P2‐112: PATTERNS OF STRUCTURAL COVARIANCE ASSOCIATED WITH ALZHEIMER'S DISEASE: A NOVEL VOXEL‐WISE GRAPH THEORETIC APPROACH

Abstract

Evidence is growing that there is structural co-variance between spatially distributed brain regions, such that neurodegenerative cortical thickness changes in one region are associated with thickness in other functionally connected regions. We hypothesize that brain structural co-variance is present at voxel-wise MRI resolution, and can be evaluated using graph theoretical analysis to identify patterns of spatially distributed volumetric brain abnormality associated with Alzheimer's disease (AD). We included ADNI T1 MRI data from 102 cognitively normal (CN) (mean-age 75.8 years, 62 male/40 female) and 92 AD (mean-age 75.5 years, 55 male/37 female) participants. All data were segmented and normalized to a standard template using VBM toolbox within SPM12. Modulated gray matter maps from all subjects in each group were used to construct group-specific voxel-based correlation matrices. Binarized adjacency matrices were generated, from which AD- and CN-specific node degree maps were constructed. Node degree within a brain region reflects similarities in voxel-wise gray matter volume with voxels in other brain regions. An independent-samples t-test was conducted to compare hippocampal node degree between AD and CN groups. The hippocampus in subjects with AD was comprised of voxels with statistically greater node degree compared to the CN group (p<0.0008) (Figure 1), suggesting greater similarity between voxels in hippocampus and voxels in other brain regions among AD participants compared to controls. VBM analyses demonstrated that these voxel-wise similarities between the hippocampus and other brain regions among the AD group were driven by diminished gray matter volume, with strong correlations between low volume gray matter voxels within the hippocampus and extra-hippocampal brain voxels, the highest proportion of which were found within basal ganglia and parasagittal frontal lobe structures according to the color-bar provided (Figure 2: red=high; black=low).