I am trying to use MRTRIX to calculate a kind of complicated case of FODs.
Is there any way to use the dwi2responce function based completely on the measured data (for example dwi.mif) and selected voxels for the kernel calulation, which does not include tensor calculation by MRTRIX?
I would assume that -in theory- the kernel can be calculated without a-priori knowledge of directionality. In case this is not possible, is the -dirs parameter (of the dwi2reponse manual) working with providing the tensor for each voxel in FSL form (eg. a volume of size [x z y 6], where 6 the tensor components)?

I have a feeling the -dirs option to dwi2response manual is what you’re after? You can provide the directions yourself if you have better estimates or simply prefer to avoid a tensor calculation.

Sorry, missed the fact you’d already spotted that option. But just to clarify: the -dirs option expects a 4D, 3 volume image of XYZ components for the directions vectors for each voxel – not tensor components.

Maybe… Some of us have looked into that over the years, but I don’t think any of us have come up with an algorithm that performs well in practice – certainly not as well as you can achieve when you know the directions…

But it remains the case that all of the approaches in dwi2reponse (manual included) currently rely on knowledge of the directions.

Dr. Tournier, Thanks a lot for the inputs.
In that case I will try to use the external tensor and produce directions using the tensor2metric funtion for the voxels of interest.