Objectives <p>Ustekinumab is a biological treatment for Crohn’s disease, but some patients do not respond. This study aimed to assess the role of radiomic techniques in predicting the treatment response by quantifying transmural inflammation in Crohn’s disease.</p> Materials and methods <p>A total of 296 patients (training cohort, <i>n</i> = 207; testing cohort, <i>n</i> = 89) were retrospectively recruited. Manual segmentation of 3D volumes of interest (VOIs) encompassing inflamed bowel wall segments was performed on arterial-phase CT enterography scans, from which radiomic features were extracted. Following feature dimensionality reduction via Pearson correlation filtering (threshold &gt; 0.9) and recursive feature elimination, the least absolute shrinkage and selection operator (LASSO) logistic regression was utilized as a classifier to construct a radiomic signature. Subsequently, to leverage both radiomic and clinical information for optimal prediction, the radiomic signature was integrated with clinically accessible variables (C-reactive protein level, prior biologic exposure) to develop a model predicting UST efficacy at 16 weeks. The predictive performance was compared using the area under the curve (AUC) and calibration curve analysis. Clinical utility was assessed by decision curve analysis.</p> Results <p>The radiomic signature, based on 1,288 features, was an independent risk factor for Ustekinumab response, with area under the curve values of 0.819 in the training cohort and 0.791 in the testing cohort. An integrated model combining the radiomic signature, C-reactive protein levels, and prior biologics exposure achieved area under the curve values of 0.847 and 0.801, respectively. The model demonstrated good calibration and clinical benefit.</p> Conclusions <p>The baseline radiomic signature is a promising biomarker for predicting Ustekinumab treatment efficacy in Crohn’s disease.</p>