Enhancing Epileptic Seizure Classification Using Entropy Measures on EEG Signals

Electroencephalogram (EEG) signals provide a non-invasive means to monitor brain activity and are crucial for the diagnosis of neurological conditions such as epilepsy. This study explores the use of entropy-based features to enhance the classification of epileptic seizures from EEG signals. Eight entropy measures–Shannon, Spectral, Renyi, Tsallis, Wavelet, Approximate, Sample, and Permutation entropy–were extracted and evaluated both individually and in various combinations. Support Vector Machine (SVM) was employed as the classification model. The results show that individual entropy features such as Sample Entropy alone achieved only 34% Accuracy in a five-class classification task, indicating limited discriminative power. In contrast, the combination of multiple entropy features significantly improved performance, with the best accuracy reaching 99.7% when the entropy features were combined with the Multidistance Signal Level Difference (MSLD) technique. This highlights the importance of entropy feature fusion and advanced preprocessing in improving EEG classification systems. The proposed approach demonstrates strong potential for future clinical implementation in seizure detection systems.