Discordance Between Spatial and Population Correlations From Human Brain Imaging Data

Area of Science:

• Neuroscience
• Brain Imaging
• Biostatistics

Background:

• Population-level brain maps are commonly used to correlate receptor binding potential (BP) and function (e.g., cerebral blood flow, CBF).
• These correlations are typically estimated across regions using data from independent cohorts, raising interpretational concerns.

Purpose of the Study:

• To investigate the validity of across-region correlations in multimodal brain imaging.
• To highlight potential biases and interpretational issues in population-level brain map analyses.

Main Methods:

• Utilized multimodal neuroimaging ([11C]Cimbi-36 PET and MRI) in 24 healthy participants to measure regional BP and CBF.
• Estimated correlations using four methods: across regions from single/separate cohorts, across participants within regions, and within a linear mixed effects model.
• Conducted simulation studies to assess the impact of regional differences on correlation estimates.

Main Results:

• Significant positive correlations were found across regions (ρ̂1.1 = 0.672, p = 0.0023; ρ̂1.2 = 0.659, p = 0.0030).
• Region-specific correlations across participants were substantially lower and non-significant (ρ̂2 mean = 0.140; ρ̂3 = 0.138, p = 0.26).
• Simulations demonstrated that regional variations in BP/CBF can bias across-region correlations and inflate Type-1 error rates.

Conclusions:

• Across-region correlations in population-level brain maps may not accurately reflect biological relationships at the participant level.
• Interpreting across-region correlations as evidence for a biological relation is problematic due to potential biases.
• Further research is needed to develop validated methods for analyzing population-level brain imaging data and mitigating biases.