University of Hong Kong (HKU-MedAI)
Multi-scale ECG-language pretraining model that aligns 12-lead ECG signals with clinical text reports at three granularities -- token, beat, and rhythm level -- rather than a single global embedding. First fine-tunes a cardiology-specialized text encoder to improve understanding of ECG report language, then trains an ECG-FM-initialized ECG encoder against it with hierarchical contrastive supervision. Outperforms prior ECG-language and self-supervised baselines including MERL, ST-MEM, and HeartLang on zero-shot classification, linear probing, and ECG report generation, with especially large gains at low label fractions. Developed at the University of Hong Kong (HKU-MedAI).
Architecture
Hybrid
Multi-scale ECG-language contrastive pretraining: ECG-FM-style transformer ECG encoder plus a cardiology-specialized text encoder, aligned via token-, beat-, and rhythm-level cross-modal contrastive losses
Framework
PyTorch
Added to catalog
2026-08-10
Apache 2.0
License for model weights only. Associated code may be licensed seperately, check code source for specific terms.
12-lead ECG multi-label arrhythmia-classification dataset released for the 2018 China Physiological Signal Challenge.
45,152 12-lead, 10-second ECGs from Chapman University / Shaoxing People's Hospital / Ningbo First Hospital with arrhythmia diagnoses; ~56% male / 44% female.
800,035 12-lead ECG-report pairs from 161,352 subjects at Beth Israel Deaconess Medical Center.
52% male / 48% female; age range 0-95 (median ~62). 21,837 10-second 12-lead ECG records.
ECG-text joint embedding used for zero-shot / linear-probe classification of cardiac abnormalities
Zero-shot and linear-probe multi-label ECG diagnostic classification (PTB-XL, CPSC2018, Chapman-Shaoxing)