3KG achieved an AUROC of 0.826 for 23-class ECG diagnosis when trained on 1% of labeled data, outperforming the best self-supervised baseline which achieved an AUROC of 0.757.
Does 3KG contrastive learning improve diagnostic performance on 12-lead ECGs compared to baseline self-supervised methods?
A physiologically-inspired contrastive learning approach using 3D augmentations of 12-lead ECGs significantly improves diagnostic performance, particularly for conduction and rhythm abnormalities, when labeled data is limited.
Absolute Event Rate: 0.826% vs 0.757%
We propose 3KG, a physiologically-inspired contrastive learning approach that generates views using 3D augmentations of the 12-lead electrocardiogram. We evaluate representation quality by fine-tuning a linear layer for the downstream task of 23-class diagnosis on the PhysioNet 2020 challenge training data and find that 3KG achieves a 9. 1\% increase in mean AUC over the best self-supervised baseline when trained on 1\% of labeled data. Our empirical analysis shows that combining spatial and temporal augmentations produces the strongest representations. In addition, we investigate the effect of this physiologically-inspired pretraining on downstream performance on different disease subgroups and find that 3KG makes the greatest gains for conduction and rhythm abnormalities. Our method allows for flexibility in incorporating other self-supervised strategies and highlights the potential for similar modality-specific augmentations for other biomedical signals.
Gopal et al. (Wed,) conducted a other in Arrhythmia (n=43,134). 3KG (physiologically-inspired contrastive learning) vs. Patient Contrastive Learning (best self-supervised baseline) was evaluated on Mean AUROC for 23-class diagnosis on 1% labeled data. 3KG achieved an AUROC of 0.826 for 23-class ECG diagnosis when trained on 1% of labeled data, outperforming the best self-supervised baseline which achieved an AUROC of 0.757.