Very large fMRI study using the IMAGEN database: sensitivity-specificity and population effect modeling in relation...

Area of Science:

• Neuroimaging
• Cognitive Neuroscience
• Statistical Analysis

Background:

• Classical Random Effect (RFX) group statistics in fMRI are commonly used for large cohort analysis.
• Analyzing large cohorts (n=1326) reveals widespread statistical significance not always aligned with practical significance in brain activation patterns.
• The need for improved methods to refine group-level fMRI analysis is evident.

Purpose of the Study:

• To investigate the utility of anatomical information in classical fMRI Random Effect (RFX) group statistics for large cohorts.
• To compare a novel matter-weighted Gaussian mixture model with the standard single-Gaussian model for analyzing BOLD contrasts.
• To assess the impact of cohort size on group effect t-values and model generalization.

Main Methods:

• Utilized data from 1326 subjects in the IMAGEN study for fMRI analysis.
• Compared a matter-weighted mixture of Gaussians model (incorporating tissue probability) against a single-Gaussian model.
• Employed a 10-fold cross-validation scheme with per-voxel parameter estimation on one subgroup and likelihood computation on a separate subgroup.

Main Results:

• Observed a broad activation pattern with increasing cohort size, highlighting the distinction between statistical and practical significance.
• Demonstrated that incorporating anatomical (matter) information consistently enhances the quantitative analysis of BOLD responses.
• Found improvements particularly in brain areas where inter-subject registration is difficult.

Conclusions:

• Injecting tissue-probability information into fMRI group analysis offers significant improvements over standard methods.
• The matter-weighted Gaussian mixture model provides a more accurate quantitative analysis of BOLD responses.
• This approach is especially beneficial for analyzing complex brain regions requiring precise anatomical alignment.