Using Blood Vessels Location Information in Optic Disk Segmentation
Publication type
Journal article
Author(s)
A.S.Semashko, A.S.Krylov, A.S.Rodin
Publication date
September 2011
Journal
16th International Conference on Image Analysis and Processing (ICIAP'2011), Lecture Notes in Computer Science
Volume
6979
Issue
part 2
Page(s)
384-393
Abstract
In this paper we present an approach to achieve high accuracy of optic disk segmentation using information on the location of blood vessels (vessel map). Morphological preprocessing is employed to remove the vessels from the image and to compute the vessel map.
Vessel map is combined with edge map to obtain robust initial approximation of OD boundary using circular Hough transform.
We use this approximation to build 2D weight function for the edge map, which is then used in the active contour model.
We introduce an additional step to perform correction of the contour; in this step, the active contour model includes pressure forces and soft elliptical constraint. Vessel map is used in calculation of the ellipse parameters and pressure values.
The method was tested on 1240 publicly available retinal images, and manual labeling of the disk boundary by medical experts was used to assess its accuracy and compare it with other optic disk segmentation methods.
