Learning an MR acquisition-invariant representation using Siamese neural networks

W.M. Kouw, M. Loog, L.W. Bartels, A.M. Mendrik

Research output: Chapter in Book/Conference proceedings/Edited volumeConference contributionScientificpeer-review

1 Citation (Scopus)

Abstract

Generalization of voxelwise classifiers is hampered by differences between MRI-scanners, e.g. different acquisition protocols and field strengths. To address this limitation, we propose a Siamese neural network (MRAI-net) that extracts acquisition-invariant feature vectors. These can consequently be used by task-specific methods, such as voxelwise classifiers for tissue segmentation. MRAI-net is tested on both simulated and real patient data. Experiments show that MRAI-net outperforms voxelwise classifiers trained on the source or target scanner data when a small number of labeled samples is available.
Original languageEnglish
Title of host publication2019 IEEE 16th International Symposium on Biomedical Imaging (ISBI 2019)
Subtitle of host publicationProceedings
Place of PublicationDanvers
PublisherIEEE
Pages364-367
Number of pages4
ISBN (Electronic)978-1-5386-3641-1
ISBN (Print)978-1-5386-3642-8
DOIs
Publication statusPublished - 2019
EventIEEE International Symposium on Biomedical Imaging, ISBI 2019 - Venice, Italy
Duration: 8 Apr 201911 Apr 2019
https://biomedicalimaging.org/2019/

Conference

ConferenceIEEE International Symposium on Biomedical Imaging, ISBI 2019
Abbreviated titleISBI'19
CountryItaly
CityVenice
Period8/04/1911/04/19
Internet address

Keywords

  • MRI
  • Acquisition-variation
  • Representation-learning
  • Siamese-neural-network

Fingerprint Dive into the research topics of 'Learning an MR acquisition-invariant representation using Siamese neural networks'. Together they form a unique fingerprint.

Cite this