基于增强CT影像组学评估肝硬化患者肝储备功能的应用

Values of enhanced CT-based radiomics in assessment of liver reserve function of patients with cirrhosis

目的:基于增强CT影像组学特征构建列线图预测模型对肝硬化患者肝储备功能进行Child-Pugh分级。方法:回顾性分析经临床证实的144例肝硬化患者,按照Child-Pugh评分标准分成Child-Pugh A级33例,B级60例,C级51例。构建Child-Pugh A vs Child-Pugh B/C及Child-Pugh A/B vs Child-Pugh C两个数据集,分别以8:2的比例随机分成训练集和测试集。在3期增强CT图像上手动勾画肝脏区域作为感兴趣区(ROI),于感兴趣区中提取并筛选特征。建立影像组学标签并构建列线图预测模型,将模型用于训练集及测试集,并绘制受试者工作特性曲线(ROC)评估其效能。结果:在Child-Pugh A vs Child-Pugh B/C数据集中,列线图在训练集与测试集中AUC分别为0.920和0.807,敏感度分别为0.933和0.741,特异度分别为0.846和0.826。在Child-Pugh A/B vs Child-Pugh C数据集中,列线图在训练集与测试集中AUC分别为0.880和0.821,敏感度分别为0.805和0.818,特异度分别为0.878和0.947。结论:基于不同肝脏储备功能肝硬化患者的腹部3期CT增强图像组学特征建立的列线图模型可作为预测Child-Pugh分级较为可靠的辅助诊断工具。

Objective:To establish nomogram model basd on enhanced CT radiomics to predict liver reserve function in patients with cirrhosis.Methods:144 clinically confirmed patients with cirrhosis were retrospectively analyzed.They were divided into three groups:Child-Pugh A(n=33),Child-Pugh B(n=60),and Child-Pugh C(n=51).Two data sets of Child-Pugh A vs Child-Pugh B/C and Child-Pugh A/B vs Child-Pugh C were constructed and randomly divided into training set and test set at a ratio of 8:2.Region of interest(ROI)was manually delineated around the liver margin on triphasic enhanced CT images,radiomics features were extracted and filtered in ROI.Radiomics signatures and scoring formula were established with features and the nomogram prediction model was constructed based on Radiomics score.The diagnostic efficiency of the model in the training set and test set was evaluated by receiver operating characteristic curve(ROC).Results:In the data set of Child-Pugh A vs Child-Pugh B/C,the AUC,sensitivity and specificity was 0.920 and 0.807,0.933 and 0.741,and 0.846 and 0.826,respectively.In the Child-Pugh A/B vs Child-Pugh C data set,the AUC,sensitivity and specificity was 0.880 and 0.821,0.805 and 0.818,and 0.878 and 0.947,respectively.Conclusion:The nomogram model based on enhanced CT radiomics can be used as a reliable diagnostic tool for predicting liver reserve function of cirrhotic patients.