Write tutorial for Cellpose (and maybe create workflow)

[kostrykin]

Cellpose is a popular tool for segmentation of cell nuclei and cytoplasm in fluorescence microscopy and histology images: https://usegalaxy.eu/root?tool_id=toolshed.g2.bx.psu.edu/repos/bgruening/cellpose/cellpose/3.0.8+galaxy0

There are two different variants for how this can be tackled.

---

1) The "insightful" variant. There are public dataset of suitable cellular images that are accompanied by hand-labeled ground truth segmentation data, which was created by human experts (manual segmentations). Examples of such datasets are given below. However, for those datasets there is no license information available. If we can (i) still use one of these datasets for the tutorial or (ii) find another such dataset with a permissive license, then we should create a tutorial which covers the following steps:

1. Loading the dataset (both the microscopy images and the ground truth images).
2. Using the Cellpose tool to segment the microscopy images.
3. Quantification of the similarity of the segmentation results with the ground truth images (i.e. assessing the segmentation performance of Cellpose on that dataset). This can be accomplished using the SegMetrics tool.

Candidate datesets:

• IEEE ISBI Cell Tracking Challenge, GFP-GOWT1 mouse stem cells,
  http://data.celltrackingchallenge.net/training-datasets/Fluo-N2DH-GOWT1.zip
• IEEE ISBI Cell Tracking Challenge, HeLa cells stably expressing H2b-GFP,
  http://data.celltrackingchallenge.net/training-datasets/Fluo-N2DL-HeLa.zip

When looking a suitable dataset, one should keep in mind that it should be a dataset which wasn't used for the training of Cellpose. Otherwise, the assessment of the segmentation performance would be rather meaningless. See the Section "Datasets" in the Cellpose paper (Stringer et al., Nat Methods 2021) for a list of datasets included in training.

---

2) The "fallback" variant. If we can’t find such a dataset, use any publicly available dataset of suitable cellular images and just apply Cellpose, for example: https://zenodo.org/record/3362976/files/B2.zip