Early Diagnosis of Heart Disease Via Google’s Auto ML

Abstract

Coronary heart disease is a leading cause of death around the world. Early and accurate detection of this disease is vital to reducing complications and preventing heart failure and heart attacks. This project creates an early-level diagnosis method that analyzes heart sound audio to detect and characterize heart anomalies including murmurs, arrhythmias, and extra heart sounds. The dataset was obtained from Peter Bently’s Classifying Heart Sounds Database. This dataset was used to train a machine learning algorithm in Google’s AutoML Vision that could distinguish between normal heart sounds, murmurs, arrhythmias, and extra heart sounds. The final model has an average precision of 0.929, thus providing accurate results. This product can be used as a cheap, accessible, and non-invasive diagnosis method for the early detection of cardiovascular diseases.