3D reconstruction of the retinal arterial tree based on fundus images using a self-calibration approach

Systemic diseases, such as hypertension and diabetes, are associated with changes in the retinal microvasculature. Although a number of studies have been performed on the quantitative assessment of the geometrical patterns of the retinal vasculature, previous work has been confined to 2 dimensional (2D) analyses. In this paper, we present an approach to obtain a 3D reconstruction of the retinal arteries from a pair of 2D retinal images acquired in vivo. A simple essential matrix based self-calibration approach was employed for the “fundus camera-eye” system. Vessel segmentation was performed using a semi-automatic approach and correspondence between points from different images was calculated. The results of 3D reconstruction show the centre-line of retinal vessels and their 3D curvature clearly. 3D reconstruction of the retinal vessels is feasible and may be useful in future studies of the retinal vasculature in disease.