Gastric intestinal metaplasia: Management and surveillance strategies

Table 1 Comparative summary of diagnostic technologies for gastric intestinal metaplasia

Ref.	Design	Population	Main findings	Strengths	Critique
Yao et al[54]	Multicenter randomized controlled trial (training evaluation)	Asian and Western centers	E-learning improved detection of early GIM features (light-blue crest, white opaque substance)	International applicability; structured training intervention	No measurement of actual cancer prevention outcomes; inter-center variability possible
Yan et al[45]	AI model development study	Asian single-center cohort	CNN achieved area under the curves approximately 0.93 for GIM detection	High diagnostic accuracy; real-time feasibility	Model trained on homogeneous data; no external Western validation
Iwaya et al[56]	AI histopathological analysis study	Asian biopsy samples	Deep learning identified IM with > 95% sensitivity	Potential to automate pathology workload; reproducible	Retrospective; lack of prospective clinical deployment; reliance on digitized slides limits real-world application
Ligato et al[55]	AI corpus-focused endoscopic model	Western European cohort	CNN detected corpus IM with area under the curves approximately 0.89	Addresses Western underrepresentation; corpus-specific model	Small sample size; limited diversity across European centers; early-stage validation only

GIM: Gastric intestinal metaplasia; AI: Artificial intelligence; CNN: Convolutional neural network; IM: Intestinal metaplasia.