Automated system for the diagnosis of glaucoma and diabetic retinopathy

Abstract

Digital color fundus imaging a popular imaging modality is used for the diagnosis of retinal diseases to assess two sight threatening diseases namely glaucoma and Diabetic Retinopathy In glaucoma diagnosis commonly used quantitative classification of the optic nerve known as Cup to Disc Ratio does not account for the focal narrowing of the neuroretinal rim and disc size As glaucoma progresses optic cup grows larger and leads newlineto changes in the respective fundus images A fundus image analysis system is therefore developed to detect glaucoma in two phases by considering a structural feature with the incorporation of rim width disc size and distribution of textural features to reflect the physiological changes in the fundus images First phase involves the segmentation of optic disc optic cup extraction of structural features and quantification of neuroretinal rim Textural features are combined with the structural features and classified using Adaptive Neuro Fuzzy Inference System in the second phase The methods described in the literature often fail to segment the optic disc boundary accurately due to fuzzy boundaries blood vessel interruption inconsistent image contrast or missing edge features To incorporate the above information in tracing the disc boundary a Differential Windowing technique in the polar coordinate domain is proposed In the DW technique representation of the images in polar coordinate facilitates the description of local image regions in terms of their radial and tangential characteristics to find a closed contour in the region of interest newline newline