CT增强影像组学对肾透明细胞癌和非透明细胞癌的鉴别诊断价值  

Diagnositic value of CT-enhanced imaging omics in clear cell renal cell carcinoma and non-clear cell renal cell carcinoma

作　　者：刘晨露 翟建[1] 陈基明[1] 喻泓清 肖国庆 LIU Chenlu;ZHAI Jian;CHEN Jiming;YU Hongqing;XIAO Guoqing(Medical Imaging Center,Yijishan Hospital,Wannan Medical College,Wuhu 241001,China)

机构地区：[1]皖南医学院弋矶山医院影像中心,安徽芜湖241001

出　　处：《沈阳医学院学报》2022年第6期611-615,共5页Journal of Shenyang Medical College

摘　　要：目的:探讨CT增强影像组学对肾透明细胞癌(ccRCC)和非透明细胞癌(non-ccRCC)的鉴别诊断价值。方法:回顾性分析我院2012年1月至2020年12月收治的经组织病理学确诊的80例ccRCC和54例non-ccRCC(包括29例乳头状细胞癌、25例嫌色细胞癌)的临床资料。患者术前均行肾脏CT动态增强扫描。使用RadiAnt-DICOM软件将图像导入,再使用ITK-SNAP软件人工逐层勾画各期相感兴趣区(ROI),生成三维感兴趣区容积(VOI),应用AK软件提取纹理特征。对各期CT影像组学特征数据进行归一化处理,以最小冗余最大相关(mRMR)和最小绝对值收敛和选择算子(LASSO)回归分析进行特征选择和建立影像组学标签,并用100次留组交叉验证(LGOCV)验证模型的可靠性;然后使用ROC曲线评价模型的鉴别诊断能力。结果:AK软件从皮髓质期、实质期和排泄期各提取1316个纹理特征,经LASSO回归降维得到皮髓质期、实质期、排泄期和三期联合最具有预测价值的特征分别为12、13、15和9个。皮髓质期、实质期、排泄期和三期联合模型在训练组和验证组中鉴别曲线下面积(AUC)分别为0.91和0.94、0.88和0.88、0.88和0.86以及0.92和0.94。结论:CT增强影像组学构建的4个模型对术前鉴别ccRCC和non-ccRCC均有较好的诊断效能,三期联合模型预测效能最好。Objective:To investigate the value of CT-enhanced imaging omics in the differential diagnosis of clear cell renal cell carcinoma(ccRCC)and non-clear cell renal cell carcinoma(non-ccRCC).Methods:The clinical data of 80 cases of ccRCC and 54 cases of non-ccRCC(including 29 papillary cell renal carcinoma and 25 chromophobe renal cell carcinoma)diagnosed by histopathology were retrospectively analyzed.All patients underwent renal dynamic enhanced CT scan before operation.RadiAnt-DICOM software was used to import the image and convert it into DICOM.Then ITK-SNAP software was used to manually delineate the region of interest(ROI)of each phase,and finally generate 3D VOI.And AK software was used to extract texture features.The CT radiomics feature data of each phase was normalized,and minimum redundancy maximum relevance(mRMR)and least absolute shrinkage and selection operator(LASSO)regression analysis were used to perform feature selection and establish radiomics tags,and 100 leave-group out cross-validation(LGOCV)was used to verify the reliability of the model.ROC curve was used to evaluate the differential diagnosis ability of the model.Results:A total of 1316 texture features were extracted by AK software from the cortico-medulla phase,parenchymal phase and excretory phase,respectively.LASSO regression analysis showed that the most predictive features of cortico-medulla phase,parenchymal phase,excretory phase and three-phase combination were 12,13,15 and 9,respectively.The area under the curve(AUC)of cortico-medulla phase,parenchymal phase,excretory phase and three-phase combination models in the training group and the testing group were 0.91 and 0.94,0.88 and 0.88,0.88 and 0.86,0.92 and 0.94,respectively.Conclusion:The four models constructed by CT texture analysis have good diagnostic efficacy in differentiation of ccRCC and non-ccRCC before operation,and three-phase combined model has the best predictive efficacy.

关 键 词：肾细胞癌 透明细胞癌 非透明细胞癌 影像组学 

分 类 号：R737.11[医药卫生—肿瘤]