Discriminative features for interictal epileptic discharges in intracerebral EEG signals

This paper extracts features and selects the most discriminate feature subset for classifying interictal epileptic discharge periods (IED) from non-IED periods in intracerebral EEG (iEEG) signals. Generalized autoregressive conditional heteroscedasticity (GARCH) model based on the student t-distribution is used to describe the wavelet coefficients of the iEEG signals. A variety of features are extracted from the coefficients of GARCH models. The Markov random field (MRF) based feature subset selection method is used to select the most discriminative features. Experimental results on real patients´ data validate the effectiveness of the selected features.