Abstract
Higher order visual perceptual deficits are a cardinal feature in individuals with cerebral visual impairment (CVI). We investigated visual search behavior in CVI participants (n=16; mean age=16.00 yrs ± 5.03 SD) and neurotypical controls (n=31; mean age=20.59 yrs ± 5.56 SD). Participants searched for a target in a series of 80 naturalistic LabelMe scenes (mixture of indoor and outdoor scenes, fixed presentation time of 4 sec) in response to either an image cue or text cue while gaze behavior was recorded with an eye tracker (Tobii 4C; 90 Hz). Target identification accuracy and reaction time (RT) were collected. Receiver operating characteristic (ROC) analyses comparing the distribution of eye gaze patterns to an image saliency (Graph-Based Visual Saliency) model and semantic saliency (Linguistic Analysis of Semantic Salience) model were also carried out. Compared to controls, CVI participants were less accurate and slower in finding the target for both the image cue (accuracycontrols=91.05% ± 11.70 SD, accuracyCVI=50.63% ± 33.15 SD, t(16.95)=4.72, p<0.001, d=1.89; RTcontrols=1.20 sec ± 0.11 SD, RTCVI=1.77 sec ± 0.34 SD, t(16.78)=-6.44, p<0.001, d=-2.59) and text cue (accuracycontrols=88.06% ± 10.03 SD, accuracyCVI=42.81% ± 30.10 SD, t(16.74)=5.84, p<0.001, d=2.35; RTcontrols=1.37 sec ± 0.18 SD, RTCVI=2.00 sec ± 0.53 SD, t(15.61)=-4.446, p<0.001, d=-1.87) tasks. For the image cue task, controls and CVI participants showed more similar ROC scores with respect to the image saliency model (ROCcontrols=0.80 ± 0.04 SD; ROCCVI=0.74 ± 0.08 SD, t(19.27)=3.31, p=0.002, d=1.23). In contrast, for the text cue task, there was a greater disparity in ROC scores with respect to the semantic model (ROCcontrols=0.83 ± 0.06 SD; ROCCVI=0.65 ± 0.12 SD, t(17.66)=5.58, p<0.001, d=2.16). These findings suggest that CVI participants tend to rely more on image saliency than semantic saliency cues during visual search of naturalistic scenes.