Neuroimaging meta regression for coordinate based meta analysis data with a spatial model.

Area of Science:

• Neuroimaging
• Statistical Modeling
• Spatial Statistics

Background:

• Coordinate-based meta-analysis synthesizes neuroimaging study evidence to map brain activation.
• A significant challenge is developing computationally efficient and statistically interpretable models for activation foci locations.
• Existing methods may lack flexibility in modeling spatial activation patterns and covariate influences.

Purpose of the Study:

• To propose a novel generative coordinate-based meta-regression (CBMR) framework.
• To approximate a smooth activation intensity function and model spatial brain activation.
• To investigate the impact of study-level covariates on activation patterns.

Main Methods:

• Utilizing a spline parameterization to capture the spatial structure of brain activation.
• Employing four distinct stochastic models to account for random variations in activation foci.
• Developing a generative framework for coordinate-based meta-regression (CBMR).

Main Results:

• The CBMR framework was applied to 20 meta-analytic datasets.
• Spatial homogeneity tests were performed at the voxel level to validate the model.
• Performance was compared against existing kernel-based and model-based meta-analysis approaches.

Conclusions:

• The proposed CBMR framework provides a statistically interpretable and computationally efficient method for coordinate-based meta-analysis.
• Spline parameterization effectively models the spatial distribution of brain activation.
• The framework successfully incorporates study-level covariates, enhancing meta-analytic insights.