Purchase this article with an account.
Robert W Massof, Daniel Laby, David Meadows, David Kirschen; Model that Summarizes the Effects of Blur on Psychophysical Measures of Visual Function with a Single Latent Variable. Invest. Ophthalmol. Vis. Sci. 2019;60(9):5923.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
To develop and validate a model of a single psychophysical outcome measure that efficiently summarizes and predicts the effects of blur on contrast sensitivity as a function of stimulus size and exposure duration.
Model. We developed a psychophysical model of contrast sensitivity as a function of stimulus size and exposure time based on an empirical model offered by Alexander et al (IOVS 1992;33:1846-1852) that was expanded to incorporate Bloch’s, Ricco’s, Blondel-Rey’s, and Piper’s laws. Combined with standard psychometric functions that asymptote at chance for probability correct, the model predicts iso-threshold surfaces in stimulus size/contrast/exposure time 3-space. A Rasch model was employed to estimate measures of a single latent variable that summarizes iso-threshold surfaces. Methods. We analyzed an existing database for 180 different Landolt rings, each with a unique combination of 6 sizes (S), 6 contrasts (C), and 5 exposure times (T), presented once on a computer display at one of 4 randomly-chosen orientations to 296 normally-sighted subjects viewing with different amounts of blur due uncorrected refractive error. Subjects had to identify the orientation of each Landolt ring and their responses were scored as correct (1) or incorrect (0).
Rasch analysis was used to estimate item measures for each stimulus, I(S,C,T), and person measures for each subject, θ, in logit units. Our psychophysical model, with the assumption of a cumulative logistic psychometric function, was used to estimate size thresholds (S’) as a function of C and T and contrast thresholds (C’) as a function of S and T for each subject using MLE. If a single latent variable can summarize iso-threshold surfaces, then θ-I(S,C,T) =a1[log S – log Sθ’(C,T)]+b1= a2[log C – log Cθ’(S,T)]+b2. The figure illustrates that this prediction is confirmed (however due to boundary limitations, the predictions fail at the extremes).
This study validates the model and confirms the hypothesis that a single latent variable can be estimated and used to summarize iso-threshold surfaces in S,C,T space for a sample of subjects who differ in blur.
This abstract was presented at the 2019 ARVO Annual Meeting, held in Vancouver, Canada, April 28 - May 2, 2019.
This PDF is available to Subscribers Only