Epileptic Seizure Detection: Approximate Entropy and Discrete Wavelet Transform based method

Epilepsy is one of the most common neurological disorders and affects almost 60 million people worldwide; many techniques were used to detect epileptic seizures in the EEG recording. In this study we have implemented the new non linear epileptic seizure detection, proposed by H.Ocak, the method is based on discrete wavelet analyse and Approximate Entropy. The detection technique is divided on four steps, the data is firstly divided on different epochs, and secondly a third level DWT decomposition was used. Then ApEn values were computed for D1 level in both normal and epileptic EEG, finally a threshold is used to detect epileptic activity. Epochs with ApEn values less then threshold (1.7) are considered as epileptic ones. Two data bases were used in this study; the first one is an EEG record for epileptic patient and the second is for a normal subject. The mean values of ApEn were respectively 1.3177 and 1.9138. The results obtained, led us to conclude that epileptic epochs are much predictable (less complex) then normal ones.