Abstract
Dementia arises from multifactorial neurodegenerative processes, with growing evidence implicating blood–brain barrier dysfunction, neuroinflammation, and vascular injury as contributors to progressive brain tissue damage. Because decline in white matter microstructural integrity is a prominent feature of brain aging, biomarkers that capture this process may improve dementia risk stratification beyond conventional clinical factors. Fractional anisotropy (FA), derived from diffusion tensor imaging, reflects white matter microstructural integrity and may provide a sensitive marker of downstream injury associated with chronic vascular and barrier dysfunction. We conducted a retrospective single-center study of 528 participants, including individuals with dementia ( n = 176), age-matched cognitively normal older adults ( n = 176), and a young healthy reference cohort ( n = 176), all of whom underwent standardized 3-Tesla diffusion MRI. Dementia prediction was evaluated using elastic net-regularized logistic regression with nested tenfold cross-validation, incorporating 16 prespecified clinical and imaging predictors. Corpus callosum FA showed the largest standardized coefficient within the penalized model, exceeding all demographic and clinical variables. The model demonstrated strong discrimination between dementia and age-matched controls (area under the receiver operating characteristic curve [AUC] = 0.970; sensitivity = 90.9%; specificity = 93.2%). FA also declined stepwise from young adults to cognitively normal older adults to participants with dementia, consistent with progressive white matter microstructural injury across the aging-dementia continuum. These findings indicate that corpus callosum FA provides additional predictive value beyond conventional risk factors and support diffusion MRI as a potentially scalable biomarker of age-related white matter microstructural injury with potential utility for dementia risk stratification and biologically informed prevention strategies.