Detection and classification of epileptic EEG signals by the methods of nonlinear dynamics

Abstract

Epilepsy is a common neurological disease caused by the hypersynchronous discharge of brain nerve cells. The scalp or intracranial Electroencephalogram (EEG) signals from the clinic usually have the characteristics of chaos, nonlinearity, etc. On the one hand, to effectively identify the epileptic signals with these characteristics, two indicators, namely, Sample Entropy (SampEn) and Higuchi's Fractal Dimension (HFD) are selected as features, the most EEG signal segments were classified automatically by using the Support Vector Machine (SVM) classifier, and by this method, the recognition accuracy reached 89.8%. On the other hand, because the complexity of some EEG signals is not obvious, it is difficult to identify them by this method, to improve the recognition accuracy of this kind of signals, the method of combining phase space reconstruction (PSR) with Poincaré section(PS) is used, and both the epileptic and non-epileptic signals were distinguished to a certain extent, the recognition rate reached more than 90%. The above results can provide theoretical guidance for the recognition or prediction of epileptic EEG signals in clinical practice.