Neural networks for the diagnosis of coronary artery disease

Abstract
Coronary stenoses produce sounds due to the turbulent blood flow in partially occluded arteries. If these sounds are reliably detected, they provide a simple, noninvasive way to detect coronary artery disease. The authors examine the utility of neural networks for extracting useful information from the diastolic heart sounds associated with coronary stenoses and the clinical examination variables. Fifty-five recordings (35 abnormal, 20 normal) were studied. Ten of the patients in the database were selected for use as training cases for the neural network. The network correctly identified 25 or 30 patients with coronary artery disease and 13 of 15 patients with no apparent occlusions. These results compare favorably with other noninvasive methods for detecting coronary artery disease.

This publication has 13 references indexed in Scilit: