Medical imaging: Requirements analysis for automating diabetic retinopathy monitoring and diagnosis

This thesis proposes a system for acquiring retinal images that are then processed using image enhancement and feature detection techniques to identify features and anomalies that indicate the presence of diabetic retinopathy.; When a person has prolonged diabetes, they are at high risk to develop diabetic retinopathy. Frequent examinations by an ophthalmologist help diagnose and monitor the onset and progression of diabetic retinopathy. However, depending on where the patient lives and their economic status, they may not be able to take advantage of frequent examinations.; The proposed system is designed to enable frequent examinations and diagnoses that can provide these examinations to anyone no matter where they live or what their economic status.; This thesis discusses the possible methods for imaging the retina of the eye in order to obtain the image data required for diagnosis. The image acquisition methods covered are CT Scan and non-mydriatic retinal imaging.; The thesis discusses the image processing techniques used to enhance the images for later analyses. The methods discussed are used in both the spatial and frequency domain. Fourier Transforms are included in the discussion to describe how images are converted to the frequency domain from the spatial domain. The steps for enhancing the images for feature detection include filters and segmentation and mathematical operations such as convolution, correlation, and morphology.; The results recorded in this thesis indicate that it is possible using the prescribed algorithms to detect changes and anomalies in retinal images sufficient for aiding in the diagnosis and monitoring of diabetic retinopathy.