Coronary CT angiography (CCTA) is a first-line diagnostic tool for coronary artery disease, can not only quantitate the severity of coronary artery stenosis, but also further evaluate the plaque composition, morphology and vulnerability. The quantitative detection of perivascular fat based on CCTA imaging can be served as vessel inflammation state for comprehensively evaluation of long-term risk of cardiovascular events. With development of artificial intelligence and radiomics technology, machine learning is more and more applied to individual risk stratification and treatment decision-making of coronary artery diseases. Machine learning can deeply mine histological imaging markers, integrate clinical, biological and CCTA imaging information, generate accurate non-invasive prediction model, and provide support for further evaluation of cardiovascular risk. This article reviews the role of CCTA in the detection and quantification of plaque vulnerability and inflammation, and briefly summarizes the research advance of machine learning algorithm combined with radiomics in the evaluation of atherosclerotic plaque risk.