Detection of Heart Murmurs in Phonocardiograms with Parallel Hidden Semi-Markov Models

Andrew McDonald, Mark Gales, Anurag Agarwal
University of Cambridge


Abstract

Introduction: We describe a recurrent neural network and hidden semi Markov model (HSMM) approach to detect heart murmurs in phonocardiogram recordings. This model forms the `CUED_Acoustics' entry to the 2022 George B. Moody PhysioNet challenge.

Methods: Segmentation of the phonocardiogram is a key pre-processing step for many heart sound algorithms. However, most previous work assumes that heart sound recordings only contain S1 and S2 sounds, leading to poorer segmentations of signals that contain a strong murmur. Our approach applies multiple HSMMs, each making different assumptions about a possible murmur, to produce multiple segmentations of the signal. By comparing the confidence of each HSMM's output, we simultaneously produce a murmur classification and robust segmentation.

Results: On the murmur detection task, our algorithm achieved a training cross-validation score of 0.799 and a validation score of 0.758 (ranked 6th out of 61 teams). On the clinical outcomes task, we achieved a training score of 11040 and a validation score of 9257 (ranked 9th out of 61 teams).

Conclusion: The algorithm is highly sensitive (92.7%) to murmurs and, compared to end-to-end models, provides interpretable results about their location and timing. This makes it a promising tool for symptomatic screening.