Research
Print page Print page
Switch language
Rigshospitalet - a part of Copenhagen University Hospital
Published

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

Research output: Contribution to journalJournal articleResearchpeer-review

Harvard

MAGNIMS Study Group and Alzheimer’s Disease Neuroimaging Initiative 2020, 'Facing privacy in neuroimaging: removing facial features degrades performance of image analysis methods' European Radiology, vol. 30, no. 2, pp. 1062-1074. https://doi.org/10.1007/s00330-019-06459-3

APA

CBE

MLA

Vancouver

Author

MAGNIMS Study Group and Alzheimer’s Disease Neuroimaging Initiative. / Facing privacy in neuroimaging : removing facial features degrades performance of image analysis methods. In: European Radiology. 2020 ; Vol. 30, No. 2. pp. 1062-1074.

Bibtex

@article{a44055ebf5874125befaef3a97710e97,
title = "Facing privacy in neuroimaging: removing facial features degrades performance of image analysis methods",
abstract = "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 {"}random{"} 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.",
keywords = "Database, Ethics, Magnetic resonance imaging, Neuroimaging, Privacy, Reproducibility of Results, Humans, Middle Aged, Magnetic Resonance Imaging/methods, Male, Multiple Sclerosis/pathology, Tumor Burden, Information Dissemination, Glioblastoma/diagnostic imaging, Brain/diagnostic imaging, Algorithms, Neuroimaging/methods, Confidentiality, Aged, 80 and over, Female, Aged, Image Interpretation, Computer-Assisted/methods, Alzheimer Disease/pathology, Face",
author = "{de Sitter}, A and M Visser and I Brouwer and Cover, {K S} and {van Schijndel}, {R A} and Eijgelaar, {R S} and M{\"u}ller, {D M J} and S Ropele and L Kappos and {\'A} Rovira and M Filippi and C Enzinger and J Frederiksen and O Ciccarelli and Guttmann, {C R G} and Wattjes, {M P} and Witte, {M G} and {de Witt Hamer}, {P C} and F Barkhof and H Vrenken and {MAGNIMS Study Group and Alzheimer’s Disease Neuroimaging Initiative}",
year = "2020",
month = "2",
doi = "10.1007/s00330-019-06459-3",
language = "English",
volume = "30",
pages = "1062--1074",
journal = "European Radiology",
issn = "0938-7994",
publisher = "Springer",
number = "2",

}

RIS

TY - JOUR

T1 - Facing privacy in neuroimaging

T2 - removing facial features degrades performance of image analysis methods

AU - de Sitter, A

AU - Visser, M

AU - Brouwer, I

AU - Cover, K S

AU - van Schijndel, R A

AU - Eijgelaar, R S

AU - Müller, D M J

AU - Ropele, S

AU - Kappos, L

AU - Rovira, Á

AU - Filippi, M

AU - Enzinger, C

AU - Frederiksen, J

AU - Ciccarelli, O

AU - Guttmann, C R G

AU - Wattjes, M P

AU - Witte, M G

AU - de Witt Hamer, P C

AU - Barkhof, F

AU - Vrenken, H

AU - MAGNIMS Study Group and Alzheimer’s Disease Neuroimaging Initiative

PY - 2020/2

Y1 - 2020/2

N2 - 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 "random" 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.

AB - 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 "random" 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.

KW - Database

KW - Ethics

KW - Magnetic resonance imaging

KW - Neuroimaging

KW - Privacy

KW - Reproducibility of Results

KW - Humans

KW - Middle Aged

KW - Magnetic Resonance Imaging/methods

KW - Male

KW - Multiple Sclerosis/pathology

KW - Tumor Burden

KW - Information Dissemination

KW - Glioblastoma/diagnostic imaging

KW - Brain/diagnostic imaging

KW - Algorithms

KW - Neuroimaging/methods

KW - Confidentiality

KW - Aged, 80 and over

KW - Female

KW - Aged

KW - Image Interpretation, Computer-Assisted/methods

KW - Alzheimer Disease/pathology

KW - Face

UR - http://www.scopus.com/inward/record.url?scp=85074811129&partnerID=8YFLogxK

U2 - 10.1007/s00330-019-06459-3

DO - 10.1007/s00330-019-06459-3

M3 - Journal article

VL - 30

SP - 1062

EP - 1074

JO - European Radiology

JF - European Radiology

SN - 0938-7994

IS - 2

ER -

ID: 58366820