Discriminating preictal from interictal states by using coherence measures

Records of brain electrical activity from depth electrodes are analysed to detect dynamical marks anticipating an impending seizure in epileptic patients who are candidate for surgery. The analysis is based on the mutual information measure that enables a quantitative characterization of the statistical coupling or dependence between two dynamical variables. We found that in the preictal periods there are significant changes of the averaged values of the measures of interdependence between EEG signals with respect to the basal ones (interictals). Such increased (or decreased) levels of coupling was revealed by the mutual information between the phases of the signals as well. The analysis of the phase differences between recorded channels confirms these results.