proc steps to dwi2fod output and next steps

,
1

Hello, I really appreciate the amount of help available on this forum, thank you!
I just started using mrtrix3 and was hoping someone could give feedback on below processing steps and dwi2fod output

The rawdata (dwi_raw.mif) has resolution 1x1x2, 66 volumes, with 10 b0s and 56-dirs with b=1000;
below is what I have done so far-

  1. dwidenoise dwi_raw.mif dwi_deno.mif -noise noise.mif

  2. mrdegibbs dwi_deno.mif dwi_deno_unr.mif -axes 0,1

  3. dwifslpreproc dwi_deno_unr.mif dwi_den_unr_preproc.mif —nocleanup rpe_none -pe_dir AP -eddy_options " --slm=linear" -fslgrad bvec bval

  4. dwibiascorrect ants dwi_den_unr_preproc.mif dwi_den_unr_preproc_unbiased.mif -bias bias.mif

5a) mrconvert dwi_den_unr_preproc_unbiased.mif dwi_den_unr_preproc_unbiased.nii

b) bet2 dwi_den_unr_preproc_unbiased.nii dwi_den_unr_preproc_unbiased_masked.nii -m -f 0.7

c) mrconvert dwi_den_unr_preproc_unbiased_masked.nii mask.mif

  1. dwi2response tournier dwi_den_unr_preproc_unbiased.mif wm_response_tournier.txt

  2. dwi2fod csd dwi_den_unr_preproc_unbiased.mif wm_response_tournier.txt fod.mif -mask mask

The output looks like below-

coronal sag axial

First question is if I should modify any steps from 1-7 to improve this output?

Secondly my goal is do prob tractography and compute connectogram using a white matter atlas; is there a webpage that details the steps? I found the https://osf.io/pm9ba/ page and https://andysbrainbook.readthedocs.io/en/latest/MRtrix/MRtrix_Introduction.html, but the former is for connectome multi-shell data, and the latter explains with freesurfer outputs; I would like to try an atlas based method.

Thanks so much!

2

Welcome @veeus18!

The main variation your listed steps have from the current recommendations are the response function estimation and FOD calculation algorithm. The dhollander algorithm is generally recommended for response function estimation. "dwi2fod csd invokes the original CSD algorithm, which involves a “soft” non-negativity constraint (negative values are penalised) rather than the “hard” non-negativity constraint (negative values are forbidden) of the MSMT CSD algorithm. Despite the naming, the “multi-shell” algorithm can still be used with what’s considered “single-shell” data: with two unique b-values the data can be decomposed into two tissues, typically WM and CSF.

As far as connectome construction using a volume-based atlas, take a look at my BIDS App and see if your target parcellation is already supported there. I’ve not yet been able to finalise and release a decent volume of changes, so it’s a little out of date, but new parcellations are not too difficult to add. Otherwise you can check out the relevant wiki post which explains the underlying logic & steps.

Cheers
Rob