Fernando Pérez-García

> @fepegar I'm happy with what ever you have time for +1 Ok, I'll open the PR and ping you. > I can split it of into a separate feature...

> I might give the ElasticDeformation a short look but my time is currently also very limited. Good luck! [The docs](https://torchio.readthedocs.io/transforms/augmentation.html#randomelasticdeformation) should help understand how the deformation field is generated...

Patch for `RandomAffine` available in `v0.18.64`.

Hi both. I just want to say that I really want to look at this issue and corresponding PR, but I want to properly dedicate time to it and these...

So it seems that you're not yet convinced with your implementation. If you'd still like to have this in the package, we could add a `contrib` submodule and put it...

Yeah this is related to what is and isn't copied during a transform. It is indeed tricky. The problem is that deep-copying can be long if the tensor is large,...

[I edited your code a bit to understand it more easily, sorry]

> I can understand the different choice for LabelMap but why then a different behaviour between RandomBlur and RandomAffine ? This was not on purpose. It's just that label maps...

> I see, it goes through transform.py __call function, and there a copy is made. Yes, but it's a shallow copy: the array is the same. I'll try to measure...

The final k-space / spectrum is filled from the others always using the last dimension: https://github.com/fepegar/torchio/blob/eddd02decf2f00777429452e6056fdf72674d8d3/torchio/transforms/augmentation/intensity/random_motion.py#L271-L274 It would be nice if this dimension was selected randomly.