Automatic Identification of Epilepsy Seizures from EEG Signals Using a Hybrid CNN and LSTM Model

Abstract

The neurological illness known as epilepsy is characterized by a disruption in the normal functioning of the brain and can be found in severe cases. It is estimated that more than 10 percent of the whole population across the entire planet is affected by this ailment Every single day. When acquiring information on the brain's electrical activity, electroencephalograms, often known as EEGs, are utilized rather frequently by researchers. In this paper, an end-to-end system is proposed that utilises a combination of two deep learning models, namely Convolutional Neural Networks (CNNs) and Long Short-Term Memory Networks (LSTM), to classify electroencephalogram (EEG) data of epilepsy disordered people into three distinct categories: preictal, normal, and seizure. The findings of the experiment were obtained by making use of a dataset that is well-known and easily available, which Bonn International University provided. Within this CNN-LSTM classification model, the tasks of feature extraction, selection, and classification are all carried out in an automated fashion. Because of this, there is no longer a requirement for a manually devised methodology for feature extraction. In this study, the performance of the CNN-LSTM model is studied and assessed with relation to specificity, sensitivity, and accuracy. This is accomplished through the usage of the 10-fold cross-validation approach. The accuracy is 99.33%, the sensitivity is 99.33%, and the specificity is 99.66% concurrently, as indicated by the findings that were gathered while the trials were being carried out and the data that were collected. The results of our research indicate that deep learning approaches are the most appropriate choices for categorization when compared to other methods that are currently deemed to be state-of-the-art.