Altered T1w/T2w-FLAIR Ratio in White Matter Hyperintensities as an Indicator of Structural Integrity Loss: Association with Alzheimer's Disease and Vascular Dementia

Background: White matter hyperintensities (WMH) are prevalent in dementia, but lesion volume does not capture their microstructural heterogeneity. The T1-weighted to fluid-attenuated inversion recovery (T1w/T2w-FLAIR) ratio is sensitive to myelin, gliosis, and tissue water. We tested whether lesion-specific T1w/T2w-FLAIR ratio, referenced to each participant's normal-appearing white matter (NAWM), differs by diagnosis and reflects distinct amyloid and vascular mechanisms in Alzheimer's disease (AD) and vascular dementia (VD). Methods: We analyzed 576 participants from the multicentre BICWALZS cohort (seven South Korean sites), spanning subjective cognitive impairment (SCI, n=71), mild cognitive impairment (n=270), AD (n=125) and VD (n=88). WMH T1w/T2w-FLAIR ratio was regressed on NAWM ratio, yielding standardized residuals as the outcome. Regression and mediation models tested diagnosis, plasma biomarkers, APOE genotype, amyloid PET and vascular risk burden, adjusting for age, sex, education and site. We conducted regression and mediation analyses after multiple imputation for missing variables. We adjusted for hierarchical models using Bonferroni correction. We tested for insensitivity to site effects by applying ComBat harmonization. Results: Older age, AD, VD, and high vascular risk burden were associated with higher residualized T1w/T2w-FLAIR ratios relative to SCI. Lower plasma amyloid-beta 42 (greater amyloid burden) was associated with lower T1w/T2w-FLAIR ratios. Greater vascular burden was associated with greater T1w/T2w-FLAIR ratios, which partially mediated the VD association with T1w/T2w-FLAIR. Lower amyloid-beta 42 (greater amyloid) was associated with lower T1w/T2w-FLAIR, which partially mediated the effect between AD and T1w/T2w-FLAIR ratio. Findings were robust to harmonization. Conclusions: Residualized WMH T1w/T2w-FLAIR ratio captures lesion-specific microstructural variation missed by volumetric measures, consistent with vascular-gliotic injury in VD and coexisting amyloid-linked demyelination in AD. Limitations include the cross-sectional design, no cognitively normal comparison group, and a predominantly Korean sample.