Epileptic seizure detection using empirical mode decomposition based fuzzy entropy and support vector machine

Abstract

A neurological condition affecting the central nerve system of people causing recurring seizure is termed as epilepsy or seizure disorder. A seizure can be described as a brief, interim disturbance in the electrical activity of the brain and the cause of its occurrence is the quick firing of many nerve cells in the brain, causing an electrical storm. Almost 50 million people worldwide have epilepsy and its studies often rely on EEG signals for analyzing brains behavior during the occurrence of seizure. Many kinds of research around the world is carried out over past few years to analyze EEG signals automatically to detect epilepsy and type of seizure present. This paper presents classification of EEG signals into healthy/inter-ictal versus ictal EEGs using EMD based fuzzy entropy method and SVM classifier. In EMD, decomposition of the EEG signal from different epileptic states takes place to obtain IMFs. Fuzzy entropy reduces the detection time by reducing the data size of the EEG data without any loss of the information. So here EMD is followed by the calculation of FuzzyEn of the reconstructed signal obtained from IMFs. Finally, feature vectors are formed using fuzzy entropy that are applied to non-linear SVM classifier for classification purpose. The classification accuracies obtained by using proposed method for Z versus S, O versus S, N versus S, and F versus S when only 50% data is applied during testing of the classifier are 99.88%, 99.38%, 98.62%, and 97% respectively.