Modeling Site and Scanner Effects in Voxel-Based Morphometry: A Study of BMI in Healthy Older Adults
Poster No:
1611
Submission Type:
Abstract Submission
Authors:
Emma Gleave1, Paul Thompson1, Priya Rajagopalan2
Institutions:
1University of Southern California, Los Angeles, CA, 2Department of Radiology, Los Angeles General Medical Center,, Los Angeles, CA
First Author:
Co-Author(s):
Introduction:
Elevated body mass index (BMI) in adults has been consistently associated with lower gray matter volumes (GMV) in frontal, temporal and cerebellar regions (García-García et al., 2019). Several neuroimaging studies have used voxel-based morphometry (VBM; Ashburner & Friston, 2000) to investigate volumetric brain differences associated with BMI. To improve statistical power to detect these effects, multisite initiatives–such as ENIGMA and COINSTAC–have pooled international multisite data via decentralized analysis, or meta-analysis of individual site summary statistics, to achieve more reliable results based on large and diverse samples (Turner et al., 2022; Bayer et al., 2022). Here, we compared several statistical models for site and scanner effects to assess how they impacted the results of a linear mixed model evaluating VBM gray matter volumetric differences associated with participant BMI. We compared these statistical modeling approaches across three subsets of healthy adult participants from the ADNI, NACC and OASIS-3 neuroimaging datasets, as well as combining them in a multisite mega-analysis.
Methods:
Data from 1,866 participants (61.5% female; 70.1 (10.1SD)) years old) were acquired across multiple centers and scanners, from 84 independent scanner sites with 37 unique scanning protocols. Scanner manufacturer, model, and field strength, were combined to create individual protocol codes for our analysis. The Computational Anatomy Toolbox (CAT12; Gaser et al., 2023) was used to perform a large-scale voxel-wise segmentation of 3D T1-weighted brain MRI data smoothed with a 6-mm kernel. The statistical relationship between voxel-wise brain gray matter segmented volume (GMV) and BMI was assessed using a linear mixed effects model, run both across the combined sample of datasets and within each individual dataset, adjusting for age, sex, age*sex interaction, total intracranial volume, and the random effects of the scan site (modeled as a confound). Four variations of the random effect were modeled separately: "None" or no random effect, Site, Protocol, and Protocol+Site, relative to the voxel-wise gray matter volumes from ADNI, NACC, OASIS-3, or the combined dataset mega-analysis.
Results:
After correcting for multiple comparisons by controlling the false discovery rate at 5%, the results of the voxel-wise brain analyses demonstrated patterns of significantly lower gray matter volumes (GMV) associated with BMI. This pattern held true for regressions of each individual dataset to varying degrees, but was most apparent when all datasets were combined in the mega-analysis. P-value significance maps for each cohort analysis using the Protocol+Site random effect are shown in Figure 1.
The resulting p-values from each within-dataset and between-dataset regression are summarized in Figure 2. Cumulative plots of p-values for the models with differing random effects, mapped by dataset, demonstrate the advantage of combining datasets for stronger associations; the p-values from the combined mega-analysis (N=1,866) show stronger significance, in the Q-Q plot, relative to effects mapped in each individual dataset (ADNI: N=702; NACC: N= 794; OASIS3: N=370). P-value plots of random effects within each dataset show that the inclusion of no random effect ("None") acts as a baseline, but Site, Protocol, and Protocol+Site each account for the site-level variance in the multisite analysis.
The resulting p-values from each within-dataset and between-dataset regression are summarized in Figure 2. Cumulative plots of p-values for the models with differing random effects, mapped by dataset, demonstrate the advantage of combining datasets for stronger associations; the p-values from the combined mega-analysis (N=1,866) show stronger significance, in the Q-Q plot, relative to effects mapped in each individual dataset (ADNI: N=702; NACC: N= 794; OASIS3: N=370). P-value plots of random effects within each dataset show that the inclusion of no random effect ("None") acts as a baseline, but Site, Protocol, and Protocol+Site each account for the site-level variance in the multisite analysis.
Conclusions:
Overall, the p-value significance maps support previous findings that lower gray matter volumes are associated with higher BMI in older adults. Higher power and more robust results were achieved through larger sample sizes, supporting the utility of mega-analyses and pooling multi-site data for neuroimaging studies. Additionally, the differences in random effect modeling further highlight the importance of considering appropriate design to account for additional variance in multisite studies, accounting for scan center and scanning parameters.
Lifespan Development:
Aging
Modeling and Analysis Methods:
Multivariate Approaches 1
Novel Imaging Acquisition Methods:
Anatomical MRI 2
Physiology, Metabolism and Neurotransmission:
Physiology, Metabolism and Neurotransmission Other
Keywords:
ADULTS
Aging
Modeling
MRI
Multivariate
NORMAL HUMAN
Other - VBM
1|2Indicates the priority used for review
Abstract Information
By submitting your proposal, you grant permission for the Organization for Human Brain Mapping (OHBM) to distribute your work in any format, including video, audio print and electronic text through OHBM OnDemand, social media channels, the OHBM website, or other electronic publications and media.
The Open Science Special Interest Group (OSSIG) is introducing a reproducibility challenge for OHBM 2025. This new initiative aims to enhance the reproducibility of scientific results and foster collaborations between labs. Teams will consist of a “source” party and a “reproducing” party, and will be evaluated on the success of their replication, the openness of the source work, and additional deliverables. Click here for more information. Propose your OHBM abstract(s) as source work for future OHBM meetings by selecting one of the following options:
Please indicate below if your study was a "resting state" or "task-activation” study.
Healthy subjects only or patients (note that patient studies may also involve healthy subjects):
Was this research conducted in the United States?
Are you Internal Review Board (IRB) certified? Please note: Failure to have IRB, if applicable will lead to automatic rejection of abstract.
Were any human subjects research approved by the relevant Institutional Review Board or ethics panel? NOTE: Any human subjects studies without IRB approval will be automatically rejected.
Were any animal research approved by the relevant IACUC or other animal research panel? NOTE: Any animal studies without IACUC approval will be automatically rejected.
Please indicate which methods were used in your research:
For human MRI, what field strength scanner do you use?
Which processing packages did you use for your study?
Provide references using APA citation style.
Bayer, J.M.M., et al. (2022). Site effects how-to and when: An overview of retrospective techniques to accommodate site effects in multi-site neuroimaging analyses. Frontiers in neurology, 13, 923988.
García-García, I., et al. (2019). Neuroanatomical differences in obesity: meta-analytic findings and their validation in an independent dataset. International Journal of Obesity, 43(5), 943-951.
Gaser, C., et al., (2023). CAT–A computational anatomy toolbox for the analysis of structural MRI data. biorxiv, 2022-06.
Turner, J.A., et al. (2022). ENIGMA + COINSTAC: Improving Findability, Accessibility, Interoperability, and Re-usability. Neuroinformatics, 20(1), 261-275.