| Objective: To investigate the value of MRI image features and texture analysis in preoperative prediction of WHO/ISUP nuclear grade in patients with clear cell renal cell carcinoma(cc RCC).Materials and Methods: MRI images of 109 patients diagnosed as cc RCC by surgical pathology from July 2016 to December 2020 in First Affiliated Hospital of Fujian Medical University were retrospectively analyzed.According to the WHO/ISUP grading system,the patients were divided into low grade group(68 cases,gradeⅠ in 4 cases and grade Ⅱ in 64 cases)and high grade group(41 cases,grade Ⅲ in 33 cases and grade Ⅳ in 8 cases).MRI radiological features were evaluated and MRI imaging texture features were extracted.The largest diameter sclice of lesion on crosssectional images was selected and ROIs were drawn on images.Quantitative texture analysis software Ma Zda was used to extract texture features,including gray-scale histogram,co-occurrence matrix,run-length matrix,gradient,autoregressive model and wavelet transform.Two algorithms,Fisher coefficient and mutual information in Mazda software,were used to select the extracted quantitative features,and then the reduced texture parameters or imaging features were tested by the independent sample t test,Mann-Whitney U test or Chi-square test.Variables with statistically significant differences were used to construct a multi-factors binary logistic regression model and the ROC curve was used to analyze its effectiveness in predicting high grade cc RCC.Results There were significant differences in tumor length,shape and margin,intratumoral hemorrhage,intratumoral vessels,peritumoral vessels,cystic-solid,renal sinus involvement,vein thrombosis,and intratumoral necrosis and 20 texture features between the low and high grade cc RCC groups.Five univariate multivariate binary logistic regression models were constructed,including MRI image features model(M1),texture feature“S(4,0)Entropy”model(M2),texture feature“S(5,-5)Entropy”model(M3),combined MRI image features with texture feature “S(4,0)Entropy”model(M4),combined MRI image features with texture feature “S(5,-5)Entropy model”(M5).The areas under the ROC curve of the diagnostic models M1,M2,M3,M4 and M5 were0.783(95% CI 0.692～0.874),0.746(95% CI 0.656～0.836),0.749(95% CI0.660～0.838),0.848(95% CI 0.774～0.922)and 0.849(95% CI 0.774～0.924),respectively.The sensitivity,specificity,positive predictive value and negative predictive value of high grade cc RCC were respectively 48.8%,85.3%,66.7%,73.4%(M1);56.1%,75.0%,57.5%,73.9%(M2);56.1%,75.0%,57.5%,73.9%(M3);65.9%,88.2%,77.1%,81.1%(M4);65.9%,86.8%,75.0%,80.1%(M5).MRI image feature combined texture feature “S(4,0)Entropy”model(M4),MRI image feature combined texture feature “S(5,-5)Entropy model”(M5).Comparison of the combined MRI image features with texture feature “S(4,0)Entropy”logistic regression model(M4)and combined MRI image features with texture feature “S(5,-5)Entropy” logistic regression model(M5)with MRI image features logistic regression model(M1)showed that the diagnostic efficiency of the two pairs of models for diagnosing high-grade cc RCC were comparable(P > 0.05).However,the diagnostic performance of the combined MRI image features with texture feature “S(4,0)Entropy”logistic regression model(M4)and combined MRI image features with texture feature “S(5,-5)Entropy” logistic regression model(M5)are superior to any single texture feature.Conclusion Based MRI image features model can preoperatively predict WHO/ISUP nuclear grade.The combination of MRI image features and texture analysis is helpful to accurately distinguish the nuclear grade of cc RCC,which has better diagnostic efficiency than the application of texture features alone.MRI texture analysis combined with imaging features is promising to be an effective preoperative noninvasive method in predicting WHO/ISUP grade of cc RCC. |
PDF Full Text Request