Research on Noninvasive Diagnosis for Coronary Heart Disease Based on Neural Network

Objective To extract characteristic parameters of ECG signals a new method of non-invasive diagnosis for coronary heart disease with artificial neural network. Methods ECG signals were digitized with A/D converter and filtered to eliminating the noise. Span of QRS interval, R-R interval,and voltage of S-T segment of filtered ECG were detected. These 3 characteristics were as the input parameters of the input layer. Samples were trained with an improved 3-layers back propagation(BP) artificial neural network, as trained samples. The non-trained samples were recognized with these BP neural networks. Results After 12 samples had been trained about 1500 times, the BP neural network could accurately distinguish samples of coronary heart disease from the trained samples and also recognize 20 non-trained samples, 19 to be correct except one. Conclusion It is showed that based on BP network and characteristic parameters of ECG, a new and promising method of non-invasive diagnosis for coronary heart disease has been found.