Artificial Intelligence-assisted Endoscopic Diagnosis of Upper Gastrointestinal Precancerous Lesions: A Systematic Review and Meta-analysis

Background and objectives: Upper gastrointestinal precancerous lesions are important targets for cancer screening and prevention.This study evaluated the diagnostic accuracy of artificial intelligence (AI)-assisted endoscopy for upper gastrointestinal precancerous lesions.Methods: PubMed, Embase, MEDLINE, Cochrane Library, and IEEE Xplore were searched from inception to March 31, 2026.Studies reporting sensitivity and specificity, or sufficient data for their calculation, were included.Two reviewers independently extracted data.Pooled sensitivity, specificity, and the area under the receiver operating characteristic curve were estimated using a bivariate random-effects model.Results: Fifteen studies comprising 24 datasets and 19,578 records were included.Pooled sensitivity and specificity were 0.930 (95% confidence interval (CI): 0.886-0.958)and 0.944 (95% CI: 0.906-0.967),respectively; the diagnostic odds ratio was 223.79 (95% CI: 92.24-542.94),and the area under the receiver operating characteristic curve was 0.9661.Substantial heterogeneity was observed (bivariate I² = 73%).Among the variables formally evaluated by meta-regression, imaging modality was associated with heterogeneity (global test P = 0.005); image-enhanced endoscopy had higher sensitivity than white-light endoscopy (0.960 vs. 0.875).No significant publication bias was detected by Egger's test (P = 0.833), Begg's test (P = 0.641), or Deeks' funnel plot asymmetry test (P = 0.053). Conclusions: AI-assisted endoscopy shows high diagnostic accuracy for upper gastrointestinal precancerous lesions.Multicenter prospective studies are needed for validation.