A Preliminary Study Of Enhanced CT Texture Analysis In Renal Cell Carcinoma

Part ? Feasibility study of enhanced CT texture parameters in differentiating malignant degree of renal clear cell carcinomaObjective:To explore the feasibility of texture analysis in preoperative assessment of pathological grade of clear cell renal cell carcinoma(ccRCC)Methods:Computed Tomography(CT)enhanced images were reviewed retrospectively in 64cases,which were divided into four groups according to their pathological grading of cancer,including 12 with? grade,20 with ? grade,22 with ? grade and 10 with ? grade.These tumors were then divided into low(?+?,32 cases)and high grade(?+?,32 cases)groups.Region of interest(ROI)was placed at axial CT enhanced images with maximum lesion and texture analysis was performed via Mazda software.The feature selection methods included mutual information(MI),Fishers coefficient,and classification error probability combined with average correlation coefficients(POE+ACC),which was used to identify the most significant texture features in discriminating high grade and low grade ccRCC and each of the method extracts 10 parameters.Finally,seven parameters were extracted that appear twice or more in these three methods.Two independent sample t-tests were conducted for parameters that conform to normal distribution,and Wilcoxon rank sum test was conducted for parameters that do not conform to normal distribution.The receiver operating characteristic curve(ROC)was established and area under curve(AUC)was obtained,and the diagnostic efficiency of each parameter was finally compared and analyzed.Result:In the 7 parameters extracted from the texture analysis among the 2 groups,WavEnLH-s-4 and Sigma had statistical difference(P<0.05).The AUC of ROC in WavEn-LH-s-4 was 0.88(66.67%sensitivity and 77.78%specificity,P<0.05).However,the AUC of ROC in sigma was 0.621(100.00%sensitivity and 37.04%specificity,P>0.05),which had no statistical significance.Conclusion:The texture analysis of CT enhanced images can provide quantitative characteristic parameters,and provide new ideas and methods for preoperative assessment of pathological grade of ccRCC.Part? Enhanced CT texture analysis combined with machine learning in the differentiation of renal cell carcinoma subtypes and pathological grading of renal clear cell carcinomaObjective:To explore the feasibility of texture analysis in preoperative assessment of pathological grade of clear cell renal cell carcinoma(ccRCC)and differentiating pathological subtypes of renal cell carcinoma.Methods:Computed Tomography(CT)enhanced images were reviewed retrospectively in 64 ccRCC cases,13 chromophobe renal carcinoma cases and 18 papillary renal cell carcinoma cases,which were divided into four groups according to their pathological grading of cancer,including 12 with ? grade,20 with ? grade,22 with? grade and 10 with ? grade.These tumors were then divided into low(?+?,32 cases)and high grade(?+?,32 cases)groups.Region of interest(ROI)was placed at axial CT enhanced images and texture analysis was performed via Mazda software.The feature parameters were selected by Fishers coefficient.The characteristic parameters were analyzed by different machine learning methods,which were verified by the remain-one cross-validation.Finally,the accuracy,sensitivity and specificity were used as the evaluation indexes of the models.Results:In the highly and poorly differentiated renal clear cell carcinoma groups,the diagnostic efficiency of decision tree was the highest,and the accuracy,sensitivity and specificity were 80.65%,76.67%and 84.38%,respectively.The diagnostic efficiency of the random forest model was the highest in the differentiation between ccRCC and renal non-transparent cell carcinoma,and the decision tree was the second.The diagnostic efficiency and sensitivity specificity of the decision tree were 88.61%,92.19%and 73.33%,respectively.The decision tree was of great value in the pathological grading of ccRCC and the differentiation of pathological subtypes of renal cell carcinoma,with high accuracy and balanced sensitivity and specificity.Conclusions:CT texture analysis combined with decision tree model contributes to evaluating the malignant degree of renal clear cell carcinoma and pathological subtype of renal cell carcinoma,which has certain clinical feasibility.