Single-Lead Abnormal ECG Screening with a Lightweight Deep Learning Model for Wearable Health Monitoring

• Estela Ribeiro USP
• Quenaz Bezerra Soares USP
• Douglas de Andrade de Almeida USP
• Fabio Jatene USP
• Marco Antonio Gutierrez USP

DOI: https://doi.org/10.5753/sbcas.2026.20274

Resumo

Electrocardiograms (ECGs) are essential for cardiac assessment, and the rising adoption of wearables has intensified the need for automated identification of abnormal ECGs on resource-constrained platforms. We evaluate LiteVGG-11, a lightweight CNN (20,365 parameters), for binary classification using Lead II signals. The model was trained on CODE15 and validated on PTB-XL and real-world wearable data, achieving AUCs of 0.736 and 0.773, respectively. Wearable validation revealed a decline in specificity due to artifacts and Lead II’s spatial constraints. Findings indicate that LiteVGG-11 is feasible for preliminary screening in continuous health monitoring, despite limitations in identifying non-specific abnormalities that require multi-lead observation.

Referências

Bahuguna, R., Upadhyay, S., Aditi, Kumar, V., Saini, A., and Jain, A. (2023). Normal and abnormal ecg signal classification using deep learning. In 2023 International Conference on Computational Intelligence and Sustainable Engineering Solutions (CISES), pages 645–649.

Dias, F. M., Ribeiro, E., Moreno, R. A., Ribeiro, A. H., Samesima, N., Pastore, C. A., Krieger, J. E., and Gutierrez, M. A. (2023). Artificial intelligence-driven screening system for rapid image-based classification of 12-lead ecg exams: A promising solution for emergency room prioritization. IEEE Access, 11:121739–121752.

Goldberger, A. L., Goldberger, Z. D., and Shvilkin, A. (2018). Chapter 5 - the normal ecg. In Goldberger, A. L., Goldberger, Z. D., and Shvilkin, A., editors, Goldberger’s Clinical Electrocardiography (Ninth Edition), pages 32–40. Elsevier, 9th edition.

Ribeiro, A. H., Paixao, G. M., Lima, E. M., Horta Ribeiro, M., Pinto Filho, M. M., Gomes, P. R., Oliveira, D. M., Meira Jr, W., Schon, T. B., and Ribeiro, A. L. P. (2021). Code-15%: a large scale annotated dataset of 12-lead ecgs.

Soares, Q. B., Andrade, D. A., Ribeiro, E., Verardino, R. G. S., Reis, T. C., Samesima, N., Monteiro, R., Jatene, F. B., and Gutierrez, M. A. (2024). Clinical Assessment of a Lightweight CNN Model for Real-Time Atrial Fibrillation Prediction in Continuous Wearable Monitoring. In 2024 46th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), pages 1–4.

Soares, Q. B., Monteiro, R., Jatene, F. B., and Gutierrez, M. A. (2022). A Lightweight Unidimensional Deep Learning Model for Atrial Fibrillation Detection. In 2022 Computing in Cardiology (CinC), volume 498, pages 1–4. IEEE.

Wagner, P., Strodthoff, N., Bousseljot, R.-D., Kreiseler, D., Lunze, F. I., Samek, W., and Schaeffter, T. (2020). PTB-XL, a large publicly available electrocardiography dataset. Scientific Data, 7(1):154.

Zhu, J., Lv, J., and Kong, D. (2022). CNN-FWS: A Model for the Diagnosis of Normal and Abnormal ECG with Feature Adaptive. Entropy, 24(4).