Classification of Heart Sounds Using Machine Learning

Artificial intelligence has become an important tool in the domain of medicine. This includes providing care to those with limited access or improving the accuracy of diagnosis. In this project a machine learning model was developed for the classification of heart sounds as either normal or abnormal. This model could be used with an electronic stethoscope to analyze a patient’s heart sounds and provide assistance to clinicians. The model was developed using the Random Forest classifier and uses the PhysioNet database. The final model could classify normal and abnormal heart sounds with an F1 score of 0.8773. Important features to the classification included Age, Murmurs, Gender and # Beats.