Depiction of hemorrhagic regions caused by diabetic retinopathy on fundus flouresan angiography images by using image enhancement and basic classification methods

Diabetes can lead very different complications to the human body. The most commonly encountered complications in the human body is eyes. Eye vessels structure corruption, giving rise to a new, abnormal blood vessels and bleeding that occurs as a complication called diabetic retinopathy. Fundus fluorescein angiography (FFA) has been widely used imaging technique for the detection of disease. In recent years, increase in the incidence of this disease has increased the number of studies on the subject.In this thesis, detection of diabetic retinopathy and to highlight areas that cause hemorrhagic images of FFA on the classification, clustering methods and image analysis methods were used. As supervised learning, back propagation neural networks, statistical methods linear discriminant analysis (LDA), as unsupervised learning techniques, K-means clustering, Gaussian mixture method and the histogram image enhancement techniques transactions were selected for analysis. These approaches have been to emphasize shadows bleeding, and then the results which is obtained by these methods are compared.