Branches filtering approach to extract retinal blood vessels in fundus image

Retinal blood vessels can give information about an abnormality or disease by examining its pathological changes. One of the abnormalities is diabetic retinopathy that is signed by a disorder of retinal blood vessels resulting from diabetes mellitus. Currently, diabetic retinopathy is one of major cause of human vision abnormalities or even blindness. Hence, early detection of such an abnormality can provide early and proper treatment, and segmentation of the abnormality provides a map of retinal vessels that can ease the assessment of the characteristics of the vessels. We propose a new method to segment blood vessels in a retinal image. In the method, Max-Tree is used to represent the image based on its gray level. Afterwards, the filtering process is done using branches filtering approach in which the tree branches is selected based on the elongation attribute of the nodes. The selection is started from the leaf nodes. This experiment was done to 40 retinal images, and utilized its manual segmentation by experts to validate the results. We obtain the accuracy of 93.95% and 94.21%, respectively to 40 images and 20 images.