dMRI data were acquired pre- and post-taVNS treatment on a Skyra 3T MRI (Siemens Healthineers, Erlangen, Germany) immediately following feeding with swaddling to induce natural sleep. A diffusional kurtosis imaging (DKI) [23 (link)] protocol utilized b-values of b = 0, 1000 and 2000 s/mm2, (3 mm)3 isotropic voxels, echo time = 122 ms, repetition time = 6700 ms with 30 diffusion-encoding directions per non-zero b-value and 40 contiguous axial slices per image volume. Additionally, 9 non-diffusion weighted (b = 0 s/mm2) images were also acquired with matching imaging parameters.
Image volumes and their corresponding diffusion-encoding directions that were corrupted by motion or signal dropout were manually removed prior to image processing. The DESIGNER pipeline [24 (link)], reformatted into Python 3 [25 ], was employed for the full DKI analysis. The pipeline includes Marchenko-Pastur-PCA denoising [26 (link)], Gibbs artifact correction [27 (link)], eddy correction [28 (link), 29 (link)], Gaussian smoothing with a kernel 1.25 times the voxel size [30 (link)] and Rician noise bias correction [31 (link)]. The output of DESIGNER is tensor-derived diffusion and kurtosis metrics; the FA was specifically focused in on for this investigation.
Image volumes and their corresponding diffusion-encoding directions that were corrupted by motion or signal dropout were manually removed prior to image processing. The DESIGNER pipeline [24 (link)], reformatted into Python 3 [25 ], was employed for the full DKI analysis. The pipeline includes Marchenko-Pastur-PCA denoising [26 (link)], Gibbs artifact correction [27 (link)], eddy correction [28 (link), 29 (link)], Gaussian smoothing with a kernel 1.25 times the voxel size [30 (link)] and Rician noise bias correction [31 (link)]. The output of DESIGNER is tensor-derived diffusion and kurtosis metrics; the FA was specifically focused in on for this investigation.
