Cortical Pooling Analysis of Perimetic Sensitivity for Conventional and Sinusoidal Stimuli

Abstract

Abstract: : Purpose: New display technologies allow perimetric testing to employ sinusoidally–modulated stimuli, such as in frequency–doubling perimetry. Sinusoidal stimuli are quite different from the luminance increments used in conventional perimetry, yet there are several reports that glaucomatous defects tend to be similar for both types of stimuli (e.g., Pan et al. ARVO 2003, #56; Harwerth et al. American Academy of Optometry, 2003). We employed a quantitative model of cortical pooling to interpret these empirical findings. Methods: Computational modeling of contrast sensitivity was performed with probability summation across arrays of spatial filters, which pooled responses of ganglion cell mosaics with varying amounts of heterogeneous damage (Swanson et al. 2004, IOVS 45:466). Sensitivity was computed for conventional size III stimuli (0.43 deg luminance increments), as well as for sinusoidal gratings windowed either at two cycles or with circular Gaussians that yielded 1 octave spatial bandwidth. Sinusodial stimuli were varied in spatial frequency from 0.25 c/deg to 4.0 c/deg. Spatial filters were varied in peak spatial frequency and in spatial and orientation bandwidth. For each array of filters, effects of heterogeneous cell loss were analyzed by comparing the relative rates of decline for perimetric sensitivity and ganglion cell number. Results: For most conditions, simulations found that perimetric sensitivity for all stimuli declined at the same rate as ganglion cell number until 90% ganglion cell loss, except when ganglion cell spacing or filter peak frequency were large relative to stimulus size (in which case perimetric sensitivity declined more slowly ganglion cell number). For ganglion loss greater than 90%, perimetric sensitivity to stimulus size III tended to decrease more rapidly than ganglion number, and sensitivity for sinusoidal stimuli decreased more slowly. As a result, depth of perimetric defect was the same across stimuli until 90% ganglion cell loss. Conclusions: Quantitative modeling of cortical pooling explains how depth of glaucomatous perimetric defects could be similar for conventional (size III) stimuli and sinusoidal stimuli, up to advanced levels of ganglion cell loss. The predictions of the model are consistent with recent empirical data comparing depth of glaucomatous defects for conventional and sinusoidal stimuli.

Keywords: visual fields • contrast sensitivity • computational modeling