Skip to Main Content
In this study, decision tree algorithms are used with promising results in a crucial and at the same time complicated classification problem concerning differential diagnosis of heart sounds. Decision tree structures are constructed, using data mining/distillation methods and then are used to classify heart sounds that were recorded from patients that have either aortic stenosis (AS) or mitral regurgitation (MR). Emphasis is given on the selection of the appropriate features that are adequately independent from the heart sound signal acquisition method. The differentiation capabilities and the classification performance of the fully expanded decision tree classifiers and the pruned decision tree classifiers are studied for this problem. For each constructed decision tree classifier the partial classification accuracies for the AS and MR auscultation findings are also estimated.