Posted in:

Description

The Facemasking pipeline is used to de-identify a subject's face.This is done by blurring facial characteristics that might make it possible to learn the identity of that subject. The process is also referred to as defacing.

Example:
FacemaskScreenshot.jpg

Can be run on

The Facemasking pipeline (FaceMasking.xml) can currently be run on MR images that have a T1 and/or T2 scan type.

Project Inputs

Name
Description
Allowed Values
usebet
Creates an exclusion brain mask
1 or 0
1=on, 0=off
maskears
Masks the ears
1 or 0
1=on, 0=off
invasiveness
Control the degree and depth of alteration of the image
0.5 to 1.5
threshold
Background intensity threshold
-1 to infinity
-1 for automatic
use_manual_roi
Determine if Manual ROI should be used
1 or 0
1=on, 0=off
rois
Masks in the reference space
Separate with underscores if more than one ROI
whole numbers
Format: x0,y0,z0-x1,y1,z1
should be 0 if previous is 0
dirs
Set direction for the manual ROI.
Separate with commas if more than one ROI
0,1,2
0=left/right, 1=top/bottom,
2=near/far
scanTypes
The label of the T1 and/or T2 in the session
scan type labels
format t1,t2
NOTE: ROI stands for Region of Interest

Launch Inputs

Name
Description
Allowed Values
Selected scans will be facemasked
Pick which scans to facemask
check one or more checkboxes
Choose reference scan
Select one scan to be used as a map
select one radio button
Generate an exclusion

check or uncheck
Enable ear masking
Select if ears should also be masked
check or uncheck
Enable manual ROI

check or uncheck
Invasiveness Coefficient
Control the degree and depth of alteration
0.5 to 1.5
Background-object intensity threshold

-1 to infinity
-1 for automatic
Send additional emails to
Enter any additional users to be notified when
facemasking is complete
email address
Separate with commas if more than one

Outputs

The following will result from a successful completion of the pipeline

  • The user who launched the pipeline will receive an email from the XNAT Team with the wording:
    "Face masked DICOM files have been generated for ..."

  • New files will be generated: DICOM files labelled DICOM_DEFACED as well as images labelled DEFACE_QC
    The files are viewable from the Manage Files area

    Deface output files.jpg


  • On the Session Details page, the Defaced snapshots will show below the original snapshots in scan types that were affected.
Facemask Session Screenshot.jpg



Additional Information

NRG Face Masking page

References

Milchenko, M. V. & Marcus, D. (2012). Obscuring surface anatomy in volumetric imaging data. Neuroinformatics. doi: 10.1007/s12021-012-9160-3