Application of LSTM Algorithm to Assist Diagnosis of Epilepsy Based on Electroencephalogram (EEG) Signals

Abstract

Epilepsy is a common disease that affects the brain's ability and has the potential to destroy the quality of life of sufferers. Diagnosis of epilepsy can be done by clinical testing and by using the electroencephalography (EEG) method. This research aims to apply artificial intelligence to improve the effectiveness and accuracy of EEG signal analysis. Epilepsy diagnosis is done automatically based on trained EEG signal files. This application can be done by applying the Long-Short Term Memory (LSTM) machine learning algorithm for recognizing patterns from brain signals that lead to epilepsy. The development was carried out using the EEG signal dataset from the University of Bonn which consists of 5 data sets. The detection process consists of the stages of data loading, augmentation, filtering, training, and classification. The developed system will be loaded into a GUI to facilitate users. The result of this research is a machine learning model with Long Short-Term Memory (LSTM) algorithm that has an accuracy rate of 91%, validation accuracy of 94% and loss of 0.2. Compared to other machine learning approaches such as SVM, KNN, and ANN, the proposed method achieves higher accuracy without the need for explicit feature extraction, highlighting its effectiveness in time-series signal classification. The model evaluation results show that this research is successful in assisting the detection of epilepsy using EEG signals with a high level of accuracy and efficiency.