Machine learning in GI endoscopy: practical guidance in how to interpret a novel field

Area of Science:

• Gastroenterology
• Artificial Intelligence
• Medical Imaging

Background:

• The increasing volume of machine learning (ML) research in endoscopy presents challenges for scientific appraisal.
• Clinicians and ML experts often lack the necessary cross-disciplinary knowledge to critically evaluate ML studies in gastroenterology.

Purpose of the Study:

• To establish comprehensive quality requirements for ML research in endoscopy.
• To guide readers and reviewers in assessing the scientific rigor and novelty of ML applications in GI journals.

Main Methods:

• A multidisciplinary, international evaluation of current trends and pitfalls in ML for endoscopy.
• Development of quality criteria across six key themes: terminology, data, algorithm description, experimental setup, results interpretation, and clinical integration.

Main Results:

• Identifies common trends and potential challenges in ML endoscopy research.
• Proposes a framework for critical appraisal of ML studies, ensuring scientific validity.

Conclusions:

• High-quality ML research in endoscopy requires standardized evaluation criteria.
• This review offers a foundational framework to enhance the scientific rigor and clinical relevance of ML applications in gastroenterology.