Objective To evaluate the value of coronary artery CT angiography(CCTA)based on artificial intelligence(AI)in the diagnosis of coronary artery stenosis.Methods The imaging data of 180 patients suspected of coronary artery disease(CAD)with CCTA and coronary angiography(ICA)were analyzed retrospectively.The CCTA-AI software of Aitrox Technology Corporation Limited was used to detect coronary artery disease automatically,and ICA was used as the gold standard to calculate the diagnostic efficacy of AI and physicians for more than moderate coronary artery stenosis and severe coronary artery stenosis by ROC curves.Kappa test was used to evaluate the consistency of AI software and the radiologists on CCTA-based coronary artery CAD-RADS scores and plaque nature.Results The AUCs for detection of more than mod-erate coronary stenosis and severe coronary stenosis by the AI software and the radiologists were 0.808,0.858,0.872 and 0.907,respectively.The difference of the AUCs between AI and radiologists were not significant(P>0.05).The agreement CAD-RADS scores were good between the AI software,ICA and radiologists(Kappa =0.643 and 0.808,respectively,both P<0.001).The results of agreement test of plaque nature evaluated by AI and physicians were good(Kappa =0.915,P<0.05).Conclusion The AI software has good performance in the diagnosis of coronary artery disease,so it is good for clinical diagnosis of CAD.
NETL
NSTL
万方数据
