ANN based Epilepsy Classification using EEG

Abstract

Brain is the most complex organ amongst all the systems in human body. ElectroEncephaloGraph (EEG) is a record of electrical activity of brain, which is used to identify the neurological disorder of brain. There are various neurological disorders like Epilepsy, brain cancer, etc. Epilepsy is one of the most common neurological disorders of brain. Objective detection efficiently is still a challenging task for many neurological disorders. Epilepsy needs to be detected efficiently using required EEG feature extraction such as: variance, power spectral density, energy and entropy. This dissertation discusses the design of the system that detects the epileptic activity with efficacy. Decomposition of the EEG signal to various sub-bands by multi- level wavelet decomposition is followed to extract features from these sub-bands. These sub-bands are delta (0-4.05 Hz), theta (4.05-8.1 Hz), alpha (8.1-12.15 Hz), beta (12.15-32.5 Hz) and gamma (>32.5 Hz). The range of the features in normal and epileptic group of 50 subjects from each data set is analyzed for data available at the Department of Epiletology, University of Bonn. These features are then classified using an artificial neural network (ANN) that detects the EEG data for epilepsy. The performance of the ANN classifier is evaluated in terms of sensitivity, specificity and classification accuracy. Two-class classification is done for normal and epileptic subjects. Normal subjects are further classified for eyes open and eyes closed. Epileptic subjects are further classified for non-seizure and seizure, and non-seizure as hippocampal formation and epileptogenic zone. It is observed that the repeated use of two-class classifier gives better accuracy i.e. in the range of 98.7% - 100%. Multi-class classification is done on normal and epileptic subjects that show lower but reasonably good classification accuracy (88.4%).