Identification of Choroidal Neovascularisation on Fluorescein Angiograms Using Gradient Vector Flow Active Contours

The application of image processing to the investigation of age-related macular degeneration (AMD) has focused on detecting focal drusen deposits in colour fundus images. This research investigates gradient vector flow (GVF) active contours for the detection of choroidal neovascularisation (CNV) from fundus fluorescein angiograms in exudative AMD, the most severe form of the disease. The method was used in the identification of hyperfluorescent regions using pre-selected angiograms with expert-identified lesion components. Using active contours that are positioned close enough to the lesion of interest by the user, the algorithm can simply deform to the lesion, providing a successful outcome. If the initial active contour is positioned too far from the lesion location, the algorithm had a tendency to identify other image components incorrectly, especially in images with significant background interference or complex lesions. The algorithm reported here is guided by experts and hence is semi-automatic. The solutions (lesion position and size) obtained were compared with those identified and measured by an expert reader in a series of 10 fundus fluorescein images.