Abstract
We developed a quantitative two-stage neural model relating perimetric sensitivity to glaucomatous ganglion cell damage. In the first stage, responses of a mosaic of ganglion cells were computed for the standard perimetric stimulus (0.43 deg. diameter white increment). In the second stage, outputs of the ganglion cell array were analyzed by psychophysical mechanisms constructed as families of orientation-tuned spatial filters. Perimetric sensitivity was computed by the probabilistic sum of sensitivities of these mechanisms. The model successfully captured two key features of data in the perimetric literature: normal spatial summation functions, and the relation between depth of defect and threshold variability in the presence of ganglion cell loss. Three main conclusions of the model were robust in the presence of variations in parameters used:
Spatial summation functions were determined primarily by the spatial tuning of the orientation-tuned filters, not by the ganglion cell receptive field parameters.
When stimulus size is similar to critical area of spatial filters, percent of ganglion cells lost is linearly related to depth of perimetric defect if defect is expressed in linear units rather than in dB units.
Increased variability in defective areas can be attributed to effects of small eye movements (less than 1 degree) interacting with inhomogeneous ganglion cell damage.