HeartBeatNet: Unleashing the Power of Attention in Cardiology

Cardiovascular disease diagnosis and prompt medical care depend critically on the classification of heart sounds. In recent years, deep learning-based approaches have shown promising results in automating the process of heart sound categorization. This paper proposes a model HeartBeatNet (an attention UNet-based system) for heart sound classification that demonstrates comparatively better performance. The proposed system combines the strengths of attention mechanisms and the UNet architecture to effectively capture relevant features and to make accurate predictions. The system is trained on the PhysioNet/Cinc 2016 dataset consisting of annotated heart sound recordings, which are first converted into Mel Spectrograms before feeding into the UNet based network. The results indicate that the proposed system achieves high accuracy of 95.14%, sensitivity of 90.00%, and specificity of around 96.72% to classify various heart sound abnormalities.