One of the largest obstacles preventing high angular resolution diffusion imaging (HARDI) and multi shell diffusion-weighted imaging from being incorporated into standard diffusion MRI (DMRI) protocols is the large time burden required to collect sufficient numbers of images with different direction encodings directions and b-values. Recently, multi-band techniques ^1,2, which excite and collect data from multiple slices simultaneously, have greatly increased the speed of acquisition and hold tremendous potential for extending the range of DMRI sequences that can be carried out within research or clinical protocols. To date, however, there has been little validation of the techniques in terms of the effects of multiband acceleration on resulting diffusion parameters. In this study, we have explored the differences of DMRI data collected with and without MB acceleration, and investigate the effects that MB acceleration has on resulting DMRI parameters. Seven healthy subjects were scanned on a 3T Siemens Skyra with a 72 direction, b=1000 s/mm² protocol. Without MB acceleration, TR/TE = 9600/88 ms. With a MB acceleration of 2, TR/TE = 5000/102 ms and with MB acceleration of 3, TR/TE = 4000/102 ms. Six b=0 images were obtained in each of the studies and two additional b=0 images were collected with reverse phase encoding direction to correct EPI distortions using the TOPUP routine within FSL ^3,4. Images were also corrected for eddy currents using the EDDY routine in FSL. LPCA denoising as described by Manjon et al. ⁵ was implemented in MATLAB. MRTrix was used to generate FOD maps as well as to perform probabilistic tractography and track density image formation ^6,7. MATLAB 2015a was used for statistical analysis and calculation of FOD coefficients for constrained spherical deconvolution as described by Tournier et al. ⁸.

As seen in Fig 1, is apparent that the diffusion images produced after preprocessing steps are quite similar across all levels of acceleration, but it should be noted that there are subtle differences throughout the dataset that distinguish images collected without multiband acceleration and those collected with it. Differences in the data become more apparent when the parameters, such as FA, are compared as shown in Fig 2. The FA from several white matter voxels in an image with no multiband acceleration is plotted against the FA from those same voxels from an image with 2x acceleration (A) and 3x acceleration (B). It is clear that the values of FA are similar, but not identical.

The fiber orientation distribution functions generated using the constrained spherical deconvolution technique also vary. Fig 3 shows the FODs from a single voxel in a region of crossing fibers (B-D) from different amounts of multiband acceleration, and from a single voxel within the corpus callosum (E-G). Of note, the primary diffusion directions seem to remain in tact in both voxels, but the FODs in the region of crossing fibers appear to lose some of their angular resolution as the multiband acceleration increases. This observation is further validated in figure 4B, which plots each of the spherical harmonic coefficients used to create the FODs in Fig 3(B-D). Lower order coefficients are very similar for all levels of multiband acceleration, but begin to deviate from as the order increases. This pattern is also seen in the coefficients of the FODs in Fig 3(E-G), plotted in Fig 4A, but the coefficients do not vary as much until higher order coefficients.

Finally, the results of probabilistic tractography appear to remain largely intact despite some of the differences discussed above in earlier processing steps. While there are some minor variations between the tractography results (Fig 5), all of the major, and most minor pathways appear to be in tact. Track density imaging results (not shown) indicate that as more tracts are traced out, the more similar the tractography results are.

While there are subtle differences between data collected with and without multiband acceleration throughout all stages of data processing and analysis, it is difficult to determine how these differences truly affect the results of a diffusion study, given that there is no gold standard other than a sequence without multiband imaging. Despite some visually apparent differences in FODs calculated at some locations in the brain, tractography results seem to remain intact.