Purchase this article with an account.
Andrew Turpin, Allison M. McKendrick, William H. Swanson; Using Reaction Time to Choose Stimuli Levels in Perimetry. Invest. Ophthalmol. Vis. Sci. 2011;52(14):5511.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Previously in perimetry, reaction time (RT) of responses to stimuli has been used to judge false responses. In this study, however, we explore the use of RT to improve stimulus placement in an adaptive thresholding procedure.
Our baseline procedure is a Zippy Estimation by Sequential Testing (ZEST) strategy parameterized for a visual field test based on Gabor patch stimuli [Hot et al, IOVS 2008]. The dynamic range is 2 log units, a uniform prior is used, and the likelihood function (LF) is 0.1+0.8*Gauss(s,0.1), where s is the stimulus value.Thresholds and RT data were collected using this procedure on 9 subjects in at least 60 locations of the visual field. We then fit this RT data with a Gamma distribution with parameters scale=exp(1+0.7*d), and shape=exp(-5+1.2*d), where d is the distance of a stimuli level from measured threshold. This distribution was truncated to the possible domain of d and normalized. Our new procedure (BURTO) alters the LF of the ZEST by multiplying by the probability of observing the reaction time obtained from positive responses. We simulated BURTO and ZEST on four populations of 1000 patients, all of whom had 3% false positive and negative rates; reaction times sampled from the third author's empirical data as a function of d; a psychometric function spread of 0.25 consistent with the literature; and true thresholds -1.8, -1.3, -0.8 and -0.3.
The table shows accuracy, precision and the number of presentations required for the four simulated patient populations. A * indicates statistically significant differences between BURTO and ZEST using a t-test on means, and an F-test on standard deviations at the p<0.01 level.
For a small loss in precision on normal thresholds, BURTO is upwards of 10% faster for normal thresholds, and 5% to 8% faster for depressed thresholds with no loss of precision. Using RT to alter stimulus placement during adaptive procedures may improve perimetric procedures.
This PDF is available to Subscribers Only