Artificial intelligence-enabled advanced endoscopic imaging to assess deep healing in inflammatory bowel disease.

Abstract

Endoscopic remission is the primary long-term therapeutic goal in inflammatory bowel disease (IBD). The assessment of this therapeutic target typically relies on white light endoscopy (WLE) combined with histological sampling. Nonetheless, distinguishing between endoscopic mild, patchy inflammation and quiescent disease can be challenging, and discrepancies have been observed between endoscopic and histological disease activity, mainly when using WLE. Recent advances in endoscopic technologies are gradually transforming clinical practice. Dye-based chromoendoscopy and virtual chromoendoscopy are currently available in the endoscopist armamentarium, enhancing the assessment of mucosal architecture and vascular patterns, improving the visualisation of patchy inflammation and helping detect subtle dysplastic colonic lesions. Moreover, novel advanced tools, including probe-based confocal laser endomicroscopy and endocytoscopy, offer the remarkable ability to investigate the deep aspect of the gastrointestinal tract in real time, including the structure and function of the intestinal barrier and inflammatory-related alterations. Thus, these techniques can bridge the gap between endoscopy and histology, enabling the integration of novel treat-to-target strategies associated with more favourable outcomes. Artificial intelligence (AI) represents a further step forward in overcoming the limitations associated with endoscopy, including subjectivity and the requirement for expertise. Their implementation in clinical practice may enable standardised, accurate and rapid disease assessment. Moreover, AI can aid in accurately predicting responses to therapy and disease outcomes by stratifying patients' risks, thereby advancing us towards patient-centred personalised medicine. This narrative review summarises the available advanced endoscopic technologies and their integration with AI to assess IBD activity, define promising therapeutic targets and predict long-term outcomes.