Publication detail

Automated Optic Disc Segmentation in Multimodal Images of Retina

Jiri Jan, Radim Chrastek, Libor Kubecka,Victor Derhartunian, Heinrich Niemann, Georg Michelson

Original Title

Automated Optic Disc Segmentation in Multimodal Images of Retina

Czech Title

Automatická segmentace optického disku v multimodálních obrazech

English Title

Automated Optic Disc Segmentation in Multimodal Images of Retina

Type

conference paper

Language

en

Original Abstract

Automatic segmentation of the optic disc margin CSLO- and colour fundus-images are registered using a fully-automatic multi-modal global registration method. It is performed by finding optimal parameters of the affine geometric transformation according to the modified mutual information similarity criterion. The segmentation of the optic disk in the color fundus photographs (CFP) is based on nonlinear filtering, the Canny edge detector and a modified Hough transform. The result of the segmentation of CFP (the initial circle contour) is transferred into the preprocessed CSLO image. By the anchored active contours (AAC) method the curve forming the sought optic nerve head (ONH) border is found. The method for automatic segmentation of the optic disc based on registration of images from the scanning laser tomograph and fundus photographs yields a 94% success-rate in respect to correct overlay. In 89% of the images the optic disc was correctly segmented.

Czech abstract

Automatická segmentace optického disku na retinálních obrazech. CSLO- and colour fundus-images are registered using a fully-automatic multi-modal global registration method. It is performed by finding optimal parameters of the affine geometric transformation according to the modified mutual information similarity criterion. The segmentation of the optic disk in the color fundus photographs (CFP) is based on nonlinear filtering, the Canny edge detector and a modified Hough transform. The result of the segmentation of CFP (the initial circle contour) is transferred into the preprocessed CSLO image. By the anchored active contours (AAC) method the curve forming the sought optic nerve head (ONH) border is found. The method for automatic segmentation of the optic disc based on registration of images from the scanning laser tomograph and fundus photographs yields a 94% success-rate in respect to correct overlay. In 89% of the images the optic disc was correctly segmented.

English abstract

Automatic segmentation of the optic disc margin CSLO- and colour fundus-images are registered using a fully-automatic multi-modal global registration method. It is performed by finding optimal parameters of the affine geometric transformation according to the modified mutual information similarity criterion. The segmentation of the optic disk in the color fundus photographs (CFP) is based on nonlinear filtering, the Canny edge detector and a modified Hough transform. The result of the segmentation of CFP (the initial circle contour) is transferred into the preprocessed CSLO image. By the anchored active contours (AAC) method the curve forming the sought optic nerve head (ONH) border is found. The method for automatic segmentation of the optic disc based on registration of images from the scanning laser tomograph and fundus photographs yields a 94% success-rate in respect to correct overlay. In 89% of the images the optic disc was correctly segmented.

Keywords

retina, optic disc, automatic segmentation, image registration

RIV year

2005

Released

25.09.2005

Publisher

DOG/SOE

Location

Berlin

Pages from

1000

Pages to

1000

Pages count

1

URL

knihovna UBMI - ( kompletní přednáška)

BibTex


@inproceedings{BUT14720,
  author="Jiří {Jan} and Libor {Kubečka} and Radim {Chrástek}",
  title="Automated Optic Disc Segmentation in Multimodal Images of Retina",
  annote="Automatic segmentation of the optic disc margin
CSLO- and colour fundus-images are registered using a fully-automatic multi-modal global registration method. It is performed by finding optimal parameters of the affine geometric transformation according to the modified mutual information similarity criterion. The segmentation of the optic disk in the color fundus photographs (CFP) is based on nonlinear filtering, the Canny edge detector and a modified Hough transform. The result of the segmentation of CFP (the initial circle contour) is transferred into the preprocessed CSLO image. By the anchored active contours (AAC) method the curve forming the sought optic nerve head (ONH) border is found. The method for automatic segmentation of the optic disc based on registration of images from the scanning laser tomograph and fundus photographs yields a 94% success-rate in respect to correct overlay. In 89% of the images the optic disc was correctly segmented.",
  address="DOG/SOE",
  booktitle="CD Proceedings of the DOG/SOE congress",
  chapter="14720",
  edition="Basic Sciences and Clinical Application",
  institution="DOG/SOE",
  year="2005",
  month="september",
  pages="1000",
  publisher="DOG/SOE",
  type="conference paper"
}