Automated static threshold perimetry using a remote eye tracker

Abstract

 

Purpose

 

Current methods of Static Threshold Perimetry require (i) an explicit, button-press response (precluding testing of infants) and (ii) expensive, specialised equipment. Here we present a novel measure that addresses these problems by combining a cheap, commercially available, eye tracker (Tobii EyeX: $135), with an ordinary desktop computer.

 

Methods

 

Luminance detection thresholds were measured monocularly in 7 healthy adults (additional data collection ongoing), using both a Humphrey Field Analyzer [HFA] and an automated remote eyetracking [ARE] procedure (Fig 1A). The eye tracker was used to present stimuli relative to the current point of fixation, and to assess whether the participant made an eye-movement towards the stimulus. In both tests, Goldman III stimuli were arranged on a 24-2 grid, and were presented individually against a 10 cd/m² white background. Participants completed each test twice (same eye) in order to assess test-retest reliability.

 

Results

 

The pointwise Coefficient of Repeatability was similar for the two tests (ARE: 8.1 dB. HFA: 6.3 dB). Differences in mean sensitivity to stimuli in the central 10° and those located more peripherally (10—24°) were observed in both the ARE (CI_95% = 0.9—2.7 dB) and the HFA (CI_95% = 2.3—4.6 dB). Furthermore, as shown in Fig 1B, the ARE was able to differentiate between the blind spot and surrounding retinal locations (t₆ = -3.1, p = 0.021).

 

Conclusions

 

An eye tracker can be used to perform Static Threshold Perimetry based on eye movement responses alone. The ARE was sensitive to normal variations in sensitivity across the healthy eye, and could isolate the blind spot. It may therefore be capable of detecting visual field deficits, including acute scotomas. Its low price and ease of use could make such a test particularly effective as a means of screening infants.  

 

Fig 1. (A) Example trial for the ARE. Here a test point is presented at [-3° -6°], relative to the current point of fixation. Fixation at trial onset was unconstrained. The monitor was a Samsung 305T LCD, gain-corrected in software for uniformity. (B) Mean threshold data from the ARE, computed from 4 participants tested with their left eye. Numbers show the final threshold estimate at each location, in dB (higher = more sensitive). Note that the blind spot was measured in an identical manner to all other points, with no prior assumptions or constraints, in order to simulate an unknown scotoma.