Manifold Learning for Density Segmentation in High Risk Mammograms

Harry Strange; Erika R. E. Denton; Minnie Kibiro; Reyer Zwiggelaar

doi:10.1007/978-3-642-38628-2_29

Manifold Learning for Density Segmentation in High Risk Mammograms

Harry Strange
, Erika R. E. Denton
, Minnie Kibiro
, Reyer Zwiggelaar

Department of Computer Science

Norfolk and Norwich University Hospitals NHS Foundation Trust

Research output: Chapter in Book/Report/Conference proceeding › Conference Proceeding (ISBN)

4 Citations (Scopus)

Abstract

There is a strong correlation between relative mammographic breast density and the risk of developing breast cancer. As such, accurately modelling the percentage of a mammogram that is dense is a pivotal step in density based risk classification. In this work, a novel method based on manifold learning is used to segment high-risk mammograms into density regions. As such, finer details are present in the segmentations and more accurate measures of breast density are produced. A set of high risk (BI-RADS IV) full field digital mammograms with density annotations obtained from radiologists are used to test the validity of the proposed approach. By exploiting the manifold structure of the input space, segmentations with average accuracy of 87% when compared with radiologists’ segmentations can be obtained. This is an increase of over 12% compared with segmentation in the high-dimensional space.

Original language	English
Title of host publication	Pattern Recognition and Image Analysis - 6th Iberian Conference, IbPRIA 2013, Proceedings
Subtitle of host publication	6th Iberian Conference, IbPRIA 2013, Funchal, Madeira, Portugal, June 5-7, 2013, Proceedings
Editors	João M. Sanches, Luisa Micó, Jaime S. Cardoso
Publisher	Springer Nature
Pages	245-252
Number of pages	8
ISBN (Electronic)	978-3-642-38628-2
ISBN (Print)	978-3-642-38627-5
DOIs	https://doi.org/10.1007/978-3-642-38628-2_29
Publication status	Published - 17 May 2013

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	7887 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

UN SDGs

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

SDG 3 Good Health and Well-being

Keywords

Breast Density Segmentation
Kernel PCA
Mammography
Manifold Learning

Access to Document

10.1007/978-3-642-38628-2_29

Permanent link

Persistent link

Cite this

Strange, H., Denton, E. R. E., Kibiro, M., & Zwiggelaar, R. (2013). Manifold Learning for Density Segmentation in High Risk Mammograms. In J. M. Sanches, L. Micó, & J. S. Cardoso (Eds.), Pattern Recognition and Image Analysis - 6th Iberian Conference, IbPRIA 2013, Proceedings: 6th Iberian Conference, IbPRIA 2013, Funchal, Madeira, Portugal, June 5-7, 2013, Proceedings (pp. 245-252). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 7887 LNCS). Springer Nature. https://doi.org/10.1007/978-3-642-38628-2_29

Strange, Harry ; Denton, Erika R. E. ; Kibiro, Minnie et al. / Manifold Learning for Density Segmentation in High Risk Mammograms. Pattern Recognition and Image Analysis - 6th Iberian Conference, IbPRIA 2013, Proceedings: 6th Iberian Conference, IbPRIA 2013, Funchal, Madeira, Portugal, June 5-7, 2013, Proceedings. editor / João M. Sanches ; Luisa Micó ; Jaime S. Cardoso. Springer Nature, 2013. pp. 245-252 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{d21d24644f414bd3a37e3defd1eb4533,

title = "Manifold Learning for Density Segmentation in High Risk Mammograms",

abstract = "There is a strong correlation between relative mammographic breast density and the risk of developing breast cancer. As such, accurately modelling the percentage of a mammogram that is dense is a pivotal step in density based risk classification. In this work, a novel method based on manifold learning is used to segment high-risk mammograms into density regions. As such, finer details are present in the segmentations and more accurate measures of breast density are produced. A set of high risk (BI-RADS IV) full field digital mammograms with density annotations obtained from radiologists are used to test the validity of the proposed approach. By exploiting the manifold structure of the input space, segmentations with average accuracy of 87\% when compared with radiologists{\textquoteright} segmentations can be obtained. This is an increase of over 12\% compared with segmentation in the high-dimensional space.",

keywords = "Breast Density Segmentation, Kernel PCA, Mammography, Manifold Learning",

author = "Harry Strange and Denton, \{Erika R. E.\} and Minnie Kibiro and Reyer Zwiggelaar",

year = "2013",

month = may,

day = "17",

doi = "10.1007/978-3-642-38628-2\_29",

language = "English",

isbn = "978-3-642-38627-5",

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

publisher = "Springer Nature",

pages = "245--252",

editor = "Sanches, \{Jo{\~a}o M.\} and Luisa Mic{\'o} and Cardoso, \{Jaime S.\}",

booktitle = "Pattern Recognition and Image Analysis - 6th Iberian Conference, IbPRIA 2013, Proceedings",

address = "Switzerland",

}

Strange, H, Denton, ERE, Kibiro, M & Zwiggelaar, R 2013, Manifold Learning for Density Segmentation in High Risk Mammograms. in JM Sanches, L Micó & JS Cardoso (eds), Pattern Recognition and Image Analysis - 6th Iberian Conference, IbPRIA 2013, Proceedings: 6th Iberian Conference, IbPRIA 2013, Funchal, Madeira, Portugal, June 5-7, 2013, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 7887 LNCS, Springer Nature, pp. 245-252. https://doi.org/10.1007/978-3-642-38628-2_29

Manifold Learning for Density Segmentation in High Risk Mammograms. / Strange, Harry; Denton, Erika R. E.; Kibiro, Minnie et al.
Pattern Recognition and Image Analysis - 6th Iberian Conference, IbPRIA 2013, Proceedings: 6th Iberian Conference, IbPRIA 2013, Funchal, Madeira, Portugal, June 5-7, 2013, Proceedings. ed. / João M. Sanches; Luisa Micó; Jaime S. Cardoso. Springer Nature, 2013. p. 245-252 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 7887 LNCS).

Research output: Chapter in Book/Report/Conference proceeding › Conference Proceeding (ISBN)

TY - GEN

T1 - Manifold Learning for Density Segmentation in High Risk Mammograms

AU - Strange, Harry

AU - Denton, Erika R. E.

AU - Kibiro, Minnie

AU - Zwiggelaar, Reyer

PY - 2013/5/17

Y1 - 2013/5/17

N2 - There is a strong correlation between relative mammographic breast density and the risk of developing breast cancer. As such, accurately modelling the percentage of a mammogram that is dense is a pivotal step in density based risk classification. In this work, a novel method based on manifold learning is used to segment high-risk mammograms into density regions. As such, finer details are present in the segmentations and more accurate measures of breast density are produced. A set of high risk (BI-RADS IV) full field digital mammograms with density annotations obtained from radiologists are used to test the validity of the proposed approach. By exploiting the manifold structure of the input space, segmentations with average accuracy of 87% when compared with radiologists’ segmentations can be obtained. This is an increase of over 12% compared with segmentation in the high-dimensional space.

AB - There is a strong correlation between relative mammographic breast density and the risk of developing breast cancer. As such, accurately modelling the percentage of a mammogram that is dense is a pivotal step in density based risk classification. In this work, a novel method based on manifold learning is used to segment high-risk mammograms into density regions. As such, finer details are present in the segmentations and more accurate measures of breast density are produced. A set of high risk (BI-RADS IV) full field digital mammograms with density annotations obtained from radiologists are used to test the validity of the proposed approach. By exploiting the manifold structure of the input space, segmentations with average accuracy of 87% when compared with radiologists’ segmentations can be obtained. This is an increase of over 12% compared with segmentation in the high-dimensional space.

KW - Breast Density Segmentation

KW - Kernel PCA

KW - Mammography

KW - Manifold Learning

UR - https://www.scopus.com/pages/publications/84883154461

U2 - 10.1007/978-3-642-38628-2_29

DO - 10.1007/978-3-642-38628-2_29

M3 - Conference Proceeding (ISBN)

SN - 978-3-642-38627-5

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

SP - 245

EP - 252

BT - Pattern Recognition and Image Analysis - 6th Iberian Conference, IbPRIA 2013, Proceedings

A2 - Sanches, João M.

A2 - Micó, Luisa

A2 - Cardoso, Jaime S.

PB - Springer Nature

ER -

Strange H, Denton ERE, Kibiro M, Zwiggelaar R. Manifold Learning for Density Segmentation in High Risk Mammograms. In Sanches JM, Micó L, Cardoso JS, editors, Pattern Recognition and Image Analysis - 6th Iberian Conference, IbPRIA 2013, Proceedings: 6th Iberian Conference, IbPRIA 2013, Funchal, Madeira, Portugal, June 5-7, 2013, Proceedings. Springer Nature. 2013. p. 245-252. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-642-38628-2_29

Manifold Learning for Density Segmentation in High Risk Mammograms

Abstract

Publication series

UN SDGs

Keywords

Access to Document

Permanent link

Other files and links

Fingerprint

Cite this