Two computer programs calculate the same 126 bootstrap interval limits for MR-ART v1.0.2.

Two independently written programs were compared on the stored MR-ART v1.0.2 MRI quality metrics and expert grades. They used the same 148 participants, 10,000 participant-bootstrap draws, definitions, NumPy version, and seed. The comparison covered seven endpoints for three metrics, raw scores and scores centered by each participant’s standard scan, and centered-minus-raw changes. All 42 AUC records, 21 changes, 63 point estimates, 126 interval limits, and 282 reported integer values agreed exactly after numeric parsing (maximum absolute difference 0.0; tolerance 1e-12); no AUC draw was undefined. For the equal mean of HM1 and HM2, centering changed AUC by -0.04695 for CJV (95% interval -0.08714 to -0.00939), +0.11057 for EFC (0.01822 to 0.20520), and -0.05861 for total SNR (-0.12229 to 0.00233). Synthetic and direct-pair checks also agreed as reported. This establishes numerical agreement for this fixed data and method scope, not independent verification, population interval coverage, or generalization across scanners or raters. The source uses one scanner, specified motion, and one expert grade per scan; participant independence is assumed.

MR-ART · MRI quality · bootstrap intervals · AUC · numerical reproduction · image quality metrics