Epilepsy detection from EEG data using 2D Convolutional Neural Network

An epileptic seizure, a disorder in brain functionality, happens when electrical bursts spread across the brain, causing the person to lose control or consciousness. Predicting epileptic seizures before they occur is useful for seizure prevention with medicine or for neural pre-surgical planning. Machine learning and computational methods are used to predict epileptic seizures from electroencephalogram (EEG) recordings. However, noise removal and feature extraction in EEG data are two important challenges that have a negative impact on the effectiveness of both time and true positive prediction rate. We offer a model in this paper that provides a reliable strategy for both preprocessing and feature extraction. Our model is based on a two-dimensional Convolution Neural Network (CNN), with EEG input provided to the CNN in the form of two-dimensional images. On the University of Bonn data set’s normal vs intericatl vs ictal instance, our proposed system achieved 97.8 percent accuracy and over 97 percent for the other parameters, which include precision, recall, F1-score, and ROC-AUC. Our findings are reproducible using the code available on github [1].