Skip to content

Latest commit

 

History

History
114 lines (109 loc) · 6.28 KB

README.md

File metadata and controls

114 lines (109 loc) · 6.28 KB
Raw

CereVessSeg model and CereVessPro contrastive learning for cerebrovascular segmentation

We build a deep learning model named CereVessSeg and a contrastive learning method CereVessPro for cerebrovascular segmentation in MRA volumes

System requirements

Operating System

Linux (Ubuntu 18.04, 20.04).

Hardware requirements

GPU and CPU are required for model training and inference. For training, RTX 3090 or stronger GPU is recommended. A strong CPU is also required to go along with the GPU (at least 20 cores).

Installation instructions

  1. Install CereVessSeg and CereVessPro in a virtual environment. We use anaconda to create and control virtual environments:

    • download the Anaconda installer using wget commands: 
      apt-get install wget
wget https://repo.anaconda.com/archive/Anaconda3-2023.09-0-Linux-x86_64.sh
      or directly download the installer from Anaconda for linux.
    • verify the hash code integrity of the package:
      sha256sum Anaconda3-2023.09-0-Linux-x86_64.sh
    • run the Anaconda bash shell script:
      bash Anaconda3-2023.09-0-Linux-x86_64.sh
    • after running the bash command, you’ll be welcomed to the Anaconda setup. However, you must review and agree to its license agreement before the installation. Hit Enter to continue. Type 'yes' when a question comes out (to use the default directory to place Anaconda), until the installer finishes and displays “Thank you for installing Anaconda3”. Close and re-open your terminal window for the installation to take effect, or enter the command source ~/.bashrc to refresh the terminal.

  2. Install CUDA 11.1 or higher. Here we use CUDA 11.1 for example, which was used in our experiments:

    • (optional) if you installed some old versions of CUDA before, it is better to remove previous installation： 
      !/bin/bash
sudo apt-get purge *nvidia*
sudo apt remove --autoremove nvidia-*
sudo rm /etc/apt/sources.list.d/cuda*
sudo apt remove --autoremove nvidia-cuda-toolkit
sudo apt-get autoremove && sudo apt-get autoclean
sudo rm -rf /usr/local/cuda*
    • verify gpu cuda availability:
      lspci | grep -i nvidia
    • verify the version of gcc compiler, which is required for CUDA usage:
      gcc --version
    • system update: 
      sudo apt-get update  
sudo apt-get upgrade
    • Install other import packages:
      sudo apt-get install g++ freeglut3-dev build-essential libx11-dev libxmu-dev libxi-dev libglu1-mesa libglu1-mesa-dev
    • get the PPA repository driver: 
      sudo add-apt-repository ppa:graphics-drivers/ppa
sudo apt-key adv --fetch-keys http://developer.download.nvidia.com/compute/cuda/repos/ubuntu1804/x86_64/7fa2af80.pub  
echo "deb https://developer.download.nvidia.com/compute/cuda/repos/ubuntu1804/x86_64/ |sudo tee /etc/apt/sources.list.d/cuda.list 
sudo apt-get update
    • Installing CUDA: 
      sudo apt-get -o Dpkg::Options::="--force-overwrite" install cuda-11-1 cuda-drivers`
    • setup path: 
      echo 'export PATH=/usr/local/cuda-11.1/bin:$PATH' >> ~/.bashrc
echo 'export LD_LIBRARY_PATH=/usr/local/cuda-11.1/lib64:$LD_LIBRARY_PATH' >> ~/.bashrc
source ~/.bashrc
sudo ldconfig
    • install cuDNN: 
      wget https://developer.nvidia.com/compute/machine-learning/cudnn/secure/8.0.4/11.1_20200923/cudnn-11.1-linux-x64-v8.0.4.30.tgz
tar -xzvf cudnn-11.1-linux-x64-v8.0.4.30.tgz
    • copy the following files into the cuda toolkit directory: 
      sudo cp -P cuda/include/cudnn*.h /usr/local/cuda-11.1/include
sudo cp -P cuda/lib64/libcudnn* /usr/local/cuda-11.1/lib64/
sudo chmod a+r /usr/local/cuda-11.1/lib64/libcudnn*
    • final check: 
      sudo apt install nvidia-cuda-toolkit
nvidia-smi
nvcc -V

  3. Install Python 3.9 or higher version by Anaconda.

    conda create -n your_env python=3.9
conda activate your_env

  4. Install Pytorch

    pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu111
conda install -c anaconda cudnn
conda install -c anaconda cudatoolkit

  5. Download this 2_CereVessSeg_CereVessPro folder to your device, then install CereVessSeg and CereVessPro

    cd Path_you_save/2_CereVessSeg_CereVessPro/CereVessSeg_CereVessPro_code
pip install -e .

  6. Setting up paths as nnUNet where you intend to save raw data, preprocessed data and trained models. Please follow the instructions here.

How to run it?

  1. Download our trained models and then unzip. Here are two ways to get the trained models: way1, way2.
  2. Put our trained models from ./netstore in the path you set for saving trained models.
  3. Copy MRA files in rawMRAs folder from Step 1 to current Linux device. You can also only CereVessSeg and CereVessPro for your own MRA data. Here we provided two MRA files in the MRA_data as examples along with their annotations. Please make sure every MRA file name ending with "_0000.nii.gz".
  4. Directly run segmentation inference by CereVessSeg. Here MRA_folder is the directory wherein MRA files are placed, and Pred_folder is the directory wherein MRA segmentation results will be saved. 
     cd PATH_you_save/2_CereVessSeg_CereVessPro/CereVessSeg_CereVessPro_code/nnunet/inference
 CUDA_VISIBLE_DEVICES=0 python predict_simple.py -i MRA_folder -o Pred_folder -t 503 -tr nnUNetTrainerMSV2 -m 3d_fullres -f 0 -p nnUNetPlansv2.1
  5. Directly run inference by finetuned CereVessSeg that was pretrained by CereVessPro: 
     cd Path_you_save/2_CereVessSeg_CereVessPro/CereVessSeg_CereVessPro_code/nnunet/inference
 CUDA_VISIBLE_DEVICES=0 python predict_simple.py -i MRA_folder -o Pred_folder -t 503 -tr nnUNetTrainerSSLTune -m 3d_fullres -f 0 -p nnUNetPlansv2.1