JupyterLab + PyTorch + 3D Slicer + VS Code Docker Image

A comprehensive Docker image combining JupyterLab, PyTorch with CUDA support, 3D Slicer medical imaging platform, and VS Code in a browser-accessible environment with VNC desktop.

Acknowledgements

This project builds upon and integrates several excellent open-source projects:

• jupyter-remote-desktop-proxy: The foundation for providing VNC desktop access through Jupyter
• PyTorch Docker Images: Official PyTorch base images with CUDA support
• 3D Slicer: Open-source medical image computing platform
• code-server: VS Code in the browser by Coder

Special thanks to the JupyterHub, PyTorch, 3D Slicer, and VS Code communities for their outstanding work.

Features

• PyTorch 2.3.1 with CUDA 12.1 and cuDNN 8 support
• JupyterLab 4.5+ with JupyterHub integration
• Python and R kernels for Jupyter notebooks
• 3D Slicer 5.10.0 medical imaging platform with dark theme
• VS Code (code-server) in the browser with Python and Jupyter extensions
• XFCE4 Desktop accessible via browser with dark theme
• Persistent authentication tokens across container restarts
• Auto-start 3D Slicer in fullscreen mode when VNC desktop opens
• Corporate proxy and certificate support
• GPU support for PyTorch and CUDA workloads
• Common development tools: vim, nano, git, tmux, htop, and more

Building the Image

Basic Build

docker build -t jupyter_rad:dev .

Build with Corporate Proxy

If you're behind a corporate proxy, use these build arguments:

docker build \
  --build-arg http_proxy=http://proxy.company.com:8080 \
  --build-arg https_proxy=http://proxy.company.com:8080 \
  --build-arg HTTP_PROXY=http://proxy.company.com:8080 \
  --build-arg HTTPS_PROXY=http://proxy.company.com:8080 \
  --build-arg no_proxy=localhost,127.0.0.1 \
  --build-arg NO_PROXY=localhost,127.0.0.1 \
  -t jupyter_rad:dev .

Build with Corporate Certificates

If you need to add corporate CA certificates:

1. Create a folder containing your .crt files (e.g., ./corporate-certs/)
2. Build with the CERTS_FOLDER argument:

docker build \
  --build-arg CERTS_FOLDER=./corporate-certs \
  -t jupyter_rad:dev .

Build with Local 3D Slicer

If you already have 3D Slicer downloaded locally and want to use it instead of downloading during build:

1. Extract your Slicer archive to a folder (e.g., ./local-slicer/Slicer-5.10.0-linux-amd64/)
2. Build with the LOCAL_SLICER_FOLDER argument:

docker build \
  --build-arg LOCAL_SLICER_FOLDER=./local-slicer \
  -t jupyter_rad:dev .

Note: The folder should contain the extracted Slicer directory (e.g., Slicer-5.10.0-linux-amd64/).

Combined: Proxy + Certificates + Local Slicer

docker build \
  --build-arg http_proxy=http://proxy.company.com:8080 \
  --build-arg https_proxy=http://proxy.company.com:8080 \
  --build-arg CERTS_FOLDER=./corporate-certs \
  --build-arg LOCAL_SLICER_FOLDER=./local-slicer \
  -t jupyter_rad:dev .

Running the Container

Basic Run

docker run --rm -it -p 8888:8888 jupyter_rad:dev

Full Example: GPU + Memory

docker run --rm -it \
  --gpus all \
  --shm-size=2g \
  -p 8888:8888 \
  jupyter_rad:dev

Docker Run Flags Explained

Flag	Purpose
--rm	Automatically remove container when it stops
-it	Interactive terminal (see logs, enable Ctrl+C)
--gpus all	Enable all GPUs for CUDA workloads
--shm-size=2g	Increase shared memory to 2GB (prevents Slicer crashes)
-p 8888:8888	Expose JupyterLab port

Accessing the Services

Access the services in your browser:

• JupyterLab: http://localhost:8888/lab?token=YOUR_TOKEN
• VNC Desktop (3D Slicer): http://localhost:8888/desktop
• VS Code: http://localhost:8888/vscode

The authentication token is displayed when the container starts.

JupyterHub Integration

Kubernetes JupyterHub Setup

This image is designed to work with JupyterHub on Kubernetes. Configure it in your values.yaml file.

Once users spawn their server, they can access:

• JupyterLab: Default interface at /user/<username>/lab
• VNC Desktop (3D Slicer): /user/<username>/desktop
• VS Code: /user/<username>/vscode

The 3D Slicer application will auto-start in the VNC desktop.

User Configuration

Default User

• Username: jovyan
• UID: 1000
• GID: 100 (users group)
• Home: /home/jovyan
• Sudo: Passwordless sudo enabled

Pre-configured Features

3D Slicer

• Auto-starts in fullscreen mode
• Dark theme pre-configured
• WebEngine disabled (prevents container crashes)
• Extensions can be installed manually
• Full write access to /opt/Slicer directory

VS Code (code-server)

• Dark theme pre-configured
• Python extension installed
• Jupyter extension installed
• Settings persisted in ~/.local/share/code-server

XFCE Desktop

• Dark Adwaita theme
• Dark background
• 3D Slicer desktop shortcut

Development

Modifying the Image

1. Edit the Dockerfile or configuration files
2. Rebuild: docker build -t jupyter_rad:dev .
3. Test: docker run --rm -it --shm-size=2g -p 8888:8888 jupyter_rad:dev

Key Configuration Files

• Dockerfile: Main build definition
• start.sh: Container entrypoint script
• SlicerRC.py: 3D Slicer startup configuration
• Slicer.ini: Slicer preferences
• code-server-settings.json: VS Code settings
• xfce4-*.xml: Desktop environment themes
• jupyter_remote_desktop_proxy/share/xstartup: VNC session startup

License

This project maintains the original licenses of its components. See individual component documentation for details.