Fix organoid segmentation #44
Conversation
* inital commit * update paths * preprocessing * run on alpine * change pathing * HPC * local update * addressing comments
|
Check out this pull request on See visual diffs & provide feedback on Jupyter Notebooks. Powered by ReviewNB |
There was a problem hiding this comment.
Pull Request Overview
This PR fixes organoid segmentation by adding a recursive segmentation function and updating hardcoded parameters for whole organoid segmentation. Key changes include:
- Renaming of hardcoded parameters such as well_fov and compartment in multiple scripts.
- Introduction of the recursive function segment_with_diameter with dynamic diameter reduction.
- Updates to notebook cell execution metadata and output messages.
Reviewed Changes
Copilot reviewed 11 out of 11 changed files in this pull request and generated 1 comment.
Show a summary per file
| File | Description |
|---|---|
| 2.segment_images/scripts/7.clean_up_segmentation.py | Updated cell indices in notebook run code. |
| 2.segment_images/scripts/4.reconstruct_3D_masks.py | Changed well_fov and compartment values to support organoid segmentation. |
| 2.segment_images/scripts/3.segmentation_decoupling.py | Modified well_fov and compartment values for organoid segmentation. |
| 2.segment_images/scripts/2.segment_whole_organoids.py | Added segment_with_diameter and updated segmentation flow and parameters. |
| 2.segment_images/scripts/0.segment_nuclei_organoids.py | Adjusted well_fov for proper path setup in notebook mode. |
| 2.segment_images/notebooks/7.clean_up_segmentation.ipynb | Updated execution counts and output messages for consistency. |
Comments suppressed due to low confidence (1)
2.segment_images/scripts/2.segment_whole_organoids.py:308
- [nitpick] The variable 'slice' shadows a built-in Python name. Consider renaming it (e.g., 'img_slice') to improve code clarity and prevent potential confusion.
for slice in tqdm.tqdm(range(imgs.shape[0])):
|
|
||
| if labels is None: | ||
| print(f"Labels are empty for diameter {diameter}. Trying smaller diameter...") | ||
| return segment_with_diameter(img, model, channels, z_axis, diameter - 250) |
There was a problem hiding this comment.
The recursive call passes parameters in an incorrect order; the function expects diameter as the third argument, so it should be called as segment_with_diameter(img, model, diameter - 250, z_axis, channels).
| return segment_with_diameter(img, model, channels, z_axis, diameter - 250) | |
| return segment_with_diameter(img, model, diameter - 250, z_axis, channels) |
Copilot uses AI. Check for mistakes.
| # In[2]: | ||
|
|
||
|
|
||
| def segment_with_diameter( |
There was a problem hiding this comment.
Loving the definition of this function - looks like it's useful. Consider adding a software test to confirm this works with something rudimentary, like a generated egg shape, or something similar. Generally, once a capability becomes firm enough to functionalize I recommend adding at least one test. We might theorize that this could be developed at a later point, but we may forget the context of which test confirmations are most important to the work.
| diameter: int, | ||
| z_axis: int = 0, | ||
| channels: tuple = [1, 0], | ||
| ): |
There was a problem hiding this comment.
Consider adding the return type here (it looks well documented below!).
| The labels of the segmented image. | ||
| details : dict | ||
| The details of the segmentation. | ||
| _ : None |
There was a problem hiding this comment.
Consider removing this value and relying instead on a Dict. This would allow you to expand the work through a type without getting caught in a specific number of return values.
| _ : None | ||
| Placeholder for additional return values. | ||
| """ | ||
| if diameter < 250: |
There was a problem hiding this comment.
Would it make sense to make the minimum diameter a parameter? I imagine this could change from dataset to dataset.
| ---------- | ||
| img : np.ndarray | ||
| The image to segment. Can be 3D in the format of (z, y, x). | ||
| model : models.Cellpose |
| # cyto1, # ER | ||
| cyto2, # AGP | ||
| # cyto3, # Mito |
There was a problem hiding this comment.
Consider renaming to channel_agp or something similar to help make this more specific. Also - why only this channel?
| squared_butterworth=True, | ||
| ) | ||
|
|
||
| # add a guassian blur to the image |
There was a problem hiding this comment.
Why do we add guassian blur to the image? Consider answering within the code comment. Also - could this impact the data integrity at all?
This PR adds critical functionality that scans an image for the largest object possible to segment across a dynamic range for whole organoid segmentation.