Estimating Thresholds in Conventional and High-Pass Resolution Perimetry Using Computer Simulation

Abstract

High-pass resolution perimetry is a new technique that uses a vanishing optotype as the test stimulus. This type of stimulus is thought to produce steeper frequency-of-seeing curves than the stimulus used in conventional automated perimetry, resulting in a lower variability in threshold estimates. The purpose of this study was to compare the stimulus properties in both conventional and high-pass resolution perimetry in computer simulation experiments. The computer model contained stimulus-response data sets obtained from frequency-of-seeing experiments in 11 normal subjects tested with each perimetric technique at four retinal locations [(0 degrees , 30 degrees ), (0 degrees , 15 degrees ), (0 degrees , - 15 degrees ), and (0 degrees , - 30 degrees )]. We programmed eight staircase procedures varying from a coarse one with one reversal to an elaborate procedure with four reversals to look at the relationship between accuracy (the difference between the estimated and actual threshold) and efficiency (number of presentations required for estimate) with each technique. Threshold estimates were made 100 times for each location, strategy, perimetric technique, and subject. Our results showed that an increase in accuracy was always coupled with a decrease in efficiency and that the relationship between these two parameters was similar between the two techniques. Some evidence, however, is presented to show that compared to conventional perimetry relatively simple staircase procedures may be used with high-pass resolution perimetry to attain threshold estimates of similar accuracy.