Optical diagnosis of histopathology- is it implementable in the world of artificial intelligence?

Colorectal cancer (CRC) remains a leading cause of cancer-related mortality in the United States, with colonoscopy serving as the gold standard for both diagnosis and early intervention. While diminutive polyps (<5 mm) constitute most findings, only a small fraction exhibit advanced histological features. Optical diagnosis, which enables real-time classification of polyp histology through new technologies and the support of new strategies to leave low risk polyps in place (diagnose-and-leave) or resect without sending for formal pathology (resect-and-discard) have been studied as a cost-saving and effective strategy for diminutive polyps. There have been advances in imaging, such as narrow band imaging (NBI), but widespread adoption has yet to occur. The integration of artificial intelligence (AI), particularly computer-aided diagnosis (CADx) systems, has emerged as a promising tool to standardize optical diagnosis, reduce interobserver variability, and improve adherence to surveillance guidelines. However, barriers to widespread implementation persist, including concerns about medicolegal liability, financial disincentives, and skepticism of CADx accuracy. The goal of article is to review the current evidence surrounding optical diagnosis, review diagnostic accuracy, and evaluate the challenges of widespread clinical adoption.