Optimized Classification of Coronary Artery Disease Using Feature Engineering and One-Dimensional Convolutional Neural Network

Coronary artery disease (CAD), which is one of the leading causes of the global morbidity and mortality rates, needs for early, accurate, and non-invasive diagnostic methods. Electro- cardiogram (ECG) signals have been used to obtain information about the electrical activity of the heart and are considered a potentiality in modality for detecting CAD. This paper proposes an enhanced diagnostic methodology that integrates advanced feature extraction and a One-Dimensional Convolutional Neural Network (1D-CNN) for automatic CAD classification. The PTB-XL ECG recordings dataset, which contains annotated multilead ECG signals with standardized diagnostic statements, has been used in this study. A binary classification task has been performed to group the ECGs of coronary artery disease related classifications to a CAD class, while all other ECGs classified as normal are assigned to a control class. The dataset featured class imbalance resembling real clinical distribution with 90% Normal and 10% CAD segments. The proposed model achieved an accuracy of 95.09% and an AUC score of 0.958. The results demonstrate that the proposed method can facilitate accurate ECG-based CAD detection, minimize the chance of manual misinterpretation, and enable real-time clinical screening and early diagnosis in a scalable manner.