The Methodology And Reliability Of Retinal Fractal Dimension Measurement Using A Novel Computer-assisted Method In An Indian Population

Abstract

Purpose: : Fractal analysis is a method to quantify the geometric branching complexity and density of the retinal vessels. This study describes the methodology of fractal analysis and evaluates the reliability of retinal fractal dimension (Df) measurement with a novel computer-assisted method.

Methods: : We developed a new computer-assisted program to measure the Fractal Dimension (Df) of the retinal vessels from digital retinal photographs for disc-centred (OD Df) and macula- centred (Mac Df) views. Retinal photographs from the Singapore Indian Eye Study (SINDI), a population-based survey of 3,400 (75.6% response) persons aged 40 to 80 years, were used for this study. A subset of 171 retinal photographs (5% of SINDI participants) were randomly selected and measured by two trained graders independently to determine the inter-grader reliability. The graders repeated the measurements after two weeks to determine the intra-grader reliability. Inter-grader and Intra-grader reliability was assessed using intra-class correlation coefficients (ICC) and Bland-Altman plots. Pearson correlation coefficients (r) were calculated to assess the associations between Df with age and systolic blood pressure.

Results: : The mean (standard deviation [SD]) OD Df and Mac Df were 1.484 (SD 0.043) and 1.453 (SD 0.060), respectively. Mac Df was significantly lower than OD Df (p<0.001). Both intragrader and intergrader reliability estimates for Df measures were high with ICCs ranging from 0.88 to 0.99. The OD Df and Mac Df were highly correlated (r=0.701, p<0.001). Both OD Df and Mac Df were significantly associated with age (r= -0.287 and-0.466 respectively, all p<0.001) and systolic blood pressure (r= -0.250 and -0.256 respectively, all p<0.001).

Conclusions: : The geometric branching complexity of the retinal vessels can be quantitatively and reliably measured from retinal photographs using this novel computer-based program and shows a strong inverse correlation with age and blood pressure.

Keywords: imaging/image analysis: clinical • blood supply • retina