Publication detail

Method for Registration of Multimodal Images of Human Retina

KUBEČKA, L., JAN, J.

Original Title

Method for Registration of Multimodal Images of Human Retina

Czech Title

Metoda pro registraci multimodálních obrazů lidské sítnice

English Title

Method for Registration of Multimodal Images of Human Retina

Type

conference paper

Language

en

Original Abstract

To improve early glaucoma detection and progression monitoring, the Heidelberg Retina Tomograph (HRT) is used. The HRT is a confocal laser scanning system, which provides 3D image data of the human retina. The standard colour photographs provide different information; they are used for vessel segmentation and for computing arterio-venous diameter ratio, which is important for early detection of arteriosclerosis and diabetic retinopathy. In both cases it is important to detect the border of the optic nerve head (optic disc) correctly. We found, that rather than analysing each modality separately to produce a robust segmentation, fusing information from both modalities to a vector-valued image is helpful. This paper describes the method of transforming one image to fit the other so that they are in register forming a fused image.

Czech abstract

Pro zlepšení brzké detekce a sledování vývoje glaukomu se používá konfokální laserový skenovací oftalmoskop - Heidelberg Retina Tomograph (HRT), který poskytuje 3D obrazová data lidské sítnice, a je tudíž vhodý pro sledování tvarových změn sítnice. Standardní barevná fotografie poskytuje jinou informaci o sítnici, je používána pro segmentaci cév a výpočet poměru tloušťky tepen a žil, což je důležitý parametr pro brzkou diagnózu arteriosklerózy a diabetické retinopatie. V obou případech je nutné správně detekovat hranici výstupu optického nervu (optický disk). Z jistili jsme, že lépe než analyzovat obě modality separátně, je vhodnější provést fúzi obou obrazů do jednoho vektorového obrazu a provádět analýzu v takto upravených datech. Tento článek se popisuje metodu transformace jednoho z obrazů tak, aby co nejlépe vyjadřoval informační obsah druhého z obrazů (tzv. registrace obrazů).

English abstract

To improve early glaucoma detection and progression monitoring, the Heidelberg Retina Tomograph (HRT) is used. The HRT is a confocal laser scanning system, which provides 3D image data of the human retina. The standard colour photographs provide different information; they are used for vessel segmentation and for computing arterio-venous diameter ratio, which is important for early detection of arteriosclerosis and diabetic retinopathy. In both cases it is important to detect the border of the optic nerve head (optic disc) correctly. We found, that rather than analysing each modality separately to produce a robust segmentation, fusing information from both modalities to a vector-valued image is helpful. This paper describes the method of transforming one image to fit the other so that they are in register forming a fused image.

Keywords

mutual information, image registration, affine transformation, simulated anneling

RIV year

2004

Released

22.06.2004

Publisher

BUT Brno Press

Location

Brno

ISBN

80-214-2633-0

Book

Proceedings of 17th Biennial Int. Conf. Biosignal 2004

Edition number

První

Pages from

255

Pages to

257

Pages count

3

BibTex


@inproceedings{BUT13244,
  author="Libor {Kubečka} and Jiří {Jan}",
  title="Method for Registration of Multimodal Images of Human Retina",
  annote="To improve early glaucoma detection and progression monitoring, the Heidelberg Retina Tomograph (HRT) is used. The HRT is a confocal laser scanning system, which provides 3D image data of the human retina. The standard colour photographs provide different information; they are used for vessel segmentation and for computing arterio-venous diameter ratio, which is important for early detection of arteriosclerosis and diabetic retinopathy. In both cases it is important to detect the border of the optic nerve head (optic disc) correctly. We found, that rather than analysing each modality separately to produce a robust segmentation, fusing information from both modalities to a vector-valued image is helpful. This paper describes the method of transforming one image to fit the other so that they are in register forming a fused image.",
  address="BUT Brno Press",
  booktitle="Proceedings of 17th Biennial Int. Conf. Biosignal 2004",
  chapter="13244",
  institution="BUT Brno Press",
  year="2004",
  month="june",
  pages="255",
  publisher="BUT Brno Press",
  type="conference paper"
}