An MRI-based radiomics nomogram in predicting histologic grade of non-muscle-invasive bladder cancer

BackgroundNon-muscle-invasive bladder cancer (NMIBC) is categorized into high and low grades with different clinical treatments and prognoses. Thus, accurate preoperative evaluation of the histologic NMIBC grade through imaging techniques is essential. ObjectivesTo develop and validate an MRI-based radiomics nomogram for individualized prediction of NMIBC grading. MethodsThe study included 169 consecutive patients with NMIBC (training cohort: n = 118, validation cohort: n = 51). A total of 3148 radiomic features were extracted, and one-way analysis of variance and least absolute shrinkage and selection operator were used to select features for building the radiomics score(Rad-score). Three models to predict NMIBC grading were developed using logistic regression analysis: a clinical model, a radiomics model and a radiomics-clinical combined nomogram model. The discrimination and calibration power and clinical applicability of the models were evaluated. The diagnostic performance of each model was compared by determining the area under the curve (AUC) in receiver operating characteristic (ROC) curve analysis. ResultsA total of 24 features were used to build the Rad-score. A clinical model, a radiomics model, and a radiomics-clinical nomogram model that incorporated the Rad-score, age, and number of tumors were constructed. The radiomics model and nomogram showed AUCs of 0.910 and 0.931 in the validation set, which outperformed the clinical model (0.745). The decision curve analysis also showed that the radiomics model and combined nomogram model yielded higher net benefits than the clinical model. ConclusionA radiomics-clinical combined nomogram model has the potential to be used as a non-invasive tool for the differentiating low-from high-grade NMIBCs.

Automated summary

Accurate preoperative evaluation of NMIBC grade is crucial for individualized treatment. A radiomics nomogram based on MRI was developed and validated for predicting NMIBC grading. The model showed high accuracy and clinical applicability in differentiating low-grade from high-grade NMIBCs.