Abstract
Purpose :
A portable Shack-Hartmann based autorefractor (SVOne by Smart Vision Labs) has recently been shown to produce accuracy levels similar to that of other subjective and objective procedures [1]. The SVOne captures 5 consecutive images about 1 second apart and processes them individually before combining the result to produce the final measurement output. In this study, we assess the variability across the 5 single-image results that make up a particular measurement.
Methods :
Raw data for 25 eyes from 25 individuals were selected from a database of SVOne measurements. Each data set contains 5 Shack-Hartmann spot-pattern images of a particular eye, and the only selection criterion was that no artifacts due to corneal reflection was visible. Pupil sizes represented ranged from 2 mm to beyond 6 mm, but only a 6-mm subpupil is analyzed for pupils larger than that. An improved image processing/analysis procedure was developed for this study. The new procedure automatically isolates the spot pattern, estimates spot locations via an iteratively weighted centroiding algorithm, and matches each found spot to its appropriate reference spot in order to determine the wavefront slope. A wavefront reconstructor comprised of only low-order Zernike modes was used to reconstruct the wavefront. Refractions were calculated individually for each of the 5 images and decomposed into its spherical equivalent (M) and crossed-cylinders (J0 and J45).
Results :
Spherical equivalent error ranged from -5.83 D to 4.60 D in the 25 eyes analyzed, and intra-measurement variability was within 0.25 D standard error (SE) for 80 percent of the cases. For the J0 and J45 terms, intra-measurement variability was well within 0.25 D SE in all cases. In fact, variability in J0 and J45 was below 0.125 D in the majority of the cases. A myopic shift due to accommodation during the measurement may explain some of the higher variabilities seen in M. For example, the measurement with the highest variability in M (0.44 D SE) contained 4 single-image measurements all within 0.11 D of each other and an outlier that is more than 1 D off from the rest.
Conclusions :
Intra-measurement variability for SVOne is generally low. Higher variations were only observed in the spherical equivalent term likely due to accommodation. 1. KJ Ciufreda and M Rosenfield, OVS, 92(12): 1133-1139
This is an abstract that was submitted for the 2016 ARVO Annual Meeting, held in Seattle, Wash., May 1-5, 2016.