Inverse Consistency by Construction for Multistep Deep Registration

Hastings Greer; Lin Tian; Francois-Xavier Vialard; Roland Kwitt; Sylvain Bouix; Raul San Jose Estepar; Richard Rushmore; Marc Niethammer

doi:10.1007/978-3-031-43999-5_65

Inverse Consistency by Construction for Multistep Deep Registration

Hastings Greer, Lin Tian, Francois-Xavier Vialard, Roland Kwitt, Sylvain Bouix, Raul San Jose Estepar, Richard Rushmore, Marc Niethammer

Artificial Intelligence and Human Interfaces

Research output: Working paper/Preprint › Preprint

Abstract

Inverse consistency is a desirable property for image registration. We propose a simple technique to make a neural registration network inverse consistent by construction, as a consequence of its structure, as long as it parameterizes its output transform by a Lie group. We extend this technique to multi-step neural registration by composing many such networks in a way that preserves inverse consistency. This multi-step approach also allows for inverse-consistent coarse to fine registration. We evaluate our technique on synthetic 2-D data and four 3-D medical image registration tasks and obtain excellent registration accuracy while assuring inverse consistency.

Original language	English
Pages	688-698
Number of pages	11
DOIs	https://doi.org/10.1007/978-3-031-43999-5_65
Publication status	Published - 28 Apr 2023

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	14229 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Bibliographical note

Publisher Copyright:
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023.

Keywords

cs.CV
Registration
Deep Learning

Fields of Science and Technology Classification 2012

102 Computer Sciences

Access to Document

10.1007/978-3-031-43999-5_65

Cite this

Greer, H., Tian, L., Vialard, F.-X., Kwitt, R., Bouix, S., Estepar, R. S. J., Rushmore, R., & Niethammer, M. (2023). Inverse Consistency by Construction for Multistep Deep Registration. (pp. 688-698). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 14229 LNCS). https://doi.org/10.1007/978-3-031-43999-5_65

@techreport{8c5ce82a0d57484cb3421adaee8118c4,

title = "Inverse Consistency by Construction for Multistep Deep Registration",

abstract = "Inverse consistency is a desirable property for image registration. We propose a simple technique to make a neural registration network inverse consistent by construction, as a consequence of its structure, as long as it parameterizes its output transform by a Lie group. We extend this technique to multi-step neural registration by composing many such networks in a way that preserves inverse consistency. This multi-step approach also allows for inverse-consistent coarse to fine registration. We evaluate our technique on synthetic 2-D data and four 3-D medical image registration tasks and obtain excellent registration accuracy while assuring inverse consistency.",

keywords = "cs.CV, Registration, Deep Learning",

author = "Hastings Greer and Lin Tian and Francois-Xavier Vialard and Roland Kwitt and Sylvain Bouix and Estepar, {Raul San Jose} and Richard Rushmore and Marc Niethammer",

note = "Publisher Copyright: {\textcopyright} The Author(s), under exclusive license to Springer Nature Switzerland AG 2023.",

year = "2023",

month = apr,

day = "28",

doi = "10.1007/978-3-031-43999-5_65",

language = "English",

isbn = "9783031439988",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

pages = "688--698",

type = "WorkingPaper",

}

Greer, H, Tian, L, Vialard, F-X, Kwitt, R, Bouix, S, Estepar, RSJ, Rushmore, R & Niethammer, M 2023 'Inverse Consistency by Construction for Multistep Deep Registration' Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 14229 LNCS, pp. 688-698. https://doi.org/10.1007/978-3-031-43999-5_65

TY - UNPB

T1 - Inverse Consistency by Construction for Multistep Deep Registration

AU - Greer, Hastings

AU - Tian, Lin

AU - Vialard, Francois-Xavier

AU - Kwitt, Roland

AU - Bouix, Sylvain

AU - Estepar, Raul San Jose

AU - Rushmore, Richard

AU - Niethammer, Marc

PY - 2023/4/28

Y1 - 2023/4/28

N2 - Inverse consistency is a desirable property for image registration. We propose a simple technique to make a neural registration network inverse consistent by construction, as a consequence of its structure, as long as it parameterizes its output transform by a Lie group. We extend this technique to multi-step neural registration by composing many such networks in a way that preserves inverse consistency. This multi-step approach also allows for inverse-consistent coarse to fine registration. We evaluate our technique on synthetic 2-D data and four 3-D medical image registration tasks and obtain excellent registration accuracy while assuring inverse consistency.

AB - Inverse consistency is a desirable property for image registration. We propose a simple technique to make a neural registration network inverse consistent by construction, as a consequence of its structure, as long as it parameterizes its output transform by a Lie group. We extend this technique to multi-step neural registration by composing many such networks in a way that preserves inverse consistency. This multi-step approach also allows for inverse-consistent coarse to fine registration. We evaluate our technique on synthetic 2-D data and four 3-D medical image registration tasks and obtain excellent registration accuracy while assuring inverse consistency.

KW - cs.CV

KW - Registration

KW - Deep Learning

UR - http://www.scopus.com/inward/record.url?scp=85174693417&partnerID=8YFLogxK

U2 - 10.1007/978-3-031-43999-5_65

DO - 10.1007/978-3-031-43999-5_65

M3 - Preprint

SN - 9783031439988

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 688

EP - 698

BT - Inverse Consistency by Construction for Multistep Deep Registration

ER -

Inverse Consistency by Construction for Multistep Deep Registration

Abstract

Publication series

Bibliographical note

Keywords

Fields of Science and Technology Classification 2012

Access to Document

Other files and links

Deep Learning mit persistenter Homologie: Deep Homological Learning

Cite this

Inverse Consistency by Construction for Multistep Deep Registration

Abstract

Publication series

Bibliographical note

Keywords

Fields of Science and Technology Classification 2012

Access to Document

Other files and links

Projects

Deep Learning mit persistenter Homologie: Deep Homological Learning

Cite this