Exploring The Correlation and Relevance of Radiomics Features for Coronary Artery Calcification and Stenosis prediction

This study aims to assess the correlation and relevance of radiomic features extracted from coronary artery branches about the presence of calcification and stenosis.To do this, contrast-enhanced computed tomography (CECT) images from a subset of 54 cases were analyzed.Coronary arteries were segmented using the Chan-Vese segmentation method, followed by postprocessing in 3D Slicer.Radiomic features were then extracted using the PyRadiomics open-source library.Based on CECT reports, labels were assigned according to the presence of calcification and plaque.Feature selection was performed using correlation analysis and the Minimum Redundancy Maximum Relevance (MRMR) method.Finally, three different machine learning classification algorithms-Logistic Regression, Random Forest, and Support Vector Machines-were employed to evaluate the performance of the selected features.The results demonstrated that some features were informative for both calcification and stenosis.The features selected by correlation analysis combined with the Logistic Regression classifier showed more stable performance than the other methods.