A new look at threshold estimation algorithms for automated static perimetry.

Automated perimetry is often associated with lengthy test times when a staircase algorithm is applied. This arises because the fixed step sizes used during threshold estimation (e.g.,4/2 dB) yield reduced test efficiency, with test times being dependent on the relative positioning of the start and endpoints, as well as the step size. Neighborhood logic may speed up the process, although several presentations are still required for normal threshold values and many more presentations are required for abnormal values. We consider whether there is any justification for using a fixed step size during the threshold procedure. We show how empirical data can be applied, within a Bayesian framework, to reduce test time with little or no loss of accuracy. Finally, we demonstrate the effect that the starting probability density function can have on test efficiency by implementing an empirically determined and bimodal probability density function that provides fast outcomes.