Real-time ischemic beat classification using backpropagation neural network

This paper explains an adaptive backpropagation neural network (NN) for the detection of ischemic beats in electrocardiogram (ECG) recordings. The proposed method consists of a preprocessing stage for QRS detection, baseline wandering removal, and noise suppression. In this stage ST segments are extracted. In the next stage, the pattern length is reduced and subtracted from the normal template. In the third stage the extracted patterns are used for training a neural network and ischemic beats are detected. The algorithm used to train the NN is an adaptive backpropagation algorithm. An adaptive algorithm attempts to keep the learning step size as large as possible while keeping learning stable and then reduces the learning time. To evaluate the methodology, a cardiac beat dataset is constructed using several recordings of the European Society of Cardiology ST-T database. Our results were high both in sensitivity and positive predictivity. Specially, the obtained sensitivity and positive predictivity were 97.22% and 97.44%, respectively. These results are better than other any previously reported ones.