Abstract: Optic disc (OD) examination is of significant interest to both ophthalmologists and to image analysts. OD reveals symptoms of various ocular diseases like Glaucoma. For image analysts, optic disc detection although given its brighter intensities and sharp contrast is surprisingly a difficult task given its innumerous variations caused by retinal pathologies and imaging conditions. In this study, we propose a method for automatically detecting OD. The method involves a hierarchical approach where retinal image undergoes five-level wavelet decomposition for coarse OD detection which is followed by shape based classifier for precise OD boundary delineation. The proposed method was evaluated on 5789 images and achieved OD detection accuracy of 97.59%. OD boundary delineation performance was evaluated on a representative sample of 28 images and achieved a performance score of 88.37%. The results demonstrated consistency of the method across different image variations and can be adopted for various CAD applications on retinal images.

Keywords: automated detection, optic disc, wavelet decomposition, pattern classifier.