A novel computed tomography enterography radiomics combining intestinal and creeping fat features could predict surgery risk in patients with Crohn’s disease

Abstract

Objective The objective of this study is to segment creeping fat and intestinal wall on computed tomography enterography (CTE) and develop a radiomic model to predict 1-year surgery risk in patients with Crohn’s disease. Methods This retrospective study included 135 Crohn’s disease patients who underwent CTE between January and December 2021 (training cohort) and 69 patients between January and June 2022 (test cohort). A total of 1874 radiomic features were extracted from the intestinal wall and creeping fat respectively on the venous phase CTE images, and radiomic models were constructed based on the selected features using the Boruta and extreme gradient boosting algorithms. The combined models were established by integrating clinical predictors and radiomic models. The receiver operating characteristic curve, calibration curve, and decision curve analyses were used to compare the predictive performance of models. Results In the training and test cohorts, the area under the curve (AUC) values of the creeping fat radiomic model for surgery risk stratification were 0.916 and 0.822, respectively, similar to the intestinal model with AUC values of 0.889 and 0.822. Moreover, the combined radiomic model was superior to the single models, showing good discrimination with the highest AUC values (training cohort: 0.963; test cohort: 0.882). Addition of clinical predictors to the radiomic models failed to significantly improve the diagnostic ability. Conclusion The CTE-based creeping fat radiomic model provided additional information to the intestinal radiomic model, and their combined radiomic model enables accurate surgery risk prediction of Crohn’s disease patients within 1 year of CTE.