A generalized estimating equations approach for resting-state functional MRI group analysis.

Area of Science:

• Neuroscience
• Biostatistics
• Medical Imaging

Background:

• Alzheimer's disease (AD) is characterized by altered brain connectivity.
• Functional Magnetic Resonance Imaging (fMRI) is crucial for studying brain activity and connectivity.
• Existing methods for comparing group connectivity, like Fisher-z transformation, have limitations.

Purpose of the Study:

• To evaluate inter-regional correlations at resting-state in Alzheimer's disease (AD) patients and healthy controls.
• To propose and compare novel statistical models, specifically Generalized Estimating Equation (GEE) transition and marginal models, for analyzing within-subject correlations.
• To assess the efficacy of these GEE models against the standard Fisher-z pooling approach for group comparisons.

Main Methods:

• Utilized resting-state fMRI data from AD subjects and healthy controls.
• Developed and applied GEE transition and marginal models to capture within-subject correlations.
• Calculated residuals from GEE models for inter-regional correlation analysis.
• Compared GEE methods with the standard Fisher-z transformation pooling approach.
• Conducted simulation studies to evaluate statistical properties of the methods.

Main Results:

• Demonstrated the application of GEE models and Fisher-z pooling in an AD connectivity study.
• Simulation studies indicated that the GEE transition model may possess superior statistical properties.
• GEE models provide a robust framework for analyzing complex within-subject correlations in neuroimaging.

Conclusions:

• GEE models offer a promising alternative for analyzing resting-state brain connectivity in group studies, particularly in neurodegenerative diseases like AD.
• The GEE transition model warrants further investigation for its enhanced statistical performance.
• Accurate statistical modeling is essential for understanding group differences in brain connectivity patterns.