June 2020
Volume 61, Issue 7
Free
ARVO Annual Meeting Abstract  |   June 2020
Using a Bayesian adaptive method to estimate psychometric function threshold and slope in visual prosthesis recipients
Author Affiliations & Notes
  • Roksana Sadeghi
    Biomedical Engineering, Johns Hopkins School of Medicine, Baltimore, Maryland, United States
  • Chris Bradley
    Wilmer Eye Institute, Johns Hopkins University School of Medicine, Baltimore, Maryland, United States
  • Arathy Kartha
    Wilmer Eye Institute, Johns Hopkins University School of Medicine, Baltimore, Maryland, United States
  • Gislin Dagnelie
    Wilmer Eye Institute, Johns Hopkins University School of Medicine, Baltimore, Maryland, United States
  • Footnotes
    Commercial Relationships   Roksana Sadeghi, None; Chris Bradley, None; Arathy Kartha, None; Gislin Dagnelie, Second Sight (P), Second Sight (C)
  • Footnotes
    Support  UG3 NS095557
Investigative Ophthalmology & Visual Science June 2020, Vol.61, 2209. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Roksana Sadeghi, Chris Bradley, Arathy Kartha, Gislin Dagnelie; Using a Bayesian adaptive method to estimate psychometric function threshold and slope in visual prosthesis recipients. Invest. Ophthalmol. Vis. Sci. 2020;61(7):2209.

      Download citation file:


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

      ×
  • Supplements
Abstract

Purpose : Electrical stimulation of neurons in the human visual pathway creates visual perception. Several pulse parameters need to be optimized to create reliable percepts, typically by measuring stimulus threshold values. The traditional approach in the Argus II, the only commercialized visual implant in USA, combines a method of constant stimuli with a method of ascending limits that increases the stimulus range as needed. This combination is known as Hybrid threshold testing; it requires ~80 trials per electrode, causing the neurons to adapt and thresholds to increase, even if testing of several electrodes is interspersed. Here, we present a Bayesian adaptive threshold test, estimating thresholds on multiple electrodes concurrently.

Methods : As in the hybrid threshold method, a trial starts with a tone followed by a stimulus, with occasional catch trials inserted to estimate the false positive rate. The new algorithm delivers the adaptive stimulus at a given electrode based on all prior responses at that electrode for both threshold and slope values (L. Kontsevitch and C. Tyler 1999). Fewer than 25 trials for each electrode are required to estimate both parameters of the psychometric function. To validate this method, it was tested in Argus II users and the results were compared to hybrid testing results.

Results : Two subjects with an Argus II implant participated. Using both hybrid and adaptive methods, the threshold values of 58 and 14 electrodes were estimated and compared in 7 and 2 sessions for S1 and S2 respectively. The new method narrowed thresholds to within a tight range in 11-25 (average 15) trials per electrode. Threshold values from both methods are highly correlated (r2 = 0.87) in both subjects; the regression line shows that Bayesian estimates fall progressively below hybrid thresholds as thresholds increase. The slope of the Weibull function was estimated as 0.72 and 0.54 for S1 and S2, respectively.

Conclusions : The Bayesian method provides precise estimates of threshold and slope values and is less prone to neuronal adaptation than the hybrid method in Argus II. By virtue of rapid convergence, the new method can be applied to visual prostheses with high numbers of electrodes to estimate visual perception thresholds concurrently and within a single session.

This is a 2020 ARVO Annual Meeting abstract.

 

The comparison between Bayesian adaptive and hybrid threshold testing in 2 Argus II users.

The comparison between Bayesian adaptive and hybrid threshold testing in 2 Argus II users.

×
×

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.

×