Improving the segmentation of anatomical structures in chest radiographs using U-net with an imagenet pre-trained encoder

Maayan Frid-Adar; Avi Ben-Cohen; Rula Amer; Hayit Greenspan

doi:https://doi.org/10.1007/978-3-030-00946-5_17

Improving the segmentation of anatomical structures in chest radiographs using U-net with an imagenet pre-trained encoder

Maayan Frid-Adar, Avi Ben-Cohen, Rula Amer, Hayit Greenspan

Department of Bio-Medical Engineering

Tel Aviv University

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

Abstract

Accurate segmentation of anatomical structures in chest radiographs is essential for many computer-aided diagnosis tasks. In this paper we investigate the latest fully-convolutional architectures for the task of multi-class segmentation of the lungs field, heart and clavicles in a chest radiograph. In addition, we explore the influence of using different loss functions in the training process of a neural network for semantic segmentation. We evaluate all models on a common benchmark of 247 X-ray images from the JSRT database and ground-truth segmentation masks from the SCR dataset. Our best performing architecture, is a modified U-Net that benefits from pre-trained encoder weights. This model outperformed the current state-of-the-art methods tested on the same benchmark, with Jaccard overlap scores of 96.1% for lung fields, 90.6% for heart and 85.5% for clavicles.

Original language	English
Title of host publication	Image Analysis for Moving Organ, Breast, and Thoracic Images - Third International Workshop, RAMBO 2018, Fourth International Workshop, BIA 2018, and First International Workshop, TIA 2018, Held in Conjunction with MICCAI 2018, Proceedings
Editors	David Snead, Emanuele Trucco, Danail Stoyanov, Zeike Taylor, Lena Maier-Hein, Nasir Rajpoot, Hrvoje Bogunovic, Francesco Ciompi, Mitko Veta, Mona K. Garvin, Xin Jan Chen, Anne Martel, Jeroen van der Laak, Yanwu Xu, Stephen McKenna
Pages	159-168
Number of pages	10
DOIs	https://doi.org/10.1007/978-3-030-00946-5_17
State	Published - 2018
Event	3rd International Workshop on Reconstruction and Analysis of Moving Body Organs, RAMBO 2018, 4th International Workshop on Breast Image Analysis, BIA 2018, and 1st International Workshop on Thoracic Image Analysis, TIA 2018, held in conjunction with 21s... - Granada, Spain Duration: 16 Sep 2018 → 20 Sep 2018

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	11040 LNCS

Conference

Conference	3rd International Workshop on Reconstruction and Analysis of Moving Body Organs, RAMBO 2018, 4th International Workshop on Breast Image Analysis, BIA 2018, and 1st International Workshop on Thoracic Image Analysis, TIA 2018, held in conjunction with 21s...
Country/Territory	Spain
City	Granada
Period	16/09/18 → 20/09/18

Keywords

Chest radiographs
Clavicle segmentation
Fully convolutional networks
Heart segmentation
Lung segmentation

All Science Journal Classification (ASJC) codes

Theoretical Computer Science
General Computer Science

Access to Document

https://doi.org/10.1007/978-3-030-00946-5_17

Cite this

Frid-Adar, M., Ben-Cohen, A., Amer, R., & Greenspan, H. (2018). Improving the segmentation of anatomical structures in chest radiographs using U-net with an imagenet pre-trained encoder. In D. Snead, E. Trucco, D. Stoyanov, Z. Taylor, L. Maier-Hein, N. Rajpoot, H. Bogunovic, F. Ciompi, M. Veta, M. K. Garvin, X. J. Chen, A. Martel, J. van der Laak, Y. Xu, & S. McKenna (Eds.), Image Analysis for Moving Organ, Breast, and Thoracic Images - Third International Workshop, RAMBO 2018, Fourth International Workshop, BIA 2018, and First International Workshop, TIA 2018, Held in Conjunction with MICCAI 2018, Proceedings (pp. 159-168). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11040 LNCS). https://doi.org/10.1007/978-3-030-00946-5_17

Frid-Adar, Maayan ; Ben-Cohen, Avi ; Amer, Rula et al. / Improving the segmentation of anatomical structures in chest radiographs using U-net with an imagenet pre-trained encoder. Image Analysis for Moving Organ, Breast, and Thoracic Images - Third International Workshop, RAMBO 2018, Fourth International Workshop, BIA 2018, and First International Workshop, TIA 2018, Held in Conjunction with MICCAI 2018, Proceedings. editor / David Snead ; Emanuele Trucco ; Danail Stoyanov ; Zeike Taylor ; Lena Maier-Hein ; Nasir Rajpoot ; Hrvoje Bogunovic ; Francesco Ciompi ; Mitko Veta ; Mona K. Garvin ; Xin Jan Chen ; Anne Martel ; Jeroen van der Laak ; Yanwu Xu ; Stephen McKenna. 2018. pp. 159-168 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{25c1fe0fb41b4f19af168b99b4a8a6d9,

title = "Improving the segmentation of anatomical structures in chest radiographs using U-net with an imagenet pre-trained encoder",

abstract = "Accurate segmentation of anatomical structures in chest radiographs is essential for many computer-aided diagnosis tasks. In this paper we investigate the latest fully-convolutional architectures for the task of multi-class segmentation of the lungs field, heart and clavicles in a chest radiograph. In addition, we explore the influence of using different loss functions in the training process of a neural network for semantic segmentation. We evaluate all models on a common benchmark of 247 X-ray images from the JSRT database and ground-truth segmentation masks from the SCR dataset. Our best performing architecture, is a modified U-Net that benefits from pre-trained encoder weights. This model outperformed the current state-of-the-art methods tested on the same benchmark, with Jaccard overlap scores of 96.1% for lung fields, 90.6% for heart and 85.5% for clavicles.",

keywords = "Chest radiographs, Clavicle segmentation, Fully convolutional networks, Heart segmentation, Lung segmentation",

author = "Maayan Frid-Adar and Avi Ben-Cohen and Rula Amer and Hayit Greenspan",

note = "Publisher Copyright: {\textcopyright} Springer Nature Switzerland AG 2018.; 3rd International Workshop on Reconstruction and Analysis of Moving Body Organs, RAMBO 2018, 4th International Workshop on Breast Image Analysis, BIA 2018, and 1st International Workshop on Thoracic Image Analysis, TIA 2018, held in conjunction with 21s... ; Conference date: 16-09-2018 Through 20-09-2018",

year = "2018",

doi = "https://doi.org/10.1007/978-3-030-00946-5_17",

language = "الإنجليزيّة",

isbn = "9783030009458",

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

pages = "159--168",

editor = "David Snead and Emanuele Trucco and Danail Stoyanov and Zeike Taylor and Lena Maier-Hein and Nasir Rajpoot and Hrvoje Bogunovic and Francesco Ciompi and Mitko Veta and Garvin, {Mona K.} and Chen, {Xin Jan} and Anne Martel and {van der Laak}, Jeroen and Yanwu Xu and Stephen McKenna",

booktitle = "Image Analysis for Moving Organ, Breast, and Thoracic Images - Third International Workshop, RAMBO 2018, Fourth International Workshop, BIA 2018, and First International Workshop, TIA 2018, Held in Conjunction with MICCAI 2018, Proceedings",

}

Frid-Adar, M, Ben-Cohen, A, Amer, R & Greenspan, H 2018, Improving the segmentation of anatomical structures in chest radiographs using U-net with an imagenet pre-trained encoder. in D Snead, E Trucco, D Stoyanov, Z Taylor, L Maier-Hein, N Rajpoot, H Bogunovic, F Ciompi, M Veta, MK Garvin, XJ Chen, A Martel, J van der Laak, Y Xu & S McKenna (eds), Image Analysis for Moving Organ, Breast, and Thoracic Images - Third International Workshop, RAMBO 2018, Fourth International Workshop, BIA 2018, and First International Workshop, TIA 2018, Held in Conjunction with MICCAI 2018, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 11040 LNCS, pp. 159-168, 3rd International Workshop on Reconstruction and Analysis of Moving Body Organs, RAMBO 2018, 4th International Workshop on Breast Image Analysis, BIA 2018, and 1st International Workshop on Thoracic Image Analysis, TIA 2018, held in conjunction with 21s..., Granada, Spain, 16/09/18. https://doi.org/10.1007/978-3-030-00946-5_17

Improving the segmentation of anatomical structures in chest radiographs using U-net with an imagenet pre-trained encoder. / Frid-Adar, Maayan; Ben-Cohen, Avi; Amer, Rula et al.
Image Analysis for Moving Organ, Breast, and Thoracic Images - Third International Workshop, RAMBO 2018, Fourth International Workshop, BIA 2018, and First International Workshop, TIA 2018, Held in Conjunction with MICCAI 2018, Proceedings. ed. / David Snead; Emanuele Trucco; Danail Stoyanov; Zeike Taylor; Lena Maier-Hein; Nasir Rajpoot; Hrvoje Bogunovic; Francesco Ciompi; Mitko Veta; Mona K. Garvin; Xin Jan Chen; Anne Martel; Jeroen van der Laak; Yanwu Xu; Stephen McKenna. 2018. p. 159-168 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11040 LNCS).

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

TY - GEN

T1 - Improving the segmentation of anatomical structures in chest radiographs using U-net with an imagenet pre-trained encoder

AU - Frid-Adar, Maayan

AU - Ben-Cohen, Avi

AU - Amer, Rula

AU - Greenspan, Hayit

N1 - Publisher Copyright: © Springer Nature Switzerland AG 2018.

PY - 2018

Y1 - 2018

N2 - Accurate segmentation of anatomical structures in chest radiographs is essential for many computer-aided diagnosis tasks. In this paper we investigate the latest fully-convolutional architectures for the task of multi-class segmentation of the lungs field, heart and clavicles in a chest radiograph. In addition, we explore the influence of using different loss functions in the training process of a neural network for semantic segmentation. We evaluate all models on a common benchmark of 247 X-ray images from the JSRT database and ground-truth segmentation masks from the SCR dataset. Our best performing architecture, is a modified U-Net that benefits from pre-trained encoder weights. This model outperformed the current state-of-the-art methods tested on the same benchmark, with Jaccard overlap scores of 96.1% for lung fields, 90.6% for heart and 85.5% for clavicles.

AB - Accurate segmentation of anatomical structures in chest radiographs is essential for many computer-aided diagnosis tasks. In this paper we investigate the latest fully-convolutional architectures for the task of multi-class segmentation of the lungs field, heart and clavicles in a chest radiograph. In addition, we explore the influence of using different loss functions in the training process of a neural network for semantic segmentation. We evaluate all models on a common benchmark of 247 X-ray images from the JSRT database and ground-truth segmentation masks from the SCR dataset. Our best performing architecture, is a modified U-Net that benefits from pre-trained encoder weights. This model outperformed the current state-of-the-art methods tested on the same benchmark, with Jaccard overlap scores of 96.1% for lung fields, 90.6% for heart and 85.5% for clavicles.

KW - Chest radiographs

KW - Clavicle segmentation

KW - Fully convolutional networks

KW - Heart segmentation

KW - Lung segmentation

UR - http://www.scopus.com/inward/record.url?scp=85053894851&partnerID=8YFLogxK

U2 - https://doi.org/10.1007/978-3-030-00946-5_17

DO - https://doi.org/10.1007/978-3-030-00946-5_17

M3 - منشور من مؤتمر

SN - 9783030009458

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

SP - 159

EP - 168

BT - Image Analysis for Moving Organ, Breast, and Thoracic Images - Third International Workshop, RAMBO 2018, Fourth International Workshop, BIA 2018, and First International Workshop, TIA 2018, Held in Conjunction with MICCAI 2018, Proceedings

A2 - Snead, David

A2 - Trucco, Emanuele

A2 - Stoyanov, Danail

A2 - Taylor, Zeike

A2 - Maier-Hein, Lena

A2 - Rajpoot, Nasir

A2 - Bogunovic, Hrvoje

A2 - Ciompi, Francesco

A2 - Veta, Mitko

A2 - Garvin, Mona K.

A2 - Chen, Xin Jan

A2 - Martel, Anne

A2 - van der Laak, Jeroen

A2 - Xu, Yanwu

A2 - McKenna, Stephen

T2 - 3rd International Workshop on Reconstruction and Analysis of Moving Body Organs, RAMBO 2018, 4th International Workshop on Breast Image Analysis, BIA 2018, and 1st International Workshop on Thoracic Image Analysis, TIA 2018, held in conjunction with 21s...

Y2 - 16 September 2018 through 20 September 2018

ER -

Frid-Adar M, Ben-Cohen A, Amer R, Greenspan H. Improving the segmentation of anatomical structures in chest radiographs using U-net with an imagenet pre-trained encoder. In Snead D, Trucco E, Stoyanov D, Taylor Z, Maier-Hein L, Rajpoot N, Bogunovic H, Ciompi F, Veta M, Garvin MK, Chen XJ, Martel A, van der Laak J, Xu Y, McKenna S, editors, Image Analysis for Moving Organ, Breast, and Thoracic Images - Third International Workshop, RAMBO 2018, Fourth International Workshop, BIA 2018, and First International Workshop, TIA 2018, Held in Conjunction with MICCAI 2018, Proceedings. 2018. p. 159-168. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: https://doi.org/10.1007/978-3-030-00946-5_17

Improving the segmentation of anatomical structures in chest radiographs using U-net with an imagenet pre-trained encoder

Abstract

Publication series

Conference

Keywords

All Science Journal Classification (ASJC) codes

Access to Document

Other files and links

Fingerprint

Cite this