Objective To predict the expression of programmed death-1(PD-1)in tumor cells of patients with rectal cancer by using multiparameter MRI radiomics nomogram.Methods A total of 168 patients with pathologically confirmed rectal adenocarcinoma were retrospectively enrolled.All patients underwent preoperative multiparametric MRI and postoper-ative PD-1 immunohistochemistry(IHC)analysis.They were randomly divided into training group(n=118)and test group(n=50)according to the ratio of 7:3.The performance of T2 WI,DWI and T2 WI+DWI radiomics models were con-structed and compared.The clinical independent risk factors were screened by univariate and multivariate logistic regression to establish the clinical model.Finally,a joint model combining radiomics score and clinical features was constructed and presented as a visual nomogram.The area under the curve(AUC),calibration curve and decision curve were used to evalu-ate the clinical value of the model.Results After feature screening,8,5 and 9 radiomics features were retained in T2WI,DWI and combined sequences,respectively.The performance of PD-1 expression in each model training set and test set was evaluated by AUC(test set,T2WI was 0.64,DWI was 0.66,T2WI+DWI was 0.74).The combined mode was significantly better than the single imaging mode.T and N stages were independent risk factors(P＜0.05),and the AUC value of the clinical model was 0.69.Finally,the combined model based on T2WI+DWI sequence combined with clinical independent predictors was 0.75,which was significantly better than radiomics and clinical models alone.Conclusion The nomogram based on multiparameter MRI radiomics characteristics combined with clinical factors can effectively predict the expression status of PD-1 in rectal cancer patients.
NETL
NSTL
维普
万方数据
