Neural network classifier for the detection of epilepsy

Abstract

Epilepsy is a neurological disorder which affects the nervous system. Epileptic seizures are due to hyperactivity in certain parts of the brain. Automatic seizure detection helps in diagnosis and monitoring of epilepsy especially during long term recordings of EEG. This paper presents the bispectrum analysis of electroencephalogram (EEG) for the detection of epilepsy. Bispectrum is a higher order spectrum. It characterizes the nonlinearities in the signal. Features extracted from the bispectrum of EEG are applied to the neural network classifier to detect normal and epileptic EEGs. The classification accuracy of 81.67% is obtained. The results demonstrate that the proposed features are more effective in differentiating epileptic EEG as compared to features from the conventional power spectrum.