Periventricular and Perivascular‐Space Diffusion MRI Measures, Amyloid Abnormality, and Executive Decline in the Alzheimer's Disease Continuum: A Longitudinal Study

BACKGROUND: Amyloid abnormality is a defining component of Alzheimer disease (AD) biology but does not fully account for variation in executive-function decline, which may also reflect white matter and periventricular injury. PURPOSE: To compare periventricular diffusivity (PVeD), diffusion tensor image analysis along the perivascular space (ALPS), and G-Along Perivascular Space (G-ALPS), an algebraic variation of ALPS, as diffusion MRI measures in relation to amyloid abnormality and longitudinal executive-function decline. STUDY TYPE: Retrospective. POPULATION: The imaging-baseline cohort included 1190 participants from the Alzheimer's Disease Neuroimaging Initiative (ADNI) (70.6 ± 7.5 years; 666 female, 524 male). The primary analytic sample included 302 participants (70.6 ± 6.1 years; 168 female, 134 male). FIELD STRENGTH/SEQUENCE: 3 T/diffusion-weighted spin-echo echo-planar imaging and 3D T1-weighted inversion-prepared gradient-echo imaging. ASSESSMENT: PVeD, ALPS, and G-ALPS were evaluated in relation to cerebrospinal fluid Aβ42, amyloid positron emission tomography (PET) Centiloid, structural MRI measures, white matter hyperintensity (WMH), and longitudinal Alzheimer's Disease Neuroimaging Initiative Executive Function composite (ADNI-EF). STATISTICAL TESTS: Group comparisons, regression, longitudinal models, and bootstrap indirect-effect analyses with Benjamini-Hochberg correction were performed. RESULTS: All three diffusion MRI measures differed across diagnostic groups, with mean values of PVeD, ALPS, and G-ALPS showing lower levels in dementia than cognitively normal participants (PVeD: 0.456 ± 0.048 vs. 0.430 ± 0.047; ALPS: 1.291 ± 0.188 vs. 1.163 ± 0.186; G-ALPS: 1.757 ± 0.512 vs. 1.393 ± 0.467). G-ALPS showed the most consistent association with ADNI-EF trajectory (β = 0.031, 95% CI: 0.003-0.059). Exploratory models identified statistical indirect effects involving inverted CSF Aβ42, ADNI-EF decline, and PVeD (-0.0518), ALPS (-0.0407), or G-ALPS (-0.0460); all survived Benjamini-Hochberg correction, with a corresponding G-ALPS effect for amyloid PET Centiloid (-0.0169). DATA CONCLUSION: Periventricular and perivascular-space diffusion MRI measures were associated with amyloid abnormality and executive-function decline, mainly in cognitively normal and mild cognitive impairment participants. EVIDENCE LEVEL: 3. TECHNICAL EFFICACY: Stage 2.