CT影像组学在肾透明细胞癌Fuhrman分级中的应用价值

Application value of CT-based radiomics in Fuhrman grading of clear cell renal cell carcinoma

目的:探讨CT影像组学在肾透明细胞癌Fuhrman分级中的应用价值。方法:选取126例肾透明细胞癌(ccRCC)患者作为研究对象,依据Fuhrman分级标准将FuhrmanⅠ、Ⅱ级ccRCC患者纳入低级别组;FuhrmanⅢ、Ⅳ级ccRCC患者纳入高级别组。按患者就诊时间不同将患者分为训练集(n=84)和验证集(n=42)。基于患者一般资料和CT图像特征构建CT特征模型,基于提取的最优CT影像组学特征构建影像组学模型,并将CT图像特征和影像组学得分融合构建融合模型,比较验证集和训练集中各模型的诊断效能。结果:经过Logistic回归分析和LASSO特征选择,最终获得3个CT图像特征和10个最优影像组学特征,成功构建CT特征模型及CT影像组学模型。ROC结果显示,训练集CT特征模型、CT影像组学模型、融合模型曲线下面积(AUC)分别为0.688、0.823、0.810,验证集CT特征模型、CT影像组学模型、融合模型AUC分别为0.689、0.853、0.846。结论:CT影像组学对肾透明细胞癌Fuhrman分级具有较好的诊断效能,临床应用价值高。

Objective:To investigate the application value of CT-based radiomics in Fuhrman grading of clear cell renal cell carcinoma(ccRCC).Methods:126 patients with ccRCC were selected as the research subjects,according to Fuhrman grading criteria,patients with Fuhrman grade I and II ccRCC were included in the low-grade group,and those with Fuhrman grade III and IV ccRCC were included in the high-grade group.Patients were divided into a training set(n=84)and a validation set(n=42)based on their visit time.A CT feature model was constructed based on the patients general information and CT image features,and a radiomics model was constructed based on the extracted optimal CT radiomics features.The CT image features and radiomics scores were combined to build a fusion model.The diagnostic performance of each model in the validation set and the training set was compared.Results:Logistic regression analysis and LASSO feature selection,3 CT image features and 10 optimal radiomics features were obtained,and the CT feature model and CT radiomics model were successfully constructed.ROC curve analysis results showed that the AUC's of the CT feature model,the CT radiomics model,and the fusion model in the training set were 0.688,0.823 and 0.810,respectively.The AUC's of the three models in the validation set were 0.689,0.853 and 0.846,respectively.Conclusion:CT-based radiomics has good diagnostic performance in Fuhrman grading of ccRCC and high clinical application value.