Normalized Gabor Entropy Analysis of iEEG for Prediction of Epileptic Seizures

A novel measure for analysis of multivariate signals in the time-frequency-space domain, the normalized Gabor entropy (NGE), is introduced and applied to multichannel intracranial EEG (iEEG) recordings hours prior to seizures onset in two patients with focal epilepsy. NGE profiles showed a statistically significant progressive decrease of NGE values at epileptogenic focus-related channels as time for seizures occurrence approached. This result implies the progressive appearance of dominant Gabor atoms in the EEG tens of minutes prior to seizures, the detection and monitoring of which could further assist with improvement of the performance of seizure prediction algorithms.