Abstract
Purpose: :
Standard perimetry suffers from high test-retest variability in the presence of scotomas. The blind spot is a model for studying scotomas, and high variability is found at the edges of dense scotomas and of the normal blind spot (Wyatt et al, ARVO 2008). We have proposed that small fixational eye movements can generate large variations in sensitivity near such boundaries; however, in spite of many similarities, we found variability at scotoma edges was approximately twice as great as variability at blind spot edges. This confirms Häfliger & Flammer (1989, 1991), who concluded that the blind spot differs fundamentally from a scotoma. But is this necessarily true? A model of perimetry, visual field, and fixation was developed in order to explore the difference.
Methods: :
Retina was modeled as two uniform areas with thresholds "30 dB" (normal) and "0 dB" (blind), separated by a boundary consisting of a sharp edge or a linear ramp. Gaze direction was a normal distribution with SD 0-1.25 deg. Intrinsic noise was modeled with a normal distribution with SD = 1 or 2 dB added to threshold. The position of test locations relative to the boundary was either random or 0-2 deg in 0.25 deg steps. "Experimental" thresholds were determined with a full-threshold algorithm employing a 4 dB / 2 dB staircase with 2 reversals. Initial test luminance was taken to be Istart = 10 dB or 20 dB. To examine overall statistics, large numbers of repeats were performed; to simulate experimental data, 4-6 repeats were performed for each test location in each eye.
Results: :
Visual field boundaries mapped with perimetry are "blurred" by fixational eye movements; the broader the distribution of gaze direction, the greater the extent of the blur. This can give rise to very similar sensitivity maps of sharp edges vs. ramps, yet markedly different variability maps.[E.g., comparing an edge to a 1 deg ramp with SD(gaze) = 0.5 deg, ΔThreshold(edge vs ramp) ≤ 2 dB ≈ 0.37 * local SD(Threshold), but Max Variability ratio(edge / ramp) ≈ 150%.] Data from simulated tests on human subjects with 4-6 repeats were very similar to experimental data from blind spots and scotoma edges. In addition, the model produced an explicit form of perimetric "starting point bias" (e.g., Spenceley & Henson, 1996; Turpin et al, 2003).
Conclusions: :
Sharp scotoma edges vs. more gradual blind spot edges could account for the variability difference between blind spots and scotomas. Further plausibility is provided by early blind spot data (Israel, 1968) and by the relevant anatomy. Testing with precisely stabilized test locations could confirm or reject this hypothesis.
Keywords: perimetry • neuro-ophthalmology: diagnosis • visual fields