A Comparative Learning and Neural Network Anecdote Architecture to Predicted Arrhythmias using ECG Wavelength

Abstract

ECG signals play a vital role in evaluating heart-connected diseases, encompassing arrhythmias. Self-regulating spotting of arrhythmias can considerably decrease the trouble for physicians and enhance sufferer results. This paper presents a deep learning-based neural network architecture for categorising ECG waves and finding arrhythmias. The proposed system dominance convolutional layers to reduce appropriate features and recurrent layers to entrap temporal dependencies in ECG waveforms. A comparative learning between prescriptive classifiers and neural networks is also fulfilled. An outcome substantiates the preeminent performance of the neural network, attaining an accuracy and significantly reducing perfidious positives linked to traditional methods.

Introduction

I. Background and Motivation

• Arrhythmias are abnormal heart rhythms that can lead to serious issues like heart attacks and strokes.

• ECG signals are vital for detecting arrhythmias, but manual interpretation is time-consuming and error-prone, especially with large datasets.

• Advances in machine learning (ML) and deep learning (DL) show great promise in improving detection accuracy.

• The study focuses on using hybrid deep learning (CNN + LSTM) to outperform traditional models in arrhythmia detection.

II. Contributions

• Proposed a hybrid CNN-LSTM deep learning model:

  • CNNs extract spatial features.

  • LSTMs capture temporal dependencies.

• Compared its performance with traditional ML models (SVM, KNN, Decision Trees, Random Forest).

III. Related Work

• Traditional ML methods:

  • Require manual feature extraction.

  • Include models like SVM, KNN, and Decision Trees.

• Deep learning:

  • Learns features automatically from raw ECG data.

  • CNNs are good for spatial analysis; LSTMs excel at time-series data like ECGs.

IV. Methodology

A. Data Collection

• ECG data from 2084 myocardial infarction patients.

• Split: 80% training, 20% testing.

• Preprocessing: Noise filtering, normalization, segmentation, and labeling.

B. Model Architecture

• 1D CNN layers: Extract spatial features from ECG signals.

• LSTM layers: Model the time-series nature of ECG data.

• Fully connected + softmax layer: Classify the signal into arrhythmia types.

C. Training

• Optimizer: Adam, learning rate = 0.001.

• Loss function: Categorical cross-entropy.

• Hyperparameters: Tuned using grid search and cross-validation.

D. Performance Metrics

• Accuracy: Correct predictions.

• Recall (Sensitivity): True positives for arrhythmia.

• Specificity: True negatives (normal rhythms).

• F1-score: Balance between precision and recall.

V. Results and Discussion

A. Performance Evaluation

• Best accuracy during training: 26.71%.

• Validation accuracy: 30%.

• Final training accuracy: ~26%.

• Confusion matrix and classification report showed:

  • High recall for one class (class 2).

  • Poor precision and recall for others.

• Model performed well on atrial fibrillation, but struggled with other classes.

B. Comparative Study

• Hybrid CNN-LSTM outperformed traditional classifiers.

• Advantage: Ability to learn both spatial and temporal features without manual extraction.

C. Error Analysis

• Misclassifications occurred, mainly due to noisy ECG data.

• Suggested improvements:

  • Better noise filtering.

  • Deeper architectures for more complex patterns.

Conclusion

This study presented a deep learning-based technique for identifying the disease arrhythmias with the help of ECG waveforms. By connecting traditional and recurrent layers, the model efficiently noted both spatial and temporal features of ECG waves, preeminent to consequence improvement in identify an accuracy.Our research explains that the proposed framework outperformed convolutional machine learning methods. Future study will explore the integration of more cosmopolitan noise-handling techniques and fine tuning of hyperparameters of further improve performance.

References

[1] P. A. Moreno-Sánchez et al., “ECG-based data-driven solutions for diagnosis and prognosis of cardiovascular diseases: A systematic review,” Comput Biol Med, vol. 172, p. 108235, Apr. 2024, doi: 10.1016/J.COMPBIOMED.2024.108235. [2] A. Gour, M. Gupta, R. Wadhvani, and S. Shukla, “ECG Based Heart Disease Classification: Advancement and Review of Techniques,” Procedia Comput Sci, vol. 235, pp. 1634–1648, Jan. 2024, doi: 10.1016/J.PROCS.2024.04.155. [3] “View of Hybrid FNN-DNN Approach for Early Detection of Cardiac Arrhythmia: A Novel Framework for Enhanced Diagnosis.” Accessed: Aug. 11, 2024. [Online]. Available: https://www.vfast.org/journals/index.php/VTCS/article/view/1781/1489 [4] L.-M. Tseng and V. S. Tseng, “Predicting Ventricular Fibrillation through Deep Learning”, doi: 10.1109/ACCESS.2020.3042782. [5] R. J. Martis, U. R. Acharya, and L. C. Min, “ECG beat classification using PCA, LDA, ICA and Discrete Wavelet Transform,” Biomed Signal Process Control, vol. 8, no. 5, pp. 437–448, 2013, doi: 10.1016/J.BSPC.2013.01.005. [6] E. Moghadas, J. Rezazadeh, and R. Farahbakhsh, “An IoT patient monitoring based on fog computing and data mining: Cardiac arrhythmia usecase,” Internet of Things, vol. 11, p. 100251, Sep. 2020, doi: 10.1016/J.IOT.2020.100251. [7] L. Zhang, H. Peng, and C. Yu, “An approach for ECG classification based on wavelet feature extraction and decision tree,” 2010 International Conference on Wireless Communications and Signal Processing, WCSP 2010, 2010, doi: 10.1109/WCSP.2010.5633782. [8] S. Wang et al., “Application of machine learning to predict the occurrence of arrhythmia after acute myocardial infarction,” BMC Med Inform Decis Mak, vol. 21, no. 1, pp. 1–14, Dec. 2021, doi: 10.1186/S12911-021-01667-8/TABLES/6. [9] V. M. Singh, V. Saran, P. Kadambi, E. Khan, and M. Gönen, “Autonomous Myocardial Infarction Detection from Electrocardiogram with a Multi Label Classification Approach,” Apr. 13, 2023, PMLR. Accessed: Jan. 25, 2024. [Online]. Available: https://proceedings.mlr.press/v189/singh23a.html [10] A. A.?; Ahmed et al., “Classifying Cardiac Arrhythmia from ECG Signal Using 1D CNN Deep Learning Model,” Mathematics 2023, Vol. 11, Page 562, vol. 11, no. 3, p. 562, Jan. 2023, doi: 10.3390/MATH11030562. [11] M. A. Khan and Y. Kim, “Cardiac Arrhythmia Disease Classiication Using LSTM Deep Learning Approach”, doi: 10.32604/cmc.2021.014682. [12] Z. Ozpolat and M. Karabatak, “Performance Evaluation of Quantum-Based Machine Learning Algorithms for Cardiac Arrhythmia Classification,” Diagnostics 2023, Vol. 13, Page 1099, vol. 13, no. 6, p. 1099, Mar. 2023, doi: 10.3390/DIAGNOSTICS13061099. [13] A. H. Khan, M. Hussain, and M. K. Malik, “Cardiac disorder classification by electrocardiogram sensing using deep neural network,” Complexity, vol. 2021, 2021, doi: 10.1155/2021/5512243.

Copyright

Copyright © 2025 R. Sudha, Dr. A. Nithya. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.