Implement create_slice_grid by constantinpape · Pull Request #68 · mobie/mobie-utils-python

constantinpape · 2022-05-24T15:51:48Z

Fixes #67

Have to leave now, will finish this later today.

martinschorb · 2022-05-24T18:29:33Z

mobie/metadata/slice_grid_view.py

+    if source_type != "imageSource":
+        raise ValueError("create_slice_grid is only supported for image sources.")
+
+    shape = _load_shape(source_data["imageData"])


Is this intended to happen in original voxel space or in physical space?
My idea was to slice along physical axes (making it compatible also with any kind of transformed sources) and simply introduce translations by source.shape[axis]/ n_slices. The case of a source only being scaled to physical space and slicing happening in a source axis would then be a subset of that general slicing operation.

I don't quite understand what you mean. But scaling must happen in physical space, since we only get the transformed version of the source here. See line 73 and following for how the transformation is computed (and I haven't double checked if this is correct yet, since I first need to implement the infrastructure to read the shape and transformation from the source metadata (i.e. bdv.xml or ome.zarr), which might be on s3)

Yes, makes total sense. My thinking was immersed in what the MoBIE viewer would do (that should know the bounds of the source); I completely forgot that we are in a separate code world here 😆

Obviously the shape we will build on is voxel-based an then all transformations will apply (not just scaling since all rotations and shear will also mess up the source shape).

Sorry, too much MoBIE-usage in the last day...

:) No worries

So we will need to apply all existing transformations to the "shape" vector. The slicing step would then be np.dot(T_all,shape)[axis]/n.

I think applying the slicing function to a single ("mother") source would be perfectly fine. You could simply add the other sources to the grid view and its transforms later once they are generated.

This makes perfect sense, but I am not sure we need the axis parameter then. Wouldn't you always slice along the z axis? (w.r.t. the initial transformation of course)

Slicing across another axis would just mean rotate it s.t. this axis becomes z

Should I add another parameter to the function so that you can pass a list of transformations that's applied before the slicing?

Maybe you could even reference an existing view and "import" the sourceTransforms from there?

This makes perfect sense, but I am not sure we need the axis parameter then. Wouldn't you always slice along the z axis? (w.r.t. the initial transformation of course)

Yes, you would then add the extra rotation in case slicing another axis. Otherwise the grid will also not be in the xy plane. Probably referencing an existing view (that users can generate in MoBIE and export) would be indeed ideal here (and then we can ditch the axis parameter). The users probably don't care how the transforms look as long as the orientation matches with what they like to see.

Maybe you could even reference an existing view and "import" the sourceTransforms from there?

Ok, but I will create two functions for this, one that takes initial transforms and one that takes a reference view.
(Having both options in the same function would require some handling of precedence if both are given, and I want to avoid that)
(And the one that takes a reference view just calls the one with initial transforms internally)

constantinpape

@martinschorb I refactored this a bit now and added two functions (one with, the other without reference view). I will call it a day now and tackle the actual implementation tomorrow.

constantinpape · 2022-05-24T20:06:43Z

mobie/metadata/slice_grid_view.py

+    write_dataset_metadata(dataset_folder, dataset_metadata)
+
+
+def create_slice_grid(


This is the function for creating a slice grid without passing an initial reference view.

constantinpape · 2022-05-24T20:07:04Z

mobie/metadata/slice_grid_view.py

+    _write_view(dataset_folder, dataset_metadata, view_file, view_name, sliced_view, overwrite)
+
+
+def create_slice_grid_with_reference_view(


This is the function for creating a slice grid from an initial reference view

constantinpape · 2022-05-25T08:59:52Z

Hey @martinschorb,
I have a first implementation now, but it is completely untested, and since there are many different code paths writing all tests and debugging will take some time and I can't do all of that today.
Can you please let me know for which source you need this exactly for the paper figure? Than I will make it work for that one today and cover the rest in the next few days.

martinschorb · 2022-05-25T09:06:25Z

Can you please let me know for which source you need this exactly for the paper figure?

that's tomo_38_hm. You could also leave out all transforms.
Alternatively we could do tomo38_lm including the fluorescence overlay. Then we would need the transforms from Figure2d.

constantinpape · 2022-05-25T09:07:45Z

that's tomo_38_hm. You could also leave out all transforms.
Alternatively we could do tomo38_lm including the fluorescence overlay. Then we would need the transforms from Figure2d.

Ok, I will give it a try after lunch and ping you once I have something.

constantinpape · 2022-05-25T11:39:28Z

Hey @martinschorb,
I have a first version that runs through and added a test view under slice-grid->slice-grid-tomo38 (on main directly).
The grid placement works, but the transformation I estimate seems to be much too small, and it shows each tomo at the same z position. I have to stop until 4 PM now, gonna have another look then.

constantinpape · 2022-05-31T21:19:34Z

Fyi @martinschorb I will probably merge this tomorrow without finishing the slice grid implementation, depending on whether we need it for the figure (might choose the continuous representation of the z-axis instead).

I still like the feature idea a lot and will implement it eventually, but I don't really have enough time right now to sit down and figure out all the transformation logic...
And I did a lot of refactoring here that I want to release asap because it significantly improves the package structure.

…splay tables

Implement create_slice_grid WIP

bc22f82

martinschorb reviewed May 24, 2022

View reviewed changes

Refactor create_slice_grid functionality WIP

d247785

constantinpape commented May 24, 2022

View reviewed changes

Complete implementation of create_slice_grid (untested)

0f33325

Fix several issues in create_slice_grid

53d3d0a

constantinpape mentioned this pull request May 26, 2022

Towards v0.3 #69

Closed

5 tasks

constantinpape added 9 commits May 28, 2022 11:11

Refactor source metadata functionality

8e12747

More refactoring

f46cf6f

Refactor view functionality

eb45e0f

Update example notebook

10f4cef

Refactor slice grid view

2b55d3a

Update example notebooks WIP

72621fe

Update citation info

5752b0c

Start working on htm project creation example

361a8fa

Move slice grid view functionality to experimental

bece248

constantinpape marked this pull request as ready for review May 31, 2022 21:17

constantinpape added 8 commits June 1, 2022 21:59

Work on htm example notebook

0b83b19

Rename source_annotation to region also in code

495aa3d

Check that columns given in colorByColumn actually exist for regionDi…

a1afdd7

…splay tables

Check colorByColumn for segmentationDisplays

17f7b09

More renaming of annotation to region

89a6c55

Add util functionality for computing clims for grid views

61c3ab5

Update htm example

023760f

Update readmes

3ed5992

constantinpape merged commit d3fef60 into master Jun 2, 2022

constantinpape deleted the slice-grid branch June 2, 2022 21:39

This was referenced Jun 2, 2022

feature request: slice_volume #67

Open

missing check, view validation #70

Closed

		write_dataset_metadata(dataset_folder, dataset_metadata)


		def create_slice_grid(

		_write_view(dataset_folder, dataset_metadata, view_file, view_name, sliced_view, overwrite)


		def create_slice_grid_with_reference_view(

Conversation

constantinpape commented May 24, 2022

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

constantinpape May 24, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

constantinpape left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

constantinpape commented May 25, 2022

Uh oh!

martinschorb commented May 25, 2022

Uh oh!

constantinpape commented May 25, 2022

Uh oh!

constantinpape commented May 25, 2022

Uh oh!

constantinpape commented May 31, 2022

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

constantinpape May 24, 2022 •

edited

Loading