Experimental design modulates variance in BOLD activation: The variance design general linear model.

Area of Science:

• Neuroimaging
• Cognitive Neuroscience
• Biostatistics

Background:

• Traditional fMRI analyses focus on mean BOLD signal or functional connectivity.
• Other BOLD signal statistics, like variance, may hold crucial neurobiological information.
• Existing methods lack a formal framework for analyzing BOLD variance effects.

Purpose of the Study:

• Introduce a novel statistical framework, the variance design general linear model (VDGLM), for analyzing BOLD variance in fMRI data.
• Provide a flexible tool applicable to various fMRI study designs.
• Enable simultaneous inference on mean and variance effects while controlling for confounds.

Main Methods:

• Developed the variance design general linear model (VDGLM) to model both mean and variance of BOLD activation.
• Designed the framework for general applicability across diverse fMRI experimental paradigms.
• Incorporated the ability to test for variance effects related to multiple conditions and noise regressors.

Main Results:

• Demonstrated the VDGLM's utility in a working memory task.
• Showed that engagement in a working memory task is associated with decreased whole-brain BOLD variance.
• Validated the framework's capability to detect variance effects distinct from mean effects.

Conclusions:

• The VDGLM offers a powerful new approach for analyzing fMRI data beyond traditional mean-based analyses.
• BOLD variance is a sensitive indicator of cognitive task engagement, decreasing during working memory tasks.
• This framework expands the analytical toolkit for neuroimaging research, potentially uncovering novel biomarkers.