Enhanced CT-based Texture Analysis And Radiomics Score In Evaluating Fuhrman Nuclear Grading Of Clear Cell Renal Carcinoma

Enhanced CT-based texture analysis in evaluating Fuhrman nuclear grading of clear cell renal carcinomaObjective To evaluate the feasibility of quantitative CT texture analysis in evaluating the nuclear grade of clear cell renal carcinoma(clear cell renal cell carcinoma,cc RCC).Methods The data of 78 patients with cc RCC confirmed by pathology were retrospectively analyzed.There were 15 cases of gradeⅠ,37 cases of grade Ⅱ,17 cases of grade Ⅲ and 9cases of grade Ⅳ.According to Fuhrman classification,I and II grades were regarded as low grade group(52 cases),and Ⅲ and Ⅳ grade as high grade group(26cases).According to Fuhrman classification,I and II grades were regarded as low grade group(52 cases),Ⅲ and Ⅳ grades were regarded as high grade group(26 cases).Matlab2014 a software was used to analyze the texture of enhanced CT cortical and medullary phase images in all patients.4 gray histogram feature parameters(Mean,SD,Kurtosis,Skewness)and 4 gray level co-occurrence matrix characteristic parameters(Contrast,Correlation,Energy,Consistency)were extracted,all the parameters are analyzed and compared.Results There were significant differences in Mean,SD and Correlation between patients with different Fuhrman grades(P<0.05),but there were no significant differences among the other 5 features(Kurtosis,Skewness,Contrast,Energy,Consistency)(P < 0.05).The correlat-ion of 10 texture feature parameters(Mean,SD,Kurtosis,Skewness,Contrast,Correlation,Energy,Consistency,homogeneity,entropy)and Fuhrman grade showed that only the Correlation was of great clinical value,with |r| value of 0.382(P<0.05).The Correlation of cc RCC in high grade group was higher than that in low grade group,the difference was statistically significant(P=0.001),but there was no significant difference between residual eigenvalues(P>0.05).The area under ROC curve of low grade clear cell renal carcinoma was 0.753.The sensitivity and specificity of diagnosis were76.92% and 71.15%,respectively.The correlation coefficient between the texture features 0.241~0.975,there is obvious correlation in some texture features(|r|≥0.5),the correlation between SD and Mean is the highest(|r|=0.975).The correlation between Mean and Energy,Kurtosis and Skewness is also higher(|r|=0.948、|r|=0.954).Conclusion There is a certain clinical value in evaluating Fuhrman grade of cc RCC with CT texture analysis.Enhanced CT-based radiomics signature combined with clinical features in evaluating nuclear grading of clear cell renal carcinomaObjectives To explore the value of enhanced CT-based radiomics signature combined with clinical features in evaluating nuclear grading of renal clear cell carcinoma.Methods The data of 101 patients with cc RCC confirmed by pathology and Fuhrman grading were analyzed retrospectively.Among them,17 were grade I,52 were grade II,19 were grade III and 13 were grade IV.According to Fuhrman grading,grade I and II were classified as low grade group(69 cases)and grade III and IV as high grade group(32cases).Using ITK-SNAP20.0 software(www.itksnap.org)to select the maximum diameter level of the lesion for ROI delineation and segmentation,all the original images were standardized after processing,using A.K software(GE health care,Analysis Kit,Version: 3.2.0.R)for feature extraction,the data were divided into training set and validationset by 7:3,m RMR were primarily applied to select the minimum redundant and maximum relevant feature subsets,then LASSO logistic regression was used to screen out the best feature subset in each phase and construct the radiomics score.classification efficiency of the integrated radiomics model was evaluated by drawing the ROC and calculating the area under the ROC curve(AUC).Fisher test was used to accurately test the distribution differences of clinical features between the two groups.The characteristics of P < 0.1 were analyzed in univariate Logistic regression,features with significant differences were further analyzed by multivariate logistic regression.The clinical features with high correlation with the rad_score were screened,combined with rad_score a nomogram was constructed.ROC curve and calibration curve were used to evaluate the performance of the model,and decision curve analysis(DCA)was used to evaluate the clinical utility of the model.Results The classification efficiency of enhanced three-stage integrated histology model was higher than that of each phase model.The AUC of cortical,parenchymal,excretory and three-stage integrated histology model in the training set were 0.82,0.76,0.74 and 0.86,respectively.The clinical features with statistically significant(p < 0.05)by multivariate logistic regression analysis were: gender,maximum diameter.The AUC,sensitivity and specificity of the combined model was 0.90,73.9%,86.4% for training set and 0.7387.0%,60.0% in validation set,respectively..DCA curve shows that the net benefit of the combined model is higher than that of the radiomics model and the clinical characteristic model,which means higher clinical utility of the combined model.Conclusion It is feasible to evaluate the nuclear grading of renal clear cell carcinoma based on enhanced CT radiomics signature combined with clinical features,which is of guiding significance for the early precise treatment and prognosis evaluation of patients with renal clear cell carcinoma.