抽象的な

Automated DR and prediction of various related diseases of retinal fundus images

B. Sumathy, S. Poornachandra

An Automated DR system is developed and prediction of various related diseases are analyzed. Digital Retinal Fundus image is analyzed for the classification of various stages of Diabetic Retinopathy (DR). This imparts much importance, since many of systemic organs disease are related to the developmental stages of DR and those could be predicted well in advance, by analyzing the fundus image itself which is a cost effective technique. This kind of cost effective automated system is needed for mass screening programs in the developing nations. The system is tested on the database obtained from a private hospital in Tamilnadu, which consists of 196 images of different diagnoses. The system outputs are validated with the Expert ophthalmologist’s ground truth images. The proposed method involves three main phases. In the first phase is preprocessing steps are done followed by next phase of removal of normal features with extraction of abnormal features and its statistical values. The third phase is classification stages which includes mild, moderate to severe, severe PDR and neovascularization etc. using neural network. The proposed method has achieved a sensitivity, specificity and accuracy as 94.87%, 90% and 93.22%. The automated detection and classification method surely gives an important and promising data and information to the doctors for further progress in the treatment.