Mammography plugin for Orthanc

This plugin for Orthanc adds a deep learning algorithm to detect masses in scanned mammograms. The inference algorithm produces DICOM-SR (structured reports) instances that are stored inside Orthanc. The DICOM-SR instance can then be displayed using a version of the Stone Web viewer for Orthanc.

Reference paper, in the BibTeX format:

@inproceedings{OrthancMIE2024,
 title = {Integrated and Interoperable Platform for Detecting Masses on Mammograms},
 author = {Chatzopoulos, Edouard and Jodogne, S\'ebastien},
 booktitle = {34th Medical Informatics Europe Conference (MIE 2024)},
 address = {Athens, Greece},
 journal = {Studies in Health Technology and Informatics},
 volume = {316},
 pages = {1103-1107},
 issn = {1879-8365},
 doi = {10.3233/SHTI240603},
 publisher = {{IOS} Press},
 year = {2024}
}

This plugin is copyrighted by the UCLouvain university, under the AGPLv3+ license. It is a deliverable of the research lab led by Sébastien Jodogne.

Setup

Make sure that Python 3 is installed with support of virtual environments. Under Debian/Ubuntu distributions, this can be done by running:

$ sudo apt update
$ sudo apt install python3 python3-venv

Evidently, you also have to install Orthanc. The mammography plugin requires the two following plugins for Orthanc to be installed:

• Python plugin.
• DICOMweb plugin, if you want to use the DICOM-SR viewer.

Configuration

Firstly, clone this repository:

$ cd /tmp/
$ git clone https://github.com/jodogne/orthanc-mammography/

Secondly, setup the Python virtual environment with the required dependencies:

$ cd /tmp/orthanc-mammography/
$ python3 -m venv venv
$ source ./venv/bin/activate
(venv) python3 -m pip install --upgrade pip
(venv) python3 -m pip install -r requirements.txt

Thirdly, create a configuration file for Orthanc with Python. A minimalist configuration file would look like (make sure to adapt the path to the virtual environment, depending on your version of Python):

{
  "Plugins" : [
    "...list of paths to the Python and DICOMweb plugins..."
  ],
  "Python" : {
    "Path" : "./mammography.py"
  },
  "Mammography" : {
    "VirtualEnv" : "./venv/lib/python3.10/site-packages/"
  }
}

Finally, start Orthanc using this configuration file.

Usage

Two samples images from the CBIS-DDSM dataset are available in the ./sample-images/ folder.

Upload DICOM images of mammograms to Orthanc.

Using the built-in Orthanc Explorer interface of Orthanc, go down in the DICOM hierarchy until you reach a DICOM instance whose modality is MG (mammogram). The yellow button Deep learning for mammography will appear.

Clicking on the button will launch the deep learning inference. Once the inference is done (this takes about 5 seconds on a standard laptop), Orthanc Explorer is automatically redirected to the newly-created DICOM series that contains the DICOM-SR instance with the results of the inference.

You can then click on the yellow button Stone Web viewer (for mammography) to display the content of the DICOM-SR.