Machine Learning for Predicting Future Surgery in Ileal Crohn’s Disease Using CT-Enterography

Researchers used machine learning (ML) to quantify cumulative ileal injury on CT-enterography (CTE) scans for patients with ileal Crohn’s disease (CD) to predict future bowel surgery. Analysis of 229 CTE scans with 8,424 mini-segments revealed strong agreement between ML and radiologists in grading bowel injury (κ=0.80), similar to inter-radiologist agreement (κ=0.87). Cumulative injury scores were higher in CD biologic users who underwent future surgery. Models using cumulative spatial metrics (AUC=0.76) outperformed those with conventional bowel measures (AUC=0.62). This approach may improve outcome prediction and aid personalized bowel assessment in CD.