基于三期增强CT影像组学和临床特征在小肾透明细胞癌鉴别诊断中的应用分析  

Application of Radiomics Based on Contrast-Enhanced CT and Clinical Features in the Differential Diagnosis of Small Clear Cell Renal Cell Carcinoma and Small Non-Clear Cell Renal Cell Carcinoma

作　　者：乔冠中 刘冬敏 余卓 姜慧杰[1] QIAO Guanzhong;LIU Dongmin;YU Zhuo(Department of Radiology,The Second Affiliated Hospital of Harbin Medical University,Harbin,Heilongjiang Province 150086,P.R.China)

机构地区：[1]哈尔滨医科大学附属第二医院影像科,150086

出　　处：《临床放射学杂志》2023年第8期1284-1291,共8页Journal of Clinical Radiology

基　　金：国家重点研发计划资助项目(编号:2019YFC0118100);国家自然科学基金面上项目(编号:81671760;81873910;62171167)。

摘　　要：目的探讨基于三期增强CT的影像组学和临床特征在小肾透明细胞癌与小肾非透明细胞癌的鉴别诊断中的应用价值。方法回顾性分析102例(透明细胞肾细胞癌75例,非透明细胞肾细胞癌27例)经病理诊断为肾细胞癌(RCC)的患者的包括性别、年龄、肿瘤最大径、肿瘤位置(左肾或右肾)、出血、钙化、囊变坏死、中心瘢痕、假包膜在内的临床信息及患者的三期增强CT图像,在每一期的图像上逐层勾画感兴趣区(ROI);提取病灶的纹理特征;利用方差阈值法、K最佳法、LASSO模型进行降维筛选,分别在皮质期、实质期、排泄期筛选出来20、11、14个最佳特征;利用这些影像组学特征和具有统计学意义的临床特征构建基于支持向量机(SVM)模型的单期、两期联合、三期联合分类器模型;通过受试者工作特征(ROC)曲线和校准曲线来评价模型的性能。结果患者仅假包膜一项临床信息与小肾癌的病理类型的差异具有统计学意义(P=0.001),其余的临床信息包括:性别、肿瘤位置、出血、钙化、囊变坏死、中心瘢痕、年龄、肿瘤最大径与小肾癌病理类型的差异均无统计学意义(P>0.05);在测试集上,实质期模型达到曲线下面积(AUC)=0.59,排泄期模型达到AUC=0.60,皮质期+实质期模型达到AUC=0.75,皮质期+排泄期模型达到AUC=0.77,实质期+排泄期模型达到AUC=0.73,三期联合模型达到AUC=0.73。皮质期模型表现最佳:AUC=0.79。结论基于皮质期增强CT的影像组学与临床特征可用于小肾透明细胞癌和小肾非透明细胞癌的鉴别诊断。Objective To investigate the value of radiomicsbased on contrast-enhanced CT and clinical features in the differential diagnosis of small clear cell renal cell carcinoma and small non-clear cell renal cell carcinoma.Methods The retrospective study included 102 cases(75 clear cell carcinoma,27 Cases of non-clear cell carcinoma).Clinical features including sex,age,maximum tumor diameter,tumor location(left or right kidney),bleeding,calcification,cystic necrosis,central scar,pseudocapsule,and enhanced third-stage CT images of patients,The ROI is outlined layer by layer on the image of each issue;Extract the texture features of the lesions;Dimension reduction screening using variance threshold method,select K best,LASSO model,20,11 and 14 best characteristics were selected in cortical,parenchymal and excretion periods;These imaging omics features and statistical significant clinical features were used to construct the single-stage,two-stage joint and three-stage joint classifier models based on the SVM model;The model performance was evaluated by the subject operating characteristic(ROC)curve,calibration curve,and decision curve.Results The difference between the patients with only one false capsule and clinical information and the pathological type of small renal carcinoma was statistically significant(P=0.001).The remaining clinical information included:gender,tumor location,bleeding,calcification,cystic necrosis,central scar,age,and tumor maximum diameter and small kidney cancer pathological type(P>0.05).On the test set,the me-dullary phase model’s AUC reached 0.59,the nephrographic phase model’s AUC reached 0.60,the cortical phase plus medullary phase model’s AUC reached 0.75,the cortical phase plus nephrographic phase model’s AUC 0.77,the medullary phase plus nephrographic phase model’s AUC 0.73,the three-phase joint model’s AUC 0.73.The cortical phase model performed best,AUC=0.79.Conclusion The radiomics based on contrast-enhanced CT and clinical features can be used in the differential diagnosis o

关 键 词：小肾癌 透明细胞肾细胞癌 非透明细胞肾细胞癌 体层摄影术 X线计算机 影像组学 

分 类 号：R737.11[医药卫生—肿瘤] R730.44[医药卫生—临床医学]