Early diagnosis of heart disease using classification and regression trees

Abstract

Early diagnosis of heart defects are very important for medical treatment. In this paper, we propose an automatic method to segment heart sounds, which applies classification and regression trees. The diagnostic system, designed and implemented for detecting and classifying heart diseases, has been validated with a representative dataset of 116 heart sound signals, taken from healthy and unhealthy medical cases. The ultimate goal of this research is to implement a heart sounds diagnostic system, to be used to help physicians in the auscultation of patients, with the goal of reducing the number of unnecessary echocardiograms and of preventing the release of newborns that are in fact affected by a heart disease. In this study, 99.14% accuracy, 100% sensitivity, and 98.28% specificity were obtained on the dataset used for experiments.