Epilepsy Detection using Time-Frequency Domain and Entropy Based EEG Analysis

Abnormal electrical activities due to brain tumor, developmental anomaly, neural-atrophy in cortical/sub-cortical brain regions cause an epileptic seizure. Electroencephalography (EEG) is an important diagnostic test used for observing waveforms such as epileptic brain activities. In this study, a new method which detects epileptic seizure from EEG signals automatically is proposed. Discrete wavelet transform and time dependent entropy based statistical features of the EEG signal are used to train artificial neural networks. The proposed method has been applied on EEG signals obtained from healthy individuals and epileptic patients for epileptic seizure detection, and accuracy of 100% has been achieved. This method has also been applied on EEG signals containing normal, interictal and ictal states, and accuracy, sensitivity and specificity of 98.6%, 96.0% and 99.3% have been achieved, respectively.