Two-Stage Multitask-Learner for PCG Murmur Location Detection

Heart murmurs are a potential symptom for cardiac diseases, which can be captured easily by smartphones or similar devices. Hence, the automated analysis of murmurs in heart sound recordings may provide a cost-efficient pre-screening method for heart conditions. In this study, we present an approach for detecting heart murmurs that utilizes a Pooling-based Artificial Neural Network (PANN) structure to extract features from audio waveforms of arbitrary lengths. It can classify single recordings based on recording location and the extracted features in an end-to-end manner. The approach is inspired by the multiple instance learning framework. We performed a 10-fold stratified cross-validation on our training set and show that the results are consistent with the evaluation on the hidden test set of the PhysioNet challenge 2022. Our team Heart2Beat was ranked 12
^th
in the murmur detection task and 11
^th
in the clinical out-come task and achieved a weighted accuracy metric score of 0.751 and a clinical outcome cost of 12244 respectively.