In this paper, we present a novel approach of quantification of retinal vascular shape changes for detecting vascular retinopathies. The method classifies vessels and quantifies their tortuosity changes in fluorescein angiographic images. The classification approach is based on a split and merge technique. Two classes of vessel are defined: artery/vein and arteriole/veinlet. In a sequence of images, we form correspondence of classified vessels. Quantifying classified vessels consists of comparing the correspondent tortuosity. We use eccentricity to define tortuosity. Eccentricity computation is based on 2D moment invariant techniques. The method is applied to images of patients suffering of diabetes and sickle cell hemoglobin C disease.
Publié le : 2002-07-04
Classification:
split and merge,
correspondence,
retinal angiography,
vascular retinopathy.,
classification,
division et fusion,
mise en correspondance,
angiographie rétinienne,
rétinopathies vasculaires.,
[INFO]Computer Science [cs],
[MATH]Mathematics [math]
@article{hal-01261703,
author = {Assogba, Kokou and Bouaoune, Yasmina and Bunel, Philippe},
title = {Classification by split and merge technique for the detection of vascular retinopathies},
journal = {HAL},
volume = {2002},
number = {0},
year = {2002},
language = {fr},
url = {http://dml.mathdoc.fr/item/hal-01261703}
}
Assogba, Kokou; Bouaoune, Yasmina; Bunel, Philippe. Classification by split and merge technique for the detection of vascular retinopathies. HAL, Tome 2002 (2002) no. 0, . http://gdmltest.u-ga.fr/item/hal-01261703/