Intra- and Peritumoral CT-Based Radiomics for Assessing Pathologic T-Staging in Clear Cell Renal Cell Carcinoma: A Multicenter Study.

Abstract

Background: A radiomics model constructed from the intratumoral region of computed tomography (CT) can predict the pathologic T stage of clear cell renal cell carcinoma (ccRCC). However, the predictive capability of the radiomics model that incorporates both intra- and peritumoral regions of CT for the pathologic T stage in ccRCC patients has not been reported to date.

Methods: This study enrolled 250 patients with ccRCC who underwent laparoscopic surgery. Three radiomics models were developed based on the intra- and peritumoral regions. The sensitivity, specificity, accuracy, and receiver operating characteristic (ROC) curves of each model were analyzed. Decision curve analysis (DCA) and calibration curves were used to assess the net benefit and calibration ability of the models. Additionally, the diagnostic performance of the different models were compared with that of radiologists.

Results: The radiomics model based on the intra- and peritumoral regions at 5 mm exhibited the strongest performance, with area under curve values of 0.91 (95 % confidence interval [CI], 0.8551-0.9650), 0.85 (95 % CI, 0.7490-0.9517), and 0.873 (95 % CI, 0.7612-0.9839) in distinguishing high and low T stages of ccRCC across the training, validation, and test sets, respectively. The model's accuracy in the training, validation, and test sets was 0.798, 0.732, and 0.769, with corresponding sensitivity values of 0.921, 0.857, and 0.882, and specificity values of 0.747, 0.690, and 0.729. The calibration curve demonstrated a high level of agreement between the predicted and actual outcomes, whereas the DCA showed that the model provided a meaningful net benefit.

Conclusions: The radiomics model based on the intra- and peritumoral regions of CT has certain value in distinguishing between high and low T stages of ccRCC.