Automated drusen segmentation in fundus images for diagnosing age related macular degeneration

Age related macular degeneration (ARMD) is an eye abnormality due to deposits of drusen on the retina and the disease may cause severe blindness. Early detection of ARMD using a computerized system can save patient´s vision. The ophthalmologists can find this system useful for screening of ARMD. This paper presents a novel method for accurate detection of drusen in colored retinal images. The system uses filter bank to extract all possible drusen regions form retinal image and eliminates false pixels which appear because of resemblance of drusen with optic disc. The proposed system represents each region with a number of features and then applies support vector machine to classify these regions as drusen and non-drusen. The performance of the proposed system is evaluated by testing it on STARE database using parameters such as sensitivity, specificity and accuracy. The validity of the proposed method is also shown by comparing it with already published methods.