WSIRadiomics

English | 中文

WSIRadiomics is a cell-level → WSI-level radiomics feature extraction toolkit for Whole Slide Images (WSI).

It supports starting from cell instance segmentation results (GeoJSON) to compute cell-level features and further aggregate them into WSI-level features, making it suitable for computational pathology, digital pathology, and multimodal research.

✨ Features

• 🧬 Cell-level → WSI-level feature computation pipeline
• 🧠 Supports basic radiomics features such as first-order and shape
• 🧩 Flexible YAML-based parameter configuration
• 🏷 Supports aggregation by cell type (cell_type)
• 📦 Modular design, easy to extend and customize
• 📝 Uses Python logging, no print statements
• 🚫 Core APIs do not write files, output is fully controlled by the user

📦 Installation

Method 1: Install via PyPI (Recommended)

pip install wsiradiomics

Suitable for direct usage, server environments, and virtual environments.

⚠️ Important notice (WSI dependency)
The OpenSlide system library must be installed manually:

# Windows / Linux (conda)
conda install -c conda-forge openslide openslide-python

Method 2: Install via Conda / Mamba Environment (Recommended for research)

The project provides a complete environment file:

conda env create -f environment.yaml
conda activate wsiradiomics

Or using mamba (faster):

mamba env create -f environment.yaml
mamba activate wsiradiomics

This method is especially suitable for WSI / OpenSlide / Linux server environments.

🚀 Quick Start

It is recommended to first refer to the example code and configuration files in the examples/ directory.

Example files overview

examples/
├── example_file.csv    # CSV example for batch processing
├── extract_from_pandas.py # Batch feature extraction example
└── params.yaml       # Parameter configuration example

1️⃣ Single WSI feature extraction (API usage)

from wsiradiomics.extractor import extract

res = extract(
    svs_path="example.svs",
    geojson_path="cells.geojson",
    params_path="params.yaml",
)

wsi_features = res["wsi_features"]

2️⃣ Batch processing (CSV-driven, recommended)

CSV example (see examples/example_file.csv):

wsi_path,mask_path
/path/to/wsi_001.svs,/path/to/wsi_001_cells.geojson
/path/to/wsi_002.svs,/path/to/wsi_002_cells.geojson

Run the example script:

python examples/extract_from_pandas.py \
  --input_csv examples/example_file.csv \
  --params examples/params.yaml \
  --out_dir result/

Output results:

result/
├── wsi_features.csv
└── run_wsi_feature_extract.log

📄 License

MIT License

📬 Contact

Feel free to submit issues or suggestions via GitHub Issues