The accuracy of retinal vessel diameter measurement based on pixels distribution and fractal representation

Abstract

We propose a method to accurately measure the retinal blood vessel widths with the final goal to be able to highlight the retinal microvascular signs and to identify related diseases such as diabetic retinopathy and hypertension. First, we use an automated diameter measurement method and then we use the histogram analysis and fractal representation to obtain retinal blood vessel diameters. The histogram method uses  pixels distribution, but vessel edges are more accurately localized using image profiles provided by the fractal technique. Sixty retinal images, divided into three groups (i.e. normal, diabetic retinopathy and hypertensive subjects) were used. Every image has been measured using all three methods. The measurements were performed in the same location and over the same retinal blood vessel by using the plug-in ImageJ. A statistical analysis has been performed to provide the variability among the various measurement methods and to determine the significance of the analyzed methods.