Heart "mur-mur" detection from auscultation recordings using wavelet transform based features and convolutional neural network: "George B. Moody Physionet Challenge 2022" experience

Abstract

Abstract: Physionet, web-based archive of biomedical and physiological signals, and Computing in cardiology conferense have organized annual cardiological origin data analysis challanges for more than 20 years. The aim of the challanges is to encourage participants to develop algorithmic approaches tackling clinically interesting questions that remain unsolved. Challenge 2022 was focused on heart murmur (an extra, unusual sound in heartbeat caused by abnormal blood flow over heart valves) detection from phonocardiogram recordings. Participants were asked to design and implement a working, open-source algorithm that, based only on the provided recordings and routine demographic data, can determine whether any murmurs are audible from a patient’s recordings. The lack of early diagnoses of these conditions represents a major public health problem, especially in underprivileged countries with high birth rates. Our approach to the problem consisted of continuous wavelet transform (CWT) technique to form features from the given data set and convolutional neural network (CNN) to classify patients to “murmur-absent” or “murmur-present”. A total of 87 teams submitted 779 algorithms during the course of the Challenge, including 81 teams with 167 successful entries during the unofficial phase and 63 teams with 306 successful entries during the official phase. Only 40 teams had a successful entry for the murmur detection task on the test set for the murmur detection task and met the other Challenge criteria for ranking. Our teams proposed algorithm was evaluated on the hidden test set and received a weighted accuracy score of 0.671 (ranked 23th out of 40 teams). The highest weighted accuracy metric score was received by team “HearHeart” 0.780. Proposed algorithms can lower healthcare costs and increase the accessibility of cardiac screening and care for patients with abnormal cardiac function in low-resourced environments.