Feature Extraction Method for Epileptic Seizure Detection Based on Cluster Coefficient Distribution of Complex Network

Automatic epileptic seizure detection has important research significance in clinical medicine. Feature extraction method for epileptic EEG occupies core position in detection algorithm, since it seriously affects the performance of algorithm. In this paper, we propose a novel epileptic EEG feature extraction method based on the statistical property of complex networks theory. EEG signal is first converted to complex network and cluster coefficients of every node in the network are computed. Through analysis of the cluster coefficient distribution, the partial sum of cluster coefficient distribution is extracted as the classification feature. A public epileptic EEG dataset was utilized for evaluating the classification performance of extracted feature. Experimental results show that the extracted feature achieves classification accuracy up to 94.50%, which indicates that it can clearly distinguish between the ictal EEG and interictal EEG. The higher classification accuracy demonstrates the extracted classification feature’s great potentiality of the real-time epileptic seizures detection. Key-Words: Epileptic Seizure Detection, Feature Extraction Method, Complex Network, Cluster Coefficient Distribution, Nonlinear Time Series Analysis, Electroencephalograph (EEG)