Automated pediatric cardiac auscultation.

Most of the relevant and severe congenital cardiac malfunctions can be recognized in the neonatal period of a child's life. Misclassification of a congenital heart defect may have serious consequences on the long-term outcome of the affected child. Experienced cardiologists can usually evaluate heart murmurs with secure confidence, whereas nonspecialists, with less clinical experience, may have more difficulty. There is an acute shortage of physicians in South Africa and many rural clinics are run by nurses. Automated screening based on electronic auscultation at clinic level could therefore be of great benefit. This paper describes an automated artificial neural network as well as a direct ratio and a wavelet analysis technique, to discriminate between pathological and nonpathological heart sounds. To test the performance of the three techniques, auscultation data and electrocardiogram (ECG)-data of 163 patients, aged between 2 mo and 16 yr, were digitized. The neural network achieved a sensitivity and specificity of 90% and 96.5%, respectively, when tested with the Jack-knife method. Statistical analysis of the input to the final sigmoid function shows that a better than 99% sensitivity and specificity can be achieved if sufficient training data are available.