This is the official code for the paper "Free Lunch in Pathology Foundation Model: Task-specific Model Adaptation with Concept-Guided Feature Enhancement".
-
Download model weights from CONCH.
-
Generate concept anchors by
python generate_concept_anchor.py. -
Use CATE-MIL (
catemil.py) to train your model or integrate the CATE (cate.py) into your MIL model.
If you find this work useful in your research, please consider citing:
@inproceedings{
huang2024free,
title={Free Lunch in Pathology Foundation Model: Task-specific Model Adaptation with Concept-Guided Feature Enhancement},
author={Yanyan Huang and Weiqin Zhao and Yihang Chen and Yu Fu and Lequan Yu},
booktitle={The Thirty-eighth Annual Conference on Neural Information Processing Systems},
year={2024},
url={https://openreview.net/forum?id=dwYekpbmYG}
}
We would like to thank the authors of CONCH for their excellent work.