Abstract
Purpose: :
In standard automated perimetry, as implemented in the Humphrey Field Analyzer, contrast sensitivities are measured down to 0dB, representing a 3180 cd/m2 stimulus presented on a 10 cd/m2 background. However, thresholds measured using these high levels in glaucoma patients are extremely variable, with 95% confidence intervals for test-retest covering 20dB or more. Therefore, it is reasonable to ask whether such results are meaningful, or just the result of high variability and random chance.
Methods: :
The integrated product of a stimulus with the receptive field of a single retinal ganglion cell (RGC) can be used to find the contrast gain of that RGC at low contrast. At higher contrasts, the RGC saturates. Its firing rate in response to a stimulus of contrast x can be modeled by a Michaelis-Menton function, Rate = Base + Max*x / (Max/Gain + x) (Kaplan & Shapley 1986, PNAS 83:2755-2757), where Base and Max are the baseline and maximum firing rates of the RGC. Using published results giving the dimensions and peak contrast gain of their receptive fields (Croner & Kaplan 1995, Vis Res 35:7-24), responses of RGCs to perimetric stimuli of differing intensities were calculated.
Results: :
At a mid-peripheral location in the visual field, modeled healthy magnocellular RGCs attained 95% of maximum firing rate in response to a 23dB stimulus covering their receptive field. For parvocellular RGCs, a 16dB stimulus produced 95% maximum firing.
Conclusions: :
Increasing the stimulus level beyond 15dB does not significantly increase the response from healthy RGCs. If the visual cortex receives the same signal from the retina, it will produce the same response. Therefore, presenting stimuli more intense than 15dB to glaucoma patients would only yield useful information if some sort of RGC dysfunction were present that reduced the contrast gain. Experiments are needed with glaucoma patients to determine whether sensitivities below 15dB can be considered to be informative.
Keywords: perimetry • ganglion cells • computational modeling