Data and code supporting "Systematic review finds keratoconus AI unproven in fellow eyes with normal topography and tomography"

This deposit contains the study roster, adjudicated evidence tables, index-provenance records, partial-identification analyses, target-substitution experiments, cross-device and demographic analyses, verification-subsample inputs, search materials, registration deviations, PRISMA materials, and Supplementary Tables S1-S10 for a methodological review of artificial-intelligence keratoconus screening covering 573 studies. Analysis code is included for the corpus-level, partial-identification, target-substitution, cross-device, and verification results; the verification upper-bound analysis, the natural target-swap analysis, and the PRISMA flow diagram run from archived files alone. All analysed data derive from published literature and from two public datasets, the CASIA cohort (Harvard Dataverse, CC0) and the Almeria Pentacam cohort (Mendeley Data, CC BY 4.0). No patient-level data from the authors' institution are included. Each numbered directory is distributed as a gzip-compressed tar archive; extract them into one directory to restore the documented layout.