Development of classification algorithm for epileptic seizures using electrocardiogram signal

Abstract

An epilepsy classification system using electrocardiogram (ECG) data will ease the process of diagnosis. In epileptic patients, the seizures affect Heart Rate Variability (HRV). This emphasizes the importance of autonomic function changes in diagnosing epilepsy. The present work proposes an algorithm that classifies a person as epileptic or nonepileptic using ECG signal. Time Domain Features (TDF) and Frequency Domain Features (FDF), derived from the R-R Intervals (RRI) of ECG signal are utilized. In addition, Statistical Features (SF) are derived from extracted TDF and FDF. The Support Vector Machines (SVM) classifier is used to classify the ECG signal as epileptic or nonepileptic based on the extracted TDF, FDF and SF. The classification accuracy of the proposed method exhibits 97.5%. Analysis on clinical data shows that the proposed combination of TDF, FDF and statistical HRV features gives excellent classification accuracy. These results indicate that the proposed method can be applied to wearable heart rate measuring devices for diagnostic purpose.