关键词: 冠状动脉CT血管成像, 自适应统计迭代重建, 人工智能, 冠心病

Abstract: Objective To investigate the clinical value of adaptive statistical iterative reconstruction-V (ASIR-V) coronary CT angiography (CCTA) combined with artificial intelligence (AI) in evaluating coronary heart disease (CHD). Methods A prospective study was conducted among 124 patients suspected of CHD who underwent CCTA. The control group was established based on the conventional CCTA parameters including a Noise Index (NI) of 11 and anterior and posterior ASIR-V of 50%, and the experimental groups were established based on the high-weighted posterior ASIR-V of 90%, combined with anterior ASIR-V settings of 60%, 70%, and 80%, respectively. Image quality was assessed subjectively using a 5-point scale, and signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR), and effective dose (ED) were calculated. The above groups were compared in terms of subjective scores, noise (SD), SNR, CNR, and ED. The group with high image quality and low ED was selected for AI analysis, and with coronary angiography (CAG) results as the gold standard, vascular stenosis ≥ 50% was defined as positive cases. The receiver operating characteristic (ROC) curve was plotted to assess the diagnostic performance of AI. Results There were significant differences between the control group and the anterior ASIR-V 80% group in subjective score (4.61± 0.50 vs 4.26± 0.58, P < 0.05), SD (36.15± 6.82 vs 42.08± 7.98, P < 0.05), AR SNR [13.56± 3.51 vs 11.26 (8.67, 13.02), P < 0.05], AR CNR (11.50± 3.14 vs 9.52± 2.70, P < 0.05), RCA SNR [12.82± 3.29 vs 10.40 (8.18, 12.06), P < 0.05], RCA CNR (10.76± 2.93 vs 8.75± 2.59, P < 0.05), and LCX SNR (12.45± 3.26 vs 10.35± 2.56, P < 0.05), and compared with the anterior ASIR-V 80% group, the control group had a significantly lower SD and significantly higher levels of the other parameters. For the anterior ASIR-V 70% group, AI results yielded an AUC of 0.911 for diagnosing vascular stenosis ≥ 50%, with a sensitivity of 0.935, a specificity of 0.888, a positive predictive value of 0.811, a negative predictive value of 0.963, and an accuracy of 0.904. Conclusion CCTA parameters of posterior ASIR-V 90% and anterior ASIR-V 70% can ensure image quality while reducing the radiation dose, and their AI analysis results have a high diagnostic value for CHD.

Key words: Coronary CT angiography, Adaptive statistical iterative reconstruction-V, Artificial intelligence, Coronary heart disease