冠脉生理功能评估软件(DEEPVESSEL FFR)与有创FFR在评估冠脉缺血中的对比研究  

Comparative study on software DEEPVESSEL FFR and invasive FFR in assessing coronary ischemia

作　　者：徐浩[1] 张治[1] 解学乾[2] 杨文艺[1] 刘少稳[1] XU Hao;ZHANG Zhi;XIE Xueqian;YANG Wenyi;LIU Shaowen(Department of Cardiology,the First People’s Hospital,Shanghai Jiao Tong University,Shanghai 200080,China;Department of Radiology,the First People’s Hospital,Shanghai Jiao Tong University,Shanghai 200080,China)

机构地区：[1]上海交通大学附属第一人民医院心内科,上海200080 [2]上海交通大学附属第一人民医院放射科,上海200080

出　　处：《诊断学理论与实践》2021年第4期384-390,共7页Journal of Diagnostics Concepts & Practice

基　　金：松江区科学技术攻关项目(农业、医药卫生类,19SJKJGG101)。

摘　　要：目的:基于冠状动脉(冠脉)计算机体层摄影血管造影(computed tomography angiography,CTA)检查结果,采用国产人工智能冠脉生理功能评估软件(DEEPVESSEL-FFR,DV-FFR)诊断冠脉功能性缺血,并评估其应用价值。方法:本研究为前瞻性、单中心、自身对照研究,共纳入18例患者(共21根血管,狭窄程度为30%~90%),同时采用有创冠脉血流储备分数(fractional flow reserve,FFR)检测和DV-FFR检查评估冠脉功能性缺血情况,以有创FFR作为金标准,评价DV-FFR的诊断效能。DV-FFR使用64排及64排以上CT设备采集冠脉CTA的DICOM格式数据,基于计算机深度学习技术进行血管分割和重建,提取血管中心线,进而计算冠脉FFR。DV-FFR采用三维几何自动量化方法计算FFR数值,评估3个主支血管[左前降支(left anterior descending,LAD)、左回旋支(left circumflex,LCX)、右冠状动脉(right coronary artery,RCA)]的狭窄缺血风险。结果:参考有创FFR结果,取FFR=0.8为切点值,DVFFR≤0.8即为有意义的功能性心肌缺血。在血管层面,DV-FFR软件诊断缺血的准确率、特异度、灵敏度、阳性预测值、阴性预测值分别为90.5%、88.9%、91.7%、91.7%和88.9%;在患者层面,DV-FFR软件的诊断准确率、特异度、灵敏度、阳性预测值、阴性预测值分别为88.9%、87.5%、90.0%、90.0%和87.5%。DV-FFR结果与有创FFR结果间一致性较好,诊断效能无差异(P=0.787)。结论:CTA检查结果显示冠脉狭窄程度为30%~90%时,采用DV-FFR诊断冠脉功能性缺血的结果与有创FFR检测结果间一致性较好,可作为评估冠脉功能性缺血的一种有效方法。Objective:To assess application value of artificial intelligence software DEEPVESSEL FFR(DV-FFR)in evaluating CTA(computed tomography angiography)of coronary ischemia.Methods:This was a prospective,single-center,and self-control study,and 21 vessels in 18 patients with coronary artery narrow(degree between 30%-90%)were included.Coronary ischemia was evaluated by DV-FFR and invasive FFR,and the efficiency of DV-FFR for diagnosing coronary ischemia was assessed by comparison with the results of golden criterion(invasive FFR).Based on data obtained from CTA DICOM,the blood vessel images were segmented and reconstructed by DV-FFR with deep learning technology,and vessel centerlines was obtained.FFR was calculated using3-D geometric auto-quantitative technology to assess ischemia in LAD(left anterior descending),LCX(left circumflex),RCA(right coronary artery).Results:When cut-off value of invasive FFR was taken as 0.8,DV-FFR≤0.8 was considered as functional coronary ischemia.On vessel level,the accuracy,specificity,sensitivity,positive predictive value,and negative predictive value of DV-FFR for diagnosing functional coronary ischemia were 90.5%,88.9%,91.7%,91.7%and 88.9%,respectively.While on patient level,the above indexes were 88.9%,87.5%,90.0%,90.0%and 87.5%,respectively.The results of the 2 methods were similar(P=0.787).Conclusions:The results of DV-FFR showed a good consistency with that of invasive FFR in evaluating coronary artery narrow degree between 30%-90%,which could be used as effective diagnostic approach.

关 键 词：CT血管造影 冠状动脉 人工智能软件 血流储备分数 

分 类 号：R541.7[医药卫生—心血管疾病]