Cardiovascular diseases (CVDs) are the leading cause of mortality worldwide, accounting for 32% of global deaths. Electrocardiography (ECG) is a widely used, cost-effective, and non-invasive diagnostic tool for detecting cardiac abnormalities. However, ECG interpretation remains challenging due to noise interference, physiological variations, and the need for expert evaluation. This study proposes a machine learning-based approach for automatic classification of cardiac conditions using ECG images. The methodology involves feature extraction using Wavelet Transform (WT) and Gray-Level Co-occurrence Matrix (GLCM), followed by feature fusion to enhance classification. A total of 928 ECG images from four categories—Myocardial Infarction (MI), Abnormal Heartbeat (ABH), History of MI (HMI), and Normal—were analyzed. The extracted features were classified using XGBoost, Random Forest, Support Vector Machine, K-Nearest Neighbors, Decision Tree, and Logistic Regression. Results showed that XGBoost achieved the highest accuracy (93.55%), followed by Random Forest (93.01%), outperforming conventional methods. The findings suggest that feature fusion enhances classification and offers an interpretable, computationally efficient alternative to deep learning. This study contributes to automated cardiac diagnostics by providing a robust framework suitable for clinical applications and wearable ECG systems.
Karaca et al. (Thu,) studied this question.