July 2018
Volume 59, Issue 9
ARVO Annual Meeting Abstract  |   July 2018
A novel Bayesian approach to testing and analyzing visual acuity
Author Affiliations & Notes
  • Luis A Lesmes
    Adaptive Sensory Technology, San Diego, California, United States
  • Footnotes
    Commercial Relationships   Luis Lesmes, Adaptive Sensory Technology Inc (I), Adaptive Sensory Technology Inc (P), Adaptive Sensory Technology Inc (E)
  • Footnotes
    Support  None
Investigative Ophthalmology & Visual Science July 2018, Vol.59, 1073. doi:
  • Views
  • Share
  • Tools
    • Alerts
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Luis A Lesmes; A novel Bayesian approach to testing and analyzing visual acuity. Invest. Ophthalmol. Vis. Sci. 2018;59(9):1073.

      Download citation file:

      © ARVO (1962-2015); The Authors (2016-present)

  • Supplements

Purpose : To develop a computational framework for improving the precision for testing visual acuity (VA), and the detection of its changes

Methods : We developed a novel Bayesian method (Fig 1) for testing and analyzing VA that: (1) uses high-density sampling of optotype size (.02 logMAR resolution); (2) describes the full acuity function (threshold and range;Fig 1a,b); (3) considers acuity behavior via composite multiple-optotype psychometric functions, Fig1c; (4) applies a Bayesian adaptive strategy to gain information about acuity threshold and range, and (5) calculates an acuity change index based on ROC analyses of Bayesian posteriors. In this proof-of-concept study, simulations compare the stimulus sampling, precision/repeatability, and sensitivity to change for Bayesian VA and e-ETDRS testing. Although VA testing is designed to estimate threshold-size for correctly reporting 3/5 optotypes- Carkeet et al (2001;2017) have demonstrated that blur/contrast conditions can affect acuity range, and thereby increase VA test variability. Simulated observers were drawn from a sample population with mean VA=.30 logMAR, (s.d., .30) and different parameters of acuity range (from .10 to .80).

Results : For e-ETDRS, varing from low to high acuity range doubled the test-retest variability from .05 to >.10, and test-retest precision decreased from 92% to 82%. The Bayesian VA algorithm continues to converge with increased test duration. For testing 10-30 rows of five optotypes (50-150 letters), test-retest variability exhibited for low and high ranges varied from .015-.033 logMAR to .07-.11 logMAR, respectively. Corresponding values of test-retest precision varied from 97% to 84%. The amount of Bayesian VA testing needed to match the precision of e-ETDRS is 5-7 rows. This precision advantage is likely provided by precise stimulus sampling that matches the full acuity function. In a simulation of a 5-letter change in VA, e-ETDRS exhibited a 85% accuracy for detecting acuity change, whereas Bayesian VA testing provided detection accuracy of 86%, 94%, and 97% with testing of 10, 20, and 30 rows.

Conclusions : This study provides a proof-of-concept for Bayesian testing of visual acuity. Estimation of the full acuity function with high-density sampling of optotype size exhibits the potential for sensitive and precise detection of changes in VA.

This is an abstract that was submitted for the 2018 ARVO Annual Meeting, held in Honolulu, Hawaii, April 29 - May 3, 2018.



This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.