search

WayScience / demo-cellprofiler

Demonstrating various CellProfiler development or implementations.
BSD 3-Clause "New" or "Revised" License
0 stars 1 forks source link

Adjust CellProfiler dataset reference ExampleHuman pipeline to improve image displays #5

Open d33bs opened 2 weeks ago

d33bs commented 2 weeks ago

I am not sure this is exactly how I envisioned it, but this is interesting!

From my perspective after looking at the example image, I find that overlaying all of these objects over the same image to be a bit too busy. As well, you have cell segmentations on the nucleus channel, where you won't be able to identify if the segmentation looks correct.

I would recommend that we focus on collecting two overlay images to represent the nuclei and the cells segmentation (cytoplasm isn't as important since it is just the product of the first two segmentations).

  • The nuclei overlay would using the DNA/nuclei channel and overlaying Nuclei.
  • The cells overlay would use the channel that was used for segmenting whole cells, and overlaying Cells on it.

These two are the only things I use. Another option, if you want to keep everything together in one overlay, is that you will need to merge/composite the two channels used for the segmentation of nuclei and cells and then add all of these object overlays on the merge channel image.

_Originally posted by @jenna-tomkinson in https://github.com/WayScience/demo-cellprofiler/pull/4#discussion_r1808970197_

d33bs commented 2 weeks ago

Also in https://github.com/WayScience/demo-cellprofiler/pull/4#discussion_r1808977744:

This image is a bit busy. I am understand the nuclei segmentations on this nuclei channel, but the whole cell segmentations do not help us since we can't tell if the segmentations are correct. Also, I don't know what the yellow represents, looks like also nuclei segmentations?

I make recommendations in the cppipe file. This looks super pretty and cool, but for the context of QC, I don't think just having this overlay will be enough.

d33bs commented 2 weeks ago

Also in https://github.com/WayScience/demo-cellprofiler/pull/4#discussion_r1808985231:

I make a comment in the cppipe file that I didn't think that the MaskImage module is used, but I think that is incorrect based on the folder name of this file.

To me, this doesn't look like a mask, but I could be misinterpreting. Also, it says MaskDNACells which to me makes it sound like this image should include both nuclei and cells segmentation, but this just looks like the nuclei channel raw image.

For all of these image_0.png images, I think that these should be double-checked since it looks like these are coming from the SQLite file but don't represent the correct output.