Purchase this article with an account.
D. Nemani, C. M. Wilson, K. A. Karp, G.-S. Ying, M. Maguire, J. B. Prince, J. Ng, A. R. Fielder, M. M. Mills, G. E. Quinn; Semi-Automated Digital Image Analysis for Eyes With and Without Retinopathy of Prematurity (ROP). Invest. Ophthalmol. Vis. Sci. 2008;49(13):1384. doi: https://doi.org/.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
To compare the use of two computer-assisted vessel analysis programs, specifically Computer Assisted Image Analysis of the Retina (CAIAR) and Vasculo-matic ala Nicola version 1.1 ( IVAN), in eyes of patients at risk for ROP.
11 good quality retinal images taken from 11 eyes using the NIDEK NM200D non-contact camera (30 degree field) were selected representing a range of peripapillary vessel findings. Five eyes with no disease to mild ROP, two eyes with pre-plus and four with plus disease were analyzed by a masked grader. The criteria for selecting images included a centered optic disc (as necessary for image analysis by IVAN), and clearly demarcated vessels including arteries and veins, preferably at least one of each in each quadrant. Vessel width is reported in arbitrary units for each program.
In the 11 images, 60 arteries and 48 veins were measurable by CAIAR, and 55 arteries and 48 veins were measurable by IVAN. The region of interest begins at the disc for CAIAR and at 900 microns from the disc edge for IVAN. Overall, vessel width values from CAIAR and IVAN did not correlate well (r= 0.13). When only veins were considered, the correlation was 0.50; for arteries, the correlation was -0.22. When the width was compared by quadrant, correlation in this small sample was best for temporal veins (TV) (0.73 superior TV, 0.62 inferior TV). CAIAR required significantly more time to delineate vessels than did IVAN.
In this preliminary study, we found that width measurements do not appear to correlate well between CAIAR and IVAN. This may be because CAIAR measures peripapillary changes (the area generally considered in clinical diagnosis), while IVAN measures slightly more peripheral retinal vessels. A comparison of measurements to clinical status with a larger sample is indicated as a next step in determining utility of computer-assisted algorithms for ROP screening.
This PDF is available to Subscribers Only