Objective To investigate the value of the radiomics model based on CT images in predicting postoperative recurrence of patients with non-muscle-invasive bladder cancer(NM1BC).Methods The clinicopathological data of 311 patients with NMIBC who underwent transurethral resection of bladder tumor(TURBT)at our institution between January 2014 and December 2017 were collected.The patients were randomly classified into training(n=218)and validation(n=93)sets in a 7∶3 ratio.Radiomics features were extracted from the preoperative CT images using the PyRadiomics package.We used the least absolute shrinkage and selection operator(LASSO)regression model to select the optimal com-bination of the radiomics features and to establish a prognostic classifier,the Rad score.The COX proportional hazards mod-el was used to identify the independent prognostic factors,and a nomogram was then constructed based on these factors.The predictive accuracy,good fit,and clinical utility of the nomogram were evaluated in the training and validation sets,and were compared with the EOTRC recurrence score.Results Using the LASSO model,we identified 5 radiomics features and de-veloped a formula to calculate the Rad score.Multivariate analysis revealed that pT stage,histological grade,and the Rad score were independent predictors of recurrence(P＜0.05).A Rad nomogram combining these 3 factors was constructed based on these factors.The nomogram had a good fit as well as higher AUC values(training set:0.762 vs.0.702;valida-tion set:0.828 vs.0.763)and better clinical utility than the EOTRC recurrence score.The Rad nomogram divided the en-tire cohort into three risk groups,with 5-year recurrence rates of 15.6％,45.2％,and 89.6％,respectively(P＜0.001).Moreover,The Rad recurrence grading can differentiate between those with different risks of recurrence in the low,interme-diate,and high risk groups of EORTC(P for trend＜0.05).Conclusion The CT radiomics model can accurately pre-dict postoperative recurrence of NMIBC patients following TURBT,which may assist in clinical decision process.
NETL
NSTL
万方数据
