基于CT影像组学特征评估肝硬化患者肝脏储备功能

目的:探讨应用CT影像组学特征评估肝硬化患者肝脏储备功能的价值。方法:回顾性收集经临床确诊的肝硬化患者121例。根据Child-Pugh分级标准分为A、B和C三个等级，其中A级51例（A组）；B和C级共70例（B组）。所有患者均采用GE Discovery CT 750 HD行平扫加增强扫描。选择增强扫描门静脉期图像，由两名放射科医生应用Shukun Radiomics V94软件在门静脉左支层面对整个肝脏进行勾画；使用组内相关系数对两名医生勾画的结果进行一致性检验。将所有患者按照7∶3的比例随机分为训练集与验证集后，对整个肝脏进行影像组学特征提取，经过降维后筛选出影像组学特征，使用逻辑回归方法建立肝硬化患者的肝脏储备功能模型。应用ROC曲线下面积（AUC）评价模型的性能。结果:两名医生对图像分割的一致性检验结果良好，ICC均大于0.75。最终用17个影像组学特征建立了评估肝硬化患者肝脏储备功能的模型，训练集的AUC为0.84，准确性为0.78，敏感性为0.79，特异性为0.77；验证集的AUC为0.77，准确性为0.71，敏感性为0.76，特异性为0.65。结论:应用CT影像组学特征，能够评估肝硬化患者的肝脏储备功能。 

Evaluation on the Hepatic Functional Reserve of Cirrhotic Patients based on CT Radiomics Characteristics

Objective: To explore the value of CT radiomics characteristics in evaluating hepatic functional reserve of patients with liver cirrhosis. Methods: A total of 121 patients with clinically confirmed liver cirrhosis were retrospectively collected, who were graded as A, B and C according to Child-Pugh standard. 51 cases were grade A (group A) and 70 cases were grade B and C (group B). All patients underwent non-contrast and contrast enhanced CT scan using GE Discovery CT 750 HD scanner. Enhanced CT images of portal venous phase were selected, and the whole liver was delineated at the level of the left portal vein by 2 radiologists using Shukun Radiomics V94 software. Intraclass correlation coefficient (ICC) was applied to test the inter-group consistency of the results obtained by the 2 radiologists. All patients were randomly divided into the training set and the validation set at a ratio of 7∶3, and the whole liver was performed extraction process for radiomics characteristics. After dimensionality reduction, radiomics characteristics were acquired and the model was established by Logistic Regression (LR). Area Under ROC Curve (AUC) was used to evaluate the performance of the model. Results: The consistency test results of image delineation by two radiologists turned out well and the ICC were greater than 0.75. Finally, 17 radiomics characteristics were used to establish the evaluation model for hepatic functional reserve of patients with liver cirrhosis. The AUC of the training set was 0.84 while the accuracy, sensitivity and specificity was 0.78, 0.79, and 0.77, respectively. The AUC of the validation set was 0.77 while the accuracy, sensitivity and specificity was 0.71, 0.76, and 0.65, respectively. Conclusion: CT radiomics characteristics could be used to evaluate the hepatic functional reserve of patients with liver cirrhosis.