Hybrid classifier based drusen detection in colored fundus images

Age related macular degeneration (ARMD) is a medical condition which results in deterioration of human retina and in particularly macula. It is caused due to deposits of drusen on the retina and the disease may cause severe blindness. It is important to detect ARMD in its early stages to save patient´s vision. This paper proposes a new technique for drusen detection from fundus images by using Gabor kernel based filter bank and eliminating spurious regions which may be confused with drusen. The proposed system represents each region with a number of features and then applies hybrid classifier as an ensemble of Naive Bayes and Support Vector Machine to classify these regions as drusen and non-drusen. The proposed system is evaluated by testing it on STARE database using performance factors like sensitivity, specificity and accuracy. The results show the comparison and validity of proposed system with existing techniques.