Objective:

To develop and clinically validate a deep learning-based framework for quantitative assessment of endoscopic biopsies in Crohn's disease, addressing the limitations of conventional histologic assessment.

Key Findings:

• AI models showed robust internal performance with Dice coefficients and AUC values exceeding 0.95.
• AI-derived crypt abscess and cryptitis ratios demonstrated good concordance with pathologist assessments, indicating potential for clinical application.
• Weak agreement was noted for submucosal plasma cell density (CCC = 0.376), which may limit its clinical utility.
• AI-derived features showed weak to moderate associations with clinical or endoscopic activity, suggesting further investigation is needed.
• Greater submucosal inflammatory infiltration was observed in CD biopsies compared to non-CD inflammatory colitis, highlighting the distinct pathology of CD.

Interpretation:

The AI-based framework provides objective and reproducible tissue-level quantification of Crohn's disease histopathology, complementing routine pathologic assessment.

Limitations:

• AI over-detection in submucosal regions with negative or low manual scores, which may lead to misinterpretation of results.
• Associations between AI-derived features and clinical indicators were generally weak to moderate, indicating the need for cautious interpretation.

Conclusion:

The framework offers a practical pathologist-supervised approach for standardizing histologic evaluation in Crohn's disease, with potential for future enhancements to improve accuracy.