Analysis and Prediction of Epilepsy Based on Visibility Graph

Abstract

To classify and predict epilepsy seizure, visibility graph method is applied to analyze epileptic EEG. This stragety transform EEG time series to a complex network, and then EEG time series is analyzed from network topology and statistical characteristics. This paper devotes to classify epileptic EEG in ictal and interictal period and predict epilepsy seizure using visibility graph with sliding window. The results show that clustering coefficient is statistically higher in ictal period than in interictal period, and there is no obvious difference for their average path length. Clustering coefficient can be regarded as a new marker of epiletic seizure and is used to predict seizure. The EEG analysis method provides a new idea for epilepsy diagnosis and prediction.