An investigation into neural networks for the detection of exudates in retinal images

Gerald Schaefer*, Edmond Leung

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

We present an approach of automatically detecting exudates in retinal images using neural networks. Exudates are one of the early indicators of diabetic retinopathy which is known as one of the leading causes for blindness. A neural network is trained to classify whether small image windows are part of exudate areas or not. Furthermore, it is shown that a pre-processing step based on histogram specification in order to deal with varying lighting conditions greatly improves the recognition performance. Application of principal component analysis is used for dimensionality reduction and speed-up of the system. Experimental results were obtained on an image data set with known exudate locations and showed good classification performance with a sensitivity of 94.78% and a specificity of 94.29%.

Original languageEnglish
Title of host publicationApplications of soft computing: updating the state of art
EditorsErel Avineri, Mario Köppen, et al
Place of PublicationBerlin (DE)
PublisherSpringer
Pages169-177
Number of pages9
ISBN (Electronic)978-3-540-88079-0
ISBN (Print)978-3-540-88078-3
DOIs
Publication statusPublished - 11 Feb 2009
Event12th Online World Conference on Soft Computing in Industrial Applications - Online
Duration: 16 Oct 200726 Oct 2007

Publication series

NameAdvances in Soft Computing
PublisherSpringer
Volume52
ISSN (Print)1615-3871
ISSN (Electronic)1860-0794

Conference

Conference12th Online World Conference on Soft Computing in Industrial Applications
Abbreviated titleWSC12
Period16/10/0726/10/07

Fingerprint

Neural networks
Principal component analysis
Lighting
Specifications
Processing

Keywords

  • diabetic retinopathy
  • exudate detection
  • histogram equalisation/specification
  • neural network
  • principal component analysis

Cite this

Schaefer, G., & Leung, E. (2009). An investigation into neural networks for the detection of exudates in retinal images. In E. Avineri, M. Köppen, & et al (Eds.), Applications of soft computing: updating the state of art (pp. 169-177). (Advances in Soft Computing; Vol. 52). Berlin (DE): Springer. https://doi.org/10.1007/978-3-540-88079-0_17
Schaefer, Gerald ; Leung, Edmond. / An investigation into neural networks for the detection of exudates in retinal images. Applications of soft computing: updating the state of art. editor / Erel Avineri ; Mario Köppen ; et al. Berlin (DE) : Springer, 2009. pp. 169-177 (Advances in Soft Computing).
@inproceedings{293c0ae5310d4389940b6a82a8268281,
title = "An investigation into neural networks for the detection of exudates in retinal images",
abstract = "We present an approach of automatically detecting exudates in retinal images using neural networks. Exudates are one of the early indicators of diabetic retinopathy which is known as one of the leading causes for blindness. A neural network is trained to classify whether small image windows are part of exudate areas or not. Furthermore, it is shown that a pre-processing step based on histogram specification in order to deal with varying lighting conditions greatly improves the recognition performance. Application of principal component analysis is used for dimensionality reduction and speed-up of the system. Experimental results were obtained on an image data set with known exudate locations and showed good classification performance with a sensitivity of 94.78{\%} and a specificity of 94.29{\%}.",
keywords = "diabetic retinopathy, exudate detection, histogram equalisation/specification, neural network, principal component analysis",
author = "Gerald Schaefer and Edmond Leung",
year = "2009",
month = "2",
day = "11",
doi = "10.1007/978-3-540-88079-0_17",
language = "English",
isbn = "978-3-540-88078-3",
series = "Advances in Soft Computing",
publisher = "Springer",
pages = "169--177",
editor = "Erel Avineri and Mario K{\"o}ppen and {et al}",
booktitle = "Applications of soft computing: updating the state of art",
address = "Germany",

}

Schaefer, G & Leung, E 2009, An investigation into neural networks for the detection of exudates in retinal images. in E Avineri, M Köppen & et al (eds), Applications of soft computing: updating the state of art. Advances in Soft Computing, vol. 52, Springer, Berlin (DE), pp. 169-177, 12th Online World Conference on Soft Computing in Industrial Applications, 16/10/07. https://doi.org/10.1007/978-3-540-88079-0_17

An investigation into neural networks for the detection of exudates in retinal images. / Schaefer, Gerald; Leung, Edmond.

Applications of soft computing: updating the state of art. ed. / Erel Avineri; Mario Köppen; et al. Berlin (DE) : Springer, 2009. p. 169-177 (Advances in Soft Computing; Vol. 52).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

TY - GEN

T1 - An investigation into neural networks for the detection of exudates in retinal images

AU - Schaefer, Gerald

AU - Leung, Edmond

PY - 2009/2/11

Y1 - 2009/2/11

N2 - We present an approach of automatically detecting exudates in retinal images using neural networks. Exudates are one of the early indicators of diabetic retinopathy which is known as one of the leading causes for blindness. A neural network is trained to classify whether small image windows are part of exudate areas or not. Furthermore, it is shown that a pre-processing step based on histogram specification in order to deal with varying lighting conditions greatly improves the recognition performance. Application of principal component analysis is used for dimensionality reduction and speed-up of the system. Experimental results were obtained on an image data set with known exudate locations and showed good classification performance with a sensitivity of 94.78% and a specificity of 94.29%.

AB - We present an approach of automatically detecting exudates in retinal images using neural networks. Exudates are one of the early indicators of diabetic retinopathy which is known as one of the leading causes for blindness. A neural network is trained to classify whether small image windows are part of exudate areas or not. Furthermore, it is shown that a pre-processing step based on histogram specification in order to deal with varying lighting conditions greatly improves the recognition performance. Application of principal component analysis is used for dimensionality reduction and speed-up of the system. Experimental results were obtained on an image data set with known exudate locations and showed good classification performance with a sensitivity of 94.78% and a specificity of 94.29%.

KW - diabetic retinopathy

KW - exudate detection

KW - histogram equalisation/specification

KW - neural network

KW - principal component analysis

UR - http://link.springer.com/chapter/10.1007%2F978-3-540-88079-0_17

UR - http://www.scopus.com/inward/record.url?scp=59549083766&partnerID=8YFLogxK

U2 - 10.1007/978-3-540-88079-0_17

DO - 10.1007/978-3-540-88079-0_17

M3 - Conference contribution

AN - SCOPUS:59549083766

SN - 978-3-540-88078-3

T3 - Advances in Soft Computing

SP - 169

EP - 177

BT - Applications of soft computing: updating the state of art

A2 - Avineri, Erel

A2 - Köppen, Mario

A2 - et al,

PB - Springer

CY - Berlin (DE)

ER -

Schaefer G, Leung E. An investigation into neural networks for the detection of exudates in retinal images. In Avineri E, Köppen M, et al, editors, Applications of soft computing: updating the state of art. Berlin (DE): Springer. 2009. p. 169-177. (Advances in Soft Computing). https://doi.org/10.1007/978-3-540-88079-0_17