背景:冠状动脉疾病报告和数据系统(CAD-RADS)2.0用于标准化冠状动脉计算机断层扫描血管造影(CCTA)结果的报告。人工智能软件可以量化斑块成分,脂肪衰减指数,和血流储备分数。
目的:分析CAD-RADS狭窄合并斑块负荷分类患者不同严重程度的斑块特征,建立随机森林分类模型。
方法:回顾性收集2021年4月至2022年2月期间接受治疗的100例患者的数据。在每位患者中观察到的最严重的斑块是目标病变。根据CAD-RADS将患者分为三组:CAD-RADS1-2P0-2,CAD-RADS3-4BP0-2和CAD-RADS3-4BP3-4。组间评估变量之间的差异和相关性。AUC,准确度,精度,召回,和F1评分用于评估诊断性能。
结果:共纳入100例患者和178条动脉。CT血流储备分数(CT-FFR)的差异(H=23.921,p<0.001),脂质成分的体积(H=12.996,p=0.002),纤维脂质成分的体积(H=8.692,p=0.013),脂质成分体积的比例(H=22.038,p<0.001),纤维脂质成分体积的比例(H=11.731,p=0.003),钙化成分体积的比例(H=11.049,p=0.004),与斑块类型(χ2=18.110,p=0.001)有统计学意义。
结论:CT-FFR,斑块的脂质和纤维脂质成分的体积和比例,钙化成分的比例,和斑块类型是有价值的CAD-RADS狭窄+斑块负荷分类,尤其是CT-FFR,volume,以及脂质和纤维脂质成分的比例。使用随机森林建立的模型优于临床模型(AUC:0.874vs.0.647)。
BACKGROUND: The coronary artery disease-reporting and data system (CAD-RADS) 2.0 is used to standardize the reporting of coronary computed tomography angiography (CCTA) results. Artificial intelligence software can quantify the
plaque composition, fat attenuation index, and fractional flow reserve.
OBJECTIVE: To analyze
plaque features of varying severity in patients with a combination of CAD-RADS stenosis and
plaque burden categorization and establish a random forest classification model.
METHODS: The data of 100 patients treated between April 2021 and February 2022 were retrospectively collected. The most severe
plaque observed in each patient was the target lesion. Patients were categorized into three groups according to CAD-RADS: CAD-RADS 1-2 + P0-2, CAD-RADS 3-4B + P0-2, and CAD-RADS 3-4B + P3-4. Differences and correlations between variables were assessed between groups. AUC, accuracy, precision, recall, and F1 score were used to evaluate the diagnostic performance.
RESULTS: A total of 100 patients and 178 arteries were included. The differences of computed tomography fractional flow reserve (CT-FFR) (H = 23.921, p < 0.001), the volume of lipid component (H = 12.996, p = 0.002), the volume of fibro-lipid component (H = 8.692, p = 0.013), the proportion of lipid component volume (H = 22.038, p < 0.001), the proportion of fibro-lipid component volume (H = 11.731, p = 0.003), the proportion of calcification component volume (H = 11.049, p = 0.004), and plaque type (χ2 = 18.110, p = 0.001) was statistically significant.
CONCLUSIONS: CT-FFR, volume and proportion of lipid and fibro-lipid components of plaques, the proportion of calcified components, and plaque type were valuable for CAD-RADS stenosis +
plaque burden classification, especially CT-FFR, volume, and proportion of lipid and fibro-lipid components. The model built using the random forest was better than the clinical model (AUC: 0.874 vs. 0.647).