Visual Search Patterns During Exploration of Naturalistic Scenes are Driven by Saliency Cues in Individuals with Cerebral Visual Impairment

Abstract

Two main sources of information have been identified to explain what image features guide gaze behavior, namely, image salience (associated with bottom-up processing) and image semantics (influenced by top-down processing). In this study, we compared the relative influence of these two factors while subjects searched for a predetermined target in a naturalistic scene. We also compared visual search performance in individuals with neurotypical development and those with cerebral visual impairment (CVI); a neurodevelopmental condition associated with higher-order visual processing deficits. Participants searched for the prompted target (presented as either an image or text cue) in a series of indoor and outdoor naturalistic scenes. Gaze behavior was recorded with an eye tracker, and behavioral outcomes were success rate, reaction time, visual search area, and number of fixations. A receiver operating characteristic (ROC) analysis was performed to compare the distribution of individual eye gaze patterns based on predictions of image salience (using Graph-Based Visual Saliency) and image semantics (using Global Vectors for Word Representations combined with Linguistic Analysis of Semantic Salience) models. We found that overall, CVI participants were less likely and were slower to find the target than controls. Visual search behavior in CVI was also associated with a larger visual search area and a greater number of fixations. ROC-based analysis revealed that saliency scores were lower in CVI than controls for both the image salience and image semantics model predictions. Furthermore, search strategies in the CVI group were not affected by cue type, although search times and accuracy were correlated with Verbal IQ scores for text-cued searches. Collectively, these results suggest that individuals with CVI do not utilize image or semantic information as readily as controls, in line with their associated deficits related to higher-order processing.