Self-organizing QRS-complex Recognition By Neural Networks

We are developing the selforganizing QRS-complex in electrocardiogram (ECG) recognition system by neural networks. The system consists of preprocessor, neural networks, and recognizer. The ART2 is employed as neural networks which self-organizes in response to the input ECG. The preprocessor divides the ECG into each cardiac cycle. One cardiac cycle of the ECG is inputted to the neural networks. The neural networks indicates the approximate locations of significant points to recognize the QRScomplex. Then, recognizer finds out exact locations of significant points around the location indicated. The selforganizing process of the ART2 stores the characteristic ECG patterns, so that the system can process the incoming ECG using these leaned pattern which will vary depend on patients. As a result, the system recognizes the QRS-complex with 97% accuracy. We are developing the handy system diagnosing the cardiac disease from electrocardiogram (ECG). The system enable us check our ECG without going to hospital. In this paper, we will demonstrate self-organizing QRS-complex recognition system based on neural networks, which is implemented by a personal computer. In the present system, the QRScomplex is recognized by obtaining both Q and S points.