Regional diffusion imaging measures to disentangle SVD-related hypertensive arteriopathy versus cerebral amyloid angiopathy

Background: Detecting and distinguishing early changes due to the two key subtypes of cerebral small vessel disease (SVD), hypertensive arteriopathy (HA) and cerebral amyloid angiopathy (CAA), has significant clinical implications. Our goal was to develop and validate dMRI signatures associated with HA-SVD and CAA-SVD proxies and assess their clinical utility using Alzheimer’s disease and SVD biomarkers, pathology, and cognition. Methods: Two independent cohorts with baseline dMRI scans, T2* gradient-echo MRI, and vascular risk measures were analyzed: Mayo Clinic Study of Aging (MCSA, N = 1080) and Alzheimer’s Disease Neuroimaging Initiative (ADNI, N = 549). In MCSA, regional dMRI measures associated with proxies of HA-SVD (hypertension) and CAA-SVD (lobar cerebral microbleeds) were identified using logistic regression models. The top regional features were then used to compute composite dMRI indices for HA-SVD and CAA-SVD. These dMRI indices were validated in ADNI and in an independent pathology sample of MCSA (N = 147). In MCSA, we also computed standard global SVD indices from diffusion and FLAIR MRI and compared them with dMRI indices to reflect SVD subtypes. Next, we evaluated the association of these indices with cognitive performance (global, attention, and memory) using regression models, after adjusting demographics, white matter hyperintensities (WMH) and amyloid. Results: Hypertension was associated with reduced microstructural integrity predominantly in fronto-parieto-projection pathways, whereas lobar microbleeds were associated with occipito-parietal damage. These differential tract association patterns with HA-SVD and CAA-SVD proxies were less pronounced in ADNI. In the community-dwelling MCSA cohort with higher prevalence of vascular disease, dMRI indices provided more differentiated associations with both proxies of SVD than global SVD indices, underscoring added value for etiology-specific identification. As expected, CAA-SVD indices were more strongly associated with occipital WMH and amyloid burden and were linked to CAA pathology scores. CAA-SVD indices also had greater association with memory performance, independent of amyloid and WMH. Conversely, HA-SVD indices were robustly associated with post-mortem Kalaria scales and attention scores. Conclusions: Using three datasets, (population-based sample, independent cohort, pathology sample), we found that regional dMRI signatures can capture distinct SVD processes of HA and CAA. These dMRI signatures offer potential for early differential identification of SVD subtypes and can aid in guiding clinical decision making and prevention. Supplementary Information: The online version contains supplementary material available at 10.1186/s13024-026-00942-4.