Vascular network tracking in SLO ocular fundus images for static and dynamic parameter extraction

In this work a tracking algorithm which extracts the vascular network structure from fundus Scanning Laser Ophthalmoscope (SLO) images is presented. The tracking algorithm is based on a priori knowledge of the vessel structure. It exploits the continuity of radius, position, direction and brightness of a blood vessel and is based on a recursive strategy. First, a main vessel is tracked and its branch points on both sides are identified. Then, the tracking process is applied again to the identified branches. The procedure is repeated till no branch points are found. The output of the algorithm is a structural description of the vascular network consisting of vessel position, radius and curvature. The presented algorithm was developed for images obtained using a contrast agent (fluoroangiography) but was also adapted to images without any contrast agent. The algorithm was tested both on simulated and real images and proved to give accurate measurement of vessel radius and position (mean errors below 1 pixel).