Modeling Alzheimer's disease: Bayesian copula graphical model from demographic, cognitive, and neuroimaging data

Area of Science:

• Neuroscience
• Biostatistics
• Medical Imaging

Background:

• Early Alzheimer's disease (AD) detection necessitates understanding complex feature relationships.
• Conditional independencies and partial correlations are key for identifying confounding and mediating variables in AD.
• Bayesian Gaussian Copula Graphical Models (GCGMs) offer a novel approach for analyzing AD-related data.

Purpose of the Study:

• To estimate conditional dependencies and partial correlations in AD using Bayesian GCGMs.
• To leverage GCGMs' ability to handle diverse data types (binary, discrete, continuous) and quantify uncertainty.
• To apply GCGMs to AD research for the first time.

Main Methods:

• Designed a GCGM to analyze relationships between brain gray matter volume, glucose uptake, amyloid-beta levels, demographics, and cognitive scores.
• Applied the GCGM to a cohort of healthy and cognitively impaired participants across AD stages.
• Utilized Bayesian inference for robust estimation of dependencies and correlations.

Main Results:

• Identified three indirect aging pathways affecting cognition: hippocampal volume loss, posterior cingulate cortex (PCC) volume loss, and amyloid-beta accumulation.
• Found a positive partial correlation between being female and cognition, dampened by lower hippocampal/PCC volume, increased amyloid-beta, and less education.
• Observed limited direct relations for glucose uptake but confirmed hippocampus and PCC volumes are linked to cognition.

Conclusions:

• Bayesian GCGMs provide valuable insights into Alzheimer's disease pathogenesis.
• The study highlights specific neuroanatomical and biological factors influencing cognitive decline in AD.
• Understanding these complex pathways can inform early detection and intervention strategies for AD.