Fractal, multifractal and lacunarity analysis applied in retinal regionsof diabetic patients with and without non-proliferative diabetic retinopathy

Abstract

The aim of this study was to evaluate the blood vascular network of the retina and its regions in diabetic patients with and without signs of mild early diabetic retinopathy by using number of bifurcation points, fractal and multifractal methods. We used 33 segmented images of retinographies, 28 corresponding to the retinographies that did not show any sign of diabetic retinopathy and 5 diagnosed with non-proliferative diabetic retinopathy (NPDR). The segmented images were obtained from the DRIVE (Digital Retinal Images for Vessel Extraction) database. The segmented images of retinal blood vessels were skeletonized and fractionated into nine equal regions. Later, the skeletonized images were assessed by using the following methods: number of bifurcation points, box-counting dimension, information dimension, lacunarity parameter and multifractal analysis. The retinas and their regions of control group (diabetic without NPDR) were statistically compared to those of the NPDR by using the Z-test and the Mann-Whitney test. The multifractal analysis showed that skeletonized images of retinal vessels and its nine regions follows a multifractal behavior. In our hands, neither the fractal method nor the number of bifurcation points disclosed statistically significant differences (p>0.05). No any difference was observed in the retinal vascular network between diabetic patients with or without NPDR