Cant create a proper FOD?

[egemen4552]

In my dataset whenever I try to create fiber orientation densities whether I use tournier or dhollander I dont get good results.

Is there a problem with preprocessing steps? What should I do in this situation

Preprocessing steps

mrconvert $PA_DWI raw_dwi.mif -fslgrad $PA_BVEC $PA_BVAL
dwidenoise raw_dwi.mif dwi_den.mif -noise noise_AP.mif
mrdegibbs dwi_den.mif dwi_den_unr.mif -axes 0,2
dwiextract dwi_den_unr.mif - -bzero | mrmath - mean mean_b0_PA.mif -axis 3
dwifslpreproc dwi_den_unr.mif dwi_preproc.mif -nocleanup -pe_dir PA -rpe_none -eddy_options " --slm=linear --data_is_shelled"
dwibiascorrect ants dwi_preproc.mif dwi_preproc_unbiased.mif -bias bias_PA.mif
dwi2mask dwi_preproc_unbiased.mif mask.mif

after these steps I do dwi2response and dwi2fod but unfortunately results are not good.

my Raw DWI

my WM-FOD

my GM-FOD

[jdtournier]

Hi @egemen4552,

Sorry if this comes a bit late – we’ve been busy preparing for and running the MRtrix workshop.

I wonder if your issue is similar to the one discussed in this post:

Mtnormalise nan values as balance factors

Yes, it looks like you have single-shell data (b = 0 & 1,000 s/mm²), which can’t directly support a 3-tissue decomposition – but your dwi2fod call does request 3 tissue types. This is why the GM segment is coming out all wrong (though I would have expected it to come out closer to straight zeros…). There are thankfully other options for dealing with single-shell data: use the original single-shell CSD method (i.e. dwif2fod csd). This will only rely on the b=1,000 s/mm² data. use the new MSMT-…

Have a look and see if this is relevant, and let us know if that doesn’t fix the issue.

All the best,
Donald