Multi-modal ensemble learning prediction of anti-TNF treatment response in pediatric patients with Crohn's disease using clinical, radiomic, and deep learning MR enterography features.

The multi-modal ensemble model integrating all magnetic resonance enterography-based features achieved an AUROC of 0.771 [95% confidence intervals (CI): 0.745, 0.797], outperforming models using individual feature types (radiologist assessment: 0.706 [0.691, 0.721]; radiomic: 0.710 [0.691, 0.729]; deep learning: 0.724 [0.700, 0.749]).

Radiomic features reflecting intensity distribution and texture heterogeneity were most predictive of treatment response.