patho-pix: An Automated Modular WSI Preprocessing Pipeline for Digital Pathology
Overview
patho-pix is a Python package designed to automate preprocessing of digital pathology images, leveraging existing processing frameworks into a modular pipeline. This package uses raw whole-slide-images (WSIs) and corresponding labels to preprocess the WSIs, It organizes and cleans up the input data into ready-to-use tiles and metadata, ensuring high-quality data for downstream machine learning analysis. Patho-pix integrates several key steps, including extraction of tissue regions ommiting background, tiling of the WSI, color normalization, artifact detection and removal and exportation of the tiles and metadata.
Features
- Tissue Region Extraction: Identifies and focuses on tissue-covered regions, excluding the background of the image.
- Tiling: Divides WSIs into tiles of dynamic size (i.e. 1024x1024).
- Color Normalization: Color normalization across tiles for consistency.
- Artifact Detection: Identifies and removes tiles with artifacts (Upcoming feature).
- Artifact Removal: The pipeline provides two options to deal with artifacts. Corrupted tiles can either be thrown away and disregarded for further analysis. Alternatively a generative AI can be applied enhancing image quality by removing artifacts and filling in the missing data (Upcoming feature).
- Export Tiles & Metadata: Saves processed tiles and associated metadata for further use.
Pipeline
Usage
To use patho pix you need to manually install the OpenSlide C-library as a dependency.
- Mac (with homebrew):
brew install openslide. - Linux (with the package manager of your choice): e.g., Ubuntu: sudo apt install
openslide-tools - Windows: Take a look at the OpenSlide Docu
Contributing
If you want to support the patho-pix project please take a look at the CONTRIBUTING guide.
For VSCode
You can use the provided Dev Container to quick start your development process of the patho pix project. Install Docker Desktop and VS Code Extension Dev Containers.