Sparse-view CT Reconstruction using Wasserstein GANs

Franz Thaler, Kerstin Hammernik, Christian Payer, Martin Urschler, Darko Štern

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

We propose a 2D computed tomography (CT) slice image reconstruction method from a limited number of projection images using Wasserstein generative adversarial networks (wGAN). Our wGAN optimizes the 2D CT image reconstruction by utilizing an adversarial loss to improve the perceived image quality as well as an L1 content loss to enforce structural similarity to the target image. We evaluate our wGANs using different weight factors between the two loss functions and compare to a convolutional neural network (CNN) optimized on L1 and the Filtered Backprojection (FBP) method. The evaluation shows that the results generated by the machine learning based approaches are substantially better than those from the FBP method. In contrast to the blurrier looking images generated by the CNNs trained on L1, the wGANs results appear sharper and seem to contain more structural information. We show that a certain amount of projection data is needed to get a correct representation of the anatomical correspondences.

LanguageEnglish
Title of host publicationMachine Learning for Medical Image Reconstruction - First International Workshop, MLMIR 2018, Held in Conjunction with MICCAI 2018, Proceedings
PublisherSpringer Verlag Heidelberg
Pages75-82
Number of pages8
Volume11074 LNCS
ISBN (Print)9783030001285
DOIs
StatusPublished - 16 Sep 2018
Event1st Workshop on Machine Learning for Medical Image Reconstruction, MLMIR 2018 Held in Conjunction with 21st Medical Image Computing and Computer Assisted Intervention, MICCAI 2018 - Granada, Spain
Duration: 16 Sep 201816 Sep 2018

Publication series

NameLecture Notes in Computer Science
Volume11074
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference1st Workshop on Machine Learning for Medical Image Reconstruction, MLMIR 2018 Held in Conjunction with 21st Medical Image Computing and Computer Assisted Intervention, MICCAI 2018
CountrySpain
CityGranada
Period16/09/1816/09/18

Fingerprint

Computed Tomography
Filtered Backprojection
Tomography
Image Reconstruction
Image reconstruction
Projection
Structural Similarity
Loss Function
Slice
Image Quality
Image quality
Learning systems
Machine Learning
Correspondence
Optimise
Neural Networks
Neural networks
Target
Evaluate
Evaluation

Keywords

  • Computed tomography
  • Convolutional neural networks
  • Generative adversarial networks
  • L1 loss
  • Sparse-view reconstruction

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)

Fields of Expertise

  • Information, Communication & Computing

Cooperations

  • BioTechMed-Graz

Cite this

Thaler, F., Hammernik, K., Payer, C., Urschler, M., & Štern, D. (2018). Sparse-view CT Reconstruction using Wasserstein GANs. In Machine Learning for Medical Image Reconstruction - First International Workshop, MLMIR 2018, Held in Conjunction with MICCAI 2018, Proceedings (Vol. 11074 LNCS, pp. 75-82). (Lecture Notes in Computer Science ; Vol. 11074). Springer Verlag Heidelberg. DOI: 10.1007/978-3-030-00129-2_9

Sparse-view CT Reconstruction using Wasserstein GANs. / Thaler, Franz; Hammernik, Kerstin; Payer, Christian; Urschler, Martin; Štern, Darko.

Machine Learning for Medical Image Reconstruction - First International Workshop, MLMIR 2018, Held in Conjunction with MICCAI 2018, Proceedings. Vol. 11074 LNCS Springer Verlag Heidelberg, 2018. p. 75-82 (Lecture Notes in Computer Science ; Vol. 11074).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Thaler, F, Hammernik, K, Payer, C, Urschler, M & Štern, D 2018, Sparse-view CT Reconstruction using Wasserstein GANs. in Machine Learning for Medical Image Reconstruction - First International Workshop, MLMIR 2018, Held in Conjunction with MICCAI 2018, Proceedings. vol. 11074 LNCS, Lecture Notes in Computer Science , vol. 11074, Springer Verlag Heidelberg, pp. 75-82, 1st Workshop on Machine Learning for Medical Image Reconstruction, MLMIR 2018 Held in Conjunction with 21st Medical Image Computing and Computer Assisted Intervention, MICCAI 2018, Granada, Spain, 16/09/18. DOI: 10.1007/978-3-030-00129-2_9
Thaler F, Hammernik K, Payer C, Urschler M, Štern D. Sparse-view CT Reconstruction using Wasserstein GANs. In Machine Learning for Medical Image Reconstruction - First International Workshop, MLMIR 2018, Held in Conjunction with MICCAI 2018, Proceedings. Vol. 11074 LNCS. Springer Verlag Heidelberg. 2018. p. 75-82. (Lecture Notes in Computer Science ). Available from, DOI: 10.1007/978-3-030-00129-2_9
Thaler, Franz ; Hammernik, Kerstin ; Payer, Christian ; Urschler, Martin ; Štern, Darko. / Sparse-view CT Reconstruction using Wasserstein GANs. Machine Learning for Medical Image Reconstruction - First International Workshop, MLMIR 2018, Held in Conjunction with MICCAI 2018, Proceedings. Vol. 11074 LNCS Springer Verlag Heidelberg, 2018. pp. 75-82 (Lecture Notes in Computer Science ).
@inproceedings{089cfb30f6ef49fb9fa046c235ac5e5f,
title = "Sparse-view CT Reconstruction using Wasserstein GANs",
abstract = "We propose a 2D computed tomography (CT) slice image reconstruction method from a limited number of projection images using Wasserstein generative adversarial networks (wGAN). Our wGAN optimizes the 2D CT image reconstruction by utilizing an adversarial loss to improve the perceived image quality as well as an L1 content loss to enforce structural similarity to the target image. We evaluate our wGANs using different weight factors between the two loss functions and compare to a convolutional neural network (CNN) optimized on L1 and the Filtered Backprojection (FBP) method. The evaluation shows that the results generated by the machine learning based approaches are substantially better than those from the FBP method. In contrast to the blurrier looking images generated by the CNNs trained on L1, the wGANs results appear sharper and seem to contain more structural information. We show that a certain amount of projection data is needed to get a correct representation of the anatomical correspondences.",
keywords = "Computed tomography, Convolutional neural networks, Generative adversarial networks, L1 loss, Sparse-view reconstruction",
author = "Franz Thaler and Kerstin Hammernik and Christian Payer and Martin Urschler and Darko Štern",
year = "2018",
month = "9",
day = "16",
doi = "10.1007/978-3-030-00129-2_9",
language = "English",
isbn = "9783030001285",
volume = "11074 LNCS",
series = "Lecture Notes in Computer Science",
publisher = "Springer Verlag Heidelberg",
pages = "75--82",
booktitle = "Machine Learning for Medical Image Reconstruction - First International Workshop, MLMIR 2018, Held in Conjunction with MICCAI 2018, Proceedings",

}

TY - GEN

T1 - Sparse-view CT Reconstruction using Wasserstein GANs

AU - Thaler,Franz

AU - Hammernik,Kerstin

AU - Payer,Christian

AU - Urschler,Martin

AU - Štern,Darko

PY - 2018/9/16

Y1 - 2018/9/16

N2 - We propose a 2D computed tomography (CT) slice image reconstruction method from a limited number of projection images using Wasserstein generative adversarial networks (wGAN). Our wGAN optimizes the 2D CT image reconstruction by utilizing an adversarial loss to improve the perceived image quality as well as an L1 content loss to enforce structural similarity to the target image. We evaluate our wGANs using different weight factors between the two loss functions and compare to a convolutional neural network (CNN) optimized on L1 and the Filtered Backprojection (FBP) method. The evaluation shows that the results generated by the machine learning based approaches are substantially better than those from the FBP method. In contrast to the blurrier looking images generated by the CNNs trained on L1, the wGANs results appear sharper and seem to contain more structural information. We show that a certain amount of projection data is needed to get a correct representation of the anatomical correspondences.

AB - We propose a 2D computed tomography (CT) slice image reconstruction method from a limited number of projection images using Wasserstein generative adversarial networks (wGAN). Our wGAN optimizes the 2D CT image reconstruction by utilizing an adversarial loss to improve the perceived image quality as well as an L1 content loss to enforce structural similarity to the target image. We evaluate our wGANs using different weight factors between the two loss functions and compare to a convolutional neural network (CNN) optimized on L1 and the Filtered Backprojection (FBP) method. The evaluation shows that the results generated by the machine learning based approaches are substantially better than those from the FBP method. In contrast to the blurrier looking images generated by the CNNs trained on L1, the wGANs results appear sharper and seem to contain more structural information. We show that a certain amount of projection data is needed to get a correct representation of the anatomical correspondences.

KW - Computed tomography

KW - Convolutional neural networks

KW - Generative adversarial networks

KW - L1 loss

KW - Sparse-view reconstruction

UR - http://www.scopus.com/inward/record.url?scp=85053906671&partnerID=8YFLogxK

U2 - 10.1007/978-3-030-00129-2_9

DO - 10.1007/978-3-030-00129-2_9

M3 - Conference contribution

SN - 9783030001285

VL - 11074 LNCS

T3 - Lecture Notes in Computer Science

SP - 75

EP - 82

BT - Machine Learning for Medical Image Reconstruction - First International Workshop, MLMIR 2018, Held in Conjunction with MICCAI 2018, Proceedings

PB - Springer Verlag Heidelberg

ER -