Geometrical and morphological analysis of vascular branches from fundus retinal images

Abstract

A semi-automatic method to measure and quantify geometrical and topological properties of complete vascular trees in fundus images is described. The method is validated by comparing automatic vs. manual measurements in 17 individual bifurcations. We also compared automatic analyses of complete vascular trees from 4 pairs of red-free and fluorescein images. Preliminary results comparing 10 hypertensive and 10 normotensive subjects show changes in geometrical properties similar to those reported in previous studies. Several topological indices show differences in the arterial, but not venous, trees in hypertensive subjects. This suggests that a combination of geometrical and topological measurements of the whole vascular network may provide more sensitive indicators of morphological changes due to diseases.