Multi-task learning for neonatal brain segmentation using 3d dense-unet with dense attention guided by geodesic distance

Toan Duc Bui; Li Wang; Jian Chen; Weili Lin; Gang Li; Dinggang Shen

doi:10.1007/978-3-030-33391-1_28

Multi-task learning for neonatal brain segmentation using 3d dense-unet with dense attention guided by geodesic distance

Toan Duc Bui, Li Wang, Jian Chen, Weili Lin, Gang Li, Dinggang Shen

Department of Brain and Cognitive Engineering

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

15 Citations (Scopus)

Abstract

The deep convolutional neural network has achieved outstanding performance on neonatal brain MRI tissue segmentation. However, it may fail to produce reasonable results on unseen datasets that have different imaging appearance distributions with the training data. The main reason is that deep learning models tend to have a good fitting to the training dataset, but do not lead to a good generalization on the unseen datasets. To address this problem, we propose a multi-task learning method, which simultaneously learns both tissue segmentation and geodesic distance regression to regularize a shared encoder network. Furthermore, a dense attention gate is explored to force the network to learn rich contextual information. By using three neonatal brain datasets with different imaging protocols from different scanners, our experimental results demonstrate superior performance of our proposed method over the existing deep learning-based methods on the unseen datasets.

Original language	English
Title of host publication	Domain Adaptation and Representation Transfer and Medical Image Learning with Less Labels and Imperfect Data First MICCAI Workshop, DART 2019 and First International Workshop, MIL3ID 2019 Shenzhen, Held in Conjunction with MICCAI 2019 Shenzhen, 2019 Proceedings
Editors	Qian Wang, Fausto Milletari, Nicola Rieke, Hien V. Nguyen, Badri Roysam, Shadi Albarqouni, M. Jorge Cardoso, Ziyue Xu, Konstantinos Kamnitsas, Vishal Patel, Steve Jiang, Kevin Zhou, Khoa Luu, Ngan Le
Publisher	Springer
Pages	243-251
Number of pages	9
ISBN (Print)	9783030333904
DOIs	https://doi.org/10.1007/978-3-030-33391-1_28
Publication status	Published - 2019
Event	1st MICCAI Workshop on Domain Adaptation and Representation Transfer, DART 2019, and the 1st International Workshop on Medical Image Learning with Less Labels and Imperfect Data, MIL3ID 2019, held in conjunction with 22nd International Conference on Medical Image Computing and Computer- Assisted Intervention, MICCAI 2019 - Shenzhen, China Duration: 2019 Oct 13 → 2019 Oct 17

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	11795 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	1st MICCAI Workshop on Domain Adaptation and Representation Transfer, DART 2019, and the 1st International Workshop on Medical Image Learning with Less Labels and Imperfect Data, MIL3ID 2019, held in conjunction with 22nd International Conference on Medical Image Computing and Computer- Assisted Intervention, MICCAI 2019
Country/Territory	China
City	Shenzhen
Period	19/10/13 → 19/10/17

Bibliographical note

Publisher Copyright:
© Springer Nature Switzerland AG 2019.

Keywords

Attention
Geodesic distance
Multi-task learning
Neonatal brain segmentation

ASJC Scopus subject areas

Theoretical Computer Science
General Computer Science

Access to Document

10.1007/978-3-030-33391-1_28

Cite this

Bui, T. D., Wang, L., Chen, J., Lin, W., Li, G., & Shen, D. (2019). Multi-task learning for neonatal brain segmentation using 3d dense-unet with dense attention guided by geodesic distance. In Q. Wang, F. Milletari, N. Rieke, H. V. Nguyen, B. Roysam, S. Albarqouni, M. J. Cardoso, Z. Xu, K. Kamnitsas, V. Patel, S. Jiang, K. Zhou, K. Luu, & N. Le (Eds.), Domain Adaptation and Representation Transfer and Medical Image Learning with Less Labels and Imperfect Data First MICCAI Workshop, DART 2019 and First International Workshop, MIL3ID 2019 Shenzhen, Held in Conjunction with MICCAI 2019 Shenzhen, 2019 Proceedings (pp. 243-251). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11795 LNCS). Springer. https://doi.org/10.1007/978-3-030-33391-1_28

Multi-task learning for neonatal brain segmentation using 3d dense-unet with dense attention guided by geodesic distance. / Bui, Toan Duc; Wang, Li; Chen, Jian et al.
Domain Adaptation and Representation Transfer and Medical Image Learning with Less Labels and Imperfect Data First MICCAI Workshop, DART 2019 and First International Workshop, MIL3ID 2019 Shenzhen, Held in Conjunction with MICCAI 2019 Shenzhen, 2019 Proceedings. ed. / Qian Wang; Fausto Milletari; Nicola Rieke; Hien V. Nguyen; Badri Roysam; Shadi Albarqouni; M. Jorge Cardoso; Ziyue Xu; Konstantinos Kamnitsas; Vishal Patel; Steve Jiang; Kevin Zhou; Khoa Luu; Ngan Le. Springer, 2019. p. 243-251 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11795 LNCS).

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

Bui, TD, Wang, L, Chen, J, Lin, W, Li, G & Shen, D 2019, Multi-task learning for neonatal brain segmentation using 3d dense-unet with dense attention guided by geodesic distance. in Q Wang, F Milletari, N Rieke, HV Nguyen, B Roysam, S Albarqouni, MJ Cardoso, Z Xu, K Kamnitsas, V Patel, S Jiang, K Zhou, K Luu & N Le (eds), Domain Adaptation and Representation Transfer and Medical Image Learning with Less Labels and Imperfect Data First MICCAI Workshop, DART 2019 and First International Workshop, MIL3ID 2019 Shenzhen, Held in Conjunction with MICCAI 2019 Shenzhen, 2019 Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 11795 LNCS, Springer, pp. 243-251, 1st MICCAI Workshop on Domain Adaptation and Representation Transfer, DART 2019, and the 1st International Workshop on Medical Image Learning with Less Labels and Imperfect Data, MIL3ID 2019, held in conjunction with 22nd International Conference on Medical Image Computing and Computer- Assisted Intervention, MICCAI 2019, Shenzhen, China, 19/10/13. https://doi.org/10.1007/978-3-030-33391-1_28

Bui TD, Wang L, Chen J, Lin W, Li G, Shen D. Multi-task learning for neonatal brain segmentation using 3d dense-unet with dense attention guided by geodesic distance. In Wang Q, Milletari F, Rieke N, Nguyen HV, Roysam B, Albarqouni S, Cardoso MJ, Xu Z, Kamnitsas K, Patel V, Jiang S, Zhou K, Luu K, Le N, editors, Domain Adaptation and Representation Transfer and Medical Image Learning with Less Labels and Imperfect Data First MICCAI Workshop, DART 2019 and First International Workshop, MIL3ID 2019 Shenzhen, Held in Conjunction with MICCAI 2019 Shenzhen, 2019 Proceedings. Springer. 2019. p. 243-251. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-030-33391-1_28

Bui, Toan Duc ; Wang, Li ; Chen, Jian et al. / Multi-task learning for neonatal brain segmentation using 3d dense-unet with dense attention guided by geodesic distance. Domain Adaptation and Representation Transfer and Medical Image Learning with Less Labels and Imperfect Data First MICCAI Workshop, DART 2019 and First International Workshop, MIL3ID 2019 Shenzhen, Held in Conjunction with MICCAI 2019 Shenzhen, 2019 Proceedings. editor / Qian Wang ; Fausto Milletari ; Nicola Rieke ; Hien V. Nguyen ; Badri Roysam ; Shadi Albarqouni ; M. Jorge Cardoso ; Ziyue Xu ; Konstantinos Kamnitsas ; Vishal Patel ; Steve Jiang ; Kevin Zhou ; Khoa Luu ; Ngan Le. Springer, 2019. pp. 243-251 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{a8f49bc9c468420dad9e1905797b68ea,

title = "Multi-task learning for neonatal brain segmentation using 3d dense-unet with dense attention guided by geodesic distance",

abstract = "The deep convolutional neural network has achieved outstanding performance on neonatal brain MRI tissue segmentation. However, it may fail to produce reasonable results on unseen datasets that have different imaging appearance distributions with the training data. The main reason is that deep learning models tend to have a good fitting to the training dataset, but do not lead to a good generalization on the unseen datasets. To address this problem, we propose a multi-task learning method, which simultaneously learns both tissue segmentation and geodesic distance regression to regularize a shared encoder network. Furthermore, a dense attention gate is explored to force the network to learn rich contextual information. By using three neonatal brain datasets with different imaging protocols from different scanners, our experimental results demonstrate superior performance of our proposed method over the existing deep learning-based methods on the unseen datasets.",

keywords = "Attention, Geodesic distance, Multi-task learning, Neonatal brain segmentation",

author = "Bui, {Toan Duc} and Li Wang and Jian Chen and Weili Lin and Gang Li and Dinggang Shen",

note = "Publisher Copyright: {\textcopyright} Springer Nature Switzerland AG 2019.; 1st MICCAI Workshop on Domain Adaptation and Representation Transfer, DART 2019, and the 1st International Workshop on Medical Image Learning with Less Labels and Imperfect Data, MIL3ID 2019, held in conjunction with 22nd International Conference on Medical Image Computing and Computer- Assisted Intervention, MICCAI 2019 ; Conference date: 13-10-2019 Through 17-10-2019",

year = "2019",

doi = "10.1007/978-3-030-33391-1_28",

language = "English",

isbn = "9783030333904",

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

publisher = "Springer",

pages = "243--251",

editor = "Qian Wang and Fausto Milletari and Nicola Rieke and Nguyen, {Hien V.} and Badri Roysam and Shadi Albarqouni and Cardoso, {M. Jorge} and Ziyue Xu and Konstantinos Kamnitsas and Vishal Patel and Steve Jiang and Kevin Zhou and Khoa Luu and Ngan Le",

booktitle = "Domain Adaptation and Representation Transfer and Medical Image Learning with Less Labels and Imperfect Data First MICCAI Workshop, DART 2019 and First International Workshop, MIL3ID 2019 Shenzhen, Held in Conjunction with MICCAI 2019 Shenzhen, 2019 Proceedings",

}

TY - GEN

T1 - Multi-task learning for neonatal brain segmentation using 3d dense-unet with dense attention guided by geodesic distance

AU - Bui, Toan Duc

AU - Wang, Li

AU - Chen, Jian

AU - Lin, Weili

AU - Li, Gang

AU - Shen, Dinggang

N1 - Publisher Copyright: © Springer Nature Switzerland AG 2019.

PY - 2019

Y1 - 2019

N2 - The deep convolutional neural network has achieved outstanding performance on neonatal brain MRI tissue segmentation. However, it may fail to produce reasonable results on unseen datasets that have different imaging appearance distributions with the training data. The main reason is that deep learning models tend to have a good fitting to the training dataset, but do not lead to a good generalization on the unseen datasets. To address this problem, we propose a multi-task learning method, which simultaneously learns both tissue segmentation and geodesic distance regression to regularize a shared encoder network. Furthermore, a dense attention gate is explored to force the network to learn rich contextual information. By using three neonatal brain datasets with different imaging protocols from different scanners, our experimental results demonstrate superior performance of our proposed method over the existing deep learning-based methods on the unseen datasets.

AB - The deep convolutional neural network has achieved outstanding performance on neonatal brain MRI tissue segmentation. However, it may fail to produce reasonable results on unseen datasets that have different imaging appearance distributions with the training data. The main reason is that deep learning models tend to have a good fitting to the training dataset, but do not lead to a good generalization on the unseen datasets. To address this problem, we propose a multi-task learning method, which simultaneously learns both tissue segmentation and geodesic distance regression to regularize a shared encoder network. Furthermore, a dense attention gate is explored to force the network to learn rich contextual information. By using three neonatal brain datasets with different imaging protocols from different scanners, our experimental results demonstrate superior performance of our proposed method over the existing deep learning-based methods on the unseen datasets.

KW - Attention

KW - Geodesic distance

KW - Multi-task learning

KW - Neonatal brain segmentation

UR - http://www.scopus.com/inward/record.url?scp=85075659372&partnerID=8YFLogxK

U2 - 10.1007/978-3-030-33391-1_28

DO - 10.1007/978-3-030-33391-1_28

M3 - Conference contribution

AN - SCOPUS:85075659372

SN - 9783030333904

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

SP - 243

EP - 251

BT - Domain Adaptation and Representation Transfer and Medical Image Learning with Less Labels and Imperfect Data First MICCAI Workshop, DART 2019 and First International Workshop, MIL3ID 2019 Shenzhen, Held in Conjunction with MICCAI 2019 Shenzhen, 2019 Proceedings

A2 - Wang, Qian

A2 - Milletari, Fausto

A2 - Rieke, Nicola

A2 - Nguyen, Hien V.

A2 - Roysam, Badri

A2 - Albarqouni, Shadi

A2 - Cardoso, M. Jorge

A2 - Xu, Ziyue

A2 - Kamnitsas, Konstantinos

A2 - Patel, Vishal

A2 - Jiang, Steve

A2 - Zhou, Kevin

A2 - Luu, Khoa

A2 - Le, Ngan

PB - Springer

T2 - 1st MICCAI Workshop on Domain Adaptation and Representation Transfer, DART 2019, and the 1st International Workshop on Medical Image Learning with Less Labels and Imperfect Data, MIL3ID 2019, held in conjunction with 22nd International Conference on Medical Image Computing and Computer- Assisted Intervention, MICCAI 2019

Y2 - 13 October 2019 through 17 October 2019

ER -

Multi-task learning for neonatal brain segmentation using 3d dense-unet with dense attention guided by geodesic distance

Abstract

Publication series

Conference

Bibliographical note

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this