Empirical and Advanced Classification for Automatic Seizure Detection in Brain Images

Abstract

Epilepsy occurs with the result of abnormal transiting disturbance and electrical relative activities that appear in human brain. Electroencephalogram (EEG) isa sufficient test measure to identify the abnormal electrical activity of neurons and it is widely used in detection and analysis of electro epileptic seizures. Based on emergency inception appear in human brain related data sets, pyramidal ensemble convolutional neural network (PECNN) is one of the approaches to classify different instances with respect attribute data. It is often complex to classify EEG data for epilepsy patients because of the presence of high amount of Epilepsy related content which makes it complex to interpret and to classify the brain data which consists epilepsy features.So, in this paper, we propose Empirical De-composition-based Classification Approach (EDCA) to analyze the abnormal epileptic seizure signal from human brain images. EDCA evaluates intrinsic mode functions (IMF), and it extracts different features obtained from IMF for classification of such abnormal epileptic components using least square support vector machine (L-SVM) classifier with different radial bias functions (RBF). RBF provides best accuracy of classification for proposed approach with respect to epileptic seizure EEG extraction from human brain images. An experimental result of proposed approach gives better and accurate results with respect to existing approaches in terms of different parameters.