Adversarial Attacks against MRI Segmentation

This repository contains code to run adversarial attacks against the state-of-the-art cardiac cardiac segmentation networks as described in this paper: An Exploration of 2D and 3D Deep Learning Techniques for Cardiac MR Image Segmentation.

The code and instructions for training the target model can be found here - https://github.com/baumgach/acdc_segmenter.

Requirements

• Python 3.5
• Tensorflow (tested with tensorflow 1.15)
• The package requirements are given in requirements.txt

Running the code locally

Open the config/system.py and edit all the paths there to match your system.

Launch attacks by running the following python evaluate acdc_logdir/unet2D_bn_modified_wxent <attack> <Add Gaussian noise>

For example, python evaluate acdc_logdir/unet2D_bn_modified_wxent fgsm False

where you have to adapt the line to match your experiment. Note that, the path must be given relative to your working directory. Giving the full path will not work.

References

@article{baumgartner2017exploration,
  title={An Exploration of {2D} and {3D} Deep Learning Techniques for Cardiac {MR} Image Segmentation},
  author={Baumgartner, Christian F and Koch, Lisa M and Pollefeys, Marc and Konukoglu, Ender},
  journal={arXiv preprint arXiv:1709.04496},
  year={2017}
}

Contributors:

1. Abhinav Garg ([email protected])
2. Anant Kandikuppa ([email protected])