Reproduce “MS-ECG-FM: Towards a More Universal Electrocardiogram Foundation Model for Health Monitoring using Multi-source Contrastive Learning” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Robert A. Lewis, I-Min Chiu, Kyle Verrier, Karthik Jayaraman Raghuram, Francoise Marvel, Salar Abbaspourazad, Anshuman Mishra, Guillermo Sapiro and 2 more
MS-ECG-FM trains ECG representations by aligning recordings with one of five report types from the same clinical encounters. It uses stochastic report selection and evaluates frozen representations with linear probes across 98 binary 12-lead labels. It reaches macro-averaged AUROC 93.6, versus 91.5 for ECGFounder and 91.2 for MELP; a report-only variant scores 92.8, so diversity alone does not explain cross-model gains.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.