Objective:

To assess whether introducing AI-assisted endoscopy was associated with changes in secondary review modifications and referral decisions in a mobile gastric cancer screening program, including differences by primary endoscopist experience.

Approach:

• Design and setting: Retrospective multicenter before–after study in Gifu Prefecture, Japan, comparing screening EGD examinations from 2021–2022 with those from 2023–2025, after AI implementation.
• Participants and intervention: The cohort included 5,013 examinations: 2,021 before AI and 2,992 during the AI period. Primary interpretation incorporated the CAD EYE real-time alert system; the existing independent secondary review process continued.
• Outcomes and analysis: The primary outcome was modification during secondary review, including new findings, diagnostic reinterpretation, atrophy reassessment, and referral recategorization. Adjusted logistic regression and interaction analyses examined associations with AI implemen…

Interpretation:

The provided context describes the study rationale, methods, and outcomes but does not report results. Secondary-review changes were treated as descriptive process outcomes, not measures of diagnostic accuracy or clinical benefit, because no external reference standard was available.

Limitations:

• The retrospective before–after design may be subject to confounding.
• No external reference standard was available to determine whether review modifications or referral decisions were correct or clinically appropriate.
• The supplied context ends during the statistical analysis section and contains no study results.

Sources:

• Article context supplied in the prompt