A Multi-Label Electrocardiogram Classification Study Integrating Clinical Metadata and Channel Attention

Jiajing Sun; Ying Sun; Jiang Zhao; Wei Yu; Qinyang Li; Shun He

doi:10.63313/JCSFT.9060

Authors

Jiajing Sun College of International Business, Zhejiang Yuexiu University, Shaoxing, 312069, China Author
Ying Sun College of International Business, Zhejiang Yuexiu University, Shaoxing, 312069, China Author
Jiang Zhao College of International Business, Zhejiang Yuexiu University, Shaoxing, 312069, China Author
Wei Yu College of International Business, Zhejiang Yuexiu University, Shaoxing, 312069, China Author
Qinyang Li College of International Business, Zhejiang Yuexiu University, Shaoxing, 312069, China Author
Shun He College of International Business, Zhejiang Yuexiu University, Shaoxing, 312069, China Author

DOI:

https://doi.org/10.63313/JCSFT.9060

Keywords:

Electrocardiogram (ECG), SE Block, Metadata Fusion, SAM Optimizer

Abstract

To address the issues of feature extraction redundancy and low recognition accuracy for ambiguous categories in multi-label electrocardiogram (ECG) automatic classification, this paper proposes a deep learning diagnostic model that integrates clinical metadata with a channel attention mechanism. Ablation experiments indicate that traditional Transformer and multi-scale convolutional modules are susceptible to feature interference in specific scenarios. In this study, the SE Block is employed to enhance channel-wise feature representation, patient physiological metadata is introduced as prior knowledge, and the SAM optimizer is utilized to improve the loss landscape. Experimental results demonstrate that the proposed combined model achieves a Macro-AUC of 0.9251 and a Macro-F1 score of 0.7254 on the PTB-XL dataset. Compared with the baseline model, it achieves significant improvements in the identification of ambiguous categories such as ventricular hypertrophy.

References

[1] Gao, T., Xie, X., Liu, H., Zhou, S., and Shu, M. (2025). a. A Multilevel Metric Fusion Framework for Few-Shot Electrocardiogram Classification. IEEE Transactions on Instrumentation and Measurement 74:1–11. doi:10.1109/TIM.2025.3553245.

[2] Prakash, A.J., and Atef, M. (2025). b. Multi-Scale Feature Extraction for ECG Beat Classification Using a CNN-Transformer Network with Imbalance Mitigation. Pp. 1–5 in 2025 IEEE International Symposium on Circuits and Systems (ISCAS). London, United Kingdom: IEEE.

[3] Jiang, A., Huang, C., Cao, Q., Xu, Y., Zeng, Z., Chen, K., Zhang, Y., and Wang, Y. (2024). c. Self-Supervised Anomaly Detection Pretraining Enhances Long-Tail ECG Diagnosis.

[4] Rahul, J., and Sharma, L.D. (2025). d. Advancements in AI for Cardiac Arrhythmia Detection: A Comprehensive Overview. Computer Science Review 56:100719. doi:10.1016/j.cosrev.2024.100719.

[5] Xie, C.-X., Wang, L.-H., Yu, Y.-T., Ding, L.-J., Yang, T., Kuo, I.-C., Wang, X.-K., Gao, J., and Abu, P.A.R. (2025). e. Clinical Sudden Cardiac Death Risk Prediction: A Grid Search Support Vector Machine Multimodel Base on Ventricular Fibrillation Visualization Features. Computers and Electrical Engineering 123:110022. doi:10.1016/j.compeleceng.2024.110022.

[6] Hu, J., Shen, L., Albanie, S., Sun, G., and Wu, E. (2020). f. Squeeze-and-Excitation Networks. IEEE Transactions on Pattern Analysis and Machine Intelligence 42(8):2011–23. doi:10.1109/TPAMI.2019.2913372.

[7] Foret, P., Kleiner, A., Mobahi, H., and Neyshabur, B. (2020). g. Sharpness-Aware Minimization for Efficiently Improving Generalization.

[8] Wei, Y., and Lian, C. (2024). h. Frequency-Enhanced Hybrid Multimodal CNN-Transformer Network for Electrocardiogram Classification. Pp. 20–24 in 2024 14th International Conference on Information Science and Technology (ICIST). Chengdu, China: IEEE.

[9] Liu, W., Pan, S., Li, Z., Chang, S., Huang, Q., and Jiang, N. (2025). i. Lead-Fusion Barlow Twins: A Fused Self-Supervised Learning Method for Multi-Lead Electrocardiograms. Information Fusion 114:102698. doi:10.1016/j.inffus.2024.102698.

[10] Sarankumar, R., Ramkumar, M., Vijaipriya, K., and Velselvi, R. (2024). j. Bidirectional Gated Recurrent Unit with Auto Encoders for Detecting Arrhythmia Using ECG Data. Knowledge-Based Systems 294:111696. doi:10.1016/j.knosys.2024.111696.

[11] He, K., Zhang, X., Ren, S., and Sun, J. (2016). k. Deep Residual Learning for Image Recognition. Pp. 770–78 in 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Las Vegas, NV, USA: IEEE.

[12] Zhu, H., Cheng, C., Yin, H., Li, X., Zuo, P., Ding, J., Lin, F., Wang, J., Zhou, B., Li, Y., Hu, S., Xiong, Y., Wang, B., Wan, G., Yang, X., and Yuan, Y. (2020). l. Automatic Multilabel Electrocardiogram Diagnosis of Heart Rhythm or Conduction Abnormalities with Deep Learning: A Cohort Study. The Lancet Digital Health 2(7):e348–57. doi:10.1016/S2589-7500(20)30107-2.

[13] Wagner, P., Strodthoff, N., Bousseljot, R.-D., Kreiseler, D., Lunze, F.I., Samek, W., and Schaeffter, T. (2020). m. PTB-XL, a Large Publicly Available Electrocardiography Dataset. Scientific Data 7(1):154. doi:10.1038/s41597-020-0495-6.

[14] Chen, S., Wang, H., Zhang, H., Peng, C., Li, Y., and Wang, B. (2024). n. A Novel Method of Swin Transformer with Time-Frequency Characteristics for ECG-Based Arrhythmia Detection. Frontiers in Cardiovascular Medicine 11:1401143. doi:10.3389/fcvm.2024.1401143.

[15] Loshchilov, I., and Hutter, F. (2017). o. SGDR: Stochastic Gradient Descent with Warm Restarts.

[16] Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A.N., Kaiser, L., and Polosukhin, I. (2023). p. Attention Is All You Need.

[17] Chen, T., Ma, Y., Pan, Z., Wang, W., and Yu, J. (2025). q. Fusion of Multi-Scale Feature Extraction and Adaptive Multi-Channel Graph Neural Network for 12-Lead ECG Classification. Computer Methods and Programs in Biomedicine 265:108725. doi:10.1016/j.cmpb.2025.108725.