Publication detail

Towards automated diagnostic evaluation of retina images

H. Niemann, R. Chrastek, T. Hothorn, B. Lausen, L. Kubecka, J. Jan, G. Michelson

Original Title

Towards automated diagnostic evaluation of retina images

Czech Title

Automatické diagnostické hodnocení obrazů sítnice.

English Title

Towards automated diagnostic evaluation of retina images

Type

journal article

Language

en

Original Abstract

In this paper we address automatic segmentation of the optic nerve head (ONH) with the long-term goal of automatic diagnosis of the early stages of glaucoma. The images discussed are average images obtained from a scanning laser ophthalmoscope (SLO). The segmentation consists of the following main steps: finding a region of interest containing the ONH, constraining the search space for final segmentation, and computing the fine segmentation by an active contour model. The agreement of “true positive pixels,” i.e., pixels attributed to the ONH by both manual and automatic segmentation, is very good. The classification results obtained from three different classifiers using manual or automatic segmentation still reveal the superiority of manual segmentation. One means to further improve automatic segmentation is to use information from an SLO as well as from a fundus camera.

Czech abstract

Morfologická analýza hlavy optického nervu je uznávanou metodou diagnozy glaukomu. Tato analýza závisí na předchozím správném nalezení hranice hlavy optického nervu. První námi vivinutá automatická metoda byla závislá na šumu v obraze, nehomogenním osvětlení a přítomnosti cév. Proto jsme inspirováni současným klinickým výzkumem vytvořili algoritmus provádějící segmentaci v registrovaných multimodálních obrazech sítnice. Multimodální přístup kombinuje tomografický obraz s barevnou fotografií sítnice pomocí registrace obrazů založené na optimalizaci podobnostního kritéria vzájemné informace. Jádrem segmentačního algoritmu jsou kotvené aktivní kontury inicializované Houghovou transformací použité na ¨morfologicky zpracovaných obrazech. Metoda byla testována na 174 multimodálních obrazových párech. Systém dosáhl 89% správně segmentovaných optických disků ve srovnání s 74% u monomodální metody. Navržený algoritmus je slibným krokem k vytvoření automatického systém skríningu glaukomu.

English abstract

In this paper we address automatic segmentation of the optic nerve head (ONH) with the long-term goal of automatic diagnosis of the early stages of glaucoma. The images discussed are average images obtained from a scanning laser ophthalmoscope (SLO). The segmentation consists of the following main steps: finding a region of interest containing the ONH, constraining the search space for final segmentation, and computing the fine segmentation by an active contour model. The agreement of “true positive pixels,” i.e., pixels attributed to the ONH by both manual and automatic segmentation, is very good. The classification results obtained from three different classifiers using manual or automatic segmentation still reveal the superiority of manual segmentation. One means to further improve automatic segmentation is to use information from an SLO as well as from a fundus camera.

Keywords

retina, glaucoma, segmentation, opric disc, optic nerve head

RIV year

2006

Released

01.01.2005

Publisher

MAIK Nauka/Interperiodica Publishing, Moscow

Location

Moscow

Pages from

273

Pages to

276

Pages count

4

URL

BibTex


@article{BUT46311,
  author="Radim {Chrástek} and Libor {Kubečka} and Jiří {Jan}",
  title="Towards automated diagnostic evaluation of retina images",
  annote="In this paper we address automatic segmentation of the optic nerve head (ONH) with the long-term
goal of automatic diagnosis of the early stages of glaucoma. The images discussed are average images obtained
from a scanning laser ophthalmoscope (SLO). The segmentation consists of the following main steps: finding
a region of interest containing the ONH, constraining the search space for final segmentation, and computing
the fine segmentation by an active contour model. The agreement of “true positive pixels,” i.e., pixels attributed
to the ONH by both manual and automatic segmentation, is very good. The classification results obtained from
three different classifiers using manual or automatic segmentation still reveal the superiority of manual segmentation.
One means to further improve automatic segmentation is to use information from an SLO as well as from
a fundus camera.",
  address="MAIK Nauka/Interperiodica Publishing, Moscow",
  chapter="46311",
  institution="MAIK Nauka/Interperiodica Publishing, Moscow",
  journal="Pattern Recognition and Image Analysis
",
  number="2",
  volume="15",
  year="2005",
  month="january",
  pages="273",
  publisher="MAIK Nauka/Interperiodica Publishing, Moscow",
  type="journal article"
}