Purchase this article with an account.
Ayeswarya Ravikumar, Jason Marsack, Yue Shi, Raymond Applegate; What is the smallest change in visual acuity that is correlated with a change in image quality?. Invest. Ophthalmol. Vis. Sci. 2013;54(15):1283.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
The 95% confidence interval for logMAR acuity testing is ~ ±5 letters (±1 line)[ Optom Vis Sci, 1998. 75(5): 342]. Metrics of image quality (IQ) can quantify 6 just noticeable differences in blur before 1 line of acuity is lost [J Cataract Refract Surg, 2011. 37(8): 1523] and visual acuity (VA) is highly correlated (R2> 0.8) to change in metrics of image quality when inducing several lines of change in acuity. However, the correlation between acuity and metric values is not known as the change in acuity approaches the test re-test reliability of acuity. The purpose here is to determine how the coefficient of determination (R2 ) between change in VA and change in IQ metric/s value changes as the range is decreased from 8 lines to the test re-test reliability of acuity measurement (1 line).
The Thibos virtual eye generator [Opththalmic Physiol opt, 2009;29(3):288-291] was used to generate 60 WFEs whose log visual Strehl ratio varied between 0.0 and -1.8 in approximately equal steps. For each of the 60 WFEs, three unique logMAR acuity charts were generated (0.8 to -0.3 logMAR). For each chart, VA was measured up to the 5th letter missed for each of 3 normal subjects. Change in visual acuity was linearly regressed against change in 29 IQ metrics to determine the coefficient of determination when the change in visual acuity was decreased in 1 line steps from 8 lines to 1 line.
See Figure. Extrapolation from the fitted line shows that R2 goes to zero at ~3 letters (0.6 of a line), which is below the test retest reliability for acuity measurement (1 line). With a 2 line loss (or one line above the 95% confidence level for the measurement of acuity), 42% of the variance in acuity is accounted for by the metrics. The function begins to asymptote between 4 and 5 lines, and little improvement in correlation is seen with additional lines of acuity change.
Image quality metrics can detect blur prior to an acuity loss (prior work), begin to be correlated with acuity below the re-test reliability for acuity testing and reach a maximum R 2 > 0.8. These two experimental observations suggest that IQ metrics can serve as an objective tool for evaluating and designing therapy intended to improve visual performance, and that correlations are primarily limited by the acuity task and not by the IQ metrics.
This PDF is available to Subscribers Only