If you feel like you keep rewriting the same code over and over again when working on Whole Slide Images and you wish there were a nicely integrated library to automate all this, you came to the right place. With PathAIA we aim to create a fast, high level and modular library to work on WSI at scale in order to perform image analysis or to create a well rounded dataset for your machine learning project.

Simple installation from PyPI

pip install pathaia

from pathaia.patches import patchify_folder_hierarchically

You can extract at multiple pyramid levels with a hierarchical structure between patches of different levels. You can control pretty much every extraction parameter you like, from patch size to interval between patches or filters to chose which patch to extract. You can also decide whether you want to save patches as png or just extract coordinates in csv.

infolder = "/path/to/input/slide/folder"
outfolder = "/path/to/output/patches/folder"
top_level = 5
low_level = 0
psize = 224
interval = {"x": 224, "y": 224}
silent = list(range(low_level, top_level+1))
extensions = [".svs"]
recurse = False
slide_filters = ["full"]
verbose = 2

With these parameters you will find all svs slides that are directly in infolder and extract patch coordinates from levels 0 to 5 with a hierarchical structure. No png image will be stored as silent lists all levels. Patches will be contiguous with size 224 and will only be extracted from tissue zone that are determined by filtering on slide thumbnails. With verbose=2 thumbnails of extracted areas are also stored on disk.

patchify_folder_hierarchically(
    infolder,
    outfolder,
    top_level,
    low_level,
    psize,
    interval,
    silent=silent,
    extensions=extensions,
    recurse=recurse,
    slide_filters=slide_filters,
    verbose=verbose,
)

Output csv will look like :

You can use more advanced features to work on slides, most notably using your custom filters. Check documentation for more info.

The lightning community is maintained by 4 core contributors from Institut Universitaire du Cancer de Toulouse - Oncopole:

• Arnaud Abreu
• Pilar Ortega
• Robin Schwob
• Kevin Cortacero

If you have any questions please:

1. Read the docs.
2. Check existing issues, or add a new issue

Please observe the GNU GPL 3.0 license that is listed in this repository.

If you want to cite the framework feel free to use this.

@article{pathaia2021,
  title={PathAIA},
  author={Abreu, A and .al},
  journal={GitHub. Note: https://github.com/ArnaudAbreu/PathAIA},
  volume={3},
  year={2021}
}