Deface raw T1 image

i am trying to set up a bids-compatible data structure. As I understand, bids currently is primarily geared to raw data (unprocessed) - and that is fine since I want to i) have a clean archive and ii) later share it.
My question: I want to deface the T1s right away, but using the tool cut off part of the brain. After I normalized the T1 into MNI space, it was fine. However, I want to do deface raw data - so normalizing is not good. I could try to just reorient them slightly to coarsely match, but before I spend too much time I thought perhaps sb who has done this before has a recommendation.
Thanks and best, Ralf.

PS: sidequestion: how can I add tags here? It wouldln’d accept the “deface” in the field below.

pydeface should work on raw data without the need to do any prior spatial transformations. If it doesn’t I would recommend opening an issue at

One thing you can try is doing affine transformation to mni, estimate the face mask there, apply inverse affine transform to the face mask, apply the transformed mask in the native space.

thank you! I opened an issue. Closing this here.
Best, Ralf.