The following is a summary of “Integrating radiomics with the vesical imaging-reporting and data system to predict muscle invasion of bladder cancer,” published in the June 2023 issue of the Urologic Oncology by Wang et al.

This study aims to develop predictive models for determining muscle invasion of bladder cancer based on integrating radionics with the Vesical Imaging-Reporting and Data System (VI-RADS). From January 2015 to March 2022, 191 patients were retrospectively included in this study. About 7:3 random split of these 121 data sets into training and validation sets. The remaining (n = 70) data was an independent testing set. On high-b-value DWI images, radio mics features were extracted from bladder carcinoma. In the training set, radionics model pipelines were trained. The optimal model was selected based on the validation set’s performance. 

The chosen model was then evaluated using the independent testing set. The VI-RADS was reviewed by two radiologists using T2WI and DWI. Reader 1 was an accomplished reader, whereas Reader 2 was a novice. Integrating the radionics signature and VI-RADS produced a clinical-radiomics model. Utilizing analysis of the receiver operating characteristic curve, the efficacy was evaluated. The histopathological findings served as the benchmark for assessing the diagnostic precision of muscle invasion. The area under the curve (AUC) for the radionics model in the training, validation, and assessment sets, respectively, was 0.801, 0.867, and 0.806. 

In the training, validation, and assessment sets, the AUC values for Readers 1/2’s VI-RADS scores were 0.831/0.781, 0.909/0.815, and 0.871/0.776, respectively. In the training, validation, and testing sets, the clinical-radiomics model for Readers 1/2 yielded AUC values of 0.888/0.854, 0.961/0.919, and 0.881/0.844, respectively. (P<0.05) The efficacy of the clinical-radionics model was superior to the VI-RADS score for Reader 2 with limited experience. The radionics model was beneficial for diagnosing bladder cancer with muscle invasion. The clinical-radiomics model incorporating radionics and VI-RADS performed better than VI-RADS alone, which was advantageous for readers with less diagnostic experience.

Source: sciencedirect.com/science/article/abs/pii/S1078143922004240