Abstract:

CHF, a grave medical condition that impacts millions of individuals globally, is becoming increasingly prevalent at a rate of 2% every year. Despite the widespread availability of sensors and technologies, automated detection of CHF remains a challenge. In response, this study proposes a new method for CHF detection using heart sounds, which combines traditional machine learning (ML) with convolutional neural networks (CNNs). By analysing recordings of heart sounds, this model can detect subtle changes in the sound patterns that may be difficult for a human to detect. The usefulness of this technique goes beyond identifying CHF in both healthy individuals and patients. The research discovered 15 specialized characteristics that can distinguish between different stages of CHF with an impressive precision of 93.2%. This development offers hope for the development of homebased CHF monitors to prevent hospitalization. This study highlights the importance of combining traditional ML with advanced techniques such as CNNs in improving the accuracy and efficiency of CHF detection. The promising results of this method demonstrate its potential for the detection of fresh CHF cases and the creation of CHF monitoring systems that can be used at home. Further research and development of this method could lead to the development of more effective and accessible CHF detection and monitoring tools.

Description:

Chronic heart failure, heart sounds, machine learning, convolutional neural networks, PCG