An Effective CAD System for Heart Sound Abnormality Detection

The study of heart sound signals is considered a helpful approach for monitoring heart diseases and for assessing heart hemodynamic condition. In fact, several cardiac disorders are tangible in heart sound signal characteristics such as intensity, time relations and spectral content. To assist cardiologists in cardiovascular pathology screening and prevention, a computer-aided system able to segment and classify phonocardiogram records is proposed. After the detection of the fundamental heart sounds, systole and diastole, various features are extracted and a correlation analysis for avoiding redundancy and for quantify the feature discrimination capacity is made. The performance of the conceived system is evaluated considering the accuracy, the sensitivity and the specificity in classifying heart sound signals as normal or abnormal and is tested adopting the entire collection of records provided by the PhysioNet/CinC Challenge 2016 database. The obtained results show the method ability to aid the interpretation of specialists during their clinical practice.