Chronic Heart Failure Diagnosis from Heart Sounds Using Machine Learning and Full-Stack Deep Learning

Abstract

There are presently approximately 26 million people suffering from chronic heart failure all over the globe. More than 1 million people are hospitalized every year because of it, and it has a significant role in the mortality rate among those with cardiovascular illnesses. Every year in North America and Western Europe. Preventative measures, enhanced early identification, and a lack of hospitalisation or even life-threatening circumstances are all possible outcomes of research into the detection of chronic heart failure. In this study, we provide a machine-learning approach to identifying chronic heart failure using ECG recordings. The process includes steps like as filtering, segmenting, feature extraction, and machine learning. Data from 122 participants were used to evaluate the strategy using a leave-one-subject-out design.The approach was 15 percentage points more accurate than a majority classifier, with a success rate of 96%. Specifically, it had an 87% recall rate among those with chronic heart failure. The research verified that chronic heart failure may be detected using powerful machine learning applied to real-world sounds acquired with a discrete digital stethoscope.