Digital image processing techniques for the detection and quantitative analysis of diagnostic features in retinal images

点击翻译按钮获取中文摘要

Several diseases such as diabetes, arteriosclerosis, and hypertension affect the blood vessels in the retina of the eye. Specifically, the shape, width, and tortuosity of the vessels are modified by the disease processes. Similar changes are also seen in premature infants, in a condition known as retinopathy of prematurity. In addition, images of the retina can reveal the presence of pathological features, such as exudates, microaneurysms, hemorrhages, venous beading, and neovascularization. Quantitative analysis of the architecture of the vasculature of the retina and changes as above could assist in the monitoring of the stage of the disease processes, as well as in evaluating the effects of treatment. We propose to develop digital image processing and pattern analysis techniques for computer-aided detection and quantitative analysis of features as above in order to assist in the assessment of the related pathological processes by opthalmologists and other clinical specialists.The project will include the development of algorithms to analyze the shape, width, and tortuosity of blood vessels using shape factors, measures of complexity, and fractal analysis. The methods will be designed to facilitate longitudinal quantitative analysis to aid follow-up of patients with retinal pathology and assessment of the effect of treatment. The methods will be tested with publicly available databases of retinal images, as well as with images obtained from  the Alberta Children's Hospital.