Install the necessary packages as follows:
pip3 install -r requirements.txt
Key features:
- Implemented a modified V-Net [https://arxiv.org/abs/1606.04797]
- Use on-the-fly data augmentation (translate, zoom, shear, flip, and rotate)
- Uses group normalization (default group size 8)
Help:
usage: run_vnet3d.py [-h] --core_tag CORE_TAG --nii_dir NII_DIR --batch_size
BATCH_SIZE --image_size IMAGE_SIZE --learning_rate
LEARNING_RATE --group_size GROUP_SIZE --f_root F_ROOT
--n_validation N_VALIDATION --n_test N_TEST --optimizer
OPTIMIZER [--print_summary_only]
Script to run VNet3D
optional arguments:
-h, --help show this help message and exit
--core_tag CORE_TAG, -ct CORE_TAG
--nii_dir NII_DIR, -I NII_DIR
--batch_size BATCH_SIZE, -bs BATCH_SIZE
--image_size IMAGE_SIZE, -is IMAGE_SIZE
--learning_rate LEARNING_RATE, -lr LEARNING_RATE
--group_size GROUP_SIZE, -gs GROUP_SIZE
--f_root F_ROOT, -fr F_ROOT
--n_validation N_VALIDATION
--n_test N_TEST
--optimizer OPTIMIZER, -op OPTIMIZER
--print_summary_only
Here, --nii_dir
should have only nii.gz files within it.
Each sample in the --nii_dir
should have the following 5 suffices:
_flair.nii.gz
_t1.nii.gz
_t1ce.nii.gz
_t2.nii.gz
_seg.nii.gz
(label)
Here, the label file with_seg.nii.gz
should have only 0, 1, 2, 4 as its value, which corresponds to the BraTS 2018 brain tumor segmentation data [https://www.med.upenn.edu/sbia/brats2018/data.html].