An automated approach to deep medullary veins quantification. Association with vascular risk factors and imaging

Although visibility of Deep Medullary Veins (DMVs) has been suggested as an imaging biomarker for cerebral small vessel disease (CSVD) and brain atrophy, qualitative visual assessment of DMVs may lack reliability. In this study, we used a rigorous, automated approach to segment DMVs on SWI and to estimate a vein voxel fraction (VVF). We hypothesized that VVF would be associated with vascular risk factors, imaging markers of CSVD, and brain volumes.

A retrospective analysis of data from participants enrolled in studies of brain aging and prediction of Alzheimer's disease. All underwent 3T MRI (T1WI, FLAIR, SWI, and arterial spin labeling), clinical evaluations, and laboratory tests to assess vascular risks. A SWI image processing pipeline included denoising, bias field correction, adaptive histogram equalization, and multi-scale Jerman filtering that yielded vesselness maps. The vein voxel fraction (VVF) was calculated as a ratio of DMVs voxels in a periventricular region to the volume of the periventricular region. White matter hyperintensities, microbleeds, brain volumes, and CBF were also assessed.

This study included 131 cognitively healthy participants, 71 (65, 76) years (median, Q1, Q3), (51%) female, and a subgroup of subjects with cognitive impairment (n=30, 69 (59, 76) years, 43% female). In the unimpaired group, the VVF positively correlated with systolic blood pressure and body mass index. It showed an inverse association with lateral ventricle volume (all at p < 0.05). It was related to white matter hyperintensities volume only in unadjusted analysis.

Vein voxel fraction was associated with increased weight and high blood pressure. Possibly, these observations reflect venous stasis. These factors should be accounted for while evaluating brain venous system with SWI. In addition, subcortical atrophy was related to less visible DMVs.