<![CDATA[Heart disorders persist a primary cause of mortality worldwide, underscoring the necessity for precise and effective diagnostic support systems. The objective of this study is to classify heart disorders employing a combination of deep learning and machine learning approaches based upon electrocardiogram (ECG) image data., The model’s performance was evaluated through 5-fold cross-validation per patient to ensure robust generalizability. The dataset comprised 486 ECG images from 284 patients. A total of six models were subjected to comparative analysis, including Support Vector Machine (SVM), VGG16, ResNet50, Custom CNN, Xception, and Inception-V3, by utilizing key evaluation metrics including accuracy, precision, recall, specificity, F1-score, and AUC-ROC. The experimental results demonstrated that Inception-V3 achieved the optimal overall performance, demonstrating a balance between sensitivity and precision. Furthermore, deep learning models generally outperformed traditional methods such as support vector machines (SVM). The mean performance across all models yielded an accuracy of approximately 78.6% and an AUC-ROC of 0.83, demonstrating reliable discrimination in cardiac disorder classification. Deep learning-based architectures, particularly Inception-V3 and Xception, demonstrated considerable potential in the development of automated and accurate diagnostic systems for the early detection of cardiac disorders. Future research could explore hybrid approaches and larger and more diverse datasets to enhance clinical applicability. This study provides improved accuracy and reliability in cardiac disorder classification by leveraging and comparing machine learning and deep learning approaches. The proposed model has been demonstrated to effectively capture complex patterns in medical data, thereby supporting early diagnosis and improving clinical decision-making.]]>
