International Journal of Engineering and Technology

Drusen are one of the early clinical findings in the development of age-related macular degeneration (ARMD), which causes irreversible vision loss. Automatic screening of individuals at risk may allow the detection of ARMD at an early stage, where it is curable. Detecting and locating the drusen in a color retinal image is a difficult task because they differ in shape, size, degree of confluence and texture. Hence, building a classifier for drusen identification is a challenging task. To address this difficulty an automated system for drusen detection is proposed with a goal of assessing the risk of the development of ARMD with increased accuracy while reducing the screening time. Our system was evaluated using fundus images, the proposed system initially detects and eliminates the optic disc as described in our previous work [1]. The system incorporates fundus image analysis techniques for image de-noising, illumination correction and normalization of contrast. A feature based fuzzy C-means clustering technique with a combination of multiresolution analysis using discrete wavelet transform (DWT), is proposed to improve the robustness of drusen detection. The gradient paths are labelled for calculating the area and location of the drusen spot. The proposed system was by comparing the drusen detected output images with the hand-labelled ground-truth images.