Assessing the neuronal network activity underlying human epileptic seizures
Zusammenfassung der Projektergebnisse
Using electroencephalographic recordings (EEG) of epilepsy patients and methods of time series analysis the purpose of the research project was to identify (i) characteristic changes of the global network activity and (ii) local parts of the neuronal network that are specifically active during seizure evolution. Significant progress was made in several directions: 1. On basis of arbitrary bivariate interrelation measures a family of matrices was defined that allows assessing the significance of measured values with respect lo appropriate surrogates. All matrix elements are set on the same scale independently. Recent findings that nonlinearities are not significant in resting-state functional magnetic resonance imaging data were confirmed. For intracranial EEG data nonlinearities are mainly significant during epileptic seizures and for epileptogenic brain regions. 2. The role of network hubs during epileptic seizure evolution was assessed using adjacency matrices obtained from data-driven thresholding and the node degree as approximation to the hub property. The temporal evolution of the hubs turned out patient specific and reproducible. Interestingly, the highest node degrees are close to but not inside the seizure onset zone. This result has good chance to find its way into clinical practice of focus localization. 3. An estimator of global correlation in multivariate systems was developed and a leave-oneout approach was used to assess the individual contribution of single channels. In high frequency intracranial EEG data the epileptogenic tissue turned out characterized by high correlation. During seizure a pronounced spatial reorganization of correlation takes place. 4. Steps towards a genuinely multivariate definition of finite-lag cross-correlation matrices were undertaken. For high dimensional optimization two cost functions and various optimizers were tested. It was found that using a cost function with low fluctuations is crucial. 5. For time series with EEG-like properties it was found that as compared with the crosscorrelation matrix the slope cross-correlation matrix has considerably suppressed fluctuations (see 4). Besides eigenvalue analysis this made time resolved analysis of eigenvectors possible for intracranial EEG before, during and after epileptic seizures. 6. Combining EEG slopes as univariate markers of epileptiform activity and eigenvectors of the slope correlation matrices of 5 as multivariate entities, a systems approach to analysis of neural correlation was undertaken, that integrates effects on various spatial scales. For intracranial EEG data of epilepsy patients it was shown that during seizures spatial correlation patterns are reorganized in a patient specific and reproducible way. The seizure onset zone can be identified by early ictal decrease of correlation, confirming the findings of the complementary method 3.
Projektbezogene Publikationen (Auswahl)
-
Data-driven estimates of the number of clusters in multivariate time series, Phys. Rev. E78, 066703 (2008)
C. Rummel, M. Müller and K. Schindler
-
Analyzing spatio-temporal patterns of genuine cross-correlations, J. Neurosci. Meth. 191, 94-100 (2010)
C. Rummel, M. Müller, G. Baier, F. Amor and K. Schindler
-
Assessing periodicity of periodic leg movements during sleep. Front. Neurosci. 4, 58 (2010)
C. Rummel, H. Gast, K. Schindler, M. Müller, F. Amor, C.W. Hess and J. Mathis
-
Peri-ictal correlation dynamics of high-frequency (80-200Hz) intracranial EEG, Epilepsy Res. 89, 72-81 (2010)
K. Schindler, F. Amor, H. Gast, M. Müller, A. Stibal, L. Mariani and C. Rummel
