Hybrid Deep-Meta Learning Framework For Robust ECG Classification Using Resnet50 Features And Multi-Level Ensemble Integration
DOI:
https://doi.org/10.29284/ftshvk32Keywords:
Electrocardiogram (ECG) signals, Hybrid Deep–Meta Learning, PCA whitening, ResNet50, Adaptive Instance Normalization, Convolutional Block Attention Module, Support Vector Machine, Light Gradient Boosting Machine, Multi-level Ensemble Meta-classifier.Abstract
The accurate classification of electrocardiogram (ECG) signals is vital for early detection and intervention in cardiovascular diseases. This study proposes a Hybrid Deep–Meta Learning Framework that integrates ResNet50-based deep feature extraction with a multi-level ensemble meta-classifier for robust ECG image classification. The preprocessing pipeline employs advanced image enhancement techniques-including Contrast Limited Adaptive Histogram Equalization (CLAHE), gridline suppression, and multi-modal data augmentation (MixUp, CutMix, and time-warping)-to generate high-quality, diverse training samples. Deep visual representations are extracted using a modified ResNet50 backbone augmented with Adaptive Instance Normalization (AdaIN) and a lightweight Convolutional Block Attention Module (CBAM) to emphasize salient morphological and temporal ECG features. The extracted feature embeddings undergo dimensionality reduction via Principal Component Analysis (PCA) whitening and are fed into a hybrid ensemble classifier comprising Support Vector Machine (SVM) and Light Gradient Boosting Machine (LightGBM) as base learners, and a Meta-Learning MLP layer as the final decision integrator. The proposed system achieves superior generalization by leveraging complementary decision boundaries across heterogeneous learners. Comprehensive experiments demonstrate that the framework achieves high accuracy, precision, recall, and F1-scores across multiple ECG categories (Normal, Abnormal Heartbeat, and Myocardial Infarction). Furthermore, a custom-designed graphical user interface (GUI) enables real-time feature extraction, visualization, and classification, facilitating clinical usability. This hybrid deep–meta learning paradigm establishes a scalable and explainable approach for automated ECG interpretation, offering substantial potential for deployment in smart healthcare systems and telecardiology platforms.
References
[1] Mudassar Khalid, C. Pluempitiwiriyawej, S. Wangsiripitak, “The Applications of Deep Learning in ECG Classification for Disease Diagnosis: A Systematic Review and Meta-Data Analysis,” Engineering Journal, vol. 28, no. 8, pp. 45-63, 2024.
[2] S. Bhatia, S.K. Pandey, A. Kumar, A. Alshuhail, “Classification of Electrocardiogram Signals Based on Hybrid Deep Learning Models,” Sustainability, vol. 14, no. 24, 16572, 2022.
[3] Y. Elmir, Y. Himeur, A. Amira, “ECG Classification Using Deep CNN and Gramian Angular Field,” arXiv preprint, 2308.02395, 2023.
[4] P.N. Malleswari et al., “Deep learning-assisted arrhythmia classification using 2-D ECG spectrograms,” EURASIP Journal on Advances in Signal Processing, vol. 2024, Article 104, 2024.
[5] T. Sadad, M. Safran, I. Khan et al., “Efficient Classification of ECG Images Using a Lightweight CNN with Attention Module and IoT,” Sensors, vol. 23, no. 18, 7697, 2023.
[6] A. Rashidul Islam, K. Fida Hasan, S. Sultana et al., “HARDC: A novel ECG-based heartbeat classification using hierarchical attention dual structured RNN and dilated CNN,” arXiv preprint, 2303.06020, 2023.
[7] K. Karthikeyan, S. Ahanmed, V. Vignesh, C. Surya, “Classification of Arrhythmia by Using Deep Learning with 2-D ECG Spectral Image Representation,” J. Popul. Therapeutics & Clin. Pharmacol., 30(7), pp. 12-17, 2023.
[8] V.K. Gautam, J. Kapoor, M.M. Khan, “ECG Signal Classification Using Machine Learning and Deep Learning: A Unified Framework,” J. Comput. Anal. Appl., vol. 33, no. 8, pp. 5682-5699, 2024.
[9] C.K.N. Chaithra, N. Neelappa, M.T. Sreedevi, A. Asha, “Hybrid Deep Learning Models for Accurate ECG Classification in Cardiovascular Disease Diagnosis,” ETASR, vol. 15, no. 5, pp. 26683-26688, 2025.
[10] Z. Wang, Y. Xu, X. Li, et al., “Two-dimensional ECG-based cardiac arrhythmia classification using DSE-ResNet,” Scientific Reports, vol. 12, no. 1, pp. 18664, 2022. DOI: 10.1038/s41598-022-18664-0.
[11] H. L. Nguyen, S. V. Pham, H. C. Le, “Blending Ensemble Learning Model for 12-Lead Electrocardiogram-Based Arrhythmia Classification,” Computers, vol. 13, no. 12, pp. 316, 2024. DOI: 10.3390/computers13120316.
[12] A. Bayani, M. Kargar, “LDCNN: A new arrhythmia detection technique with ECG signals using a linear deep convolutional neural network,” Physiological Reports, vol. 12, no. 17, e16182, 2024. DOI: 10.14814/phy2.16182.
[13] X. Bai, X. Dong, Y. Li, R. Liu, H. Zhang, “A hybrid deep learning network for automatic diagnosis of cardiac arrhythmia based on 12-lead ECG,” Scientific Reports, vol. 14, Article 24441, 2024. DOI: 10.1038/s41598-024-24441-4.
[14] S. Sattar, M. Saima, A. Khan et al., “Cardiac Arrhythmia Classification Using Advanced Deep Learning Framework,” Sensors (Basel), vol. 24, no. 8, 2484, 2024. DOI: 10.3390/s24082484.
[15] B. Omarov, M. Baikuvekov, D. Sultan, N. Mukazhanov, M. Suleimenova, M. Zhekambayeva, “Ensemble Approach Combining Deep Residual Networks and BiGRU with Attention Mechanism for Classification of Heart Arrhythmias,” Computers, Materials & Continua, vol. 80, no. 1, pp. 341-359, 2024. DOI: 10.32604/cmc.2024.052437.
[16] Y. D. Daydulo, B. L. Thamineni, A. A. Dawud, “Cardiac arrhythmia detection using deep learning approach and time-frequency representation of ECG signals,” BMC Medical Informatics and Decision Making, vol. 23, Article 232, 2023. DOI: 10.1186/s12911-023-02326-W.
[17] F. Zabihi, A. Fazeli, H. Sheikhazad et al., “An electrocardiogram signal classification using a hybrid deep learning and machine learning model,” Computers in Biology and Medicine, vol. 176, 2024. DOI: 10.1016/j.compbiomed.2024.107509.
[18] S. Lamba, S. Kumar, M. Diwakar, “FADLEC: Feature extraction and arrhythmia classification using deep learning from electrocardiograph signals,” Discover Artificial Intelligence, vol. 5, Article 82, 2025. DOI: 10.1007/s44163-025-00290-0.
[19] D. Tenepalli, V. Gupta, R. P. Kumar, “Advancing cardiac diagnostics: high-accuracy arrhythmia detection using multi-lead ECG and ensemble learning,” Frontiers in Physiology, vol. 16, Article 1613812, 2025. DOI: 10.3389/fphys.2025.1613812.
[20] A. A. Ahmed, W. Ali, T. A. A. Abdullah, S. J. Malebary, “Classifying Cardiac Arrhythmia from ECG Signal Using 1D CNN Deep Learning Model,” Mathematics, vol. 11, no. 3, article 562, 2023. DOI: 10.3390/math11030562.
[21] Z. Huang, Z. Wen, H. Li, “Research on Arrhythmia Classification and Risk Degree Prediction based on Deep Neural Network and Convolutional Neural Network,” Frontiers in Computing and Intelligent Systems, vol. 5, no. 3, article 13848, 2023. DOI: 10.54097/fcis.v5i3.13848.
[22] Atiaf A. Rawi, M. K. Elbashir, A. M. Ahmed, “ECG Heartbeat Classification Using CONVXGB Model,” Electronics, vol. 11, no. 15, article 2280, 2022. DOI: 10.3390/electronics11152280.
[23] S. L. Abdul Rofar, Z. F. Mohd Apandi, N. S. Aziz, W. R. Wan Othman, “Deep Learning Architectures for ECG Classification in Cardiac Arrhythmia Detection: A Review,” Environment-Behaviour Proceedings Journal, vol. 10, Special Issue 32, pp. 67-74, 2023. DOI: 10.21834/e-bpj.v10iSI32.7056.
[24] https://www.kaggle.com/datasets/jayaprakashpondy/ecgimages
Downloads
Published
Issue
Section
License
This work is licensed under a Creative Commons Attribution 4.0 International License.
This work is licensed under a Creative Commons Attribution 4.0 International License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
