Intelligent Diagnosis of Coronary Artery Disease Using Radiomic Features from Magnetocardiography

Coronary artery disease (CAD), a prevalent cardiovascular condition caused by arterial narrowing or blockage that restricts blood flow to the heart muscle, continues to be a major global health concern with vital mortality rates. This research develops an intelligent diagnostic system integrating SERF (Spin Exchange Relaxation Free) magnetocardiography with radiomics analysis for early, noninvasive CAD detection. The study analyzed cardiac magnetic signals from 663 clinically confirmed CAD cases, stratified into severe and non-severe stenosis groups. Using 0.9 times the maximum positive and negative magnetic field intensities of the ST segment and T wave as the region of interest, through advanced radiomics processing of 2D temporal isomagnetic maps, we extracted comprehensive feature sets including morphological patterns, first-order statistics, and textural characteristics. Seven machine learning models algorithms - Logistic Regression, Support Vector Machine, k-Nearest Neighbors, Naive Bayes, Decision Tree, Random Forest, and XGBoost - were systematically evaluated, with experimental results showing substantial performance gains over conventional methods - achieving a 9.2% improvement in accuracy, a 9% increase in F1-score, and a 0.118 rise in AUC. These findings demonstrate the clinical potential of combining SERF technology with radiomics for enhanced CAD diagnosis, offering a radiation-free alternative to traditional imaging modalities while maintaining diagnostic reliability. The proposed methodology represents a significant advancement in cardiac diagnostics, particularly for early-stage disease detection where timely intervention is most critical.