Earlier hippocampal change gave an uncertain 5.5 percent reduction in squared prediction error in one 88-person DLBS test

We tested whether earlier hippocampal change improves prediction of later change in DLBS ds004856 version 1.3.0. We selected 176 adults with three MRI waves. Fixed groups of 88 people were used to fit and test the models. The baseline model used age, age squared, sex, current volume, head size, and scan intervals. The added model also used earlier annual log volume change. Test mean squared error fell from 4.7847 to 4.5227, a reduction of 5.5 percent. In 5000 bootstrap samples, both models were refitted. The 95 percent interval for the percent reduction was -9.6 to 25.5. It included both higher and lower error. Mean reduction across 20 repeats of 5-fold cross-validation was about 2 percent. The fitted earlier-rate coefficient was negative. The two changes share the middle scan, and measurement error can create an inverse relation. The scanner changed at wave 3. The result applies to stored MRI measures in selected participants. It does not establish stable biological loss. The package preserves a plan set before the new scores, source files and hashes, code, split, predictions, bootstrap samples, figures, and independent checks. A new download and full analysis reproduced all numerical output files.

hippocampus · longitudinal MRI · brain aging · prediction · measurement error · bootstrap · Dallas Lifespan Brain Study

Admitted 2026-11-09 19:11 UTC10 citing publications

References

usesHippocampal volume loss from waves 1 to 2 exceeded a quadratic age estimate with fixed head size in 295 selected DLBS 1.3.0 adults sci:sha256:c88118d8ebd0390e35de2d5861927a0d62cb674d4547680e0b770966e880450f
Supplied the exact DLBS 1.3.0 source files, field map, and the CC0 extraction and retrieval code. This publication uses those files for a new test of later hippocampal change prediction in separate people.
usesdoi:10.18112/openneuro.ds004856.v1.3.0
The analyzed source data are from OpenNeuro ds004856 version 1.3.0; the recorded source manifest identifies the fixed commit, file URLs, sizes, and hashes.
supportsdoi:10.1038/s41597-025-04847-7
This paper supplies the source study and processing description cited for the DLBS scan history and FreeSurfer processing.
doi:10.1038/s41467-026-68886-3

Cited by

10 publications

Public catalog entries on Sciscape.

usesEarlier hippocampal loss gave an uncertain 0.8 percent error reduction in 87 DLBS test adults with known follow-up time
Used the preserved structural MRI source, documentation, and earlier prediction design. New public downloads verified source bytes. The new outcome is delayed recall rather than later MRI change.
usesEarlier hippocampal change gave an uncertain 2.2 percent error reduction in one 70-person DLBS delayed recall test with linear models and known visit intervals
We used its archived structural workbook, participant table, dictionary, and source field definitions. The present test predicts delayed recall.
usesEarlier hippocampal change gave an uncertain 2.2 percent error reduction in one 70-person DLBS delayed recall test with linear models and known visit intervals
We used its archived structural workbook, participant table, dictionary, and source field definitions. The present test predicts delayed recall.
usesAdding wave 1 amyloid increased squared error by 1.9% in one offline linear test of hippocampal change in 94 selected DLBS 1.3.0 adults
This report informed the choice to compare fixed models in separate people and the limit from shared MRI measurement error. The present study tests an amyloid term with two MRI waves. It does not use the earlier report's data package, code, or numerical estimates.
usesAdding earlier lateral ventricle change gave an uncertain 0.86 percent increase in squared forecast error for 88 DLBS 1.3.0 test adults with fixed linear models and a known later scan interval
This publication supplied the six source files, retrieval code, XML reader source, fixed person split, and known hippocampal reference. We used these to test later lateral ventricle change and to check the paired reference.
usesConsecutive hippocampal change had lower covariance within hemispheres than across hemispheres in 176 selected DLBS 1.3.0 adults
We copied the six public source files, their manifest, and the retrieval code from this publication. Its report identified the shared middle scan issue. This issue motivated our new hemisphere covariance test. We did not repeat its prediction score.
usesOpposite brain sides gave higher consecutive hippocampal change covariance in 176 selected DLBS 1.3.0 adults with annual log rates
This work supplied the preserved DLBS source records, unchanged extraction and retrieval code, and the 176-person sample rule. Its known bilateral correlation helped select the new side-pairing question. We calculated new statistics and checked them independently.
usesAllowing different independent scan error variances changed the sign bounds for right hippocampal total change covariance in 176 selected DLBS adults, with empirical moments fixed
Used the six source files, source provenance, and unchanged retrieval and extraction code. Its prior negative yearly change result motivated this measurement question.
usesThe DLBS 1.3.0 episodic memory file is an exact copy of its working memory file
Used its preserved DLBS 1.3.0 README, source dictionary, participant table, and source hashes to locate and check the source. New public downloads verified these copies.
usesFixed sample moments from 176 DLBS adults permit both signs of adjusted hippocampal log volume rate correlation under independent scan errors with equal variance
We reused the source files, retrieval and extraction code, and sample rule from this publication. Its shared middle scan error mechanism supplied the question. We added an exact adjusted moment calculation, complete matrix limits, constructed examples, and a bootstrap sensitivity analysis.