Surformer : an interpretable pattern-perceptive survival transformer for cancer survival prediction from histopathology whole slide images

This is the official pytorch implementation of Surformer Surformer : an interpretable pattern-perceptive survival transformer for cancer survival prediction from histopathology whole slide images.

Pipline

Experiment Results on five TCGA tumor datasets

Requirements

Installation

Please install pytorch version >=1.2

Dataset Preparation

Please download the official TCGA datasets of BRCA, BLCA, GBMLGG, LUAD, and UCEC. For more details of pre-processing, please refer to CLAM.

Model training and testing

before training and testing, please update configs. Generally, we train the model with one 12 GB memory GPU.

  python main.py 

Citation

If you find this code useful for your research, please cite our paper

@article{wang2023surformer,
  title={Surformer: an interpretable pattern-perceptive survival transformer for cancer survival prediction from histopathology whole slide images},
  author={Wang, Zhikang and Gao, Qian and Yi, Xiaoping and Zhang, Xinyu and Zhang, Yiwen and Zhang, Daokun and Li{\`o}, Pietro and Bain, Chris and Bassed, Richard and Li, Shanshan and others},
  journal={Computer Methods and Programs in Biomedicine},
  pages={107733},
  year={2023},
  publisher={Elsevier}
}

@article{wang2023targeting,
  title={Targeting tumor heterogeneity: multiplex-detection-based multiple instance learning for whole slide image classification},
  author={Wang, Zhikang and Bi, Yue and Pan, Tong and Wang, Xiaoyu and Bain, Chris and Bassed, Richard and Imoto, Seiya and Yao, Jianhua and Daly, Roger J and Song, Jiangning},
  journal={Bioinformatics},
  volume={39},
  number={3},
  pages={btad114},
  year={2023},
  publisher={Oxford University Press}
}

Contact

If you have any question, please feel free to contact us. E-mail: zhikang.wang@monash.edu