Abstract
Purpose :
Evidence suggests that features associated with retinal blood vessels near the optic disc are early biomarkers of hypertension. We hypothesise that differences in vessel width measured farther away from the OD in ultra-widefield (UWF) retinal images also indicate the presence of hypertension.
Methods :
152 eyes of 76 age-matched males (aged 49-59 years) were selected from subjects recruited in the SCOT-Heart trial [1] for undilated UWF retinal imaging with the Optos P200C scanning laser ophthalmoscope. 38 individuals were classified with hypertension (HT) (mean age 54.4 ± 2.5 years) and 38 with normotension (NT) (54.7 ± 2.9 years). An automated vessel segmentation algorithm [2] (VAMPIRE project) was used to extract vessel pixels in a concentric annular segment subtending 180° nasal to the optic disc (OD), see Figure 1. Manual classification of arterioles and venules was carried out by a trained operator. The average width of each vessel segment was measured and the arteriole-venule width ratio (AVR) was calculated for each image. The AVR was compared between the NT and HT groups using the p-value and confidence interval (CI) of the difference between population means determined with a two-tailed Student’s t-test and the area under the receiver operator characteristics curve (AUC).
Results :
AVR evaluated from the averages of widths from the 5 widest arterioles and the 5 widest venules showed a significant difference between the NT and HT groups in the left eye (p=0.006, CI=0.017, 0.104, AUC=0.676) but not in the right eye (p=0.137, CI=-0.009, 0.063, AUC=0.583). A significant result was also found in the left eye using the widest arteriole and venule (p=0.000, CI=0.034, 0.104, AUC=0.753), but not in the right (p=0.356, CI=-0.019 0.054, AUC=0.542).
Conclusions :
Our results indicate that vessel width characteristics measured with UWF imaging may have the same predictive property for HT as on conventional imaging. Increasing the sample size will allow us to explore the observed difference in AVR measurements in the right and left eye. Extending the analysis to more individuals, to different age groups and to females is needed to validate the clinically utility of this image analysis method.
1. Newby, D. E., et al, 2012, Trials, 13(1), 184.
2. Pellegrini, E., et al, 2014, Biomed. Opt. Express 5, 4329-37.
This is an abstract that was submitted for the 2016 ARVO Annual Meeting, held in Seattle, Wash., May 1-5, 2016.