This repository provides a command-line interface for Mesmer segmentation of MIBI and OME-XML TIFFs.
Install this repo. You can do so using pip install git+https://github.com/WEHI-SODA-Hub/MibiSegmentation.git or by cloning the repo and running pip install ..
Don't forget to obtain an API key (see here) and export it like so:
export DEEPCELL_ACCESS_TOKEN="YOURTOKEN"Usage: mesmer-segment [OPTIONS] TIFF
Segments a MIBI or OME-XML TIFF using Mesmer, and prints the result to stdout. Note that you will need to obtain and export a DeepCell API key as explained here: https://deepcell.readthedocs.io/en/master/API-key.html
╭─ Arguments ──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────╮
│ * tiff PATH Path to the TIFF input file. │
│ [default: None] │
│ [required] │
╰──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────╯
╭─ Options ────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────╮
│ * --nuclear-channel TEXT Name of the nuclear channel. │
│ [default: None] │
│ [required] │
│ * --membrane-channel TEXT Name(s) of the membrane channels (can be repeated)Ensure that channels with spaces are quoted. │
│ [default: None] │
│ [required] │
│ --compartment [whole-cell|nuclear] Compartment to segment (whole-cell or nuclear). │
│ [default: whole-cell] │
│ --combine-method [prod|max] Method to use for combining channels (prod or max). │
│ [default: prod] │
│ --segmentation-level INTEGER RANGE [-1<=x<=10] Segmentation level between 0-10 where 0 is less segmentation and 10 is more. Set to -1 to use maxima_threshold instead. (This option is for backwards compatibility with an old tool.) │
│ [default: -1] │
│ --maxima-threshold FLOAT RANGE [x>=0] Controls segmentation level directly in mesmer, not sure scaling via segmentation_level (lower values = more cells, higher values = fewer cells). Provide a value >0 to use this parameter. │
│ [default: 0.1] │
│ --interior-threshold FLOAT Controls how conservative model is in distinguishing cell from background (lower values = larger cells, higher values = smaller cells). │
│ [default: 0.3] │
│ --maxima-smooth FLOAT RANGE [x>=0] Controls what is considered a unique cell (lower values = more separate cells, higher values = fewer cells). │
│ [default: 0] │
│ --min-nuclei-area INTEGER RANGE [x>=0] Minimum area of nuclei to keep. │
│ [default: 15] │
│ --remove-cells-touching-border --no-remove-cells-touching-border Whether to remove cells touching the border of the image. │
│ [default: remove-cells-touching-border] │
│ --pixel-expansion INTEGER RANGE [x>=0] Specify a manual pixel expansion after segmentation. │
│ [default: 0] │
│ --padding INTEGER RANGE [x>=0] Number of pixels to crop the image by before segmentation. │
│ [default: 0] │
│ --install-completion Install completion for the current shell. │
│ --show-completion Show completion for the current shell, to copy it or customize the installation. │
│ --help Show this message and exit. │
╰──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────╯
A typical usage example is:
mesmer-segment image.tiff \
--nuclear-channel dsDNA \
--membrane-channel panCK \
--membrane-channel "MHC I" \
--compartment whole-cell > result.tiffNote: make sure to put quotes around channel names with spaces, e.g., --membrane-channel "MHC I (HLA A B C)".
Also note that QuPath may display these channels with underscores, e.g., MHC_I _(HLA_A_B_C).