A new approach of detection and segmentation of blood vessels for the classification of healthy and diseased retinal images

Retinal image analysis is increasing popularity for the detection of eye disease like diabetic retinopathy, glaucoma, cardiovascular disease. In modern ophthalmology, the structure of retinal image with proper segmentation has achieved much interest for disease detection. This paper presents a method with some new features extraction methodology after applying segmentation procedure in retinal images so that the retinal disease can be identified properly. At first we have extracted the vascular tree from the retinal image and then segmented the image for feature generation. These features have been used for the training of the ANN classifier by Backpropagation algorithm. The method described in this paper is applied to the High Resolution Fundus (HRF) image dataset which is a new publicly available one. The detection of the diabetic retinopathy, glaucoma and healthy condition of the image are 97%, 99% and 96% respectively.