Segmentation of Exudate in Fundus Images Using Graph-Based Algorithm

Diabetes mellitus is a medical term for the designation of the disease known as diabetes. The disease is known to cause a variety of complications in the eyes, and the most common is diabetic retinopathy retinal disorders. One of the early signs of diabetic retinopathy is the emergence microaneurysms. Microaneurysms arise due to weakening of the vessel wall or the smallest blood vessels. Injuries that occur on the vessel causes the formation of exudate that is a point that looks yellowish around the retina. Application of digital retinal image segmentation to aid the detection of exudate is a system that created to segmenting retinal image, so that the resulting of image that can be used as input to the system identification exudate. In this study, include preprocessing operations such as a morphological closing operation to remove the blood vessels, using erosion and dilation operations to eliminate the boundary circumference retina, contrast enhancement, using a logical AND to delete the object instead exudate. While the segmentation method used is graph-based method. Tests conducted with a sample of data taken from Messidor database (http://messidor.crihan.fr) and accuracy was calculated by comparing the result from manually tagging of two people with image segmentation results of the testing program. From the test, the percentage average of accuracy obtains 64.46%.