Focusnet++: Attentive aggregated transformations for efficient and accurate medical image segmentation

Chaitanya Kaul; Nick Pears; Hang Dai; Roderick Murray-Smith; Suresh Manandhar

doi:10.1109/ISBI48211.2021.9433918

Focusnet++: Attentive aggregated transformations for efficient and accurate medical image segmentation

Chaitanya Kaul
, Nick Pears
, Hang Dai
, Roderick Murray-Smith
, Suresh Manandhar

Mohamed bin Zayed University of Artificial Intelligence

University of Glasgow
University of York
NAAMII

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

8 Scopus citations

Abstract

We propose a new residual block for convolutional neural networks and demonstrate its state-of-the-art performance in medical image segmentation. We combine attention mechanisms with group convolutions to create our group attention mechanism, which forms the fundamental building block of our network, FocusNet++. We employ a hybrid loss based on balanced cross entropy, Tversky loss and the adaptive logarithmic loss to enhance the performance along with fast convergence. Our results show that FocusNet++ achieves state-of-the-art results across various benchmark metrics for the ISIC 2018 melanoma segmentation and the cell nuclei segmentation datasets with fewer parameters and FLOPs.

Original language	English
Title of host publication	2021 IEEE 18th International Symposium on Biomedical Imaging, ISBI 2021
Publisher	IEEE Computer Society
Pages	1042-1046
Number of pages	5
ISBN (Electronic)	9781665412469
DOIs	https://doi.org/10.1109/ISBI48211.2021.9433918
State	Published - 13 Apr 2021
Event	18th IEEE International Symposium on Biomedical Imaging, ISBI 2021 - Virtual, Online, France Duration: 13 Apr 2021 → 16 Apr 2021

Publication series

Name	Proceedings - International Symposium on Biomedical Imaging
Volume	2021-April
ISSN (Print)	1945-7928
ISSN (Electronic)	1945-8452

Conference

Conference	18th IEEE International Symposium on Biomedical Imaging, ISBI 2021
Country/Territory	France
City	Virtual, Online
Period	13/04/21 → 16/04/21

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

SDG 3 Good Health and Well-being

Keywords

Group Attention
Medical Image Segmentation
Residual Learning

Access to Document

10.1109/ISBI48211.2021.9433918

Cite this

Kaul, C., Pears, N., Dai, H., Murray-Smith, R., & Manandhar, S. (2021). Focusnet++: Attentive aggregated transformations for efficient and accurate medical image segmentation. In 2021 IEEE 18th International Symposium on Biomedical Imaging, ISBI 2021 (pp. 1042-1046). Article 9433918 (Proceedings - International Symposium on Biomedical Imaging; Vol. 2021-April). IEEE Computer Society. https://doi.org/10.1109/ISBI48211.2021.9433918

@inproceedings{a46b82f289db45628dd9ce889ba16ef0,

title = "Focusnet++: Attentive aggregated transformations for efficient and accurate medical image segmentation",

abstract = "We propose a new residual block for convolutional neural networks and demonstrate its state-of-the-art performance in medical image segmentation. We combine attention mechanisms with group convolutions to create our group attention mechanism, which forms the fundamental building block of our network, FocusNet++. We employ a hybrid loss based on balanced cross entropy, Tversky loss and the adaptive logarithmic loss to enhance the performance along with fast convergence. Our results show that FocusNet++ achieves state-of-the-art results across various benchmark metrics for the ISIC 2018 melanoma segmentation and the cell nuclei segmentation datasets with fewer parameters and FLOPs.",

keywords = "Group Attention, Medical Image Segmentation, Residual Learning",

author = "Chaitanya Kaul and Nick Pears and Hang Dai and Roderick Murray-Smith and Suresh Manandhar",

note = "Publisher Copyright: {\textcopyright} 2021 IEEE.; 18th IEEE International Symposium on Biomedical Imaging, ISBI 2021 ; Conference date: 13-04-2021 Through 16-04-2021",

year = "2021",

month = apr,

day = "13",

doi = "10.1109/ISBI48211.2021.9433918",

language = "English",

series = "Proceedings - International Symposium on Biomedical Imaging",

publisher = "IEEE Computer Society",

pages = "1042--1046",

booktitle = "2021 IEEE 18th International Symposium on Biomedical Imaging, ISBI 2021",

address = "United States",

}

Kaul, C, Pears, N, Dai, H, Murray-Smith, R & Manandhar, S 2021, Focusnet++: Attentive aggregated transformations for efficient and accurate medical image segmentation. in 2021 IEEE 18th International Symposium on Biomedical Imaging, ISBI 2021., 9433918, Proceedings - International Symposium on Biomedical Imaging, vol. 2021-April, IEEE Computer Society, pp. 1042-1046, 18th IEEE International Symposium on Biomedical Imaging, ISBI 2021, Virtual, Online, France, 13/04/21. https://doi.org/10.1109/ISBI48211.2021.9433918

Focusnet++: Attentive aggregated transformations for efficient and accurate medical image segmentation. / Kaul, Chaitanya; Pears, Nick; Dai, Hang et al.
2021 IEEE 18th International Symposium on Biomedical Imaging, ISBI 2021. IEEE Computer Society, 2021. p. 1042-1046 9433918 (Proceedings - International Symposium on Biomedical Imaging; Vol. 2021-April).

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

TY - GEN

T1 - Focusnet++

T2 - 18th IEEE International Symposium on Biomedical Imaging, ISBI 2021

AU - Kaul, Chaitanya

AU - Pears, Nick

AU - Dai, Hang

AU - Murray-Smith, Roderick

AU - Manandhar, Suresh

PY - 2021/4/13

Y1 - 2021/4/13

N2 - We propose a new residual block for convolutional neural networks and demonstrate its state-of-the-art performance in medical image segmentation. We combine attention mechanisms with group convolutions to create our group attention mechanism, which forms the fundamental building block of our network, FocusNet++. We employ a hybrid loss based on balanced cross entropy, Tversky loss and the adaptive logarithmic loss to enhance the performance along with fast convergence. Our results show that FocusNet++ achieves state-of-the-art results across various benchmark metrics for the ISIC 2018 melanoma segmentation and the cell nuclei segmentation datasets with fewer parameters and FLOPs.

AB - We propose a new residual block for convolutional neural networks and demonstrate its state-of-the-art performance in medical image segmentation. We combine attention mechanisms with group convolutions to create our group attention mechanism, which forms the fundamental building block of our network, FocusNet++. We employ a hybrid loss based on balanced cross entropy, Tversky loss and the adaptive logarithmic loss to enhance the performance along with fast convergence. Our results show that FocusNet++ achieves state-of-the-art results across various benchmark metrics for the ISIC 2018 melanoma segmentation and the cell nuclei segmentation datasets with fewer parameters and FLOPs.

KW - Group Attention

KW - Medical Image Segmentation

KW - Residual Learning

UR - https://www.scopus.com/pages/publications/85107199632

U2 - 10.1109/ISBI48211.2021.9433918

DO - 10.1109/ISBI48211.2021.9433918

M3 - Conference contribution

AN - SCOPUS:85107199632

T3 - Proceedings - International Symposium on Biomedical Imaging

SP - 1042

EP - 1046

BT - 2021 IEEE 18th International Symposium on Biomedical Imaging, ISBI 2021

PB - IEEE Computer Society

Y2 - 13 April 2021 through 16 April 2021

ER -

Kaul C, Pears N, Dai H, Murray-Smith R, Manandhar S. Focusnet++: Attentive aggregated transformations for efficient and accurate medical image segmentation. In 2021 IEEE 18th International Symposium on Biomedical Imaging, ISBI 2021. IEEE Computer Society. 2021. p. 1042-1046. 9433918. (Proceedings - International Symposium on Biomedical Imaging). doi: 10.1109/ISBI48211.2021.9433918

Focusnet++: Attentive aggregated transformations for efficient and accurate medical image segmentation

Abstract

Publication series

Conference

UN SDGs

Keywords

Access to Document

Other files and links

Fingerprint

Cite this