钙化斑块对冠状动脉CTA诊断准确性的影响  

Impact of Segment Calcification on Diagnostic Performance of Coronary CTA

作　　者：许丽雪 李芳[1] 罗南 郭宁 化冰[3] 刘清波 贺毅 王振常[1] 杨正汉[1] XU Lixue;LI Fang;LUO Nan;GUO Ning;HUA Bing;LIU Qingbo;HE Yi;WANG Zhenchang;YANG Zhenghan(Department of Radiology,Beijing Friendship Hospital,Capital Medical University,Beijing 100050,China;不详)

机构地区：[1]首都医科大学附属北京友谊医院放射科,北京100050 [2]数坤(北京)网络科技有限公司,北京100102 [3]首都医科大学附属北京友谊医院心内科,北京100050

出　　处：《中国医学影像学杂志》2020年第10期741-745,共5页Chinese Journal of Medical Imaging

基　　金：国家重点研究与发展计划(2019YFE0107800);北京市科技计划(Z201100005620009)。

摘　　要：目的以冠状动脉造影(CAG)为参考标准,分析不同钙化负荷对视觉与深度学习(DL)算法诊断冠状动脉CT血管成像(CCTA)管腔狭窄准确性的影响。资料与方法对110例同时行CAG与256排CCTA的住院患者进行回顾性分析。采用视觉法与DL算法诊断CCTA主干管腔狭窄情况,以管腔狭窄≥50%作为诊断界值。冠状动脉各节段的钙化负荷通过视觉评估分为无、轻度、中度及重度钙化。采用受试者工作特征曲线比较各节段水平不同钙化负荷下两种方法的诊断效能。结果随着钙化负荷增加,从无钙化节段至重度钙化节段,两种方法对冠状动脉狭窄的诊断准确度显著减低,视觉法由94.5%降至78.3%(P<0.001),DL算法由93.3%降至58.7%(P<0.001)。在不同钙化负荷下,DL算法诊断CCTA主干管腔狭窄的曲线下面积与视觉法比较,差异均无统计学意义(P均>0.05)。结论不同钙化负荷下,DL算法对256排CCTA主干管腔狭窄的诊断效能与视觉评估相当。随着钙化负荷增加,两种方法的诊断准确性均降低。Purpose Taking coronary angiography(CAG)as the reference standard,to investigate the impact of segment calcification on diagnostic performance of visual evaluation and a deep learning based(DL)algorithm for coronary CT angiography(CCTA).Materials and Methods 110 inpatients who underwent CAG and 256 rows of CCTA at the same time were retrospectively analyzed.Taking luminal stenosis≥50%as a cut-off,per-segment diagnostic performance of coronary CTA was evaluated by visual assessment and the DL algorithm.Severity of segment calcification was visually rated as no,mild,moderate and severe calcification.The receiver operating characteristic curve was used to compare the diagnostic efficacy of the two methods under different levels of calcification in each segment.Results As the segment calcification increased,the diagnostic accuracy of the two methods for coronary artery stenosis decreased significantly from the noncalcified segment to the severely calcified segment,the visual method decreased from 94.5%to 78.3%(P<0.001),the DL algorithm decreased from 93.3%to 58.7%(P<0.001).Under different segment calcification,there was no significant difference in the area under the ROC curve for diagnosing CCTA segment stenosis between DL algorithm and the visual method(P>0.05).Conclusion Under different segment calcification,the diagnostic efficiency of DL algorithm for 256-row CCTA segment stenosis is equivalent to that of visual evaluation.As the segment calcification increases,the diagnostic accuracy of both methods decreases.

关 键 词：冠状动脉狭窄 冠状血管 血管造影术 体层摄影术 螺旋计算机 深度学习 钙化负荷 

分 类 号：R445.3[医药卫生—影像医学与核医学] R543.3[医药卫生—诊断学]