Adaptive kinetic perimetry of the peripheral visual field

Abstract

Purpose : Automated kinetic perimetry (AKP) often shows erroneous “spikes” in isopters. Through computer simulations, we designed a simple but efficient approach for estimating more precise isopter positions.

Methods : A series of prior visual fields (n=149) obtained with manual Goldmann perimetry in patients with glaucoma served as models of the “true” visual field. The simulated AKP strategy estimated isopter positions from at least two responses, and additional stimuli (up to 3, for a total of 5) were presented until the distance between the two closest responses was smaller than a criterion value. “Heavy tailed” response errors were modelled through a mixed Gaussian distribution fitted to experimental data from 10 healthy observers (1200 responses to a V-4e Goldmann stimulus moving at 5°/s; Fig 1A, B). Low, moderate, and high variability observers were simulated by scaling this distribution with factors of 1, 3, and 5. The accuracy and precision of the estimated isopter positions was derived through comparison to the “true” visual field (Fig 1C).

Results : With a criterion value of 5° isopter positions could be estimated to within a precision of 1.2°, 3.3° and 5.4° (standard deviation of difference between estimated and “true” isopter), in observers with low, medium, and high variability, respectively, and systematic error of < 0.01° from the true isopter, in observers with low, moderate, and high variability, respectively. On average, this required 2, 2.5 and 3 presentations per isopter location (range, 2 to 5).

Conclusions : Our simulations suggest that a simple strategy of adding additional stimuli in areas of high variability performed well throughout a large range of observer variability. Useful estimates of the biologically important inferior and temporal visual field periphery should be feasible with fewer than 20 kinetic presentations (~3 minutes). This needs to be confirmed with real patients.

This abstract was presented at the 2019 ARVO Annual Meeting, held in Vancouver, Canada, April 28 - May 2, 2019.

 

Figure 1: Panels A, B show the distribution of responses from observers (blue) and the best-fitting normal distribution (red). Some responses occur far from the expected value (“heavy tails”). Panel C shows simulated partial isopter to Goldmann V-4e stimulus, of the inferior temporal visual field, for an observer with moderate variability. The dark blue curve shows the estimated isopter, derived from the median response (red points). The confidence band contains 80% of responses. A plot of Goldmann perimetry shown in the background.