Automated Drusen Segmentation in Fundus Images for Diagnosing Age Related Macular Degeneration (T-0673) (MFN 4020)

Abstract

Age related macular degeneration (ARMD) is an eye related disease which is caused due to deposits of bright lesions called drusen. Drusen are formed at retinal level and could affect eyesight as well. More severe form of ARMD can end up in complete loss of vision which is irreversible. Many research’s’ have been done in the field of medical care. Lots of researchers are also focusing on introducing automated detection systems for retinal diseases as well. Early diagnosis of drusen with the help of automatic system can save vision loss. This system could really be useful for ophthalmologists for screening of ARMD. This paper basically presents a novel method for detection of drusen from colored retinal images. Filter bank has been used to separate drusen regions and eradicates undesired regions which could be mistaken as drusen because of their bright appearance on fundus image. LS-SVM classifier is used to distinguish drusen regions and non drusen regions. This is done with the help of feature vector i.e. each region is represented by a number of features. The performance of the proposed system is estimated using performance measures like accuracy, sensitivity and specificity. STARE database was used to evaluate the results. The competence of given system is calculated by doing comparison with previous methods.