- Create a folder to hold all your CT scans
- Within this folder, create a separate subfolder for each CT scan (e.g., casename00001, casename00002).
- Place your new CT scan file in its corresponding subfolder and name the CT scan "ct.nii.gz".
/path/to/your/CT/scan/folders
├── casename00001
│ └── ct.nii.gz
├── casename00002
│ └── ct.nii.gz
├── casename00003
│ └── ct.nii.gz
...
inputs_data=/path/to/your/CT/scan/folders
outputs_data=/path/to/your/output/folders
# If using singularity
wget https://huggingface.co/qicq1c/SuPreM/resolve/main/suprem_final.sif
SINGULARITYENV_CUDA_VISIBLE_DEVICES=0 singularity run --nv -B $inputs_data:/workspace/inputs -B $outputs_data:/workspace/outputs suprem_final.sif
# If using docker
docker pull qchen99/suprem:v1
sudo docker container run --gpus "device=0" -m 128G --rm -v $inputs_data:/workspace/inputs/ -v $outputs_data:/workspace/outputs/ qchen99/suprem:v1 /bin/bash -c "sh predict.sh"git clone https://github.com/MrGiovanni/SuPreM
cd SuPreM/direct_inference/pretrained_checkpoints/
wget https://huggingface.co/MrGiovanni/SuPreM/resolve/main/supervised_suprem_swinunetr_2100.pth
wget https://huggingface.co/MrGiovanni/SuPreM/resolve/main/supervised_suprem_unet_2100.pthconda create -n suprem python=3.9
source activate suprem
pip install torch==1.11.0+cu113 torchvision==0.12.0+cu113 torchaudio==0.11.0 --extra-index-url https://download.pytorch.org/whl/cu113
pip install monai[all]==0.9.0
cd SuPreM/
pip install -r requirements.txtdatarootpath=/path/to/your/CT/scan/folders
# NEED MODIFICATION!!!
backbone=unet # or swinunetr
pretrainpath=./pretrained_checkpoints/supervised_suprem_unet_2100.pth # or ./pretrained_weights/supervised_suprem_swinunetr_2100.pth
savepath=./inference
cd SuPreM/direct_inference/
python -W ignore inference.py --save_dir $savepath.$backbone --checkpoint $pretrainpath --data_root_path $datarootpath --backbone $backbone --store_result --suprem