Individual Differences in Cognitive Performance Are Better Predicted by Global Rather Than Localized BOLD Activity...

Area of Science:

• Neuroimaging
• Cognitive Neuroscience
• Biomarker Development

Background:

• Neuroimaging research struggles to find reliable biomarkers for cognitive processes and clinical outcomes.
• Current methods often identify localized brain regions that explain little phenotypic variation, limiting translational utility.
• Behavioral traits may be influenced by globally distributed neuroimaging variations not captured by traditional thresholded analyses.

Purpose of the Study:

• To develop a novel multivariate prediction method for behavioral prediction using neuroimaging data.
• To aggregate numerous small effects across the cortex into a summary score.
• To improve the prediction of phenotypic variation compared to existing methods.

Main Methods:

• Developed the Bayesian polyvertex score, a multivariate prediction method.
• Transformed unthresholded statistical parametric maps into a summary score.
• Assumed a globally distributed effect size pattern and operated on mass univariate summary statistics.

Main Results:

• The Bayesian polyvertex score achieved higher out-of-sample variance explained than mass univariate and popular multivariate methods.
• The method preserves the interpretability of a generative model.
• Demonstrated improved predictive power for behavioral outcomes.

Conclusions:

• Complex behaviors may be rooted in the global patterning of neuroimaging effect sizes, analogous to polygenicity in genetics.
• Future research should consider global patterns rather than solely localized brain regions.
• This approach offers a more robust method for developing neuroimaging biomarkers.