Raidionics processing backend for performing segmentation and computation of standardized report (RADS)

The code corresponds to the Raidionics backend for running processing pipelines over MRI/CT scans. The segmentation of a few organs or tumor types, as well as the generation of standardized reports are included.
The module can either be used as a Python library, as CLI, or as Docker container.

Installation

pip install git+https://github.com/dbouget/raidionics_rads_lib.git

Continuous integration

Operating System	Status
Windows
Ubuntu
macOS
macOS ARM

Getting started

Notebooks

Below are Jupyter Notebooks including different examples on how to get started with the segmentation and reporting tasks, either over the brain or mediastinal area.

Notebook	Clinical target	Colab	GitHub
Preoperative Segmentation	🧠 Neuro
Preoperative Segmentation	🫁 Mediastinum
Postoperative Segmentation	🧠 Neuro
Preoperative Reporting	🧠 Neuro
Postoperative Reporting	🧠 Neuro

CLI

raidionicsrads -c CONFIG (-v debug)

CONFIG should point to a configuration file (*.ini), specifying all runtime parameters, according to the pattern from blank_main_config.ini.

Python module

from raidionicsrads.compute import run_rads
run_rads(config_filename="/path/to/main_config.ini")

Docker

When calling Docker images, the --user flag must be properly used in order for the folders and files created inside the container to inherit the proper read/write permissions. The user ID is retrieved on-the-fly in the following examples, but it can be given in a more hard-coded fashion if known by the user.

⚠️ The Docker image can only perform inference using the CPU, there is no GPU support at this stage.

docker pull dbouget/raidionics-rads:v1.1-py38-cpu

For opening the Docker image and interacting with it, run:

docker run --entrypoint /bin/bash -v /home/<username>/<resources_path>:/workspace/resources -t -i --runtime=nvidia --network=host --ipc=host --user $(id -u) dbouget/raidionics-rads:v1.1-py38-cpu

The /home/<username>/<resources_path> before the column sign has to be changed to match a directory on your local machine containing the data to expose to the docker image. Namely, it must contain folder(s) with images you want to run inference on, as long as a folder with the trained models to use, and a destination folder where the results will be placed.

For launching the Docker image as a CLI, run:

docker run -v /home/<username>/<resources_path>:/workspace/resources -t -i --runtime=nvidia --network=host --ipc=host --user $(id -u) dbouget/raidionics-rads:v1.1-py38-cpu -c /workspace/resources/<path>/<to>/main_config.ini -v <verbose>

The <path>/<to>/main_config.ini must point to a valid configuration file on your machine, as a relative path to the /home/<username>/<resources_path> described above. For example, if the file is located on my machine under /home/myuser/Data/RADS/main_config.ini, and that /home/myuser/Data is the mounted resources partition mounted on the Docker image, the new relative path will be RADS/main_config.ini.
The <verbose> level can be selected from [debug, info, warning, error].

How to cite

If you are using Raidionics in your research, please cite the following references.

The final software including updated performance metrics for preoperative tumors and introducing postoperative tumor segmentation:

@article{bouget2023raidionics,
    author = {Bouget, David and Alsinan, Demah and Gaitan, Valeria and Holden Helland, Ragnhild and Pedersen, André and Solheim, Ole and Reinertsen, Ingerid},
    year = {2023},
    month = {09},
    pages = {},
    title = {Raidionics: an open software for pre-and postoperative central nervous system tumor segmentation and standardized reporting},
    volume = {13},
    journal = {Scientific Reports},
    doi = {10.1038/s41598-023-42048-7},
}

For the preliminary preoperative tumor segmentation validation and software features:

@article{bouget2022preoptumorseg,
    title={Preoperative Brain Tumor Imaging: Models and Software for Segmentation and Standardized Reporting},
    author={Bouget, David and Pedersen, André and Jakola, Asgeir S. and Kavouridis, Vasileios and Emblem, Kyrre E. and Eijgelaar, Roelant S. and Kommers, Ivar and Ardon, Hilko and Barkhof, Frederik and Bello, Lorenzo and Berger, Mitchel S. and Conti Nibali, Marco and Furtner, Julia and Hervey-Jumper, Shawn and Idema, Albert J. S. and Kiesel, Barbara and Kloet, Alfred and Mandonnet, Emmanuel and Müller, Domenique M. J. and Robe, Pierre A. and Rossi, Marco and Sciortino, Tommaso and Van den Brink, Wimar A. and Wagemakers, Michiel and Widhalm, Georg and Witte, Marnix G. and Zwinderman, Aeilko H. and De Witt Hamer, Philip C. and Solheim, Ole and Reinertsen, Ingerid},
    journal={Frontiers in Neurology},
    volume={13},
    year={2022},
    url={https://www.frontiersin.org/articles/10.3389/fneur.2022.932219},
    doi={10.3389/fneur.2022.932219},
    issn={1664-2295}
}

Models

The trained models are automatically downloaded when running Raidionics or Raidionics-Slicer. Alternatively, all existing Raidionics models can be browsed here directly.

Developers

git clone https://github.com/dbouget/raidionics_rads_lib.git --recurse-submodules

For running inference on GPU through the raidionics_seg_lib backend, your machine must be properly configured (cf. here)

The ANTs library can be manually installed (from source) and be used as a cpp backend rather than Python. Visit https://github.com/ANTsX/ANTs.