Drop support for native-to-standard space transforms

[tsalo]

Summary

This stems from Brainhack-Donostia/ica-aroma-org#41 (comment). We will drop support for native-to-standard space transforms for now, and all native-space data must be accompanied with tissue probability maps for now.

Next Steps

1. Remove all transforms.
2. Drop FEAT directory support too, maybe?

[tsalo]

I think the key problem here is that nitransforms doesn't support nonlinear warps. Or at least I can't find any information on warps in their documentation.

[CesarCaballeroGaudes]

Can we ask @oesteban?

[oesteban]

I think the key problem here is that nitransforms doesn't support nonlinear warps. Or at least I can't find any information on warps in their documentation.

Nonlinear warps from FSL?

• ANTs/ITK displacements fields are supported (some minor tweaks loading some kinds of h5 files might be necessary, but everything is doable ATM).
• FSL will require taking ENH: I/O of FSL displacements fields nipy/nitransforms#51 to the endline (mini-BHD project?)
• AFNI is not yet supported, although considering the consistency of their implementation I don't think it will be difficult once we pin down ANTs and FSL.

[tsalo]

@oesteban I think FSL is the most likely source of transforms (since ICA-AROMA is designed for FSL-processed data), but we are hoping to support data coming from anywhere. It doesn't sound like that's possible yet, though.

We could try to support FSL if and when nipy/nitransforms#51 is merged, though, and add support for each package as it is incorporated.

[oesteban]

@oesteban I think FSL is the most likely source of transforms (since ICA-AROMA is designed for FSL-processed data), but we are hoping to support data coming from anywhere. It doesn't sound like that's possible yet, though.

Not for fMRIPrep - I know this is not the only pipeline capable of running ICA-AROMA but probably the best test-bed for a standalone implementation of AROMA.

For the time being, c3d converts FSL distortion maps into ITK's.

[eurunuela]

Okay, so shall we drop all these dependencies and use probability maps until poldracklab/nitransforms#51 is merged?

[oesteban]

I think that was @tsalo's original proposal - do not bother at this point about how the TPMs are generated in the native space, just assume they are defined there.

[eurunuela]

I think that was @tsalo's original proposal - do not bother at this point about how the TPMs are generated in the native space, just assume they are defined there.

Yes. I wanted to make sure. I was starting to get lost in the conversation having now two different repos after the Brainhack Donostia tutorial 😅