Adaptive Super-Candidate Based Approach for Detection and Classification of Drusen on Retinal Fundus Images

Abstract. Identification and characterization of drusen is essential for the severity assessment of age-related macular degeneration (AMD). Pre-sented here is a novel super-candidate based approach, combined with robust preprocessing and adaptive thresholding for detection of drusen, resulting in accurate segmentation with the mean lesion-level overlap of 0.75, even in cases with non-uniform illumination, poor contrast and con-founding anatomical structures. We also present a feature based lesion-level discrimination analysis between hard and soft drusen. Our method gives sensitivity of 80 % for high specificity above 90 % and high sensitiv-ity of 95 % for specificity of 70 % on representative pathological databases (STARE and ARIA) for both detection and discrimination.