- Top of page
Summary: Purpose: Two subtypes of temporal lobe epilepsy (TLE) can be defined through clinical observations and analysis of hippocampal tissue resected during surgical procedures for intractable TLE: (a) mesial temporal sclerosis (MTS), which is characterized by extensive changes to the hippocampus and good surgical outcome; and (b) paradoxical temporal lobe epilepsy (PTLE), which is characterized by minimal cell loss and comparatively poorer surgical outcome. Patients in both subtypes have seizures that appear to begin in the medial temporal lobe, but documented differences in substrate and outcome between these subtypes has defined a need to distinguish MTS and PTLE patients before surgery. This report describes a retrospective study to investigate the feasibility of doing so during intracranial monitoring.
Methods: Background EEG epochs, 5 min in duration, were recorded from the anterior hippocampus in 14 (10 MTS and four PTLE) patients with consistent localization of seizure onset to medial temporal structures. The power spectral density (PSD) of the EEG epochs was calculated by a Fourier spectral estimator, and the total signal power and power of the delta, theta, alpha, beta, and gamma frequency bands were submitted to group-to-group comparison.
Results: Spectral peaks were observed in the delta band in all PSD estimates and in the theta band in nine of 14 (seven MTS, two PTLE) estimates. The MTS and PTLE subtypes could be distinguished by the total signal power and delta band power. These power measurements were greater in the PTLE subtype.
Conclusions: Both delta and theta spectral components are present in hippocampal background EEGs recorded from patients with TLE. The results indicate that group differences exist in spectral measures of background hippocampal signals recorded from MTS and PTLE subtypes. This suggests both that substrate differences in cellular composition and connectivity are reflected in hippocampal background EEGs and that spectral measurements of these signals may hold promise for tests to identify the group membership of individual patients.
When epilepsy is uncontrolled by traditional anticonvulsant therapy, surgical resection of the region generating the seizures may be performed. The most common uncontrolled epilepsy treated surgically is of medial temporal origin. Despite similar EEG seizure localization, five subtypes of medial temporal lobe epilepsy have been described by the Yale Epilepsy Program through clinical observations and analysis of hippocampal tissue resected during surgical procedures for intractable epilepsy (3–5). These subtypes are:
Paradoxical temporal lobe epilepsy (PTLE): Characterized by a uniform (∼25%) loss of principal neurons in all CA fields and dentate granule cells, no sprouting of immunoreactive fibers, and relatively low hyperexcitability of dentate granule cells (6)
. In addition, there is no mass lesion or identifiable etiology in this patient subtype.
Mass associated temporal lobe epilepsy (MaTLE): These are patients with an extrahippocampal temporal lobe mass lesion (low-grade gliomas, hamartomas, and vascular lesions). The features observed in the hippocampal tissue in this subtype are similar to those observed in the PTLE subtype.
Mesial temporal lobe epilepsy (MTLE): There is ≥50% loss of principal neurons in all CA fields and dentate granule cells, and a selective loss of hilar peptidergic neurons in this subtype. Further, there is reorganization of the dentate with sprouting of dynorphin, somatostatin, neuropeptide Y, and substance P reactive fibers, and cellular electrophysiology measurements demonstrate the highest levels of hyperexcitability in dentate granule cells of all TLE subtypes.
MTLE/Dynorphin negative (MTLE/DYN−): The features observed in the hippocampal tissue in this subtype are similar to those observed in the MTLE subtype, without sprouting of dynorphin-reactive fibers (7)
and with decreased cellular hyperexcitability in comparison with the MTLE subtype.
CA1 cell loss: The features observed in this subtype are similar to those observed in the MTLE subtype in the loss of neurons in the CA1 region, but dissimilar in showing no sprouting of immunoreactive fibers or selective interneuron loss in the area dentata, and decreased hyperexcitability of dentate granule cells in comparison with the MTLE subtype.
The relatively normal hippocampus and poorer surgical outcome (8,9) in the PTLE subtype has prompted the suggestion that seizures may not start in the hippocampus in this subtype of TLE; that in PTLE patients, seizure onset may occur outside the hippocampus and spread to involve it during early stages of seizures (5).
This study was conducted to investigate the feasibility of identifying PTLE patients during intracranial monitoring. To this end, patients from the MTLE, MTLE/DYN−, and CA1 cell loss subtypes were pooled, and this collective group, characterized by ≥50% loss of principal neurons in Ammon's horn, and varying amounts of sprouting and cellular hyperexcitability was defined as mesial temporal sclerosis (MTS). MaTLE patients were not included in the study because they can be identified during presurgical evaluation.
Previous intracranial EEG studies with similar patients have demonstrated correlations between seizure-onset intracranial EEG patterns and neuronal and glial counts and cellular electrophysiologic measures (10–13). The present study was limited to background EEG activity. Because of the profound differences in the cellular composition and connectivity of PTLE and MTS hippocampi, it was supposed there would be a measurable difference between hippocampal background EEGs recorded in these subtypes. Previous studies of hippocampal background EEGs have demonstrated task-related changes of the power spectral density (PSD) (14–18); and a difference in the PSD of EEGs recorded from epileptogenic and nonepileptogenic hippocampi (16). Previous studies have failed to establish clearly the presence of prominent rhythmic components in hippocampal background EEGs, however. It was not clear from the literature what form the PSD of these signals would take, or what differences could be expected between the PSD of signals recorded from PTLE and MTS patients. A very general possibility was that there would be comparatively lower signal power in MTS patients because of the lower neuronal counts. It also was possible, however, that the greater excitability in MTS would contribute, albeit in an unknown manner, to greater signal power.
- Top of page
The Yale–New Haven Hospital intracranial EEG archive includes data from 1991 to the present date. The following criteria were used to identify patients from this archive: (a) parasagittal placement of a depth electrode sampling the anterior body of the epileptogenic hippocampus, (b) seizure onset consistently localized to medial temporal structures, (c) anteromedial temporal resection (AMTR) (19), and (d) postresection analysis including quantitative neuronal counts, immunohistochemistry, and cellular electrophysiology. Both parasagittal and orthogonal sampling of the hippocampus have been used in the past at Yale–New Haven Hospital. Parasagittal placement of the hippocampal electrode was used as a criterion for this study, as this resulted in the identification of a larger number of patients. A total of 27 (four PTLE, 18 MTLE, two MTLE/DYN−, and three CA1 cell loss) patients met the stated criteria. All PTLE, MTLE/DYN−, and CA1 cell loss patients identified were selected for the study. Five MTLE patients were selected for this study by alphabetizing the list of MTLE patients and picking the first five entries. Seven of the 14 selected patients were women (three MTS and four PTLE), and seven of the patients underwent a left anterotemporal resection (six MTS and one PTLE). The average age at surgery was 33 years (MTS, 34 years, and PTLE, 32 years).
The electrodes used for the intracranial EEG recordings were multicontact flexible depth electrodes introduced under stereotactic guidance from the occipital lobe and targeting the anterior hippocampus (20). The placement of the electrodes was verified with postimplantation magnetic resonance imaging (MRI). A contact in the anterior body of the hippocampus proximate to the area where samples were taken for neuronal and glial counts, immunohistochemistry, and cellular electrophysiology was selected from each patient for analysis. The selected contact was either the seizure onset site or was involved in seizures at a very early stage. Background EEG epochs, 5 min in duration, where the patients appeared to be in a quiet awake state and were presumably watching television, were identified from split screen CCTV-EEG recordings obtained before the first recorded seizure during intracranial EEG monitoring. This patient state was selected because it was common. The average time from implantation was 5 days (five MTS and four PTLE). The intracranial EEGs were measured with reference to a scalp electrode contact placed on the contralateral mastoid process. The signals were digitized at 200 samples per second and stored with commercially available EEG recording equipment (Telefactor Corporation, Conshohocken, PA, U.S.A.). The signals were low-pass filtered from 0 to 50 Hz.
The EEG epochs were divided into 300 nonoverlapping 1-s segments. Segments containing interictal activity and muscle and movement artifact introduced through the reference channel were visually identified from the continuous EEG records and deleted. After the deletion of these EEG segments, the total number of data segments, in the 14 patients, ranged between 251 and 299. The first 251 interictal activity and artifact-free EEG segments from each patient were selected for analysis. The signal segments were mean-deleted, weighted with a Hann window, and the PSD of the EEGs was estimated by the Welch spectral estimator. The power in traditional scalp defined delta (0–4 Hz), theta (4–8 Hz), alpha (8–12 Hz), beta (12–30 Hz), and gamma (35–45 Hz) bands were measured from the PSD estimates, and these measures and the total signal power were log transformed and subjected to an analysis of variance (ANOVA).
- Top of page
The gross features of the hippocampal background EEG PSD were similar in the two subtypes of TLE. All PSD estimates contained a distinct peak in the delta band, and a majority (nine of 14, seven MTS and two PTLE) contained a second peak in the theta band. With one exception (MTS) in which a double peak was observed in the delta band, no other peaks were observed in the PSD estimates. Further, with two exceptions (one each MTS and PTLE), the peak in the theta band was smaller in magnitude than the peak in the delta band. Example PSDs from PTLE and MTS patients, with spectral peaks in the delta and theta bands, are displayed in Fig. 1. The theta peak often resulted from robust activity in this frequency band over the entire recording duration. Delta and theta spectral peaks also could be observed at other recording locations in the ipsilateral and contralateral hippocampus. Example PSDs of background EEGs recorded from multiple points of the contralateral hippocampus in a patient with unilateral epilepsy are displayed in Fig. 2.
Figure 1. Example of power spectral density (PSD) estimates of hippocampal background EEGs recorded from a mesial temporal sclerosis (MTS) and a paradoxical temporal lobe epilepsy (PTLE) patient. These examples demonstrate three commonly observed characteristics of the PSD of hippocampal background EEGs: peaks in the delta and theta frequency bands, and a steady decrease of signal power from the lower frequencies to the higher frequencies. Note, in these examples, PTLE signal power is greater than MTS signal power at all frequencies. Note also the difference in the peak frequency of the peak in the theta frequency band. The peak frequency of spectral peaks in the theta frequency band ranged between 4 and 8 Hz.
Download figure to PowerPoint
Figure 2. An example of delta and theta activity recorded from the nonepileptogenic hippocampus of a patient with medial temporal lobe epilepsy. Contacts 2 and 3 of the electrode lie anterior to the hippocampus, and contacts 4–7 sample the long axis of the hippocampus. Delta and theta spectral power varies independently along the length of the hippocampus; delta power is maximal at contact 5, whereas theta power is maximal at contact 7.
Download figure to PowerPoint
The average power measurements (power in the delta, theta, alpha, beta, and gamma bands, and total signal power) are shown in Table 1. A separation of MTS and PTLE patient groups was obtained with total signal power (p < 0.01) and delta power (p < 0.01), and a strong difference was obtained with theta power (p = 0.059). All three of these comparisons, total signal power and signal power in the delta and theta bands, were significant when controlled for sex (women, four PTLE vs. three MTS; total power, p < 0.01; delta power, p < 0.01; theta power, p < 0.05) and side of surgery (left AMTR, three PTLE vs. four MTS; total power, p < 0.01; delta power, p < 0.01; theta power, p < 0.05). Other comparisons were not possible because of a lack of data samples.
Table 1. Hippocampal background EEG power