Face Masking

The identifying or sensitive anatomical features in MR and CT images used in research raise patient privacy concerns when such data are shared. The Face Masking package implements anatomical surface modification algorithm customized to de-identify MR head images that minimizes the impact on the resulting image statistics.

When to use

Face Masking obscures facial features from high-resolution MR scans that need to be shared, to reduce the research subject identification risk. Face Masking is minimally invasive and is designed to run with other processing tools such as skull stripping, MR artifact correction or segmentation.

Prerequisites

• Linux, Mac OS X or cygwin operating system
• MATLAB
• FSL 5.0 software library. FSL’s bin directory should be a part of your path. Please note that using FSL 6.0 may currently result in left-right flip of output DICOM’s.

Download and installation

• Download the Face Masking zipped package for your system and extract.
• Make sure your path and environment variables are set. Assuming the package is installed to /etc/maskface:bash/sh:

MASKFACE_HOME=/etc/maskface
PATH=${MASKFACE_HOME}/bin:${PATH}
export PATH MASKFACE_HOMEcsh/tcsh:
setenv MASKFACE_HOME /etc/maskface
setenv PATH ${MASKFACE_HOME}/bin:${PATH}

Running

For typical usage and running options, please refer to the mask_face usage manual page.

XNAT pipeline

Read this to configure face masking as an XNAT pipeline.

Running in Docker container

Assuming ‘subject_dicom’ is a directory in current directory with a DICOM series to be defaced:

>docker pull mohanar/facemasking:v1.12
>docker run –rm -v `pwd`:/tmp mohanar/facemasking:v1.12 /bin/bash -c “cd /tmp && pwd && source /opt/facemasking_launch_script/maskface_setup.sh && mask_face_nomatlab subject_dicom -b 1 && chmod -R 777 *”

Presentations

This video tutorial shows the use of the anonymization tools, DicomBrowser and Face Masking, for sharing data in XNAT.