A Generalizable and Robust Deep Learning Algorithm for Mitosis Detection in Multicenter Breast Histopathological Images(Medical Image Analysis)

Journal Link

This is the source code for the first place solution to the MICCAI 2021 MIDOG microscopy domain generalization challenge. Our Algorithms and Weights is already on the platform, ready to run directly, you can get the results by applying, on your image.Please note that your grand challenge account has to be certified by the school or business, you will have a green tick once you pass it, so we can agree, then you can upload the image and run the code.

Please open new threads or address all questions to xiyue.wang.scu@gmail.com

Hardware

• 32GB of RAM
• 4*Nvidia V100 32G GPUs

Updates / TODOs

Please follow this GitHub for more updates.

• Add training code
• Add inference code for evaluation.
• Add model.
• Add fourier-based data augmentation.

1.Preparations

• Data Preparation

  • Download training challenge MIDOG 2021 data

  • External independent datasets : 1.AMIDA13 2.MITOSIS14 3.TUPAC-auxiliary 4.MIDOG2022

2.Get fourier-based data augmentation

Step 1: Apply FFT to source and target images.

Step 2: Replace the low frequency part of the source amplitude with that from the target.

Step 3: Apply inverse FFT to the modified source spectrum.

Here are some images we grabbed randomly

python get_fda_image.py

3.Get instance mask

please see the HoVer-Net,get the cell mask.And then intersect with the mitotic bbox to get the mitotic mask, and finally preprocess 512*512 patches. Here is our processed some image and the corresponding mask

Inference

test image is on the path ./test/007.tiff

python process.py

License

This code(FMDet) is released under the GPLv3 License and is available for non-commercial academic purposes.

Citation

Please use below to cite this paper if you find our work useful in your research.

@article{WANG2022102703,
title = {A generalizable and robust deep learning algorithm for mitosis detection in multicenter breast histopathological images},
author = {Xiyue Wang and Jun Zhang and Sen Yang and Jingxi Xiang and Feng Luo and Minghui Wang and Jing Zhang and Wei Yang and Junzhou Huang and Xiao Han},
journal = {Medical Image Analysis},
pages = {102703},
year = {2022},
issn = {1361-8415},
doi = {https://doi.org/10.1016/j.media.2022.102703}
}