The Effect of Stimulus Size on the Effective Dynamic Range of Perimetry

Abstract

 

Purpose

 

When sensitivities from static automated perimetry vary between 0dB and 15-19dB this may reflect the random nature of stimulus responses rather than true change (Goren et al, ARVO 2013). This conforms to predictions from neurophysiology: retinal ganglion cells approach a maximum firing rate, and so further increasing stimulus contrast would not increase the response probability. In this experiment, we examine the effect of increasing stimulus size on this limit.

 

Methods

 

34 subjects with moderate to advanced glaucoma were tested with Size III and Size V stimuli using the method of constant stimuli, with 35 presentations at each of 7 contrasts, on an Octopus perimeter controlled by the Open Perimetry Interface. Four locations were chosen per eye. At the two most damaged locations the maximum available contrast (equivalent to 3.7dB on the HFA perimeter, the units used in this report) was among those tested. The false positive rate was estimated from 50 catch trials. A cumulative Gaussian curve was fit to the response probabilities at each location, with free parameters: mean, standard deviation and maximal response probability (the probability the subject would respond to a very high contrast stimulus). In many cases the maximal response probability was <50%, so sensitivity as conventionally defined based on 50% response does not exist. Therefore the fitted curve was used to calculate Sens₂₅, the contrast giving a 25% response probability.

 

Results

 

As seen in Figure 1, when Sens₂₅ is low the response probability for the maximum available contrast is well below 100%, making perimetric sensitivities unreliable. Using segmented orthogonal regression, this probability began to decline at Sens₂₅=25.2dB for Size III (95% confidence interval 23.2 - 29.4dB from bootstrapping), and 25.4dB for Size V (22.8 - 26.9dB). As shown by the regression line in Figure 2, among locations with Sens₂₅≥19dB for both stimulus sizes, the sensitivity averaged 5.5dB higher for Size V stimuli (p<0.001, GEE regression)

 

Conclusions

 

The maximum response probability starts to decline when Sens₂₅ is below around 25dB. This decline defines the lower limit of the effective dynamic range of perimetry, which we have found to occur at 15-19dB. The limit is similar for size III and V stimuli. However, when using Size V stimuli, sensitivities are usually greater, and so locations remain within the effective dynamic range later into the disease.

 

 

Figure 1

 

 

Figure 2

 

Keywords: 642 perimetry • 758 visual fields • 531 ganglion cells