Artificial intelligence in endoscopy: navigating risk, responsibility and ethical challenges

Abstract

Artificial intelligence (AI) technologies, particularly computer-assisted detection (CADe) and computer-assisted diagnosis (CADx), are increasingly being introduced into routine gastrointestinal endoscopy, especially in colonoscopy. CADe systems assist in real-time polyp detection, while CADx offers in-vivo optical characterisation to guide resection and surveillance decisions. Robust evidence supports CADe’s ability to improve adenoma detection rate, a critical quality metric linked to reduced postcolonoscopy colorectal cancer. However, these clinical gains must be weighed against recognised limitations, including false positives and risks of operator over-reliance. While CADx holds theoretical appeal, recent data have questioned its incremental value over optical diagnosis done by experienced endoscopists, highlighting the ongoing challenges in human–AI interaction and system generalisability. This review synthesises the latest evidence and examines the ethical and practical implications of AI integration in endoscopic practice. We focus on two emerging domains of responsibility: forward-looking responsibility—encompassing clinicians’ roles in understanding, applying and communicating AI use—and outcome responsibility, which considers how accountability is shared across clinicians, developers and institutions when adverse events occur. As these technologies continue to evolve, their successful implementation will depend on clear clinical guidance, robust training programmes and thoughtful governance. CADe and CADx not only enhance detection and diagnostic consistency but also require a re-evaluation of endoscopists’ professional responsibilities in a technologically mediated environment. Supporting clinicians in using these systems safely and ethically will be essential for ensuring they contribute meaningfully to patient care.

Ähnliche Arbeiten