Refacing: Reconstructing Anonymized Facial Features Using Gans

Abramian, David; Eklund, Anders

doi:10.1101/447102

Cited by 22 publications

(27 citation statements)

References 23 publications

(24 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…However, it would be important to establish guidelines on how to make participants unrecognizable, specifically which parts of the face should be removed or otherwise processed to ensure participants' privacy. Moreover, for protecting participants' privacy, it may be important to take into account that reconstruction of removed or deformed facial features may be possible [31].…”

Section: Discussionmentioning

confidence: 99%

Facing privacy in neuroimaging: removing facial features degrades performance of image analysis methods

Sitter

Visser²,

Brouwer

et al. 2019

Eur Radiol

View full text Add to dashboard Cite

Background Recent studies have created awareness that facial features can be reconstructed from high-resolution MRI. Therefore, data sharing in neuroimaging requires special attention to protect participants' privacy. Facial features removal (FFR) could alleviate these concerns. We assessed the impact of three FFR methods on subsequent automated image analysis to obtain clinically relevant outcome measurements in three clinical groups. Methods FFR was performed using QuickShear, FaceMasking, and Defacing. In 110 subjects of Alzheimer's Disease Neuroimaging Initiative, normalized brain volumes (NBV) were measured by SIENAX. In 70 multiple sclerosis patients of the MAGNIMS Study Group, lesion volumes (WMLV) were measured by lesion prediction algorithm in lesion segmentation toolbox. In 84 glioblastoma patients of the PICTURE Study Group, tumor volumes (GBV) were measured by BraTumIA. Failed analyses on FFR-processed images were recorded. Only cases in which all image analyses completed successfully were analyzed. Differences between outcomes obtained from FFR-processed and full images were assessed, by quantifying the intra-class correlation coefficient (ICC) for absolute agreement and by testing for systematic differences using paired t tests.Results Automated analysis methods failed in 0-19% of cases in FFR-processed images versus 0-2% of cases in full images. ICC for absolute agreement ranged from 0.312 (GBV after FaceMasking) to 0.998 (WMLV after Defacing). FaceMasking yielded higher NBV (p = 0.003) and WMLV (p ≤ 0.001). GBV was lower after QuickShear and Defacing (both p < 0.001). Conclusions All three outcome measures were affected differently by FFR, including failure of analysis methods and both Brandom^variation and systematic differences. Further study is warranted to ensure high-quality neuroimaging research while protecting participants' privacy. Key Points • Protecting participants' privacy when sharing MRI data is important.• Impact of three facial features removal methods on subsequent analysis was assessed in three clinical groups.• Removing facial features degrades performance of image analysis methods.

show abstract

Section: Discussionmentioning

confidence: 99%

Facing privacy in neuroimaging: removing facial features degrades performance of image analysis methods

Sitter

Visser²,

Brouwer

et al. 2019

Eur Radiol

View full text Add to dashboard Cite

show abstract

“…Moreover, two recent developments further complicate matters. First, even defaced data may be re-identified, particularly if facial features are only blurred, as opposed to completely removed (Abramian & Eklund, 2019). Second, recent research indicates that defacing unfortunately may degrade the performance of image processing algorithms, possibly affecting the quality of measurements obtained (de Sitter et al, 2020).…”

Section: De-identificationmentioning

confidence: 99%

Mega-Analysis Methods in ENIGMA: The Experience of the Generalized Anxiety Disorder Working Group

Zugman¹,

Harrewijn²,

Cardinale³

et al. 2020

Preprint

View full text Add to dashboard Cite

The ENIGMA group on Generalized Anxiety Disorder (ENIGMA-Anxiety/GAD) is part of a broader effort to investigate anxiety disorders using imaging and genetic data across multiple sites worldwide. The group is actively conducting a mega-analysis of a large number of brain structural scans. In this process, the group was confronted with many methodological challenges related to study planning and implementation, between-country transfer of subject-level data, quality control of a considerable amount of imaging data, and choices related to statistical methods and efficient use of resources. This report summarizes the background information and rationale for the various methodological decisions, as well as the approach taken to implement them. The goal is to document the approach and help guide other research groups working with large brain imaging data sets as they develop their own analytic pipelines for mega-analyses.

show abstract

“…Future technological advances are a potential threat to the subject’s data privacy. Deep learning–based inpainting, which could undo retrospective image defacing to some extent, 18 could be applied to CHARISMA. Furthermore, access to privileged data repositories could uncover the subject’s identity by matching brain images or dental records.…”

Section: Discussionmentioning

confidence: 99%

Chemical shift–based prospective k‐space anonymization

Mattern

Knoll

Lüsebrink

et al. 2020

Magnetic Resonance in Med

View full text Add to dashboard Cite

Purpose Publicly available data provision is an essential part of open science. However, open data can conflict with data privacy and data protection regulations. Head scans are particularly vulnerable because the subject’s face can be reconstructed from the acquired images. Although defacing can impede subject identification in reconstructed images, this approach is not applicable to k‐space raw data. To address this challenge and allow defacing of raw data for publication, we present chemical shift–based prospective k‐space anonymization (CHARISMA). Methods In spin‐warp imaging, fat shift occurs along the frequency‐encoding direction. By placing an oil‐filled mask onto the subject’s face, the shifted fat signal can overlap with the face to deface k‐space during the acquisition. The CHARISMA approach was tested for gradient‐echo sequences in a single subject wearing the oil‐filled mask at 7 T. Different fat shifts were compared by varying the readout bandwidth. Furthermore, intensity‐based segmentation was used to test whether the images could be unmasked retrospectively. Results To impede subject identification after retrospective unmasking, the signal of face and shifted oil should overlap. In this single‐subject study, a shift of 3.3 mm to 4.9 mm resulted in the most efficient masking. Independent of CHARISMA, long TEs induce signal decay and dephasing, which impeded unmasking. Conclusion To our best knowledge, CHARISMA is the first prospective k‐space defacing approach. With proper fat‐shift direction and amplitude, this easy‐to‐build, low‐cost solution impaired subject identification in gradient‐echo data considerably. Further sequences will be tested with CHARISMA in the future.

show abstract

Refacing: Reconstructing Anonymized Facial Features Using Gans

Cited by 22 publications

References 23 publications

Facing privacy in neuroimaging: removing facial features degrades performance of image analysis methods

Facing privacy in neuroimaging: removing facial features degrades performance of image analysis methods

Mega-Analysis Methods in ENIGMA: The Experience of the Generalized Anxiety Disorder Working Group

Chemical shift–based prospective k‐space anonymization

Contact Info

Product

Resources

About