CaraNet: Context Axial Reverse Attention Network for Segmentation of Small Medical Objects

Ange Lou; Shuyue Guan; Hanseok Ko; Murray Loew

doi:10.1117/12.2611802

CaraNet: Context Axial Reverse Attention Network for Segmentation of Small Medical Objects

Ange Lou, Shuyue Guan, Hanseok Ko, Murray Loew

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

118 Citations (Scopus)

Abstract

Segmenting medical images accurately and reliably is important for disease diagnosis and treatment. It is a challenging task because of the wide variety of objects’ sizes, shapes, and scanning modalities. Recently, many convolutional neural networks (CNN) have been designed for segmentation tasks and achieved great success. Few studies, however, have fully considered the sizes of objects, and thus most demonstrate poor performance for small objects segmentation. This can have a significant impact on the early detection of diseases. This paper proposes a Context Axial Reserve Attention Network (CaraNet) to improve the segmentation performance on small objects compared with several recent state-of-the-art models. We test our CaraNet on brain tumor (BraTS 2018) and polyp (Kvasir-SEG, CVC-ColonDB, CVC-ClinicDB, CVC-300, and ETIS-LaribPolypDB) segmentation datasets. Our CaraNet achieves the top-rank mean Dice segmentation accuracy, and results show a distinct advantage of CaraNet in the segmentation of small medical objects. Codes available: https://github.com/AngeLouCN/CaraNet

Original language	English
Title of host publication	Medical Imaging 2022
Subtitle of host publication	Image Processing
Editors	Olivier Colliot, Ivana Isgum, Bennett A. Landman, Murray H. Loew
Publisher	SPIE
ISBN (Electronic)	9781510649392
DOIs	https://doi.org/10.1117/12.2611802
Publication status	Published - 2022
Event	Medical Imaging 2022: Image Processing - Virtual, Online Duration: 2021 Mar 21 → 2021 Mar 27

Publication series

Name	Progress in Biomedical Optics and Imaging - Proceedings of SPIE
Volume	12032
ISSN (Print)	1605-7422

Conference

Conference	Medical Imaging 2022: Image Processing
City	Virtual, Online
Period	21/3/21 → 21/3/27

Bibliographical note

Publisher Copyright:
© 2022 SPIE

Keywords

Attention
Brain tumor
Colonoscopy
Context axial reverse
Small object segmentation

ASJC Scopus subject areas

Electronic, Optical and Magnetic Materials
Atomic and Molecular Physics, and Optics
Biomaterials
Radiology Nuclear Medicine and imaging

Access to Document

10.1117/12.2611802

Cite this

Lou, A., Guan, S., Ko, H., & Loew, M. (2022). CaraNet: Context Axial Reverse Attention Network for Segmentation of Small Medical Objects. In O. Colliot, I. Isgum, B. A. Landman, & M. H. Loew (Eds.), Medical Imaging 2022: Image Processing Article 120320D (Progress in Biomedical Optics and Imaging - Proceedings of SPIE; Vol. 12032). SPIE. https://doi.org/10.1117/12.2611802

CaraNet: Context Axial Reverse Attention Network for Segmentation of Small Medical Objects. / Lou, Ange; Guan, Shuyue; Ko, Hanseok et al.
Medical Imaging 2022: Image Processing. ed. / Olivier Colliot; Ivana Isgum; Bennett A. Landman; Murray H. Loew. SPIE, 2022. 120320D (Progress in Biomedical Optics and Imaging - Proceedings of SPIE; Vol. 12032).

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

Lou, A, Guan, S, Ko, H & Loew, M 2022, CaraNet: Context Axial Reverse Attention Network for Segmentation of Small Medical Objects. in O Colliot, I Isgum, BA Landman & MH Loew (eds), Medical Imaging 2022: Image Processing., 120320D, Progress in Biomedical Optics and Imaging - Proceedings of SPIE, vol. 12032, SPIE, Medical Imaging 2022: Image Processing, Virtual, Online, 21/3/21. https://doi.org/10.1117/12.2611802

@inproceedings{f96e9a5a54fb48c1959b6f71abc5dd4d,

title = "CaraNet: Context Axial Reverse Attention Network for Segmentation of Small Medical Objects",

abstract = "Segmenting medical images accurately and reliably is important for disease diagnosis and treatment. It is a challenging task because of the wide variety of objects{\textquoteright} sizes, shapes, and scanning modalities. Recently, many convolutional neural networks (CNN) have been designed for segmentation tasks and achieved great success. Few studies, however, have fully considered the sizes of objects, and thus most demonstrate poor performance for small objects segmentation. This can have a significant impact on the early detection of diseases. This paper proposes a Context Axial Reserve Attention Network (CaraNet) to improve the segmentation performance on small objects compared with several recent state-of-the-art models. We test our CaraNet on brain tumor (BraTS 2018) and polyp (Kvasir-SEG, CVC-ColonDB, CVC-ClinicDB, CVC-300, and ETIS-LaribPolypDB) segmentation datasets. Our CaraNet achieves the top-rank mean Dice segmentation accuracy, and results show a distinct advantage of CaraNet in the segmentation of small medical objects. Codes available: https://github.com/AngeLouCN/CaraNet",

keywords = "Attention, Brain tumor, Colonoscopy, Context axial reverse, Small object segmentation",

author = "Ange Lou and Shuyue Guan and Hanseok Ko and Murray Loew",

note = "Publisher Copyright: {\textcopyright} 2022 SPIE; Medical Imaging 2022: Image Processing ; Conference date: 21-03-2021 Through 27-03-2021",

year = "2022",

doi = "10.1117/12.2611802",

language = "English",

series = "Progress in Biomedical Optics and Imaging - Proceedings of SPIE",

publisher = "SPIE",

editor = "Olivier Colliot and Ivana Isgum and Landman, {Bennett A.} and Loew, {Murray H.}",

booktitle = "Medical Imaging 2022",

}

TY - GEN

T1 - CaraNet

T2 - Medical Imaging 2022: Image Processing

AU - Lou, Ange

AU - Guan, Shuyue

AU - Ko, Hanseok

AU - Loew, Murray

PY - 2022

Y1 - 2022

N2 - Segmenting medical images accurately and reliably is important for disease diagnosis and treatment. It is a challenging task because of the wide variety of objects’ sizes, shapes, and scanning modalities. Recently, many convolutional neural networks (CNN) have been designed for segmentation tasks and achieved great success. Few studies, however, have fully considered the sizes of objects, and thus most demonstrate poor performance for small objects segmentation. This can have a significant impact on the early detection of diseases. This paper proposes a Context Axial Reserve Attention Network (CaraNet) to improve the segmentation performance on small objects compared with several recent state-of-the-art models. We test our CaraNet on brain tumor (BraTS 2018) and polyp (Kvasir-SEG, CVC-ColonDB, CVC-ClinicDB, CVC-300, and ETIS-LaribPolypDB) segmentation datasets. Our CaraNet achieves the top-rank mean Dice segmentation accuracy, and results show a distinct advantage of CaraNet in the segmentation of small medical objects. Codes available: https://github.com/AngeLouCN/CaraNet

AB - Segmenting medical images accurately and reliably is important for disease diagnosis and treatment. It is a challenging task because of the wide variety of objects’ sizes, shapes, and scanning modalities. Recently, many convolutional neural networks (CNN) have been designed for segmentation tasks and achieved great success. Few studies, however, have fully considered the sizes of objects, and thus most demonstrate poor performance for small objects segmentation. This can have a significant impact on the early detection of diseases. This paper proposes a Context Axial Reserve Attention Network (CaraNet) to improve the segmentation performance on small objects compared with several recent state-of-the-art models. We test our CaraNet on brain tumor (BraTS 2018) and polyp (Kvasir-SEG, CVC-ColonDB, CVC-ClinicDB, CVC-300, and ETIS-LaribPolypDB) segmentation datasets. Our CaraNet achieves the top-rank mean Dice segmentation accuracy, and results show a distinct advantage of CaraNet in the segmentation of small medical objects. Codes available: https://github.com/AngeLouCN/CaraNet

KW - Attention

KW - Brain tumor

KW - Colonoscopy

KW - Context axial reverse

KW - Small object segmentation

UR - http://www.scopus.com/inward/record.url?scp=85131918407&partnerID=8YFLogxK

U2 - 10.1117/12.2611802

DO - 10.1117/12.2611802

M3 - Conference contribution

AN - SCOPUS:85131918407

T3 - Progress in Biomedical Optics and Imaging - Proceedings of SPIE

BT - Medical Imaging 2022

A2 - Colliot, Olivier

A2 - Isgum, Ivana

A2 - Landman, Bennett A.

A2 - Loew, Murray H.

PB - SPIE

Y2 - 21 March 2021 through 27 March 2021

ER -

CaraNet: Context Axial Reverse Attention Network for Segmentation of Small Medical Objects

Abstract

Publication series

Conference

Bibliographical note

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this