lxmwust/ad3dmil

Attention Based Deep 3D Multiple Instance Learning

Attention Based Deep 3D Multiple Instance Learning

Prerequisites:

• Python3
• PyTorch ==1.4.0 (with suitable CUDA and CuDNN version)
• torchvision == 0.5.0
• Numpy
• argparse
• PIL
• tqdm
• SimpleITK
• cv2
• skimage
• scipy
• pydicom

Dataset:

You need to provide the text lists of training, validation, and testing raw 3D CT files in "./dataset".

Training:

1. Segmentation of lung masks: preprocess/seg.py for binary classification and preprocess/seg-multi-class.py for multi-class classification;
2. Training AD3D-MIL models: train.py for binary classification and train-mc.py for multi-class classification;
3. Testing: test.py for binary classification and test-mc.py for multi-class classification;

Note: You may modify the path or parameters in the corresponding locations.

Citation:

If you use this code for your research, please consider citing:

@ARTICLE{9098062, author={Z. {Han} and B. {Wei} and Y. {Hong} and T. {Li} and J. {Cong} and X. {Zhu} and H. {Wei} and W. {Zhang}}, journal={IEEE Transactions on Medical Imaging}, title={Accurate Screening of COVID-19 Using Attention-Based Deep 3D Multiple Instance Learning}, year={2020}, volume={39}, number={8}, pages={2584-2594},}

Contact

If you have any problem about our code, feel free to contact [email protected].