Default algorithm for dwi2response in 0.3.12?

1

I’m trying to process some data to match an older dataset we have that was generated using 0.3.12. In that original pipeline, the dwi2response command was:

dwi2response ${diffdir}/data.nii.gz ./CSD10_b3000_response.txt -lmax 10 -fslgrad ${diffdir}/bvecs ${diffdir}/bvals -mask ./eromask.nii.gz -sf ./sf_mask.nii.gz -nthreads 8

When I try to run this on the latest mrtrix, I need to specify an algorithm, and I can’t figure out which algorithm is equivalent to whatever was used in that older version. There doesn’t appear to be documentation up for that version either. I don’t see the -sf flag in any of the new docs. Is that equivalent to -voxels now?

2

That would have used the tax algorithm – as originally proposed in this paper. It’s still available in more recent versions of MRtrix3, but there are differences in the implementation so the output is unlikely to be exactly identical – so I definitely wouldn’t combine data analysed with 0.3.12 and more recent versions in the same analysis.

We decided to move away from that approach to the dwi2reponse tournier algorithm after quite a bit of discussion. Nowadays the recommended approach is the dwi2response dhollander algorithm, which seems to perform well both single and multi-shell data.

3

There doesn’t appear to be documentation up for that version either. I don’t see the -sf flag in any of the new docs. Is that equivalent to -voxels now?

Correct. The interface is different in part because the implementation has been re-done from scratch: back in 0.3.12, dwi2response was a C++ binary command, whereas now the dwi2response wrapper and all of the individual algorithms are written in Python.