Abstract
Purpose :
Previous studies show that some visual field (VF) defects are detectable via analysis of visual search behaviour, for example: when watching video, or performing daily living tasks. Here, we developed an alternate approach that measures the number of fixations to find targets on a background with spatial frequency content similar to natural scenes (1/f noise). We chose this approach because there is an established theoretical framework linking target detectability in the image to the predicted number of fixations[1]. We conducted a proof-of-concept experiment to test if the method can detect VF loss.
Methods :
21 older adults [aged:61-79 years] and 20 people with a clinical diagnosis of glaucoma [aged:59-84 years] participated. Participants visually searched for a Gabor (6c/°) that appeared in one of 25 possible locations within a 15° (visual angle) 1/f noise background (RMS contrast: 0.20). Gabor contrast was set at the 95% probability of seeing limit estimated from frequency of seeing curves in 28 older controls. The outcome measure was the number of fixations required to find the target, measured using a Gazepoint GP3 eye tracker. Each target was presented 20 times, in randomized and interleaved order. Normative limits (NL: 95%,98%,99%) for the number of fixations at each location was estimated. Procedure performance was assessed by calculating sensitivity (SE) and specificity (SP) for different combinations: NL, target locations with fixations outside NL (varied: 0-25) and number of repeated target presentations (varied: 1-20). The highest Area Under Curve (AUC) (computed as partial AUC (pAUC) for SP higher than 80%) was chosen; from this curve the criteria with highest SE for SP greater than 95% was selected.
Results :
Older adults made a median of 2-3 fixations (Interquartile range (IQR):2-4) to locate the target on the 1/f noise for all locations. The VF was flagged “abnormal” when the fixation number was greater than the 99% NL for three or more locations (85% SE, 95.2% SP, 88% pAUC) averaged over two repeated presentations. The median time for controls to perform the task was 46 (IQR:33-65) seconds.
Conclusions :
Our prototype test demonstrated effective and efficient screening of abnormal areas in central vision and was well tolerated. While we included people with glaucoma here, the method is designed to generalize to any form of central VF loss.
1.Najemnik & Geisler. Nature, 2005:434.
This is a 2021 ARVO Annual Meeting abstract.