Key result
The MINA deep learning model achieved a PR-AUC of 0.9436 for predicting atrial fibrillation from ECG signals, outperforming the best baseline model by 5.51% while maintaining interpretability and robustness against noise.
Why the study?
Deep learning models for ECG data are mostly black-box and lack the interpretability required for clinical usage.
Effect estimate: 5.51% improvement over best baseline
Absolute Event Rate: 0.9436% vs 0.8943%
The MINA deep learning model provides highly interpretable and robust prediction of heart diseases from ECG signals, outperforming baseline models.
Electrocardiography (ECG) signals are commonly used to diagnose various cardiac abnormalities. Recently, deep learning models showed initial success on modeling ECG data, however they are mostly black-box, thus lack interpretability needed for clinical usage. In this work, we propose MultIlevel kNowledge-guided Attention networks (MINA) that predict heart diseases from ECG signals with intuitive explanation aligned with medical knowledge. By extracting multilevel (beat-, rhythm- and frequency-level) domain knowledge features separately, MINA combines the medical knowledge and ECG data via a multilevel attention model, making the learned models highly interpretable. Our experiments showed MINA achieved PR-AUC 0.9436 (outperforming the best baseline by 5.51%) in real world ECG dataset. Finally, MINA also demonstrated robust performance and strong interpretability against signal distortion and noise contamination.
No takes yet. Share an insight, caveat, or question.
Hong et al. (2019) studied Atrial Fibrillation (n=8,528). MINA (Multilevel Knowledge-Guided Attention networks) vs. Baseline machine learning models (ExpertLR, ExpertRF, CNN, CRNN, ACRNN) was evaluated on Precision-Recall Area Under the Curve (PR-AUC) for atrial fibrillation prediction (5.51% improvement over best baseline). The MINA deep learning model achieved a PR-AUC of 0.9436 for predicting atrial fibrillation from ECG signals, outperforming the best baseline model by 5.51% while maintaining interpretability and robustness against noise.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: