Spatial-temporal interactions between white matter hyperintensities and multiple pathologies across the Alzheimer's disease continuum

Abstract

Introduction: The interactive relationships between Alzheimer's disease (AD) and white matter hyperintensities (WMHs) in multiscale brain structural networks still need to be clarified.

Methods: Based on subjects enrolled from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database, regional WMHs, amyloid beta (Aβ) accumulation, and microstructural changes detected by diffusion weighted imaging (DWI) in multiscale brain networks were modeled by time-evolving graphs; their interactive relationships were further investigated using Granger causality after constructing pseudo-time subject sequences.

Results: In up to 86% of the extracted pseudo-time subject sequences, Aβ was determined to be the Granger cause of WMHs in the structural connectivity of the inferior longitudinal fasciculus (ILF). Meanwhile WMHs were significantly correlated with microstructural changes measured by reduced fractional anisotropy in the inferior fronto-occipital fasciculus, ILF, and cingulum, which Granger causality pathways detected in 91%, 94%, and 93% of pseudo-time subject sequences, respectively.

Discussion: These findings provide novel insights for understanding the multiscale space-time interactions between WMHs and AD pathologies.

Highlights: This study proposed time-evolving graph modeling of heterogeneous disease markers (amyloid beta [Aβ], white matter hyperintensities [WMHs], and microstructural changes of white matter tracts) across the Alzheimer's disease (AD) continuum to investigate their complex interactions in multiscale brain structural networks. Regional accumulation of Aβ promoted WMH progression in subnetworks connected by the inferior longitudinal fasciculus (ILF). Regional WMHs were strongly associated with bundle-specific microstructural changes in the ILF, inferior fronto-occipital fasciculus, and cingulum. These results might provide novel insights for understanding the interactive relationship between cerebral small vessel disease and AD.

Keywords: Alzheimer's disease; brain structural connectomes; causal inference; time‐evolving graph; white matter hyperintensities.

FIGURE 1

Spatial modeling of heterogeneous imaging markers. A, Modeling the spatial distribution of WMHs, and (B) modeling the microstructural changes in brain structural network. DWI, diffusion weighted imaging; FA, fractional anisotropy; FLAIR, fluid‐attenuated inversion recovery; MD, mean diffusivity; MRI, magnetic resonance imaging; SN, spatially normalized; TEG, time‐evolving graph; WMHs, white matter hyperintensities

FIGURE 2

Flowchart of constructing pseudo‐longitudinal subject order of Aβ progression. Aβ, amyloid beta; CL, Centiloid; PET, positron emission tomography.

FIGURE 3

Flowchart of Granger causality inference based on pseudo‐time subject sequences of time‐evolving graphs (GC‐pTEGs). Brain structural connectome b was constructed on whole‐brain or 20 white matter tracts. Aβ, amyloid beta; WMHs, white matter hyperintensities

FIGURE 4

Results of constructing pseudo‐longitudinal subject order of Aβ progression. A, The pseudo‐longitudinal subject order of Aβ progression. B, Pseudo‐longitudinal spreading order of regional Aβ; here we presented several brain regions along the spreading order for example. Aβ, amyloid beta; ADE, Alzheimer's disease dementia; CN, cognitively normal; Hippo, Hippocampus; INS, insula; LG, lingualis gyrus; MCI, mild cognitive impairment; OG, orbital gyrus; PrCu, precuneus

FIGURE 5

Validation of pseudo‐longitudinal subject order. A, The distribution of pMCI+ and sMCI+ subjects. B, Linear regression between years to dementia diagnosis and proportion of regions suffering from abnormal Aβ accumulation in pMCI+ group, ***p < 0.001. Aβ, amyloid beta; pMCI+, progressive mild cognitive impairment; sMCI+, stable mild cognitive impairment