A new calculation gives the same 155,480 EFC predictions as the MR-ART source exclusion report.

This computational reproduction compares a new exact-rational calculation with the source exclusion results for supplied entropy focus criterion (EFC) values and expert image quality grades in OpenNeuro ds004173, version 1.0.2. All 155,480 prediction rows and 1,120 confusion tables agree, with zero differences in the specified fields. Only the HM2-to-HM1 transfer using subtracted EFC changed predictions under source-participant exclusion: six of 144 exclusions changed one or more predictions, with a maximum of 9 of 134 target predictions (6.716418%). The other three source and score combinations had no changed predictions. The 155,480 rows include repeated scans across calculations and two optimum rules; they are not independent observations. This is agreement for one fixed sample and rule, not a population result. The analysis did not calculate EFC from images or estimate uncertainty intervals. Missing scans, class imbalance, instructed rather than measured motion, and supplied expert grades limit interpretation; the results provide no evidence of biological causality or clinical performance.

MR-ART · MRI · entropy focus criterion · source exclusion · threshold transfer · balanced accuracy · computational reproduction · OpenNeuro