Usage

SCIP can be used as a command line interface (CLI) or its modules can be imported in a custom script. See the package API documentation for the latter option.

Configuration

The CLI runs a workflow that loads images and performs projection, illumination correction, segmentation, masking, and feature extraction. The CLI can be configured via options passed on the command line (for runtime configuration) and via a YAML config file (for pipeline configuration). For an overview of the command line options, run scip --help.

The YAML config file has the following specification:

load:
    format: {tiff, czi, multiframe_tiff, zarr}
    channels: []
    channel_names:
        - channel 1
        - channel 2
        - ...
    kwargs:
        regex: "^(?P<key_to_extract_from_filename>[0-9]+)$"
        scenes: {[], scene_name, regex_pattern} # only for czi loader
project:
    method: {op}
    settings:
        op: {median, max}
illumination_correction:
    method: {jones_2006}
    key: grouping_key
    export: {true, false}
    settings:
        median_filter_size: integer
        downscale: {none, integer}
segment:
    method: {cellpose}
    export: {true, false}
    settings:
        cell_diameter: {none, integer}
        dapi_channel_index: index to channels list
        parent_channel_index: index to channels list
        substract:
            left_index: index to channels list
            right_index: index to channels list
            for_index: index to channels list
mask:
    combined_indices: list of indices to channel list, masks of these channels will be combined
        for the combined features
    main_channel_index: index to channels list
    filters:
        - method: normaltest
          channel_indices: list of channel indices on which to apply filter
          settings:
        - method: std
          channel_indices: list of channel indices on which to apply filter
          settings:
            threshold: minimum required std
    methods:
        - method: "threshold"
          name: "threshold-name"
          export: {true, false}
          kwargs:
              smooth: [0, 0, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0, 0.5, 0.5, 0.5]
        - method: "spot"
          name: "spot-name"
          export: {true, false}
          kwargs:
              spotsize: 5
# with masks
feature_extraction:
    threshold-name: ["shape", "intensity", "bbox", "texture"]
    spot-name: ["shape", "intensity", "bbox", "texture"]
# without masks
feature_extraction: ["shape", "intensity", "bbox", "texture"]
export:
    format: {parquet, anndata}
    filename: name

You can use the example above as a start for your configuration.

Note that all top-level keys have to be present. You can disable a step by only providing the top-level key, like so:

mask:

Command-line

This package exposes one command-line client (CLI): scip

Call

scip --help

to get an overview of all available options.

The structure of the CLI is as follows:

scip [OPTIONS] OUTPUT CONFIG [PATHS]...

All runtime options, such as the number of workers are passed in the OPTIONS section. All pipeline settings are passed using the configuration file passed through CONFIG.