A Novel Feature Extraction Method for Epileptic EEG Based on Degree Distribution of Complex Network

Automatic seizure detection is significant in relieving the heavy workload of inspecting prolonged electroencephalograph (EEG). Feature extraction method for automatic epileptic seizure detection has important research significance because the extracted feature seriously affects the detection algorithm performance. Recently complex network theory shows its advantages to analyze the nonlinear and nonstationary signals. In this paper, we propose a novel feature extraction method for epileptic EEG based on a statistical property of complex network. The EEG signal is first converted to complex network and the degree of every node in the network is computed. By analyzing the degree distribution, the weighted mean value of degree distribution is extracted as classification feature. A public dataset was utilized for evaluating the classifying performance of the extracted feature. Experimental results show that the extracted feature achieves not only higher classification accuracy up to 96.50% but also a very fast computation speed, which indicate the extracted feature can clearly distinguish the ictal EEG from interictal EEG and has great potentiality of realtime epileptic seizures detection. Key-Words: Feature Extraction Method, Epileptic Seizure Detection, Electroencephalograph (EEG), Degree Distribution, Complex Network, Nonlinear Time Series Analysis