Accuracy and repeatability of a prism-based technique of gaze position calibration in 1st Purkinje-image based eye trackers

Title Accuracy and repeatability of a prism-based technique of gaze position calibration in 1st Purkinje-image based eye trackers
Author, Co-Author Swaathi Balaji, Shrikant Bharadwaj
Topic Binocular Vision/Pediatrics
Year
2016
Day
Thursday
Program Number
165020
Room
Ballroom A-B
Affiliation
Abstract

PURPOSE: Accuracy of gaze position measured using 1st purkinje-image (PI) based eye trackers critically depend on the calibration of the individual’s Hirschberg ratio (HR). This study evaluated the repeatability and accuracy of a prism-based technique of determining HR, vis-à-vis, the traditional technique and theoretical predictions.

METHODS: Short- (5min) and long-term (1wk) repeatability of prism-based HR of 30 subjects (18 to 28yrs) was obtained by placing base-out and base-in prisms (0-24Δ in 4Δ steps) before their IR filter occluded left eye while the right eye fixated on a distance target. In the traditional technique, these subjects fixated on distant eccentric targets (±15° in 5° steps). Gaze position was recorded at 50Hz using the PowerRef3® photorefractor and eye tracker. HR was obtained from the linear regression slope of recorded gaze position for each prism power or target eccentricity. Theoretical HR was calculated from the eye’s corneal curvature and AC depth (Jagini et al., 2014).

RESULTS:Intersubject variability of HR ranged from 10.5 to 15.5°/mm, 9.5 to 14.9°/mm and 10.1 to 13.3°/mm in the prism, traditional and theoretical techniques, respectively. The short- and long-term repeatability of prism-based HR was -2.5 to 2.3°/mm and -1.8 to 2.5°/mm, respectively (p=0.4). Prism-based HR was larger than the traditional and theoretical values by 0.77°/mm (range: -3.5 to 3.2°/mm) and 0.67°/mm (-2.7 to 0.9°/mm), respectively (p

CONCLUSION: The intra-subject repeatability of prism-based HR is similar in range to the inter-subject range of HR obtained in this study. The prism-based HR is also significantly overestimated, relative to the other two techniques. These issues may limit the utility of the prism-based technique for calibrating the individual’s HR. 

Affiliation of Co-Authors L V Prasad Eye Institute, Hyderabad, India
Outline