HAI Book 2025 - Flipbook - Page 333
Laymon, Charles
69
Multi element MRI template method for analysis of Down syndrome
PET
Charles Laymon1, Davneet Minhas1, Jaiden Heffley1, Ally Gilmore1, Ritvik Kondaveeti1,
Alexandra Gogola1, Sarah Royse1, Sai Sravanthi Joshi1, Brian Lopresti1, Weiquan Luo1, Hitesh
Davuluri1, Cristy Matan1, Dana Tudorascu1, Bradley Christian2, Shahid Zaman3, Ances Beau4,
Ann Cohen1, Benjamin Handen1
1
University of Pittsburgh, Pittsburgh, PA, US
University of Wisconsin, Madison, MI, US
3
University of Cambridge, Cambridge, GB
4
Washington University, St. Louis, MO, US
2
Background: A component of the Alzheimer9s Biomarkers Consortium-Down Syndrome (ABC-DS) is tau brain
PET. A common analysis method is tracer quantification through application of a subject-space atlas of regionsof-interest (ROIs) determined by parcellating each subject9s T1-MR scan using FreeSurfer (FS). There are
drawbacks associated with this for Down Syndrome (DS) subjects who exhibit variations in anatomy from FS-atlas
priors and are prone to motion during MR acquisition. We evaluate a multi-element template (MET), DS-specific
approach for improving results.
Method: In the standard analysis, ABC-DS MR scans were processed through the FS (5.3) pipeline. Parcellation
results (atlases) were evaluated by experienced analysts and usually edited to increase conformity with
anatomy. The final atlas was either accepted or rejected.
For the MET analysis twelve MR scans (10-DS subjects, 2 controls) were chosen as MR templates based on the
quality of their final FS-based atlas from the standard analysis. Subject processing consisted of warping all
template MRs to the subject MR using an ANTs-based pipeline and then co-warping each of the 12 associated
atlases resulting in the assignment of 12 regional labels to every voxel. A MET atlas for each subject was produced
by assigning final voxel labels based on the plurality of template labels. Final MET atlases were accepted or
rejected based on unedited results.
For subjects with acceptable final FS and MET atlases, [18F]AV1451 was sampled using ROIs from both methods
and results were compared.
Result: Out of 330 MRs, standard processing resulted in 53 unacceptable atlases whereas the MET processing
resulted in 9. Correlation of [18F]AV1451 SUVR between the two methods was high (Figure 1).
Conclusion: Unedited MET is frequently able to provide parcellations for cases in which edited FreeSurfer fails.
High correlation between AV1451 measures from the two pipelines suggests that the methods can be used
interchangeably.
HAI2025 - 333
