Produce mrtrix (grad) and FSL (bval/bvec) style output files for DWI data

[gdevenyi]

The current outputs are nice, but most people's first step is probably to feed into a program that expects certain formats.

For mrtrix you just add the bvals as another column after the vectors.
https://mrtrix.readthedocs.io/en/latest/concepts/dw_scheme.html

For FSL, it uses the insane format:
bvecs

x1 x2 x3 ... xn
y1 y2 y3 ... yn
z1 z2 z3 ... zn

bvals

b1 b2 b3 b4 b5 ... bn

https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/FDT/UserGuide

[SebastianoF]

Thanks for raising this issue.

No matter the efforts, flags and options, an interface between the converted data and the subsequent intended analysis that the user would like to perform will always be needed.

As you said every software for the subsequent analysis has different requirements.

For that reason the converter stops one step before data analysis, and tries to explain how the data were converted: if I start customise for particular subsequent analysis, the problem is when do I stop?

You can load the b-vals and b-vect and then transpose or concatenate according to your needs. You can do it both with Matlab (loading the .csv) or with python, loading the .npy.

Hope this help. Feel free to re-open the issue for further on this thread.

[gdevenyi]

Thanks, I know how to produce and handle what I need, just suggesting the smoothest way possible less-than-technical users.

[neurolabusc]

@SebastianoF I just converted the DWI sample from bruker2nifti_qa and it creates the text files Lego_2_DwGradVec.txt and Lego_2_DwEffBval.txt. Your message above suggests that by simply transposing the row/columns the DwGradVec.txt will be in FSL format (as described here and here). I know that NIfTI uses different spatial coordinates than DICOM, but I am not sure about the raw Bruker data. I just want to make sure the vectors have been rotated to match the expectations of the popular FSL tools.

[SebastianoF]

In theory vectors have been rotated as expected, so you should have only to feed fsl with no permutations. In practice there are still some open issues on orientation that I did not manage to solve (see issues #11, #33), and some permutation of the b-vects columns may be required. To find confirmation, look at the anatomical regions that are expected to have a predominant direction (corpus callosum), and see if they match the expectations.