Proliferative Diabetic Retinopathy Classification from Retinal Fundus Images Using Fractal Analysis

Abstract

Diabetic retinopathy is a complication of diabetes mellitus that affects the retinal tissue in the eye This disease is one of the leading causes of blindness in the world. Proliferative diabetic retinopathy is the most dangerous type of diabetic retinopathy (PDR). PDR is characterized by the development of neovascularization. Many studies have been conducted to identify PDR automatically. In this study, the authors used a retinal blood vascular structure approach to detect neovascularization on images. This strategy is implemented using fractal analysis. The wavelet transform segmentation method with 2D-Gabor wavelet was used in this study to provide optimal fractal feature values for classifying PDR. The maximum red lesions probability feature was also used in this study to detect PDR symptoms other than neovascularization. The most significant feature is the fractal analysis's shanon entropy in combination with the maximum red lesion probability, which yielded AUC values of 0.9335, with a sensitivity of 93.38 percent and a specificity of 81.17 percent. This method produces test results that show that as image resolution decreases, PDR classification remains stable, whereas PDR classification degrades with poor image quality.