Population-weighted Image-on-scalar Regression Analyses of Large Scale Neuroimaging Data.

Area of Science:

• Neuroscience
• Population Genetics
• Biostatistics

Background:

• Population-based neuroscience research requires accounting for subgroup heterogeneity.
• Integrating survey methodology with neuroimaging is crucial for generalizability.
• The Adolescent Brain Cognitive Development (ABCD) Study provides large-scale neuroimaging data but requires careful handling of subgroup differences.

Purpose of the Study:

• To develop and validate a method for enhancing population generalizability in neuroimaging research.
• To address discrepancies between imaging subsamples and the baseline cohort in large-scale studies.
• To improve the estimation of associations between brain activity and cognitive performance using functional Magnetic Resonance Imaging (fMRI) data.

Main Methods:

• Developed new population weights specific to an imaging subsample.
• Applied image-on-scalar regression models incorporating adjusted weights.
• Validated the approach using synthetic simulations and real fMRI data from the ABCD Study.

Main Results:

• The developed population weighting adjustments effectively captured active brain areas associated with cognitive performance.
• The adjusted weights improved the validity and generalizability of findings from the ABCD Study.
• Standard ABCD base weights were insufficient to address subsample discrepancies.

Conclusions:

• Population weighting adjustments are essential for accurate and generalizable neuroimaging research.
• This methodology enhances the integration of survey science principles into neuroscience.
• The findings support the use of tailored weighting strategies in large-scale developmental neuroimaging studies.