Biexponential and diffusional kurtosis imaging, and generalised diffusion-tensor imaging (GDTI) with rank-4 tensors: a study in a group of healthy subjects

Abstract

Clinical diffusion imaging is based on two assumptions of limited validity: that the radial projections of the diffusion propagator are Gaussian, and that a single directional diffusivity maximum exists in each voxel. The former can be removed using the biexponential and diffusional kurtosis models, the latter using generalised diffusion-tensor imaging. This study provides normative data for these three models. Eighteen healthy subjects were imaged. Maps of the biexponential parameters D (fast), D (slow) and f (slow), of D and K from the diffusional kurtosis model, and of diffusivity D' were obtained. Maps of generalised anisotropy (GA) and scaled entropy(SE) were also generated, for second and fourth rank tensors. Normative values were obtained for 26 regions. In grey versus white matter, D (slow) and D' were higher and D (fast), f (slow) and K were lower. With respect to maps of D', anatomical contrast was stronger in maps of D (slow) and K. Elevating tensor rank increased SE, generally more significantly than GA, in: anterior limb of internal capsule, corpus callosum, deep frontal and subcortical white matter, along superior longitudinal fasciculus and cingulum. The values reported herein can be used for reference in future studies and in clinical settings.