Correlation and reproducibility of retinal vascular geometric measurements for stereoscopic retinal images of the same eyes.

PURPOSE

To assess the correlation and reproducibility of retinal vascular geometric measurements obtained from two stereo-paired fundus images.

METHODS

Thirty stereoscopic pairs of color optic disc-centered photographs from the Blue Mountains Eye Study were analyzed. Side-by-side grading was performed by a single grader, using semi-automated computer software to quantify the following retinal geometric parameters: (1) retinal arteriolar/venular caliber (CRAE/CRVE); (2) arteriole-to-venule ratio (AVR); (3) branching angle; and (4) tortuosity. We used Pearson correlation (r), intra-class correlation coefficient (ICC), and Bland-Altman plots to assess within-pair correlation and reproducibility for each parameter measured.

RESULTS

Inter- and intra-grader r and ICC were high (all r > 0.90 and ICC > 0.90), except for branching angle (ICCs between 0.69-0.83). There was no significant difference between within-pair means of all retinal vascular geometric parameters, before and after excluding poor quality images. CRAE, CRVE, AVR, and arteriolar and venular tortuosity showed very high within-pair correlation and agreement (all r > 0.80 and all ICC > 0.90 respectively). Arteriolar and venular branching angles demonstrated moderate within-pair correlation (r = 0.65 and r = 0.62, respectively) and within-pair agreement (ICC = 0.76 and ICC = 0.77, respectively).

CONCLUSIONS

Use of computer-assisted software to measure retinal vascular geometric parameters from paired fundus images was highly repeatable and is robust to differences in photographic angles of paired stereo images. Such measurements can be applied to evaluate temporal changes in longitudinal studies.