Logo PTI Logo FedCSIS

Position Papers of the 21st Conference on Computer Science and Intelligence Systems

Annals of Computer Science and Information Systems, Volume 48

Cross-Dataset Component Ablation of a Hybrid Model for Robust Beat-Level ECG Arrhythmia Classification

, , , , ,

DOI: http://dx.doi.org/10.15439/2026F6565

Citation: Jonas Mindaugas Rimšelis, , , , ,

Full text

Abstract. This study presents a systematic cross-dataset ablation analysis of a hybrid model for beat-level electrocardiogram (ECG) arrhythmia classification under domain shift. The proposed multi-branch architecture combines a mixture-of-experts RR encoder with waveform branches that capture local QRS morphology, global beat context, and frequency-domain information. The full model and its ablated variants were trained over five random seeds on the MIT-BIH Arrhythmia Database under the inter-patient paradigm and evaluated on both internal test subsets and an external dataset. The results indicate that the RR encoder contributes more to robustness than the waveform encoder, especially under stronger domain shift. Furthermore, RR interval-based Feature-wise Linear Modulation (FiLM) and temporal attention have a limited impact on within-dataset evaluation but become more important in cross-dataset evaluation.

References

  1. Q. Xiao, K. Lee, S. A. Mokhtar, I. Ismail, A. L. b. M. Pauzi, Q. Zhang, and P. Y. Lim, “Deep learning-based ECG arrhythmia classification: A systematic review,” Applied Sciences, vol. 13, no. 8, 2023. https://dx.doi.org/10.3390/app13084964
  2. F. yan Zhou, Y. hao Sun, and Y. wen Wang, “Inter-patient ECG arrhythmia heartbeat classification network based on multiscale convolution and FCBA,” Biomedical Signal Processing and Control, vol. 90, p. 105789, 2024. https://dx.doi.org/10.1016/j.bspc.2023.105789. [Online]. Available: https://www.sciencedirect.com/science/article/pii/S1746809423012223
  3. J. Cai, J. Song, and B. Peng, “Enhancing ECG heartbeat classification with feature fusion neural networks and dynamic minority-biased batch weighting loss function,” Physiological Measurement, vol. 45, no. 7, p. 075002, jul 2024. https://dx.doi.org/10.1088/1361-6579/ad5cc0
  4. Z. Qiu, S. Tang, H. Liu, X. Zhao, and J. Lin, “Cdlr-net: a ecg classification network based on deep residual shrinkage networks and lstm,” Computer methods in biomechanics and biomedical engineering, pp. 1–15, 09 2025. https://dx.doi.org/10.1080/10255842.2025.2554260
  5. F. Zhou and J. Wang, “Heartbeat classification method combining multibranch convolutional neural networks and transformer,” iScience, vol. 27, no. 3, p. 109307, 2024. https://dx.doi.org/10.1016/j.isci.2024.109307
  6. H. Di and X. Li, “Arrhythmia detection algorithm based on domain knowledge and spatio-temporal network model,” in 2025 IEEE 5th International Conference on Electronic Technology, Communication and Information, ICETCI 2025, 2025. https://dx.doi.org/10.1109/ICETCI64844.2025.11084081 p. 252 – 255.
  7. R. Li, Y. Aierken, Y. Xu, J. Liu, and Y. Tang, “Research on crossdataset cardiac signal domain generalization and feature interpretability,” Scientific Reports, pp. 1–23, 2026. https://dx.doi.org/10.1038/s41598-025-33057-9
  8. Z. Kotsialou, N. Makris, and S. Gall, “Fundamentals of the electrocardiogram and common cardiac arrhythmias,” Anaesthesia & Intensive Care Medicine, vol. 25, no. 3, pp. 219–222, 2024. https://dx.doi.org/10.1016/j.mpaic.2023.11.014
  9. H. Wang, H. Shi, X. Chen, L. Zhao, Y. Huang, and C. Liu, “An Improved Convolutional Neural Network Based Approach for Automated Heartbeat Classification,” Journal of medical systems, vol. 44, no. 2, p. 35, 2019. https://dx.doi.org/10.1007/s10916-019-1511-2
  10. T. Lan, Q. Hu, X. Liu, K. He, and C. Yang, “Arrhythmias classification using short-time fourier transform and GAN based data augmentation,” in Proceedings of the Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBS, vol. 2020-July, 2020, Conference paper. https://dx.doi.org/10.1109/EMBC44109.2020.9176733 p. 308 – 311.
  11. J. Zang, J. Wang, Z. Zhang, Y. Zheng, and C. Xue, “Improving ECG classification performance by using an optimized one-dimensional residual network model,” Applied Sciences, vol. 12, no. 24, 2022. https://dx.doi.org/10.3390/app122412957
  12. A. J. Prakash and M. Atef, “A lightweight deep learning approach for patient-specific electrocardiogram beat classification using local and long-term dependencies,” Engineering Applications of Artificial Intelligence, vol. 152, no. April, 2025. https://dx.doi.org/10.1016/j.engappai.2025.110754
  13. S. Chon, K.-W. Ha, S. Park, and S. Jung, “An ECG beat classification method using multi-kernel resnet with transformer,” in Proceedings 2023 IEEE International Conference on Big Data and Smart Computing, BigComp 2023, 2023. https://dx.doi.org/10.1109/BigComp57234.2023.00031 p. 140 – 144.
  14. P. de Chazal, M. O’Dwyer, and R. Reilly, “Automatic classification of heartbeats using ECG morphology and heartbeat interval features,” IEEE Transactions on Biomedical Engineering, vol. 51, no. 7, pp. 1196–1206, 2004. https://dx.doi.org/10.1109/TBME.2004.827359
  15. Z. Ran, M. Jiang, Y. Li, Z. Wang, Y. Wu, and W. Ke, “Arrhythmia classification based on multi-feature multi-path parallel deep convolutional neural networks and improved focal loss,” vol. 21, no. March, pp. 5521–5535, 2024. https://dx.doi.org/10.3934/mbe.2024243
  16. M. M. Farag, “A tiny matched filter-based CNN for inter-patient ecg classification and arrhythmia detection at the edge,” Sensors, vol. 23, no. 3, 2023. https://dx.doi.org/10.3390/s23031365