BreastGAN

BreastGAN, second experiment.

Paper

Injecting and removing malignant features in mammography with CycleGAN: Investigation of an automated adversarial attack using neural networks

Anton S. Becker 1,2), Lukas Jendele* 3), Ondrej Skopek* 3), Nicole Berger 1), Soleen Ghafoor 1,4), Magda Marcon 1), Ender Konukoglu 5)

1. Institute of Diagnostic and Interventional Radiology, University Hospital of Zurich; Zurich, Switzerland

2. Department of Health Sciences and Technology, ETH Zurich; Zurich, Switzerland

3. Department of Computer Science, ETH Zurich

4. Department of Radiology, Memorial Sloan Kettering Cancer Center, New York City, USA

5. Computer Vision Laboratory, Department of Information Technology and Electrical Engineering, ETH Zurich

In arXiv, 2018. (* joint contribution)

Correspondence to: Anton S. Becker, Institute of Diagnostic and Interventional Radiology, UniversitätsSpital Zürich, Raemistrasse 100, CH-8091 Zürich

E-Mail: [email protected]

Citation

If you use this code for your research, please cite our paper:

@article{BreastGAN2018,
  title={{Injecting and removing malignant features in mammography with CycleGAN: Investigation of an automated adversarial attack using neural networks}},
  author={Becker, Anton S and Jendele, Lukas and Skopek, Ondrej and Berger, Nicole and Ghafoor, Soleen and Marcon, Magda and Konukoglu, Ender},
  journal={arXiv preprint arXiv:1811.07767},
  year={2018}
}

Related Projects

CycleGAN: Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks. Software that can generate photos from paintings, turn horses into zebras, perform style transfer, and more.

Requirements and versions:

• Python 3.5

• Git

• Tensorflow 1.12.0

Important: When committing, remember to be in the virtual environment, for hooks to work.

NOTE: All code in Jupyter Notebooks is purely experimental. Use at your own risk.

Setup

Make sure there is no venv/ directory in your repository. If there is, remove it. Run the following commands:

./setup/create_venv.sh
source venv/bin/activate

Important: For all commands here, we assume you are sourced into the virtual environment: source venv/bin/activate

Running the experiments

Image conversion

Put all data into the directories in data_in/. Supported are: 1_BCDR/, 2_INbreast/, 3_zrh/, cbis.

1. ./local/convert_images_all.sh

2. ./local/merge_images_all.sh

3. ./local/split_images_all.sh

4. ./local/treval_split.sh

GAN training

1. ./local/run.sh. Wait 24 hours.

2. ./local/infer.sh. Make sure to enter the correct checkpoint number here and below.

3. ./local/to_png.sh. Make sure to change the paths in notebooks/inference_tfrecord_to_png.py.

Jupyter notebooks

NOTE: All code in Jupyter Notebooks is purely experimental. Use at your own risk.

Save notebooks in the notebooks/ directory. These can also be worked on locally using Jupyter. In the project root directory, you can run either:

• jupyter notebook,

• or jupyter lab.

Add the following cell to your notebook, ideally in a "section":

# noqa
import os
wd = %pwd
print('Current directory:', wd)
if wd.endswith('notebooks'):
    %cd ..

Docker / Custom runner

After you have converted images and built a Docker image using: make build-cpu or make build-gpu (or pulling one from the remote Docker hub), you can use the Docker wrapper:

./run_docker.sh password jupyter  # Jupyter Notebooks
./run_docker.sh password lab  # Jupyter Lab
./run_docker.sh password model  # Train model
./run_docker.sh password modelboard  # Train model + run TensorBoard

Directory structure

• cluster/ — scripts for running the training/evaluation on the cluster

• data_in/ — input data and associated scripts/configs

• data_out/ — output data and logs + associated scripts/configs

• docker/ — setup and configs for running stuff inside and outside of Docker

• local/ — scripts for running the training/evaluation locally

• models/ — scripts defining the models + hyperparameters

• notebooks/ — data exploration and other rapid development notebooks

  • Models from here should eventually be promoted into models/

• resources/ — Python utilities

• setup/ — environment setup and verification scripts in Python/Bash

• venv/ — the (local) Python virtual environment

Formatting

Run: ./setup/clean.sh. A Git hook will tell you if any files are misformatted before committing.

Third Party code used in this project

ICNR

https://github.com/kostyaev/ICNR by Dmitry Kostyaev

Licensed under the MIT Licence.

In: models/utils/icnr.py

Tensor2Tensor

https://github.com/tensorflow/tensor2tensor by The Tensor2Tensor Authors.

Licensed under the Apache License Version 2.0.

In: models/breast_cycle_gan

TensorFlow, TensorFlow Models

https://github.com/tensorflow/tensorflow, https://github.com/tensorflow/models by The TensorFlow Authors.

Licensed under the Apache License Version 2.0.

In: models/breast_cycle_gan