De-Identification of Medical Imaging Data: A Comprehensive Tool for Ensuring Patient Privacy

Getting Started • Usage • Citation

---

Important

The package is now available on PyPI: pip install mede

Note

MEDE now supports the Enhanced DICOM format!

---

This repository contains the De-Identification of Medical Imaging Data: A Comprehensive Tool for Ensuring Patient Privacy, which enables users to anonymize a wide variety of medical imaging types, including:

• Magnetic Resonance Imaging (MRI)
• Computer Tomography (CT)
• Ultrasound (US)
• Whole Slide Images (WSI)
• MRI raw data (twix)

This tool combines multiple anonymization steps, including metadata deidentification, defacing, and skull-stripping, while being faster than current state-of-the-art deidentification tools.

---

Getting Started

You can install the anonymization tool directly via pip or Docker.

Installation via pip

Our tool is available via pip. You can install it with the following command:

pip install mede

Additional Dependencies for Text Removal

Warning

Since version 0.0.11 we use EasyOCR instead of Tesseract for text removal. If you want to use the text removal feature, you need to install EasyOCR and its dependencies. EasyOCR is installed automatically with the newer version of MEDE.

We also implement a manual text removal feature that can be used to remove any remaining text from the images. This feature is optional and can be enabled with the additional --refine flag. This will open an interactive window where you can manually select any remaining text/artifacts and remove them by drawing a bounding box around them.

To draw a bounding box, click and hold the left mouse button, drag to create a rectangle around the text/artifact you want to remove, and then release the mouse button. The selected area will be filled with black pixels to effectively remove the text/artifact from the image.

To reset the image to its original state, press the r key while the interactive window is open.

To save the changes, press space or enter.

---

The following installation is only necessary if you want to use the old text removal feature and have an older version of MEDE installed:

If you want to use the old text removal feature, you also need to install Google's Tesseract OCR engine. Follow the installation instructions for your operating system here.

• On Ubuntu:

  sudo apt install tesseract-ocr
  sudo apt install libtesseract-dev

• On macOS (via Homebrew):

  brew install tesseract

---

Installation via Docker

Alternatively, this tool is distributed via Docker. You can find the Docker images here. The Docker image is available for Linux-based (including macOS) amd64 and arm64 platforms.

Steps:

1. Pull the Docker image:

   docker pull morrempe/mede:[tag]   # Replace [tag] with either arm64 or amd64

2. Run the Docker container with an attached volume:
   Your data will be mounted in the data folder:

   docker run --rm -it -v [Path/to/your/data]:/data morrempe/mede:[tag]

3. Run the script with the corresponding CLI parameters:

   mede-deidentify [your flags]

---

Usage

De-Identification CLI

The mede-deidentify command-line interface (CLI) allows you to de-identify medical imaging data with various options. Below is the detailed usage guide:

mede-deidentify [-h] [-v | --verbose] [-t | --text-removal] [-i | --input]
                [-o OUTPUT] [--gpu] [-s | --skull_strip] [-de | --deface]
                [-tw | --twix] [-w | --wsi] [-r | --rename]
                [-p PROCESSES] 
                [-d {basicProfile,cleanDescOpt,cleanGraphOpt,cleanStructContOpt,
                     rtnDevIdOpt,rtnInstIdOpt,rtnLongFullDatesOpt,
                     rtnLongModifDatesOpt,rtnPatCharsOpt,rtnSafePrivOpt,
                     rtnUIDsOpt} ...]

Options

Option	Description
-h, --help	Show the help message and exit.
-v, --verbose	Enable verbose output.
-i INPUT, --input INPUT	Path to the input data.
-o OUTPUT, --output OUTPUT	Path to save the output data.
--gpu GPU	Specify the GPU device number (default: 0).
-s, --skull_strip	Perform skull stripping.
-de, --deface	Perform defacing.
-tw, --twix	Process MRI raw data (twix format) and anonymize metadata.
-w, --wsi	Process Whole Slide Images (WSI).
-t, --text-removal	Perform text removal.
--refine	Enable interactive refinement of text removal results.
-r, --rename	Rename files during processing.
-p PROCESSES, --processes PROCESSES	Number of processes to use for multiprocessing.
-d, --deidentification-profile	Specify one or more DICOM deidentification profiles to apply (see below).

---

De-Identification Profiles

The -d or --deidentification-profile option allows you to specify one or more DICOM deidentification profiles. Available profiles include:

• basicProfile
• cleanDescOpt
• cleanGraphOpt
• cleanStructContOpt
• rtnDevIdOpt
• rtnInstIdOpt
• rtnLongFullDatesOpt
• rtnLongModifDatesOpt
• rtnPatCharsOpt
• rtnSafePrivOpt
• rtnUIDsOpt

You can specify multiple profiles by separating them with spaces. For example:

mede-deidentify -d basicProfile cleanDescOpt

---

Example Usage

Here’s an example of how to use the CLI:

mede-deidentify -i /path/to/input -o /path/to/output -s -d basicProfile

This command will:

1. Take input data from /path/to/input.
2. Save the output to /path/to/output.
3. Apply skull stripping.
4. Use the basicProfile deidentification profile.

---

Citation

If you use our tool in your work, please cite us with the following BibTeX entry.

@article{rempe2025identification,
  title={De-identification of medical imaging data: a comprehensive tool for ensuring patient privacy},
  author={Rempe, Moritz and Heine, Lukas and Seibold, Constantin and H{\"o}rst, Fabian and Kleesiek, Jens},
  journal={European Radiology},
  pages={1--10},
  year={2025},
  publisher={Springer}
}