Previously, we introduced the double volumetric navigated diffusion sequence (DvNav) (Alhamud et al., 2014) to measure and correct in real time for subject motion and B0 distortion (in terms of zero- and first order shims). Our results demonstrated that the initial shim prepared by the MRI scanner changes during the DTI acquisition. These changes may occur due to subject motion but are often present even in stationary scans. Currently, the DvNav technique performs shimming over the whole DTI volume, which may not be optimal for 2D multislice sequences. In the current work, we investigated, in the presence of subject motion, whether shimming parameters differ from region to region and for different sized regions across the DTI volume.

The DvNav sequence acquires two navigators with different echo times immediately after each DWI volume to reconstruct in real time two different field maps: one for the navigator volume and one for the DTI volume. For accurate motion estimation the navigator field of view (FOV) must cover the whole object being imaged. The field map of the DTI volume is generated by mapping the location of the DTI volume onto the navigator field map. Here we create a field map for groups of slices (slabs) by mapping the position of the slab onto the navigator field map. We considered slabs of different thicknesses as illustrated in Figure 1. Slab 1 includes all DTI slices (for example 73 slices). For each successive slab, two DTI slices on either end are removed. The n-th slab therefore includes the (75-2n) central DTI slices, where n=1-35. Since the thickness of each partition (slice) in the navigator is 8 mm, the thickness of the slab to be shimmed in the DTI volume has to be greater than 8 mm so that multiple field map points are available for fitting in the through-plane direction. Following the shim calculation for all slabs, the sequence receives and stores these parameters in addition to the motion estimates. The DvNavSlab sequence applies for each slice the shim values of the smallest slab that it belongs to.

All scans were performed on a Siemens Allegra 3T scanner (Siemens Healthcare, Erlangen, Germany). Two healthy subjects (25 and 30 years) were scanned with a T1W and different diffusion sequences. These include the standard twice-refocused 2D diffusion pulse sequence (Std) (Reese et al., 2003), DvNavW (W refers to shimming across the whole DTI volume) and DvNavSlab. The standard sequence was acquired without subject motion (NoMo), while in the other two scans the subjects were instructed to move (Mo) during the acquisitions. The following parameters were used for each diffusion acquisition: TR = 11163 ms, TE = 86 ms, 73 slices, matrix size 112 x 112, in-plane FOV 224 x 224 mm2, slice thickness 2 mm, 30 non-collinear diffusion gradient directions, diffusion weighting = 1000 s mm-2, eight b=0 scans. Retrospective motion correction and co-registration to the T1W image were performed using FLIRT in FSL (FMRIB Software Library; http://www.fmrib.ox.ac.uk/fsl). The DTI tensor was calculated using dtifit and a white matter (WM) mask was calculated from the T1W image using Fast segmentation.

Figure 2 shows the changes in the zero- and first order shims for two different slabs. Slab 1 includes all DTI slices (from 1 to 73; thickness 146 mm), while slab 30 shims over a smaller DTI volume (slices 30 to 44; thickness 30 mm). The results clearly demonstrate that the pattern of shim changes during motion differ from one region to another within the same DTI volume as indicated by arrows in Fig 2. The effects of subject motion and resulting changes in B0 on DTI data are compared for standard, DvNavW and DvNavSlab acquisitions in one subject in Figure 3. Despite motion, the FA map of the DvNavSlab acquisition shows less distortion and better alignment with T1W WM, especially in the anterior regions of the brain.The patterns of in-plane shim changes are similar for both slabs but with different magnitudes. However, the through-plane shim changes are different in both shape and magnitude between the two slabs and are much larger for slab 30 (~ 13 µT/m). With Slab-by-Slab shim correction, there is a more accurate DTI estimation. This is particularly apparent where the DvNavSlab acquisition showed reduced image distortion in the anterior region of the brain (Fig. 3c).While the MRI scanner performs shimming over the whole DTI volume, the current study shows that shimming differs significantly between different regions in the DTI volume in the presence subject motion.