RESTING STATE NETWORK MODULARITY ALONG THE PRODROMAL LATE ONSET ALZHEIMER’S DISEASE CONTINUUM

Abstract

Alzheimer's disease (AD) results in a disconnection syndrome that can be reliably detected using whole-brain functional connectivity (FC) network analyses. We used multiscale modularity analysis [1] of resting state fMRI (rsfMRI) data to investigate changes in connectivity among resting state networks (RSNs) across four different stages in the AD continuum. Participants included 78 older adults from the Indiana Memory and Aging Study (22-subjective cognitive decline (SCD), 16-MCI, 13-AD, and 29-cognitively normal (CN) controls). All underwent rsfMRI preprocessed using an in-house pipeline. For each participant, FC was expressed as a matrix of Pearson correlations between pairs of time series from 278 gray matter brain regions [2]. To assess community structure across multiple scales, we systematically varied the resolution parameter γ and optimized a modularity metric at each scale. Consensus solutions were aggregated across all values of γ to derive a multi-scale co-assignment matrix. Modules were compared to a standard RSNs partition [3] with subcortical and cerebellar regions added for a total of nine networks. Additionally, the average density of FC within and between RSNs was calculated for each group providing four 9x9 matrices (containing 45 sub-blocks). Significant sub-blocks were further analyzed using linear correlation regression models against average memory, executive function, executive function with language, and general cognition scores. Age, sex, and education were treated as nuisance variables. Retained sub-blocks were determined with the following dual criteria: The block must have achieved p<0.01 for both FC and module cluster co-assignment. Modularity analysis revealed that connections between the frontoparietal attention (FP) and default mode (DMN) RSNs varied systematically with diagnosis (Figure 1). Additionally, linear regression analysis revealed negative correlations between all cognitive domains and average FC between DMN and FP networks.