Xiangling Lv, Chen Wu
Inflammatory bowel disease (IBD) is a group of chronic inflammatory bowel disorders characterized by complex etiology and significant clinical heterogeneity. With the evolution of the "treat-to-target" (T2T) concept, progressive endpoints such as endoscopic mucosal remission, histological remission, and deep remission have become key outcomes in the management of IBD. Traditional endoscopic assessment of IBD suffers from issues such as high subjectivity, lack of consistency, limited quantitative capabilities, and reliance on specialist experience; artificial intelligence (AI) is driving the evolution of endoscopic assessment toward standardization, objectivity, and real-time analysis. AI has made significant progress in areas such as UC activity scoring, CD ulcer identification, small bowel capsule endoscopy image analysis, relapse prediction, and tumor monitoring; in selected datasets or experimental settings, some of its performance metrics approach expert levels. Research trends are gradually shifting from single-image analysis toward multimodal decision-support systems that integrate endoscopic, pathological, biomarker, and clinical information, which are expected to enhance the objectivity of IBD diagnosis, treatment, and efficacy evaluation. However, current challenges include data heterogeneity, inconsistent standards, insufficient external validation, limited interpretability, and inadequate ethical and regulatory frameworks. This article reviews the progress, technical approaches, clinical value, and future directions of AI applications in IBD endoscopy, focusing on the transition from subjective scoring to AI-based digital inflammation phenotyping.