University of Oxford / University of Surrey
Model weights not public. Contact creators for more information.
Dual Bayesian ResNet model for heart murmur detection from multi-location phonocardiogram (PCG) recordings, developed for the George B. Moody PhysioNet Challenge 2022. Each patient's PCG recordings are segmented into overlapping log-mel spectrograms, which are passed through two Bayesian ResNet binary classifiers running simultaneously (present vs. unknown-or-absent, and unknown vs. present-or-absent); the two outputs are aggregated into a patient-level present/unknown/absent murmur classification. An optional second-stage XGBoost model integrates the DBRes output with demographic data and hand-crafted signal features. On the Challenge's official hidden test set, DBRes achieved a weighted accuracy of 0.771 for the murmur-detection task, placing 4th among all teams.
Architecture
CNN (2D)
Dual Bayesian ResNet operating on overlapping log-mel spectrogram segments of PCG recordings, with two simultaneous binary classification heads (present vs. rest, unknown vs. rest) aggregated into a 3-class patient-level prediction; optionally fused with demographic/signal features via XGBoost
Framework
PyTorch
Added to catalog
2026-08-12
Phonocardiogram recordings from up to four auscultation locations (aortic, pulmonic, tricuspid, mitral valve areas) collected with a Littmann 3200 digital stethoscope during two pediatric screening campaigns in Paraiba, Brazil; largely pediatric population (neonates to adolescents) plus some pregnant adults, with expert murmur and clinical-outcome annotations.
Three-class classification of heart murmur as present, absent, or unknown from multi-location phonocardiogram recordings