Noninvasive detection and classification of coronary artery occlusions using wavelet analysis of heart sounds with neural networks

Coronary artery disease (CAD) is a main cause of death around the world. Most of olds somehow suffer a kind of cardiovascular disorder. So the on-time diagnosis and treatment of coronary occlusions is very crucial to them. It has been widely reported that coronary stenoses produce sounds due to the turbulent blood flow in partially occluded arteries. In this paper, we make use of wavelet analysis and artificial neural networks to analyze heart sounds for detection and classification of CAD's. Our heart sound signals are recorded synchronously with ECG and are sampled at 4 k-sample per second with 12 bit resolution. Results show that successive classification of normal and coronary occluded patients is fulfilled with a resolution of 90% for normal group and with a resolution of 85% for abnormal group increasing performance over previous systems.