ECG-Based Arrhythmia Classification with an Explainable Two-Stage Hybrid Learning Approach

Abstract

Cardiovascular diseases are one of the leading causes of mortality worldwide, and arrhythmia detection is crucial for early diagnosis. In this study, an explainable and computationally efficient two-stage hybrid learning model is proposed for arrhythmia detection from ECG (Electrocardiogram) signals. The proposed framework combines classical statistical feature extraction methods (wavelet, Welch, RR interval), a CatBoost-based explainable classifier, and a Mini 1D-CNN-based deep learning model with a decision-level soft-weighted fusion strategy. In the first stage, the model successfully distinguished normal and arrhythmic beats with 95.37% accuracy and an AUC of 0.9866. In the second stage, arrhythmic subtypes (L, R, V, A) were classified with 91.24% accuracy, Macro F1 = 0.90, and Macro AUC = 0.971. The CatBoost component provides statistical generalization power and explainability, while the Mini 1D-CNN effectively learned structural patterns in heart waveforms (P, QRS, T), increasing morphological sensitivity. The results show that the proposed method achieves similar accuracy levels with six times fewer parameters compared to complex deep hybrid models and is applicable in real-time systems with low computational cost. This study is considered an innovative step toward the development of explainable AI-based cardiac diagnostic systems.

Keywords

• ECG
• Arrhythmia Classification
• CatBoost
• Hybrid Learning
• Explainable Artificial Intelligence (XAI)
• Two-Stage Model

Açıklanabilir İki Aşamalı Hibrit Öğrenme Yaklaşımıyla ECG Tabanlı Aritmi Sınıflandırması

Abstract

ECG, Arrhythmia Classification, CatBoost, Hybrid Learning, Explainable Artificial Intelligence (XAI), Two-Stage Model

Keywords

• ECG
• Aritmi Sınıflandırması
• CatBoost
• Hibrit Öğrenme
• Açıklanabilir Yapay Zekâ (XAI)
• İki Aşamalı Model

References

1. Singh Kathayat N, Renold AP. Hybrid Deep Learning Model for Scalogram-Based ECG Classification of Cardiovascular Diseases. IEEE Access. 2025;(13):159628–159638. doi: 10.1109/ACCESS.2025.3605279.
2. Ye Y, Chipusu K, Ashraf MA, Ding B, Huang Y, Huang J. Hybrid CNN-BLSTM architecture for classification and detection of arrhythmia in ECG signals. Sci Rep. 2025;15(1):34510. doi: 10.1038/s41598-025-17671-1.
3. Srinivas P, Katarya R. hyOPTXg: OPTUNA hyper-parameter optimization framework for predicting cardiovascular disease using XGBoost. Biomed Signal Process Control. 2022;(73):103456. doi: 10.1016/j.bspc.2021.103456.
4. Moreno-Sánchez PA, García-Isla G, Corino VDA, Vehkaoja A, Brukamp K, van Gils M, Mainardi L. ECG-based data-driven solutions for diagnosis and prognosis of cardiovascular diseases: A systematic review. Comput Biol Med. 2024;(172):108235. doi: 10.1016/j.compbiomed.2024.108235. Epub 2024 Feb 28. PMID: 38460311.
5. Ye Y, Chipusu K, Ashraf MA, Ding B, Huang Y, Huang J. Hybrid CNN-BLSTM architecture for classification and detection of arrhythmia in ECG signals. Sci Rep. 2025;15(1):34510. doi: 10.1038/s41598-025-17671-1.
6. Mavaddati S. ECG arrhythmias classification based on deep learning methods and transfer learning technique. Biomed Signal Process Control. 2025;(101):107236. doi: 10.1016/j.bspc.2024.107236.
7. Singh Kathayat N, Renold AP. Hybrid Deep Learning Model for Scalogram-Based ECG Classification of Cardiovascular Diseases. IEEE Access. 2025;(13):159628–159638. doi: 10.1109/ACCESS.2025.3605279.
8. Alamatsaz N, Tabatabaei L, Yazdchi M, Payan H, Alamatsaz N, Nasimi F. A lightweight hybrid CNN-LSTM explainable model for ECG-based arrhythmia detection. Biomed Signal Process Control. 2024;(90):105884. doi: 10.1016/j.bspc.2023.105884.

9. Ye Y, Chipusu K, Ashraf MA, Ding B, Huang Y, Huang J. Hybrid CNN-BLSTM architecture for classification and detection of arrhythmia in ECG signals. Sci Rep. 2025;15(1):34510. doi: 10.1038/s41598-025-17671-1.
10. Singh Kathayat N, Renold AP. Hybrid Deep Learning Model for Scalogram-Based ECG Classification of Cardiovascular Diseases. IEEE Access. 2025;(13):159628–159638. doi: 10.1109/ACCESS.2025.3605279.
11. Gurler Ari B. Efficient ECG Beat Classification Using SMOTE-Enhanced SimCLR Representations and a Lightweight MLP. Symmetry (Basel). 2025;17(10):1677. doi: 10.3390/sym17101677.
12. Mohonta SC, Motin MA, Kumar DK. Electrocardiogram based arrhythmia classification using wavelet transform with deep learning model. Sens Biosensing Res. 2022;37:100502. doi: 10.1016/j.sbsr.2022.100502.
13. Mavaddati S. ECG arrhythmias classification based on deep learning methods and transfer learning technique. Biomed Signal Process Control. 2025;101:107236. doi: 10.1016/j.bspc.2024.107236.
14. Lopes UK, Valiati JF. Pre-trained convolutional neural networks as feature extractors for tuberculosis detection. Comput Biol Med. 2017;89:135–143. doi: 10.1016/j.compbiomed.2017.08.001.
15. Thomas JA, Perez-Alday EA, Hamilton C, Kabir MM, Park EA, Tereshchenko LG. The utility of routine clinical 12-lead ECG in assessing eligibility for subcutaneous implantable cardioverter defibrillator. Comput Biol Med. 2018;102:242–250. doi: 10.1016/j.compbiomed.2018.05.002.
16. Alamatsaz N, Tabatabaei L, Yazdchi M, Payan H, Alamatsaz N, Nasimi F. A lightweight hybrid CNN-LSTM explainable model for ECG-based arrhythmia detection. Biomed Signal Process Control. 2024;90:105884. doi: 10.1016/j.bspc.2023.105884.
17. Srinivas P, Katarya R. hyOPTXg: OPTUNA hyper-parameter optimization framework for predicting cardiovascular disease using XGBoost. Biomed Signal Process Control. 2022;73:103456. doi: 10.1016/j.bspc.2021.103456.
18. Moody GB, Mark RG. The impact of the MIT-BIH Arrhythmia Database. IEEE Engineering in Medicine and Biology Magazine. 2001;20(3):45–50. doi: 10.1109/51.932724.
19. Acharya UR, et al. A deep convolutional neural network model to classify heartbeats. Comput Biol Med. 2017;89:389–396. doi: 10.1016/j.compbiomed.2017.08.022.
20. Addison PS. Wavelet transforms and the ECG: a review. Physiol Meas. 2005;26(5):R155–R199. doi: 10.1088/0967-3334/26/5/R01.
21. Welch P. The use of fast Fourier transform for the estimation of power spectra: A method based on time averaging over short, modified periodograms. IEEE Transactions on Audio and Electroacoustics. 1967;15(2):70–73. doi: 10.1109/TAU.1967.1161901.
22. Shaffer F, Ginsberg JP. An Overview of Heart Rate Variability Metrics and Norms. Front Public Health. 2017;5. doi: 10.3389/fpubh.2017.00258.
23. Prokhorenkova L, Gusev G, Vorobev A, Dorogush AV, Gulin A. CatBoost: unbiased boosting with categorical features. [Online]. Github; [cited 2025 Dec 3]. Available from: https://github.com/catboost/catboost
24. Kingma DP, Jimmy BA. Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980, 2014.
25. He K, Zhang X, Ren S, Sun J. Deep Residual Learning for Image Recognition. in 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), IEEE 2016. Las Vegas: NV, USA; 2016 pp. 770–778. doi: 10.1109/CVPR.2016.90.
26. Yıldırım Ö, Pławiak P, Tan R-S, Acharya UR. Arrhythmia detection using deep convolutional neural network with long duration ECG signals. Comput Biol Med. 2018;102:411–420. doi: 10.1016/j.compbiomed.2018.09.