CT Radiomic Analysis Of Bladder Cancer For The Assessment Of Molecular Characteristics And Intratumor Heterogeneity

Objective The purpose of this study is to obtain the molecular characteristics and intratumor heterogeneity of the bladder cancer by immunohistochemical profile of the tumor tissue of patients,and analyze the quantitative characteristics extracted from CT imaging by radiomics,so as to study the correlation between the molecular characteristics of bladder urothelial carcinoma and CT radiomics,To explore the intratumor heterogeneity of CT radiomics texture analysis in differentiation of bladder urothelial carcinoma.Methods Ninety-six patients with clinicopathologically confirmed BC which admitted to the Department of Urology of Baotou Central Hospital from June 2017 to June 2021 were included in this retrospective study.Including 51 high-grade urothelial carcinomas(HGUC)and 45 low-grade urothelial carcinomas(LGUC).Including 54 non-muscle invasive bladder carcinomas(NMIBCs,stage <T2)and 42 muscle invasive ones(MIBCs,stage >/=T2).Including 34 Luminal subtypes and 62 Basal subtypes.These patients were treated by transurethral resection of bladder tumor(TURBT),and the postoperative pathological results confirmed bladder urothelial carcinoma.We used two combinations of immunohistochemical markers,i.e.,cytokeratin(CK)5 with CD44 and CK20 with GATA3,to distinguish subtypes.Among them,there were 27 patients with high expression of CK5/6,62 patients with high expression of CD44,26 patients with high expression of CK20,and 82 patients with high expression of GATA-3.The patients with pathologically confirmed bladder caner underwent unenhanced and contrast enhanced CT imaging preoperatively with the same scanning parameters.Regions of interests(ROIs)of computerized tomography(CT)images were manually segmented and saved,and the selected image is imported into matlab2020 a in DICOM format.Texture analysis was performed on unenhanced and contrast enhanced CT images,respectively.The mean gray-level intensity(Mean),standard deviation,entropy,skewness,and kurtosis were obtained.One-way analysis of variance or independent-samples t-test(normality),independent-samples Kruskal-Wallis test or Mann-Whitney U test(non-normality),Chi-square test,binary Logistic regression analysis(multivariable),receiver operating characteristic(ROC)curve analysis were used for statistical analyses.Statistical analysis took bilateral tests,and significance test level P=0.05.Reaults Our results showed that the mean and entropy of HGUC texture parameters were higher than LGUC(P=0.000).The mean and entropy of MIBC texture parameters were higher than NMIBC(P=0.001,P=0.000).The standard deviation and kurtosis of HGUC texture parameters were lower than LGUC(P=0.000,P=0.006).The entropy of CK5/6 high expression group was higher than CK5/6 low expression group(P=0.013).The mean and entropy of CD44 high expression group was higher than CD44 low expression group(P=0.000).The skewness and kurtosis of CD44 high expression group was lower than CD44 low expression group(P=0.019,P=0.005).The entropy of GATA3 high expression group was lower than GATA3 low expression group(P=0.000).The mean and entropy of Basal subtypes group was higher than Luminal subtypes group(P=0.016,P=0.000),and the kurtosis of Basal subtypes group was lower than Luminal subtypes group(P=0.015).For identifying the molecular subtypes of bladder cancer,entropy showed good performance with area under ROC curve area under curve(AUC)of 0.790(95%CI: 0.685-0.895),The‘Youden index’ is 0.638,and the CUT-OFF is 4.73265,respectively.Conclusion Differences of the CT imaging features were observed between molecular characteristics in our study.CT texture analysis of radiomics could evaluate the histologic grade and staging of bladder urothelial carcinoma.We have shown that CT imaging features could predict the molecular subtypes of bladder cancer,entropy was the optimal diagnostic parameter.