Classification of bioprosthetic valve closure sounds by a neural network using linear prediction coefficients

A three layers feed-forward back-propagation neural network was trained to classify bioprosthetic valve closure sounds from 47 patients with a porcine bioprosthetic valve inserted in the aortic position. Twenty-four patients had a normal valve and 23 a degenerated one. Twelve linear prediction coefficients computed from the closure sounds were used as the network input The neural network yielded 89% correct classification in an evaluation using the leave-one-out method. This study confirmed the potential of heart sound classification by using a neural network.