Key result
New protocol details self-supervised JEPA pretraining aligning 12-lead ECG segments with clinical reports.
Why the study?
Labeled ECGs are scarce, whereas unlabeled recordings are abundant.
This preprint describes a research protocol for self-supervised JEPA pretraining of ECG representations aligned with clinical reports, though full experimental results are currently pending.
Answers are grounded in this paper and cite their sources.
Pilot ECG-text pretraining shows feasibility; leaves open scalability for low-label cardiac tasks pending full validation.
Labeled electrocardiograms (ECG) are scarce, while unlabeled recordings are abundant. We pretrain a joint-embedding predictive architecture (JEPA) on unlabeled ECG segments by predicting the embeddings of masked leads and segments, and we align the resulting representations with clinical report text. We evaluate on a public 12-lead benchmark with a linear probe and a low-label regime, and on synthetic ECG with controlled anomalies. This is a research protocol, not a clinical tool. A synthetic pilot is too easy to discriminate, so the pretraining results are pending. Preprint. Results in this version come from small-scale pilot experiments; the full experimental protocol is described in the text and is pending.
No takes yet. Share an insight, caveat, or question.
Dheiver Francisco Santos (2026) studied this question. Joint-embedding predictive architecture (JEPA) pretraining on unlabeled ECG segments aligned with clinical report text was evaluated on Linear probe and low-label regime evaluation on a public 12-lead ECG benchmark and synthetic ECG. Self-supervised JEPA pretraining on 12-lead ECG segments aligned with clinical reports is described in a research protocol, with full pretraining results currently pending.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: