Fractal Analysis Of The Normal Human Retinal Vasculature

Abstract

Purpose: It has been proposed that retinal vessels are fractal and that fractal analysis could be applied in automated diagnosis of retinal diseases. Nevertheless, no standardized method has been proposed for image acquisition, processing and segmentation or for calculating the fractal dimension, resulting in great variability among the diverse research reports. The present study was designed to identify a suitable method for fractal analysis of retinal vessels without ocular disease and propose a standard for image acquisition, processing and fractal analysis. Patients and Methods Sixty-degree red-free retinographies with the fovea centralized were analysed. Images were segmented to extract their vascular trees using the wavelet transforms method and the fractal dimension of vascular trees was calculated by the boxcounting method. Results: The box-counting fractal dimension of vascular trees followed a Gaussian distribution, with mean 1.43 and standard deviation 0.04. Conclusion: A standardized method was developed to accomplish fractal analysis of retinal vessels and is suggested for subsequent studies designed to evaluate the use of fractal dimensions in retinal neovascularization modeling and in developing methods for automated diagnosis of retinal vascular diseases.