Combining independent component analysis and backpropagation neural network for ECG beat classification.

We propose a method that uses independent component analysis (ICA) and backpropagation neural network to classify electrocardiogram (ECG) signals. In this study, ICA is used to extract important features from ECG signals. A backpropagation neural network follows to classify the input ECG beats into one of eight beat types. The independent components are calculated from the training ECG beats and serve as the ICA bases of the system. The ECG beat samples are then projected on the bases to build the ICA features for different beat types. The features based on ICA and the time interval between successive ECG beats are constituted into a feature vector and serve as inputs to the backpropagation neural network. In the study, 9800 QRS samples, including eight different ECG types, were extracted from the MIT-BIH arrhythmia database for experiments. Half of the samples were used in the training phase and the other half in the testing phase. The experiments showed an impressive highest accuracy of 98.37% under the condition that only 23 independent components were used. The results demonstrate the capability of the proposed method in the computer-aided diagnosis of heart diseases based on ECG signals.