Advances in Image Processing Techniques for Drusens Detection and Quantification in Fundus Images

Age-Related Macular Degeneration (ARMD) is considered the leading cause of irreversible blindness in developed countries. One of its risk factors is the presence of drusens, which are retina abnormalities appearing as yellowish spots in fundus images. In this article a methodology using image processing techniques for the quantification of drusens is presented. The method uses splines combined with a contrast normalization to correct uneven illumination, followed by a drusen detection and modelling algorithm. The detection uses a gradient based segmentation algorithm that isolates drusens. They are then fitted by Gaussian functions, producing a model that is used to compute the area affected. To validate the methodology, 22 images were marked by three ophthalmologists and compared to the automated method. The sensitivity and specificity for the automated process (0.664 and 0.963) were comparable to that obtained among the specialists (0.656 and 0.971). Also, the Intraclass Correlation Coefficient showed an agreement of 74.9% between the processed images and the specialists’ analysis.