Epilepsy detection using wavelet transform, genetic algorithm, and decision tree classifier

Abstract

This work presents a unique detection approach for classifying epilepsy using the CHB_MIT dataset. The suggested system utilizes the discrete wavelet transform (DWT) technique, genetic algorithm (GA), and decision tree (DT). This model consists of three distinct steps. In the first one, we present a feature extraction method that uses a DWT of four levels on electroencephalogram (EEG) and electrocardiogram (ECG) signals. The second step is the process of feature selection, which entails the elimination of irrelevant features in order to produce datasets of superior quality. This is achieved via the use of correlation and GA techniques. The reduction in dimensionality of the dataset serves to decrease the complexity of the training process and effectively addresses the problem of overfitting. The third step utilizes a DT algorithm to make predictions based on the data of epileptic patients. The performance evaluation layer encompasses the implementation of our prediction model on the CHB-MIT dataset. The results achieved from this implementation show that using feature selection techniques and an ECG signal as additional information increases the detection model's performance. The averaging accuracy is 98.3%, the sensitivity is 96%, and the specificity is 99%.