Provide a brainmask as an input argument

[jhlegarreta]

According to the docs: https://qsiprep.readthedocs.io/en/latest/usage.html

users cannot provide a brainmask to the qsiprep steps. I am wondering whether maintainers would see any potential benefit of this, or whether adding such an option could be considered in the mid term.

[smeisler]

Hi @jhlegarreta, what would be the use case you have in mind with this option? I believe the brain mask is calculated on the preprocessed DWI, so I am not sure what masking would influence during preprocessing. Perhaps something like that could be useful for postprocessing, when it would make sense to limit processes to a mask. But, in that case, you can just create your own mask and name it the same way QSIPrep does, and QSIRecon will use it, so probably not worth making an input argument for that.

[jhlegarreta]

Hi @smeisler.

what would be the use case you have in mind with this option? I believe the brain mask is calculated on the preprocessed DWI, so I am not sure what masking would influence during preprocessing.

OK, can this (whether it is used or not, and how it is computed) be please confirmed and documented @mattcieslak ?

Perhaps something like that could be useful for postprocessing, when it would make sense to limit processes to a mask.

That is the use case, but I ignore whether the computed mask is involved in qsiprep's preprocessing, though. e.g. FSL's eddy says that the brain tissue needs to be masked ("Running eddy" section).

But, in that case, you can just create your own mask and name it the same way QSIPrep does, and QSIRecon will use it, so probably not worth making an input argument for that.

I am not using qsirecon: if the mask is being used in the preprocessing, and if we see that the one computed and used by qsiprep is not accurate enough, the question is whether we would be able to compute the mask with another tool, and feed that to qsiprep.