An Artificial Neural Network Model for Epilepsy Seizure Detection

Epilepsy, neurological disorder, exposed by recrudescent seizures affects many people throughout the entire world. Visual inspection of encephalography (EEG) recorded by sensing devices is the common practice in detection of epilespsy.But, this approach requires long time inspection which is laborious and burden for a neurologist and in turns degrade the performances. Also, brain stimulation is an emerging technology for the diagnosis of many neurological problem such as epilepsy, dementia. The modern invention of electroencephalogram (EEG) signals recorder makes easy to interpret brain signals automatically for extraction of discriminating pattern of signals. So, automatic seizure detection by EEG recordings could replace the visual inspection. Therefore, this paper presents a model to detect epilepsy by recording EEG signals. The approach is based on signal processing in time frequency domain by multilevel Discrete Wavelet Transformation (DWT) and nonlinear Artificial Neural Network (ANN) model. Statistical features are considered to level the EEG signals. Then, the ANN model is trained, validated and tested. To validate the model, it is benchmarked on an epilepsy dataset recorded by the University of Bonn. By showing remarkable accuracy in predicting EEG signals of three classes such as normal, inter-ictal and ictal, the proposed model proves the prevalence. Also, the model is juxtapose with other models from the literature and shows higher accuracy.