Application of empirical mode decomposition algorithm for epileptic seizure detection from scalp EEG

Abstract

The present study investigates the effectiveness of Empirical Mode Decomposition (EMD) for real-time epileptic seizure detection from scap electroencephalogram (EEG). The EMD algorithm is used to decompose the scalp EEG signal into a finite number of intrinsic mode functions (IMFs). These intrinsic mode functions are used to obtain features that are tested using a support vector machine (SVM) based classifier. For simplicity, the mean frequency of the first and the last intrinsic mode function components within each two-second seizure epoch is used as the feature for classification. The dataset consists of a total of 198 seizures from 23 pediatric patients and one adult patient. A 3-fold-cross-validation method resulted in 70.72% mean sensitivity, 6.33 seconds mean latency and 95.37% mean specificity. These results can potentially be improved using features from more or all IMF components for each seizure epoch.