csi_images: image and data utilities for CSI-Cancer

This package is a library for interacting with data, mainly from immunoflourescent microscopy, such as cells and the images thereof. While much of the functionality is specific to the CSI-Cancer organization, some of the functionality and structure may be beneficial for the broader community. Other packages in the CSI-Cancer organization may depend on this package.

Install with:

pip install csi_images

or

pip install csi_images[imageio,rds,dev]

The base version of the package includes only core data structures and manipulations.

Optional dependencies include:

• imageio: for reading and writing images, include .czi files.
• rds: for reading and writing RDS files, such as OCULAR outputs.
• dev: for development dependencies, including documentation, tests, building, etc.

Structure

This package contains these modules:

1. csi_scans.py: a module for interacting with scan-level files, such as .czi files.
   • Scan: a class that contains all of the scan metadata. for interacting with scan metadata, such as the slide ID, the path to the scan, and scan parameters. Recommend importing via from csi_images.csi_scans import Scan
2. csi_tiles.py: a module for interacting with tiles, which have a particular (x, y) position in the scan. Tiles have several frames taken at the same position.
   • Tile: a class for containing a tile's positional data. Includes a reference to a Scan. This class unifies multiple scanners' tile positioning to convert between index and (x, y). Recommend importing via from csi_images.csi_tiles import Tile
3. csi_frames.py: a module for interacting with frames, which are individual images. Imports csi_scans.py and csi_tiles.py. Recommend importing via from csi_images.csi_frames import Frame
   • Frame: a class for containing a frame's metadata. Includes a reference to a Tile. For a list of frames for a tile, use get_frames(tile). For all frames in a scan, use get_all_frames(scan). Each frame in a tile has a different channel, or light spectrum. The frame only contains metadata, but enables gathering of the image data through the get_image() method. For many frames, use [frame.get_image() for frame in frames]. Recommend importing via from csi_images.csi_frames import Frame
4. csi_events.py: a module for interacting with individual events. Imports csi_scans.py, csi_tiles.py, and csi_frames.py.
   • Event: a class for containing a single event's metadata and feature data. Key metadata (scan, tile, x, y) is required; the others are optional and flexible. Contains functions for determining event position in different coordinate frames and performing crop and montage image manipulations, as well as name conventions for saving crops and images. Recommend importing via from csi_images.csi_events import Event
   • EventArray: a class for containing a list of events, holding their data in Pandas DataFrames. Contains functions converting back and forth from Events and files. Recommend importing via from csi_images.csi_events import EventArray
5. csi_images.py: a module for image manipulation functions. These functions are developed to be portable and contain all of the image-related dependencies, making it possible to use the other modules without image-related dependencies. Recommend importing via from csi_images import csi_images

Documentation

For more detailed documentation, check the API docs.

Alternatively, once you have cloned the repository, you can open up docs/index.html in your browser.

To regenerate the documentation, ensure that you have installed the package with development dependencies and then run:

docs/make_docs.sh

Development Installation

1. Activate your conda (conda activate yourenv) or venv (source path/to/your/venv/bin/activate) environment first.
2. Clone csi_images and install:

cd ~/path/to/your/repositories
git clone git@github.com:CSI-Cancer/csi_images.git
pip install -r requirements.txt

This installs the package in "editable" mode, as well as all dependencies and development dependencies (e.g. pytest, pdoc, etc.).

You may also use the requirements-versioned.txt file to install known good versions.