Docker pLGG_Segmentation

Main repo: https://github.com/AIM-KannLab/pLGG_Segmentation

If you find this useful, please cite:

@article{doi:10.1148/ryai.230254,
author = {Boyd, Aidan and Ye, Zezhong and Prabhu, Sanjay P. and Tjong, Michael C. and Zha, Yining and Zapaishchykova, Anna and Vajapeyam, Sridhar and Catalano, Paul J. and Hayat, Hasaan and Chopra, Rishi and Liu, Kevin X. and Nabavizadeh, Ali and Resnick, Adam C. and Mueller, Sabine and Haas-Kogan, Daphne A. and Aerts, Hugo J. W. L. and Poussaint, Tina Y. and Kann, Benjamin H.},
title = {Stepwise Transfer Learning for Expert-level Pediatric Brain Tumor MRI Segmentation in a Limited Data Scenario},
journal = {Radiology: Artificial Intelligence},
volume = {6},
number = {4},
pages = {e230254},
year = {2024},
doi = {10.1148/ryai.230254},
    note ={PMID: 38446044},
URL = { https://doi.org/10.1148/ryai.230254
}

Prerequisites:

• Install docker: https://docs.docker.com/engine/install
• Clone repo: git clone
• GPU: Nvidia GPU with CUDA support (tested on Nvidia A6000)
• The pretrained models can be downloaded at the following Drive link https://drive.google.com/file/d/1cbi3p9IoKWjKR-pl3yXde6ISx4hZy2DB/view?usp=sharing. Unzip this file, the unzipped folder should be named nnUnet_trained_models and placed in 'nnUnet/' folder.

How to run the docker image inference on single MRI T2w image

0. Put all images in the 'example_input' folder
1. Build the docker image: sudo docker build -t segmentation . Note: this will take a while to build the docker image
2. Run the docker image: sudo docker run -i --gpus=all --ipc=host -v ${PWD}/output:/output/preprocessed -t segmentation Note: this will take a while to run the docker image, depending on how many images you have in the 'example_input' folder/ how many GPUs you have. Your output will be in the 'preprocessed' folder.

!To pass parameters to docker(modify input/output paths), when running the docker image, use the following command: sudo docker run -i --gpus=all --ipc=host -v ${PWD}/output:./output/ -t segmentation --CUDA_VISIBLE_DEVICES=0 --T2W_dir /example_input --output_path /output/

Optional: to clean dockers sudo docker image prune -a or docker image prune -a --filter "until=24h" or docker system prune

Dataset

Example dataset (note, this is not the example MRI with brain tumor; this is just a sample dataset to test the pipeline): https://openneuro.org/datasets/ds000228/versions/1.1.0 To run on your own data, place T2w .nii.gz files in the 'example_input' folder.