Learning how to recognize objects with a simulated scotoma: an eye-tracking analysis

Abstract

Purpose : People with glaucoma or macular degeneration suffer from a scotoma in the visual field that leads to severe impairments in visual processing, including face recognition and object recognition. Here, we investigate how peoples’ eye movements adapt and optimize in response to a simulated scotoma in an object recognition task using a gaze-contingent display.

Methods : Ten participants (4 males, mean age=26.2y) with normal visual acuity (0.1±0.1logMAR) and right eye dominance were recruited. In each of the 120 trials per session, they had to indicate whether a randomly placed object was either a non-sense (novel) object or a known (familiar) object. Participants sat 60cm away from a calibrated, monitor-based eye tracker (Tobii PRO TX-300) with their head on a chin rest. Foveal input was blocked by applying a grey Gaussian mask with width 3.15deg to the fovea. Vision was left clear in a window defined by a Gaussian mask of width 1.94deg that was placed at 6.3deg to the left of the fovea. The remaining visual field was blurred by a Gaussian of 7deg width to simulate peripheral, blurred input.
Participants had some practice trials to become comfortable with the altered visual input before the experiment. Training proceeded with sessions twice a week until 9 sessions were done. No feedback about performance was given during the training. Eye movements were recorded at 300Hz as well as task accuracy and response time.

Results : Participants’ median performance started with 85% correct and reached close to 100% performance in the final session with response times staying level around 10s. Importantly, participants made significantly shorter saccades both in length and duration between the first and the last session, as well as fewer blinks.

Conclusions : Participants learned to solve the task relatively quickly with performance plateauing after the 7^th session. The difference in saccade metrics shows that eye movements adapted to the scotoma presence by shortening their saccades - presumably to exert more fine-grained control with the limited visual input afforded. Current analyses are focused on spatial statistics of fixations and temporal modeling of saccade and fixation dynamics.

This is an abstract that was submitted for the 2018 ARVO Annual Meeting, held in Honolulu, Hawaii, April 29 - May 3, 2018.

 

Fig. 1a) Recognition accuracy across sessions.
Fig. 1b) Response times across sessions.

 

Fig. 2 Eye movement metrics for the first and the last session. Figure titles show the result of a paired Wilcoxon test.