Apply equalization to suite2p motion projection images

[wbwakeman]

From #152, compute histogram to determine equalization to apply to projection images generated by suite2p motion correction

Tasks:

• [ ] Find settings that minimize difference between suite2p motion images and current production cell segmentation images
• [ ] Apply those settings in ophys_etl_registration pipeline
• [ ] Decision point: implement these or do not (in the case that the outputs are very far from our expectations)

Validation criteria:

• [ ] Decision has been made whether to include this equalization or not
• [ ] Output some images
• [ ] Demonstrate code that describes what differences were implemented

[njmei]

Notes about where max and avg image types are saved and loaded from LIMS:

Sources:

ophys_motion_correction_strategy:

• max_downsample_4Hz_0.png max projection produced by production motion correction saved with ophys_experiment.maximum_intensity_projection_image

suite2p_motion_correction_strategy:

• max_downsample_4Hz_0.png max projection produced by suite2p motion correction saved with ophys_experiment.maximum_intensity_projection_image
• <experiment_id>_average_projection.png average projection produced by suite2p motion correction attached to ophys_experiment as 'OphysAverageIntensityProjectionImage' WKF

ophys_cell_segmentation_strategy:

• avgInt_a1x.png avg projection produced by production segmentation attached to ophys_cell_segmentation_run as 'OphysAverageIntensityProjectionImage' WKF
• maxInt_a13a.png max projection produced by production segmentation attached to ophys_cell_segmentation_run as 'OphysMaxIntImage' WKF

Consumers:

behavior_ophys_write_nwb_strategy:

• max_projection_file consumes ophys_experiment.maximum_intensity_projection_image
• average_intensity_projection_image_file consumes 'OphysAverageIntensityProjectionImage' WKF attached to ophys_experiment

ophys_nway_cell_matching_strategy:

• (Will take precedence) ophys_average_intensity_projection_image consumes 'OphysAverageIntensityProjectionImage' WKF attached to ophys_cell_segmentation_run
• (If not attached to the ophys_cell_segmentation_run) ophys_average_intensity_projection_image consumes 'OphysAverageIntensityProjectionImage' WKF attached to ophys_experiment

[njmei]

@djkapner For average images it looks to me like the new avg projections look better than the old ones...

Though, as you pointed out, if an 'OphysAverageIntensityProjectionImage' is attached to the ophys_cell_segmentation_run that appears to be preferred when putting together the inputs for the nway strategy.

[njmei]

Max projections looks completely different from one another, the new max projections look a lot noisier...

[djkapner]

Can you compare max projections on some mesoscope experiments? In your last examples, I see what looks like dewarping artifacts. With mesoscope examples, we can at least eliminate the possibility that dewarping has broken since those first ones were created.

[njmei]

@djkapner You might be on to something, I don't see those banding artifacts for the mesoscope max projections:

[wbwakeman]

FWIW, I am dewarping experiments 918566488,1008737608, 1063089965 with Taylor's updated dewarping code from PR #131 for issue #122 .

[njmei]

So yeah... looks like those bands feature quite prominently in the dewarped image:

Looks like this was not detected because the old production segmentation applied a gaussian filter when creating its "max intensity image".

[wbwakeman]

Just to complete the thought...here are the max projection images of the three Scientifica experiments processed through the refactored dewarping code and then suite2p motion correction

[njmei]

Final conclusions:

• Dewarped ophys movies appear to have some vertical stripe artifacts.
• The production (Leonard's) cell segmentation appears to be doing some gaussian filtering to the max and avg projection images.
• Max and Avg projection images used in the behavior_ophys_write_nwb module should be from the cell segmentation module.

LIMS ticket for updating behavior_ophys_write_nwb_strategy: http://jira.corp.alleninstitute.org/browse/PBS-2618