A Systematic Review of Artificial Intelligence for Endoscopic and Histologic Assessment in Ulcerative Colitis

Artificial intelligence (AI) shows significant potential to standardize the assessment of ulcerative colitis (UC), but a comprehensive synthesis of its performance across the care pathway is needed. This systematic review critically appraises AI applications in endoscopic video analysis, digital histopathology, and multimodal approaches for UC. We systematically searched MEDLINE, Embase, Scopus, and other databases from ۲۰۱۹ to ۲۰۲۵ for studies applying AI to human UC care. We focused on diagnostic accuracy, prognosis, and operational deployment. Study selection, data extraction, and risk-of-bias assessment (using JBI and PROBAST tools) were performed by two independent reviewers. A narrative synthesis was prespecified due to heterogeneity. From ۱,۸۴۷ identified records, ۳۲ studies were included. Video-level AI models for endoscopic scoring (e.g., Mayo, UCEIS) demonstrated strong agreement with central readers (κ up to ۰.۹۲) and are being evaluated against regulatory standards. Histology-based AI systems accurately classified activity using indices like PHRI and Nancy (sensitivity/specificity ۸۹%/۸۵%) and significantly stratified ۱۲-month flare risk (HR ۴.۶۴). Multimodal fusion of endoscopy and histology outperformed single-modality approaches for assessing histologic remission and treatment response in clinical trials. Diagnostic differentiation of UC from Crohn's disease achieved high accuracy (۹۹.۱% in one study), while complication detection (e.g., CMV colitis) remains emergent. Common limitations included frame-level analysis with patient-level claims, infrequent external validation, and sparse reporting on calibration and fairness. AI demonstrates expert-level proficiency in standardizing UC activity assessment and shows prognostic value, particularly with histology and multimodal data. For successful clinical integration, future work must prioritize robust external validation, standardized outcome reporting, and prospective evaluation of decision-making impact.

Keywords:

Artificial intelligence , Colitis , Ulcerative , Systematic review , Endoscopy , Gastrointestinal , Machine Learning