Unsupervised Similarity Learning for Image Registration with Energy-Based Models

Daniel Grzech, Loïc Le Folgoc, Mohammad Farid Azampour, Athanasios Vlontzos, Ben Glocker, Nassir Navab, Julia Schnabel, Bernhard Kainz

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

We present a new model for deformable image registration, which learns in an unsupervised way a data-specific similarity metric. The proposed method consists of two neural networks, one that maps pairs of input images to transformations which align them, and one that provides the similarity metric whose maximisation guides the image alignment. We parametrise the similarity metric as an energy-based model, which is simple to train and allows us to improve the accuracy of image registration compared to other models with learnt similarity metrics by taking advantage of a more general mathematical formulation, as well as larger datasets. We also achieve substantial improvement in the accuracy of inter-patient image registration on MRI scans from the OASIS dataset compared to models that rely on traditional functions.

Original languageEnglish
Title of host publicationBiomedical Image Registration - 11th International Workshop, WBIR 2024, Held in Conjunction with MICCAI 2024, Proceedings
EditorsMarc Modat, Žiga Špiclin, Alessa Hering, Ivor Simpson, Wietske Bastiaansen, Tony C. W. Mok
PublisherSpringer Science and Business Media Deutschland GmbH
Pages229-240
Number of pages12
ISBN (Print)9783031734793
DOIs
StatePublished - 2024
Event11th International Workshop on Biomedical Image Registration, WBIR 2024, held in conjunction with the 27th International Conference on Medical Image Computing and Computer Assisted Intervention, MICCAI 2024 - Marrakesh, Morocco
Duration: 6 Oct 20246 Oct 2024

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume15249 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference11th International Workshop on Biomedical Image Registration, WBIR 2024, held in conjunction with the 27th International Conference on Medical Image Computing and Computer Assisted Intervention, MICCAI 2024
Country/TerritoryMorocco
CityMarrakesh
Period6/10/246/10/24

Keywords

  • energy-based models
  • image registration

Fingerprint

Dive into the research topics of 'Unsupervised Similarity Learning for Image Registration with Energy-Based Models'. Together they form a unique fingerprint.

Cite this