WIP - NF - Tracking with Initial Directions and other tracking parameters

[gabknight]

WIP - it needs a first review by @StongeEtienne and @jchoude. Does this will do for SET?

The aim of this PR is to add three new optional parameters to LocalTracking:

• initial_directions : specify the initial direction for tractography to follow at the initial seed position. If initial_directions is None (default), LocalTracking calls DirectionGetter.initial_direction(seed_position) to get the initial directions (current behavior). This functions return the list of peak at the seed position. In a similar fashion, initial_directions contains a list of directions for every seed positions.
  • The tracking works with directions that correspond to vertices on the sphere. Thus, I had to select the vertice the closest to the input direction instead of the actual input direction. i.e. the closest sphere vertice is used as initial direction instead of the given initial directions.
• unidirectional: perform the tractography in a single direction from the seeding position, instead of both (default). This is especially useful in conjunction with initial_direction, where, for instance, those initial directions could be pointing inward from the cortical surface or outward from subcortical regions.
• randomized_forward_direction: Useful in conjunction with unidirectional, this parameter allows the tracking to start randomly either in the initial direction or in its opposite direction. If undirectional is False, this parameter will only reverse the points order of some streamlines.

[pep8speaks]

Hello @gabknight, Thank you for updating !

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

Comment last updated on January 17, 2019 at 13:39 Hours UTC

[codecov-io]

Codecov Report

Merging #1627 into master will increase coverage by 3.29%.
The diff coverage is 98.69%.

@@            Coverage Diff             @@
##           master    #1627      +/-   ##
==========================================
+ Coverage   87.32%   90.62%   +3.29%     
==========================================
  Files         245      234      -11     
  Lines       32148    28441    -3707     
  Branches     3493     3018     -475     
==========================================
- Hits        28073    25774    -2299     
+ Misses       3242     2049    -1193     
+ Partials      833      618     -215

Impacted Files	Coverage Δ
dipy/tracking/local/localtracking.py	98.27% <100%> (+0.52%)	⬆️
dipy/tracking/local/tests/test_tracking.py	97.07% <98.26%> (+1.48%)	⬆️
dipy/testing/decorators.py	64.51% <0%> (-29.04%)	⬇️
dipy/workflows/segment.py	72.44% <0%> (-23.21%)	⬇️
dipy/workflows/tests/workflow_tests_utils.py	83.33% <0%> (-16.67%)	⬇️
dipy/data/fetcher.py	33.24% <0%> (-12.11%)	⬇️
dipy/testing/__init__.py	72.72% <0%> (-9.1%)	⬇️
dipy/data/__init__.py	83.33% <0%> (-8.83%)	⬇️
dipy/utils/optpkg.py	64.28% <0%> (-7.15%)	⬇️
dipy/io/vtk.py	14.58% <0%> (-6.58%)	⬇️
... and 78 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update a32e164...b48139b. Read the comment docs.

[gabknight]

@StongeEtienne, do you think those tracking parameters will do for your future mesh seeding integration?

[StongeEtienne]

@gabknight I will soon test them, and see if there is things to be changed/improved,
I might work on surfaces integration in Dipy, at Dipy workshop https://workshop.dipy.org/

[guaje]

@gabknight I was checking your code and I realized you are importing HemiSphere but you are not using it. Can you please check that?

[skoudoro]

What is the status of this PR @gabknight @StongeEtienne?

[gabknight]

This needs a bit of work and not required for the next release. let keep this for Dipy 1.1

[skoudoro]

Ok Sounds like plan. Thank you for the feedback.

[gabknight]

Closed in favor of PRs #2821 and #2825.