TSUNG-HSIEN CHIANGHsu, Yen-NingYen-NingHsuChen, Min-HanMin-HanChenChen, Yi-RuYi-RuChenCheng, Hsiu-ChiHsiu-ChiChengChen, Mei-JinMei-JinChenLee, Fu-JenFu-JenLeeChang, Chi-YangChi-YangChangChang, Chun-ChaoChun-ChaoChangBair, Ming-JongMing-JongBairJYH-MING LIOUChen, Chiuan-JungChiuan-JungChenChen, Yen-ChungYen-ChungChenChiang, HungHungChiangShun, Chia-TungChia-TungShunLiu, Jui-HsuanJui-HsuanLiuLin, Jaw-TownJaw-TownLinGuo, Ruey-ShanRuey-ShanGuoHAN-MO CHIUMING-SHIANG WUYu, Jiun-YuJiun-YuYuLee, Yi-ChiaYi-ChiaLeeChen, Chu-SongChu-SongChen2026-01-222026-01-222025-11-26https://www.scopus.com/pages/publications/105023292985?inwardhttps://scholars.lib.ntu.edu.tw/handle/123456789/735530Diagnosing infection and premalignant gastric conditions typically requires C urea breath testing or histological assessment, which are often unavailable in remote areas. A rural-to-center artificial intelligence (AI) model was developed and implemented to automatically evaluate upper endoscopy images from routine clinical practice.Endoscopic images were collected from a rural hospital on Matsu Islands and a tertiary center across Taiwan Strait. During model development (2020-2022), AI algorithms were trained, validated, and tested to exclude low-quality and non-gastric images, segment gastric regions, and enhance mucosal features for detecting infection and premalignant conditions. During model implementation (2023-2024), endoscopic images from a rural hospital were transmitted to the medical center for AI analyses, with results promptly returned.In the development phase, diagnostic accuracies were 92.8% (95%CI 88.9%-96.6%) for , 88.6% (95%CI 87.2%-90.0%) for atrophic gastritis, and 88.0% (95%CI 86.5%-89.5%) for intestinal metaplasia. In the implementation phase, 3518 rural residents underwent C urea breath testing or pepsinogen testing; 421 with positive results underwent endoscopy. No significant differences were observed between AI-predicted and clinically observed prevalence: (13.9% vs. 12.9%; = 0.55), atrophic gastritis (15.7% vs. 11.9%; = 0.34), and intestinal metaplasia (27.6% vs. 22.4%; = 0.32). Implementation-phase diagnostic accuracies were 91.3% (95%CI 88.0%-94.6%), 79.9% (95%CI 72.1%-86.3%), and 63.4% (95%CI 54.7%-71.6%), respectively.AI enabled physicians in resource-limited settings to rapidly assess gastric health using routinely captured endoscopic images, bridging gaps in access and expertise.enjournal article10.1055/a-2721-655241082919