Explicit Topological Priors for Deep-Learning Based Image Segmentation Using Persistent Homology

James R. Clough; Ilkay Oksuz; Nicholas Byrne; Julia A. Schnabel; Andrew P. King

doi:10.1007/978-3-030-20351-1_2

Explicit Topological Priors for Deep-Learning Based Image Segmentation Using Persistent Homology

James R. Clough^*, Ilkay Oksuz, Nicholas Byrne, Julia A. Schnabel, Andrew P. King

^*Corresponding author for this work

King's College London

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

39 Citations (Scopus)

Abstract

We present a novel method to explicitly incorporate topological prior knowledge into deep learning based segmentation, which is, to our knowledge, the first work to do so. Our method uses the concept of persistent homology, a tool from topological data analysis, to capture high-level topological characteristics of segmentation results in a way which is differentiable with respect to the pixelwise probability of being assigned to a given class. The topological prior knowledge consists of the sequence of desired Betti numbers of the segmentation. As a proof-of-concept we demonstrate our approach by applying it to the problem of left-ventricle segmentation of cardiac MR images of subjects from the UK Biobank dataset, where we show that it improves segmentation performance in terms of topological correctness without sacrificing pixelwise accuracy.

Original language	English
Title of host publication	Information Processing in Medical Imaging - 26th International Conference, IPMI 2019, Proceedings
Editors	Siqi Bao, Albert C.S. Chung, James C. Gee, Paul A. Yushkevich
Publisher	Springer Verlag
Pages	16-28
Number of pages	13
ISBN (Print)	9783030203504
DOIs	https://doi.org/10.1007/978-3-030-20351-1_2
Publication status	Published - 2019
Externally published	Yes
Event	26th International Conference on Information Processing in Medical Imaging, IPMI 2019 - Hong Kong, China Duration: 2 Jun 2019 → 7 Jun 2019

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	11492 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	26th International Conference on Information Processing in Medical Imaging, IPMI 2019
Country/Territory	China
City	Hong Kong
Period	2/06/19 → 7/06/19

Bibliographical note

Publisher Copyright:
© 2019, Springer Nature Switzerland AG.

Keywords

Cardiac MRI
Persistent homology
Segmentation
Topological data analysis
Topology

Access to Document

10.1007/978-3-030-20351-1_2

Cite this

Clough, J. R., Oksuz, I., Byrne, N., Schnabel, J. A., & King, A. P. (2019). Explicit Topological Priors for Deep-Learning Based Image Segmentation Using Persistent Homology. In S. Bao, A. C. S. Chung, J. C. Gee, & P. A. Yushkevich (Eds.), Information Processing in Medical Imaging - 26th International Conference, IPMI 2019, Proceedings (pp. 16-28). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11492 LNCS). Springer Verlag. https://doi.org/10.1007/978-3-030-20351-1_2

Clough, James R. ; Oksuz, Ilkay ; Byrne, Nicholas et al. / Explicit Topological Priors for Deep-Learning Based Image Segmentation Using Persistent Homology. Information Processing in Medical Imaging - 26th International Conference, IPMI 2019, Proceedings. editor / Siqi Bao ; Albert C.S. Chung ; James C. Gee ; Paul A. Yushkevich. Springer Verlag, 2019. pp. 16-28 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{059884c5a1f34659a671aba4f3dd72b8,

title = "Explicit Topological Priors for Deep-Learning Based Image Segmentation Using Persistent Homology",

abstract = "We present a novel method to explicitly incorporate topological prior knowledge into deep learning based segmentation, which is, to our knowledge, the first work to do so. Our method uses the concept of persistent homology, a tool from topological data analysis, to capture high-level topological characteristics of segmentation results in a way which is differentiable with respect to the pixelwise probability of being assigned to a given class. The topological prior knowledge consists of the sequence of desired Betti numbers of the segmentation. As a proof-of-concept we demonstrate our approach by applying it to the problem of left-ventricle segmentation of cardiac MR images of subjects from the UK Biobank dataset, where we show that it improves segmentation performance in terms of topological correctness without sacrificing pixelwise accuracy.",

keywords = "Cardiac MRI, Persistent homology, Segmentation, Topological data analysis, Topology",

author = "Clough, {James R.} and Ilkay Oksuz and Nicholas Byrne and Schnabel, {Julia A.} and King, {Andrew P.}",

note = "Publisher Copyright: {\textcopyright} 2019, Springer Nature Switzerland AG.; 26th International Conference on Information Processing in Medical Imaging, IPMI 2019 ; Conference date: 02-06-2019 Through 07-06-2019",

year = "2019",

doi = "10.1007/978-3-030-20351-1_2",

language = "English",

isbn = "9783030203504",

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

publisher = "Springer Verlag",

pages = "16--28",

editor = "Siqi Bao and Chung, {Albert C.S.} and Gee, {James C.} and Yushkevich, {Paul A.}",

booktitle = "Information Processing in Medical Imaging - 26th International Conference, IPMI 2019, Proceedings",

address = "Germany",

}

Clough, JR, Oksuz, I, Byrne, N, Schnabel, JA & King, AP 2019, Explicit Topological Priors for Deep-Learning Based Image Segmentation Using Persistent Homology. in S Bao, ACS Chung, JC Gee & PA Yushkevich (eds), Information Processing in Medical Imaging - 26th International Conference, IPMI 2019, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 11492 LNCS, Springer Verlag, pp. 16-28, 26th International Conference on Information Processing in Medical Imaging, IPMI 2019, Hong Kong, China, 2/06/19. https://doi.org/10.1007/978-3-030-20351-1_2

Explicit Topological Priors for Deep-Learning Based Image Segmentation Using Persistent Homology. / Clough, James R.; Oksuz, Ilkay; Byrne, Nicholas et al.
Information Processing in Medical Imaging - 26th International Conference, IPMI 2019, Proceedings. ed. / Siqi Bao; Albert C.S. Chung; James C. Gee; Paul A. Yushkevich. Springer Verlag, 2019. p. 16-28 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11492 LNCS).

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

TY - GEN

T1 - Explicit Topological Priors for Deep-Learning Based Image Segmentation Using Persistent Homology

AU - Clough, James R.

AU - Oksuz, Ilkay

AU - Byrne, Nicholas

AU - Schnabel, Julia A.

AU - King, Andrew P.

PY - 2019

Y1 - 2019

N2 - We present a novel method to explicitly incorporate topological prior knowledge into deep learning based segmentation, which is, to our knowledge, the first work to do so. Our method uses the concept of persistent homology, a tool from topological data analysis, to capture high-level topological characteristics of segmentation results in a way which is differentiable with respect to the pixelwise probability of being assigned to a given class. The topological prior knowledge consists of the sequence of desired Betti numbers of the segmentation. As a proof-of-concept we demonstrate our approach by applying it to the problem of left-ventricle segmentation of cardiac MR images of subjects from the UK Biobank dataset, where we show that it improves segmentation performance in terms of topological correctness without sacrificing pixelwise accuracy.

AB - We present a novel method to explicitly incorporate topological prior knowledge into deep learning based segmentation, which is, to our knowledge, the first work to do so. Our method uses the concept of persistent homology, a tool from topological data analysis, to capture high-level topological characteristics of segmentation results in a way which is differentiable with respect to the pixelwise probability of being assigned to a given class. The topological prior knowledge consists of the sequence of desired Betti numbers of the segmentation. As a proof-of-concept we demonstrate our approach by applying it to the problem of left-ventricle segmentation of cardiac MR images of subjects from the UK Biobank dataset, where we show that it improves segmentation performance in terms of topological correctness without sacrificing pixelwise accuracy.

KW - Cardiac MRI

KW - Persistent homology

KW - Segmentation

KW - Topological data analysis

KW - Topology

UR - http://www.scopus.com/inward/record.url?scp=85066130921&partnerID=8YFLogxK

U2 - 10.1007/978-3-030-20351-1_2

DO - 10.1007/978-3-030-20351-1_2

M3 - Conference contribution

AN - SCOPUS:85066130921

SN - 9783030203504

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

SP - 16

EP - 28

BT - Information Processing in Medical Imaging - 26th International Conference, IPMI 2019, Proceedings

A2 - Bao, Siqi

A2 - Chung, Albert C.S.

A2 - Gee, James C.

A2 - Yushkevich, Paul A.

PB - Springer Verlag

T2 - 26th International Conference on Information Processing in Medical Imaging, IPMI 2019

Y2 - 2 June 2019 through 7 June 2019

ER -

Clough JR, Oksuz I, Byrne N, Schnabel JA, King AP. Explicit Topological Priors for Deep-Learning Based Image Segmentation Using Persistent Homology. In Bao S, Chung ACS, Gee JC, Yushkevich PA, editors, Information Processing in Medical Imaging - 26th International Conference, IPMI 2019, Proceedings. Springer Verlag. 2019. p. 16-28. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-030-20351-1_2

Explicit Topological Priors for Deep-Learning Based Image Segmentation Using Persistent Homology

Abstract

Publication series

Conference

Bibliographical note

Keywords

Access to Document

Other files and links

Fingerprint

Cite this