Scan-specific ICV division reversed 8.59% of yearly hippocampal change pair orders in 295 selected DLBS 1.3.0 adults at waves 1 and 2

We tested whether division by estimated intracranial volume (ICV) at each scan preserves the order of yearly hippocampal log change between people. We used OpenNeuro DLBS 1.3.0 tables from separate FreeSurfer 5.3 scans. A fixed denominator was the reference. In 295 selected adults at waves 1 and 2, the scan-specific denominator reversed 3,726 of 43,365 pairs (8.59%). The approximate 95% interval from 10,000 whole-person bootstrap samples was 6.93% to 10.55%. Change signs differed for 20 people. The two most negative groups of 59 shared 49 people. Six planned checks gave 5.24% to 10.40% reversed pairs. The later-wave check used the same people and is not an independent replication. The package gives a reusable score audit tool and the formula for the correction exponent at each pair crossing. It preserves the source files, fixed commit and hashes, selection ledger, all primary pairs, bootstrap counts and draws, code, figures, and instructions. Independent source and scalar checks passed. A fresh download and repeat matched all 25 result files byte for byte. This audit shows that the denominator changed individual comparisons in these measured records. It does not show which calculation is closer to true anatomical change.

intracranial volume · hippocampal volume · longitudinal MRI · rank reversal · normalization · bootstrap · Dallas Lifespan Brain Study