Does the signal arising from a single fascicle significantly deviates from a monoexponential decay with a clinical scanner ?

Introduction. Multiple works have shown that the diffusion attenuation in a voxel significantly deviates from a monoexponential decay, providing evidence that the Gaussian assumption underlying diffusion tensor imaging (DTI) is not appropriate to accurately represent the diffusion-weighted (DW) signal in the voxel. The biophysical mechanisms responsible for this non-monoexponential behavior are numerous and remain under investigation [Yablonskiy2003,Schwarcz2004,Assaf2005, Minati2007,Mulkern2009]. Particularly, it can arise from the presence of multiple fascicles with heterogeneous orientations in voxels (intra-voxel orientation heterogeneity, IVOH) [Minati2007]. Additionally, when using short diffusion pulse duration δ, it may reflect the mixing of unrestricted (free) diffusion, hindered diffusion (arising from water molecules bouncing on the membranes of the cells in the extra-axonal space) and restricted diffusion (arising from the water molecules in the intra-axonal space) for each fascicle [Assaf2005]. However, when imaging with a clinical scanner with long δ, long echo time (TE) and low signal-to-noise ratio, it is not clear that a nonmonoexponential decay can be observed for a single fascicle. In this work we hypothesize that, with a clinical scanner, the observed signal arising from a single fascicle may be monoexponential and that the non-monoexponential behavior observed in voxels may reflect predominantly the IVOH and the presence of an additional compartment that is macroscopically isotropic at the measured diffusion scale. We focused on imaging a region containing a single fascicle orientation, the body of the corpus callosum (CC, Fig.1), to eliminate the effect of IVOH. We estimated the parameters of an unrestricted diffusion compartment and investigated the residual diffusion decay when subtracting the contribution of unrestricted diffusion. Method. Diffusion-weighted magnetic resonance imaging was carried out on a healthy volunteer using a Siemens 3T Trio clinical scanner with a 32 channel head coil (FOV=220mm, matrix=128x128, in-plane resolution=1.7x1.7mm, slice-thickness=2.0mm, TE=126ms, TR=3sec). We first performed an experiment to determine the optimal DWI protocol that ensures uniform SNR across increasing b-values. To this end, we acquired various b-value images between 500 and 5000 in which we measured the SNR. We performed a regression analysis to describe the SNR as a function of the b-value and determined the number of repetitions Nrepetition(b) necessary to ensures uniform SNR across all b-values, given by: