Detailed instructions on how to install and run MONAI Label can be found on the official MONAI website (https://docs.monai.io/projects/label/en/latest/installation.html), but the below instructions will get you started quickly.
GPU support
If you have a decent NVidia GPU, you should first download and install the CUDA Toolkit 11.3 at https://developer.nvidia.com/cuda-11.3.0-download-archive?target_os=Windows.
If not, you can continue without installing this and all models will run just fine on the CPU.
Install MONAI Label (Windows)
Download and install the Windows version of Python 3.9 (not a newer version!) from https://www.python.org/downloads/release/python-390/, which will be indicated as “Windows x86-64 web-based installer” .
Open a Command Prompt in Windows and make a folder where you want MONAI Label to place its files such as C:\MONAI in the below example.
Navigate to the root of the C: drive:
C:\Users\Name>cd/
Make a directory named MONAI:
C:\>md MONAI
Navigate into the directory:
C:\>cd MONAI
Install pip:
C:\MONAI>python -m pip install --upgrade pip setuptools wheel
Install pytorch:
C:\MONAI>pip install torch==1.12.1+cu113 torchvision==0.13.1+cu113 torchaudio==0.12.1 --extra-index-url https://download.pytorch.org/whl/cu113
Install Monailabel:
C:\MONAI>pip install monailabel
Download all the files required for running a model from the MONAI Model Zoo
C:\MONAI>monailabel apps --name monaibundle --download --output apps
Start the MONAI Label server
Download the TotalSegmentator model and start a local server:
C:\MONAI>monailabel start_server --app apps/monaibundle --studies none --conf models wholeBody_ct_segmentation
Run MONAI Label in ScanXm
In the Segment panel there is a section named MONAI Label. Here the URL should be set as http://localhost:8000.
When you press the Reset button, the available models will appear in the dropdown box.
Select the wholeBody_ct_segmentation option in the dropdown box and press the run button to let MONAI Label generate the segmentation. Within a few minutes the segmentation will appear in ScanXm.
