Problems with digital phantom

Hey guys,

So I have worked with MRtrix3 before and already have built a working pipeline for real data but I am encountering some problems with a digital phantom I am trying to test.

My generally used pipeline is something like:
Preprocessing(denoise,degibbs,AP-PAcorr,dwifslpreproc) → dwi2response (dhollander) → dwi2fod (msmt_csd) → 5tt seg for GM/WM boundary → tckgen with 5tt for act and gm/wm-border for seed

Now for this phantom I have:

  • Nifti (crossing pipes in a sphere representing fiber bundles)
  • Bvecs+Bvals file (all dimensions correspond with nifti)
  • Mask (just simple binary mask of sphere)
  • Rois file (begin and end points of fiber bundles, labelled 1-begin and 2-end of fiber bundle 1, 3-begin 4-end for fiber bundle 2 etc.)

Im a little bit lost (which is stupid as this should be way simpler than with a whole brain) on how to correctly set up this test. I have left out any PA-AP correction and ofcourse can not do 5tt segmentation but what do I use for the correct seed image if I do not have gm/wm border so I thought use the ROIs file as seed. Now I do this, which gives bad results:

Preprocessing (denoise,degibbs,dwifslpreproc) → dwi2response (tournier) → dwi2fod (csd) → tckgen (rois img as seed, mask img as mask)

Thanks in advance!