[Comment] A decade of artificial intelligence in endoscopy: from proof of concept to clinical reality

Abstract

A decade ago, artificial intelligence (AI) in gastroenterology remained largely theoretical. Convolutional neural networks had just begun to show promise in image recognition tasks, but clinical applications were limited to small proof-of-concept studies. The prospect of real-time polyp detection during colonoscopy—let alone autonomous diagnosis—seemed distant. Today, AI systems assist endoscopists in routine practice, regulatory agencies have approved multiple devices, and large language models (LLMs) are being explored across clinical workflows.