Brain tumor segmentation using two-stage convolutional neural network for federated evaluation

Kamlesh Pawar; Shenjun Zhong; Zhaolin Chen; Gary Egan

doi:10.1007/978-3-031-09002-8_43

Brain tumor segmentation using two-stage convolutional neural network for federated evaluation

Kamlesh Pawar
, Shenjun Zhong
, Zhaolin Chen
, Gary Egan

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

1 Citation (Scopus)

Abstract

A deep learning method is proposed for brain tumor segmentation using a two-stage encoder-decoder convolutional neural network (CNN). To improve the generalization of the proposed network for federated evaluation, we propose a two-stage encoder-decoder CNN that performs coarse segmentation at stage-I and fine segmentation at stage-II. Stage-I consists of an ensemble of three predictions on the orthogonal slices of a subject. In stage-II, the predictions of the first stage are used to crop the region of interest consisting of the tumor region and a fine grain segmentation is performed on the cropped image. A single ResUNet was used for stage-I and seven different networks were used for stage-II. Heavy data augmentation consisting of geometric transformation and random contrast was used to avoid overfitting and improve the generalization. The mean dice scores on 21 imaging sites evaluated in a federated manner achieved dice scores of 0.8659, 0.7708, and 0.7714 for the whole tumor, tumor core, and enhancing tumor respectively. The method ranked second in the federated evaluation task.

Original language	English
Title of host publication	Brainlesion
Subtitle of host publication	Multiple Sclerosis, Stroke and Traumatic Brain Injuries
Editors	Alessandro Crimi, Spyridon Bakas
Place of Publication	Cham Switzerland
Publisher	Springer
Pages	494-505
Number of pages	12
Edition	1st
ISBN (Electronic)	9783031090028
ISBN (Print)	9783031090011
DOIs	https://doi.org/10.1007/978-3-031-09002-8_43
Publication status	Published - 2022
Event	International MICCAI Brain Lesion Workshop 2021 - Online, United States of America Duration: 27 Sept 2021 → 27 Sept 2021 Conference number: 7th https://link.springer.com/book/10.1007/978-3-031-08999-2 (Proceedings)

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	12963 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	International MICCAI Brain Lesion Workshop 2021
Abbreviated title	BrainLes 2021
Country/Territory	United States of America
Period	27/09/21 → 27/09/21
Other	held in conjunction with the Medical Image Computing for Computer Assisted Intervention, MICCAI 2021
Internet address	https://link.springer.com/book/10.1007/978-3-031-08999-2 (Proceedings)

Keywords

Brain tumor segmentation
Convolutional neural network
Medical imaging

Access to Document

10.1007/978-3-031-09002-8_43

Cite this

Pawar, K., Zhong, S., Chen, Z., & Egan, G. (2022). Brain tumor segmentation using two-stage convolutional neural network for federated evaluation. In A. Crimi, & S. Bakas (Eds.), Brainlesion: Multiple Sclerosis, Stroke and Traumatic Brain Injuries (1st ed., pp. 494-505). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 12963 LNCS). Springer. https://doi.org/10.1007/978-3-031-09002-8_43

Pawar, Kamlesh ; Zhong, Shenjun ; Chen, Zhaolin et al. / Brain tumor segmentation using two-stage convolutional neural network for federated evaluation. Brainlesion: Multiple Sclerosis, Stroke and Traumatic Brain Injuries. editor / Alessandro Crimi ; Spyridon Bakas. 1st. ed. Cham Switzerland : Springer, 2022. pp. 494-505 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{4a4117f55a144bc0b220c4586dd6000e,

title = "Brain tumor segmentation using two-stage convolutional neural network for federated evaluation",

abstract = "A deep learning method is proposed for brain tumor segmentation using a two-stage encoder-decoder convolutional neural network (CNN). To improve the generalization of the proposed network for federated evaluation, we propose a two-stage encoder-decoder CNN that performs coarse segmentation at stage-I and fine segmentation at stage-II. Stage-I consists of an ensemble of three predictions on the orthogonal slices of a subject. In stage-II, the predictions of the first stage are used to crop the region of interest consisting of the tumor region and a fine grain segmentation is performed on the cropped image. A single ResUNet was used for stage-I and seven different networks were used for stage-II. Heavy data augmentation consisting of geometric transformation and random contrast was used to avoid overfitting and improve the generalization. The mean dice scores on 21 imaging sites evaluated in a federated manner achieved dice scores of 0.8659, 0.7708, and 0.7714 for the whole tumor, tumor core, and enhancing tumor respectively. The method ranked second in the federated evaluation task.",

keywords = "Brain tumor segmentation, Convolutional neural network, Medical imaging",

author = "Kamlesh Pawar and Shenjun Zhong and Zhaolin Chen and Gary Egan",

note = "Publisher Copyright: {\textcopyright} 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.; International MICCAI Brain Lesion Workshop 2021, BrainLes 2021 ; Conference date: 27-09-2021 Through 27-09-2021",

year = "2022",

doi = "10.1007/978-3-031-09002-8\_43",

language = "English",

isbn = "9783031090011",

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

publisher = "Springer",

pages = "494--505",

editor = "Alessandro Crimi and Spyridon Bakas",

booktitle = "Brainlesion",

address = "Switzerland",

edition = "1st",

url = "https://link.springer.com/book/10.1007/978-3-031-08999-2",

}

Pawar, K, Zhong, S , Chen, Z & Egan, G 2022, Brain tumor segmentation using two-stage convolutional neural network for federated evaluation. in A Crimi & S Bakas (eds), Brainlesion: Multiple Sclerosis, Stroke and Traumatic Brain Injuries. 1st edn, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 12963 LNCS, Springer, Cham Switzerland, pp. 494-505, International MICCAI Brain Lesion Workshop 2021, United States of America, 27/09/21. https://doi.org/10.1007/978-3-031-09002-8_43

Brain tumor segmentation using two-stage convolutional neural network for federated evaluation. / Pawar, Kamlesh; Zhong, Shenjun ; Chen, Zhaolin et al.
Brainlesion: Multiple Sclerosis, Stroke and Traumatic Brain Injuries. ed. / Alessandro Crimi; Spyridon Bakas. 1st. ed. Cham Switzerland: Springer, 2022. p. 494-505 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 12963 LNCS).

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

TY - GEN

T1 - Brain tumor segmentation using two-stage convolutional neural network for federated evaluation

AU - Pawar, Kamlesh

AU - Zhong, Shenjun

AU - Chen, Zhaolin

AU - Egan, Gary

N1 - Conference code: 7th

PY - 2022

Y1 - 2022

N2 - A deep learning method is proposed for brain tumor segmentation using a two-stage encoder-decoder convolutional neural network (CNN). To improve the generalization of the proposed network for federated evaluation, we propose a two-stage encoder-decoder CNN that performs coarse segmentation at stage-I and fine segmentation at stage-II. Stage-I consists of an ensemble of three predictions on the orthogonal slices of a subject. In stage-II, the predictions of the first stage are used to crop the region of interest consisting of the tumor region and a fine grain segmentation is performed on the cropped image. A single ResUNet was used for stage-I and seven different networks were used for stage-II. Heavy data augmentation consisting of geometric transformation and random contrast was used to avoid overfitting and improve the generalization. The mean dice scores on 21 imaging sites evaluated in a federated manner achieved dice scores of 0.8659, 0.7708, and 0.7714 for the whole tumor, tumor core, and enhancing tumor respectively. The method ranked second in the federated evaluation task.

AB - A deep learning method is proposed for brain tumor segmentation using a two-stage encoder-decoder convolutional neural network (CNN). To improve the generalization of the proposed network for federated evaluation, we propose a two-stage encoder-decoder CNN that performs coarse segmentation at stage-I and fine segmentation at stage-II. Stage-I consists of an ensemble of three predictions on the orthogonal slices of a subject. In stage-II, the predictions of the first stage are used to crop the region of interest consisting of the tumor region and a fine grain segmentation is performed on the cropped image. A single ResUNet was used for stage-I and seven different networks were used for stage-II. Heavy data augmentation consisting of geometric transformation and random contrast was used to avoid overfitting and improve the generalization. The mean dice scores on 21 imaging sites evaluated in a federated manner achieved dice scores of 0.8659, 0.7708, and 0.7714 for the whole tumor, tumor core, and enhancing tumor respectively. The method ranked second in the federated evaluation task.

KW - Brain tumor segmentation

KW - Convolutional neural network

KW - Medical imaging

UR - https://www.scopus.com/pages/publications/85135143596

U2 - 10.1007/978-3-031-09002-8_43

DO - 10.1007/978-3-031-09002-8_43

M3 - Conference Paper

AN - SCOPUS:85135143596

SN - 9783031090011

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

SP - 494

EP - 505

BT - Brainlesion

A2 - Crimi, Alessandro

A2 - Bakas, Spyridon

PB - Springer

CY - Cham Switzerland

T2 - International MICCAI Brain Lesion Workshop 2021

Y2 - 27 September 2021 through 27 September 2021

ER -

Pawar K, Zhong S , Chen Z, Egan G. Brain tumor segmentation using two-stage convolutional neural network for federated evaluation. In Crimi A, Bakas S, editors, Brainlesion: Multiple Sclerosis, Stroke and Traumatic Brain Injuries. 1st ed. Cham Switzerland: Springer. 2022. p. 494-505. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-031-09002-8_43

Brain tumor segmentation using two-stage convolutional neural network for federated evaluation

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

ARC Centre of Excellence for Integrative Brain Function

Cite this

Brain tumor segmentation using two-stage convolutional neural network for federated evaluation

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Projects

ARC Centre of Excellence for Integrative Brain Function

Cite this