Technical University of Munich / Imperial College London
Deep learning strategy for cost-effective, comprehensive cardiac screening from ECG alone, by transferring domain-specific structural information from cardiac magnetic resonance (CMR) imaging into ECG representations. Combines multimodal contrastive learning with masked data modelling during pretraining on paired ECG-CMR data, then uses only ECG at inference. On 40,044 UK Biobank subjects, the multimodal pretraining improved subject-specific CVD risk prediction by up to 12.19% and cardiac phenotype prediction by up to 27.59% versus ECG-only baselines, with learned ECG representations shown to incorporate information from CMR regions of interest.
Architecture
Hybrid
Multimodal contrastive learning combined with masked data modelling (MAE-style) to transfer CMR-derived structural information into an ECG encoder during pretraining; only the ECG branch is used at inference
Framework
PyTorch
Added to catalog
2026-08-14
MIT
License for model weights only. Associated code may be licensed seperately, check code source for specific terms.
UK Biobank Paired ECG-CMR Subset (MMCL-ECG-CMR)
Subjects with paired 12-lead ECG and CMR from UK Biobank
ECG-derived representation used for CVD risk prediction and cardiac phenotype (CMR-derived measurement) prediction from ECG alone