concept and design: Bonilha, Pereira, Rio, Cendes, and Li. Acquisition of data: Bonilha, Castellano, Pereira, Rio, Cendes, and Li. Analysis and interpretation of data: Bonilha, Rorden, and Li. Drafting of the manuscript: Bonilha, Pereira, Rio, and and Li. Critical revision of the manuscript for important intellectual content: Bonilha, Rorden, Castellano, Cendes, and Li. Statistical expertise: Bonilha, Rorden, Castellano, and Li. Obtained funding: Bonilha, Cendes, and Li. Administrative, technical and material support: Pereira, Rio, and Li. Study supervision: Cendes and Li. Funding/Support:This research is supported by grant 00/ from the Foundation of Support to the Research of the State of São Paulo, São Paulo, Brazil. REFERENCES 1. Babb TL, Brown WJ. Pathological findings in epilepsy. In: Engel J Jr, ed. Surgical Treatment of the Epilepsies. New York, NY: Raven Press; 1987: Margerison JH, Corselis JAN. Epilepsy and the temporal lobes: a clinical electroencephalographic and neuropathological study of the brain in epilepsy, with particular reference to the temporal lobes. Brain. 1966;89: Jack CR, Sharbrough FW, Twomey CK, et al. Temporal lobe seizures: lateralization with MR volume measurements of the hippocampal formation. Radiology. 1990;175: Cendes F, Andermann F, Gloor P, et al. MRI volumetric measurements of amygdala and hippocampus in temporal lobe epilepsy. Neurology. 1993;43: Lencz T, McCarthy G, Bronen RA, et al. Quantitative magnetic resonance imaging in temporal lobe epilepsy: relationship to neuropathology and neuropsychological function. Ann Neurol. 1992;31: Squire LR, Zola-Morgan S. The medial temporal lobe memory system. Science. 1991;253: Bernasconi N, Bernasconi A, Andermann F, Dubeau F, Feindel W, Reutens DC. Entorhinal cortex in temporal lobe epilepsy: a quantitative MRI study. Neurology. 1999;52: Bernasconi N, Bernasconi A, Caramanos Z, Andermann F, Dubeau F, Arnold DL. Morphometric MRI analysis of the parahippocampal region in temporal lobe epilepsy. Ann N Y Acad Sci. 2000;911: Bernasconi N, Bernasconi A, Caramanos Z, et al. Entorhinal cortex atrophy in epilspsy patients exhibiting normal hippocampal volumes. Neurology. 2001; 56: Bernasconi N, Bernasconi A, Caramanos Z, Antel SB, Andermann F, Arnold DL. Mesial temporal damage in temporal lobe epilepsy: a volumetric MRI study of 1383

6 the hippocampus, amygdala and parahippocampal region. Brain. 2003;126: Jutila L, Ylinen A, Partanen K, et al. MR volumetry of the entorhinal, perirhinal, and temporopolar cortices in drug-refractory temporal lobe epilepsy. AJNR Am J Neuroradiol. 2001;22: Salmenpera T, Kalviainen R, Partanen K, Mervaala E, Pitkanen A. MRI volumetry of the hippocampus, amygdala, entorhinal cortex, and perirhinal cortex after status epilepticus. Epilepsy Res. 2000;40: Salmenpera T, Kalviainen R,Partanen K, Pitkanen A.Quantitative MRI volumetry of the entorhinal cortex in temporal lobe epilepsy. Seizure. 2000;9: Spencer SS. Neural networks in human epilepsy: evidence of and implications for treatment. Epilepsia. 2002;43: Wennberg R, Arruda F, Quesney LF, Olivier A. Preeminence of extrahippocampal structures in the generation of mesial temporal seizures: evidence from human depth electrode recordings. Epilepsia. 2002;43: Commission on Classification and Terminology of the International League Against Epilepsy. Proposal for revised classification of epilepsies and epileptic syndromes. Epilepsia. 1989;30: Kobayashi E, D Agostino MD, Lopes-Cendes I, et al. Hippocampal atrophy and T2-weighted signal changes in familial mesial temporal lobe epilepsy. Neurology. 2003;60: Kobayashi E, Lopes-Cendes I, Guerreiro CA, Sousa SC, Guerreiro MM, Cendes F. Seizure outcome and hippocampal atrophy in familial mesial temporal lobe epilepsy. Neurology. 2001;56: Rorden C, Brett M. Stereotaxic display of brain lesions. Behav Neurol. 2000;12: Friston KJ, Holmes AP, Worsley KJ, Poline JB, Frith CD, Frackowiak RSJ. Statistical parametric maps in functional imaging: a general linear approach. Hum Brain Mapp. 1995;2: Good CD, Johnsrude IS, Ashburner J, Henson RN, Friston KJ, Frackowiak RS. A voxel-based morphometric study of ageing in 465 normal adult human brains. Neuroimage. 2001;14: Genovese CR, Lazar NA, Nichols T. Thresholding of statistical maps in functional neuroimaging using false discovery rate. Neuroimage. 2002;15: Keller SS, Wieshmann UC, Mackay CE, Denby CE, Webb J, Roberts N. based morphometry of grey matter abnormalities in patients with medically intractable temporal lobe epilepsy: effects of side of seizure onset and epilepsy duration. J Neurol Neurosurg Psychiatry. 2002;73: Keller SS, Mackay CE, Barrick TR, Wieshmann UC, Howard MA, Roberts N. based morphometric comparison of hippocampal and extrahippocampal abnormalities in patients with left and right hippocampal atrophy. Neuroimage. 2002; 16: Woermann FG, Free SL, Koepp MJ, Ashburner J, Duncan JS. -by-voxel comparison of automatically segmented cerebral gray matter a rater-independent comparison of structural MRI in patients with epilepsy. Neuroimage. 1999; 10: Natsume J, Bernasconi N, Andermann F, Bernasconi A. MRI volumetry of the thalamus in temporal, extratemporal, and idiopathic generalized epilepsy. Neurology. 2003;60: Dreifuss S, Vingerhoets FJ, Lazeyras F, et al. Volumetric measurements of subcortical nuclei in patients with temporal lobe epilepsy. Neurology. 2001;57: Ashburner J, Friston KJ. -based morphometry the methods. Neuroimage. 2000;11: Baron JC, Chetelat G, Desgranges B, et al. In vivo mapping of gray matter loss with voxel-based morphometry in mild Alzheimer s disease. Neuroimage. 2001; 14: Maguire EA, Gadian DG, Johnsrude IS, et al. Navigation-related structural change in the hippocampi of taxi drivers. Proc Natl Acad Sci U. S. A. 2000;97: Kelly ME, McIntyre DC. Perirhinal cortex involvement in limbic kindled seizures. Epilepsy Res. 1996;26: McIntyre DC, Kelly ME. The parahippocampal cortices and kindling. Ann N Y Acad Sci. 2000;911: Bertram EH, Mangan PS, Zhang D, Scott CA, Williamson JM. The midline thalamus: alterations and a potential role in limbic epilepsy. Epilepsia. 2001;42: Bertram EH, Scott C. The pathological substrate of limbic epilepsy: neuronal loss in the medial dorsal thalamic nucleus as the consistent change. Epilepsia. 2000; 41(suppl 6):S3-S Bertram EH. Functional anatomy of spontaneous seizures in a rat model of limbic epilepsy. Epilepsia. 1997;38: Lothman EW, Bertram EH III, Stringer JL. Functional anatomy of hippocampal seizures. Prog Neurobiol. 1991;37: Zhang DX, Bertram EH. Midline thalamic region: widespread excitatory input to the entorhinal cortex and amygdala. J Neurosci. 2002;22: