Applying joint graph embedding to study Alzheimer's neurodegeneration patterns in volumetric data

Area of Science:

• Neuroimaging
• Network Science
• Biostatistics

Background:

• Volumetric MRI is a potential Alzheimer's Disease (AD) biomarker but lacks specificity.
• Analyzing whole-brain spatial patterns of neurodegeneration can improve specificity.

Approach:

• Extended a graph embedding algorithm to analyze morphometric connectivity from structural MRI volume changes over years.
• Modeled data using multiple random eigengraphs and implemented a modified multigraph embedding algorithm.
• Developed a novel statistical testing procedure for group differences, controlling for confounders and identifying significant structures using permutation testing.

Key Points:

• The enhanced algorithm provides meaningful finite-sample results and estimates maximum likelihood edge probabilities.
• Identified networks dominated by known AD-associated structures, validating the framework's promise.
• Discovered novel network-structure tuples not detectable by traditional methods.

Conclusions:

• The novel network-based approach enhances the study of neurodegeneration patterns in Alzheimer's Disease.
• This method offers improved specificity and potential for identifying new AD biomarkers.
• The framework provides a promising tool for advancing Alzheimer's Disease research.