Transform points with DiffeormorphicMap

[skoudoro]

This PR is just a follow-up and close #936. Many people ask for this feature and we never merged it.

it adds 2 functions to apply warp fields on point: transform_points and transform_points_inverse. After registering two images, you can use this function to transform your streamlines/points.

You might be interested @vigji @wjy73 @dPys. Feel free to test this PR and give feedback/review. Thanks in advance!

fixes #2327
fixes #2313

TODO

• Create or Improves the current tutorial.
• Need to compare with deform_streamlines in dipy.tracking.streamline before merging. Performance should be better. Results should be similar.
• After comparison, remove duplicate? or make sure it uses the same codebase?

[pep8speaks]

Hello @skoudoro, Thank you for updating !

Cheers ! There are no PEP8 issues in this Pull Request. 🍻

Comment last updated at 2021-10-31 18:54:02 UTC

[arokem]

I think that we should go ahead and merge this, adding a documentation example on a separate PR.

[arokem]

@skoudoro : I believe the CI failures are unrelated (and probably fixed in #2413?)

Would you mind rebasing this one?

[skoudoro]

thank you @arokem. rebase done. I will try to write a tutorial this week. If I don't succeed, I will create an issue as a reminder and you will be able to merge this.

[codecov]

Codecov Report

Merging #2369 (4d9c9bc) into master (08b8993) will decrease coverage by 0.05%.
The diff coverage is 54.83%.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #2369      +/-   ##
==========================================
- Coverage   84.24%   84.19%   -0.06%     
==========================================
  Files         130      130              
  Lines       17919    17948      +29     
  Branches     3053     3059       +6     
==========================================
+ Hits        15096    15111      +15     
- Misses       2106     2118      +12     
- Partials      717      719       +2     

Impacted Files	Coverage Δ
dipy/align/imwarp.py	95.56% <53.33%> (-2.87%)	⬇️
dipy/tracking/streamline.py	93.80% <100.00%> (ø)

[arokem]

OK - looks like everything is passing now.

Looks like some of the new code is not currently tested. Do you want to use deform_streamlines to test it?

That could bring us a step closer towards unifying the code base across these two functions.

[alexrockhill]

How's this going? Looks like maybe ready to merge, happy to help write an example/documentation if that's of any use.

[skoudoro]

Hi @alexrockhill,

Thank you for pinging me on this PR.

The last step is to compare with deform_streamlines in dipy.tracking.streamline and to update the tests. Just need to find time to finalize this.

Concerning tutorials and documentation, it can be done on another PR. Thank you for proposing your help and feel free to do a PR over my branch in my fork.

[alexturcea]

Hello @skoudoro , is there any plan for merging this?

[skoudoro]

Hi @alexturcea,

I will have time this month to finalize this PR. So it will be on the release this month.

Hi @arokem,

I compared deform_streamlines and transform_points and we get the exact same results. Slightly better performance with transform_points. I added a test on 1a90b2b

if all CI's come back green, this is ready to go.

[skoudoro]

PS: I will update and create a tutorial on a new PR

[skoudoro]

I will wait until wednesday, and if no news, I will go ahead and merge this PR, is that ok @arokem and @Garyfallidis?

[arokem]

Yeah: 👍 from my end.

[Garyfallidis]

Nice!