High-resolution magnetization-transfer imaging of post-mortem marmoset brain: Comparisons with relaxometry and histology.

Cell membranes and macromolecules or paramagnetic compounds interact with water proton spins, which modulates magnetic resonance imaging (MRI) contrast providing information on tissue composition. For a further investigation, quantitative magnetization transfer (qMT) parameters (at 3T), including the ratio of the macromolecular and water proton pools, F, and the exchange-rate constant as well as the (observed) longitudinal and the effective transverse relaxation rates (at 3T and 7T), R1obs and R2*, respectively, were measured at high spatial resolution (200 µm) in a slice of fixed marmoset brain and compared to histology results obtained with Gallyas' myelin stain and Perls' iron stain. R1obs and R2* were linearly correlated with the iron content for the entire slice, whereas distinct differences were obtained between gray and white matter for correlations of relaxometry and qMT parameters with myelin content. The combined results suggest that the macromolecular pool interacting with water consists of myelin and (less efficient) non-myelin contributions. Despite strong correlation of F and R1obs, none of these parameters was uniquely specific to myelination. Due to additional sensitivity to iron stores, R1obs and R2* were more sensitive for depicting microstructural differences between cortical layers than F.

Orientation dependence of magnetization transfer parameters in human white matter.

Quantification of magnetization-transfer (MT) experiments is typically based on a model comprising a liquid pool "a" of free water and a semisolid pool "b" of motionally restricted macromolecules or membrane compounds. By a comprehensive fitting approach, high quality MT parameter maps of the human brain are obtained. In particular, a distinct correlation between the diffusion-tensor orientation with respect to the B0-magnetic field and the apparent transverse relaxation time, T2(b), of the semisolid pool (i.e., the width of its absorption line) is observed. This orientation dependence is quantitatively explained by a refined dipolar lineshape for pool b that explicitly considers the specific geometrical arrangement of lipid bilayers wrapped around a cylindrical axon. The model inherently reduces the myelin membrane to its lipid constituents, which is motivated by previous studies on efficient interaction sites (e.g., cholesterol or galactocerebrosides) in the myelin membrane and on the origin of ultrashort T2 signals in cerebral white matter. The agreement between MT orientation effects and corresponding forward simulations using empirical diffusion imaging results as input as well as results from fits employing the novel lineshape support previous suggestions that the fiber orientation distribution in a voxel can be modeled as a scaled Bingham distribution.