A Decision Support System to Diagnose Heart Diseases in Newborns

摘要

The task of identifying the occurrence of a heart disease is very important for newborns, as those seriously affected may be under the risk of death. In this study, a computer-aided method is proposed to diagnose newborns potentially affected by a heart disease. An ensemble of decision trees has been trained and tested with phonocardiogram data, represented in a feature space that includes Shannon Energy and Wigner Bispectrum. High classification accuracy, sensitivity and specificity have been obtained on the given data. The validation process has been performed on a balanced dataset of 110 heart sound signals -taken from healthy and unhealthy medical cases.