Abstract
Purpose :
There is evidence that patients with glaucoma exhibit inaccurate eye movements in relation to their glaucomatous visual field damage (VFD). In this study, we analyzed gaze response to a series of visual stimuli with an eye-tracker and attempted to devise a model to detect glaucoma solely based on gaze response.
Methods :
55 eyes from 39 glaucoma patients and 69 eyes from 36 healthy controls were examined. Participants underwent a monocular gaze response test that requires the subjects to naturally follow a white dot appearing on the screen which was aligned with the standard automated perimetry. Garway-Heath criteria partitioned an eye into five sectors (excluding nasal) and divided into healthy sectors (group 0), healthy sectors with VFDs elsewhere in the same eye (group 1), and glaucoma sectors with VFDs elsewhere in the same eye (group 2). A gaze response based sectorized VFD prediction model was developed using a Random Forest algorithm.
Results :
The disparity between stimulus and gaze response significantly increased as it progressed across groups 0, 1, 2 (all p<0.01). There was no statistical difference in VF sensitivity between Group 0 and Group 1 (p=0.12), but gaze response parameters were statistically different (p<0.01). The disparity parameters had significant correlations with the VF sensitivity and OCT thickness across every sectorized unit (all P<0.01). A Random Forest model yielded an area under the curve value of 0.92.
Conclusions :
This intuitive, user-friendly gaze response test achieved sufficient diagnostic values in detecting glaucoma per sectorized unit.
This abstract was presented at the 2024 ARVO Annual Meeting, held in Seattle, WA, May 5-9, 2024.