VinDr-Mammo: A large-scale benchmark dataset for computer-aided diagnosis in full-field digital mammography

Description

This repository provides code used for de-identification and stratification of the VinDr-Mammo dataset, which can be downloaded via our project on Physionet . Python script for visualization of DICOM image is also provided.

Installation

To install required package via Pip, run

pip install -r requirements.txt

De-identification

See the deidentification.py file for more details.

Data Stratification

Please refer to the stratification.py file and the split_data.ipynb notebook. You may need to change the GLOBAL_PATH and LOCAL_PATH variables in split_data.ipynb to proper paths to the annotations files.

Visualization

Change the dicom_path variable in the visualize.py file to your desired DICOM file for visualization.

python visualize.py

License

This source code in released under Apache 2.0 License.

Citing

If you use the VinDr-Mammo dataset in your research please use the following BibTeX for citation:

@article{Nguyen2022.03.07.22272009,
  author={Nguyen, Hieu T. 
    and Nguyen, Ha Q. 
    and Pham, Hieu H. 
    and Lam, Khanh 
    and Le, Linh T. 
    and Dao, Minh 
    and Vu, Van},
  title={VinDr-Mammo: A large-scale benchmark dataset for computer-aided diagnosis in full-field digital mammography},
  year={2022},
  doi={10.1101/2022.03.07.22272009},
  URL={https://www.medrxiv.org/content/early/2022/03/10/2022.03.07.22272009},
  journal={medRxiv}
}