This repository is an official PyTorch implementation of the paper "Diagnose Like A Pathologist: Weakly-Supervised Pathologist-Tree Network for Slide-Level Immunohistochemical Scoring" [paper] from AAAI 2021.

• Python 3.6
• PyTorch >= 1.5.0
• torch-geometric
• numpy
• sklearn
• openslide

• Train the PTree-Net with your HER2 WSI dataset:

python ./train.py 

If you find our work useful in your research or publication, please cite our work:

@inproceedings{chen2021diagnose,
  title={Diagnose Like A Pathologist: Weakly-Supervised Pathologist-Tree Network for Slide-Level Immunohistochemical Scoring},
  author={Chen, Zhen and Zhang, Jun and Che, Shuanlong and Huang, Junzhou and Han, Xiao and Yuan, Yixuan},
  booktitle={Proceedings of the AAAI Conference on Artificial Intelligence},
  year={2021}
}

• GCN implementation with PyTorch Geometric.