Evaluation and interpretation of DTI-ALPS, a proposed surrogate marker for glymphatic clearance, in a large population-based sample

Abstract

Background: The diffusion tensor imaging along perivascular spaces index (DTI-ALPS), which measures diffusivity in the perivascular spaces along the medullary veins, has gained popularity and controversy as a surrogate marker of glymphatic clearance. The goal of this work is to automatically estimate DTI-ALPS in a large population-based sample, evaluate the correlates of the signal observed in the context of aging and dementia biomarkers, and evaluate its clinical usefulness.

Methods: We identified 2715 participants aged 30 + years in the population-based Mayo Clinic Study of Aging with diffusion MRI. We calculated DTI-ALPS through a modified pipeline of previously published methods. We evaluated DTI-ALPS using different protocols and scanners and reported ICC for agreement. We examined the predictors of longitudinal DTI-ALPS with demographics (age, sex), vascular risk, clinical data (diagnosis, global cognition), and imaging markers (white matter hyperintensity (WMH), global amyloid load from PIB-PET, and temporal meta-ROI Tau-PET SUVR) in a subset of participants aged 50 + years using Pearson correlations, ANCOVA with adjustments for age and sex, and linear mixed effect models. We also compared the utility of DTI-ALPS with WMH for prediction of cognitive decline.

Results: With modifications to the automated DTI-ALPS pipeline, consistent measurements can be made from data obtained with different protocols on different scanners. DTI-ALPS was negatively correlated with age, vascular risk, and WMH burden and was positively correlated with cognitive scores and higher in females. In the longitudinal models, WMH explained the greatest variability in decline of DTI-ALPS. The age and sex adjusted associations with AD biomarkers (amyloid and tau) were minimal. DTI-ALPS had a significant interaction with WMH on the rate of cognitive decline.

Conclusions: DTI-ALPS can be reliably automated in large samples. The computed DTI-ALPS was associated with vascular dysfunction (vascular risk and WMH) and may provide additional complementary information about cognitive decline. The low associations with AD biomarkers suggest that DTI-ALPS may be a poor surrogate of AD.

Keywords: Amyloid deposition; DTI-ALPS; Vascular dysfunction.

Fig. 1

(A) RGB encoded FA image showing the placement of ROIs. (B) Change from mean to median and direction-based voxel selection helped in cases where the ROIs overlap with other tracts or ventricles. In the left image, the right superior corona radiata (RSCR, arrow B1) and left superior corona radiata (LSCR, arrow B2) overlap with the ventricles. In the right image, the right superior longitudinal fascicle (RSLF, arrow B3) and left superior longitudinal fascicle (LSLF, arrow B4) spheres have voxels with dominant right-left (RL) diffusion, violating the DTI-ALPS assumption and are therefore excluded in our implementation of DTI-ALPS. (C) Distribution across scans of the estimated fraction of voxels rejected. We excluded voxels based on the formulae in Sect. 2 below. (D) Distributions of the number of voxels per ROI of the diffusivities in the included (top row) and excluded voxels (bottom row) for the projection and association areas. The vertical axes are in 10⁶ voxels with units , but note that the excluded projection voxels have been scaled up by 10 to make them visible. Note that exclusion has more impact on the association regions than the projection regions

Fig. 2

(A) Different components of the DTI-ALPS formulae before 1000 extraction from Siemens (B) After 1000 extraction from Siemens (C) Change in different components of the DTI-ALPS with Age. We observed better harmonization of DTI-ALPS numerator and denominator from (A) to (B) The diffusivity values in Y PROJ had the largest regression coefficient with age in C

Fig. 3

DTI-ALPS by age stratified with sex, CMC, and imaging biomarkers (Amyloid PIB and WMH)

Fig. 4

Comparison of DTI-ALPS between groups defined by imaging biomarkers of AD, WMH, and cognitive status

Fig. 5

Correlation of DTI-ALPS with PVS fractions. We report the slope, Pearson-r and p-value of the regression fit in the title of the respective plots before [top] and after [bottom] accounting for age. We only consider non-zero fractions in these plots (A) PVS voxel fraction measured in the DTI-ALPS regions (B) Number of PVS voxels per risk area voxel in the whole MRI (C) Volume fraction of PVS to the volume of risk in frontal and parietal regions where the DTI-ALPS regions overlap

Fig. 6

Predicted cognitive trajectories of an average 70-year-old male with CMC = 2 and baseline cycle number = 2. DTI-ALPS, CMC, and WMH were important predictors of cognition