Thesis (M.Sc.) - Cairo University - Faculty of Engineering - Department of Systems and Biomedical Engineering

The main purpose of this research is to build a system capable of identifying the normal anatomical structure of the retina and the bright lesions which are caused by diabetic retinopathy. This is done by digital image processing technologies. In the first step Gabor filters were employed to increase the intensity contrast between the blood vessel and the background. Next we used entropic threshold, in order to classify each pixel in the blood vessels and other areas. Then to determine the optical disk location, we used iterative thresholding. Then we applied the geometric model that relied on the implicit active contour in order to accurately determine the boundary of the optical disk. In the final stage entropic threshold was applied to determine bright lesions such as hard exudates and cotton patches in the fundus images. This method achieved a sensitivity and specificity of 95.8% and 90.5%, respectively

Issued also as CD