Robust Transfer Learning for Paper ECG Recognition

2026-09-30

AI Summary

This study addresses the recognition challenges of real-world paper electrocardiograms (ECGs) arising from heterogeneous layouts, physical artifacts, and label scarcity by proposing the RobECG-CL framework. The method constructs progressive degradation views and introduces rank-aware contrastive learning with multi-view consistency constraints, driving the model to simultaneously achieve intra-record invariance and degradation-aware ranking for robust representation learning. Experiments demonstrate that this framework effectively overcomes heterogeneous layouts and severe degradation interference. It significantly enhances robustness on the CODE-II and EchoNext benchmarks, surpasses existing waveform foundation models in few-shot scenarios with only 1% labeled data, and achieves state-of-the-art macro-average AUROC on real-world hospital datasets.

Abstract

Paper ECG recognition is challenging because real-world ECG images vary in layout, physical artifacts, and label availability. We introduce RobECG-CL, a rank-aware contrastive learning framework for robust paper ECG representation learning. Starting from standard 12-lead ECG recordings, we construct progressively degraded paper ECG views with heterogeneous layouts and train the model to balance same-recording invariance with degradation-aware ordering. Across synthetic stress tests on CODE-II and EchoNext, RobECG-CL improves robustness under severe degradation and few-shot transfer, outperforming contrastive learning baselines and surpassing the waveform-based foundation model, ECG-FM, in the 1% labeled setting. On 312 samples of hospital data with 37 labels, RobECG-CL achieves the best macro AUROC.

Problem

• Paper ECG recognition
• Transfer learning
• Physical artifacts
• Layout variation
• Label scarcity

Innovation

• Contrastive Learning
• Rank-Aware
• Paper ECG Recognition
• Transfer Learning
• Few-Shot