EEG Signal Analysis for Epilepsy Detection Using Deep Neural Networks

Epilepsy is a neurologically life-threatening illness characterized by spontaneous, repeated seizures, which are described as a rapid surge of aberrant electrical activity in the cerebral cortex. Previously, invasive techniques were used to obtain electrical readings from the brain, but now the most frequently used test for epilepsy diagnosis is electroencephalography, which is a non-invasive and painless procedure where electrodes are placed on the scalp to monitor abnormal patterns in the brain's electrical activity. The research suggests designing and implementing a classification model based on convolutional neural networks and a hybrid LSTM and bi-LSTM models. Utilizing the publicly accessible Bonn University dataset, these deep learning algorithms forecast anomalous cerebral activity. The method was tested on ten individuals, five of whom were healthy and five who were epileptic. The dataset was tested for a CNN model, hybrid CNN, LSTM hybrid model as well as hybrid CNN, bi-LSTM model for epilepsy detection. The proposed models show a significant increase in sensitivity and accuracy when utilizing hybrid models to detect abnormal brain activity.