摘要
目的:探索人工智能(AI)辅助检测冠状动脉CT血管造影(CCTA)在冠状动脉狭窄检测中的应用价值。方法:选取医院收治的行CCTA检查及冠状动脉造影检查的53例疑患有冠状动脉粥样硬化的患者,所有患者均行常规CCTA扫描,将图像上传至后处理ISP星云工作站,由两名高年资医师完成手动冠状动脉图像后处理及报告,另将CCTA图像传至AI冠状动脉CT血管成像工作站,由AI辅助辨识图像中的左前降支(LAD)、左回旋支(LCX)和右冠主支(RCA)主要血管及病变并生成报告,以评估AI与医师对冠状动脉狭窄的诊断价值。结果:在53例患者的CCTA图像中共解析159支血管节段,AI的平均后处理和诊断时间明显少于医师诊断时间。AI对冠状动脉LAD、LCX和RCA病变的判断与医师的诊断,差异均无统计学意义,AI检测冠状动脉狭窄的灵敏度、特异度、准确率、阳性预测值和阴性预测值分别为92.08%、87.93%、77.99%、93.00%和86.44%;医师的诊断分别为95.05%、91.38%、84.91%、95.05%和91.38%,AI对诊断冠状动脉狭窄程度与医师诊断一致性较好(Kappa=0.844)。结论:AI辅助的CCTA检测冠状动脉狭窄有较高且精准的判断,并显著减少了后处理时间,可以作为医师诊断冠状动脉狭窄的有效辅助工具。
Objective:To investigate the application value of coronary computed tomography angiography(CCTA)with the assistance of artificial intelligence(AI)in diagnosing coronary artery stenosis.Methods:53 patients with suspected coronary atherosclerosis who admitted to hospital and underwent CCTA and coronary arteriography contrast examination were enrolled in this research,and all patients underwent routine CCTA examination.The images of them were uploaded to post processing ISP nebula workstation,and two senior physicians completed the manual post processing and report of coronary artery images.At the same time,the CCTA images were uploaded to AI-assisted CCTA workstation to distinguish the main blood vessels and lesion of left anterior descending(LAD)branch,left circumflex artery(LCX)and right coronary artery(RCA),and produce reporter by the assistant AI.And then,the diagnostic values of AI and physician on the coronary artery stenosis were assessed.Results:A total of 159 vessel segments from the CCTA images of 53 patients were analyzed,and the average post processing time and diagnostic time of AI were significantly shorter than those of physicians.The differences of the judgements on the lesions of coronary artery LAD,LCX and RCA between AI and physician were no significant.The sensitivity,specificity,accuracy,positive predictive value and negative predictive value of AI were respectively 92.08%,87.93%,77.99%,93.00%and 86.44%.Those indicators of physician were 95.05%,91.38%,84.91%,95.05%and 91.38%.The diagnostic consistency of coronary artery stenosis between AI and physician was favorable(Kappa=0.844).Conclusion:AI-assisted CCTA detection can provide higher and more accurate judgment for coronary artery stenosis,and it significantly shortens the time of post processing.It can be used as an effectively assistant tool of physician in diagnosing the coronary artery stenosis.
作者
左晨
刘畅
付丽媛
ZUO Chen;LIU Chang;FU Li-yuan(Department of Radiology and Diagnosis,The 900th Hospital of People’s Liberation Army Joint Service Support Force,Fuzhou 350025,China;不详)
出处
《中国医学装备》
2022年第12期11-14,共4页
China Medical Equipment
基金
福建省科技计划项目(2021I0037)“基于智能化平台的CT质量控制检测评价系统的设计与实现”
联勤保障部队第900医院科研计划(2019Q02)“CT成像设备质量控制检测智能评价系统的研发”。
