Abstract
Purpose: :
If, as previously proposed, a substantial part of test-retest variability results from fixational eye movements (Wyatt et al, Vis Res 2007; ARVO 2009), could this be corrected post hoc with eye movement information? Preliminary testing is described.
Methods: :
The right eyes of 7 normal subjects were tested with a rectangular test array (4x7; 2 deg spacing), extending from 11 to 17 deg in the temporal field, encroaching on the blind spot. Additional control locations were placed in nasal, superior, and inferior visual field. Visual stimuli (size III, 0.1 sec) were presented on a CRT (Radius) with a background luminance of 5 cd/m^2. Testing employed a 2 dB/1 dB two-reversal staircase. Gaze direction was recorded continuously with a video-based eyetracker (ISCAN). Subjects participated in at least two testing sessions.
Results: :
Variability of gaze direction could be described by a normal distribution with SD between 0.2 and 0.5 deg. For each subject, one or more test locations showing substantial test-retest variability were selected for further analysis. 15 such test locations were selected in the 7 eyes. For a given location, gaze direction at the time of each test flash was determined from the eyetracker record. For 5 of the eyes, vertical gaze direction data were contaminated by lid intrusion; to the extent possible, test locations were selected near vertical blind spot edges for those eyes. Stimuli which elicited "seen" and "not-seen" responses were plotted as functions of gaze direction, and the data were examined for evidence of a gaze dependence of visual sensitivity: 2- or 3-dimensional plots of seen and not-seen luminances vs test location (x or x-y) were constructed and fitted with simple functions S(x) or S(x,y) describing sensitivity. Of the 15 locations analyzed, data for 10 locations could be plausibly accounted for by a gaze dependence appropriate to the test location, 1 location did not show a clear dependence, 2 locations produced data that were contrary to an appropriate dependence, 1 location produced 2 solutions (1 appropriate), and 1 location gave inadequate data. Including gaze information in analysis of test-retest variability reduced variability by an average of 2/3.
Conclusions: :
The results suggest that near steep edges of sensitivity, visual sensitivity is often gaze-dependent, suggesting that eyetracker data could be used to reduce variability. Although Henson et al (1996) did not find a relationship between test-retest variability and extent of fixational eye movement, such as might be expected from the present findings, the power of an uncontrolled variable -- test location relative to the sensitivity edge -- may account for that.
Keywords: perimetry • neuro-ophthalmology: diagnosis • visual fields