冠周脂肪影像组学鉴别冠状动脉狭窄程度初探

Preliminary study of pericoronary adipose tissue based on radiomic model in coronary artery stenosis

目的初步探讨基于冠脉CTA(CCTA)图像上冠状动脉周围脂肪(PCAT)的影像组学模型对冠状动脉狭窄程度的鉴别诊断。方法回顾性分析319例接受CCTA检查者的临床、影像资料,分为三组,包括正常组(101例202支经CCTA评估为正常者),轻中度组[74例74支经数字减影血管造影(DSA)评估为轻中度狭窄者],重度组(144例144支经DSA评估为重度狭窄者)。于左前降支(LAD)及右冠状动脉(RCA)最狭窄处的斑块周围勾画、分割PCAT的兴趣区(ROI),提取ROI内的影像组学特征,采用逻辑回归以及最小绝对收缩与选择算子(LASSO)进行特征筛选并建立影像组学模型,使用受试者工作曲线(ROC)评估正常-病变组及轻中度狭窄组-重度狭窄组之间的诊断效能。结果构建的两个影像组学模型(正常-病变模型、轻中度-重度模型)分别得出16个和9个最优特征参数,其中LogarithmGLCM;ogarithm;lusterShade为两者共有的特征参数。正常与病变组影像组学模型在训练组、测试组的ROC曲线下面积(AUC)分别为0.964和0.935,轻中度与重度组影像组学模型在训练组、测试组的AUC分别为0.69,0.644。结论基于CCTA图像上PCAT的影像组学模型在正常与冠状动脉狭窄之间具有较好的鉴别诊断效能,但区分冠脉狭窄程度的鉴别能力较低。

Objective To investigate the radiomic model of pericoronary adipose tissue(PCAT)based on coronary computed tomography angiography(CCTA)for differentiating the severity of coronary artery stenosis.Methods The clinical data and CCTA images of 319 patients in our institution from January 2018 to June 2020 were analyzed retrospectively.The patients were divided into control group of normal coronary arteries(101 patients with 202 vessels)on CCTA,mild to moderate stenosis group(74 patients with 74 coronary arteries),and severe stenosis group(144 patients with 144 coronary arteries)on digital subtraction angiography(DSA).Regions of interest(ROIs)of PCAT around the most stenosed plaque in the left anterior descending coronary artery(LAD)and right coronary artery(RCA)were segmented.The radiomic features in ROIs were extracted,and then selected by logistic regression and least absolute shrinkage and selection operator(LASSO)analysis,and the radiomic models were established.The diagnostic efficiency between different stenosis rates was evaluated by receiver operating characteristic(ROC)curve.Results 16 and 9 optimal characteristic parameters were obtained in the constructed radiomics normal-disease model and mild to moderate-severe stenosis model,respectively.Logarithmglcm_logarithm_Clustershade is a characteristic parameter in both groups.The areas under the ROC curves(AUCs)of the normal-disease model were 0.964 in training and 0.935 in the test cohort whereas the AUCs of the mild to moderate-severe stenosis model were 0.690 in training and 0.644 in the test cohort.Conclusion The radiomic model of PCAT based on CCTA image has good diagnostic efficiency between normal and coronary artery stenosis,but relatively low efficiency in differentiating the degree of coronary artery stenosis.