对比人工与人工智能图像后处理冠状动脉CT血管造影显示冠状动脉解剖及其病变  

Comparison on manual and artificial intelligence image postprocessing of coronary CT angiography for displaying anatomy and lesions of coronary arteries

作　　者：赵福琳 陈澜菁 刘俊七 兰永树 ZHAO Fulin;CHEN Lanjing;LIU Junqi;LAN Yongshu(Department of Radiology,Affiliated Hospital of Southwest Medical University,Luzhou 646000,China)

机构地区：[1]西南医科大学附属医院放射科,四川泸州646000

出　　处：《中国介入影像与治疗学》2022年第12期782-786,共5页Chinese Journal of Interventional Imaging and Therapy

摘　　要：目的对比人工与人工智能(AI)后处理用于冠状动脉CT血管造影(CCTA)显示冠状动脉解剖及病变的效能。方法回顾性分析158例疑诊冠心病患者的CCTA资料,分别以人工(人工组,含1名负责图像后处理的影像科主任医师和2名负责分析图像的影像科主治医师)及深睿、数坤AI软件(记为A_(1)组、A_(2)组)行后处理,于容积再现(VR)图像中测量冠状动脉分支数目,于曲面重组(CPR)图像中测量左前降支(LAD)、左回旋支(LCX)及右冠状动脉(RCA)管腔拉直长度;观察LAD(近、中、远段)、LCX(近、中远段)、RCA(近、中、远段)有无管腔狭窄及其程度、有无斑块及其性质,分析3组评估结果的差异及一致性。结果3组所示冠状动脉分支数总体差异有统计学意义(P<0.01),A_(1)组、A_(2)组所示数目均多于人工组(P=0.04、<0.01)。A_(1)组(P=0.04、0.03)、A_(2)组(P均<0.01)所测LAD和RCA长度均大于人工组,其余组间LAD和RCA长度差异均无统计学意义(P均>0.05)。3组所测LCX长度(P=0.18)及评估LAD、LCX及RCA狭窄程度及斑块性质结果差异均无统计学意义(P均>0.05)。A_(1)组、A_(2)组评估LAD、LCX、RCA狭窄程度(Kappa=0.58、0.51、0.57),以及A_(1)组(Kappa=0.55、0.57、0.62)、A_(2)组(Kappa=0.56、0.58、0.67)与人工组评估结果的一致性均为一般。结论CCTA图像经AI后处理后显示冠状动脉分支较人工后处理更为完整;人工与AI后处理评估冠状动脉狭窄程度及钙化结果的一致性均不高。Objective To compare the efficacy of manual and artificial intelligence(AI)postprocessing of coronary CT angiography(CCTA)for displaying anatomy and lesions of coronary arteries.Methods CCTA data of 158 patients with suspected coronary heart disease were retrospectively analyzed.CCTA images were postprocessed manually by a chief physician and then analyzed by 2 attending physicians(manual group),while Shenrui and Shukun AI software(denoted as A_(1) and A_(2) group)were used for the same processions.The number of segments coronary arteries were measured on volume reconstruction(VR)images,and the lumen straightening length of left anterior descending artery(LAD),left circumflex artery(LCX)and right coronary artery(RCA)were measured on curved planar reformation(CPR)images.LAD(proximal,middle and distal segments),LCX(proximal and mid-distal segments),RCA(proximal,middle and distal segments)were observed,the degree of luminal stenosis and the nature of plaque were evaluated,the results were compared among 3 groups,and the consistencies were analyzed.Results significant differences of numbers of coronary artery segments were found among 3 groups(P<0.01),which in A_(1) and A_(2) group were more than in manual group(P=0.04,<0.01).The lengths of LAD and RCA in A_(1)(P=0.04,0.03)and A_(2) group(both P<0.01)were all longer than in manual group,while no significant difference was found between other groups(both P>0.05).There was no significant difference of LCX lengths(P=0.18),of the degree of stenosis of LAD,LCX and RCA,nor of plaque nature among 3 groups(all P>0.05).The consistency of A_(1) group and A_(2) group for the stenosis degree of LAD,LCX and RCA(Kappa=0.58,0.51,0.57),as well as of A_(1) group(Kappa=0.55,0.57,0.62)and A_(2) group(Kappa=0.56,0.58,0.67)with manual group for the nature of LAD,LCX and RCA plaques were all general.Conclusion AI post-processing of CCTA images could show more complete coronary branches than manual post-processing.The consistencies of AI and manual post-processing for evaluating the degree

关 键 词：冠状动脉疾病 体层摄影术 X线计算机 图像后处理 人工智能 

分 类 号：R541.4[医药卫生—心血管疾病] R814.42[医药卫生—内科学]