HAI Book 2025 - Flipbook - Page 386
Johns, Emily
92
Exploring the predictive power of graph-theoretic metrics for tau
pathology in cognitively unimpaired older adults
Emily Johns1,2, Raina Vin1,2,3, Hamid Abuwarda1,2, Carolyn Fredericks1,2
1
Department of Neurology, Yale University School of Medicine, New Haven, CT, US
Interdepartmental Neuroscience Program, Yale University School of Medicine, New Haven, CT, US
3
Department of Psychiatry, Yale University School of Medicine, New Haven, CT, US
2
Background: Tau pathology is a hallmark feature of Alzheimer's disease (AD), yet its impact on brain network
organization in cognitively unimpaired older adults remains unclear. Resting-state functional MRI (rs-fMRI) allows
for assessing brain connectivity through graph-theoretic metrics, quantifying network structure and function.
This study evaluated the predictive power of these metrics and moderating factors in key networks affected in AD
such as the default mode network (DMN) and salience network (SN) for tau pathology in cognitively unimpaired
amyloid-positive and amyloid-negative older adults.
Methods: Functional connectivity matrices were generated from Pearson correlations between regions of
interest defined by the Brainnetome Atlas in 380 participants in the A4 study. Graph-theoretic metrics4local
efficiency, degree, betweenness centrality, and closeness centrality4were calculated for each network.
Regression models assessed the relationship of these metrics for regional tau SUVRs. Additionally, we explored
the moderating effects of age, sex, education, amyloid PET burden, ethnicity, and APOE genotype on these
relationships.
Results: In the DMN, higher local efficiency in the hippocampus, cingulate, precuneus, superior frontal,
parahippocampal, and temporal gyri was consistently associated with lower tau PET signal in Braak stage 5
(adjusted-pvalue=0.047). In the SN, higher betweenness centrality in the medial amygdala predicted early-stage
Braak pathology (Braak stage 1/2, p
