Misconceptions in the use of the General Linear Model applied to functional MRI: a tutorial for junior neuro-imagers

Area of Science:

• Neuroimaging
• Cognitive Neuroscience
• Biomedical Engineering

Background:

• Functional Magnetic Resonance Imaging (fMRI) relies on the General Linear Model (GLM) for data analysis.
• Misconceptions regarding GLM application can lead to inaccurate fMRI results.
• Understanding GLM nuances is crucial for reliable neuroimaging research.

Purpose of the Study:

• To address common misconceptions in applying the GLM to fMRI data.
• To educate researchers using practical examples and Matlab code.
• To clarify the impact of parameterization, hemodynamic modeling, and signal change computation on fMRI outcomes.

Main Methods:

• Tutorial-based approach using illustrative examples and Matlab code.
• Analysis of controlled block and alternating block designs.
• Examination of periodic vs. random event-related designs.
• Focus on model parameterization, hemodynamic basis functions, and percentage signal change calculation.

Main Results:

• Modeling "baseline" can lead to over-parameterization and affect effect sizes, though not always statistical significance.
• Hemodynamic model choices (e.g., basis functions, derivatives, orthogonalization) influence parameter estimates.
• Proper understanding of these factors is essential for accurate percentage signal change computation.

Conclusions:

• Clarifies GLM application pitfalls in fMRI, particularly concerning parameterization and hemodynamic modeling.
• Provides practical guidance for improving the accuracy of fMRI analyses and group-level inferences.
• Emphasizes the importance of careful model specification for robust neuroimaging findings.