A Robust Deep Learning Framework for Real-Time Denoising of Heart Sound

The heart sound signal captured via a digital stethoscope is often corrupted with environmental and physiological noise, altering its salient and critical properties. The problem is exacerbated in crowded low-resource hospital settings with high noise levels. In this study, inspired by the heart sound’s quasi-periodic nature, we present a novel deep encoder-decoder based LU-Net architecture to suppress ambient hospital and internal lung sound noises that corrupt the heart sounds. Training is done using a large benchmark PCG dataset mixed with physiological noise, i.e., breathing sounds. Two different noisy datasets were created for experimental evaluation by mixing unseen lung sounds and hospital ambiance noises with the clean heart sound recordings. We also use the inherently noisy portion of the PASCAL heart sound dataset for evaluation. Experimental results show that the proposed framework can effectively suppress background noise in both real-world and synthetically generated noisy heart sound recordings, improving the signal-to-noise ratio (SNR) level by 5.575 dB on an average using only 1.32 M parameters. The study reveals that the proposed model outperforms the current state-of-the-art U-Net model with an average SNR improvement of 5.613 dB and 5.537 dB in the presence of lung sound and hospital noise, respectively. LU-Net also outperformed the state-of-the-art Fully Convolutional Network (FCN) by 1.750 dB and 1.748 dB for lung sound and hospital noise conditions, respectively. In addition, the proposed denoising method model improves classification accuracy by 38.93% in the noisy portion of the PASCAL heart sound dataset. The proposed enhancement scheme can thus play a vital role in deploying automated cardiac screening systems in low-resource and underserved communities.