Abstract: Objective: To investigate the impact of an artificial intelligence (AI)-assisted system on the accuracy of radiology residents in evaluating coronary artery stenosis severity and the diagnostic efficacy for plaques of different natures. Materials and Methods: A retrospective analysis was performed on 108 patients with coronary atherosclerosis who underwent coronary computed tomography angiography (CCTA) and digital subtraction angiography (ICA) from August to October 2024 (58 males, 50 females; mean age 64.00±11.45 years). Two residents independently interpreted baseline CCTA images to grade the stenosis severity at the most narrowed sites of four coronary arteries. Plaque characteristics (calcified/non-calcified/mixed) were determined by senior physicians. After a 4-week interval, the residents re-interpreted the images with the assistance of the AI system. The paired McNemar test was used to compare the differences in diagnostic accuracy of stenosis severity before and after AI assistance, and the chi-square test was applied to analyze differences among different plaque types and diagnostic methods. Results: The analysis of 282 stenotic vessels showed that the diagnostic accuracy of AI alone was significantly higher than that of the residents’ independent diagnosis. AI assistance significantly improved the residents’ diagnostic efficacy, but remained lower than that of senior physicians. No significant difference in diagnostic accuracy was observed among different plaque types with AI assistance. Conclusion: The AI-assisted system significantly improved the accuracy of radiology residents in evaluating CCTA stenosis, with improvements in the diagnosis of plaques of different natures, thus demonstrating clinical application value.