Self-supervised Multi-scale Consistency for Weakly Supervised Segmentation Learning

Gabriele Valvano; Andrea Leo; Sotirios A. Tsaftaris

doi:10.1007/978-3-030-87722-4_2

Self-supervised Multi-scale Consistency for Weakly Supervised Segmentation Learning

Gabriele Valvano^*, Andrea Leo, Sotirios A. Tsaftaris

^*Corresponding author for this work

School of Engineering

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

Abstract

Collecting large-scale medical datasets with fine-grained annotations is time-consuming and requires experts. For this reason, weakly supervised learning aims at optimising machine learning models using weaker forms of annotations, such as scribbles, which are easier and faster to collect. Unfortunately, training with weak labels is challenging and needs regularisation. Herein, we introduce a novel self-supervised multi-scale consistency loss, which, coupled with an attention mechanism, encourages the segmentor to learn multi-scale relationships between objects and improves performance. We show state-of-the-art performance on several medical and non-medical datasets. The code used for the experiments is available at https://vios-s.github.io/multiscale-pyag.

Original language	English
Title of host publication	Domain Adaptation and Representation Transfer, and Affordable Healthcare and AI for Resource Diverse Global Health
Subtitle of host publication	3rd MICCAI Workshop, DART 2021, and 1st MICCAI Workshop, FAIR 2021, Held in Conjunction with MICCAI 2021, Proceedings
Editors	Shadi Albarqouni, M. Jorge Cardoso, Qi Dou, Konstantinos Kamnitsas, Bishesh Khanal, Islem Rekik, Nicola Rieke, Debdoot Sheet, Sotirios Tsaftaris, Daguang Xu, Ziyue Xu
Publisher	Springer Science and Business Media Deutschland GmbH
Pages	14-24
Number of pages	11
ISBN (Print)	9783030877217
DOIs	https://doi.org/10.1007/978-3-030-87722-4_2
Publication status	Published - 21 Sep 2021
Event	3rd MICCAI Workshop on Domain Adaptation and Representation Transfer, DART 2021, and the 1st MICCAI Workshop on Affordable Healthcare and AI for Resource Diverse Global Health, FAIR 2021, held in conjunction with 24th International Conference on Medical Image Computing and Computer Assisted Intervention, MICCAI 2021 - Virtual, Online Duration: 27 Sep 2021 → 1 Oct 2021

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	12968 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	3rd MICCAI Workshop on Domain Adaptation and Representation Transfer, DART 2021, and the 1st MICCAI Workshop on Affordable Healthcare and AI for Resource Diverse Global Health, FAIR 2021, held in conjunction with 24th International Conference on Medical Image Computing and Computer Assisted Intervention, MICCAI 2021
City	Virtual, Online
Period	27/09/21 → 1/10/21

Keywords

Segmentation
Self-supervised learning
Shape prior

Access to Document

10.1007/978-3-030-87722-4_2

Cite this

Valvano, G., Leo, A., & Tsaftaris, S. A. (2021). Self-supervised Multi-scale Consistency for Weakly Supervised Segmentation Learning. In S. Albarqouni, M. J. Cardoso, Q. Dou, K. Kamnitsas, B. Khanal, I. Rekik, N. Rieke, D. Sheet, S. Tsaftaris, D. Xu, & Z. Xu (Eds.), Domain Adaptation and Representation Transfer, and Affordable Healthcare and AI for Resource Diverse Global Health: 3rd MICCAI Workshop, DART 2021, and 1st MICCAI Workshop, FAIR 2021, Held in Conjunction with MICCAI 2021, Proceedings (pp. 14-24). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 12968 LNCS). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-3-030-87722-4_2

Valvano, Gabriele ; Leo, Andrea ; Tsaftaris, Sotirios A. / Self-supervised Multi-scale Consistency for Weakly Supervised Segmentation Learning. Domain Adaptation and Representation Transfer, and Affordable Healthcare and AI for Resource Diverse Global Health: 3rd MICCAI Workshop, DART 2021, and 1st MICCAI Workshop, FAIR 2021, Held in Conjunction with MICCAI 2021, Proceedings. editor / Shadi Albarqouni ; M. Jorge Cardoso ; Qi Dou ; Konstantinos Kamnitsas ; Bishesh Khanal ; Islem Rekik ; Nicola Rieke ; Debdoot Sheet ; Sotirios Tsaftaris ; Daguang Xu ; Ziyue Xu. Springer Science and Business Media Deutschland GmbH, 2021. pp. 14-24 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{7079d9e964f34ae9b04832736d52eb88,

title = "Self-supervised Multi-scale Consistency for Weakly Supervised Segmentation Learning",

abstract = "Collecting large-scale medical datasets with fine-grained annotations is time-consuming and requires experts. For this reason, weakly supervised learning aims at optimising machine learning models using weaker forms of annotations, such as scribbles, which are easier and faster to collect. Unfortunately, training with weak labels is challenging and needs regularisation. Herein, we introduce a novel self-supervised multi-scale consistency loss, which, coupled with an attention mechanism, encourages the segmentor to learn multi-scale relationships between objects and improves performance. We show state-of-the-art performance on several medical and non-medical datasets. The code used for the experiments is available at https://vios-s.github.io/multiscale-pyag.",

keywords = "Segmentation, Self-supervised learning, Shape prior",

author = "Gabriele Valvano and Andrea Leo and Tsaftaris, {Sotirios A.}",

note = "Proxy DOA to exclude from REF Funding Information: Acknowledgments. This work was partially supported by the Alan Turing Institute (EPSRC grant EP/N510129/1). S.A. Tsaftaris acknowledges the support of Canon Medical and the Royal Academy of Engineering and the Research Chairs and Senior Research Fellowships scheme (grant RCSRF1819\8\25). Publisher Copyright: {\textcopyright} 2021, Springer Nature Switzerland AG.; 3rd MICCAI Workshop on Domain Adaptation and Representation Transfer, DART 2021, and the 1st MICCAI Workshop on Affordable Healthcare and AI for Resource Diverse Global Health, FAIR 2021, held in conjunction with 24th International Conference on Medical Image Computing and Computer Assisted Intervention, MICCAI 2021 ; Conference date: 27-09-2021 Through 01-10-2021",

year = "2021",

month = sep,

day = "21",

doi = "10.1007/978-3-030-87722-4_2",

language = "English",

isbn = "9783030877217",

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

publisher = "Springer Science and Business Media Deutschland GmbH",

pages = "14--24",

editor = "Shadi Albarqouni and Cardoso, {M. Jorge} and Qi Dou and Konstantinos Kamnitsas and Bishesh Khanal and Islem Rekik and Nicola Rieke and Debdoot Sheet and Sotirios Tsaftaris and Daguang Xu and Ziyue Xu",

booktitle = "Domain Adaptation and Representation Transfer, and Affordable Healthcare and AI for Resource Diverse Global Health",

address = "Germany",

}

Valvano, G, Leo, A & Tsaftaris, SA 2021, Self-supervised Multi-scale Consistency for Weakly Supervised Segmentation Learning. in S Albarqouni, MJ Cardoso, Q Dou, K Kamnitsas, B Khanal, I Rekik, N Rieke, D Sheet, S Tsaftaris, D Xu & Z Xu (eds), Domain Adaptation and Representation Transfer, and Affordable Healthcare and AI for Resource Diverse Global Health: 3rd MICCAI Workshop, DART 2021, and 1st MICCAI Workshop, FAIR 2021, Held in Conjunction with MICCAI 2021, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 12968 LNCS, Springer Science and Business Media Deutschland GmbH, pp. 14-24, 3rd MICCAI Workshop on Domain Adaptation and Representation Transfer, DART 2021, and the 1st MICCAI Workshop on Affordable Healthcare and AI for Resource Diverse Global Health, FAIR 2021, held in conjunction with 24th International Conference on Medical Image Computing and Computer Assisted Intervention, MICCAI 2021, Virtual, Online, 27/09/21. https://doi.org/10.1007/978-3-030-87722-4_2

Self-supervised Multi-scale Consistency for Weakly Supervised Segmentation Learning. / Valvano, Gabriele; Leo, Andrea; Tsaftaris, Sotirios A.

Domain Adaptation and Representation Transfer, and Affordable Healthcare and AI for Resource Diverse Global Health: 3rd MICCAI Workshop, DART 2021, and 1st MICCAI Workshop, FAIR 2021, Held in Conjunction with MICCAI 2021, Proceedings. ed. / Shadi Albarqouni; M. Jorge Cardoso; Qi Dou; Konstantinos Kamnitsas; Bishesh Khanal; Islem Rekik; Nicola Rieke; Debdoot Sheet; Sotirios Tsaftaris; Daguang Xu; Ziyue Xu. Springer Science and Business Media Deutschland GmbH, 2021. p. 14-24 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 12968 LNCS).

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

TY - GEN

T1 - Self-supervised Multi-scale Consistency for Weakly Supervised Segmentation Learning

AU - Valvano, Gabriele

AU - Leo, Andrea

AU - Tsaftaris, Sotirios A.

N1 - Proxy DOA to exclude from REF Funding Information: Acknowledgments. This work was partially supported by the Alan Turing Institute (EPSRC grant EP/N510129/1). S.A. Tsaftaris acknowledges the support of Canon Medical and the Royal Academy of Engineering and the Research Chairs and Senior Research Fellowships scheme (grant RCSRF1819\8\25). Publisher Copyright: © 2021, Springer Nature Switzerland AG.

PY - 2021/9/21

Y1 - 2021/9/21

N2 - Collecting large-scale medical datasets with fine-grained annotations is time-consuming and requires experts. For this reason, weakly supervised learning aims at optimising machine learning models using weaker forms of annotations, such as scribbles, which are easier and faster to collect. Unfortunately, training with weak labels is challenging and needs regularisation. Herein, we introduce a novel self-supervised multi-scale consistency loss, which, coupled with an attention mechanism, encourages the segmentor to learn multi-scale relationships between objects and improves performance. We show state-of-the-art performance on several medical and non-medical datasets. The code used for the experiments is available at https://vios-s.github.io/multiscale-pyag.

AB - Collecting large-scale medical datasets with fine-grained annotations is time-consuming and requires experts. For this reason, weakly supervised learning aims at optimising machine learning models using weaker forms of annotations, such as scribbles, which are easier and faster to collect. Unfortunately, training with weak labels is challenging and needs regularisation. Herein, we introduce a novel self-supervised multi-scale consistency loss, which, coupled with an attention mechanism, encourages the segmentor to learn multi-scale relationships between objects and improves performance. We show state-of-the-art performance on several medical and non-medical datasets. The code used for the experiments is available at https://vios-s.github.io/multiscale-pyag.

KW - Segmentation

KW - Self-supervised learning

KW - Shape prior

UR - http://www.scopus.com/inward/record.url?scp=85116479020&partnerID=8YFLogxK

U2 - 10.1007/978-3-030-87722-4_2

DO - 10.1007/978-3-030-87722-4_2

M3 - Conference contribution

AN - SCOPUS:85116479020

SN - 9783030877217

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

SP - 14

EP - 24

BT - Domain Adaptation and Representation Transfer, and Affordable Healthcare and AI for Resource Diverse Global Health

A2 - Albarqouni, Shadi

A2 - Cardoso, M. Jorge

A2 - Dou, Qi

A2 - Kamnitsas, Konstantinos

A2 - Khanal, Bishesh

A2 - Rekik, Islem

A2 - Rieke, Nicola

A2 - Sheet, Debdoot

A2 - Tsaftaris, Sotirios

A2 - Xu, Daguang

A2 - Xu, Ziyue

PB - Springer Science and Business Media Deutschland GmbH

T2 - 3rd MICCAI Workshop on Domain Adaptation and Representation Transfer, DART 2021, and the 1st MICCAI Workshop on Affordable Healthcare and AI for Resource Diverse Global Health, FAIR 2021, held in conjunction with 24th International Conference on Medical Image Computing and Computer Assisted Intervention, MICCAI 2021

Y2 - 27 September 2021 through 1 October 2021

ER -

Valvano G, Leo A, Tsaftaris SA. Self-supervised Multi-scale Consistency for Weakly Supervised Segmentation Learning. In Albarqouni S, Cardoso MJ, Dou Q, Kamnitsas K, Khanal B, Rekik I, Rieke N, Sheet D, Tsaftaris S, Xu D, Xu Z, editors, Domain Adaptation and Representation Transfer, and Affordable Healthcare and AI for Resource Diverse Global Health: 3rd MICCAI Workshop, DART 2021, and 1st MICCAI Workshop, FAIR 2021, Held in Conjunction with MICCAI 2021, Proceedings. Springer Science and Business Media Deutschland GmbH. 2021. p. 14-24. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/978-3-030-87722-4_2

Self-supervised Multi-scale Consistency for Weakly Supervised Segmentation Learning

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Canon Medical / RAEng Senior Research Fellow in Healthcare AI

Canon Medical / RAEng Senior Research Fellow in Healthcare AI

Cite this

Self-supervised Multi-scale Consistency for Weakly Supervised Segmentation Learning

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Projects

Canon Medical / RAEng Senior Research Fellow in Healthcare AI

Canon Medical / RAEng Senior Research Fellow in Healthcare AI

Cite this