AI in Endoscopy: ESGE Curriculum for Safe Use- Endoscopy Feb. 26

Introduction

Artificial intelligence is rapidly entering routine GI endoscopy—particularly for polyp detection, lesion characterization, and quality assurance. While early studies show improved detection and efficiency, AI also introduces new risks: inappropriate trust, cognitive bias, deskilling, and uncertainty about how and when AI should be used in clinical decision-making.

Recognizing this gap, European Society of Gastrointestinal Endoscopy (ESGE) has issued a formal Position Statement defining a structured curriculum for the safe, effective, and responsible use of AI in endoscopy.

This document is not about which AI to buy—it is about how clinicians should be trained to use AI properly.

⸻

The core problem

AI tools are being deployed faster than training standards can keep up.

Key unanswered questions in daily practice include:

• Who should be allowed to use AI in endoscopy?
• What level of endoscopic skill is required before using AI?
• How do we prevent over-reliance on AI?
• How do we monitor whether AI improves—or harms—real-world performance?

Until now, no formal competency framework existed.

⸻

What ESGE proposes: a 3-phase curriculum

1️⃣ Before adoption (Preadoption phase)

• AI is not a shortcut for poor endoscopy

Endoscopists must first be competent in standard endoscopic skills (scope handling, lesion visualization, interpretation).

• AI literacy is mandatory

Clinicians should understand:

• what AI can and cannot do,
• how algorithms are trained,
• where bias and errors may occur.

👉 Message: AI supports good endoscopists—it does not replace fundamentals.

⸻

2️⃣ Training phase

• Hands-on training with approved AI systems is essential.
• Education must go beyond buttons and alerts to include:
• recognition of automation bias (blind trust in AI),
• algorithm aversion (rejecting AI after seeing errors),
• anchoring and conservatism bias.

👉 Message: Human–AI interaction is the new technical skill.

⸻

3️⃣ Independent use & quality assurance

• AI must never be used in isolation for clinical decisions.
• Key quality indicators (e.g. ADR in colonoscopy) must be monitored:
• before AI adoption,
• during implementation, and
• after routine use.
• If performance worsens or unintended effects appear, AI de-implementation should be considered.

👉 Message: AI use must be auditable, reversible, and accountable.

⸻

What this means for clinicians

• AI is an assistant, not an authority.
• Final responsibility always remains with the endoscopist.
• Safe AI use requires:
• baseline endoscopic competence,
• structured education,
• awareness of cognitive traps,
• continuous performance monitoring.

This position statement reframes AI from a device issue to a professional competency issue.

⸻

Bottom-line takeaway for GastroAGI

ESGE makes it clear: the success of AI in endoscopy depends less on algorithms and more on how clinicians are trained to use them.

AI can improve quality—but only when embedded within a structured curriculum that prioritizes skills, judgment, and accountability.