The Research On Automatic Analysis Of Glaucomatous Morphological Characteristics From Color Fundus Image

Glaucoma is the second leading cause of vision loss in the world. Glaucoma is associated with progressive changes of the optic nerve head (optic disk and optic cup) and retinal nerve fiber layer. A well-established method for diagnosis of glaucoma is the examination of the optic nerve head based on funds image as glaucomatous patients tend to have larger cup-to-disc ratios and are more likely to have peripapillary atrophy (PPA) and retinal fiber layer defects. Automatic retinal image analysis is an important screening tool for early detection of eye diseases, such as diabetic retinopathy, age-related macular degeneration and glaucoma. Glaucoma screening is a very complicate and difficult as there are lots of symptoms in different glaucomatous individuals with different stages. Moreover, there is not widely accepted golden standard yet. At present, the doctors make decisions based on their experiences after the patient finish some time consuming and expensive examines. The screening for glaucoma is very necessary as there are lots of glaucomatous individuals who are not unaware of their illness while there are not enough doctors to make diagnosis to the larger population. Robust and accurate detection the optic nerve head and PPA is very important and necessary step in automatic retinal screening systems including glaucoma. The glaucoma risk factors such as cup to disk ratio, notching, and ISNT rule is based on the accurate optic nerve head segmentation.Lots of studies have been done before about the optic nerve head segmentation. The difficulty in optic segmentation is due to the fuzzy boundaries and PPA noises. In this thesis a novel method for optic nerve head segmentation and PPA detection is proposed. First it uses template matching to find the region of interest (ROI).The way to remove vessels in the ROI is based on PDE inpainting. In order to find a better initial contour in the optic disk segmentation step it uses circle Hough transform to find an estimated center and radius. Two novel optic disk segmentation approaches using image texture are explored in this thesis. One of the optic disk segmentation methods is using image texture based on fuzzy c means (FCM). The other optic segmentation approach is a balloon snake model with edge cluster. Both of these methods can improve the disk segmentation accuracy. The optic cup segmentation is based on the vessel bends and pallor region. A multi-stage corner detection method is employed in this thesis to improve the cup segmentation accuracy. At last the glaucoma risk factors are proposed based on the optic disk and optic cup segmentation...