Early Detection of Chronic Heart Failure from Phonocardiography Data: A Machine Learning and Deep Learning Approach

Abstract

The ability of an experienced physician to detect the progression of chronic heart failure (CHF)
primarily relies on patient examination and changes in heart failure biomarkers, determined from
blood tests. Unfortunately, the clinical deterioration of a CHF patient often signals a fully developed
CHF episode that may necessitate hospitalization. In some cases, distinctive changes in heart sounds
can accompany heart failure progression and are detectable through phonocardiography. Leveraging
recent advancements in machine learning and deep learning, this project introduces an early detection
system for chronic heart failure using phonocardiography (PCG) data. The system utilizes an end-toend average aggregate recording model that incorporates features extracted from both machine
learning and deep learning techniques. The proposed ChronicNet model is compared with individual
machine learning and deep learning models, demonstrating its effectiveness in early CHF detection.