Semi-supervised learning for network-based cardiac MR image segmentation

Wenjia Bai; Ozan Oktay; Matthew Sinclair; Hideaki Suzuki; Martin Rajchl; Giacomo Tarroni; Ben Glocker; Andrew King; Paul M. Matthews; Daniel Rueckert

doi:10.1007/978-3-319-66185-8_29

Semi-supervised learning for network-based cardiac MR image segmentation

Wenjia Bai
, Ozan Oktay
, Matthew Sinclair
, Hideaki Suzuki
, Martin Rajchl
, Giacomo Tarroni
, Ben Glocker
, Andrew King
, Paul M. Matthews
, Daniel Rueckert

Imperial College London
King's College London

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

435 Scopus citations

Abstract

Training a fully convolutional network for pixel-wise (or voxel-wise) image segmentation normally requires a large number of training images with corresponding ground truth label maps. However, it is a challenge to obtain such a large training set in the medical imaging domain, where expert annotations are time-consuming and difficult to obtain. In this paper, we propose a semi-supervised learning approach, in which a segmentation network is trained from both labelled and unlabelled data. The network parameters and the segmentations for the unlabelled data are alternately updated. We evaluate the method for short-axis cardiac MR image segmentation and it has demonstrated a high performance, outperforming a baseline supervised method. The mean Dice overlap metric is 0.92 for the left ventricular cavity, 0.85 for the myocardium and 0.89 for the right ventricular cavity. It also outperforms a state-of-the-art multi-atlas segmentation method by a large margin and the speed is substantially faster.

Original language	English
Title of host publication	Medical Image Computing and Computer Assisted Intervention − MICCAI 2017 - 20th International Conference, Proceedings
Editors	Pierre Jannin, Simon Duchesne, Maxime Descoteaux, Alfred Franz, D. Louis Collins, Lena Maier-Hein
Publisher	Springer Verlag
Pages	253-260
Number of pages	8
ISBN (Print)	9783319661841
DOIs	https://doi.org/10.1007/978-3-319-66185-8_29
State	Published - 2017
Externally published	Yes
Event	20th International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2017 - Quebec City, Canada Duration: 11 Sep 2017 → 13 Sep 2017

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	10434 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	20th International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2017
Country/Territory	Canada
City	Quebec City
Period	11/09/17 → 13/09/17

Access to Document

10.1007/978-3-319-66185-8_29

Cite this

Bai, W., Oktay, O., Sinclair, M., Suzuki, H., Rajchl, M., Tarroni, G., Glocker, B., King, A., Matthews, P. M., & Rueckert, D. (2017). Semi-supervised learning for network-based cardiac MR image segmentation. In P. Jannin, S. Duchesne, M. Descoteaux, A. Franz, D. L. Collins, & L. Maier-Hein (Eds.), Medical Image Computing and Computer Assisted Intervention − MICCAI 2017 - 20th International Conference, Proceedings (pp. 253-260). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 10434 LNCS). Springer Verlag. https://doi.org/10.1007/978-3-319-66185-8_29

Bai, Wenjia ; Oktay, Ozan ; Sinclair, Matthew et al. / Semi-supervised learning for network-based cardiac MR image segmentation. Medical Image Computing and Computer Assisted Intervention − MICCAI 2017 - 20th International Conference, Proceedings. editor / Pierre Jannin ; Simon Duchesne ; Maxime Descoteaux ; Alfred Franz ; D. Louis Collins ; Lena Maier-Hein. Springer Verlag, 2017. pp. 253-260 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{a9462a7a56534c7293b6482eaa6f9892,

title = "Semi-supervised learning for network-based cardiac MR image segmentation",

abstract = "Training a fully convolutional network for pixel-wise (or voxel-wise) image segmentation normally requires a large number of training images with corresponding ground truth label maps. However, it is a challenge to obtain such a large training set in the medical imaging domain, where expert annotations are time-consuming and difficult to obtain. In this paper, we propose a semi-supervised learning approach, in which a segmentation network is trained from both labelled and unlabelled data. The network parameters and the segmentations for the unlabelled data are alternately updated. We evaluate the method for short-axis cardiac MR image segmentation and it has demonstrated a high performance, outperforming a baseline supervised method. The mean Dice overlap metric is 0.92 for the left ventricular cavity, 0.85 for the myocardium and 0.89 for the right ventricular cavity. It also outperforms a state-of-the-art multi-atlas segmentation method by a large margin and the speed is substantially faster.",

author = "Wenjia Bai and Ozan Oktay and Matthew Sinclair and Hideaki Suzuki and Martin Rajchl and Giacomo Tarroni and Ben Glocker and Andrew King and Matthews, \{Paul M.\} and Daniel Rueckert",

note = "Publisher Copyright: {\textcopyright} Springer International Publishing AG 2017.; 20th International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2017 ; Conference date: 11-09-2017 Through 13-09-2017",

year = "2017",

doi = "10.1007/978-3-319-66185-8\_29",

language = "English",

isbn = "9783319661841",

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

publisher = "Springer Verlag",

pages = "253--260",

editor = "Pierre Jannin and Simon Duchesne and Maxime Descoteaux and Alfred Franz and Collins, \{D. Louis\} and Lena Maier-Hein",

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

}

Bai, W, Oktay, O, Sinclair, M, Suzuki, H, Rajchl, M, Tarroni, G, Glocker, B, King, A, Matthews, PM & Rueckert, D 2017, Semi-supervised learning for network-based cardiac MR image segmentation. in P Jannin, S Duchesne, M Descoteaux, A Franz, DL Collins & L Maier-Hein (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. 10434 LNCS, Springer Verlag, pp. 253-260, 20th International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2017, Quebec City, Canada, 11/09/17. https://doi.org/10.1007/978-3-319-66185-8_29

Semi-supervised learning for network-based cardiac MR image segmentation. / Bai, Wenjia; Oktay, Ozan; Sinclair, Matthew et al.
Medical Image Computing and Computer Assisted Intervention − MICCAI 2017 - 20th International Conference, Proceedings. ed. / Pierre Jannin; Simon Duchesne; Maxime Descoteaux; Alfred Franz; D. Louis Collins; Lena Maier-Hein. Springer Verlag, 2017. p. 253-260 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 10434 LNCS).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - Semi-supervised learning for network-based cardiac MR image segmentation

AU - Bai, Wenjia

AU - Oktay, Ozan

AU - Sinclair, Matthew

AU - Suzuki, Hideaki

AU - Rajchl, Martin

AU - Tarroni, Giacomo

AU - Glocker, Ben

AU - King, Andrew

AU - Matthews, Paul M.

AU - Rueckert, Daniel

N1 - Publisher Copyright: © Springer International Publishing AG 2017.

PY - 2017

Y1 - 2017

N2 - Training a fully convolutional network for pixel-wise (or voxel-wise) image segmentation normally requires a large number of training images with corresponding ground truth label maps. However, it is a challenge to obtain such a large training set in the medical imaging domain, where expert annotations are time-consuming and difficult to obtain. In this paper, we propose a semi-supervised learning approach, in which a segmentation network is trained from both labelled and unlabelled data. The network parameters and the segmentations for the unlabelled data are alternately updated. We evaluate the method for short-axis cardiac MR image segmentation and it has demonstrated a high performance, outperforming a baseline supervised method. The mean Dice overlap metric is 0.92 for the left ventricular cavity, 0.85 for the myocardium and 0.89 for the right ventricular cavity. It also outperforms a state-of-the-art multi-atlas segmentation method by a large margin and the speed is substantially faster.

AB - Training a fully convolutional network for pixel-wise (or voxel-wise) image segmentation normally requires a large number of training images with corresponding ground truth label maps. However, it is a challenge to obtain such a large training set in the medical imaging domain, where expert annotations are time-consuming and difficult to obtain. In this paper, we propose a semi-supervised learning approach, in which a segmentation network is trained from both labelled and unlabelled data. The network parameters and the segmentations for the unlabelled data are alternately updated. We evaluate the method for short-axis cardiac MR image segmentation and it has demonstrated a high performance, outperforming a baseline supervised method. The mean Dice overlap metric is 0.92 for the left ventricular cavity, 0.85 for the myocardium and 0.89 for the right ventricular cavity. It also outperforms a state-of-the-art multi-atlas segmentation method by a large margin and the speed is substantially faster.

UR - https://www.scopus.com/pages/publications/85029504302

U2 - 10.1007/978-3-319-66185-8_29

DO - 10.1007/978-3-319-66185-8_29

M3 - Conference contribution

AN - SCOPUS:85029504302

SN - 9783319661841

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

SP - 253

EP - 260

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

A2 - Jannin, Pierre

A2 - Duchesne, Simon

A2 - Descoteaux, Maxime

A2 - Franz, Alfred

A2 - Collins, D. Louis

A2 - Maier-Hein, Lena

PB - Springer Verlag

T2 - 20th International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2017

Y2 - 11 September 2017 through 13 September 2017

ER -

Bai W, Oktay O, Sinclair M, Suzuki H, Rajchl M, Tarroni G et al. Semi-supervised learning for network-based cardiac MR image segmentation. In Jannin P, Duchesne S, Descoteaux M, Franz A, Collins DL, Maier-Hein L, editors, Medical Image Computing and Computer Assisted Intervention − MICCAI 2017 - 20th International Conference, Proceedings. Springer Verlag. 2017. p. 253-260. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-319-66185-8_29

Semi-supervised learning for network-based cardiac MR image segmentation

Abstract

Publication series

Conference

Access to Document

Other files and links

Fingerprint

Cite this