Moyamoya disease involves progressive narrowing or occlusion of the terminal internal carotid artery and/or proximal anterior or middle cerebral arteries, together with abnormal basal collateral vessels on vascular imaging. The study included only people without clinically overt ischemic stroke or intracranial hemorrhage before enrollment, making the cohort a model of chronic cerebral hypoperfusion with less interference from major infarction or hemorrhagic injury. The retrospective observational design and complete-case analysis mean that the results cannot establish causation or determine whether either MRI measure predicts subsequent cognitive change. The researchers also noted that asymptomatic moyamoya disease is not necessarily free of structural white-matter injury or subtle cognitive impairment.

Participants had a mean age of 44.22 years, and the cohort included 34 women and 25 men. Mean MoCA and MMSE scores were 22.63 and 26.12, respectively. The mean ALPS index was 1.42, and the mean PVeD index was 0.48. Angiographic involvement was bilateral in 43 patients, or 72.9%, and unilateral in 16, or 27.1%. Maximum hemispheric Suzuki stage ranged from 1 to 4, with a median of 3. Median left and right Suzuki stages were both 2, and a paired Wilcoxon signed-rank test did not show a significant left-right difference in angiographic severity, with P = 0.095. The total white-matter hyperintensity score had a median of 3, an interquartile range of 2–5 and an observed range of 0–9.

MRI was acquired on a 3.0-tesla Siemens MAGNETOM Prisma scanner using a 64-channel head and neck coil. The protocol included three-dimensional T1-weighted imaging, axial T2-weighted imaging, fluid-attenuated inversion recovery, susceptibility-weighted imaging and diffusion tensor imaging. DTI data underwent correction for susceptibility-related distortion, eddy-current effects and participant motion. Tensor maps were then registered to MNI152 standard space for ALPS region-of-interest placement and parameter extraction. Visual quality control excluded images with substantial motion, signal loss, severe distortion, incomplete brain extraction, failed correction or unacceptable registration.

For DTI-ALPS, fixed bilateral regions of interest were placed at the level of the bodies of the lateral ventricles in projection- and association-fiber regions. The left and right indices were calculated as the mean of diffusivity along the x axis in the projection and association regions divided by the mean of y-axis diffusivity in the projection region and z-axis diffusivity in the association region. The mean ALPS index was the arithmetic average of the bilateral values. PVeD was calculated with the automated EstPVeD pipeline in MATLAB R2024a and SPM12. It registers a lateral-ventricle template to diffusion space, constructs an anatomically adaptive periventricular region and uses tissue segmentation to exclude cerebrospinal fluid and non-brain tissue. The transverse tensor ratio was calculated as Dxx divided by the square root of Dxx squared plus Dyy squared plus Dzz squared, with median values extracted from the left and right periventricular regions.

Before adjustment for confounding, mean ALPS was positively correlated with MoCA at Spearman rho = 0.493 and with MMSE at rho = 0.326. The left ALPS index correlated with MoCA at rho = 0.522 and MMSE at rho = 0.372, while right ALPS correlated with MoCA at rho = 0.335 but not MMSE. After adjustment for age and sex, the partial correlation for mean ALPS was 0.494 with MoCA and 0.316 with MMSE. Left ALPS remained associated with both scores, with partial rho values of 0.508 and 0.346, and right ALPS remained associated with MoCA at partial rho = 0.349. Right ALPS was not significantly associated with MMSE after correction, with partial rho = 0.203 and q = 0.2229. These results indicate that the imaging-cognition relationship was stronger and more consistent for ALPS than for PVeD in this sample.

PVeD showed a less consistent pattern in the unadjusted analysis. Left PVeD correlated with MMSE at rho = 0.320, while its association with MoCA was close to the corrected threshold, with q = 0.0526. Mean PVeD had a nominal association with MMSE at rho = 0.257, P = 0.0496, but it did not survive correction, with q = 0.0743. Mean PVeD was not significantly associated with MoCA, with rho = 0.246 and q = 0.0805. Following age and sex adjustment, none of the six PVeD associations met the false-discovery-rate threshold. Mean PVeD had partial rho values of 0.147 with MoCA and 0.182 with MMSE. Left PVeD and MMSE had a partial rho of 0.243, but q = 0.1365, while the other adjusted PVeD correlations were also non-significant after correction.

Multivariable linear regression produced a similar pattern. In models adjusted for age and sex, mean ALPS was associated with MoCA with an unstandardized beta of 11.716, a 95% confidence interval of 4.930 to 18.502, P = 0.0011 and adjusted R-squared of 0.233. Its association with MMSE was also nominally significant, with beta = 5.213, 95% confidence interval 0.619 to 9.807 and P = 0.0269. Mean PVeD was not significantly associated with MoCA, with beta = 34.662, 95% confidence interval −1.697 to 71.022 and P = 0.0613. Its MMSE association was nominal, with beta = 23.944, 95% confidence interval 0.735 to 47.153 and P = 0.0434. The regression P values were exploratory and were not additionally adjusted for multiple testing.

Adding total white-matter hyperintensity burden attenuated the ALPS-MMSE relationship to non-significance, with beta = 3.739, 95% confidence interval −0.897 to 8.375 and P = 0.1117. The ALPS-MoCA association remained statistically significant, with beta = 9.281, 95% confidence interval 2.512 to 16.051, P = 0.0081 and adjusted R-squared of 0.301. Mean PVeD was not associated with either cognitive score after WMH adjustment. In an extended model adding maximum hemispheric Suzuki stage to age, sex and WMH burden, mean ALPS remained associated with MoCA, although the estimate was smaller: beta = 7.790, 95% confidence interval 0.816 to 14.765, P = 0.0293 and adjusted R-squared of 0.317. ALPS-MMSE and both PVeD associations were non-significant in those extended models. The sensitivity analyses therefore suggest that structural white-matter injury and angiographic severity account for part, but not necessarily all, of the imaging-cognition relationships.

Several limitations constrain the findings. The sample was relatively small and cross-sectional, so the associations cannot establish causality, longitudinal predictive value or the direction of any imaging-cognition relationship. APOE genotype and amyloid-beta or tau biomarkers were unavailable, meaning occult Alzheimer’s pathology, mixed vascular-neurodegenerative disease and APOE-related effects could not be excluded. Education level was also unavailable and may have influenced MoCA and MMSE performance. Although WMH burden and maximum hemispheric Suzuki stage were included, potential confounders including ventricular volume, brain atrophy, lacunes, cerebral microbleeds, quantitative cerebral perfusion and cerebrovascular reserve were not quantitatively modeled. MoCA and MMSE are global screening instruments rather than comprehensive neuropsychological tests, so the absence of a PVeD association with global scores does not exclude relationships with executive function, processing speed, attention or memory.

The biological meaning of the two metrics also requires caution. In chronic hypoperfusion, periventricular diffusion may reflect white-matter injury, demyelination, tissue rarefaction, ventricular enlargement, altered tissue water content, ischemia-related changes and potentially interstitial fluid transport. PVeD was introduced as a marker associated with amyloid burden and cognition in the Alzheimer’s disease continuum, but the present study shows that its relationship with global cognition was less consistent in a chronic-ischemic population. ALPS is not a direct measure of glymphatic flow either: it can be influenced by fiber architecture, crossing fibers, region-of-interest placement, scanner parameters and disease-related tissue changes. Agreement between ALPS and tracer-based MRI measures of cerebrospinal fluid distribution and clearance remains limited. The study therefore supports a disease-specific, structurally informed interpretation of diffusion-based markers rather than a claim that one index is a superior measure of true glymphatic function. Prospective studies combining longitudinal cognition, perfusion and vascular-injury measures, Alzheimer’s biomarkers and tracer-based fluid assessments will be needed to clarify what each index reflects and whether either predicts cognitive trajectories.

Subject of Research: Cross-sectional comparison of DTI-ALPS and periventricular diffusivity with cognition in asymptomatic moyamoya disease

Article Title: Disease-context dependence of periventricular diffusivity and DTI-ALPS in asymptomatic moyamoya disease

Article References: Liu, H., Liu, X., Ji, H., Xiao, A., & Liu, Y. (2026). Disease-context dependence of periventricular diffusivity and DTI-ALPS in asymptomatic moyamoya disease. Frontiers in Neurology, 17, 1927018. https://doi.org/10.3389/fneur.2026.1927018

Image Credits: AI Generated

DOI: 10.3389/fneur.2026.1927018

Keywords: moyamoya disease, DTI-ALPS, periventricular diffusivity, diffusion MRI, chronic hypoperfusion, brain fluid transport, cognitive impairment, white matter injury, vascular cognitive impairment, glymphatic system, Disease-context, dependence