Enhanced EEG functional connectivity in mesial temporal lobe epilepsy

Transcription

1 Epilepsy Research (2008) 81, journal homepage: Enhanced EEG functional connectivity in mesial temporal lobe epilepsy Gaelle Bettus a, Fabrice Wendling e, Maxime Guye a,b,c, Luc Valton c, Jean Régis a,b,d, Patrick Chauvel a,b,c, Fabrice Bartolomei a,b,c, a INSERM, U751, Laboratoire de Neurophysiologie et Neuropsychologie, Marseille F-13005, France b Université delaméditerranée, Faculté demédecine, Marseille F-13005, France c CHU Timone, Service de Neurophysiologie Clinique, Assistance Publique des Hôpitaux de Marseille, Marseille F-13005, France d CHU Timone, Service de Neurochirurgie Fonctionnelle et Stéréotaxique, Assistance Publique des Hôpitaux de Marseille, Marseille F-13005, France e INSERM, U642, Rennes F-35000, France Received 16 December 2007; received in revised form 18 April 2008; accepted 22 April 2008 Available online 10 June 2008 KEYWORDS EEG; Temporal lobe epilepsy; Functional connectivity; Interictal; Synchrony Summary Purpose: To analyze and compare spectral properties and interdependencies of intracerebral EEG signals recorded during interictal periods from mesial temporal lobe structures in two groups of epileptic patients defined according to the involvement of these structures in the epileptogenic zone (EZ). Methods: Interictal EEG activity in mesial temporal lobe (MTL) structures (hippocampus, entorhinal cortex and amygdala) was obtained from intracerebral recordings performed in 21 patients with drug-resistant mesial temporal lobe epilepsy (MTLE group). This group was compared with a control group of patients (non-mtle group) in whom depth-eeg recordings of MTL show that seizures did not start from the MTL. Comparison criteria were based on spectral properties and statistical coupling (nonlinear correlation coefficient h 2 ) of MTL signals. Results: Power spectral density analysis showed a significant decrease in the theta frequency sub-band (p = 0.01) in the MTLE group. Nonlinear correlation (h 2 ) values were found to be higher in the MTLE group than in the NMTLE group (p = ). This effect was significant for theta, alpha, beta and gamma frequencies. Correlation values were not correlated with the frequency of interictal spikes (IS) and significant differences between groups were still measureable even when spikes were suppressed from analyzed EEG periods. Discussion: This study shows that, during the interictal state, the EZ in MTLE is characterized by a decrease of oscillations in the theta sub-band and by a general increase of signal interdependencies. This last finding suggests that the EZ is characterized by network of neuronal assemblies with a reinforced functional connectivity Elsevier B.V. All rights reserved. Corresponding author at: Service de Neurophysiologie Clinique, CHU Timone-264 Rue st Pierre, F Marseille, France. address: fabrice.bartolomei@ap-hm.fr (F. Bartolomei) /$ see front matter 2008 Elsevier B.V. All rights reserved. doi: /j.eplepsyres

2 Enhanced EEG functional connectivity in mesial temporal lobe epilepsy 59 Introduction Focal (or partial) epilepsies are characterized by recurrent seizures generated in an abnormal region of the brain, the epileptogenic zone (EZ). Approximately 1/3 of cases are resistant to antiepileptic drugs. In this situation, surgical resection of the EZ is the only therapeutic option able to suppress seizures or, at least, to significantly reduce their frequency. The localization and the definition of the EZ are therefore crucial issues in epileptology and are addressed through detailed analysis of anatomo-functional data acquired in epileptic patients during pre-surgical evaluation. Among investigation methods used during this evaluation, intracerebral exploration remains the only way to directly record the electrophysiological activity (depth- EEG) from brain structures and to formulate hypotheses about their potential involvement in epileptogenic processes. In this context, a large number of studies have been dedicated to the analysis of depth-eeg signals. Based of the estimation of interdependences (i.e. statistical coupling) between signals recorded from distinct sites, some reports have demonstrated that the areas involved in the generation of seizures (defining the EZ) are characterized by synchronous oscillations at seizure onset (Bartolomei et al., 2005, 2001b, 2004, 1999; Duckrow and Spencer, 1992; Gotman and Levtova, 1996; Le Van Quyen et al., 1998; Lieb et al., 1987). The synchronization of activities recorded from brain structures is therefore an important phenomenon that may be used for identifying epileptogenic networks (i.e. promoting the initiation of seizures). Other studies based on nonlinear associations in multivariate signals (Guye et al., 2006) have also reported that long distance functional connectivity is dramatically altered during seizures, or indicated that the topology of networks changes as ictal activity develops (Ponten et al., 2007). In contrast, knowledge about the properties of electrophysiological signals recorded from the EZ during interictal periods remains elusive. It is usually observed that brain structures involved in the EZ produce abnormal transient electrophysiological (interictal spikes). However, the spectral characteristics of depth-eeg signals as well as the properties of the functional connectivity of involved networks are poorly described. Generally speaking, the analysis of statistical couplings between signals generated from distant sites has been proposed as an approach to assess this functional connectivity. For instance, in cognitive studies, synchronization processes in the theta frequency band has been associated with memory processes (Sarnthein et al., 1998; Stam et al., 2002a). Synchronization in the gamma band has been shown to relate to the representation of complex information in conscious perception (Csibra et al., 2000; Micheloyannis et al., 2003; Rodriguez et al., 1999; Tallon-Baudry and Bertrand, 1999) and also to memory processes (Fell et al., 2003; Summerfield and Mangels, 2005; Tallon-Baudry et al., 2001). Functional connectivity may also serve as a potentially useful marker of brain disease (Cordes et al., 2001; Lowe et al., 1998; Salvador et al., 2005; Stam et al., 2002b; Uhlhaas and Singer, 2006). It is indeed expected that diseases inducing changes in synaptic efficiency may alter the communication within and between neuronal networks and thus, induce neurological disturbances. In the field of epilepsy, two studies have recently shown an increase of the local synchrony calculated from monochannel signals (using the autocorrelation function) (Monto et al., 2006) or from multichannel signals (using the mean phase coherence method) (Schevon et al., 2007), in the vicinity of the epileptogenic zone. In these studies, recordings were performed from a neocortical focus using intracranial grids and led authors to suggest that functional connectivity could be altered during the interictal state of partial epilepsies. In the present study, we investigated the spectral properties and the possible changes in functional connectivity affecting the mesial temporal region in patients with MTLE. Intracerebral EEG signals were collected in patients with MTLE undergoing presurgical evaluation. They were recorded from the mesial temporal structures using stereoelectroencephalography (SEEG). Moreover, in order to confront quantities measured in MTLE with those obtained under normal condition, we also studied depth-eeg signals recorded from mesial structures in patients in whom the EZ was localized outside the mesial temporal lobe. Methods Patients Patients undergoing pre-surgical evaluation of drug-resistant epilepsy were retrospectively selected in the database of our epilepsy unit (CHU Timone, Marseille) for the present study. All patients had a comprehensive evaluation including detailed history and neurological examination, neuropsychological testing, conventional presurgical MRI, surface EEG and depth-eeg recording (stereo-eeg using intracerebral electrodes implanted according to a stereotactic approach) as previously reported (Bartolomei et al., 2002; Guye et al., 2006). Patients were selected for the present study if they satisfied the following criteria: (i) a clearly localized, unilateral epileptogenic zone defined by the regions primarily involved in seizures (generally characterized by a fast ictal discharge at onset) and (ii) at least two mesial temporal lobe structures explored by intracerebral electrodes. Patients were divided into two groups depending on the participation of the mesial regions at seizure onset: (i) The first group, denoted by the acronym MTLE, included patients (n = 21) with seizures that initially involved mesial temporal lobe. (ii) The second group, denoted by the acronym NMTLE, included patients (n = 14) with a non-mesial temporal lobe epilepsy. These patients had frontal lobe epilepsy (n = 5), lateral temporal lobe epilepsy (n = 4) and occipital lobe epilepsy (n = 5). Table 1 gives a summary of the patient s data. SEEG recordings Depth-EEG recordings were performed according to the stereoelectroencephalographic approach in which intracerebral multiple contact electrodes (10 15 contacts, length: 2 mm, diameter: 0.8 mm, 1.5 mm apart) are placed intracranially according to Talairach s stereotactic method (Talairach et al., 1974). The positioning of electrodes was established in each patient based upon hypotheses about the localization of the epileptogenic zone formulated from available non-invasive information. The implantation

3 Table 1 Clinical data of patients included in the study Patients Gender Age (year) Epilepsy duration (year) Aetiology (MRI) EZ localization (SEEG) Side Seizure outcome (Engel s class) Explored MTL regions MTLE group Ar F HS M R I A, Ha, Hp, EC Ba M Post-traumatic lesion M R I A, Ha, Hp, EC (Temporopolar cortex) Bi F HS M L I A, Ha, Hp, EC Br F Normal M L II A, Ha, Hp, EC Ca M HS M L I A, Ha, Hp, EC Ch M 34 8 Cavernoma M R NO A, Ha, Hp, EC Cl M HS M L II A, Ha, Hp, EC Cu M Normal M R II A, Ha, Hp, EC Fl M MT CD M R II A, Ha, Hp, EC Gon F HS M L I A, Ha, Hp, EC Gos F HS M L I A, Ha, Hp, EC Gu M HS M L I A, Ha, Hp, EC Maz M HS M L I A, Ha, Hp, EC Mar M HS M L I A, Ha, Hp, EC Pa F HS M R I A, Ha, Hp, EC Ph F Normal M R I A, Ha, Hp, EC Tal M HS M R NO A, Ha, Hp, EC Due F HS M R I A, Ha, Hp, EC Gar M HS M R I A, Ha, Hp, EC Sc F SH M L I A, Ha, Hp, EC Ta M Normal M L II A, Ha, Hp, EC NMTLE group Be F 15 7 Occipital CD Occipital R I Ha, Hp, EC Bu M Occipital CD Occipital R II A, Ha, EC Do F 14 8 Normal Occipital R I Ha, EC Fe M Normal Orbito-frontal L I A, Ha, Hp, EC Ga M Normal STG R I A, Ha, Hp, EC Jo M 20 3 Normal Prefrontal mesial R-L A, EC La F STG CD STG R I A, Ha, Hp, EC Me F STG CD STG R I A, Ha, Hp, EC Mi F Normal Occipital R II Ha, Hp, EC Mo F STG atrophy STG R I A, Ha, Hp Ro F Occipital CD Internal occipital R III (gamma knife) Ha, Hp, EC Tal M Normal Prefrontal R I A, EC Buf F Normal Prefrontal R I A, Ha, EC To M Post traumatic lesions (orbitofrontal cortex) Mesial prefrontal L I A, EC MTLE: patients with mesial temporal lobe seizures; NMTLE: patients with non-mesial temporal lobe seizures; F: female; M: male; CD: cortical dysplasia, STG: superior temporal gyrus; HS: hippocampal sclerosis; M: mesial, R: right; L: left, EZ: epileptogenic zone, MTL: mesial temporal lobe, A: amygdala, Ha: anterior hippocampus, Hp: posterior hippocampus, EC: entorhinal cortex, NO: not operated. 60 G. Bettus et al.

4 Enhanced EEG functional connectivity in mesial temporal lobe epilepsy 61 Figure 1 (a) Schematic diagram of SEEG electrodes placement on a lateral view of Talairach s basic referential system in a patient with MTLE. A: Electrode exploring the amygdala (medial leads) and the anterior part of the middle temporal gyrus (MTG) (lateral leads); B: electrode exploring the anterior hippocampus (medial leads) and the mid part of MTG (lateral leads); TB: electrode exploring the entorhinal cortex (internal contacts) and the anterior part of the inferior temporal gyrus (external contacts); C: electrode exploring the posterior hippocampus (medial leads) and the posterior part of MTG (lateral leads); T: electrode exploring the insula (medial leads) and the anterior part of the superior temporal gyrus (STG) (lateral leads) O: electrode exploring the temporo-occipital junction, P: parietal electrode. (b) Reconstruction of the trajectory of the electrode named TB and reaching the entorhinal cortex. The contacts #1 and #2 of the electrodes are located in the entorhinal cortex and are used for signal analysis. (c) Bipolar interictal signals (raw signals with no filter) recorded from the most internal contacts of electrodes TB (1 2, entorhinal cortex), A (1 2, amygdala), B (1 2, anterior hippocampus) and C (1 2, posterior hippocampus) were used in this study. In MTLE these regions are the seat of numerous interictal spikes. On the right side of the figure, the same signal is represented after removal of the interictal spikes discharge. Note that there is no artefact generated by the procedure. Interval between each vertical line represents 1 s. Amplitude is 20 V/mm for the four signals. accuracy was per-operatively controlled by telemetric X-ray imaging. A post-operative computerized tomography (CT) scan without contrast medium was then used to verify both the absence of bleeding and the location of each contact. After removal of intracerebral electrodes, a MRI was performed, permitting visualization of the trajectory of each electrode. Finally, CT-scan/MRI data fusion was achieved to anatomically locate each contact along the electrode trajectory. An example of the SEEG implantation scheme is illustrated in Fig. 1a. Signals were recorded on a 128 channel Deltamed TM system. They were sampled at 256 Hz and recorded on a hard disk (16 bits/sample) using no digital filter. The acquisition system includes two hardware filters. The first one is a high-pass filter with cut-off frequency equal to 0.16 Hz at 3 db. It removes very slow variations that may contaminate the baseline. The second one is a 1st order low-pass anti-aliasing filter (cut-off frequency equal to 97 Hz at 3 db). SEEG was carried out as part of our patients normal clinical care, and patients were informed that their data might be used for research purposes. Analyzed SEEG recordings corresponded to periods of time in which patients were at rest but in a waking state (30 min, eyesclosed, resting state, recording performed between 07:00 and 12:30). It is noteworthy that studied interictal recordings were temporally distant from the preceding and the following seizure by 2 h, at least. In addition, recordings were chosen at least two days after the electrode implantation surgical procedure in order to limit possible effect of the general anesthesia. Signal analysis Estimation of the relative power of SEEG signals in frequency sub-bands The power spectral density (PSD) was estimated from time-series signals using the periodogram method (also known as the Welch s method (Welch, 1967)) which consists in dividing the signal into successive blocks and in averaging squared-magnitude discrete Fourier transforms (DFTs) computed on each signal block. DFTs were estimated using a standard Fast Fourier Transform (FFT) algorithm

5 62 G. Bettus et al. on 5 s blocks (1280 samples). In order to decrease the variance of the PSD estimate, an overlapping of 50% (2.5 s) between blocks was chosen. The power of SEEG signals in classical EEG frequency sub-bands delta ( Hz), theta ( Hz), alpha ( Hz), beta ( Hz) and gamma (24 97 Hz) was computed by integrating the PSD over the corresponding sub-band: P B = f2 f 1 S(f)df where S(f) denotes the PSD and P B denote the signal power in frequency sub-band B, ranging from frequency f 1 to frequency f 2. In order to compare values measured on signals recorded from different patients, P B was normalized by dividing by the total signal power: f2 S(f)df f RP B = 1 fs/2 S(f)df 0 where f s is the sampling frequency. Therefore, RP B (0 RP B 1) is an estimate of the Relative Power of the signal in frequency sub-band B where relative means with respect to the total signal power. As an example, if f 1 = 3.4 and f 2 = 7.4 in the above equation and if the result is RP B = 0.3 then it can be said that 30% of the total signal power is due to the contribution of oscillations in the theta sub-band. In the present study, for each signal recorded from each mesial temporal lobe structure, power spectral densities were calculated by averaging discrete Fourier transforms over the 30 min period of time selected in each patient. Therefore, PSDs were calculated by averaging 720 DFTs (30 60/2.5) as computed on 5 s blocks overlapping by 50%. The relative power in each frequency sub-band (RP B ) was then estimated from PSDs (Eq. (1)). Consequently, for each patient and for each recorded structure (Table 1), we obtained five values indicating the relative contribution to the total signal power of oscillations taking place into the five considered sub-bands (delta, theta, alpha, beta and gamma). Estimation of nonlinear correlations between SEEG signals In this study, statistical couplings between SEEG signals recorded from distinct mesial structures were estimated using nonlinear regression analysis. This method, introduced by Pijn and Lopes Da Silva (1993) in the field of EEG analysis, has been extended (Wendling et al., 2001) and largely used by our group to interpret the degree and the direction of functional couplings in networks of neuronal populations from which SEEG signals are recorded (Bartolomei et al., 2001a). Nonlinear regression analysis is aimed at quantifying the relationship (or association) between two SEEG signals X(t) and Y(t) recorded from two neuronal populations P x and P y, without making any assumption on nature (linear or nonlinear) of this relationship. The method provides two quantities, namely the nonlinear correlation coefficient (h 2 xy ) and the time delay ( xy), respectively, calculated from signals X(t) and Y(t). The nonlinear correlation coefficient h 2 xy is obtained by considering the amplitude y of signal Y(t + ) as a perturbed function of the amplitude x of signal X(t) (i.e. the conditional mean of Y(t + ) given X(t)=x). The variance of the perturbation formally corresponds to the conditional variance of Y(t + ) given X(t) (i.e. the reduction of variance of Y(t + ) obtained by predicting y values from x values). This conditional variance can be estimated from the construction of a piecewise linear regression curve Y(t + )=h(x(t)). For convenience, it is also normalized by the variance of Y(t + ) and the normalized value is subtracted from 1 such that values are comprised between 0 (Y is independent of X) (1) and1(x and Y are linearly or nonlinearly dependent) (Wendling et al., 2001). For the sake of simplicity, the nonlinear correlation coefficient h 2 xy will be simply denoted by h2 in the following. In the present study, h 2 values were computed on both broadband signals (providing a global estimation of nonlinear interdependencies) and on signals filtered in the EEG sub-bands described. Hamming finite impulse response (FIR) filters were chosen for their linear phase that is more appropriate for the computation of correlations in selected sub-bands. Filter order was equal to 256 (corresponding to one second duration at a sampling rate of 256 Hz). Interdependencies between artifact-free signals from at least two areas of interest (amygdala, anterior hippocampus, entorhinal cortex and posterior hippocampus) were estimated over the 30 min period selected for each patient of the two groups (MTLE and NMTLE) (Fig. 2). The localization of electrode contacts was done using post-implantation MRI. For each patient, analyzed structures are indicated in Table 1. In order to limit the influence of the common reference, bipolar signals (obtained from subtraction of monopolar signals recorded on two adjacent contacts in the structure) were preferred to monopolar signals. h 2 values were computed over a 5 s sliding window for all possible pairs (for instance, 6 possible interactions are taken into account when 4 mesial structures are considered). Values were averaged over time in order to get a single estimate (mean ± S.D.) of the degree of coupling between mesial structures, either broadband or in each EEG sub-band. Detection of interictal spikes (IS) The potential influence of IS on nonlinear correlation values was also studied by computing the values on (i) original SEEG recordings and on (ii) the same recordings, but after removal of IS. In order to detect IS, we used a previously described algorithm (Bourien et al., 2005). This method starts from the fact that interictal epileptic spikes include a sharp component corresponding to a transient wave of high amplitude and short duration compared to background activity. This component is characterized a specific signature in the time-frequency plane, i.e. an increase of energy in higher frequency bands (typically from 20 to 40 Hz). Based on this observation, the detection method included two main steps. In the first stage, each EEG signal was decomposed on a wavelet filter bank. The mean value q(t) of the squared modulus of filter outputs was computed at each sample time. This random quantity q(t) exhibited high mean value during spike or polyspikes periods and low mean value during background EEG. In the second stage, a Page-Hinkley algorithm (Basseville and Nikiforov, 1993) was used to automatically estimate time instants corresponding to abrupt changes of q(t), i.e. corresponding to IS occurrences. For each patient, detection parameters (bias and threshold) were adapted by the neurologist who qualitatively defined the best compromise between false alarms and non-detections by interactive visual analysis of detection results (Fig. 2b). Statistical analysis A non-parametric Mann Whitney test was performed in order to compare the relative power in frequency sub-bands (RP B values) as well as the nonlinear correlations (h 2 values) between the two groups of patients (MTLE and NMTLE). Correlation between h 2 values and the frequency of interictal spikes was done using a nonparametric measure of correlation (Spearman s rank correlation coefficient). A p-value 0.05 was considered as significant. For the analysis in each EEG sub-band a Bonferroni correction was applied for multiple comparisons. Therefore, as the number of sub-bands is equal to 5, a p-value 0.01 was considered as significant.

6 Enhanced EEG functional connectivity in mesial temporal lobe epilepsy 63 Figure 2 (a) Example of nonlinear regression analysis (broad band, non-filtered signals) of the functional coupling between three mesial structures, amygdala (A), anterior hippocampus (Ha) and entorhinal cortex (EC) over a period of 30 s characterized by to the nonlinear correlations coefficients h 2 xy obtained from the three SEEG signals. Three curves represent the temporal course of the h 2 xy values over time. Red curve: nonlinear correlations between A and Ha SEEG signals. Blue curve: nonlinear correlation between A and EC SEEG signals. Green curve nonlinear correlation between HA and EC SEEG signals. Signals are recorded in a patient from MTLE group. Interval between each vertical line represents 1 s. Amplitude is 10 V/mm for the signals in a and 5 V/mm for signals in b. (b) Example of spike detection result obtained from the algorithm described in Bourien et al. (2005) is shown. Spike detection remains a difficult issue in which the false alarm rate and the false negative rate must be controlled. In the present study, spike detection is performed according to an interactive method (semi-automatic): through a software user interface, the operator can modify the detection parameters (biais and threshold) and can have immediate feedback on detection results. (1) A 20-s segment of intracerebral EEG recorded from the hippocampus. (2) Detection results: green bars automatically appear on detected spikes for given parameter setting. In this example, one can notice that low-amplitude spikes are detected (green arrow). However some are also missed (red arrow). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of the article.) Results Relative power of SEEG signals in frequency sub-bands The relative power in each frequency sub-band (RP B ) values were first compared on a structure per structure basis between the two groups of patients. Mean values and standard deviations of RP B values computed on signals recorded from the anterior hippocampus, the posterior hippocampus, the entorhinal cortex, and the amygdala are provided in Fig. 3a for each frequency band. This figure shows that the relative power associated to oscillations in the theta band is reduced for all structures studied in the MTLE group. To a lesser extent, the same comment holds for the alpha band. For other frequency bands (delta, beta and gamma), results were found to vary from structure to structure: as the structure changes, either a decrease or an increase of the relative power (in a given frequency band) can be observed between the two groups of patients. These values were then averaged over recorded structures within the two groups of patients (MTLE and NMTLE) in order to globally assess the inter-group variability. Mean values and standard deviations obtained for each frequency band are compared in Fig. 3b. Statistical analysis showed that there was no difference between the two groups for delta (p = 0.38), alpha (p = 0.34), beta (p = 0.25) and gamma (p = 0.12) sub-bands. In contrast, values for the theta band were found to be significantly lower (p = 0.01) in the MTLE group than in the NMTLE group. The question that is raised is whether this decrease just reflects a general and unspecific slowing of recorded EEG signals or whether it is linked to the epileptogenic nature of the recorded structure. To answer

7 64 G. Bettus et al. Figure 3 (a) Relative power of SEEG signals in delta ( Hz), theta ( Hz), alpha ( Hz), beta ( Hz) and gamma ( Hz) frequency bands. (b) These values are averaged over recorded structures within the two groups of patients (MTLE and NMTLE) in order to globally assess the inter-group variability. Differences in theta band are significant ( * p = 0.01). MTLE: group of patients with a mesial temporal lobe epilepsy; NMTLE: group of patients with a non-mesial temporal lobe epilepsy. this question, we assumed that a general slowing of EEG activity would also result in an increase of the relative power in the delta sub-band. Therefore, we checked whether or not the delta and the theta values were correlated, using the Spearman rank correlation test. That was not the case since we did not observe any tendency corroborating this hypothesis (p = 0.93). This result strongly suggests that theta rhythm is altered in the MTLE group as the power of theta oscillations (relatively to the total power) present in signals recorded from epileptogenic mesial temporal structure is decreased compared to that measured in non-epileptogenic structures. Functional couplings among epileptogenic versus non-epileptogenic structures We first compared the nonlinear correlation values (h 2 ) measured on broadband signals and averaged over the 6 possible combinations between the four recorded structures in the two groups of patients (MTLE and NMTLE). A representation of mean values and standard deviations is provided in Fig. 4a (left part). As depicted, the averaged nonlinear correlation was found to be higher in the first group (0.233 ± 0.042) than in the second group (0.200 ± 0.015). This difference was found to be statistically significant (p = ). We then looked for differences in considered EEG sub-bands as displayed in Fig. 4b, which provides averaged h 2 values computed for each sub-band in the two groups. Visual analysis of histograms revealed a general tendency: nonlinear correlation values computed in patients of the MTLE group were higher than those computed in NMTLE patients for all frequency sub-bands. This tendency was further confirmed by statistical analysis which showed that values significantly differed between the two groups of patients in the theta band (p = ), alpha band (p = ), beta band (p = ) and gamma band (p = ). They were however found to be equivalent in the delta band (p = ). The above results obtained on both broadband and filtered signals suggest that the differences between the two groups can be associated to the existence of altered (i.e. enhanced) functional couplings between mesial structures in the MTLE group, as characterized by abnormally high nonlinear correlation values measured among signals from these structures. In order to determine if this increase in correlation values involves some particular interactions between structures, we compared the following interactions between the two groups: anthip-ec; anthip-am and anthip-posthip. Fig. 5 shows that for these three interactions, coupling values appeared to be higher in the MTLE group. Difference is highly significant for EC-Hip interactions (p = 0.004) but not for anthip-am (p = 0.06) and for anthip-posthip (p = 0.36).

8 Enhanced EEG functional connectivity in mesial temporal lobe epilepsy 65 Figure 4 (a) Mean h 2 values (EEG broad band ( Hz)) in MTLE and NMTLE groups of patients. h 2 values were computed from native SEEG signals (unprocessed data) or were computed on SEEG signals in which interictal spikes were withdrawn (discharges removed). Correlation values are significantly higher in the MTLE group for both conditions ( * p 0.01). (b) Mean h 2 values from mesial temporal lobe SEEG signals in each frequency sub-band. Correlation values are significantly higher in the MTLE group for theta, alpha, beta and gamma bands ( * p 0.01). Influence of interictal spikes in correlation values The presence of interictal spikes in signals recorded from mesial structures is a potential factor that could influence h 2 values and therefore explain the differences observed between the two groups of patients. In order to check this hypothesis, we first searched for the existence of a correlation between IS frequency and h 2 values. IS were automatically detected in each structure over the 30 min period selected in each patient (Fig. 2b). As expected, the average IS frequency was found to be higher in epileptogenic mesial temporal regions (MTLE group, about 14 spikes/min) compared to non-epileptogenic (NMTLE group, about 5.4 spikes/min). This inter-group difference was statistically significant (p < ). The correlation between h 2 values and mean spikes rates was then analyzed using the Spearman s correlation coefficient. In the NMTLE group, the correlation between the two variables was found to be statistically significant (p = ). In the MTLE group, the same tendency was observed although the p-value did not reach the Bonferroni-corrected statistical threshold (p = 0.07). The above result suggests that h 2 values are correlated to spiking activity. However, interictal spikes may not have the same influence on h 2 values measured in the two groups of patients. Consequently, one must still determine whether the observed difference in h 2 values is due to the effect of IS (that would be different in the two groups) or to enhanced connectivity in the MTLE group, as hypothesized above. In order to more accurately evaluate the effect of IS on estimated nonlinear correlation values, we re-computed these values on the same recordings performed in the 35 patients, but after suppression of time intervals containing interictal spikes (Fig. 1c). Spike-free signals represented approximately 45% and 70% of the data for analysis in, respectively, the MTLE and the NMTLE groups. Results are displayed in Fig. 4a (right part) which represents the mean h 2 values estimated on broad band signals from which interictal spikes were withdrawn for the two groups of patients. They show that a slight decrease of h 2 values (about 5%) is observed due to the absence of transients in the signals. However, the difference between the MTLE and NMTLE groups is conserved and is still significant (p = 0.002). The nonlinear correlations in EEG sub-bands are displayed in Fig. 6b, which provides averaged h 2 values computed for each sub-band in the two groups after spikes removal. We Figure 5 Mean h 2 values (EEG broad band ( Hz)) in MTLE and NMTLE groups of patients. Three interactions are considered (as available in all patients): anterior hippocampusamygdala (anhip-am); anterior hippocampus-entorhinal cortex (anthip-ec) and anterior hippocampus posterior hippocampus (anthip-posthip). Differences are significant for the interactions anthip-ec between the two groups. Figure 6 Relative power of signals computed on SEEG signals in which interictal spikes were withdrawn, in delta ( Hz), theta ( Hz), alpha ( Hz), beta ( Hz) and gamma (24 97 Hz) frequency bands. Mean h 2 values from mesial temporal lobe SEEG signals in which interictal spikes were withdrawn, in each frequency sub-band. Correlation values are significantly higher in the MTLE group for alpha and beta bands ( * p 0.01).

9 66 G. Bettus et al. still observe the previous general tendency since nonlinear correlation values computed in MTLE patients are higher than those computed in NMTLE patients for all frequency sub-bands. The difference is significant for the alpha band (p = ) and beta (p = ) but does not reach the significant threshold for delta band (p = 0.038) theta band (p = 0.06) and gamma (p = 0.25). Influence of interictal spikes on relative power of SEEG signals in frequency sub-bands As for h 2 values, we have studied the influence of IS on the power of SEEG signals. Mean values and standard deviations of RP B values were computed on signals recorded from mesial temporal structures after spikes removal. These values were averaged over recorded structures within the two groups of patients (MTLE and NMTLE). Mean values and standard deviations obtained for each frequency band are indicated in Fig. 6a. Statistical analysis showed that there was no difference between the two groups for Delta (p = 0.25), Alpha (p = 0.80), Beta (p = 0.80) and Gamma (p = 0.22) sub-bands. We still observed a decrease in the theta band that did not, however, reach statistical significance (p = 0.07). Results suggest that the decrease in theta band is related to the interictal spikes, at least in part. Discussion This study investigated the alterations of EEG oscillations recorded from the mesial temporal lobe (MTL) during the interictal period in two groups of epileptic patients (with and without mesial temporal lobe epilepsy). Two main findings were obtained. First, a decreased signal power in the theta frequency band was observed in MTL structures when these structures were epileptogenic. Secondly, the functional coupling between these structures was increased during the interictal state, compared with that measured in patients with non-epileptogenic mesial structures. Decrease of signal power in the theta band As previously described (Zaveri et al., 2001), we found that theta and delta activities are the prominent oscillations recorded in human mesial temporal lobe. Zaveri et al. (2001) found a decrease in absolute power of the signal in patients with hippocampal sclerosis in comparison with patients with normal MRI. Our approach was different since we studied the relative content of each sub-band and compared epileptogenic MTL (in patients in whom seizures were generated in these structures) versus non-epileptogenic MTL (in patients in whom seizure were generated outside the MTL). The only significant difference between these two groups was a decrease of the power in the theta band (relatively to the total signal power). The functional consequences of the theta decrease are still unknown despite abundant literature dedicated to theta rhythm and its potential functions. In rat, theta oscillations are recorded in several limbic structures, particularly the hippocampus, and are associated with different states from REM sleep to various types of locomotor activities (review in Buzsaki, 2002). In human, theta oscillations are thought to be involved in memory-related cognitive processes, since they have been linked to spatial navigation, working memory or episodic memory process (Basar et al., 2001; Ekstrom et al., 2005; Klimesch et al., 2005; Mormann et al., 2005). Theta oscillations could support the links between cortical and hippocampal regions during memory recollections (Barbeau et al., 2005; Guderian and Duzel, 2005). Therefore, we can speculate that the alteration of theta oscillations found in our epileptic population may have negative consequences on the function associated to brain regions studied here. In particular, the reduction in theta noted within the MTLE group could be an important factor involved in the reduction of episodic memory, often observed in such patients (Wieser, 2004). A direct comparison of the decrease in theta activity at rest and memory alterations of the studied patients however remains to be made. In agreement with this hypothesis, a recent study has reported a power decrease in theta and delta bands on the side of the epileptogenic zone in TLE patients performing a word memorization task. Findings of this study suggested a reduced availability of recruitable neural assemblies in the hippocampus and in the rhinal cortex during memory formation (Mormann et al., 2007). Enhanced EEG connectivity in the epileptogenic zone The second main finding was an increased degree of interdependencies between signals in epileptogenic mesial temporal lobe structures during the interictal state. These interdependencies were characterized by computing the nonlinear cross-correlation of signals recorded from structures of interest. During the past decades, numerous methods allowing the computation of indexes of synchronization between EEG signals have been proposed, both linear and nonlinear. Although results may differ from one method to the other, several recent studies have tended to show that, qualitatively, results are similar (Ansari-Asl et al., 2006; Quian Quiroga et al., 2002). As in our previous studies, we used a nonlinear method based on nonlinear regression analysis (Lopes da Silva et al., 1989; Pijn and Lopes Da Silva, 1993). This method is known to provide a robust estimate of the statistical coupling between two signals (h 2 coefficient) (Ansari-Asl et al., 2006). We characterized the functional coupling between mesial structures during the interictal state by comparing the MTLE group to a control group including patients with extra-temporal seizures. Results showed that functional connectivity was enhanced in the epileptogenic zone in this group of patients (MTLE). This effect was significant for the theta, alpha, beta and gamma bands and to a large part independent of the occurrence of interictal spiking. This result is in agreement with the study of Mormann et al. (2000) showing a local increase in interictal synchrony using the mean phase coherence in a group of 17 MTLE patients. It is also in line with a recent study which estimated the interdependencies between signals recorded by subdural grids in nine patients during presurgical evaluation with neocortical epilepsy (Schevon et al., 2007). Inter-electrode synchrony was quantified using the mean phase coherence

10 Enhanced EEG functional connectivity in mesial temporal lobe epilepsy 67 algorithm and revealed areas of elevated local synchrony that may be a marker of epileptogenic cortex. The functional significance of an increase of functional EEG connectivity between structures forming the EZ is unclear. It has long been proposed that the EZ in human epilepsies is characterized by a network of structures connected through abnormally reinforced links (Wendling et al., 1997). This hypothesis was historically established on the basis of seizure analysis and according to SEEG recordings that allow simultaneous analysis of multiple brain structures. It has been more recently proven that signals between EZ structures may be synchronized at seizure onset and during the course of the seizure (Bartolomei et al., 2005, 2001b, 2004, 1999). The course of the synchronization of EEG signals in the EZ is particular, generally characterized by a sequence of synchrony desynchrony pattern (Bartolomei et al., 2001b, 2004; Wendling et al., 2003). We propose that even during the interictal state, abnormal long-range reinforced EEG connectivity between structures forming the EZ network is a fundamental characteristic of this region. The underlying mechanisms are unknown but are probably related to the synaptic alterations observed during secondary epileptogenesis process (Khalilov et al., 2003). The features observed in the networked organization of the EZ are indeed reminiscent of those seen in secondary epileptogenesis processes which occur not only at the site of a lesion (hippocampal sclerosis, cortical dysplasia, etc.), but also in distant areas, suggesting that the anatomically distant areas undergo a physiological change consequent to neuronal alterations at the primary insult site (Morrell, 1989). Of particular interest, in a recent study using the amygdala kindling model in the rat (an animal model of secondary epileptogenesis), an increase of coherence values between amygdala and frontal cortex was observed. Therefore, enhanced connectivity appears to be a marker of the kindling phenomenon at the EEG level (Blumenfeld et al., 2007). This result suggests that this particular state can facilitate seizure occurrence by priming the system to synchronize during seizures. It has indeed been demonstrated that in mtle, during the few seconds preceding the ictal onset (fast discharge) there is a reinforcement of couplings between structures (Bartolomei et al., 2004). This phenomenon has been suggested to be the initial event necessary to recruit all the epileptogenic structures in seizure genesis. This view is however challenged by other studies suggesting that a decrease of synchronization could be a more favorable state in which the probability of seizure occurrence is increased (Mormann et al., 2003). Whatever the mechanisms involved, the results of the present study indicate that the EZ exhibits particular longlasting EEG features during the interictal state. This adds further evidence that the organization of the epileptogenic zone in MTLE is not reduced to a single focus and that a more complex network configuration is responsible for the epileptogenicity. References Ansari-Asl, K., Senhadji, L., Bellanger, J.J., Wendling, F., Quantitative evaluation of linear and nonlinear methods characterizing interdependencies between brain signals. Phys. Rev. E Stat. Nonlin. Soft Matter Phys. 74, Barbeau, E., Wendling, F., Regis, J., Duncan, R., Poncet, M., Chauvel, P., et al., Recollection of vivid memories after perirhinal region stimulations: synchronization in the theta range of spatially distributed brain areas. Neuropsychologia 43, Bartolomei, F., Guye, M., Gavaret, M., Regis, J., Wendling, F., Raybaud, C., et al., The presurgical evaluation of epilepsies. Rev. Neurol. (Paris) 158, 4S55 64S55. Bartolomei, F., Khalil, M., Wendling, F., Sontheimer, A., Regis, J., Ranjeva, J.P., et al., Entorhinal cortex involvement in human mesial temporal lobe epilepsy: an electrophysiologic and volumetric study. Epilepsia 46, Bartolomei, F., Wendling, F., Bellanger, J., Regis, J., Chauvel, P., 2001a. Neural networks involved in temporal lobe seizures: a nonlinear regression analysis of SEEG signals interdependencies. Clin. Neurophysiol. 112, Bartolomei, F., Wendling, F., Bellanger, J.J., Regis, J., Chauvel, P., 2001b. Neural networks involving the medial temporal structures in temporal lobe epilepsy. Clin. Neurophysiol. 112, Bartolomei, F., Wendling, F., Regis, J., Gavaret, M., Guye, M., Chauvel, P., Pre-ictal synchronicity in limbic networks of mesial temporal lobe epilepsy. Epilepsy Res. 61, Bartolomei, F., Wendling, F., Vignal, J., Kochen, S., Bellanger, J., Badier, J., et al., Seizures of temporal lobe epilepsy: identification of subtypes by coherence analysis using stereo-electro-encephalography. Clin. Neurophysiol. 110, Basar, E., Basar-Eroglu, C., Karakas, S., Schurmann, M., Gamma, alpha, delta, and theta oscillations govern cognitive processes. Int. J. Psychophysiol. 39, Basseville, M., Nikiforov, I., Detection of Abrupt Changes: Theory and Application. Prentice-Hall. Inc. Blumenfeld, H., Rivera, M., Vasquez, J.G., Shah, A., Ismail, D., Enev, M., et al., Neocortical and thalamic spread of amygdala kindled seizures. Epilepsia 48, Bourien, J., Bartolomei, F., Bellanger, J.J., Gavaret, M., Chauvel, P., Wendling, F., A method to identify reproducible subsets of co-activated structures during interictal spikes, application to intracerebral EEG in temporal lobe epilepsy. Clin. Neurophysiol. 116, Buzsaki, G., Theta oscillations in the hippocampus. Neuron 33, Cordes, D., Haughton, V.M., Arfanakis, K., Carew, J.D., Turski, P.A., Moritz, C.H., et al., Frequencies contributing to functional connectivity in the cerebral cortex in resting-state data. AJNR Am. J. Neuroradiol. 22, Csibra, G., Davis, G., Spratling, M.W., Johnson, M.H., Gamma oscillations and object processing in the infant brain. Science 290, Duckrow, R., Spencer, S., Regional coherence and the transfer of ictal activity during seizure onset in the medial temporal lobe. Electroencephalogr. Clin. Neurophysiol. 82, Ekstrom, A.D., Caplan, J.B., Ho, E., Shattuck, K., Fried, I., Kahana, M.J., Human hippocampal theta activity during virtual navigation. Hippocampus 15, Fell, J., Klaver, P., Elfadil, H., Schaller, C., Elger, C., Fernandez, G., Rhinal-hippocampal theta coherence during declarative memory formation: interaction with gamma synchronization? Eur. J. Neurosci. 17, Gotman, J., Levtova, V., Amygdala-hippocampus relationships in temporal lobe seizures: a phase coherence study. Epilepsy Res. 25, Guderian, S., Duzel, E., Induced theta oscillations mediate large-scale synchrony with mediotemporal areas during recollection in humans. Hippocampus 15,

11 68 G. Bettus et al. Guye, M., Regis, J., Tamura, M., Wendling, F., McGonigal, A., Chauvel, P., et al., The role of corticothalamic coupling in human temporal lobe epilepsy. Brain 129, Khalilov, I., Holmes, G.L., Ben-Ari, Y., In vitro formation of a secondary epileptogenic mirror focus by interhippocampal propagation of seizures. Nat. Neurosci. 6, Klimesch, W., Schack, B., Sauseng, P., The functional significance of theta and upper alpha oscillations. Exp. Psychol. 52, Le Van Quyen, M., Adam, C., Baulac, M., Martinerie, J., Varela, F., Nonlinear interdependencies of EEG signals in human intracranially recorded temporal lobe seizures. Brain Res. 792, Lieb, J., Hoque, K., Skomer, C., Song, X., Interhemispheric propagation of human medial temporal lobe seizures: a coherence/phase analysis. Electroencephal. Clin. Neurophysiol. 67, Lopes da Silva, F., Pijn, J.P., Boeijinga, P., Interdependence of EEG signals: linear vs. nonlinear associations and the significance of time delays and phase shifts. Brain Topogr. 2, Lowe, M.J., Mock, B.J., Sorenson, J.A., Functional connectivity in single and multislice echoplanar imaging using resting-state fluctuations. Neuroimage 7, Micheloyannis, S., Vourkas, M., Bizas, M., Simos, P., Stam, C.J., Changes in linear and nonlinear EEG measures as a function of task complexity: evidence for local and distant signal synchronization. Brain Topogr. 15, Monto, S., Vanhatalo, S., Holmes, M.D., Palva, J.M., Epileptogenic neocortical networks are revealed by abnormal temporal dynamics in seizure-free subdural EEG. Cereb. Cortex. Mormann, F., Fell, J., Axmacher, N., Weber, B., Lehnertz, K., Elger, C.E., et al., Phase/amplitude reset and theta-gamma interaction in the human medial temporal lobe during a continuous word recognition memory task. Hippocampus 15, Mormann, F., Fernandez, G., Klaver, P., Weber, B., Elger, C.E., Fell, J., Declarative memory formation in hippocampal sclerosis: an intracranial event-related potentials study. Neuroreport 18, Mormann, F., Kreuz, T., Andrzejak, R.G., David, P., Lehnertz, K., Elger, C.E., Epileptic seizures are preceded by a decrease in synchronization. Epilepsy Res. 53, Mormann, F., Lehnertz, K., David, P., Elger, C.E., Mean phase coherence as a measure for phase synchronization and its application to the EEG of epilepsy patients. Physica. D 144, Morrell, F., Varieties of human secondary epileptogenesis. J. Clin. Neurophysiol. 6, Pijn, J., Lopes Da Silva, F., Propagation of electrical activity: nonlinear associations and time delays between EEG signals. In: Zschocke, Speckmann (Eds.), Basic Mechanisms of the EEG. Birkauser, Boston. Ponten, S., Bartolomei, F., Stam, C., Small-world networks and epilepsy: graph theoretical analysis of intracerebrally recorded mesial temporal lobe seizures. Clin. Neurophysiol. 118, Quian Quiroga, R., Kraskov, A., Kreuz, T., Grassberger, P., Performance of different synchronization measures in real data: a case study on electroencephalographic signals. Phys. Rev. E Stat. Nonlin. Soft Matter. Phys. 65, Rodriguez, E., George, N., Lachaux, J.P., Martinerie, J., Renault, B., Varela, F.J., Perception s shadow: long-distance synchronization of human brain activity. Nature 397, Salvador, R., Suckling, J., Coleman, M.R., Pickard, J.D., Menon, D., Bullmore, E., Neurophysiological architecture of functional magnetic resonance images of human brain. Cereb. Cortex. Sarnthein, J., Petsche, H., Rappelsberger, P., Shaw, G., Stein, A., Synchronization between prefrontal and posterior association cortex during human working memory. Proc. Natl. Acad. Sci. U.S.A., 95. Schevon, C.A., Cappell, J., Emerson, R., Isler, J., Grieve, P., Goodman, R., et al., Cortical abnormalities in epilepsy revealed by local EEG synchrony. Neuroimage. Stam, C.J., van Cappellen van Walsum, A.M., Micheloyannis, S., 2002a. Variability of EEG synchronization during a working memory task in healthy subjects. Int. J. Psychophysiol. 46, Stam, C.J., van Cappellen van Walsum, A.M., Pijnenburg, Y.A., Berendse, H.W., de Munck, J.C., Scheltens, P., et al., 2002b. Generalized synchronization of MEG recordings in Alzheimer s Disease: evidence for involvement of the gamma band. J. Clin. Neurophysiol. 19, Summerfield, C., Mangels, J.A., Functional coupling between frontal and parietal lobes during recognition memory. Neuroreport 16, Talairach, J., Bancaud, J., Szickla, G., Bonis, A., Geier, S., Approche nouvelle de la chirurgie de l épilepsie: methodologie stéréotaxique et résultats thérapeutiques. Neurochirurgie 20 (Suppl. 1), Tallon-Baudry, C., Bertrand, O., Oscillatory gamma activity in humans and its role in object representation. Trends Cogn. Sci. 3, Tallon-Baudry, C., Bertrand, O., Fischer, C., Oscillatory synchrony between human extrastriate areas during visual shortterm memory maintenance. J. Neurosci. 21, RC177. Uhlhaas, P.J., Singer, W., Neural synchrony in brain disorders: relevance for cognitive dysfunctions and pathophysiology. Neuron 52, Welch, P., The use of the fast fourier transform for the estimation of spectra: a method based on time averaging over short, modified periodograms. IEEE Trans. Acoust. Electroacoust 15, Wendling, F., Badier, J., Chauvel, P., Coatrieux, J., A method to quantify invariant information in depth-recorded epileptic seizures. Electroenceph. Clin. Neurophysiol. 102, Wendling, F., Bartolomei, F., Bellanger, J., Chauvel, P., Interpretation of interdependencies in epileptic signals using a macroscopic physiological model of EEG. Clin. Neurophysiol. 112, Wendling, F., Bartolomei, F., Bellanger, J.J., Bourien, J., Chauvel, P., Epileptic fast intracerebral EEG activity: evidence for spatial decorrelation at seizure onset. Brain 126, Wieser, H.G., ILAE Commission Report. Mesial temporal lobe epilepsy with hippocampal sclerosis. Epilepsia 45, Zaveri, H.P., Duckrow, R.B., de Lanerolle, N.C., Spencer, S.S., Distinguishing subtypes of temporal lobe epilepsy with background hippocampal activity. Epilepsia 42,