Study-level dataset for spatial artificial intelligence in gastric endoscopy

This dataset contains the study-level data underlying a task-specific systematic review of spatial artificial intelligence in gastric endoscopy for gastric premalignant lesions and early gastric cancer. It includes 39 unique study families and records gastric target, endoscopic modality, spatial task, reference annotation, pathology–spatial linkage, unit of analysis, validation setting, patient-level data separation, external and multicenter validation, prospective and video evaluation, live spatial evaluation, reader comparison, procedure-altering spatial use, margin or resection-planning evaluation, code and trained-model availability, failure analysis, and representative spatial performance as reported by the original studies. Companion reports from the same underlying study were linked and represented as a single study family. Spatial performance measures were retained in their original form and were not converted or quantitatively pooled.