摘要
目的探讨人工智能(artificial intelligence,AI)在辅助冠状动脉CTA(coronary CT angiograph,CCTA)病变诊断中的价值。方法回顾性分析临床怀疑冠心病行CCTA和冠状动脉血管造影(CAG)检查的160例患者的临床资料及影像资料。将CCTA图像分析结果和AI辅助诊断软件结果进行对比分析,计算AI及放射科医师对冠状动脉狭窄程度的诊断效能及数据处理时间。同时基于CCTA图像,评价AI和放射科医师对斑块性质、斑块长度的诊断。结果①以CAG为金标准,AI诊断冠状动脉轻度狭窄、中度狭窄、重度狭窄的AUC值分别为0.8、0.85、0.89,敏感度分别为85.8%、90.2%、91.6%,特异度分别为79.1%、82.4%、86.6%,放射科医师诊断冠状动脉轻度狭窄、中度狭窄、重度狭窄的AUC值分别为0.83、0.88、0.90,敏感度分别为87.2%、91.8%、92.9%,特异度81.7%、80.7%、85.8%,AI与放射科医师对冠状动脉狭窄诊断的AUC值及敏感性、特异性之间的差异无统计学意义(P>0.05);②AI及放射科医师对斑块性质结果一致性良好(Kappa值为0.804,P<0.05),斑块长度之间分析无统计学差异(P>0.05);③AI软件自动处理图像、重建分析并生成报告,用时约为人工用时的3.0%。结论AI在辅助CCTA血管狭窄方面具有良好的诊断效能,对斑块性质及斑块长度具有良好的评估,且能显著缩短图像处理及报告生成时间,是一种较好的辅助放射科医师诊断CCTA病变的工具。
Objective To evaluate the value of artificial intelligence(AI)in the diagnosis of coronary artery CTA(CCTA).Methods The clinical data and imaging data of 160 patients suspected of coronary artery disease(CAD)with CCTA and coronary angiography(CAG)were analyzed retrospectively.The image analysis results of CCTA and AI software were compared and ana⁃lyzed to calculate the diagnostic effectiveness of AI and radiologists on the degree of coronary artery stenosis and the data process⁃ing time.To evaluate plaque properties and length by AI and radiologists based on CCTA images.Results①Taking CAG as gold standard,The AUC values of AI in the diagnosis of mild,moderate and severe coronary artery stenosis were 0.8,0.85 and 0.89,the sensitivity was 85.8%,90.2%and 91.6%,and the specificity was 79.1%,82.4%and 86.6%,respectively,The AUC values of radiologists in the diagnosis of mild,moderate and severe coronary artery stenosis were 0.83,0.88 and 0.90,respectively.The sensitivity was 87.2%,91.8%and 92.9%,and the specificity was 81.7%,80.7%and 85.8%,respectively,There was no no sta⁃tistical difference between AI and radiologist's in AUC value,sensitivity and specificity(P>0.05).②The results of AI and radi⁃ologists on plaque properties were consistent(Kappa value was 0.804,P<0.05),which had no statistical differences(P>0.05).③AI-assisted software automatically reconstructed and analyzed CCTA images and generated reports,which took about 3%of the manual time.Conclusion AI-assisted diagnosis has good evaluation of plaque properties and length,and can significantly short⁃en the time of image processing and report generation,which is a good tool to assist radiologists in the diagnosis of CCTA lesions.
作者
黄方方
张卉
窦允龙
王亚洲
程国飞
王道清
HUANG Fang-fang;ZHANG Hui;DOU Yun-long;WANG Ya-zhou;CHENG Guo-fei;WANG Dao-qing(Radiology Department,the First Affiliated Hospital of Henan University of Traditional Chinese Medicine,Zhengzhou 450000,China)
出处
《医药论坛杂志》
2024年第1期22-25,共4页
Journal of Medical Forum
基金
河南省中医药科学研究专项课题(20-21ZY1021)。
