Purpose: To investigate the clinical accuracy and determine the reproducibility of static, semiautomated retinal vessel analysis to supplement established vascular risk factors.
Methods: Manual blood pressure measurements and calibrated retinal photographs were obtained, after informed consent, on subjects without any eye disease aged >50 years. A total of 48 subjects without systemic hypertension or any other vascular disease and 54 subjects with confirmed hypertension were enrolled. Analysis was performed on retinal photographs taken by a retinal thickness analyzer (Talia Technology, Israel). The arteriovenous ratio (AVR) was calculated by a semiautomated vessel tracking VSL software (Talia Technology). Reproducibility was determined for software tracking, intra-, and intergrader selection as well as intra- and intervisit for 20 subjects. The effects of image quality degradation and decentration were investigated.
Results: Validation showed an excellent agreement between semiautomated software and manual vessel measurements. In the 102 subjects analyzed, retinal AVR was only correlated with established systemic hypertension (p=0.01) and gender (p=0.01). There was no effect of age on AVR. Other risk factors such as diabetes, smoking, body mass index, and current blood pressure showed some trends on multifactorial analysis. When limiting the number of vessels selected, software tracking induced no variability. The mean standard deviation for AVR was 0.02 for intra- as well as intergrader and 0.01 for inter- and intravisit effects. Image decentration only increased variability and the algorithm was robust against reducing image resolution and noise. Improper image focus, however, caused incorrect measurements of AVR.
Conclusions: Analysis of AVR can be performed reproducibly on routine retina photographs. Retinal AVR appears to be a relatively independent risk factor to assess systemic cerebrovascular disease.