THALIA - An automatic hierarchical analysis system to detect drusen lesion images for amd assessment

Age-related macular degeneration (AMD) is a leading cause of permanent blindness. In its early stage AMD is characterized by drusen which are extracellelur deposits in the retina. In this paper, we present THALIA, an automatic system for the detection of drusen images for AMD assessment. First, the macular region of interest is detected using a seeded mode tracking approach. The macular region of interest is then mapped into a new representation using a hierarchicial word transform (HWI). In HWI, dense sampling is first carried out to generate structured pixels which embed local context. These structured pixels are then clustered using hierarchical k-means. The HWI image is subsequently classified using a SVM-based classifier. We have tested THALIA on a dataset of 350 images and obtained an accuracy of 95.46%. Results are promising for further validation of the THALIA system.