Explainable Offline Classification of Kalman Filter Preprocessed ECG Signals Using Fuzzy Membership-Based SVM with LIME

Abstract

The precise and rapid classification of electrocardiogram (ECG) signals is critically important for the early diagnosis of cardiovascular diseases. In this study, an offline, explainable, and lightweight artificial intelligence framework is proposed for ECG signal classification. The proposed hybrid approach integrates signal processing, fuzzy modeling, and machine learning to achieve both high performance and interpretability. First, Kalman filter-based preprocessing is applied to suppress noise and artifacts in raw ECG signals, improving signal quality and reliability. Then, Fuzzy C-Means (FCM) clustering is employed to derive fuzzy membership degrees, enabling the transformation of conventional features into a representation that captures uncertainty and transitional patterns inherent in ECG signals. These fuzzy membership features are subsequently used as inputs to a Support Vector Machine (SVM) classifier, resulting in highly accurate classification performance. To enhance transparency, the model’s decision-making process is interpreted using the Local Interpretable Model-Agnostic Explanations (LIME) technique, providing insight into feature contributions for individual predictions. In addition, the proposed system is designed within an offline artificial intelligence paradigm, eliminating the need for internet connectivity while maintaining low computational complexity. Experimental results demonstrate that the proposed Kalman–FCM–SVM framework achieves high classification accuracy and strong discriminative capability across multiple evaluation metrics. The integration of fuzzy membership representation and explainable AI significantly improves both model interpretability and reliability. The proposed approach shows strong potential for deployment in portable healthcare devices and real-time clinical decision support systems, particularly in resource-constrained environments.

Keywords

• ECG
• offline classifaction
• kalman filter
• fuzzy C means
• SVM
• LIME

Ethical Statement

This study does not involve any experiments on human or animal subjects. The ECG data used in this study were obtained from the publicly available MIT-BIH Arrhythmia Database (PhysioNet), which is fully anonymized and does not contain any personally identifiable information. Therefore, ethical approval and informed consent were not required.

Kalman Filtresi ile Ön İşlenmiş ECG Sinyallerinin LIME ile Açıklanabilir Bulanık Üyelik Tabanlı SVM Kullanılarak Çevrimdışı Sınıflandırılması

Abstract

Elektrokardiyogram (ECG) sinyallerinin doğru ve hızlı bir şekilde sınıflandırılması, kardiyovasküler hastalıkların erken teşhisi açısından kritik öneme sahiptir. Bu çalışmada, ECG sinyallerinin sınıflandırılması için çevrimdışı (offline), açıklanabilir ve hafif bir yapay zekâ çerçevesi önerilmektedir. Önerilen hibrit yaklaşım, hem yüksek performans hem de yorumlanabilirlik sağlamak amacıyla sinyal işleme, bulanık modelleme ve makine öğrenmesi yöntemlerini entegre etmektedir. İlk olarak, ham ECG sinyallerindeki gürültü ve artefaktları bastırmak için Kalman filtresi tabanlı ön işleme uygulanmış, böylece sinyal kalitesi ve güvenilirliği artırılmıştır. Daha sonra, Fuzzy C-Means (FCM) kümeleme algoritması kullanılarak bulanık üyelik dereceleri elde edilmiş ve bu sayede geleneksel özellikler, ECG sinyallerinin doğasında bulunan belirsizlik ve geçiş durumlarını yakalayabilen bir temsile dönüştürülmüştür. Elde edilen bulanık üyelik özellikleri, sınıflandırma aşamasında Destek Vektör Makineleri (SVM) modeline girdi olarak verilmiş ve yüksek doğrulukta sınıflandırma performansı elde edilmiştir. Modelin karar verme sürecinin şeffaflığını artırmak amacıyla, sınıflandırma sonuçları Local Interpretable Model-Agnostic Explanations (LIME) yöntemi kullanılarak açıklanmış ve her bir tahmin için özellik katkıları analiz edilmiştir. Ayrıca önerilen sistem, çevrimdışı yapay zekâ yaklaşımı çerçevesinde tasarlanmış olup internet bağlantısına ihtiyaç duymadan çalışabilmekte ve düşük hesaplama maliyeti sunmaktadır. Deneysel sonuçlar, önerilen Kalman–FCM–SVM tabanlı yaklaşımın farklı performans ölçütleri açısından yüksek sınıflandırma doğruluğu ve güçlü ayırt edicilik sağladığını göstermektedir. Bulanık üyelik temsili ile açıklanabilir yapay zekânın entegrasyonu, modelin hem yorumlanabilirliğini hem de güvenilirliğini önemli ölçüde artırmaktadır. Önerilen yaklaşım, özellikle taşınabilir sağlık cihazları ve gerçek zamanlı klinik karar destek sistemlerinde, kaynak kısıtlı ortamlarda kullanım için yüksek potansiyele sahiptir.

Keywords

• EKG
• çevrim dışı sınıflandırma
• kalman filtresi
• fuzzy c means
• DVM
• LIME

Ethical Statement

Bu çalışmada insan veya hayvan deneyi yapılmamıştır. Kullanılan ECG verileri, anonimleştirilmiş ve herkese açık olan MIT-BIH Arrhythmia veri setinden (PhysioNet) elde edilmiştir. Bu nedenle etik kurul onayı ve bilgilendirilmiş onam gerekmemektedir.

References

1. [1] E. H. Houssein, M. Kilany, and A. E. Hassanien, “ECG signals classification: a review,” International Journal of Intelligent Engineering Informatics, vol. 5, no. 4, p. 376, 2017. doi:10.1504/ijiei.2017.087944
2. [2] E. Kim, J. Kim, J. Park, H. Ko, and Y. Kyung, “TinyML-based classification in an ecg monitoring embedded system,” Computers, Materials & Continua, vol. 75, no. 1, p. 1751–1764, 2023. https://doi.org/10.32604/cmc.2023.031663.
3. [3] S. Yoo, S. Hong, D. Kam, and Y. Lee, “A lightweight ml-based ecg classification system using resource-constrained edge devices,” IEEE J Biomed Health Inform, vol. 29, no. 10, p. 7274-7284, 2025.doi: 10.1109/JBHI.2025.3589142.
4. [4] S. Guzman-Alfaro, A. Acuña-Correa, et al., “Real-time cardiac arrhythmia classification using tinyml,” Biomechanics, vol. 13, no. 5, p. 532, 2026.
5. [5] G. Silva, P. Silva, G. Moreira, V. Freitas, J. Gertrudes, and E. Luz, “A Systematic Review of ECG Arrhythmia Classification: Adherence to Standards, Fair Evaluation, and Embedded Feasibility,” arXiv, 2025, https://doi.org/10.48550/arXiv.2503.07276
6. [6] A. Das, S. Singh, J. Kim, Tariq Ahamed Ahanger, and Anil Audumbar Pise, “Enhanced EEG signal classification in brain computer interfaces using hybrid deep learning models,” Scientific Reports, vol. 15, no. 1, Jul. 2025, doi: https://doi.org/10.1038/s41598-025-07427-2.
7. [7] A. E. Mansour, El-Sayed Atlam, A. Ahmed, M. A. Atwa, E. M. Abdelrahim, and A. I. Siam, “Interpretable deep learning models for arrhythmia classification based on ecg signals using ptb-x dataset,” Diagnostics, vol. 15, no. 15, pp. 1950–1950, Aug. 2025, doi: https://doi.org/10.3390/diagnostics15151950.
8. [8] Lomoio U., P. Vizza, R. Giancotti, S. Petrolo, et al., “A convolutional autoencoder framework for ecg signal analysis,” Heliyon, vol. 11, no. 2, pp. e41517–e41517, Jan. 2025, doi: https://doi.org/10.1016/j.heliyon.2024.e41517.

9. ‌[9] H. Danandeh Hesar and A. Danandeh Hesar, “Adaptive dual augmented extended Kalman filtering of ECG signals,” Measurement, vol. 239, p. 115457, Jan. 2025, doi: https://doi.org/10.1016/j.measurement.2024.115457. [10] A. Pullin, J. Stuchbury-Wass, M. Ciliberto, KJ. Butkow, et al., “Ear-ecg denoising using heart sounds and the extended kalman filter,” 2025 IEEE 21st International Conference on Body Sensor Networks (BSN), pp. 1–4, Nov. 2025, doi: https://doi.org/10.1109/bsn66969.2025.11337841.
10. ‌[11] Z. Samandari and S. F. Molaeezadeh, “Noninvasive estimation of blood potassium concentration using ECG and FCM-ANFIS model,” Research on Biomedical Engineering, vol. 40, no. 3–4, pp. 647–658, Jul. 2024, doi: https://doi.org/10.1007/s42600-024-00371-6.
11. [12] A. Zeroual, A. Benzaoui, and L. Boubchir, “Hybrid learning paradigm for automatic electrocardiograph heartbeat anomaly detection,” 2025 IEEE International Conference on Big Data (BigData), pp. 8080–8085, Dec. 2025,doi:https://doi.org/10.1109/bigdata66926.2025.11402155.
12. [13] K. Adalarasu, B. Raghavan, B. Madhavan, S. Venkatesh, and R. Amirtharajan, “An explainable machine learning (XAI) framework to enhance types of cardiovascular disease diagnosis and prognosis,” Physical and Engineering Sciences in Medicine, Sep. 2025, doi: https://doi.org/10.1007/s13246-025-01653-8.
13. [14] K. Eunchan, K. Jaehyuk, P. Juyoung, K. Haneul, Y. Kyung, “TinyML-based classification in an ECG monitoring embedded system,” Computers, Materials, & Continua, vol. 75, no. 1, p. 1751, 2023. doi:10.32604/cmc.2023.031663
14. [15] H. Yelchuri, R. Rashmi, “A review of TinyML.” arXiv preprint arXiv:2211.04448, 2022, https://doi.org/10.48550/arXiv.2211.04448
15. [16] G. Moody and R. Mark, “MIT-BIH arrhythmia database v1.0.0,” physionet.org, Feb. 24, 2005. https://physionet.org/content/mitdb/1.0.0/
16. [17] C. Hajiyev, “Robust kalman filter with recursive measurement noise covariance estimation against measurement faults,” International Journal of Prognostics and Health Management, vol. 16, no. 1, Jan. 2025, doi: https://doi.org/10.36001/ijphm.2025.v16i1.4204.
17. [18] H. D. Hesar, “A novel ar-based kalman filtering framework for ecg enhancement,” 2025 32nd National and 10th International Iranian Conference on Biomedical Engineering (ICBME), pp. 1286–1292, Nov. 2025, doi: https://doi.org/10.1109/icbme68496.2025.11392248.
18. [19] C. E. Heil, and D. F. Walnut, “Continuous and discrete wavelet transforms,” SIAM review, vol. 31, no. 4, pp. 628-666, 1989.
19. [20] A. D Aghdam, N. J. Dabanloo, F. N. Rahatabad, and K. Maghooli, “A mathematical approach to intelligent control in artificial heart pacemaker design using anfis with fcm, ga, and pso,” Mathematical Sciences, vol. 19, no. 2, 2025, doi: https://doi.org/10.57647/mathsci.2025.1902.06.
20. [21] H. Adarsha, N. Vasantha Kumari, N. Kalidas, J. Sidhu, B. K. Pattanayak, and H. Kalra, “Classification and diagnosis of eeg signals for epilepsy detection using a proposed cnn-fcm approach,” 2025 IEEE International Conference on Emerging Technologies and Applications (MPSec ICETA), pp. 1–6, Feb. 2025, doi: https://doi.org/10.1109/mpseciceta64837.2025.11118765.
21. [22] V. Rai and N. Tiwari, “brain tumor classification using fcm, densenet, and svm with transfer learning techniques,” EPJ Web of Conferences, vol. 328, p. 01035, 2025, doi: https://doi.org/10.1051/epjconf/202532801035.
22. [23] A. Salih, I.B. Galazzo, A.M. Lee, K. Lekadir, et al., “Explainable artificial ıntelligence and cardiac ımaging: toward more ınterpretable models,” Circulation: Cardiovascular Imaging, Apr. 2023, doi: https://doi.org/10.1161/circimaging.122.014519.
23. ‌[24] S. Bayrak, “Unveiling intrusions: explainable svm approaches for addressing encrypted wi-fi traffic in uav networks,” Knowledge and Information Systems, vol. 66, no. 11, pp. 6675–6695, Jul. 2024, doi: https://doi.org/10.1007/s10115-024-02181-9.
24. [25] M. Rahman and B. I. Morshed, “Resource-constrained on-chip ai classifier for beat-by-beat real-time arrhythmia detection with an ecg wearable system,” Electronics, vol. 14, no. 13, p. 2654, Jun. 2025, doi: https://doi.org/10.3390/electronics14132654.