Discussion On The Clinical Application Of Automatic Post-processing Of Coronary Artery CTA Based On Artificial Intelligence

Part I:Research on the application of the post-processing technology of coronary artery CTA based on artificial intelligence to evaluate the degree of coronary artery stenosisObjectiveTo explore the clinical value of artificial intelligence(AI)post-processing technology applied to coronary artery CT angiography(CCTA)in evaluating the efficacy and feasibility of coronary artery stenosis.MethodsThe CCTA images of 145 patients with coronary atherosclerotic heart disease in our hospital from January 2016 to August 2017 were analyzed retrospectively.All patients underwent CCTA and coronary DSA(Digital Subtraction Angiography,DSA)within 3 months.All CCTA images were imported into manual post-processing and AI post-processing analysis workstations.According to the different post-processing methods,CCTA images were divided into two groups:manual post-processing analysis group(manual analysis group)and AI post-processing analysis group(AI group).The Left main coronary artery(LM),left anterior descending branch(LAD),left circumflex(LCX)and Right coronary artery(RCA)were analyzed as follows:consistency analysis of the accuracy of coronary plaque detection and differentiation between manual analysis group and AI group.Results1.The time of post-processing and analysis of CCTA images by AI was significantly shorter than that by manual analysis(P<0.05).2.In the diagnosis of coronary artery stenosis:using DSA as a reference standard,the consistency of the judgment of coronary artery stenosis in the AI group was lower than that in the manual analysis group in the four blood vessels(9 segments);There are different degrees of overdiagnosis in the degree of stenosis(the overdiagnosis rates in the AI group were 7.02%and 6.06%,respectively;the overdiagnosis rates in the manual analysis group were 8.19%and 6.06%).3.The consistency between the AI group and the manual analysis group in the judgment of the degree of stenosis is good(Kappa≥ 0.75)for proximal segment of RCA,and general(0.4 ≤ Kappa<0.75)for LM、middle segment of RCA,proximal middle segment of LAD and proximal segment of LCX.ConclusionAI has higher post-processing efficiency for CCTA images,and a good consistency compared by manual analysis in coronary artery stenosis diagnosis for proximal segment of RCA.The false-positive rate of coronary artery stenosis diagnosis by AI is higher than that of manual analysis.The accuracy of CCTA automatic post-processing technology based on artificial intelligence still needs to be further improved.Part Ⅱ:Research on the application of the post-processing technology of coronary artery CTA based on artificial intelligence to evaluate coronary plaqueObjectiveTo explore the clinical value of artificial intelligence(AI)post-processing technology applied to coronary artery CT angiography(CCTA)in detecting coronary plaque and judging plaque properties.MethodsThe CCTA images of 145 patients with coronary atherosclerotic heart disease in our hospital from January 2016 to August 2017 were analyzed retrospectively.All patients underwent CCTA and coronary DSA(Digital Subtraction Angiography)within 3 months.All CCTA images were imported into manual post-processing and AI post-processing analysis workstations.According to the different post-processing methods,CCTA images were divided into two groups:manual post-processing analysis group(manual analysis group)and AI post-processing analysis group(AI group).The Left main coronary artery,left anterior descending branch,left circumflex and Right coronary artery were analyzed as follows:Using the manual analysis as the reference standard,the consistency and differential rate analysis between the manual analysis group and the AI group on the detection of coronary plaque and the judgment of plaque properties was performed.Results1.The amount of atherosclerotic plaques in 145 patients were separately detected:249 coronary arteries(337 segments)in the DSA group;298 coronary arteries(358 segments)in the manual analysis group and 353 coronary arteries(485 segments)in the AI group.2.There was statistical significance(P<0.05)between the AI group and the manual analysis group in the differential diagnosis of calcified and non-calcified plaques.3.The misdiagnosis rate of calcified plaque in the AI group was higher(36.17%).ConclusionThe differential diagnosis by AI for different plaque properties needs to be further improved.