SPIE2023

Towards Lifting the Trade-off between Accuracy and Adversarial Robustness of Deep Neural Networks with Application on COVID 19 CT Image Classification and Medical Image Segmentation

For classification:

requirements:

python3.8.10

pytorch1.9.0

guidance:

1. go to "DNNRobustness/app/COVID19a/" for the COVID19 experiment:

1.1 run train.py to get baseline model;

1.2 run train_adv.py to get model trained with our proposed adversarial training.

2. go to "DNNRobustness/app/MedMNIST/" for the COVID19 experiment:

2.1 run train.py to get baseline model;

2.2 run train_adv.py to get model trained with our proposed adversarial training.

For segmentation:

1. Please go to this link(https://github.com/MIC-DKFZ/nnUNet) for instructions on installing and configuring nnUnet and dependent libraries.

2. All the experimental data can be downloaded from http://medicaldecathlon.com/ or as shown in paper(Isensee, F., Jaeger, P. F., Kohl, S. A., Petersen, J., & Maier-Hein, K. H. (2020). nnU-Net: a self-configuring method for deep learning-based biomedical image segmentation. Nature Methods, 1-9.)

3. Go to "nnunet/run/"

4. use "python run_training --task id" to get the baseline model

5. use "python run_our_training --task id" to get the model with our proposed adversarial training

The experiment was run on Tesla V100 GPUs, CentOS system.

Based on original nnUnet, we did modifications on:

1.nnunet/training/network_training/network_trainer.py

2.nnunet/training/network_training/nnUNetTrainer.py

3.nnunet/training/network_training/nnUNetTrainerV2.py

4.nnunet/training/loss_functions/dice_loss.py

5.nnunet/training/loss_functions/crossentropy.py

6.nnunet/training/loss_functions/deep_supervision.py

7.nnunet/training/dataloading/dataset_loading.py

8.nnunet/training/data_augmentation/data_augmentation_moreDA.py

9.nnunet/utilities/to_torch.py

contact

Should you have any questions, please feel free to contact:

[email protected]

[email protected]