OSIRIS_validation_dataset

Dataset summary • Source: single-center refractive registry, 2,533 eyes. Population: myopic and myopic astigmatic eyes undergoing keratorefractive lenticule extraction (preoperative manifest spherical equivalent -12.75 to -0.60 D); hyperopic, previously operated, and pseudophakic eyes are not represented. • Selected raw OSIRIS scan rows: 23,718, forming 10,416 unique within-session sessions (pa- tient, laterality, date, time), of which 9,824 contain at least two scans. • Cross-device agreement pairs (OSIRIS whole-eye versus MS-39 total-corneal, and OSIRIS 6- mm versus 4-mm): 26,170 exported pairs. • Metrics: HOA RMS, coma RMS, trefoil RMS, spherical aberration, selected Zernike coeffi- cients, objective sphere/cylinder/axis/spherical-equivalent, Strehl ratio, MASI, and maximum wavefront diameter, at 4-mm and 6-mm apertures where applicable. • Privacy: no names, dates of birth, medical record numbers, or calendar dates in any row-level export. Patient and eye identifiers are opaque hashed tokens. Visit timing is expressed as a relative label and days-from-surgery integer rather than calendar dates. Data records The package contains the following. Relational database.- OSIRIS_validation_study.sqlite: consolidates all analysis inputs and result tables in a single queryable database. Analysis inputs (anonymized, scan- and pair-level).- selected_repeat_scan_data_anonymized.csv: scan-level records for selected repeated OSIRIS scans (37 fields), including patient and eye to- kens, laterality, visit label, days from surgery, session identifier, maximum wavefront diameter, pupil-support flags, 4-mm and 6-mm wavefront metrics, objective refraction, and validity and zero-code flags. - agreement_pair_data_anonymized.csv: pair-level records for agreement analysis (comparison, metric, aperture, stratum, patient and eye tokens, visit label, and the two paired values). Pre-computed result tables.- repeatability_summary.csv: full within-session repeatabil- ity by metric, aperture, cohort, visit stratum, and pupil-support stratum (217 rows). - pri- mary_repeatability_results.csv: compact paper-ready primary repeatability results with confidence intervals. - agreement_summary.csv: Bland-Altman, correlation, and concordance results for OSIRIS versus MS-39, OSIRIS 6-mm versus 4-mm, and OSIRIS objective versus clinical manifest (102 rows). Quality-control tables.- pupil_gating_qc.csv: session-level 6-mm and 4-mm wavefront availabil- ity and zero-coding by maximum-wavefront-diameter band. - pupil_support_effect.csv: cluster- robust comparison of absolute within-session differences below versus at or above a 6-mm maxi- mum wavefront diameter. - clinical_rounding_independence_qc.csv: test of whether clinical re- fraction is a rounded derivative of the objective measurement. - pipeline_reconstruction_qc.csv: reconstruction error between means of linked raw repeats and registry selected-scan aver- ages. - study_flow_qc.csv: aggregate study-flow and zero-code/pupil-availability counts. - data_dictionary.csv: field and file dictionary. - build_summary.json: integrity check and record counts. Figures. - Repeatability Bland-Altman plots for 6-mm HOA and 6-mm spherical aberration; scatter and Bland-Altman plots for OSIRIS versus MS-39 HOA and spherical aberration; study flow; pupil- support gate; astigmatic power vectors; and axis agreement versus cylinder magnitude. Code.- scripts/run_osiris_validation.py: end-to-end pipeline. - scripts/finalize_osiris_validation.py, scripts/bootstrap_primary_repeatability.py, scripts/bootstrap_masi.py: finalization and cluster-bootstrap routines.