Epilepsy seizure detection using SWT features and ANN classifier

Abstract

Epilepsy is a common life-threatening neurological disorder, which is unpredictable and affects around 50 million people worldwide. Electroencephalography (EEG) is a monitoring method used to diagnose epilepsy. It records fluctuations in electrical activities of brain. In this study, Stationary Wavelet Transform (SWT) is used to analyse the EEG signal for classification of focal and normal EEG signal. SWT decomposes EEG signal into approximate and detailed coefficients. From each coefficient the statistical features are extracted and are fed into Artificial Neural Network (ANN) classifier for the discrimination of focal and normal EEG signal. These methods are verified by using Karunya EEG database. This method is achieved an overall accuracy of 98.75%, sensitivity of 98.39%, specificity of 99.16%, positive predictive value (PPV) of 99.16%, negative predictive value (NPV) of 98.33%.