dwipreproc can be used for EPI distortion correction if supplied with matching reversed phase-encoded EPI images - this is done using FSL’s TOPUP. It looks like you don’t have suitable data for this step, and furthermore you pass the -rpe_none option to dwipreproc, which explicitly disables EPI distortion correction.
Well, it means all your data were acquired with the phase encode direction running along the negative y axis. But what we mean when we talk of reversed phase encode images is the available of images acquired with the opposite phase encode direction to the rest of the DWI data. That would generally be a separate set of images to the main DWI series.
OK, my point was that these nodes are not defined on the basis of functionality. If you want the node coordinates, maybe the simplest thing is to generate an image of the coordinates of each voxel (you can use warpinit for that), and then measure the mean value within your node ROI. This depends on your ROI being a binary mask, rather than an image of integer labels. But you can convert it to a binary image using mrcalc.
To provide an example of what I mean, assuming your parcellation is an image of labels, and you want to extract the mean position for node index 15, first generate an image of voxel positions (you only need to do this once):
warpinit nodes_image.nii positions.mif
then compute the mean value of these positions within the voxels that have value 15:
mrcalc nodes_image.nii 15 -eq - | mrstats positions.mif -mask - -output mean
(above not tested, but hopefully close enough as a starting point…)