Address correspondence and reprint requests to Dr. F. Bartolomei at Service de Neurophysiologie Clinique, CHU Timone-264 Rue St. Pierre, 13005-Marseille, France. E-mail: email@example.com
Summary: Purpose: Several studies have demonstrated diminution in the volume of entorhinal cortex (EC) ipsilateral to the pathologic side in patients with temporal lobe epilepsy (TLE). The relation between the degree of EC atrophy and the epileptogenicity of this structure has never been directly studied. The purpose of the study was to determine whether atrophy of the EC evaluated by the quantitative magnetic resonance imaging (MRI) method is correlated with the epileptogenicity of this structure in TLE.
Methods: Intracerebral recordings (SEEG method) of seizures from 11 patients with mesial TLE were analyzed. Seizures were classified according to patterns of onset: pattern 1 was the emergence of a low-frequency, high-amplitude rhythmic spiking followed by a tonic discharge, and pattern 2 was the emergence of a tonic discharge in the mesial structures. A nonlinear measure of SEEG signal interdependencies was used to evaluate the functional couplings occurring between hippocampus (Hip) and EC at seizure onset. MRI volumetric analysis was performed by using a T1-weighted three-dimensional gradient-echo sequence in TLE patients and 12 healthy subjects.
Results: Significant interactions between Hip and Ec were quantified at seizure onset. The EC was found to be the leader structure in most of the pattern 2 seizures. Volumetric measurements of EC demonstrated an atrophy in 63% of patients ipsilateral to the epileptic side. A significant correlation between the strength of EC–Hip coupling and the degree of atrophy was found. In addition, in those patients that had a normal EC volume, the EC was never the leader structure in Ec–Hip coupling.
Conclusions: These results validate the potential role of volumetry to predict the epileptogenesis of the EC in patients with hippocampal sclerosis and MTLE.
Mesial temporal lobe epilepsy (MTLE) is the most frequent type of drug-resistant TLE. It is frequently associated with hippocampal neuronal loss and gliosis, particularly in the CA1 subfield and the dentate hilus (Ammon's horn sclerosis). It is classically thought that the hippocampus is the most important structure in the generation of MTLE seizures (1). However, increasing evidence suggests that mesial temporal structures other than the hippocampus participate in seizure generation, and in particular, the entorhinal cortex (EC) (2–6). The EC is located in the medial part of the temporal lobe and plays a central role in processing high-level sensory information to the hippocampus (7). The extensive reciprocal connections between the EC, the hippocampus, and other brain areas make it a potential candidate for generation and propagation of MTLE seizures. Indeed, numerous in vitro experimental studies using hippocampal–entorhinal preparations have shown that the EC is able to generate spontaneous ictal events and that the EC may have a lower threshold for seizure generation than does the hippocampus (8–13). In addition, neuropathologic studies (14) have reported a characteristic pattern of neuropathologic change consisting of neuronal loss in layer III of the anterior portion of the EC. Recently several neuroradiologic studies (15–21) demonstrated diminution in the volume of EC ipsilateral to the epileptic side in patients with TLE. The rate of patients exhibiting a significant reduction in EC volumes ranges from 52% (18) to 96% (22).
However, the relation between the degree of EC atrophy and the epileptogenicity of this structure has never been directly studied. The purpose of the study was to determine whether the degree of atrophy of the EC, as demonstrated by quantitative magnetic resonance imaging (MRI), is correlated with the role of this structure in mesial temporal lobe seizures, particularly its tendency to generate epileptic activity and its degree of interaction with the hippocampus.
In this study, the activity of EC and hippocampus was studied by using intracerebral recordings (stereoelectroencephalography, SEEG) performed in patients with MTLE undergoing presurgical evaluation. We then used a statistical measure of SEEG signal interdependencies (also referred to as degree of “synchrony” or “association”) that provides an estimation of the degree and direction of coupling between EC and hippocampus. We made the assumption that the degree of interdependency between the EC and hippocampus SEEG activity correlates to the degree of involvement of the EC in the generation of epileptic activity, thus providing a means of quantification.
Patients and stereoelectroencephalography study
Eleven patients were selected from a group undergoing evaluation for surgical treatment of intractable TLE from May 1999 to May 2002. In these cases, SEEG was considered to be necessary as part of presurgical evaluation because of either the need to define the role of langauge regions (thus explaining the higher number of left-sided cases) or the need to ascertain lateralization where this remained unclear after noninvasive investigations. All had (a) seizures that involved mesial temporal regions at the onset, (b) intracerebral electrodes that explored the EC in addition to the hippocampus and other parts of the temporal lobe, and (c) available high-resolution MRI for volumetric study.
All patients had a comprehensive evaluation including detailed history and neurologic examination, neuropsychological testing, routine MRI study, surface electroencephalography (EEG), and SEEG recording of seizures. The latter was performed during long-term video-EEG monitoring.
SEEG recordings were performed by using intracerebral multiple contacts electrodes (10 to 15 contacts; length, 2 mm; diameter, 0.8 mm; 1.5 mm apart) placed intracranially according to Talairach's stereotactic method (23). The positioning of electrodes was determined in each patient from available noninvasive information and hypotheses about the localization of the epileptogenic zone (defined as the set of cerebral regions where simultaneous discharges are observed at seizure onset). The implantation accuracy was peroperatively controlled by telemetric x-ray imaging. A postoperative computed tomography (CT) scan without contrast material was then used to verify both the absence of bleeding and the precise 3D location of each lead. Intracerebral electrodes were then removed, and an MRI was performed, on which the trajectory of each electrode remains visible. Finally, a CT scan/MRI data fusion was performed to locate each lead anatomically along each electrode trajectory.
Several distinct functional regions of the temporal lobe were explored via an orthogonal implantation of depth electrodes (Fig. 1a). All the selected patients had electrodes that spatially sampled mesial/limbic regions (amygdala, EC, and hippocampus) and lateral/neocortical regions of the middle temporal gyrus (MTG). The EC is the rostral part of the parahippocampal cortex extending from the limen insulae (anteriorly) to the anterior part of the lateral geniculate nucleus (posteriorly) and from the sulcus semiannularis in its rostral part and the hippocampal fissure in its caudal part (medially) to the medial bank of the collateral sulcus (laterally) (24). EC was sampled with an electrode passing through the anterior temporobasal region of the temporal lobe. This electrode recorded the lateral MTG cortex, lateral and mesial walls of the occipitotemporal and collateral sulci, and then ended in the EC (Fig. 1b).
Signals were recorded on a Deltamed system (Paris, France) on a maximum number of channels equal to 128. They were sampled at 256 Hz and recorded to hard disk (16 bits/sample) by using no digital filter. The only filter present in the acquisition procedure is a hardware-analog high-pass filter (cut-off frequency equal to 0.16 Hz) used to remove very slow variations that sometimes contaminate the baseline.
Analysis of entorhinal and hippocampal cortex involvement during TLE
Analysis of ictal patterns at seizure onset
Seizure onset was defined by the first electrophysiologic modifications observed before the emergence of clinical signs of seizure. SEEG recordings were first visually analyzed and classified into two categories according to the two main patterns of ictal onset defined in previous studies (25–27)
– Pattern 1: seizure onset characterized by the emergence of a low-frequency (<2 Hz); high-amplitude rhythmic spiking followed by a tonic discharge in one or several mesial structures.
– Pattern 2: seizure onset characterized by the emergence of a tonic discharge in one or several mesial structures without prior spiking activity.
Evaluation of the degree of coupling between hippocampus and entorhinal cortex during seizures by nonlinear regression analysis of SEEG signals
To quantify interactions between EC and hippocampus, a nonlinear regression-analysis method was applied to SEEG signals recorded from both structures. This method, which has previously been described (6,28,29), computes the statistical relation (degree of “association”) between two signals without a priori assumption on the shape of this relation (linear or nonlinear).
Nonlinear regression analysis provides two parameters. The first of these is a nonlinear correlation coefficient, called h2, that describes how a signal X from a given channel is associated with a signal Y from another channel. The estimation of parameter h2 is performed on a temporal window of fixed duration and sliding on both signals to track the temporal evolution of statistical relation. h2 takes its values in [0, 1]. Low values denote that signals X and Y are independent. These can be interpreted as an indicator of low interaction between considered structures (29). Conversely, high values of h2 mean that signal Y may be explained by a transformation (possibly nonlinear) of signal X (i.e., signals X and Y are dependent). They are interpreted as the result of strong interactions between structures (29).
The second parameter is a direction index, called D, which provides information on the causal property of the relation. Parameter D takes into account both the estimated time delay between signals X and Y and the asymmetrical nature of the nonlinear correlation coefficient (generally the h2 value is not the same when computed from X to Y and Y to X). Its values range from –1.0 (X is driven by Y) to 1.0 (Y is driven by X). Statistical properties of estimated quantities h2 and D (bias, variance) also were studied in a physiological model of EEG generation (29) to evaluate their performances and interpret their behavior in real situations.
Optimal parameter setting for h2 and D computation on signals sampled at 256 Hz is obtained for a 5-s window sliding every 500 ms.
Maximal values of h2 obtained at seizure onset including 10 s before and 10 s after the onset of the tonic discharge in mesial temporal lobe structures, between bipolar traces recorded from hippocampus (Hip) and EC, were retained. Three seizures showing the same pattern of onset in each patient were studied. For patients having more than three recorded spontaneous seizures, we arbitrarily chose to analyze the three first seizures recorded during video-SEEG monitoring session. These values were considered to be significant if greater than interictal average values plus 2 standard deviations (6).
MRI acquisition and volumetric analysis
The 11 selected patients (six female and five male patients; mean age, 30 years (±8 years); range, 14–43 years) were compared with 12 healthy individuals (six women and six men) with a mean age of 28 years (±5; range, 18–32 years). Patients with lesions other than HS were excluded from this study.
Patients and control subjects underwent imaging with a 1.5-T Magnetom Siemens scanner (Erlangen, Germany) with a standard head coil and a 3-D magnetization-prepared rapid acquisition gradient-echo (MP-RAGE) sequence with TR/TE/TI of 9.7 ms/4 ms/250 ms, flip angle, 12 degrees; field of view, 320 mm; matrix, 256 × 256. This resulted in 128 contiguous sagittal T1-weighted images with an isotropic voxel of 1.25 × 1.25 × 1.25 mm.
EC volumetry was performed by using the histology-based volumetric method of Insausti et al. (30). Boundaries of the EC were traced by using the interactive mouse-driven software, Brain Voyager. Anatomic guidelines for outlining the hippocampus were those described in previous studies (31,32) (Fig. 3). The intrarater reliability for EC segmentation was evaluated by two measurements made at 3-month interval, by the same observer (M.K.) in 10 control subjects. All hippocampal and EC segmentation was performed by the same investigator (M.K.), who was blinded to clinical findings, and therefore interrater variability was not evaluated for the purposes of this study. Intrarater reliability was calculated by using the absolute value of the difference between two measures divided by the first, expressed as a percentage according to the method used in a previous study (33). The intrarater variability thus calculated was 3.9%. To minimize interindividual variability due to differences in whole intracranial volume (ICV), the volumes were normalized by using two steps. First, the ICV for each patient and control was measured according to Eritaia et al. (34). Second, normalized cortical area volume was calculated with the formula
(Overall mean intracranial area/Intracranial area for each subject) × Subject's measured cortical raw volume according to Insausti et al. (30).
Comparison of quantitative data (comparison of frequencies, comparison of volumes between right and left sides) was made by using nonparametric tests (Mann–Whitney test).
Each patient's volume measurements were standardized relative to the value of normal controls by using a z-score transformation. For individual analysis of volumetric measurements, a z-score of −2.0 on any volumetric measure indicates a raw value that is 2 SD below the mean of normal controls on that measure and was considered significant.
Electrophysiological comparison between patterns 1 and 2 was made by using Fisher's exact test. A p value <0.05 was considered to be statistically significant. Linear regression analysis was performed to determine if a statistically significant correlation exists between h2 values and volumetric data.
Analysis of SEEG recordings
Eleven patients from a series of 110 patients explored with depth electrodes in our epilepsy unit fulfilled the inclusion criteria of this study.
A seizure onset corresponding to pattern 1 was observed in patients 1 through 6). The number of recorded seizures and patterns observed is indicated in Table 1. The rhythmic discharge of high-amplitude spikes predominated in the hippocampus. The mean duration of this spiking was 17.5 s (range, 7–25 s). This pattern of spiking was followed by a tonic discharge affecting the medial temporal structures: the involvement of the EC was simultaneous with the hippocampus in five cases or slightly delayed (patient P4). Spectral analysis showed that the mean frequency of the discharge at the onset of the tonic discharge was 13.6 Hz (range, 6–30 Hz) in the hippocampus and 13.8 (range, 7–32) in the EC.
Table 1. Patient data
Type of epilepsy
Number of spontaneous recorded seizures
FS at 8 mo (encephalitis)
Bilateral HS Left > Right
Type 2 (3/4)
FS at l0 mo
Type 1 (4/4)
FS at 15 mo
Right HS Lelt frontal arachnoidian cyst
Type 1 (5/6)
FS at 2 y
Type 2 (3/3)
Encephalitis at age 7
Type 2 (7/7)
Type 1 (6/7)
FS from 1 to 3 y
Type I (8/8)
FS at 11 mo
Type I (3/3)
FS from 1 to 2 y Familial history of epilepsy
Bilateral HS Left > Right
Type 1 (3/3)
Type 2 (3/3)
A seizure onset corresponding to pattern 2 was observed in seizures from five patients (cases 7–11; Table 1). In this situation, the EC and the hippocampus were simultaneously involved at the onset of the tonic discharge, without preictal spiking. Time-frequency analysis of discharges disclosed a mean frequency of 19.6 Hz (range, 7–35 Hz) in the hippocampus and 20.3 Hz (range, 10–36 Hz) in the EC (Table 2). Frequency distributions were not statistically different in the two pattern groups (p = 0.45, Mann–Whitney test).
Table 2. Ictal patterns in the 11 patients and results of nonlinear correlation analysis (h2) between hippocampus (Hip) and entorhinal cortex (EC)
Background h2 values mean (SD)
Mean values of h2 max (SD)
Ipsilateral EC atrophy (z-score,volumetry)
The background interictal mean values (interictal period) and the mean of maximal h2XY values reached at seizure onset. These values were analyzed from three seizures in each patient. The leader structure was determined according to the D index (see Methods). EC volumes ipsilateral to the epileptic side also are indicated with the corresponding Z-score values.
Analysis of signal interdependencies between hippocampus and entorhinal cortex
Table 2 reports the main results observed in the two groups (patterns 1 and 2). Unidirectional coupling was decided for values of D≥0.2 when h2 values were statistically different from those computed over the interictal period (>2 standard deviations). These thresholds were set according to previous simulation experiments (29).
In comparison with values measured on interictal background activity, the onset of seizures was marked by an increase in nonlinear correlation between signals from hippocampus and EC in seizures. Table 2 indicates the values measured during the interictal period and the maximal values obtained at seizure onset for three seizures in each patient (mean of the maximal values and SD)
Figures 4 and 5 show typical examples of correlation studies in pattern 1 and 2 seizures. The hippocampus was found to be the leader in six of 11 cases and the EC in four cases. In one case, bidirectional coupling was found, and the leader structure was therefore not determined. The EC was the leader in the coupling in four of five of pattern 2 seizures and never in pattern 1 seizures (p = 0.015, Fisher's exact test).
Entorhinal cortex and hippocampus volumetry
The values obtained in patients and controls are indicated in Table 3. Taken as a whole, the mean value of the hippocampal volume was 3,280 mm3 within the epileptic side in patients and significantly lower in comparison with the control group (5,250 right and 5,008 left; p = 0.005).
Table 3. Volumetric values (mean and standard deviation) of the entorhinal cortex (EC) and the hippocampus (Hip) after normalization by using the intracranial volume measurements method
R-TLE, right mesial temporal lobe epilepsy; L-TLE, left mesial temporal lobe epilepsy.
TLE patients (11)
Figure 6 shows the z-score representation of the values of hippocampal volumes ispilateral and controlateral to the epileptic side for the 11 patients. All the patients had a significant reduction (z-score <2) of ipsilateral hippocampal volumes; seven had a bilateral reduction in hippocampal volumes in comparison with controls.
Patients with left TLE had significant reduction of left hippocampal volumes in comparison with controls (p = 0.017; Mann–Whitney) as well as reduction of right hippocampal volumes (p = 0.041; Mann–Whitney; Table 3).
Patients with right TLE had significant reduction of right hippocampal volumes (p = 0.036) but normal volumes for left hippocampus (p = 0.33; Table 3).
Entorhinal cortex volumes
The values obtained in patients and controls are indicated in Table 3. Taken as a whole, the volume of the EC in TLE patients in the epileptic side was lower (1,187 ± 242 mm3) than in controls (mean, 1,580 ± 184 mm3), representing a 25% reduction in EC volumes (p = 0.005). Volume measured from the contralateral side was also reduced in comparison with control values (p = 0.03).
Figure 7 shows the z-score representation of the values of entorhinal volumes both ipsilateral and controlateral to the epileptic side for the 11 patients. Seven (63%) patients had significant reduction (z-score <2) of ipsilateral entorhinal volumes, and four (36%) had bilateral reduction in entorhinal volumes.
Patients with left TLE had significant reduction in left (p = 0.0005) and right (p = 0.023) EC volumes. Patients with right TLE did not have statistical reduction in left EC (p = 0.22) or right EC (p = 0.053), although this last result almost reaches significance (Table 3).
Correlations between EC volumes and electrophysiology
The mean ipsilateral volume of the EC in patients that had pattern 2 seizures was found to be lower (1,095 mm3) than mean EC volume in patients with pattern 1 seizures (1,217 mm3). However, this difference did not reach statistical significance (p = 0.34). In the seven patients with EC atrophy, nonlinear regression analysis demonstrated a leader role of the EC in four (57%). In patients without atrophy of the EC (four patients), this analysis showed that the EC was never the leader. These results suggest that EC atrophy was more frequent in those patients in whom the EC might play a leader role in initiating seizures.
To estimate the possible relation between EC atrophy and the involvement of the EC at seizure onset, a linear regression analysis was performed between h2 values and EC volumes. Figure 8a shows that ipsilateral EC volumes and h2 values are correlated (r= 0.76) and that EC volumes decrease for increasing h2 values. No significant correlation was found between h2 values and hippocampal volume (Fig. 8b).
The main result of this study is the demonstration of a relation between EC atrophy and the role of this structure in MTLE seizure genesis. We found a significant correlation between the degree of coupling between EC and hippocampal signals at seizure onset and the degree of atrophy of the EC. In addition, in those patients with normal EC volume, the EC was never the leader structure in EC–hippocampal coupling. Several aspects of these results are discussed later.
The study of the relation between electrophysiologic signals generated at seizure onset in the hippocampus and EC is in line with previous work showing that increased correlation between structures involved in the epileptogenic zone can be observed at seizure onset (6,29).
Our results suggest that the EC is involved in different ways depending on the electrophysiologic organization of ictal onset. In pattern 1 seizures (periodic spiking preceding the rapid discharge), we found a coupling between EC and the hippocampus in which the hippocampus was always the leader. Pattern 2 seizures, in which tonic discharge simultaneously affected the hippocampus and EC at the onset, appeared to be initiated by interactions between the EC and hippocampus in which the EC was found frequently to play a leader role. Our results were consistent with those of the previous study of Spencer and Spencer (4), who also demonstrated an important role of EC in some human TLE seizures. They found that the EC was mainly involved when tonic discharges occurred and suggested that the so-called “preictal” spiking (“periodic preictal spiking”) was not related to EC activity but rather was of hippocampal origin. Results also are in accordance with those observed in a chronic in vivo model of epilepsy (unilateral intrahippocampal injections of kainic acid in rats), in which hypersynchronous electrographic ictal-onset patterns (similar to pattern 1) and low-voltage fast activity at seizure onset (similar to human pattern 2) were observed. As in human MTLE, hypersynchronous ictal onsets originated predominantly in the hippocampus, whereas low-voltage fast ictal onsets more often involved extrahippocampal structures (35).
Volumetric analysis and relation with electrophysiology
Several recent studies quantified the EC volume in patients with MTLE (15,17–19,21). All these studies demonstrated a statistical reduction in the volume of the EC ipsilateral to the epileptic side. Whereas a correlation existed between entorhinal and hippocampal volume loss in one study (18), another study showed that decrease in EC volume can be observed in patients with normal hippocampal volumes (17). No association has so far been found with clinically or pathologically identified causes of epilepsy, duration of epilepsy, or age at onset of epilepsy (18).
In keeping with these recent studies, we also found that mean EC volume ipsilateral to the epileptic side was reduced and that a significant reduction (<2 DS) of EC volume was observed in 63% of patients who had hippocampal atrophy. This reduction was more marked in patients with left than with right MTLE, but this could be due to the low number of patients with right MTLE in this study.
No previous study correlated the relation between atrophy of the EC and its functional role as determined by electrophysiology. Our results are the first to demonstrate a relation between EC volumes and the involvement of this structure in MTLE seizure genesis. The EC was found to be the leader in the hippocampal–EC couplings only in those patients with EC atrophy, a situation more often associated with pattern 2 seizures. The degree of EC atrophy was found to be correlated with the degree of involvement of the EC in the networks that generate seizures.
These results suggest that the more atrophic the EC, the more it is likely to interact with the hippocampus at seizure onset. This result may be interpreted in the light of studies dealing with the entorhinal pathology in TLE. The most robust histologic correlation between entorhinal atrophy and histopathology is a preferential cell loss in superficial layers of the EC. Animal studies with experimental models of MTLE found an entorhinal cell loss, preferentially located in the pyramidal cells of the layer III (36). Such a layer III cell loss has also been described in human TLE (37). The pyramidal layer III of the EC sends projection to area CA1 of the hippocampus, and this projection includes excitation of inhibitory interneuron (38). Thus it has been proposed [review in (39)] that the cell loss in entorhinal layer III could have a disinhibitory effect on CA1 hippocampal region and promote a reverbatory excitatory “dialogue” between EC and the different hippocampal areas.
Whatever the pathophysiologic mechanisms involved, results of this study indicate that volumetric analysis of the EC may help to predict the epileptogenic role of this structure in MTLE. The study adds futher evidence that the organization of the epileptogenic zone in MTLE can not readily be reduced to a single “focus” and that a more complex network configuration is probably responsible for the initiation of seizure activity.
However, studies including a larger population of patients are mandatory to confirm these results and to establish whether differences may be observed in these different subtypes of MTLE, particularly with regard to ultimate surgical outcome.