MACS - a new SPM toolbox for model assessment, comparison and selection

Area of Science:

• Cognitive Neuroscience
• Neuroimaging Analysis

Background:

• Functional magnetic resonance imaging (fMRI) data analysis commonly employs general linear models (GLMs).
• Assessing the quality and statistical inference of GLMs in fMRI remains a significant challenge due to a lack of formal measures.

Purpose of the Study:

• Introduce a new toolbox, Model Assessment, Comparison and Selection (MACS), for evaluating GLMs in fMRI data.
• Provide a comprehensive suite of methods for model assessment, comparison, and selection within the SPM software environment.

Main Methods:

• The MACS toolbox integrates classical, information-theoretic, and Bayesian approaches to GLM assessment.
• It incorporates recent advancements in model selection and model averaging techniques for fMRI data analysis.
• The toolbox is built upon the Statistical Parametric Mapping (SPM) software, ensuring ease of use and extensibility.

Main Results:

• The MACS toolbox successfully reproduces model selection and model averaging results from prior fMRI studies.
• It offers a validated and reusable computational resource for GLM assessment in fMRI.

Conclusions:

• The MACS toolbox addresses the limitations of previous fMRI model diagnosis tools, which are often discontinued or not readily available.
• Emphasizing GLM quality assessment with MACS is expected to reduce false-positive rates in cognitive neuroscience.
• Increased adoption of MACS will enhance the reproducibility and replicability of GLM-based fMRI studies.