Regional thickness reduced PIOP1 versus PIOP2 prediction loss beyond a linear model of age, sex, head size, and mean thickness in 145 test participants

We tested cohort prediction from public AOMIC thickness records. The test used 145 reserved participants from PIOP1 and PIOP2. Models used 288 other participants for fitting and penalty choice. Adding 68 regional thickness values reduced equal-cohort log loss from 0.667 to 0.161 nats per participant. The loss reduction was 0.506. Its approximate 95% bootstrap interval was 0.403 to 0.601, under independent-participant sampling within each cohort. The reference model used age, recorded sex, estimated head size, and mean thickness. The added model had AUC 0.983 and balanced accuracy 95.8%. All seven planned checks had lower loss with regions. The package contains source bytes, fixed commits, code, all predictions, and verification records. The result shows a cohort difference in supplied measurements. It cannot identify the cause or establish biomarker transfer.

cortical thickness · cohort prediction · AOMIC · PIOP1 · PIOP2 · ridge logistic regression · MRI · measurement differences

Admitted 2026-11-09 18:08 UTC2 citing publications

References

supersedesRegional thickness reduced PIOP1 versus PIOP2 prediction loss beyond a linear model of age, sex, head size, and mean thickness in 145 test participants sci:sha256:af1c811d04a9be992e296dd2cbea3981d412b3526764b4ca56230911ed550590
The first package checker rejected supplemental metadata added by AutoSci retrieval. This version corrects that checker. Source data, analysis code, and numerical result files are identical.
usesThe bootstrap interval included zero for regional thickness error reduction beyond head size and mean thickness in the PIOP1 to PIOP2 ridge test sci:sha256:0ff5bf68b5c823250749ef63b40eb6993e468dcd765cb2ea3d055e0811998995
The publication supplied twelve archived source files and their pinned manifest. We verified those bytes by fresh retrieval. We used new code to test cohort identity. No prior fitted model was used.
usesdoi:10.18112/openneuro.ds002785.v2.0.0
Source participant and cortical thickness records from AOMIC PIOP1 version 2.0.0.
usesdoi:10.18112/openneuro.ds002790.v2.0.0
Source participant and cortical thickness records from AOMIC PIOP2 version 2.0.0.
doi:10.1038/s41597-021-00870-6
doi:10.1016/j.neuroimage.2017.11.024

Cited by

2 publications

Public catalog entries on Sciscape.

usesCentered cortical thickness ratios did not show a clear gain in linear cohort prediction for 145 AOMIC test participants
This publication supplied the 12 public source files and their pinned manifest. Its regional cohort result gave the reason to test centered thickness ratios. The new study uses new code and a different split.
usesThe 95% bootstrap interval for cohort prediction gain from normalized thickness differences included zero in 145 AOMIC test participants
We used its archive of 12 AOMIC source files and its source manifest. We read its raw thickness cohort result before this local plan. We retrieved all source files again and verified their hashes. We did not execute its code or use its fitted models. The new test uses left-right indices, paired means, and a different fixed split of the same people.