Electrocardiogram image classification for six classes of heart diseases

Heart diseases are among the leading causes of death worldwide, as reported by the World Health Organization. Electrocardiograms (ECGs) are essential for diagnosing various heart conditions.While much of the existing research primarily focuses on accuracy as the key performance metric for ECG classification, this approach overlooks the critical importance of sensitivity, specificity, and F1-score, particularly when dealing with imbalanced datasets. In this study, we propose a hybrid model that integrates machine learning techniques (Support Vector Machines and Random Forest) with deep learning (VGG16) to classify six heart conditions: abnormal heart rhythm, atrial fibrillation, ischaemic heart disease, myocardial infarction, normal heart rhythm, and sinus bradycardia. We utilised a hybrid dataset consisting of 2848 ECG images sourced from both the Kaggle online database and a cardiac centre. The proposed model achieved state-of-the-art performance, attaining 95% accuracy, 99% specificity, 96% precision, 92% sensitivity, and 95% F1-score, thereby surpassing existing benchmarks such as MINA, ExpertRF, and CRNN from the PhysioNet Challenge dataset. These results underscore the model’s potential for facilitating early diagnosis, enabling cardiologists to deliver timely and personalised treatment.