Power calculations for multicenter imaging studies controlled by the false discovery rate.

Area of Science:

• Neuroimaging
• Biostatistics
• Neuroscience

Background:

• Magnetic resonance imaging (MRI) is vital for brain imaging research, enabling exploration of structural and functional neural network changes.
• Multicenter investigations enhance participant recruitment for neuroimaging studies.
• Accurate power calculations are essential for designing effective multisubject brain imaging experiments.

Purpose of the Study:

• To describe image-based power calculations for two-group, cross-sectional neuroimaging designs.
• To determine minimum sample size and minimum effect size based on power thresholds, false discovery rate (FDR), and network size.
• To evaluate the impact of within-center variance on sample size requirements across different brain regions.

Main Methods:

• Developed image-based power calculations for cross-sectional neuroimaging studies.
• Estimated within-center variance using repeat MRI scans from 12 healthy participants across five centers.
• Simulated multicenter studies with varying recruitment proportions and analyzed the effect of FDR and true positive rates.

Main Results:

• Within-center variance in gray matter distribution was consistent across participating centers.
• Detecting effects in subcortical regions requires 3-6 times larger sample sizes than in the neocortex.
• Recruiting across five centers showed minimal sample size penalty compared to a single low-variance center.

Conclusions:

• Image-based power calculations provide a framework for optimizing sample size in neuroimaging research.
• Multicenter studies can efficiently increase recruitment rates with minimal impact on statistical power.
• Approximately 80 participants per group are needed to detect gray matter differences in high-variance regions at 80% power.