This is the official implement of Artifact Restoration in Histology Images with Diffusion Probabilistic Models (MICCAI2023) Arxiv

This is the first attempt at a denoising diffusion probabilistic model for histological artifact restoration, called ArtiFusion. Specifically, ArtiFusion formulates the artifact region restoration as a gradual denoising process, and its training relies solely on artifact-free images to simplify the training complexity. Furthermore, to capture local-global correlations in the regional artifact restoration, a novel Swin-Transformer denoising architecture is designed, along with a time token scheme. Our extensive evaluations demonstrate the effectiveness of ArtiFusion as a pre-processing method for histology analysis, which can successfully preserve the tissue structures and stain style in artifact-free regions during the restoration.

The dataset is a subset of Camelyon17. You may download from the following link. here

The proposed ArtiFusion learns the capability of generating local tissue representation from contextual information during the training stage. We follow the training procedure in guided-diffusion

Run train.sh to train a DDPM model

sh train.sh

Run sample.sh for sampling from trained model

sh sample.sh

python test.py --conf_path confs/XXXX.yml

We develop our code based on the implementation of RePaint and guided-diffusion. And thanks to Yiqing Shen for the contribution of the codes for down-sample classification tasks.

You may cite us as

@InProceedings{ArtiFusion,
  author="He, Zhenqi
  and He, Junjun
  and Ye, Jin
  and Shen, Yiqing",
  title="Artifact Restoration in Histology Images with Diffusion Probabilistic Models",
  booktitle="Medical Image Computing and Computer Assisted Intervention -- MICCAI",
  year="2023",
  pages="518--527"
}