Abstracts

Rationale: Epileptic seizure is a brain disorder where neuronal networks exhibit abnormal synchronization and its prediction is important and still a challenging issue in epilepsy investigations. We believe that novel multivariate electroencephalographic (EEG) analysis methods for inferring neural connectivity could be a valuable tool for characterizing normal, as well as, ictal, and perictal brain activities. This is a preliminary study aimed at characterizing effective neural connectivity pattern, using ordinary coherence and partial directed coherence (PDC) analysis, among brain areas related to epileptic seizure. The PDC is the frequency domain counterpart of Granger causality concept, and gives a measure of effective connectivity between channels at a given frequency.Methods: The EEG scalp recordings were obtained using 10-20 international electrodes positioning system with temporal 10-10 extension and sphenoidal electrodes in a patient with temporal lobe epilepsy due to hippocampal sclerosis evaluated along interictal, perictal, ictal and postictal epochs. Analyses were performed using F7, F8, FT7, FT8, T7, and T8 electrodes signals spanning perictus period, including 60 sec pre- and 60 sec post-ictus epochs, using 2-sec sliding window with 95% overlapping among adjacent segments in order to capture time-resolved change of brain dynamics. For each 2-sec-long EEG segment, vector autoregressive model was fitted, then coherence functions and PDC were estimated as a measure of dependence in between combinations of all channels. For visualization, color-coded PDC and coherence plots were obtained for connectivity dynamics analysis.Results: Upon PDC and coherence plotting visualization, a clearly identifiable patterns of connectivity emerged showing transition between pre-ictus, ictus and post-ictus periods. As expected ordinary coherence function was highest in between adjacent electrodes signals in all frequency range. For pair of electrodes, ordinary coherence function was high and significant in all frequency range (high synchronicity) during the ictus. In the post-ictus period we observed significant coherence peaked in low frequency, < 3 Hz, in between ipsi-, as well as contra lateral electrode pairs. As for PDC analysis, it showed a strongest flow of information mostly, in high frequency range (>40Hz), during the pre-ictus period, directed from focus site hemisphere toward opposite hemisphere. Conclusions: Our results clearly show a change in connectivity patterns in the transitions from interictal to postictal intervals – marked by significant spectral coherence increase between all areas. These findings support our hypothesis that functional connectivity analysis using PDC could be an important tool for the study of pathophysiology and, eventually for prediction of seizure onset.