EEG-Based Epileptic Seizure Detection and Classification Using Machine Learning

Pratik Rajendra Saindane, Dr. Ayesha Siddiqui

Abstract: In this research, there is a new technique for the identification of epileptic seizures using the strength of ML and DL algorithms in EEG signals. Epilepsy seizures are specific neurological disorders that can be characterized through unique features present in Electroencephalography (EEG). Therefore, these are highly credible with researchers. Recently, ML and DL algorithms have proven to be very effective tools in feature extraction and classification in EEG signals. There are many research works, where EEG signals have been converted into images/feature vectors in the time-frequency domain and subsequently classified. On the contrary, this study will focus on classifying EEG signal representation in time series through the use of machine learning classifiers in terms of parameter tuning along with the deep learning technique of one-dimensional convolutional neural network (1D CNN). The primary objective of this study is to find the best classifier, while at the same time emphasizing the importance of some significant parameters like sensitivity, precision, and accuracy, which play an important role in medical studies. UCI Epileptic Seizure Recognition is the dataset used for this study that contains time series data points in EEG signals. On processing, the dataset is fed to various classifiers including XGBoost, Tabnet, RF, and One- Dimensional Convolutional Neural Network (1D CNN), achieving an accuracy rate of 98%, 96%, 98% and 99%, respectively.

Keywords: XGBoost, TabNet, Deep learning (DL), Machine Learning (ML), Random Forest (RF), Epileptic seizures, 1D CNN, data points, Time series