A New Approach for Epileptic Seizure Detection from EEG and ECG Signals Using Wavelet Decomposition

Early detection of epileptic seizures is critical in clinical diagnosis to prevent disease progression. Seizures can be detected by measuring muscle and brain activity, oxygen levels, heart rate, visual EEG, ECG, motion, or audio and video recordings of the human body. In this paper, we present a new seizure detector using the combined EEG and ECG signals as a module of a health care system. Particularly, the module is based on discrete wavelet decomposition with time-domain and frequency-domain features and classification using an artificial neural network. The seizure detection module was evaluated on the CHB-MIT Scalp EEG database, available on Physionet. Simultaneous EEG and ECG recordings were accomplished on a single subject. Four recordings are used in the test. Each contains approximately 3 h of digitized EEG signals recorded with the ECG signal (23 channels for the EEG signal and 1 channel for the ECG signal). First, the EEG and ECG signal recordings are preprocessed and decomposed using DWT. Then the features are extracted from the decomposed signals using entropy measures, standard deviation, energy, mean, maximum value, minimum value, standard deviation, and variance. The multilayer perceptron is examined to classify the resulting features. The best results achieved are 98.7%, 100%, and 96% for accuracy, specificity, and sensitivity, respectively, using the fusion of discrete wavelet decomposition and an artificial neural network. Experimental analysis shows that using the ECG signal as additional information has a high classification capacity.