Digital Mammography Correctness Assessment Tool examines submitted images for radio-opaque artifacts and provides an assessment of the quality of the data.
A GPU is recommended for performing inference using the pipeline.
To set up the MammoQC environment, first clone this repository and navigate to the project directory:
git clone https://github.com/DIDSR/MammoQC.git
cd MammoQCCreate a virtual environment and install dependencies from the provided requirements.txt:
conda create --name MammoQC_env python=3.12
conda activate MammoQC_env
pip install -r requirements.txtTested Environment:
- Linux (Ubuntu 22.04 LTS recommended)
- Python 3.12
Some key dependencies include:
opencv-python==4.13.0.90
torch==2.10.0
torchvision==0.25.0
pydicom==3.0.1
flask==3.1.3
Werkzeug==3.1.5
scikit-learn==1.8.0
transformers==5.0.0
scipy==1.17.0
notebook==7.3.2
ipykernel==6.29.5(See requirements.txt for the full list.)
There are several ways of using the MammoQC pipeline
The MammoQC_Demo.ipynb notebook can be used to test the inference pipeline using the pretrained model checkpoint. If the DICOM files from the EMBED dataset specified in data/test_data.csv are available in data/files, both the quality report and classification performance metrics can be generated. Otherwise, data_dir can be set to any directory containing DICOM or PNG files and artifact detection will be performed on the files found. The levels parameter in create_image_df() can be used to increase the nesting level of the search.
The main.py script is a streamlined and simple method of performing inference (on the test_data.csv or a specified data_dir).
The app.py script starts a browser-based Flask application (at the address http://localhost:5050/) which provides a user-friendly interface for using the MammoQC pipeline.
If you utilize MammoQC in your research or applications, please cite the repository:
@misc{MammoQC,
author = {Seyed M. Kahaki, Tahsin Rahman},
title = {MammoQC},
year = {2026},
publisher = {GitHub},
journal = {GitHub Repository},
howpublished = {\url{https://github.com/DIDSR/MammoQC}},
}Instructions for obtaining the EMBED dataset is available at the Emory-HITI/EMBED_Open_Data Github repository.
Radio-opqaue artifact labels for the EMBED dataset is available at the biomedia-mira/mammo-artifacts Github repository.
The finetuned DINOv2 model checkpoint will be available on Huggingface.
For any inquiries, suggestions, or collaborative opportunities, please contact Seyed Kahaki or Tahsin Rahman either via this GitHub repo or via email (seyed.kahaki@fda.hhs.gov;Tahsin.Rahman@fda.hhs.gov).
We warmly welcome pull requests and issues to enhance the project's capabilities and documentation.
The enclosed tool is part of the Catalog of Regulatory Science Tools, which provides a peer-reviewed resource for stakeholders to use where standards and qualified Medical Device Development Tools (MDDTs) do not yet exist. These tools do not replace FDA-recognized standards or MDDTs. This catalog collates a variety of regulatory science tools that the FDA’s Center for Devices and Radiological Health’s (CDRH) Office of Science and Engineering Labs (OSEL) developed. These tools use the most innovative science to support medical device development and patient access to safe and effective medical devices. If you are considering using a tool from this catalog in your marketing submissions, note that these tools have not been qualified as Medical Device Development Tools and the FDA has not evaluated the suitability of these tools within any specific context of use. You may request feedback or meetings for medical device submissions as part of the Q-Submission Program. For more information about the Catalog of Regulatory Science Tools, email RST_CDRH@fda.hhs.gov.
• RST Reference Number: RSTXXXX
• Date of Publication: XX/XX/XXXX
• Recommended Citation:
U.S. Food and Drug Administration. (2026). MammoQC: Digital Mamography Assessment Tool (xxxx.01). https://cdrh-rst.fda.gov/XXXX