Bayesian linear discriminant analysis with hybrid ABC-PSO classifier for classifying epilepsy from EEG signals

Abstract

One of the serious neurological disorders affecting millions of people in the world is epilepsy. Epilepsy is a disease of the brain where the neurons behave in an abnormal fashion thereby leading to seizures. To monitor the activities of the brain, Electroencephalography (EEG) signals play a vital role especially for the analysis and diagnosis of epilepsy. The neurophysiologist visually examines the signals of the brain for the identification of epileptic seizures and abnormalities in it. It is quite a difficult procedure because of its prolonging time consumption and erroneous decisions made by humans and so automatic seizure detection and classification systems came into existence. In this paper, the epilepsy classification was done in two stages, firstly with Bayesian Linear Discriminant Analysis (BLDA) and the output obtained from it is further optimized and classified with Hybrid Artificial Bee Colony - Particle Swarm Optimization (ABC-PSO) classifier. Results show that an average classification accuracy of 94% is obtained when classified with BLDA classifier and on further optimization and classification, an average classification accuracy of 97.91% is obtained when classified with ABC-PSO classifier.