ndcn / infer-subc Goto Github PK

The bulk of the documentation is okay. This thread will enumerate

• gaps
• simplifications to be made
• clarifications
• reorganization

Initial todo:

• version releases. git tags engaging publish to pypi
• environment setup / dependencies. including Windows, Mac, Linux caveats
• connections to organelle-segmenter-plugin
• ??

Current "output" tiffs leave the metadata behind

Need to
[ ] . dump metadata to output for completeness. e.g. source file-name, channel origin. infer_subc version, etc.
[ ]. add a top-level summary table when processing image for stats containing metadata

QC logging.
We need to fail nicely when the segmentations fail. We will do this by collecting the regionprops from the segmented objects and comparing with prior expection.

• regionprops. this should be straighforward. at first its just "area" and numbers of organelles, and "total" areas.
• priors. need a class wrapper or something to contain the priors for the

This repo was started from a template and has some leftover crud in it.

Namely:

• CLI stubs
• Makefile
• actions
• etc.

Priority include:

• static documentation site
• pypi deployment

the paradigm and tooling that aics-segmentation used for managing workflows (specifically in a Napari

** step 1 **
make an example notebook that leverages aicssegmentation.Workflows to define/encapsulate our organelle segmentation, and correctly execute

** step 2 **.
Use it to instantiate a napari plugin to interactively develop single workflows and

** step 3 **
BUild "batch processing" of the developed workflows.

QUESTIONS

• Will we need to make our own Workflows tools, as many it contains default workflows and thumbnails specific to aics-segmentation?
• Rather, might we subclass and duck our own defaults?
• bridge to Dask graph representation (e.g. napari-workflows)

Need to add/update links in the README.

• Add in link for scipy
• Add in link for napari
• Change notebook template link from link to: ./notebooks
• Change infer golgi complex link from ./notebooks/06_golgi.ipynb to ./notebooks/06_infer_golgi.ipynb
• Change infer peroxisomes link from ./notebooks/07_peroxisome.ipynb to ./notebooks/07_infer_peroxisome.ipynb
• Change infer endoplasmic reticulum link from ./notebooks/08_endoplasmic_reticulum.ipynb to ./notebooks/08_infer_endoplasmic_reticulum.ipynb
• Change infer lb link from ./notebooks/09_lipid_bodies.ipynb to ./notebooks/09_infer_lipid_body.ipynb

noteooks maintenance is tricky. the metadata/output cells change constantly.

https://github.com/jupyter/nbdime
https://nbdev.fast.ai/

What is the correct order of the steps in the analysis, or does it matter? The documentation and linked output seem a bit confusing.

@ergonyc or @shanrhoads -- you have a much better handle on this than I do.

The current workflows need to fail nicely and report metrics.
If any step of the organelle segmentaion workflow fails we need to know and flag it for followup "by hand"

proposed solution

• try/except wrappers
• logging
• leverage aics-segmentation code

Is your feature request related to a problem? Please describe.
Currently the code can write "ome-tiff"s but they do not include the FULL meta-data including the microscope meta data which is improperly imported from the .czi file.

Describe the solution you'd like
The tiff format should be able to encapsulate ALL the metadata in an OME format... but need to investigate which are the proper fields in the OME schema to do things.

SPECIFICALLY: -> which Z-slice are we setting to "optimal" for downstream... this is crucial