EEG Band Power Feature Based Electrode Combination Analysis for Epilepsy Detection

Abstract

Epilepsy is a chronic neurological disorder characterised by seizures caused by abnormal electrical activity in the brain. Today, the diagnosis of epilepsy is largely based on the analysis of electroencephalogram (EEG) signals. However, the nonlinear and chaotic nature of EEG signals makes visual inspection a time-consuming, labour-intensive, and specialised process. In this context, this study proposes a computer-aided diagnostic method that aims to analyse EEG signals more efficiently. Within the scope of the study, power values belonging to different frequency bands were obtained from EEG signals collected from 121 participants through 35 electrodes. In determining the electrode combinations, the individual classification performance of each electrode was analysed, the electrode with the highest performance was selected and combined with the other electrodes in the vector space, thus creating effective electrode combinations. The resulting feature vectors were presented as input to the decision tree classifier, and the model's performance was evaluated with metrics such as accuracy, F1 score, sensitivity, and specificity. Using a combination of theta band power features obtained from F3C3, C3P3, T6O2, T5O1, FP2A2, O2A2 and P4A2 electrodes, the model showed a high success in epilepsy detection with 84.6% accuracy and 81.1% F1 score. © 2025 IEEE.