HookNet-TLS

HookNet-TLS is a deep learning algorithm designed to accurately detect Tertiary Lymphoid Structures and Germinal Centers (GC) within whole-slide pathology images. Building on the foundation of the HookNet architecture, HookNet-TLS is a useful tool for pathologists and researchers examining TLSs and GCs.

Quick Start

Installation

Ensure you have Docker installed and running on your system. Clone this repository and build the Docker image:

git clone https://github.com/DIAGNijmegen/pathology-hooknet-tls.git
cd hooknet-tls
docker build -t hooknet-tls .

Usage

Note. The algorithm expects that the input whole-slide-image contains the spacing corresponding to approximately 0.5µm and 2.0µm.

python3 -m hooknettls \
    hooknettls.default.image_path=/tmp/TCGA-21-5784-01Z-00-DX1.tif \
    hooknettls.default.mask_path=/tmp/TCGA-21-5784-01Z-00-DX1_tb_mask.tif

Related packages

HookNet-TLS uses the following packages

• WholeSlideData
• HookNet

Data

• HookNet-TLS Weights: https://zenodo.org/records/10614942
• HookNet-TLS Annotations: https://zenodo.org/records/10614928
• HookNet-TLS Annotations Masks: https://zenodo.org/records/10635034

Support

If you are having issues, please let us know or submit a pull request.

License

This project is licensed under the MIT License