Deformable image registration using convolutional neural networks

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

8 Citations (Scopus)

Abstract

Deformable image registration can be time-consuming and often needs extensive parameterization to perform well on a specific application. We present a step towards a registration framework based on a three-dimensional convolutional neural network. The network directly learns transformations between pairs of three-dimensional images. The outputs of the network are three maps for the x, y, and z components of a thin plate spline transformation grid. The network is trained on synthetic random transformations, which are applied to a small set of representative images for the desired application. Training therefore does not require manually annotated ground truth deformation information. The methodology is demonstrated on public data sets of inspiration-expiration lung CT image pairs, which come with annotated corresponding landmarks for evaluation of the registration accuracy. Advantages of this methodology are its fast registration times and its minimal parameterization.

LanguageEnglish
Title of host publicationMedical Imaging 2018 Image Processing
Place of PublicationBellingham
PublisherSPIE
Number of pages6
ISBN (Electronic)9781510616370
DOIs
StatePublished - 15 Mar 2018
Event2018 SPIE Medical Imaging: Image Processing - Houston, United States
Duration: 10 Feb 201815 Feb 2018

Publication series

NameProceedings of SPIE
Volume10574

Conference

Conference2018 SPIE Medical Imaging: Image Processing
CountryUnited States
CityHouston
Period10/02/1815/02/18

Fingerprint

Image registration
Parameterization
Neural networks
Three-Dimensional Imaging
parameterization
Splines
expiration
methodology
inspiration
Lung
landmarks
ground truth
thin plates
splines
lungs
education
grids
evaluation
output
Datasets

Keywords

  • convolutional networks
  • deformable image registration
  • machine learning
  • thoracic CT

Cite this

Eppenhof, K. A. J., Lafarge, M. W., Moeskops, P., Veta, M., & Pluim, J. P. W. (2018). Deformable image registration using convolutional neural networks. In Medical Imaging 2018 Image Processing [105740S] (Proceedings of SPIE; Vol. 10574). Bellingham: SPIE. DOI: 10.1117/12.2292443
Eppenhof, Koen A.J. ; Lafarge, Maxime W. ; Moeskops, Pim ; Veta, Mitko ; Pluim, Josien P.W./ Deformable image registration using convolutional neural networks. Medical Imaging 2018 Image Processing. Bellingham : SPIE, 2018. (Proceedings of SPIE).
@inproceedings{27852a303c084e77ac69168ecabd48f7,
title = "Deformable image registration using convolutional neural networks",
abstract = "Deformable image registration can be time-consuming and often needs extensive parameterization to perform well on a specific application. We present a step towards a registration framework based on a three-dimensional convolutional neural network. The network directly learns transformations between pairs of three-dimensional images. The outputs of the network are three maps for the x, y, and z components of a thin plate spline transformation grid. The network is trained on synthetic random transformations, which are applied to a small set of representative images for the desired application. Training therefore does not require manually annotated ground truth deformation information. The methodology is demonstrated on public data sets of inspiration-expiration lung CT image pairs, which come with annotated corresponding landmarks for evaluation of the registration accuracy. Advantages of this methodology are its fast registration times and its minimal parameterization.",
keywords = "convolutional networks, deformable image registration, machine learning, thoracic CT",
author = "Eppenhof, {Koen A.J.} and Lafarge, {Maxime W.} and Pim Moeskops and Mitko Veta and Pluim, {Josien P.W.}",
year = "2018",
month = "3",
day = "15",
doi = "10.1117/12.2292443",
language = "English",
series = "Proceedings of SPIE",
publisher = "SPIE",
booktitle = "Medical Imaging 2018 Image Processing",
address = "United States",

}

Eppenhof, KAJ, Lafarge, MW, Moeskops, P, Veta, M & Pluim, JPW 2018, Deformable image registration using convolutional neural networks. in Medical Imaging 2018 Image Processing., 105740S, Proceedings of SPIE, vol. 10574, SPIE, Bellingham, 2018 SPIE Medical Imaging: Image Processing, Houston, United States, 10/02/18. DOI: 10.1117/12.2292443

Deformable image registration using convolutional neural networks. / Eppenhof, Koen A.J.; Lafarge, Maxime W.; Moeskops, Pim; Veta, Mitko; Pluim, Josien P.W.

Medical Imaging 2018 Image Processing. Bellingham : SPIE, 2018. 105740S (Proceedings of SPIE; Vol. 10574).

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

TY - GEN

T1 - Deformable image registration using convolutional neural networks

AU - Eppenhof,Koen A.J.

AU - Lafarge,Maxime W.

AU - Moeskops,Pim

AU - Veta,Mitko

AU - Pluim,Josien P.W.

PY - 2018/3/15

Y1 - 2018/3/15

N2 - Deformable image registration can be time-consuming and often needs extensive parameterization to perform well on a specific application. We present a step towards a registration framework based on a three-dimensional convolutional neural network. The network directly learns transformations between pairs of three-dimensional images. The outputs of the network are three maps for the x, y, and z components of a thin plate spline transformation grid. The network is trained on synthetic random transformations, which are applied to a small set of representative images for the desired application. Training therefore does not require manually annotated ground truth deformation information. The methodology is demonstrated on public data sets of inspiration-expiration lung CT image pairs, which come with annotated corresponding landmarks for evaluation of the registration accuracy. Advantages of this methodology are its fast registration times and its minimal parameterization.

AB - Deformable image registration can be time-consuming and often needs extensive parameterization to perform well on a specific application. We present a step towards a registration framework based on a three-dimensional convolutional neural network. The network directly learns transformations between pairs of three-dimensional images. The outputs of the network are three maps for the x, y, and z components of a thin plate spline transformation grid. The network is trained on synthetic random transformations, which are applied to a small set of representative images for the desired application. Training therefore does not require manually annotated ground truth deformation information. The methodology is demonstrated on public data sets of inspiration-expiration lung CT image pairs, which come with annotated corresponding landmarks for evaluation of the registration accuracy. Advantages of this methodology are its fast registration times and its minimal parameterization.

KW - convolutional networks

KW - deformable image registration

KW - machine learning

KW - thoracic CT

UR - http://www.scopus.com/inward/record.url?scp=85047327907&partnerID=8YFLogxK

U2 - 10.1117/12.2292443

DO - 10.1117/12.2292443

M3 - Conference contribution

T3 - Proceedings of SPIE

BT - Medical Imaging 2018 Image Processing

PB - SPIE

CY - Bellingham

ER -

Eppenhof KAJ, Lafarge MW, Moeskops P, Veta M, Pluim JPW. Deformable image registration using convolutional neural networks. In Medical Imaging 2018 Image Processing. Bellingham: SPIE. 2018. 105740S. (Proceedings of SPIE). Available from, DOI: 10.1117/12.2292443