6-way classification model on TUH EEG data set

Abstract

Epilepsy is a neurological disorder that affects the brain, causing seizures to people globally. According to World Health Organisation (WHO), there are more than 50 million epileptic patients worldwide. Electroencephalogram (EEG) are widely used in the medical industry to detect electrical activities of the brain and primarily used by neurologists to diagnose brain-related disorders. Spikes detected during the EEG test indicates that a patient has epilepsy. However, experts who analyse the results may have different interpretation of the results. The project aims to develop a model using traditional classification techniques to detect spikes and epileptiform discharges from background artefacts accurately. The data set used in this project is obtained from Temple University Hospital (TUH). Cross validation of 4-folds is applied onto the data set using Support Vector Machine (SVM), K-Nearest neighbours (KNN), Decision Tree (DT) and Random Forest (RF) separately and compared using confusion matrix to obtain the best training model. KNN has the best accuracy for the targeted data set. Lack of sufficient training data makes the model not usable for actual implementation.