Orientation Dependence of R 2 ’ in the White Matter: Digital Characterization, Modelling and Implications for Studying Brain Physiology
Discuss this preprint
Start a discussion What are Sciety discussions?Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
Purpose
R 2 *, the transverse relaxation rate, reflects local magnetic field inhomogeneities from susceptibility differences with R2’, the reversible component sensitive to blood oxygenation. Orientation dependence of R 2 and R 2 * in white matter (WM) are attributed primarily to myelin, with vascular contributions to R2’ uncharacterized. This study examined WM R 2 ’ orientation dependence, evaluated existing models, and developed an improved model combining myelin and blood.
Methods
Simulations used BOLDswimsuite with 2D WM voxels (5,000 fibres). Spin-echo (TE = 70ms) and gradient-echo (TE = 35ms) signals were simulated across 30 fibre orientations (0°–90°). R 2 ′ was calculated as R 2 * − R 2 . Oxygenation, cerebral blood volume (CBV), vessel size, and vessel geometry were varied. Four published models and a novel Myelin-Blood model were fitted to R 2 ′ data and compared using R 2 and RMSE.
Results
R 2 and R 2 * showed strong orientation dependence. Parallel and mixed vessel geometries produced greater R 2 ′ amplitude and orientation dependence than random geometries; decreasing oxygenation and increasing CBV amplified orientation effects. Vessel size altered peak locations. Existing vascular models performed poorly, and the Empirical Myelin Model erred near the magic angle. The Myelin-Blood model provided near-perfect fits (mean R 2 = 0.999, RMSE = 0.007 Hz), reducing RMSE by ~74%.
Discussion
WM R 2 ′ cannot be explained by vascular or myelin effects alone. Myelin is the primary determinant of orientation dependence, but vascular contributions were evident near the magic angle and low oxygenation. The Myelin-Blood model improves WM R 2 ’ characterisation and may reduce orientation-dependent bias and improve interpretation of WM BOLD fMRI signals.
