Key result
TransECG-SSL improves ECG transfer balanced accuracy by ~16 percentage points over source-supervised pretraining.
Why the study?
Cross-dataset generalisation remains a major challenge for automated ECG arrhythmia classification due to differences across datasets, motivating the development of transferable representation learning methods that reduce reliance on dataset-specific labelled data.
Does a hybrid self-supervised Transformer representation learning framework (TransECG-SSL) improve cross-dataset ECG arrhythmia classification performance compared to single-objective SSL, source-supervised, and randomly initialized baselines under patient-disjoint evaluation?
Does a hybrid self-supervised Transformer representation learning framework (TransECG-SSL) improve cross-dataset ECG arrhythmia classification performance compared to single-objective SSL, source-supervised, and randomly initialized baselines under patient-disjoint evaluation?
Hybrid self-supervised learning combining masked signal reconstruction and contrastive learning enables label-efficient, transferable ECG arrhythmia representations across independent datasets, though domain shift and severe class imbalance remain major challenges.
Answers are grounded in this paper and cite their sources.
Enhances ECG transfer learning; leaves open clinical utility pending prospective validation.
Cross-dataset generalisation remains a major challenge for automated electrocardiogram (ECG) arrhythmia classification because of differences in patient populations, acquisition protocols, annotation practices, and class distributions. Learning representations that remain informative across independent datasets could reduce reliance on dataset-specific labelled data. We propose TransECG-SSL, a self-supervised framework for transferable heartbeat representation learning from unlabelled ECG data. A Transformer encoder is pretrained using a hybrid objective combining masked signal reconstruction and contrastive learning, then frozen and evaluated by linear probing under patient-disjoint evaluation. Experiments across three public ECG databases and seven source--target settings show that Hybrid SSL consistently outperforms randomly initialised Transformers and achieves higher balanced accuracy than architecture-matched source-supervised pretraining. Reconstruction-only SSL remains competitive with Hybrid SSL, whereas contrastive-only pretraining performs substantially worse, highlighting the importance of masked reconstruction for transferability. Cross-dataset transfer nevertheless reduces balanced accuracy relative to target-specific self-supervised pretraining, while minority arrhythmia classes remain challenging under severe class imbalance. These findings demonstrate the potential of label-efficient self-supervised learning for cross-dataset ECG representation transfer while highlighting persistent limitations from domain shift and class imbalance.
No takes yet. Share an insight, caveat, or question.
Ayyub et al. (2026) studied Cardiac arrhythmia (n=157). TransECG-SSL (Hybrid self-supervised Transformer combining masked signal reconstruction and contrastive learning) vs. Architecture-matched source-supervised, reconstruction-only SSL, contrastive-only SSL, and randomly initialized frozen Transformer baselines was evaluated on Balanced accuracy under patient-disjoint linear-probe evaluation (INCART to SVDB cross-dataset transfer). TransECG-SSL achieved higher balanced accuracy than source-supervised pretraining across all 7 ECG transfer routes, peaking at 0.589 vs 0.429 for INCART-to-SVDB transfer.
Synapse has enriched 10 related papers. Consider them for comparative context: