Publication detail

Texture Analysis Of Nerve Fibre Layer In Retinal Images Via Local Binary Patterns And Gaussian Markov Random Fields

NOVOTNÝ, A. ODSTRČILÍK, J. KOLÁŘ, R. JAN, J.

Original Title

Texture Analysis Of Nerve Fibre Layer In Retinal Images Via Local Binary Patterns And Gaussian Markov Random Fields

Czech Title

Texture Analysis Of Nerve Fibre Layer In Retinal Images Via Local Binary Patterns And Gaussian Markov Random Fields

English Title

Texture Analysis Of Nerve Fibre Layer In Retinal Images Via Local Binary Patterns And Gaussian Markov Random Fields

Type

conference paper

Language

en

Original Abstract

The proposed contribution deals with the texture analysis of colour retinal images provided by digital fundus camera. The main purpose is focused on the analysis of texture representing retinal nerve fibre layer. An early detection of changes in the texture caused by nerve fibres atrophy is important in diagnosis of glaucoma. The proposed texture analysis is based on local binary patterns and Gaussian Markov random fields approaches providing textural features allowing quantitative representation of retinal nerve fibre layer texture. The features are presented and supervised classification experiments are performed with satisfactory results. The cross-validation procedures were implemented in order to evaluate classification ability. The classification error reaches less than 3% allowing discrimination between healthy and glaucomatous tissue of the retinal nerve fibre layer. An experiment providing detection of the retinal nerve fibre layer loss in diagnostically important region around the optic nerve head was performed and the qualitative results comparable with experienced ophthalmologists diagnosis was achieved.

Czech abstract

Příspěvek se zabývá texturní analýzou vrstvy nervových vláken na fundus fotografiích sítnice. Jako metody texturní analýzy jsou využity Markovská náhodná pole a lokální binární vzory. Publikovaná metoda přispívá k možnostem včasné diagnostiky glaukomu.

English abstract

The proposed contribution deals with the texture analysis of colour retinal images provided by digital fundus camera. The main purpose is focused on the analysis of texture representing retinal nerve fibre layer. An early detection of changes in the texture caused by nerve fibres atrophy is important in diagnosis of glaucoma. The proposed texture analysis is based on local binary patterns and Gaussian Markov random fields approaches providing textural features allowing quantitative representation of retinal nerve fibre layer texture. The features are presented and supervised classification experiments are performed with satisfactory results. The cross-validation procedures were implemented in order to evaluate classification ability. The classification error reaches less than 3% allowing discrimination between healthy and glaucomatous tissue of the retinal nerve fibre layer. An experiment providing detection of the retinal nerve fibre layer loss in diagnostically important region around the optic nerve head was performed and the qualitative results comparable with experienced ophthalmologists diagnosis was achieved.

Keywords

glaucoma, retinal nerve fibre layer, markov random fields, local binnary patterns, fundus photographs, retinal images, texture analysis

RIV year

2010

Released

27.06.2010

Publisher

EURASIP

Location

Brno

ISBN

978-80-214-4106-4

Book

20th Biennial International EURASIP Conference BIOSIGNAL 2010

Edition

EURASIP

Pages from

308

Pages to

315

Pages count

8

BibTex


@inproceedings{BUT35919,
  author="Adam {Novotný} and Jan {Odstrčilík} and Radim {Kolář} and Jiří {Jan}",
  title="Texture Analysis Of Nerve Fibre Layer In Retinal Images Via Local Binary Patterns And Gaussian Markov Random Fields",
  annote="The proposed contribution deals with the texture analysis of colour retinal images provided by digital fundus camera. The main purpose is focused on the analysis of texture representing retinal nerve fibre layer. An early detection of changes in the texture caused by nerve fibres atrophy is important in diagnosis of glaucoma. The proposed texture analysis is based on local binary patterns and Gaussian Markov random fields approaches providing textural features allowing quantitative representation of retinal nerve fibre layer texture. The features are presented and supervised classification experiments are performed with satisfactory results. The cross-validation procedures were implemented in order to evaluate classification ability. The classification error reaches less than 3% allowing discrimination between healthy and glaucomatous tissue of the retinal nerve fibre layer. An experiment providing detection of the retinal nerve fibre layer loss in diagnostically important region around the optic nerve head was performed and the qualitative results comparable with experienced ophthalmologists diagnosis was achieved.",
  address="EURASIP",
  booktitle="20th Biennial International EURASIP Conference BIOSIGNAL 2010",
  chapter="35919",
  edition="EURASIP",
  howpublished="online",
  institution="EURASIP",
  number="1",
  year="2010",
  month="june",
  pages="308--315",
  publisher="EURASIP",
  type="conference paper"
}