基于影像组学的冠状动脉周围脂肪组织特征预测冠状动脉狭窄的血流动力学显著性
回顾性纳入92例临床疑似冠心病患者(121个病变),基于CCTA手动分割PCAT提取1116个影像组学特征,建立机器学习模型预测FFR≤0.8的血流动力学显著狭窄。CCTA联合决策树影像组学模型的AUC为0.812,显著高于单独CCTA的AUC 0.599(p=0.015),训练集和测试集NRI分别为0.820和0.775。
要点
- PCAT影像组学联合CCTA诊断血流动力学显著狭窄的AUC为0.812,高于单独CCTA的0.599(p=0.015)。
- 两个入选影像组学特征在FFR≤0.8的冠状动脉中显著高于FFR>0.8者。
- 联合模型训练集NRI为0.820、测试集NRI为0.775。
核心结论
- 在CCTA基础上加入PCAT影像组学特征可识别血流动力学显著的冠状动脉狭窄。
研究设计与样本
- 回顾性研究,纳入92例临床疑似冠心病患者共121个病变,均在1个月内接受CCTA、有创冠状动脉造影(ICA)和FFR检查,病变随机分配至训练集和测试集;基于PCAT手动分割计算1116个影像组学特征,经稳健性评估和特征选择后采用不同机器学习算法建立影像组学模型,以FFR为参考标准。
关键影像表现/征象
- 经排除后选出两个影像组学特征,二者在FFR≤0.8的冠状动脉中均显著高于FFR>0.8者。
诊断标准与关键数值
- 以FFR≤0.8定义为血流动力学显著狭窄。
- CCTA联合决策树影像组学模型的AUC为0.812,单独CCTA的AUC为0.599(p=0.015)。
- 联合模型在训练集的NRI为0.820,在测试集的NRI为0.775。
局限性
- 原文未报告局限性。
以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。
英文原文(摘要)· 点击展开
AIM: To investigate the diagnostic performance of the radiomics features of pericoronary adipose tissue (PCAT) in determining haemodynamically significant coronary artery stenosis as evaluated by fractional flow reserve (FFR).
MATERIALS AND METHODS: A total of 92 patients with clinically suspected coronary artery disease who underwent coronary computed tomography (CT) angiography (CCTA), invasive coronary angiography (ICA), and FFR examination within 1 month were included retrospectively, and 121 lesions were randomly assigned to the training and testing set. Based on manual segmentation of PCAT, 1,116 radiomics features were computed. After radiomics robustness assessment and feature selection, radiomics models were established using the different machine-learning algorithms. The area under the receiver operating characteristic (ROC) curve (AUC) and net reclassification index (NRI) were analysed to compare the discrimination and reclassification abilities of radiomics models.
RESULTS: Two radiomics features were selected after exclusions, and both were significantly higher in coronary arteries with FFR ≤0.8 than those with FFR >0.8. ROC analysis showed that the combination of CCTA and decision tree radiomics model achieved significantly higher diagnostic performance (AUC: 0.812) than CCTA alone (AUC: 0.599, p=0.015). Furthermore, the NRI of the combined model was 0.820 and 0.775 in the training and testing sets, respectively, suggesting the radiomics features of PCAT had were effective in classifying the haemodynamic significance of coronary stenosis.
CONCLUSIONS: Adding PCAT radiomics features to CCTA enabled identification of haemodynamically significant coronary artery stenosis.
引用
- Wen, Xu, An, et al. Predicting haemodynamic significance of coronary stenosis with radiomics-based pericoronary adipose tissue characteristics. Clin Radiol. 2021. PMID: 34852918. doi:10.1016/j.crad.2021.10.019
供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。