Automatic brain tumor segmentation with domain adaptation

Lutao Dai; Tengfei Li; Hai Shu; Liming Zhong; Haipeng Shen; Hongtu Zhu

doi:10.1007/978-3-030-11726-9_34

Automatic brain tumor segmentation with domain adaptation

Lutao Dai, Tengfei Li, Hai Shu, Liming Zhong, Haipeng Shen, Hongtu Zhu

Biostatistics

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

30 Scopus citations

Abstract

Deep convolution neural networks, in particular, the encoder-decoder networks, have been extensively used in image segmentation. We develop a deep learning approach for tumor segmentation by combining a modified U-Net and its domain-adapted version (DAU-Net). We divide training samples into two domains according to preliminary segmentation results, and then equip the modified U-Net with domain adaptation structure to obtain a domain invariant feature representation. Our proposed segmentation approach is applied to the BraTS 2018 challenge for brain tumor segmentation, and achieves the mean dice score of 0.91044, 0.85057 and 0.80536 for whole tumor, tumor core and enhancing tumor, respectively, on the challenge’s validation data set.

Original language	English (US)
Title of host publication	Brainlesion
Subtitle of host publication	Glioma, Multiple Sclerosis, Stroke and Traumatic Brain Injuries - 4th International Workshop, BrainLes 2018, Held in Conjunction with MICCAI 2018, Revised Selected Papers
Editors	Mauricio Reyes, Spyridon Bakas, Theo van Walsum, Alessandro Crimi, Farahani Keyvan, Hugo Kuijf
Publisher	Springer Verlag
Pages	380-392
Number of pages	13
ISBN (Print)	9783030117252
DOIs	https://doi.org/10.1007/978-3-030-11726-9_34
State	Published - 2019
Event	4th International MICCAI Brainlesion Workshop, BrainLes 2018 held in conjunction with the Medical Image Computing for Computer Assisted Intervention Conference, MICCAI 2018 - Granada, Spain Duration: Sep 16 2018 → Sep 20 2018

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	11384 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	4th International MICCAI Brainlesion Workshop, BrainLes 2018 held in conjunction with the Medical Image Computing for Computer Assisted Intervention Conference, MICCAI 2018
Country/Territory	Spain
City	Granada
Period	9/16/18 → 9/20/18

Keywords

Brain tumor
Confusion loss
Domain adaptation
Encoder-decoder network
Segmentation

ASJC Scopus subject areas

Theoretical Computer Science
General Computer Science

Access to Document

10.1007/978-3-030-11726-9_34

Cite this

Dai, L., Li, T., Shu, H., Zhong, L., Shen, H., & Zhu, H. (2019). Automatic brain tumor segmentation with domain adaptation. In M. Reyes, S. Bakas, T. van Walsum, A. Crimi, F. Keyvan, & H. Kuijf (Eds.), Brainlesion: Glioma, Multiple Sclerosis, Stroke and Traumatic Brain Injuries - 4th International Workshop, BrainLes 2018, Held in Conjunction with MICCAI 2018, Revised Selected Papers (pp. 380-392). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11384 LNCS). Springer Verlag. https://doi.org/10.1007/978-3-030-11726-9_34

Automatic brain tumor segmentation with domain adaptation. / Dai, Lutao; Li, Tengfei; Shu, Hai et al.
Brainlesion: Glioma, Multiple Sclerosis, Stroke and Traumatic Brain Injuries - 4th International Workshop, BrainLes 2018, Held in Conjunction with MICCAI 2018, Revised Selected Papers. ed. / Mauricio Reyes; Spyridon Bakas; Theo van Walsum; Alessandro Crimi; Farahani Keyvan; Hugo Kuijf. Springer Verlag, 2019. p. 380-392 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11384 LNCS).

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

Dai, L, Li, T, Shu, H, Zhong, L, Shen, H & Zhu, H 2019, Automatic brain tumor segmentation with domain adaptation. in M Reyes, S Bakas, T van Walsum, A Crimi, F Keyvan & H Kuijf (eds), Brainlesion: Glioma, Multiple Sclerosis, Stroke and Traumatic Brain Injuries - 4th International Workshop, BrainLes 2018, Held in Conjunction with MICCAI 2018, Revised Selected Papers. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 11384 LNCS, Springer Verlag, pp. 380-392, 4th International MICCAI Brainlesion Workshop, BrainLes 2018 held in conjunction with the Medical Image Computing for Computer Assisted Intervention Conference, MICCAI 2018, Granada, Spain, 9/16/18. https://doi.org/10.1007/978-3-030-11726-9_34

Dai L, Li T, Shu H, Zhong L, Shen H, Zhu H. Automatic brain tumor segmentation with domain adaptation. In Reyes M, Bakas S, van Walsum T, Crimi A, Keyvan F, Kuijf H, editors, Brainlesion: Glioma, Multiple Sclerosis, Stroke and Traumatic Brain Injuries - 4th International Workshop, BrainLes 2018, Held in Conjunction with MICCAI 2018, Revised Selected Papers. Springer Verlag. 2019. p. 380-392. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-030-11726-9_34

Dai, Lutao ; Li, Tengfei ; Shu, Hai et al. / Automatic brain tumor segmentation with domain adaptation. Brainlesion: Glioma, Multiple Sclerosis, Stroke and Traumatic Brain Injuries - 4th International Workshop, BrainLes 2018, Held in Conjunction with MICCAI 2018, Revised Selected Papers. editor / Mauricio Reyes ; Spyridon Bakas ; Theo van Walsum ; Alessandro Crimi ; Farahani Keyvan ; Hugo Kuijf. Springer Verlag, 2019. pp. 380-392 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{999687a4a4d94373a62ac23933683c9b,

title = "Automatic brain tumor segmentation with domain adaptation",

abstract = "Deep convolution neural networks, in particular, the encoder-decoder networks, have been extensively used in image segmentation. We develop a deep learning approach for tumor segmentation by combining a modified U-Net and its domain-adapted version (DAU-Net). We divide training samples into two domains according to preliminary segmentation results, and then equip the modified U-Net with domain adaptation structure to obtain a domain invariant feature representation. Our proposed segmentation approach is applied to the BraTS 2018 challenge for brain tumor segmentation, and achieves the mean dice score of 0.91044, 0.85057 and 0.80536 for whole tumor, tumor core and enhancing tumor, respectively, on the challenge{\textquoteright}s validation data set.",

keywords = "Brain tumor, Confusion loss, Domain adaptation, Encoder-decoder network, Segmentation",

author = "Lutao Dai and Tengfei Li and Hai Shu and Liming Zhong and Haipeng Shen and Hongtu Zhu",

note = "Funding Information: This research was partially supported by Ministry of Science and Technology Major Project of China 2017YFC1310903, University of Hong Kong Stanley Ho Alumni Challenge Fund, University Research Committee Seed Funding Award 104004215, US National Science Foundation grants DMS-1407655, and NIH grants MH086633 and MH116527. Funding Information: Acknowlegement. This research was partially supported by Ministry of Science and Technology Major Project of China 2017YFC1310903, University of Hong Kong Stanley Ho Alumni Challenge Fund, University Research Committee Seed Funding Award 104004215, US National Science Foundation grants DMS-1407655, and NIH grants MH086633 and MH116527. Publisher Copyright: {\textcopyright} Springer Nature Switzerland AG 2019.; 4th International MICCAI Brainlesion Workshop, BrainLes 2018 held in conjunction with the Medical Image Computing for Computer Assisted Intervention Conference, MICCAI 2018 ; Conference date: 16-09-2018 Through 20-09-2018",

year = "2019",

doi = "10.1007/978-3-030-11726-9_34",

language = "English (US)",

isbn = "9783030117252",

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

publisher = "Springer Verlag",

pages = "380--392",

editor = "Mauricio Reyes and Spyridon Bakas and {van Walsum}, Theo and Alessandro Crimi and Farahani Keyvan and Hugo Kuijf",

booktitle = "Brainlesion",

}

TY - GEN

T1 - Automatic brain tumor segmentation with domain adaptation

AU - Dai, Lutao

AU - Li, Tengfei

AU - Shu, Hai

AU - Zhong, Liming

AU - Shen, Haipeng

AU - Zhu, Hongtu

N1 - Funding Information: This research was partially supported by Ministry of Science and Technology Major Project of China 2017YFC1310903, University of Hong Kong Stanley Ho Alumni Challenge Fund, University Research Committee Seed Funding Award 104004215, US National Science Foundation grants DMS-1407655, and NIH grants MH086633 and MH116527. Funding Information: Acknowlegement. This research was partially supported by Ministry of Science and Technology Major Project of China 2017YFC1310903, University of Hong Kong Stanley Ho Alumni Challenge Fund, University Research Committee Seed Funding Award 104004215, US National Science Foundation grants DMS-1407655, and NIH grants MH086633 and MH116527. Publisher Copyright: © Springer Nature Switzerland AG 2019.

PY - 2019

Y1 - 2019

N2 - Deep convolution neural networks, in particular, the encoder-decoder networks, have been extensively used in image segmentation. We develop a deep learning approach for tumor segmentation by combining a modified U-Net and its domain-adapted version (DAU-Net). We divide training samples into two domains according to preliminary segmentation results, and then equip the modified U-Net with domain adaptation structure to obtain a domain invariant feature representation. Our proposed segmentation approach is applied to the BraTS 2018 challenge for brain tumor segmentation, and achieves the mean dice score of 0.91044, 0.85057 and 0.80536 for whole tumor, tumor core and enhancing tumor, respectively, on the challenge’s validation data set.

AB - Deep convolution neural networks, in particular, the encoder-decoder networks, have been extensively used in image segmentation. We develop a deep learning approach for tumor segmentation by combining a modified U-Net and its domain-adapted version (DAU-Net). We divide training samples into two domains according to preliminary segmentation results, and then equip the modified U-Net with domain adaptation structure to obtain a domain invariant feature representation. Our proposed segmentation approach is applied to the BraTS 2018 challenge for brain tumor segmentation, and achieves the mean dice score of 0.91044, 0.85057 and 0.80536 for whole tumor, tumor core and enhancing tumor, respectively, on the challenge’s validation data set.

KW - Brain tumor

KW - Confusion loss

KW - Domain adaptation

KW - Encoder-decoder network

KW - Segmentation

UR - http://www.scopus.com/inward/record.url?scp=85063466383&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85063466383&partnerID=8YFLogxK

U2 - 10.1007/978-3-030-11726-9_34

DO - 10.1007/978-3-030-11726-9_34

M3 - Conference contribution

AN - SCOPUS:85063466383

SN - 9783030117252

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

SP - 380

EP - 392

BT - Brainlesion

A2 - Reyes, Mauricio

A2 - Bakas, Spyridon

A2 - van Walsum, Theo

A2 - Crimi, Alessandro

A2 - Keyvan, Farahani

A2 - Kuijf, Hugo

PB - Springer Verlag

T2 - 4th International MICCAI Brainlesion Workshop, BrainLes 2018 held in conjunction with the Medical Image Computing for Computer Assisted Intervention Conference, MICCAI 2018

Y2 - 16 September 2018 through 20 September 2018

ER -

Automatic brain tumor segmentation with domain adaptation

Abstract

Publication series

Conference

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this