CTA-DEFACE

Deep learning-based defacing tool for CT angiography: CTA-DEFACE

This repository provides an easy to use tool based on nnUNet for automated de-identification of CT angiography images.

If you are using CTA-DEFACE, please cite the following publication:

Mahmutoglu MA, Rastogi A, Schell M, Foltyn-Dumitru M, Baumgartner M, Maier-Hein KH, Deike-Hofmann K, Radbruch A, Bendszus M, Brugnara G, Vollmuth P. 
Deep learning-based defacing tool for CT angiography: CTA-DEFACE. 
Eur Radiol Exp. 2024 Oct 9;8(1):111. 

doi: 10.1186/s41747-024-00510-9.

The example image above was rendered in 3D Slicer software using "CT-Muscle" display preset.

Key points:

• The developed ANN model (CTA-DEFACE) automatically generates facemasks for CT angiography images and subtracts them py replacing the mask with the 10th percentile in the CT image (representing void space/air).
• By means of graphics processing unit optimization, our model ensures rapid processing of medical images.
• Our model underwent external validation, underscoring its reliability for real-world application.

Installation Instructions

Since our model is heavily dependend on nnUNet, please visit their repository for installation instructions and also cite their paper:

Isensee, F., Jaeger, P. F., Kohl, S. A., Petersen, J., & Maier-Hein, K. H. (2021). nnU-Net: a self-configuring 
method for deep learning-based biomedical image segmentation. Nature methods, 18(2), 203-211.

How to use it

Please install nnunetv2 following the instructions here:

https://github.com/MIC-DKFZ/nnUNet

Clone this repository and add your images to the input folder. Create input, output and model folders in the same folder as the run_CTA-DEFACE.py.

mkdir input
mkdir output
mkdir model

Download the trained model from the following link and put the Dataset001_DEFACE folder inside the model folder.

https://drive.google.com/drive/folders/1k4o35Dkl7PWd2yvHqWA2ia-BNKrWBrqg?usp=sharing

Make sure the CT or CTA input image names end with _0000.nii.gz, which is important to be recognized by the model.

python run_CTA-DEFACE.py -i input -o output

The above command will look for all nifti files (*.nii.gz) in the input folder and save the defaced NIfTI files and the face mask in the output folder.

CAVE: Our model and python code was designed to run on Ubuntu in a preinstalled nnunet environment, please adjust accordingly in case you intend to use the code in other operating systems.