Epilepsy Seizure Detection and Classification Analysis using Residual Neural Network

Abstract

Epilepsy is a form of neurological brain disorder. It is identified by the frequent occurrence of symptoms called epileptic seizure due to abnormal activities. Using an electroencephalogram (EEG), a diagnosis of epilepsy can be done. For detection and classification purpose, there are many techniques applied in detecting epilepsy seizure such as machine learning, and nowadays deep learning algorithms are most famous to biomedical research. However, most of the deep learning methods are only analyze the epilepsy classification performance based on accuracy percentages. In term of elapsed time or learning rate analysis, it is become a rare study. Therefore, this paper proposes an epilepsy seizure detection and classification using several Residual Neural Network (ResNet) architectures and identify which ResNet architecture gives the best performance. For comparison purpose, the EEG performance analysis will be analyzed using other convolution neural network (CNN) architecture, namely GoogLeNet. Based on the results obtained, ResNet architecture give the best performance analysis for seizure detection and classification with superb performance of 100% accuracy and shortest elapsed time which only recorded 1 minute and 25 seconds