ML-Driven Coronary Artery Disease Detection via ECG Image Analysis

Abstract

Cardiovascular disease (CVD) encompasses severe conditions that affect the heart and can be life-threatening. To improveearly diagnosis and treatment, researchers are leveraging advanced machine learning (ML) techniques to analyze electronic health data accurately. This study explores multiple ML approaches for predicting heart diseases using critical patient health factors. The implemented classification models include Convolutional Neural Networks (CNN), Multilayer Perceptron (MLP), Support Vector Machine (SVM), Random Forest (RF), and Naïve Bayes (NB). Before model training, data preprocessing and feature selection were performed toenhance prediction accuracy. The models wereevaluated using accuracy, precision,recall, and F1-score metrics, with the SVM model achieving the highest accuracy of 91.67%.

Introduction

Cardiovascular disease (CVD)—including conditions like coronary artery disease and stroke—is the leading cause of death globally, with India contributing nearly 20% of global heart attack deaths. Major risk factors include poor diet, sedentary lifestyle, smoking, and alcohol consumption. Traditional diagnostic methods like ECGs are prone to manual errors and limited availability, especially in underserved areas.

AI & Machine Learning in CVD Detection

To address these challenges, the study proposes an AI-based diagnostic system using machine learning (ML) and deep learning (DL) models for early and accurate CVD detection via ECG analysis.

Key Features of the Proposed System:

• Uses classification algorithms: SVM, MLP, Random Forest, Naïve Bayes, CNN.

• SVM achieved the highest accuracy: 91.67%.

• Incorporates explainability tools like SHAP and Grad-CAM for transparency.

• Deployed via Streamlit, offering a user-friendly interface for real-time ECG classification.

Literature Survey – Notable Contributions

1. SVM achieved 97.14% accuracy in automated ECG diagnosis.

2. CNN detected coronary artery disease with 95.6% accuracy.

3. Random Forest predicted heart risk using health records with 93.8% precision.

4. Hybrid Models (NB + Decision Trees) improved sensitivity by 12%.

5. ANN analyzed echocardiographic images with 94.2% accuracy.

6. Gradient Boosted Trees used lifestyle and genetic data for personalized prediction (F1-score: 92.5%).

System Architecture

• Data Preprocessing: Cleans, augments ECG images.

• CNN Model: Detects subtle cardiac patterns from ECG images.

• Deployment: Real-time diagnostic interface via Streamlit.

• Feedback Loop: Enables ongoing performance improvements.

Algorithms Used

• CNN: For image-based ECG classification.

• Random Forest: Ensemble learning for robustness.

• Naïve Bayes: Fast, probabilistic classification.

• SVM: High-accuracy classification with clear boundary separation.

• K-NN: Pattern recognition based on proximity.

• XGBoost: High-performance gradient boosting for feature-rich predictions.

Model Performance

• Overall Accuracy: 94% on ECG data.

• Best Classifier: Myocardial Infarction (Precision & Recall: 100%).

• Challenging Class: Abnormal Heartbeat (Recall: 84%).

Conclusion

The proposed model demonstrates strong diagnostic capability (94% accuracy) in classifying ECG signals. Key achievements include: PerfectMIdetection (100%precision/recall). Highnormalrhythmidentification(F1: 0.95). Future work should address recall gaps in Abnormal Heartbeats via expanded datasets and advanced architectures. This research underscores AI’s potentialto augment cardiac diagnostics, enabling scalable, early disease detection.

References

[1] Zhou, X., Liu, B., Zhang, Q., & Wang, H. “Deep Learning-Based Automated Diagnosis ofCardiovascular Diseases Using ECG Signals.” IEEE Transactions on BiomedicalEngineering, vol. 68, no. 5,pp.1234-1245,2022. [2] Kumar, S., Patel, R., & Sharma, A. “A Hybrid Machine Learning Model for Predicting Heart Disease Using ECG and Clinical Data.” International Journal of Medical Informatics, vol. 156, pp. 104678, 2023. [3] Ghosh, R., Banerjee, T., & Dutta, P. “Explainable AI for Heart Disease Diagnosis: A Comparative Study of CNN, XGBoost, and SVM.” Journal of Artificial Intelligence in Medicine, vol. 85, pp. 102214, 2024. [4] L. Yang et al., “Study of Cardiovascular Disease Prediction Model Based on Random Forest in Eastern China,” Sci. Rep., vol. 10, no. 1, 2020. [5] NutwaraChumrit, Chalita Weangwan, NattapolAunsria, “ECG-based Arrhythmia Detection Using AverageEnergyandZero-crossingFeatureswith SupportVectorMachine”inproceedingsofFourth InternationalConference(2020 IEEE). [6] M. Haider Abu Yazid, Muhammad Haikal Satria, Shukor Talib, Novi Azman, “Artificial Neural Network Parameter Tuning Framework for Heart Disease Classification,” in proceedings of EECSI 2018, Malang-Indonesia, 16-18 Oct2018. [7] P. Ghosh et al., “Efficient Prediction of Cardiovascular Disease Using Machine Learning Algorithms,” IEEE Access, vol. 9, pp. 19304–19326, 2021. [8] Ramya G. Franklin, Dr. B. Muthukumar, “Survey of Heart Disease Prediction and Identification using Machine Learning Approaches” in Proceedings of the Third International Conference on Intelligent Sustainable Systems [ICISS 2020].

Copyright

Copyright © 2025 Prof. Dilip Kale, Sejal Ghadge, Ketan Chavan, Amaan Ansari, Arin Pawar. 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.