An attention residual u-net with differential preprocessing and geometric postprocessing: Learning how to segment vasculature including intracranial aneurysms
@article{mu2023attention,
title={An attention residual u-net with differential preprocessing and geometric postprocessing: Learning how to segment vasculature including intracranial aneurysms},
author={Mu, Nan and Lyu, Zonghan and Rezaeitaleshmahalleh, Mostafa and Tang, Jinshan and Jiang, Jingfeng},
journal={Medical Image Analysis},
volume={84},
pages={102697},
year={2023},
publisher={Elsevier}
}
Some core dependencies:
- torch == 1.11.0
- pydensecrf == 1.0rc2 # for performing Fully-Connected CRFs
More details can be found in <./requirements.txt>
- Please follow the instructions for placing the data in
dataset/, which should be structured as follows:
IAData
├── train
| ├── ct
| | ├── volume-0.nii.gz
| | └── volume-1.nii.gz
| | ...
| | └── volume-14.nii.gz
| └── seg
| ├── segmentation-0.nii.gz
| └── segmentation-1.nii.gz
| ...
| └── segmentation-14.nii.gz
└── test
├── ct
| ├── volume-0.nii.gz
| └── volume-1.nii.gz
| ...
| └── volume-7.nii.gz
└── seg
├── segmentation-0.nii.gz
└── segmentation-1.nii.gz
...
└── segmentation-7.nii.gz
-
Set the paths to the training set, testing set and trained model in
parameter.py -
Test the validity of maximum and minimum thresholds for training data using
data_analysis/get_threshold.py
python data_analysis/get_threshold.py
- Crop the 3D data for the training and testing sets using
data_prepare/get_training_set.pyanddata_prepare/get_testing_set.pyrespectively
python data_prepare/get_training_set.py
python data_prepare/get_testing_set.py
- Run
train.pyto start the training
python train.py
- Run
test.pyto start the testing
python test.py
- Run
denseCRF/3D-CRF.pyto perform CRF post-processing
python denseCRF/3D-CRF.py
If you find our code useful for your research, please cite our paper:
N. Mu, Z. Lyu, M. Rezaeitaleshmahalleh, J. Tang, and J. Jiang, "An attention residual u-net with differential preprocessing and geometric postprocessing: Learning how to segment vasculature including intracranial aneurysms", Medical Image Analysis, vol. 84, 2023.
In case of any questions, please contact the corresponding author J. Jiang at jjiang1@mtu.edu