Automated diagnosis of ARMD

S.Sivasankari, K.Narasimhan, S.Ramya School of Electrical and Electronics Engineering SASTRA university, Thanjavur, Tamilnadu, India [email protected] [email protected] [email protected] Abstract— Retinal Image analysis plays the important role in identifying retinal diseases and acts as aid for ophthalmologist. One of the retinal pathology which mainly affects the elder persons is Age Related Macular Degeneration. In retinal fundus images detection and segmentation of drusen, which helps to diagnose and grade the level of the Age related macular degeneration plays the major role. In this paper, we proposed the novel approach in which we used wavelet based sub band energy as a feature vector to discriminate normal and abnormal images. We used DB3, Symlet, RBio( 3.3,3.5,3.7) and we extracted the energy signature of various sub bands. Feature ranking and selection is done by using chi square test and consistency subset evaluation method. Thirteen features we used to classify the image using Support Vector Machine Classifer. We have collected the images from Vasan Eye Care Hospital, Thanjavur, Tamilnadu, India and with the guidance of ophthalmologist we have bifurcated normal and abnormal images for testing the proposed method. We obtained the accuracy of 93% with the combination of RBIO and SVM. Among the sub band energies we got high discriminatory power for diagonal and vertical sub band energies. KeywordARMD, Drusen, Features, Feature ranking, SVM classifier