Differential Diagnosis of Malignancy in Renal Clear Cell Carcinoma using Enhanced CT Imaging Omics
Objective To explore the value of imaging radiomics based on CT enhanced images in differentiating the malignancy degree of renal clear cell carcinoma.Methods The enhanced CT images of 192 patients of renal clear cell carcinoma(CCRCC)confirmed by pathology were analyzed retrospectively,including the poorly differentiated group(grade Ⅰ-Ⅱ.n=111)and the highly differentiated group(gradeⅢ-Ⅳ.n=81).the radiomics features were extracted from the enhanced CT images of cortical medullary phase(CMP),renal parenchymal phase(NP),excretory phase(EP)and the combination of the three phases,the dimensionality was reduced by the least absolute shrinkage and selection operator(LASSO)regression method,and the value radiomics features were divided into the training group and the test group by 50％cross-validation.The training group was established by using support vector machine(SVM)and logistic regression(LR)classifiers.CMP、NP、EP and three-phase combined imaging models were used to evaluate the diagnostic efficacy of the imaging model for the malignancy degree of CCRCC by using the area under the subject operating characteristic curve(AUC),accuracy,sensitivity,specificity and accuracy,and further verified by the test group.Results The imaging model based on CMP、NP、EP and three-phase combined images was significantly correlated with the malignancy degree of CCRCC and the CMP imaging mode had the highest diagnostic efficiency for the malignancy degree of CCRCC(R=0.831,0.801).The area under ROC curve(AUC)values of CMP,NP,EP and three-phase combined diagnostic efficacy of SVM classifier model test group were 0.819、0.785、0.808、0.812 respectively.The CMP、NP、EP and three-phase combined AUC values of LR classifier model test group were 0.860、0.789、0.808、0.799 respectively.In SVM and LR classifiers,there were significant differences in AUC values between CMP and EP、CMP and NP image radiomics models(P＜0.05),and no differences between NR and EP models(P＞0.05).Both kinds of classifiers have better predictive performance,and the performance of the model established by SVM classifier is more stable and comprehensive.Conclusion The imaging model based on enhanced CT imaging features has a clinical guiding role in distinguishing the malignant degree of renal clear cell carcinoma,and the performance of the imaging model established by SVM classifier is more stable and comprehensive
万方数据
维普
