Daily Arxiv

This is a page that curates AI-related papers published worldwide.
All content here is summarized using Google Gemini and operated on a non-profit basis.
Copyright for each paper belongs to the authors and their institutions; please make sure to credit the source when sharing.

TolerantECG: A Foundation Model for Imperfect Electrocardiogram

Created by
  • Haebom

Author

Huynh Dang Nguyen, Trong-Thang Pham, Ngan Le, Van Nguyen

Outline

This paper proposes TolerantECG, a noise-tolerant, operational-ready baseline model that addresses the issues of noise and lead absence in electrocardiogram (ECG) signals. Combining contrastive learning and self-supervised learning frameworks, TolerantECG learns ECG signal representations, corresponding knowledge-retrieval-based textual descriptions, and signals with damaged or missing leads. Experimental results demonstrate excellent performance across a variety of ECG signal conditions and class levels on the PTB-XL dataset and the MIT-BIH arrhythmia database.

Takeaways, Limitations

•
Takeaways:
◦
Contributes to improving diagnostic accuracy by proposing an electrocardiogram analysis-based model that is robust to noise and lead absence.
◦
Improving ECG signal representation learning and generalization performance through an effective combination of contrastive learning and self-supervised learning.
◦
Superior performance compared to existing methods was verified on the PTB-XL and MIT-BIH Arrhythmia Databases.
•
Limitations:
◦
Further validation of the generalization performance of the proposed model is needed (in various datasets and clinical settings).
◦
Further research is needed to determine the interpretability and reliability of the model.
◦
Additional validation and clinical trials are needed for application in real clinical settings.
👍
Made with Slashpage