Using Entropies for the Analysis of Brain Rhythms

Abstract

Epilepsy is a chronic neurological disorder characterized by seizures. It involves abnormal discharging of neurons that effects smaller section of the brain, referred to as partial epilepsy or larger section of the brain resulting in generalized epilepsy. Sometimes these abnormal activities spread from smaller section to the larger section of the brain resulting in secondary generalized epilepsy. Hence, it is important to detect and control epileptic seizure in an early stage. In this work, we design a system that classifies interictal (period between the seizure) and ictal (after onset of seizure) signals by extracting subtle information from the EEG rhythms: gamma, beta, alpha, theta and delta. The following system also aims to determine the sensitivity of these EEG rhythms towards epileptic seizure. In this research, we have used entropy methods namely: Shannon entropy, approximate entropy and sample entropy to extract the subtle information from the EEG rhythms. Classifiers namely: k-nearest neighbor, support vector machine and linear discriminant analysis is utilized to distinguish interictal and ictal signals with a classification accuracy of 94%, 95.5% and 97.5%.