Adversarial training and dilated convolutions for brain MRI segmentation

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

21 Citations (Scopus)

Abstract

Convolutional neural networks (CNNs) have been applied to various automatic image segmentation tasks in medical image analysis, including brain MRI segmentation. Generative adversarial networks have recently gained popularity because of their power in generating images that are difficult to distinguish from real images. In this study we use an adversarial training approach to improve CNN-based brain MRI segmentation. To this end, we include an additional loss function that motivates the network to generate segmentations that are difficult to distinguish from manual segmentations. During training, this loss function is optimised together with the conventional average per-voxel cross entropy loss. The results show improved segmentation performance using this adversarial training procedure for segmentation of two different sets of images and using two different network architectures, both visually and in terms of Dice coefficients.

LanguageEnglish
Title of host publicationDeep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support - 3rd International Workshop, DLMIA 2017 and 7th International Workshop, ML-CDS 2017 Held in Conjunction with MICCAI 2017, Proceedings
PublisherSpringer
Pages56-64
Number of pages9
ISBN (Print)9783319675572
DOIs
StatePublished - 2017
Event3rd International Workshop on Deep Learning in Medical Image Analysis, DLMIA 2017 and 7th International Workshop on Multimodal Learning for Clinical Decision Support, ML-CDS 2017 held in Conjunction with 20th International Conference on Medical Image Computing and Computer Assisted Intervention, MICCAI 2017, 10-14 September Quebec, Canada - Quebec City, Canada
Duration: 10 Sep 201714 Sep 2017
http://www.miccai2017.org/

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume10553 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference3rd International Workshop on Deep Learning in Medical Image Analysis, DLMIA 2017 and 7th International Workshop on Multimodal Learning for Clinical Decision Support, ML-CDS 2017 held in Conjunction with 20th International Conference on Medical Image Computing and Computer Assisted Intervention, MICCAI 2017, 10-14 September Quebec, Canada
Abbreviated titleDLMIA2017
CountryCanada
CityQuebec City
Period10/09/1714/09/17
Internet address

Fingerprint

Convolution
Magnetic resonance imaging
Brain
Segmentation
Neural networks
Loss Function
Network architecture
Image segmentation
Image analysis
Entropy
Entropy Loss
Neural Networks
Medical Image Analysis
Dice
Cross-entropy
Voxel
Network Architecture
Image Segmentation
Training
Coefficient

Keywords

  • Adversarial networks
  • Brain MRI
  • Convolutional neural networks
  • Deep learning
  • Dilated convolution
  • Medical image segmentation

Cite this

Moeskops, P., Veta, M., Lafarge, M. W., Eppenhof, K. A. J., & Pluim, J. P. W. (2017). Adversarial training and dilated convolutions for brain MRI segmentation. In Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support - 3rd International Workshop, DLMIA 2017 and 7th International Workshop, ML-CDS 2017 Held in Conjunction with MICCAI 2017, Proceedings (pp. 56-64). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 10553 LNCS). Springer. DOI: 10.1007/978-3-319-67558-9_7
Moeskops, P. ; Veta, M. ; Lafarge, M.W. ; Eppenhof, K.A.J. ; Pluim, J.P.W./ Adversarial training and dilated convolutions for brain MRI segmentation. Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support - 3rd International Workshop, DLMIA 2017 and 7th International Workshop, ML-CDS 2017 Held in Conjunction with MICCAI 2017, Proceedings. Springer, 2017. pp. 56-64 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{f7ee0dee7b6e4736b47bb0bf06b2149f,
title = "Adversarial training and dilated convolutions for brain MRI segmentation",
abstract = "Convolutional neural networks (CNNs) have been applied to various automatic image segmentation tasks in medical image analysis, including brain MRI segmentation. Generative adversarial networks have recently gained popularity because of their power in generating images that are difficult to distinguish from real images. In this study we use an adversarial training approach to improve CNN-based brain MRI segmentation. To this end, we include an additional loss function that motivates the network to generate segmentations that are difficult to distinguish from manual segmentations. During training, this loss function is optimised together with the conventional average per-voxel cross entropy loss. The results show improved segmentation performance using this adversarial training procedure for segmentation of two different sets of images and using two different network architectures, both visually and in terms of Dice coefficients.",
keywords = "Adversarial networks, Brain MRI, Convolutional neural networks, Deep learning, Dilated convolution, Medical image segmentation",
author = "P. Moeskops and M. Veta and M.W. Lafarge and K.A.J. Eppenhof and J.P.W. Pluim",
year = "2017",
doi = "10.1007/978-3-319-67558-9_7",
language = "English",
isbn = "9783319675572",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer",
pages = "56--64",
booktitle = "Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support - 3rd International Workshop, DLMIA 2017 and 7th International Workshop, ML-CDS 2017 Held in Conjunction with MICCAI 2017, Proceedings",
address = "Germany",

}

Moeskops, P, Veta, M, Lafarge, MW, Eppenhof, KAJ & Pluim, JPW 2017, Adversarial training and dilated convolutions for brain MRI segmentation. in Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support - 3rd International Workshop, DLMIA 2017 and 7th International Workshop, ML-CDS 2017 Held in Conjunction with MICCAI 2017, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 10553 LNCS, Springer, pp. 56-64, 3rd International Workshop on Deep Learning in Medical Image Analysis, DLMIA 2017 and 7th International Workshop on Multimodal Learning for Clinical Decision Support, ML-CDS 2017 held in Conjunction with 20th International Conference on Medical Image Computing and Computer Assisted Intervention, MICCAI 2017, 10-14 September Quebec, Canada, Quebec City, Canada, 10/09/17. DOI: 10.1007/978-3-319-67558-9_7

Adversarial training and dilated convolutions for brain MRI segmentation. / Moeskops, P.; Veta, M.; Lafarge, M.W.; Eppenhof, K.A.J.; Pluim, J.P.W.

Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support - 3rd International Workshop, DLMIA 2017 and 7th International Workshop, ML-CDS 2017 Held in Conjunction with MICCAI 2017, Proceedings. Springer, 2017. p. 56-64 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 10553 LNCS).

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

TY - GEN

T1 - Adversarial training and dilated convolutions for brain MRI segmentation

AU - Moeskops,P.

AU - Veta,M.

AU - Lafarge,M.W.

AU - Eppenhof,K.A.J.

AU - Pluim,J.P.W.

PY - 2017

Y1 - 2017

N2 - Convolutional neural networks (CNNs) have been applied to various automatic image segmentation tasks in medical image analysis, including brain MRI segmentation. Generative adversarial networks have recently gained popularity because of their power in generating images that are difficult to distinguish from real images. In this study we use an adversarial training approach to improve CNN-based brain MRI segmentation. To this end, we include an additional loss function that motivates the network to generate segmentations that are difficult to distinguish from manual segmentations. During training, this loss function is optimised together with the conventional average per-voxel cross entropy loss. The results show improved segmentation performance using this adversarial training procedure for segmentation of two different sets of images and using two different network architectures, both visually and in terms of Dice coefficients.

AB - Convolutional neural networks (CNNs) have been applied to various automatic image segmentation tasks in medical image analysis, including brain MRI segmentation. Generative adversarial networks have recently gained popularity because of their power in generating images that are difficult to distinguish from real images. In this study we use an adversarial training approach to improve CNN-based brain MRI segmentation. To this end, we include an additional loss function that motivates the network to generate segmentations that are difficult to distinguish from manual segmentations. During training, this loss function is optimised together with the conventional average per-voxel cross entropy loss. The results show improved segmentation performance using this adversarial training procedure for segmentation of two different sets of images and using two different network architectures, both visually and in terms of Dice coefficients.

KW - Adversarial networks

KW - Brain MRI

KW - Convolutional neural networks

KW - Deep learning

KW - Dilated convolution

KW - Medical image segmentation

UR - http://www.scopus.com/inward/record.url?scp=85029815618&partnerID=8YFLogxK

U2 - 10.1007/978-3-319-67558-9_7

DO - 10.1007/978-3-319-67558-9_7

M3 - Conference contribution

SN - 9783319675572

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

SP - 56

EP - 64

BT - Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support - 3rd International Workshop, DLMIA 2017 and 7th International Workshop, ML-CDS 2017 Held in Conjunction with MICCAI 2017, Proceedings

PB - Springer

ER -

Moeskops P, Veta M, Lafarge MW, Eppenhof KAJ, Pluim JPW. Adversarial training and dilated convolutions for brain MRI segmentation. In Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support - 3rd International Workshop, DLMIA 2017 and 7th International Workshop, ML-CDS 2017 Held in Conjunction with MICCAI 2017, Proceedings. Springer. 2017. p. 56-64. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). Available from, DOI: 10.1007/978-3-319-67558-9_7