Deep adversarial networks for biomedical image segmentation utilizing unannotated images

Yizhe Zhang, Lin Yang, Jianxu Chen, Maridel Fredericksen, David Peter Hughes, Danny Z. Chen

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

28 Citations (Scopus)

Abstract

Semantic segmentation is a fundamental problem in biomedical image analysis. In biomedical practice, it is often the case that only limited annotated data are available for model training. Unannotated images, on the other hand, are easier to acquire. How to utilize unannotated images for training effective segmentation models is an important issue. In this paper, we propose a new deep adversarial network (DAN) model for biomedical image segmentation, aiming to attain consistently good segmentation results on both annotated and unannotated images. Our model consists of two networks: (1) a segmentation network (SN) to conduct segmentation; (2) an evaluation network (EN) to assess segmentation quality. During training, EN is encouraged to distinguish between segmentation results of unannotated images and annotated ones (by giving them different scores), while SN is encouraged to produce segmentation results of unannotated images such that EN cannot distinguish these from the annotated ones. Through an iterative adversarial training process, because EN is constantly “criticizing” the segmentation results of unannotated images, SN can be trained to produce more and more accurate segmentation for unannotated and unseen samples. Experiments show that our proposed DAN model is effective in utilizing unannotated image data to obtain considerably better segmentation.

Original languageEnglish (US)
Title of host publicationMedical Image Computing and Computer Assisted Intervention − MICCAI 2017 - 20th International Conference, Proceedings
EditorsLena Maier-Hein, Alfred Franz, Pierre Jannin, Simon Duchesne, Maxime Descoteaux, D. Louis Collins
PublisherSpringer Verlag
Pages408-416
Number of pages9
ISBN (Print)9783319661780
DOIs
StatePublished - Jan 1 2017
Event20th International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2017 - Quebec City, Canada
Duration: Sep 11 2017Sep 13 2017

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume10435 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Other

Other20th International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2017
CountryCanada
CityQuebec City
Period9/11/179/13/17

Fingerprint

Image segmentation
Image Segmentation
Segmentation
Image analysis
Evaluation
Network Model
Semantics
Image Analysis
Experiments
Model

All Science Journal Classification (ASJC) codes

  • Theoretical Computer Science
  • Computer Science(all)

Cite this

Zhang, Y., Yang, L., Chen, J., Fredericksen, M., Hughes, D. P., & Chen, D. Z. (2017). Deep adversarial networks for biomedical image segmentation utilizing unannotated images. In L. Maier-Hein, A. Franz, P. Jannin, S. Duchesne, M. Descoteaux, & D. L. Collins (Eds.), Medical Image Computing and Computer Assisted Intervention − MICCAI 2017 - 20th International Conference, Proceedings (pp. 408-416). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 10435 LNCS). Springer Verlag. https://doi.org/10.1007/978-3-319-66179-7_47
Zhang, Yizhe ; Yang, Lin ; Chen, Jianxu ; Fredericksen, Maridel ; Hughes, David Peter ; Chen, Danny Z. / Deep adversarial networks for biomedical image segmentation utilizing unannotated images. Medical Image Computing and Computer Assisted Intervention − MICCAI 2017 - 20th International Conference, Proceedings. editor / Lena Maier-Hein ; Alfred Franz ; Pierre Jannin ; Simon Duchesne ; Maxime Descoteaux ; D. Louis Collins. Springer Verlag, 2017. pp. 408-416 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{1aec5d0988e74a4f8ed637d769b9ae6a,
title = "Deep adversarial networks for biomedical image segmentation utilizing unannotated images",
abstract = "Semantic segmentation is a fundamental problem in biomedical image analysis. In biomedical practice, it is often the case that only limited annotated data are available for model training. Unannotated images, on the other hand, are easier to acquire. How to utilize unannotated images for training effective segmentation models is an important issue. In this paper, we propose a new deep adversarial network (DAN) model for biomedical image segmentation, aiming to attain consistently good segmentation results on both annotated and unannotated images. Our model consists of two networks: (1) a segmentation network (SN) to conduct segmentation; (2) an evaluation network (EN) to assess segmentation quality. During training, EN is encouraged to distinguish between segmentation results of unannotated images and annotated ones (by giving them different scores), while SN is encouraged to produce segmentation results of unannotated images such that EN cannot distinguish these from the annotated ones. Through an iterative adversarial training process, because EN is constantly “criticizing” the segmentation results of unannotated images, SN can be trained to produce more and more accurate segmentation for unannotated and unseen samples. Experiments show that our proposed DAN model is effective in utilizing unannotated image data to obtain considerably better segmentation.",
author = "Yizhe Zhang and Lin Yang and Jianxu Chen and Maridel Fredericksen and Hughes, {David Peter} and Chen, {Danny Z.}",
year = "2017",
month = "1",
day = "1",
doi = "10.1007/978-3-319-66179-7_47",
language = "English (US)",
isbn = "9783319661780",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer Verlag",
pages = "408--416",
editor = "Lena Maier-Hein and Alfred Franz and Pierre Jannin and Simon Duchesne and Maxime Descoteaux and Collins, {D. Louis}",
booktitle = "Medical Image Computing and Computer Assisted Intervention − MICCAI 2017 - 20th International Conference, Proceedings",
address = "Germany",

}

Zhang, Y, Yang, L, Chen, J, Fredericksen, M, Hughes, DP & Chen, DZ 2017, Deep adversarial networks for biomedical image segmentation utilizing unannotated images. in L Maier-Hein, A Franz, P Jannin, S Duchesne, M Descoteaux & DL Collins (eds), Medical Image Computing and Computer Assisted Intervention − MICCAI 2017 - 20th International Conference, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 10435 LNCS, Springer Verlag, pp. 408-416, 20th International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2017, Quebec City, Canada, 9/11/17. https://doi.org/10.1007/978-3-319-66179-7_47

Deep adversarial networks for biomedical image segmentation utilizing unannotated images. / Zhang, Yizhe; Yang, Lin; Chen, Jianxu; Fredericksen, Maridel; Hughes, David Peter; Chen, Danny Z.

Medical Image Computing and Computer Assisted Intervention − MICCAI 2017 - 20th International Conference, Proceedings. ed. / Lena Maier-Hein; Alfred Franz; Pierre Jannin; Simon Duchesne; Maxime Descoteaux; D. Louis Collins. Springer Verlag, 2017. p. 408-416 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 10435 LNCS).

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

TY - GEN

T1 - Deep adversarial networks for biomedical image segmentation utilizing unannotated images

AU - Zhang, Yizhe

AU - Yang, Lin

AU - Chen, Jianxu

AU - Fredericksen, Maridel

AU - Hughes, David Peter

AU - Chen, Danny Z.

PY - 2017/1/1

Y1 - 2017/1/1

N2 - Semantic segmentation is a fundamental problem in biomedical image analysis. In biomedical practice, it is often the case that only limited annotated data are available for model training. Unannotated images, on the other hand, are easier to acquire. How to utilize unannotated images for training effective segmentation models is an important issue. In this paper, we propose a new deep adversarial network (DAN) model for biomedical image segmentation, aiming to attain consistently good segmentation results on both annotated and unannotated images. Our model consists of two networks: (1) a segmentation network (SN) to conduct segmentation; (2) an evaluation network (EN) to assess segmentation quality. During training, EN is encouraged to distinguish between segmentation results of unannotated images and annotated ones (by giving them different scores), while SN is encouraged to produce segmentation results of unannotated images such that EN cannot distinguish these from the annotated ones. Through an iterative adversarial training process, because EN is constantly “criticizing” the segmentation results of unannotated images, SN can be trained to produce more and more accurate segmentation for unannotated and unseen samples. Experiments show that our proposed DAN model is effective in utilizing unannotated image data to obtain considerably better segmentation.

AB - Semantic segmentation is a fundamental problem in biomedical image analysis. In biomedical practice, it is often the case that only limited annotated data are available for model training. Unannotated images, on the other hand, are easier to acquire. How to utilize unannotated images for training effective segmentation models is an important issue. In this paper, we propose a new deep adversarial network (DAN) model for biomedical image segmentation, aiming to attain consistently good segmentation results on both annotated and unannotated images. Our model consists of two networks: (1) a segmentation network (SN) to conduct segmentation; (2) an evaluation network (EN) to assess segmentation quality. During training, EN is encouraged to distinguish between segmentation results of unannotated images and annotated ones (by giving them different scores), while SN is encouraged to produce segmentation results of unannotated images such that EN cannot distinguish these from the annotated ones. Through an iterative adversarial training process, because EN is constantly “criticizing” the segmentation results of unannotated images, SN can be trained to produce more and more accurate segmentation for unannotated and unseen samples. Experiments show that our proposed DAN model is effective in utilizing unannotated image data to obtain considerably better segmentation.

UR - http://www.scopus.com/inward/record.url?scp=85029530040&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85029530040&partnerID=8YFLogxK

U2 - 10.1007/978-3-319-66179-7_47

DO - 10.1007/978-3-319-66179-7_47

M3 - Conference contribution

AN - SCOPUS:85029530040

SN - 9783319661780

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

SP - 408

EP - 416

BT - Medical Image Computing and Computer Assisted Intervention − MICCAI 2017 - 20th International Conference, Proceedings

A2 - Maier-Hein, Lena

A2 - Franz, Alfred

A2 - Jannin, Pierre

A2 - Duchesne, Simon

A2 - Descoteaux, Maxime

A2 - Collins, D. Louis

PB - Springer Verlag

ER -

Zhang Y, Yang L, Chen J, Fredericksen M, Hughes DP, Chen DZ. Deep adversarial networks for biomedical image segmentation utilizing unannotated images. In Maier-Hein L, Franz A, Jannin P, Duchesne S, Descoteaux M, Collins DL, editors, Medical Image Computing and Computer Assisted Intervention − MICCAI 2017 - 20th International Conference, Proceedings. Springer Verlag. 2017. p. 408-416. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/978-3-319-66179-7_47