Simulate rotation out of imaging plane with the Rotator

[lauraporta]

Description

What is this PR

• Bug fix
• Addition of a new feature
• Other

Why is this PR needed?
While working on finding the centre of rotation (see #30) I noticed that the rotating cells describe an ellipse and not a circle. This kind of artefact could be due to a mismatch between the plane of rotation and the imaging plane.
Therefore, to prove it I expand the Rotator class to create a new type of synthetic rotation movies.

You can notice in the max projection image that the trajectory of the cells is elliptical as expected.

Works also for rotations out of plane in another orientation:

And with shifted center of rotation:

This new feature could be described also as a "3D rotation".

I can now use the new Rotator to generate synthetic movies to build an improved derotation algorithm. It will come in a separate PR.

What does this PR do?
Further methods to Rotation class and related tests.
Renaming of classes:

• basic_rotator -> line_scanning_microscope (sorry the diff shows as if it is a whole new class 😭)
• test_finding_center_of_rotation_by_joining_two_pipelines -> test_finding_center_of_rotation

Moves functions to fit an ellipse to a dedicated module (fit_ellipse.py)

References

See #26 #27

How has this PR been tested?

Made parametrised tests

Is this a breaking change?

No

Checklist:

• The code has been tested locally
• Tests have been added to cover all new functionality
• The documentation has been updated to reflect any changes
• The code has been formatted with pre-commit

[codecov]

Codecov Report

Attention: Patch coverage is 83.46457% with 21 lines in your changes missing coverage. Please review.

Project coverage is 47.26%. Comparing base (c88975e) to head (4129dac).
Report is 25 commits behind head on main.

Files with missing lines	Patch %	Lines
derotation/analysis/fit_ellipse.py	64.00%	18 Missing ⚠️
...tation/analysis/incremental_derotation_pipeline.py	62.50%	3 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #31      +/-   ##
==========================================
+ Coverage   43.89%   47.26%   +3.36%     
==========================================
  Files           6        7       +1     
  Lines         631      677      +46     
==========================================
+ Hits          277      320      +43     
- Misses        354      357       +3     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[IgorTatarnikov]

Looks great!

Just a few comments here and there. I could use some more information on the construction of the homography_matrix, why that specific form? From my understanding the current implementation adds a scaling factor in the second dimension based on cos(rotation_plane_angle), is that correct?

Might also be nice to add "noise/simulated activity" to the bright/dim cell objects. Add a flat random noise to simulate shot noise and then an underlying "ground truth" activity as well! Understand that this is outside of the scope of this PR.

[lauraporta]

I could use some more information on the construction of the homography_matrix, why that specific form? From my understanding the current implementation adds a scaling factor in the second dimension based on cos(rotation_plane_angle), is that correct?

Yes it is correct. Your comment makes me wonder if it is an overkill to use affine_transform and a 3x3 matrix given that the transformation is only on one of the dimensions... 🤔

[IgorTatarnikov]

Looks good 👍 LGTM!