Optimizing Reduced-Lead ECG Diagnosis: An Interpretable Pipeline for Lead Selection and Model Adaptation

• Luisa G. Porfírio UFMG
• Guilherme H. G. Evangelista UFMG
• Pedro B. Rigueira UFMG
• Caio Souza Grossi UFMG
• Artur Xavier UFMG
• Victoria Andrade Flores de Mello UFMG
• Raquel Teodoro UFMG
• Pedro Dutenhefner UFMG
• Gabriela M. M. Paixão UFMG
• Gisele L. Pappa UFMG
• Antonio Ribeiro UFMG
• Wagner Meira Jr. UFMG

DOI: https://doi.org/10.5753/eniac.2025.14543

Abstract

The 12-lead electrocardiogram (ECG) is vital for heart diagnosis but mostly limited to clinics due to its complexity. Wearable devices use fewer electrical viewpoints (leads), democratizing cardiac care by enabling monitoring at home or in low-resource settings. This shift raises two key challenges: selecting the most informative leads and adapting AI models to keep accuracy. We tackle this with a data-driven pipeline that ranks leads by combining multiple model-interpretability methods. Our evaluation shows that a model architecturally adapted to use only the top two leads (V1, I) achieves a macro F1-score of 0.885, matching the full 12-lead baseline. This work provides a framework for efficient, powerful AI systems, advancing accessible cardiac diagnostics.

References

Buzelin, A., Dutenhefner, P. R., Rezende, T., Porfirio, L. G., Bento, P., Aquino, Y., Fernandes, J., Santana, C., Miana, G., Pappa, G. L., Ribeiro, A., and Jr, W. M. (2025). A cnn-based local-global self-attention via averaged window embeddings for hierarchical ecg analysis. arXiv preprint, arXiv:2504.16097.

Gradowski, T. and Buchner, T. (2025). Deep learning model for ecg reconstruction reveals the information content of ecg leads. arXiv preprint, arXiv:2502.00559.

Grand View Research (2025). Smart wearable ecg monitors market size, share & trends analysis report, 2030. Accessed 23 June 2025.

Hannun, A. Y., Rajpurkar, P., Haghpanahi, M., Tison, G. H., Bourn, C., Turakhia, M. P., and Ng, A. Y. (2019). Cardiologist-level arrhythmia detection and classification in ambulatory electrocardiograms using a deep neural network. Nature Medicine, 25(1):65–69.

Oh, J., Chung, H., myoung Kwon, J., gyun Hong, D., and Choi, E. (2022). Lead-agnostic self-supervised learning for local and global representations of electrocardiogram. arXiv preprint, arXiv:2203.06889.

Ramirez, E., Ruiperez-Campillo, S., Casado-Arroyo, R., and Merino, J. L. (2024). The art of selecting the ECG input in neural networks to classify heart diseases: maximizing information and reducing redundancy. Frontiers in Physiology, 15:1452829.

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.

Ribeiro, A. H., Ribeiro, M. H., Paixão, G. M., Oliveira, D. M., Gomes, P. R., Canazart, J. A., Ferreira, M. P., Andersson, C. R., Macfarlane, P. W., Meira Jr., W., Schön, T. B., and Ribeiro, A. L. P. (2020). Code-test: An annotated 12-lead ecg dataset.

Rigueira, P. B., Evangelista, G. H. G., Porfírio, L. G., Grossi, C. S., Buzelin, A., Pappa, G. L., Paixão, G. M. M., Ribeiro, A., and Jr., W. M. (2024). Optimizing ecg audits: Clustering-based identification of ambiguous exams. Encontro Nacional de Inteligência Artificial e Computacional (ENIAC), 21:61–72.

Sundararajan, M., Taly, A., and Yan, Q. (2017). Axiomatic attribution for deep networks. Proceedings of the 34th International Conference on Machine Learning, 70:3319–3328.

Tsouri, G. R. and Ostertag, M. H. (2014). Patient-specific 12-lead ecg reconstruction from sparse electrodes using independent component analysis. IEEE Journal of Biomedical and Health Informatics, 18(2):476–481.

Vijayarangan, S., Murugesan, B., R, V., P., P. S., Joseph, J., and Sivaprakasam, M. (2020). Interpreting deep neural networks for single-lead ecg arrhythmia classification. arXiv preprint, arXiv:2004.05399.

Wagner, P., Strodthoff, N., Bousseljot, R.-D., Samek, W., and Schaeffter, T. (2020). Ptb-xl, a large publicly available electrocardiography dataset. Scientific Data, 7(1):154.

World Health Organization (2024). Cardiovascular diseases (cvds) – fact sheet. Accessed 23 June 2025.