cosmosec/cosmo-cagan-stainnorm

This open-source project contains the source code for the paper titled 'Colour Adaptive Generative Networks for Stain Normalisation of Histopathology Images' as published in 'Medical Image Analysis' in 2022.

Cosmo-CAGAN-StainNorm

Configs

All configs are located configs folder

Training

For training the models section, simply run python3 main.py dataset=cam16. A specific dataset can be chosen using dataset=[tcga|breakhis|cam16|cam17].

Testing

For testing the model, simply run python3 main.py dataset=cam16 run=test. A specific dataset can be chosen using dataset=[tcga|breakhis|cam16|cam17]. For classifier training, simply run python3 classifier.py dataset=cam17 model=classifier run.opt_run.n_epoch=40. A specific dataset can be chosen using dataset=[tcga|breakhis|cam17].

Trained Packages

We provide several pretrained trained models. We also offer our train-test splits.

method	Dataset	url
CAGAN	TCGA-IDH	model
CAGAN	BreakHis	model
CAGAN	CAMELYON16	model

Citing Paper

In your research, if this code is useful, please cite the following paper:

@article{cong2022colour, title={Colour adaptive generative networks for stain normalisation of histopathology images}, author={Cong, Cong and Liu, Sidong and Di Ieva, Antonio and Pagnucco, Maurice and Berkovsky, Shlomo and Song, Yang}, journal={Medical Image Analysis}, pages={102580}, year={2022}, publisher={Elsevier} }