BETA-AMYLOID-RELATED ALTERATIONS OF METABOLIC CONNECTIVITY IN ADNI MCI SUBJECTS: A MULTIVARIATE COVARIANCE ANALYSIS

Abstract

Specific alterations of metabolic connectivity have been shown to be intimately related to β-amyloid plaque burden. While seed-based correlation analysis has been proven to be useful for revealing metabolic connectivity patterns, the relationship between voxelwise covariance and variance terms may remain undetected. We have implemented a new, multivariate approach, called “seed-based covariance analysis”, to identify potential alterations in metabolic connectivity, which may not be revealed by conventional correlation analysis. In this study, we have performed a seed-based covariance analysis using [18F]FDG positron emission tomography (PET) images acquired from Alzheimer's Disease Neuroimaging Initiative (ADNI) subjects with different levels of cortical β-amyloid. [18F]florbetapir PET, [18F]FDG PET, and 3D T1-weighted MR images were obtained from ADNI-GO/-2 study subjects diagnosed with mild cognitive impairment (MCI). PET volumes were registered to a customized MRI template in MNI stereotaxic space, and standardized uptake value ratio (SUVR) images were generated using Biospective's fully-automated PIANO TM image processing software. The amyloid burden for each subject was determined from a composite region-of-interest (ROI) on [18F]florbetapir images, and subjects were categorized into Amyloid-Low (Aβ L) and Amyloid-High (Aβ H) groups. We generated a set of hierarchical likelihood ratio tests to assess between-groups differences in metabolic connectivity patterns arising from (1) alterations in seed-based correlations and (2) alterations in seed-based covariances and variances with stable seed-based correlations. We observed statistically significant differences in metabolic correlations between the Aβ L and Aβ H groups for multiple cortical seeds regions, including the angular gyrus and the inferior temporal gyrus. The seed-based covariance analysis identified connectivity patterns in particular brain regions (e.g. precuneus) that were not detected by classical seed-based correlation analysis. We have introduced a new, multivariate metabolic connectivity analysis technique to examine disruptions of the cortical correlation architecture as a function of β-amyloid burden. The novel approach employed in this study may be generalized to other connectivity measures, such as functional connectivity derived from BOLD fMRI, and may provide unique insights into disease-related alterations of the connectome.