Evidence Theory-Based Approach for Epileptic Seizure Detection Using EEG Signals

Electroencephalogram (EEG) is one of the potential physiological signals used for detecting epileptic seizure. Discriminant features, representing different brain conditions, are often extracted for diagnosis purposes. On-line detection necessitates that these features are to be computed efficiently. In this work, an evidence theory-based approach for epileptic detection, using such features, and several classifiers, is proposed. Within the framework of the evidence theory, each of these classifiers is considered a source of information and as such, it may have its own local view of the current brain state. To reach a global view, these sources are fused using the Dempster's rule of combination. Each classifier is given a certain weight, during the fusion process, based on both its overall classification accuracy as well as its precision rate for the respective class. Experimental work is done where five time domain features are obtained from EEG signals and used by a set classifiers, namely, Bayesian, K-nearest neighbor, neural network, linear discriminant analysis, and support vector machine classifiers. Higher classification accuracy of 89.5% is achieved by the proposed approach compared to 75.07% and 87.71% accuracy obtained from the worst and best classifier from the used set of classifiers, and those are linear discriminant analysis and support vector machine, respectively.