Classification of Epileptic Seizures using Artificial Neural Network with Adaptive Momentum

Epilepsy seizures are neurological diseases caused by brain disorders. Electroencephalogram (EEG) is a common clinical method to detect epilepsy seizure from EEG signals. Several automated techniques have been proposed for classifying different classes of epilepsy report satisfactory results, but still there is room available for improving the classification accuracy. The proposed work uses a new artificial neural network (ANN)-based algorithm with adaptive momentum that can improve the classification accuracy of the classifier. Backpropagation with variable adaptive momentum (BPVAM) achieves high diagnostic accuracy and has high stability and robustness against data variations. The algorithm has also shown the ability to achieve the results in a highly efficient time compared to other well-known techniques. In addition, principal component analysis (PCA) has widely been used as a feature extraction method. Therefore, these two methods have been combined to form a robust classifier known as BPVAM-PCA that can improve the classification accuracy. To illustrate the ability of the proposed method, several experiments were performed for epilepsy cases identification from EEG signals. Moreover, a comparative analysis was also performed with some well-known automated techniques.