Peking University (PKUDigitalHealth)
ECG-guided photoplethysmography (PPG) foundation model pretrained on over 100,000 hours of synchronized PPG-ECG recordings from 58,796 subjects across five clinical and wearable sources, using a CLIP-style contrastive alignment framework so the PPG encoder inherits physiologically grounded structure from paired ECG. Achieves state-of-the-art performance on 13 of 15 conventional physiological-analysis tasks across eight datasets, and shows meaningful discriminative capability (AUC >= 0.70) for 307 ICD-10-coded phenotypes across 16 phecode chapters, including many non-cardiovascular conditions.
Architecture
CNN (1D)
1D ResNet (Net1D, 64->512 channel progression across 6 stages) PPG encoder, CLIP-style contrastively pretrained against a synchronized ECG encoder on >100,000 hours of paired PPG-ECG recordings
Framework
PyTorch
Added to catalog
2026-08-10
Multi-Source PPG-ECG Pretraining Corpus (CFS/HSP/MC-MED/MESA/PulseDB)
Aggregate of five clinical/wearable sources (Cleveland Family Study, Human Sleep Project, MC-MED emergency-department dataset, MESA, PulseDB) totaling over 100,000 hours of synchronized PPG-ECG recordings from 58,796 subjects, used for ECG-guided contrastive pretraining of the AnyPPG encoder.
General-purpose PPG representation embedding (ECG-guided contrastive pretraining)
Atrial fibrillation detection from PPG (linear probing on pretrained embedding)
Blood pressure estimation from PPG (linear probing on pretrained embedding)