Spatiotemporal Characterization of White Matter Hyperintensity Pathophysiology

2170 

Abstract Submission 

Olivier Parent¹, Gabriel Devenyi², Aurélie Bussy³, Grace Pigeau³, Manuela Costantino³, Jérémie Fouquet¹, Daniela Quesada-Rodriguez³, Mahsa Dadar³, Mallar Chakravarty⁴

¹Douglas Mental Health University Institute, Montreal, QC, ²McGill University, Montreal, Quebec, ³McGill University, Montreal, QC, ⁴Brain Imaging Centre, Douglas Research Centre, Montreal, Quebec

Olivier Parent  
Douglas Mental Health University Institute
Montreal, QC

Gabriel Devenyi  
McGill University
Montreal, Quebec

Aurélie Bussy  
McGill University
Montreal, QC

Grace Pigeau  
McGill University
Montreal, QC

Manuela Costantino  
McGill University
Montreal, QC

Jérémie Fouquet  
Douglas Mental Health University Institute
Montreal, QC

Daniela Quesada-Rodriguez  
McGill University
Montreal, QC

Mahsa Dadar  
McGill University
Montreal, QC

Mallar Chakravarty, PhD  
Brain Imaging Centre, Douglas Research Centre
Montreal, Quebec

The sparse vascularization of the deep white matter makes it particularly vulnerable to vascular dysfunction, resulting in areas of ischemia/hypoxia and blood-brain barrier leakage frequently seen in elderly individuals. These events are detectable with magnetic resonance imaging (MRI) and appear as white matter hyperintensities (WMHs) on Fluid-Attenuated Inversion Recovery (FLAIR) T2-weighted images [1]. Numerous studies have demonstrated that the pathophysiology underlying WMHs is highly heterogeneous, with edema, demyelination, axonal loss, oligodendrocyte loss, and inflammation being present at various degrees or even absent [2]. It remains unclear if WMHs in different spatial locations all represent similar pathophysiology and etiology [3]. Here, we demonstrate a framework developed to estimate WMH pathophysiology in vivo using microstructural MRI and normative models, which allowed us to precisely characterize spatiotemporal patterns of WMH tissue alterations and assess the added value of that information in predicting cognitive function.

We used data from 32,014 UK Biobank (UKB) participants. T1-weighted (T1w) and FLAIR images were used for identifying WMH and normal-appearing white matter (NAWM) regions using the Brain tISsue segmentatiON (BISON) pipeline [4]. Diffusion-weighted and susceptibility-weighted images were used to derive fluid-sensitive, fiber-sensitive, and myelin- and iron-sensitive microstructural markers [5]. Diffusion tensor imaging (DTI) and neurite orientation dispersion and density imaging (NODDI) were used to model the diffusion MRI signal. Multispectral registration to a custom unbiased UKB template was performed using T1w and fractional anisotropy (FA) maps as inputs [6]. In this common space, we calculated voxel-wise normative models of NAWM using Bayesian linear regression with sex and age modelled with 4th-order B-splines with the PCN toolkit (Fig. 1A) [7]. We then normalized WMH microstructural maps using these age- and sex-specific atlases of expected NAWM microstructure (Fig 1B), resulting in subject-wise estimates of WMH pathophysiology (i.e., change in microstructure as the tissue transitioned from NAWM to WMH) (Fig. 1C).

First, from between-subject averages of WMH microstructural abnormality (Fig. 2A), we used spectral clustering [8] to derive spatial patterns of WMHs that share similar pathophysiological properties (Fig. 2B). The first cluster (periventricular) has low abnormality on all metrics. The second (posterior) and third (anterior) clusters both show fluid accumulation, fiber alterations, and myelin and iron loss, while the anterior cluster shows higher abnormality on most metrics. Second, we calculated the subject-wise median WMH microstructural abnormality value within each spatial cluster. Using the Subtype and Stage Inference (SuStaIn) algorithm [9], we determined the temporal sequence at which these markers become abnormal in each WMH location (Fig. 2C) and assigned a stage of WMH pathophysiological progression for each subject in each WMH region. We did not find evidence for sub-trajectories. Third, using random forest models, we assessed the predictive power of different combinations of predictors (WMH volumes, WMH sustain stages, median WMH abnormality for all metrics, and their combinations) on cognitive function (Fig. 2D).

Using a novel framework to assess WMH pathophysiological processes in vivo, we uncovered spatiotemporal patterns of WMH tissue alterations. Our results separating anterior and posterior WMHs are consistent with accumulating evidence showing that posterior WMHs may be linked to Alzheimer's pathology, whereas anterior WMHs are more associated with vascular pathologies [10]. We further demonstrated that the pathophysiological severity of WMH can be adequately summarized into one stage of disease with SuStaIn and that this information increased the predictive power of WMHs on cognitive tests, particularly those sensitive to processing speed.

Neurodegenerative/ Late Life (eg. Parkinson’s, Alzheimer’s) ²

Aging

Bayesian Modeling

Diffusion MRI Modeling and Analysis

White Matter Anatomy, Fiber Pathways and Connectivity ¹

Aging

Cerebrovascular Disease

Cognition

Computational Neuroscience

Demyelinating

Machine Learning

Modeling

White Matter