Unsupervised MR harmonization by learning disentangled representations using information bottleneck theory.
Disentangle
Harmonization
Image synthesis
Image-to-image translation
Magnetic resonance imaging
Journal
NeuroImage
ISSN: 1095-9572
Titre abrégé: Neuroimage
Pays: United States
ID NLM: 9215515
Informations de publication
Date de publication:
11 2021
11 2021
Historique:
received:
06
05
2021
revised:
11
08
2021
accepted:
07
09
2021
pubmed:
11
9
2021
medline:
22
1
2022
entrez:
10
9
2021
Statut:
ppublish
Résumé
In magnetic resonance (MR) imaging, a lack of standardization in acquisition often causes pulse sequence-based contrast variations in MR images from site to site, which impedes consistent measurements in automatic analyses. In this paper, we propose an unsupervised MR image harmonization approach, CALAMITI (Contrast Anatomy Learning and Analysis for MR Intensity Translation and Integration), which aims to alleviate contrast variations in multi-site MR imaging. Designed using information bottleneck theory, CALAMITI learns a globally disentangled latent space containing both anatomical and contrast information, which permits harmonization. In contrast to supervised harmonization methods, our approach does not need a sample population to be imaged across sites. Unlike traditional unsupervised harmonization approaches which often suffer from geometry shifts, CALAMITI better preserves anatomy by design. The proposed method is also able to adapt to a new testing site with a straightforward fine-tuning process. Experiments on MR images acquired from ten sites show that CALAMITI achieves superior performance compared with other harmonization approaches.
Identifiants
pubmed: 34506916
pii: S1053-8119(21)00842-9
doi: 10.1016/j.neuroimage.2021.118569
pmc: PMC10473284
mid: NIHMS1748167
pii:
doi:
Types de publication
Journal Article
Research Support, N.I.H., Intramural
Research Support, Non-U.S. Gov't
Langues
eng
Sous-ensembles de citation
IM
Pagination
118569Subventions
Organisme : Intramural NIH HHS
ID : ZIA AG000191
Pays : United States
Informations de copyright
Copyright © 2021. Published by Elsevier Inc.
Références
IEEE Trans Med Imaging. 2013 Dec;32(12):2348-63
pubmed: 24058022
Inf Process Med Imaging. 2021 Jun;12729:321-333
pubmed: 35173402
Neuroimage. 2010 Oct 1;52(4):1215-23
pubmed: 20441796
Mol Psychiatry. 2014 Jun;19(6):659-67
pubmed: 23774715
Neuroimage. 2019 Jul 1;194:105-119
pubmed: 30910724
Neuroimage. 2020 Mar;208:116450
pubmed: 31821869
Med Phys. 2020 Jan;47(1):89-98
pubmed: 31660621
IEEE Trans Image Process. 2004 Apr;13(4):600-12
pubmed: 15376593
Cereb Cortex. 2000 May;10(5):464-72
pubmed: 10847596
Med Image Anal. 2015 Aug;24(1):63-76
pubmed: 26072167
Neuroimage. 2020 Sep;218:116819
pubmed: 32438049
Neuroimage. 2017 Nov 1;161:149-170
pubmed: 28826946
Proc SPIE Int Soc Opt Eng. 2019 Mar;10949:
pubmed: 31551645
IEEE Trans Med Imaging. 2021 Mar;40(3):805-817
pubmed: 33170776
Neuroradiology. 1997 Aug;39(8):537-45
pubmed: 9272488
Magn Reson Med. 2020 Oct;84(4):2174-2189
pubmed: 32250475
Med Image Anal. 2019 Dec;58:101535
pubmed: 31351230
IEEE Trans Med Imaging. 2010 Jun;29(6):1310-20
pubmed: 20378467
Neural Netw. 2020 Nov;131:50-63
pubmed: 32759031
Proc IEEE Int Symp Biomed Imaging. 2021 Apr;2021:
pubmed: 34567419
Neuroimage. 2020 Oct 15;220:117127
pubmed: 32634595
Med Image Anal. 2021 Aug;72:102136
pubmed: 34246070
Magn Reson Imaging. 2019 Dec;64:160-170
pubmed: 31301354