Purpose: To analyze device-dependent variability of two non-mydriatic fundus cameras to obtain arterio-venous ratio (AVR), central retinal arteriolar equivalent (CRAE), and central retinal venular equivalent (CRVE) in static vessel analysis (SVA).
Methods: We examined 53 participants (29 men, 24 women; median age 46 years) of the Study of Health in Pomerania (SHIP). We took 45° optic-disc-centered fundus images of the right eye with two different non-mydriatic fundus cameras. The first photograph was obtained from the TRC-NW 200, the second from the OCT 2000 (both Topcon Corporation, Tokyo, Japan). One experienced grader graded image quality from 1 "ideal quality" to 5 "not analyzable" and determined AVR, CRAE, and CRVE with the software Vesselmap3 (Imedos, Jena, Germany).
Results: Average image quality was 1.8 for the TRC-NW 200 and 1.6 for the OCT 2000. AVR could not be determined in 5 images of the TRC-NW 200 due to low image quality, while six images of the OCT 2000 were not analyzable. The difference between AVR taken from two different non-mydriatic cameras was 0.01 ± 0.03 in Bland-Altman plots. The difference between CRAE was 0.17 ± 10.15 and between CRVE was -2.32 ± 11.76.
Conclusions: The two different non-mydriatic cameras showed good agreement with respect to image quality. When using the same reading software, AVR, CRAE, and CRVE agreed well. Thus, funduscopy and SVA seem to be robust against inter-device variability. As a result, device dependency can remain unconsidered in follow-up examinations with different technical equipment. However, variability might impact more with devices from different manufacturers.
Keywords: Arterio-venous ratio; bias; diagnostic errors; inter-device variations; static vessel analysis.