Abstract
To develop magnetic resonance imaging (MRI)-based radiomics models for the prediction of the pathological grade and histological variant of bladder cancer. A total of 227 patients who underwent bladder MRI and had histopathologically confirmed grades and variants were included retrospectively from January 2017 to March 2022. They were assigned to a training set (n=131) and a testing set (n=96) based on the MRI system. MRI-based radiomics features were extracted from manually segmented volumes of interest from high-b-value DWI images and ADC maps. The radiomics models were trained with all possible pipelines in the training set. One optimal model was selected using the fivefold cross-validation method and verified by the testing set according to the pathological results. Univariate and multivariate analyses were performed to identify the significant clinical and imaging factors for developing clinical-radiomics models. The radiomics model for grade prediction had area under the curve (AUC) values of 0.784, 0.786, and 0.733 in the training, cross-validation, and testing sets, respectively. The radiomics model for variant prediction had AUC values of 0.748, 0.757, and 0.789 in the training, cross-validation, and testing sets, respectively. The performance of the clinical-radiomics model was significantly improved compared with the radiomics models alone for the total dataset (AUC for grade: 0.846 versus 0.756; AUC for variant: 0.810 versus 0.757, p<0.05). MRI-based radiomics models could be used to predict the pathological grade and histological variants of bladder cancer with relatively good performance.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have