Epileptic Seizure Detection using EEG Signals

Abstract

Epilepsy is a common neurological disorder which can be diagnosed by neurologists or physicians by using electroencephalogram or EEG signals. Since the manual examination of EEG for this purpose is very time consuming and requires trained professionals, it calls for the need of an automatic seizure detection method. In this study, time and frequency domain features are extracted from the EEG signals after preprocessing the raw EEG data and then using machine learning algorithms such as Logistic Regression, Decision Tree, Support Vector Machines, etc. to detect generalized seizures in the Temple University Hospital (TUH) corpus. A detailed account of the TUH dataset is also given. This work summarizes and compares the results of each of the algorithm trained, in terms of the performance metrics. Using the proposed approach, SVM obtained the highest accuracy of 92.7% in binary classification.