Hippocampal effective synchronization values are not pre-seizure indicator without considering the state of the onset channels.

In this paper, a model-based approach is presented to quantify the effective synchrony between hippocampal areas from depth-EEG signals. This approach is based on the parameter identification procedure of a realistic Multi-Source/Multi-Channel (MSMC) hippocampal model that simulates the function of different areas of hippocampus. In the model it is supposed that the observed signals recorded using intracranial electrodes are generated by some hidden neuronal sources, according to some parameters. An algorithm is proposed to extract the intrinsic (solely relative to one hippocampal area) and extrinsic (coupling coefficients between two areas) model parameters, simultaneously, by a Maximum Likelihood (ML) method. Coupling coefficients are considered as the measure of effective synchronization. This work can be considered as an application of Dynamic Causal Modeling (DCM) that enables us to understand effective synchronization changes during transition from inter-ictal to pre -ictal state. The algorithm is first validated by using some synthetic datasets. Then by extracting the coupling coefficients of real depth-EEG signals by the proposed approach, it is observed that the coupling values show no significant difference between ictal, pre-ictal and inter-ictal states, i.e. either the increase or decrease of coupling coefficients has been observed in all states. However, taking the value of intrinsic parameters into account, pre-seizure state can be distinguished from inter-ictal state. It is claimed that seizures start to appear when there are seizure-related physiological parameters on the onset channel, and its coupling coefficient toward other channels increases simultaneously. As a result of considering both intrinsic and extrinsic parameters as the feature vector, inter-ictal, pre-ictal and ictal activities are discriminated from each other with an accuracy of 91.33% accuracy.

A method to identify reproducible subsets of co-activated structures during interictal spikes. Application to intracerebral EEG in temporal lobe epilepsy.

OBJECTIVE: We present a novel quantitative method to statistically analyze the distribution of multichannel intracerebral interictal spikes (multi-IIS) in stereoelectroencephalographic (SEEG) recordings. The method automatically extracts groups of brain structures conjointly and frequently involved in the generation of interictal activity. These groups are referred to as 'subsets of co-activated structures' (SCAS). We applied the method to long duration interictal recordings in patients with mesial temporal lobe epilepsy (MTLE) and analyzed the reproducibility of subsets of structures involved in the generation of multi-IIS for each patient and among patients. METHODS: Fifteen patients underwent long-term intracerebral EEG recording (SEEG technique) using depth electrodes. A 1 h period of continuous interictal EEG recording was selected for each patient with precautions regarding the time after anesthesia pre-SEEG, the temporal distance with respect to seizures, the vigilance state of the patient, and the anti-epileptic drug withdrawal. A research of SCAS was conducted on each recording using the developed method that includes 3 steps: (i) automatic detection of monochannel intracerebral interictal spikes (mono-IIS), (ii) formation of multi-IIS using a temporal sliding window, and (iii) extraction of SCAS. In the third step, statistical tests are used to evaluate the frequency of multi-IIS as well as their significance (with respect to the 'random distribution of mono-IIS' case). RESULTS: In each patient, several thousands of multi-IIS (mean+/-SD, 3322+/-2190) were formed and several SCAS (mean+/-SD, 3.80+/-1.47) were automatically extracted. Results show that reproducible subsets of brain structures are involved in the generation of interictal activity. Although SCAS were found to be variable from one patient to another, some invariant information was pointed up. In all patients, multi-IIS distribute over two distinct groups of structures: mesial structures (15/15) and lateral structures (7/15). Moreover, two particular structures, the internal temporal pole and the temporo-basal cortex, may be conjointly involved with either the first or the second group. Finally, some extracted SCAS seem to match well-defined anatomo-functional circuits of the temporal lobe. CONCLUSIONS AND SIGNIFICANCE: During interictal activity in MTLE, similar subsets of temporal lobe structures are involved in the generation of spikes. This paper brings statistical evidence for the existence of these subsets and presents a method to automatically extract them from SEEG recordings. Interictal activity is spatially organized in the temporal lobe and preferentially involves two functional systems of the temporal lobe (either mesial or lateral).

Neural networks involving the medial temporal structures in temporal lobe epilepsy.

OBJECTIVES: In a previous study using the averaged coherence technique to study interactions between medial/limbic and lateral/neocortical regions, we observed that epileptogenic networks in temporal lobe epilepsy seizures (TLES) could be divided into 4 subtypes, i.e. medial (M), medial-lateral (ML), lateral-medial (LM), and lateral (L). In the ML and LM subtypes, medial structures and the anterior temporal neocortex are co-activated at the onset of seizures. However, using this approach, we were unable to determine the direction of coupling and may have overlooked non-linear variations in interdependency. The purpose of the present study using non-linear regression for analysis of stereoelectroencephalographic (SEEG) signal pairs was to measure the degree and direction of coupling between medial and neocortical areas during TLES in patients with the M, ML, and LM subtypes. METHODS: Eighteen patients with drug-resistant TLEs who underwent SEEG recording were studied. We used a non-linear correlation method as a measure of the degree and the direction of coupling on SEEG signal pairs. Patients with pure lateral TLEs were not studied. We analyzed the functional coupling between 3 regions of the temporal lobe: the anterior temporal neocortex, the amygdala, and the anterior hippocampus. A physiological model of EEG generation was used to validate the non-linear quantification method and assess its applicability to real SEEG signals. RESULTS: Results are first based on a physiological model of EEG data in which both degree and direction of coupling are explicitly represented, thus allowing construction of the neural systems inside which causality relationships are controlled and generation of multichannel EEG signals from these systems. These signals provide an objective way of studying the performance of non-linear regression analysis on real signals. In medial networks (10 patients), the ictal discharge is limited to the medial limbic structures and may propagate secondarily to the cortex. Quantified results demonstrated no significant coupling between medial and lateral structures at the beginning of the seizures. Conversely, almost constant unidirectional or bidirectional coupling was observed between hippocampus and amygdala. In medial-lateral (5 patients) and lateral-medial (3 patients) networks, the initial ictal discharge includes both limbic and neocortical regions. A rapid "tonic" discharge is observed over the temporal neocortex at the onset of seizure. Quantitative analysis showed an initial increase in the non-linear correlation coefficient between neocortex and medial structures. Quantification of the coupling direction demonstrated influence of medial over lateral structures (medial-lateral) or of the lateral neocortex over medial structures (lateral-medial). CONCLUSIONS: These results confirm the existence of several generic and organized networks involving the medial structures during TLE seizures.

Interpretation of interdependencies in epileptic signals using a macroscopic physiological model of the EEG.

This paper presents a neurophysiologically relevant model in which vectorial epileptiform electroencephalographic (EEG) signals are produced from multiple coupled neural populations. This model is used to evaluate the performances of non-linear regression analysis as a method to characterize couplings between neural populations from EEG signals they produce. Two quantities, estimated on generated signals, namely the non-linear correlation coefficient and the direction index, are related to the degree and direction of coupling parameters of the model. Their statistical behavior is first studied on a set of signals simulated for relevant configurations of the model. They are then measured on real stereoelectroencephalographic (SEEG) signals. Results obtained in three patients suffering from temporal lobe epilepsy (TLE) show that abnormal functional couplings between cerebral structures, that establish during seizures, can be interpreted in terms of causality. Perspectives are oriented to the identification of epileptogenic networks in TLE.

Seizures of temporal lobe epilepsy: identification of subtypes by coherence analysis using stereo-electro-encephalography.

OBJECTIVES: Two subtypes of temporal lobe epilepsy (TLE) according to the structures initially involved during seizures are currently recognized: medial TLE (MTLE) and lateral (or neocortical) TLE (LTLE). A few reports have suggested that the classification of TLE subtypes might be larger according to variations in the interactions between medial structures and the neocortex. In this study, we analyzed these interactions using coherence analysis of stereo-encephalographic (SEEG) signals during spontaneous seizures. METHODS: Twenty-seven patients with drug-resistant TLE, diagnosed from ictal SEEG recordings obtained during pre-surgical evaluation, were studied. Orthogonally implanted depth electrodes with multiple leads according to Talairach's method were used to sample medial and neocortical structures. Coherence analysis of ictal discharges was performed between two SEEG bipolar signals from adjacent leads located either in medial structures (amygdala and hippocampus) or in neocortical regions of the temporal lobe. A new algorithm, which was designed to reduce the bias inherent in coherence estimation, was used to compute the coherence. RESULTS: We were able to classify TLE seizures (TLES) into 4 distinct categories: (1) 'medial' TLES, characterized by medial onset with later involvement of the neocortex in the form of a 'phasic' discharge. High ictal coherence values were observed between medial structures; (2) 'medial-lateral' TLES which started in medial structures with a fast low-voltage discharge (FLVD) which rapidly affects the neocortex (< or = 3 s). High coherence values were observed between medial and lateral structures; (3) 'lateral-medial' TLES, which are different from medial-lateral TLES in that the FLVD starts in the lateral neocortex and involves the amygdala and/or hippocampus almost immediately after; (4) 'lateral' TLES: characterized by a neocortical onset, a delayed involvement of medial structures (when present), and high coherence values between neocortical structures. CONCLUSIONS: These results demonstrate the existence of numerous interactions between medial limbic structures and the neocortex during TLE seizures. Such findings could have implications for surgical strategies and the prognosis of epilepsy surgery, particularly when limited resection is indicated.

Extraction of spatio-temporal signatures from depth EEG seizure signals based on objective matching in warped vectorial observations.

In the field of epilepsy, the analysis of stereoelectroencephalographic (SEEG) signals recorded with depth electrodes provides major information on interactions between brain structures during seizures. A comprehensive methodology of comparing SEEG seizure recordings is presented. It proceeds in three steps: 1) segmentation of SEEG signals; 2) characterization and labeling of segments; and 3) comparison of observations coded as sequences of symbol vectors. The third step reports a vectorial extension of the Wagner and Fischer's algorithm to first, quantify similarities between observations and second, extract invariant sequences of events, referred to as spatiotemporal signatures. The study shows that two observations of nonequal duration can be matched by deforming the first one to optimally fit the second, under cost constraints. Results show that the methodology allows to exhibit signatures occurring during epileptic seizures and to point out different types of seizure patterns. The study brings objective results on reproducible interactions between brain structures during ictal periods and may help in the understanding of epileptogenic networks.

Mining reproducible activation patterns in epileptic intracerebral EEG signals: application to interictal activity.

The study of interictal transient events may substantially complement the analysis of seizures in the presurgical evaluation of intractable epilepsy. A comprehensive methodology of quantifying reproducibility of activation patterns in intracerebral electroencephalography signals is presented. It may be applied to various forms of transient epileptic events under the assumption that a time of occurrence may be assigned to them. In this paper, the method is used on two different forms of interictal events (interictal spikes or sharpwaves and transient bursts of fast activity). The methodology is based on signal processing and data mining algorithms and proceeds in three steps: 1) detection of transient paroxysmal events (monochannel event); 2) identification of quasisynchronous transient paroxysmal events (multichannel events); and 3) automatic extraction of similar activation patterns. Results show that the methodology allows reproducible sequential activation sets to be identified from signals recorded in four patients. Potential advantages of the method are discussed with respect to other approaches.

Time-frequency matching of warped depth-EEG seizure observations.

A methodology of comparing depth-EEG seizure recordings is presented. The approach is based on an extension of Wagner and Fischer's algorithm to N x 2-dimensional sets, allowing a confrontation of nonequal duration observations characterized by their time-frequency distributions. It proceeds by time and frequency warping on the first observation to match the second, under cost constraints. Preliminary results show that relevant signatures can be extracted from recordings.

[Identification of epileptogenic networks from modeling and nonlinear analysis of SEEG signals]. / Identification de réseaux épileptogènes par modélisation et analyse non linéaire des signaux SEEG.

This work is focused on the study of the organization of the epileptogenic zone (E.Z.) in humans based on the analysis of stereo-electroencephalographic (SEEG) signals with signal processing methods, and more especially those dedicated to the estimation of signal interdependencies. In order to evaluate quantities provided by these methods and in order to relate them to the notion of functional coupling between cerebral structures, we developed a neurophysiologically relevant model able to generate EEG signals from organized networks of neural populations. We showed that the model can produce realistic multichannel epileptiform signals (when compared to real SEEG signals) under certain conditions (excitation/inhibition ratio within populations, uni/bi-directional coupling between populations). In this paper, the model framework is used to evaluate the performances of nonlinear regression analysis as a method to characterize couplings between cerebral structures from the SEEG signals they produce. Two quantities, a nonlinear correlation coefficient and a direction index, respectively related to coupling parameters in the model (degree/direction) are presented. These two quantities are measured on real SEEG signals recorded in patients suffering from temporal lobe epilepsy and candidate to surgical treatment. Results show that the characterization of functional couplings leads to the identification of networks referred to as 'epileptogenic networks', which might be responsible for the triggering of seizures. These results also corroborate our previous results on the classification of temporal lobe epilepsies, showing that there exist recurrent seizure patterns that can be classified on the basis of interactions between medial and lateral neocortical structures.

Heart signal recognition by Hidden Markov Models: the ECG case.

Wave recognition in ECG signals by Hidden Markov Models (HMMs) relies on the stationary assumption for the set of parameters used to describe ECG waves. This approach seems unnatural and consequently generates severe errors in practice. A new class of HMMs called Modified Continuous Variable Duration HMMs is proposed to account for the specific properties of the ECG signal. An application of the latter, coupled with a multiresolution front-end analysis of the ECG is presented. Results show these methods can increase the performance of ECG recognition compared to classical HMMs.

Exploring neural directed interactions with transfer entropy based on an adaptive kernel density estimator.

This paper aims at estimating causal relationships between signals to detect flow propagation in autoregressive and physiological models. The main challenge of the ongoing work is to discover whether neural activity in a given structure of the brain influences activity in another area during epileptic seizures. This question refers to the concept of effective connectivity in neuroscience, i.e. to the identification of information flows and oriented propagation graphs. Past efforts to determine effective connectivity rooted to Wiener causality definition adapted in a practical form by Granger with autoregressive models. A number of studies argue against such a linear approach when nonlinear dynamics are suspected in the relationship between signals. Consequently, nonlinear nonparametric approaches, such as transfer entropy (TE), have been introduced to overcome linear methods limitations and promoted in many studies dealing with electrophysiological signals. Until now, even though many TE estimators have been developed, further improvement can be expected. In this paper, we investigate a new strategy by introducing an adaptive kernel density estimator to improve TE estimation.

Model-based measurement of epileptic tissue excitability.

In the context of pre-surgical evaluation of epileptic patients, depth-EEG signals constitute a valuable source of information to characterize the spatiotemporal organization of paroxysmal interictal and ictal activities, prior to surgery. However, interpretation of these very complex data remains a formidable task. Indeed, interpretation is currently mostly qualitative and efforts are still to be produced in order to quantitatively assess pathophysiological information conveyed by signals. The proposed EEG model-based approach is a contribution to this effort. It introduces both a physiological parameter set which represents excitation and inhibition levels in recorded neuronal tissue and a methodology to estimate this set of parameters. It includes Sequential Monte Carlo nonlinear filtering to estimate hidden state trajectory from EEG and Particle Swarm Optimization to maximize a likelihood function deduced from Monte Carlo computations. Simulation results illustrate what it can be expected from this methodology.

Detection of synchronized firings in multivariate neural spike trains during motor tasks.

This paper describes and compares two classical methods for the detection of neuron groups which exhibit synchronized firings in multivariate spike trains. These methods were compared on experimental and randomized data corresponding to the firing activity of 104 neurons located in motor, premotor, and parietal cortices in a monkey during movement tasks. Both methods exhibited high false positive rates in randomized data, but results showed that this rate can be advantageously reduced with a simple postprocessing. Otherwise, one method permitted to detect a significant number of synchronized groups of neurons related to the behavioral task.

A level set method for vessel segmentation in coronary angiography.

This paper presents a level set technique to extract the vascular structures in coronary angiography. It makes use of the Mumford-Shah functional to extract contours that are not necessary defined by gradient. A shape artery simulator was implemented to test and evaluate the detection method. Experimental results are presented on simulated data and real images successively.

Relevance of nonlinear lumped-parameter models in the analysis of depth-EEG epileptic signals.

In the field of epilepsy, the analysis of stereoelectroencephalographic (SEEG, intra-cerebral recording) signals with signal processing methods can help to better identify the epileptogenic zone, the area of the brain responsible for triggering seizures, and to better understand its organization. In order to evaluate these methods and to physiologically interpret the results they provide, we developed a model able to produce EEG signals from "organized" networks of neural populations. Starting from a neurophysiologically relevant model initially proposed by Lopes Da Silva et al. [Lopes da Silva FH, Hoek A, Smith H, Zetterberg LH (1974) Kybernetic 15: 27-37] and recently re-designed by Jansen et al. [Jansen BH, Zouridakis G, Brandt ME (1993) Biol Cybern 68: 275 283] the present study demonstrates that this model can be extended to generate spontaneous EEG signals from multiple coupled neural populations. Model parameters related to excitation, inhibition and coupling are then altered to produce epileptiform EEG signals. Results show that the qualitative behavior of the model is realistic; simulated signals resemble those recorded from different brain structures for both interictal and ictal activities. Possible exploitation of simulations in signal processing is illustrated through one example; statistical couplings between both simulated signals and real SEEG signals are estimated using nonlinear regression. Results are compared and show that, through the model, real SEEG signals can be interpreted with the aid of signal processing methods.

Epileptic fast activity can be explained by a model of impaired GABAergic dendritic inhibition.

This paper focuses on high-frequency (gamma band) EEG activity, the most characteristic electrophysiological pattern in focal seizures of human epilepsy. It starts with recent hypotheses about: (i) the behaviour of inhibitory interneurons in hippocampal or neocortical networks in the generation of gamma frequency oscillations; (ii) the nonuniform alteration of GABAergic inhibition in experimental epilepsy (reduced dendritic inhibition and increased somatic inhibition); and (iii) the possible depression of GABA(A,fast) circuit activity by GABA(A,slow) inhibitory postsynaptic currents. In particular, these hypotheses are introduced in a new computational macroscopic model of EEG activity that includes a physiologically relevant fast inhibitory feedback loop. Results show that strikingly realistic activity is produced by the model when compared to real EEG signals recorded with intracerebral electrodes. They show that, in the model, the transition from interictal to fast ictal activity is explained by the impairment of dendritic inhibition.

Epileptic fast intracerebral EEG activity: evidence for spatial decorrelation at seizure onset.

Low-voltage rapid discharges (or fast EEG ictal activity) constitute a characteristic electrophysiological pattern in focal seizures of human epilepsy. They are characterized by a decrease of signal voltage with a marked increase of signal frequency (typically beyond 25 Hz). They have long been observed in stereoelectroencephalographic (SEEG) signals recorded with intra-cerebral electrodes, generally occurring at seizure onset and simultaneously involving distinct brain regions. Spectral properties of rapid ictal discharges as well as spatial correlations measured between SEEG signals generated from distant sites before, during and after these discharges were studied. Cross-correlation estimates within typical EEG sub-bands and statistical tests performed in 10 patients suffering from partial epilepsy (frontal, temporal or fronto-temporal) reveal that SEEG signals are significantly de-correlated during the discharge period compared with periods that precede and follow this discharge. These results can be interpreted as a functional decoupling of distant brain sites at seizure onset followed by an abnormally high re-coupling when the seizure develops. They lead to the concept of 'disruption' that is complementary of that of 'activation' (revealed by significantly high correlations between signals recorded during seizures), both giving insights into our understanding of pathophysiological processes involved in human partial epilepsies as well as in the interpretation of clinical semiology.

Comparison of two estimators of time-frequency interdependencies between nonstationary signals: application to epileptic EEG.

Numerous works have been dedicated to the development of signal processing methods aimed at measuring the degree of association between EEG signals. This interdependency parameter is often used to characterize the functional coupling between different brain structures or regions during either normal or pathological processes. In this paper we focus on the time-frequency characterization of interdependencies between nonstationary signals. Particularly, we propose a novel estimator based on the cross correlation of narrow band filtered signals. In a simulation framework, results show that this estimator may exhibit higher statistical performances (bias and variance) compared to a more classical estimator based on the coherence function. On real data (intracerebral EEG signals), they show that this estimator enhances the readability of the time-frequency representation of the relationship and can thus improve the interpretation of nonstationary interdependencies in EEG signals. Finally, we illustrate the importance of characterizing the relationship in both time and frequency domains by comparing with frequency-independent methods (linear and nonlinear).