Detection of lesions on the retina image

Lesion areas in retinal images are important in the diagnosis of diabetic retinopathy. In this study, these areas are determined by the algorithm which includes the image processing process. The adaptive histogram equalization and median filtered image are divided into 5 separate areas by multi-level Otsu method. The Markov Random Field of the lesioned region on the 4th floor is formed and analyzed by Iterated Conditional Model. An analysis of 80 images taken from the DRIVE and STARE database showed a 90% success rate. This result can support the experts in this area at the point of making an early diagnosis.