DEGA-290-custom-segmentation-landscape-files-fix

[jaspreetishar]

This PR includes a few minor changes across the pre and qc modules:

1. In pre/init.py, fixed the order of function calls in add_custom_segmentation.
2. Took this opportunity to make some small fixes in the qc module as well.
3. In pre/boundary_tile.py, updated the logic to retain cell geometries in micron space instead of overriding them. This is necessary to compute average cell area in the qc module.

The changes have been tested by successfully running the following notebooks:

1. BNB_Streamlined_Pre_processing_Xenium_V1_Human_Pancreas_FFPE.ipynb
2. Custom_Segmentation.ipynb
3. Segmentation_QC.ipynb

[cornhundred]

Looks good. I just asked for a bit of clarification. Can you also merge in the changes from main?

It would be great to turn this notebook, Xenium_V1_human_Pancreas_FFPE_outs.zip, into an automated test but the file sizes are too large.

[cornhundred]

Thanks for the changes, I just requested a small change to one of the files. You can add some unit tests if you think it makes sense otherwise you can merge.

[huanlity]

Great work! I will look in more detail later, but I have successfully verified the three notebooks with some adjustments.

The Segmentation_QC.ipynb threw an error:
FileNotFoundError: [Errno 2] No such file or directory: 'data/raw/Xenium_V1_human_Pancreas_FFPE_outs/micron_to_image_transform.csv'

I believe this file is generated in pre and saved in the landscape folder and not the raw data folder. If you need it to qc raw data without the need to pre-process. You can simply add this line to your notebook:

dega.pre.write_xenium_transform(base_path, base_path)

[huanlity]

Another required file is missing: segmentation_parameters.json in the raw data folder. To avoid manually moving files around, you can create it in the QC notebook instantly or even write a quick function to do it if you see it being used frequently in the future.

This is the content in the json:
{ "technology": "Xenium", "segmentation_approach": "default", "dataset_name": "xenium_pancreas" }

I suggest something like this: def write_default_seg_json(technology, dataset)

[huanlity]

Underscore this helper function or remove the function and just implement it in the code as a one-liner.

celldega/src/celldega/qc/__init__.py

Line 15 in 8dcdcbb

def get_largest_polygon(geometry): # TODO: underscore

[huanlity]

@jaspreetishar All three notebooks worked on my end. One final question: is this Unknown technology expected?

segmentation_parameters.json does not exist, creating one...
Unknown technology: None

[huanlity]

LGTM ✅

[Copilot]

Pull Request Overview

This PR implements fixes across pre-processing and quality control modules to address custom segmentation landscape files issues. The changes ensure proper handling of cell geometries in micron space and correct function call ordering.

• Fixed function call ordering in add_custom_segmentation to ensure proper gene metadata handling
• Updated cell geometry handling to retain micron space coordinates for accurate QC calculations
• Improved parameter handling and file path resolution across QC and pre-processing modules

Reviewed Changes

Copilot reviewed 4 out of 5 changed files in this pull request and generated 5 comments.

File	Description
src/celldega/qc/init.py	Enhanced QC module with better parameter handling, geometry processing, and data visualization improvements
src/celldega/pre/run_pre_processing.py	Added transformation matrix copying for MERSCOPE and fixed cluster creation logic
src/celldega/pre/boundary_tile.py	Updated to retain cell geometries in micron space and improved path handling
src/celldega/pre/init.py	Fixed function call ordering in custom segmentation and improved DZI file handling