In 1849, Robert Bentley Todd postulated that each epileptic seizure “does some amount of damage to the brain” resulting in long term neuropathological changes.1 Subsequently, investigators confirmed the association of neuropathological changes, particularly hippocampal sclerosis (HS),2 in patients with epileptic seizures. However, many of the issues surrounding the origins of HS and its causal relationship with epilepsy remain uncertain. Recently, MRI has provided new insights into the clinical issues and questions associated with HS.3

A previous study demonstrated significant clinicopathological, structural and functional imaging differences between temporal lobe epilepsy (TLE) associated with hippocampal sclerosis (HS+ve) and TLE with no evidence of hippocampal sclerosis on MRI (HS−ve).4 Apart from the lack of evidence of the defining hippocampal sclerosis (HS) on MRI, HS−ve patients in general showed lateralised but more widespread temporal hypometabolism on fluorodeoxyglucose–positron emission tomography (FDG-PET) with blinded visual assessment. This group was therefore dubbed “MRI negative PET positive TLE.” Other findings in the HS−ve group included a less frequent history of febrile convulsions; slower rhythms at ictal EEG onset; less frequent histopathological HS; and a similarly good post surgical outcome even in the subgroup of HS−ve patients who had undergone a hippocampal sparing procedure.

In the current study, we applied large deformation high dimensional mapping (HDM-LD) techniques to generate detailed hippocampal surface maps of the HS+ve and HS−ve groups. We have documented previously the validity of this deformation based technique in patients with mesial temporal sclerosis (MTS).5 HDM-LD measurements of hippocampal shape in subjects with MTS using intrasubject comparisons6 and comparisons with controls7 show significant shape changes, with marked inward deviation in the Sommer sector of the MTS hippocampi. In the hippocampi contralateral to MTS, the inferior surface of the hippocampal body shows inward deformation in the medial aspect of the subiculum, with minimal involvement of the Sommer sector. Additionally, using analysis of principal dimensions to quantify hippocampal shape change suggests that MTS, after accounting for normal right–left asymmetries, affects the right and left hippocampal surface structure very symmetrically.8

Because past studies in other neuropsychiatric diseases using HDM-LD maps of the hippocampus show significant changes despite lack of total hippocampal volume changes,9^–11 we postulate that there will be HDM-LD defined changes associated with the HS−ve group. Comparing HDM-LD hippocampal maps of the HS−ve and HS+ve groups will also provide some insights into the associated pathophysiology of the hippocampus in each of these temporal lobe epilepsy syndromes.

METHODS

This study involved the same epilepsy patient cohort as that from a previously reported case control study.4 Cases comprised 30 consecutive patients with clinically and video-EEG defined non-lesional TLE and well lateralised ictal EEG changes, but without evidence of HS on MRI, including hippocampal volumetry (HS−ve). All patients had been admitted for a comprehensive inpatient assessment, including video-EEG monitoring between 1996 and 2002 at one of three tertiary referral hospitals. Another group, labelled as hippocampal sclerosis positive (HS+ve), were 30 patients with well lateralised unilateral ictal and/or interictal epileptiform EEG discharges and concordant unequivocal evidence of hippocampal sclerosis on MRI, confirmed by MR volumetry. The HS+ve subjects were selected by starting with the most recent patient and moving consecutively retrospectively through the epilepsy monitoring database, matching as appropriate for age and sex. HS−ve patients represented close to 20% of medically refractory partial epilepsy patients assessed at these institutions. As part of a comprehensive evaluation for refractory epilepsy, all subjects underwent evaluation for concomitant psychiatric disease, and were excluded if they had a history of active major psychiatric illnesses. The study was approved by the institutional ethics committees of The Royal Melbourne Hospital, St Vincent’s Hospital, The Peter MacCallum Cancer Institute and St Louis University.

Historical features in the subjects have been described previously.4 However, to control for normal right–left asymmetries of hippocampal surface structure, subjects were regrouped based on laterality of onset of epileptic seizures and presence of HS. The gender ratio, age, duration of epilepsy and preoperative seizure frequency scores classified according to a 12 point seizure frequency score12 were calculated for each of the four groups.

For HDM-LD comparison, the epilepsy subject groups were compared with a group of 27 normal controls. The control group consisted of a consecutively acquired series of MR studies acquired from normal control subjects. The volunteers for normal control MR studies had no history of central nervous system disease, significant head trauma or alcohol abuse.

MRI methods

Imaging of epilepsy subjects was performed on a 1.5 T Signa scanner (General Electric, Milwaukee, Wisconsin, USA). Whole brain acquisitions were performed in the coronal plane using a T1 weighted MPRAGE technique. Voxel dimensions were 0.98 mm×0.98 mm×1.5 mm. Field of view was 24 cm×24 cm. Matrix size was 256×256.

Control subjects were acquired on a separate scanner. MR scanning was performed with a 1.5 T Signa scanner (General Electric, Milwaukee, Wisconsin, USA). Whole brain acquisitions were obtained in the coronal plane with a fast spoiled grass technique, TR = 14, TE = 3, flip angle = 30°. Voxel dimensions were 0.859 mm×0.859 mm×1.5 mm. Field of view was 22 cm×22 cm. Matrix size was 256×256.

All groups, as well as control subjects, underwent measurement of the intracranial area in the midsagittal plane, which was performed as described by Free and colleagues,13 tracing along the inner limit of the subcutaneous fat over the convexity, along the margins of the cerebral hemispheres at the base of the brain, and including the brainstem to the foramen magnum. Intracranial area was used as a surrogate measure for intracranial volume. Because intracranial area was significantly larger for the right HS−ve group, at a ratio of 0.9485 as compared with controls, MR studies in the right HS−ve group were reduced in size by a ratio of 0.9485 before HDM-LD of the hippocampi.

We performed deformation segmentations as previously described.5 The deformation segmentation procedure resulted in generation of coordinates for transformation of a normal template hippocampal segmentation (from a single normal subject, otherwise not included in the study) into the shape of the hippocampi of the epilepsy and control subjects. We therefore generated coordinates for transformation for each subject in the study. To generate an “average” hippocampus, we generated a mean transformation for the deformation images in each group. This mean transformation was then applied to the atlas itself to generate the “average” hippocampi for each group.6

To further quantify the difference between the HS+ve and HS−ve hippocampi and control subjects, we calculated a minimum mean squared error estimation by coregistering the “average” hippocampal surfaces of the epilepsy groups to the control group.6 By calculating differences in hippocampal transformation coregistration using the minimum mean squared error, we made a direct comparison between the shape of the hippocampi in the ipsilateral cerebral hemisphere between the epilepsy groups and the control group. Final results were projected on the epilepsy group hippocampal surfaces. To define shape changes in the hippocampus, we calibrated a “flame” scale, which represents the distance between the matched hippocampi. The maximum and minimum values of the “flame” scale were set using the maximum value of inward deformation. Deformation patterns were visually assessed, comparing differences between the right and left TLE groups.

RESULTS

Sex, duration of epilepsy and seizure severity

There were 15 subjects in each epilepsy group (R HS+ve, R HS−ve, L HS+ve, L HS−ve). There were 27 normal controls. The female/male ratios of the groups were: R HS+ve 5/10, R HS−ve 5/10, L HS+ve 8/7, L HS−ve 8/7 and normal controls 13/14. Mean (SD) duration of epilepsy (in years) was: R HS+ve 19.1 (12.2), R HS−ve 18.4 (12.6), L HS+ve 18.9 (10.4) and L HS−ve 18.5 (9.3). Unpaired t test values comparing duration of epilepsy for the left TLE groups was p = 0.91 and for the right TLE groups, p = 0.89. Seizure severity scores according to a 12 point seizure frequency score, comparing the left HS+ve and HS−ve groups were p = 0.56, and comparing the right HS+ve and HS-ve groups, p = 0.44.

Results for mean age, intracranial size and hippocampal volumes are presented in table 1.

HDM-LD results

Figures 1 and 2 show the results of hippocampal surface deformations. The hippocampi affected by HS show similar patterns in both the right and left HS+ve groups, with maximal involvement of the lateral surface in the region of the Sommer sector. The contralateral hippocampal surfaces in both the HS+ve groups show differing patterns, without maximal involvement in the Sommer sector.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1 The hippocampal surfaces representing the left temporal lobe epilepsy (TLE) groups. Bold labels describe the groups as follows: (A) left hippocampus of the left temporal lobe epilepsy with hippocampal sclerosis (HS+ve) group; (B) left hippocampus of the left temporal lobe epilepsy without hippocampal sclerosis (HS−ve) group; (C) right hippocampus of the left HS+ve group; and (D) right hippocampus of the left HS−ve group. The left and right hippocampal groups are labelled. Each group of three surfaces represents different three dimensional views of the same composite hippocampus; a superior view on the left, an inferior view on the right and a lateral view positioned below the other two surfaces. The flame scales show the colour representations of positive and negative surface deformations compared with controls (in mm). The anterior (A) and posterior (P) orientations of the hippocampal surfaces are marked for the hippocampal surface representations of groups A and C. The surface maps of group A, which represent the hippocampus affected by HS, show maximal involvement over the lateral aspect of the hippocampus in the region of the Sommer sector, and is labelled with the open arrow. Other surface maps do not show maximal involvement in this region, which is best illustrated by the lateral views in each group. In the mesial temporal sclerosis (MTS)−ve group, both the left and right hippocampi show maximal inward deformation over the superior surface of the medial hippocampal head region, near the uncinate gyrus, which is marked in each MTS−ve hippocampus with a closed arrow.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2 The hippocampal surfaces representing the right temporal lobe epilepsy (TLE) groups. Bold labels describe the groups as follows: (A) right hippocampus of the right temporal lobe epilepsy with hippocampal sclerosis (HS+ve) group; (B) right hippocampus of the right temporal lobe epilepsy without hippocampal sclerosis (HS−ve) group; (C) left hippocampus of the right HS+ve group; and (D) left hippocampus of the right HS-ve group. Views of the hippocampi and flame scales in each group are as described in fig 1. As in the left TLE groups, the lateral views best illustrate the involvement of the Sommer sector, which is maximally involved in group A, the hippocampus affected by HS, and is marked with the open arrow. This region is not maximally involved in the other groups. In the HS−ve group, both hippocampi showed maximal inward deformation over the superior surface of the hippocampal body, which is marked in each mesial temporal sclerosis negative hippocampus with a closed arrow.

Visual inspection showed relative symmetry of involvement of the ipsilateral/contralateral hippocampi within the HS−ve groups. However, comparing patterns between the right HS−ve to left HS−ve groups revealed different patterns. In the right HS−ve group, both hippocampi showed maximal inward deformation over the superior surface of the hippocampal body. In the left HS−ve group, both hippocampi showed maximal inward deformation over the superior surface of the medial hippocampal head region, near the uncinate gyrus.

In comparing the contralateral HS+ve hippocampus with the HS−ve hippocampi, the left-sided TLE groups showed different patterns, with the contralateral HS+ve hippocampus showing maximal involvement over the superior surface of the hippocampal body. However, in the right TLE groups, the overall deformation pattern of the contralateral HS+ve hippocampus and HS−ve hippocampi were very similar.

DISCUSSION

The syndrome of mesial TLE due to MTS is well defined clinically, and includes febrile seizures, onset of complex partial seizures in adolescence or early adulthood, and poor response to antiepileptic medications.2 Histopathologically, the hippocampal regions which are most susceptible to involvement in MTS are the Sommer sector and CA4. Often, there is relative sparing of the CA2 region and the dentate gyrus.14 15 Epilepsy caused by MTS is often surgically remediable, with seizure free rates of approximately 70% after anterior temporal lobectomy and selective amygdalo-hippocampectomy.16

The previous study of this patient population defined a patient cohort of MRI negative PET positive temporal lobe epilepsy, in which subjects had no significant hippocampal volumetric asymmetry on MRI, and compared this population with a matched group of subjects with mesial TLE caused by MTS.4 There were some differences between the groups, such as less frequent febrile seizures and more widespread hypometabolism on [18F]FDG-PET in the HS−ve compared with the HS+ve group. However, the groups showed many clinical similarities with respect to their seizures, including duration and severity of epilepsy, as well as surgical outcome after epilepsy surgery. Because of the similar histories of chronic seizures in HS−ve and HS+ve groups, hippocampal changes as a result of chronic seizures would presumably be the same. Theories of seizure induced hippocampal damage causing MTS,17 and ideas of secondary epileptogenesis and “mirror foci” of involvement of homologous structures bilaterally,18 would predict that patterns of structural hippocampal changes would be similar between groups and in hippocampi contralateral to the epileptogenic side. Evaluation using hippocampal volumes alone would not necessarily detect these structural changes, as they may be insensitive to the presence of MTS.19 Therefore, to further define hippocampal structural involvement in the HS−ve and HS+ve groups, we compared each group with normal controls using computational anatomical techniques.

Post-acquisition automated image processing allows a more objective and sensitive interpretation of three dimensional shape and volume of neuroanatomical structures. In the emerging field of computational anatomy, general pattern theory20 and other mathematical principles provide an analytic framework and tools for studying structures, such as the hippocampus, using HDM-LD.20 HDM-LD generates highly reproducible results in patients with mesial TLE and pathologically verified MTS.5 HDM-LD defines hippocampal surfaces at a subvoxel resolution, and allows depiction of the “averaged” hippocampal surface structure for a group of patients onto a single three dimensional hippocampal map.21 The HDM-LD defined maps of the hippocampus can then be compared between groups.6

Because of normal shape and size differences between the right and left hippocampi,20 in the current study the epilepsy groups were further divided depending on hemisphere of onset of seizures, and each group was compared with the same control group using HDM-LD. The HS+ve hippocampi showed patterns of maximal volume loss over the lateral hippocampal surface, in the region of the Sommer sector, which is consistent with previous HDM-LD defined patterns of MTS.7 Interestingly, none of the contralateral hippocampi in the HS+ve groups, or the hippocampi in the HS−ve groups, showed a maximal deformation in the Sommer sector. This finding would argue that involvement of hippocampal subregions, and possibly the underlying pathophysiology, is different in HS+ve compared with the contralateral hippocampi. The similarity in duration and severity of chronic seizures in the groups suggests that the hippocampal structural differences between the groups are not caused by ongoing seizures. However, febrile seizures were more common in the HS+ve group, suggesting some relationship of acute seizures early in life with HS, which is consistent with the findings of past investigators.3 22 23

In a previous study of mesial TLE and MTS, which included a different set of subjects with mesial TLE than those in the current study, we compared HDM-LD patterns of hippocampal surface structure, and found differences in HS and contralateral hippocampi.7 There is significant volume loss in both the HS and contralateral hippocampi. There is also global brain parenchymal volume loss in subjects with chronic TLE.24^–26 HS hippocampi show a volume loss proportionately greater than global brain volume loss, while contralateral hippocampal volume changes occur in proportion to total brain parenchymal changes.7 The proportionately similar involvement of the entire brain and contralateral hippocampus is suggestive that the hippocampi contralateral to MTS are affected by a global brain process, related to chronic epilepsy. If there is a consistent subtle effect of chronic TLE on hippocampal structure, comparison of the HDM-LD results of the contralateral HS+ve hippocampi with the HS−ve hippocampi offers an initial evaluation of this effect.

There was relative symmetry of involvement of the ipsilateral/contralateral hippocampi within the HS−ve groups. However, comparisons between HS−ve groups showed asymmetrical hippocampal surface patterns. Comparison of contralateral HS+ve with HS−ve hippocampal surfaces was also inconsistent, with the left TLE groups showing different patterns, and the right TLE groups showing similar patterns. Therefore, there was no consistent pattern of involvement in the contralateral HS+ve and HS−ve hippocampi. Several factors may be responsible for this inconsistency, including an underlying difference in the pathophysiology of right and left TLE, or technical issues in MRI acquisition.

Maximal inward deformation patterns of hippocampal surface structure in the contralateral HS+ve and HS−ve hippocampi were similar in degree, and ranged between 1.2 and 2.0 mm. The degree of hippocampal deformation is comparable with that in other neuropsychiatric diseases9 27 28 which show statistically significant changes. Therefore, despite symmetrical hippocampal volumes in the HS−ve groups, there are more subtle hippocampal surface changes defined by HDM-LD. Our findings of HDM-LD hippocampal changes in the HS−ve TLE groups illustrate the promise for further study of correlations between clinical epilepsy and three dimensional neuroanatomical structural changes.

There are several important clinical aspects of correlating structural hippocampal changes with clinical features of epilepsy. In newly diagnosed patients, decreases in hippocampal volume can occur during the initial years of treatment, suggesting that measuring structural hippocampal changes can provide a surrogate marker of the epileptic process.29 A better understanding of the implications of hippocampal structural changes in refractory epilepsy is especially important in the evaluation of subjects for epilepsy surgery. Persistent seizures after failed surgery for TLE can arise from either ipsilateral or contralateral seizure foci. In the case of MTS, some subjects will fail surgery despite electroclinical localisation to the mesial temporal region, and MRI evidence of unilateral hippocampal atrophy.30 Recent studies indicate that structural hippocampal mapping can predict surgical outcomes in TLE. Subjects with ongoing seizures after surgery have more diffuse atrophy in the epileptogenic hippocampus and more segmental atrophy in the anterior and lateral contralateral hippocampus compared with postoperative seizure free subjects.31 While our study does not directly address the issues of ongoing hippocampal volume loss or response to epilepsy surgery, it does add important information about the pathophysiology of hippocampal structure in different TLE syndromes. Further study to understand the natural history of HDM-LD defined hippocampal structural changes and their relationship to progression and response to therapy in epilepsy holds promise for additional clinical applications.

There are limitations to our techniques because the epilepsy and normal control subject MRI scans were acquired on different MRI scanners. Numerous past studies have documented hippocampal volume loss in both MTS and contralateral hippocampi when compared with controls.7 32 The lack of significant contralateral hippocampal volume loss in either MTS group compared with the control group in the current study is likely because of different volume calibrations in the scanners for epilepsy and control groups. Because of this, we chose to visually evaluate relative patterns of hippocampal surface changes between the groups rather than map patterns of statistically significant regional volume loss, as in past HDM-LD studies.7 20 Relative patterns of surface change between groups are less likely to be affected by MRI based volume calibration differences of the epilepsy groups, especially given that all epilepsy patients were compared with the same control group. However, biasing of our findings as a result of using different MRI scanners is possible, and should be considered in interpreting our hippocampal surface structure results. Other major factors which may influence hippocampal volume results, such as age33 or overall head size,13 were matched or corrected for between the groups.

In conclusion, our results show that hippocampal surface structure in the epileptogenic temporal lobe of the HS+ve and HS−ve groups differ. The HS+ve hippocampi showed accentuated changes in the region of the Sommer sector while the HS−ve hippocampi did not show changes in this region. Because seizure duration and severity were similar in the HS+ve and HS−ve groups, differences in segmental volume loss between the two groups are suggestive that the underlying pathophysiology of hippocampal changes in the HS−ve and HS+ve groups is different, and not related to chronic seizure duration or severity.