Parsing Heterogeneity in Autism Spectrum Disorders Using Measures of Dynamic Visual Scanning

Background: Eye-tracking technology has been widely adopted in autism research to evaluate social attention, but at present, most work has focused on summary measures of visual fixations on specific regions-of-interest, such as the eyes or mouth. While these measures have generally revealed group differences between children with an autism spectrum disorder (ASD) and typically-developing (TD) children, they miss some aspects of how children engage with and respond to dynamically changing social environments. Recent research with school-age children with ASD revealed that while children with different IQ profiles had similar overall amounts of visual fixation on regions-of-interest, the social adaptive value of what they fixated significantly differed based on IQ profile. To better understand how those differences emerge and are manifest in dynamic visual scanning, our laboratory has developed a novel approach to map and quantify the time-varying visual salience of stimuli in the environment, allowing us to determine not just overall amount of eyes fixation, but whose eyes were fixated and at which moments in time.

Objectives: The aims of the current study are (1) to assess alterations in time-varying visual salience as a diagnostic predictor of ASD and (2) to investigate how different patterns of time-varying visual salience are related to an individual’s level of social and cognitive functioning.

Methods: Eye-tracking data were collected while school-age children with ASD (n=109) and chronological age-matched TD peers (n=26) viewed video scenes of children and adults engaged in social interaction within naturalistic visual settings. The ASD sample represented a broad range of level of social disability and cognitive functioning. A subset of children with ASD (n=37) matched to TD peers on verbal, nonverbal, and full-scale IQ were used for between-group comparisons whereas the full ASD group was used for within-group examinations of ASD heterogeneity. Examining data from TD participants, the visual salience of all areas of the onscreen image was calculated by kernel density analysis through the duration of the videos to create a continuous measure of normative time-varying visual salience. We derived measures of deviation therefrom for each participant and then compared across matched diagnostic groups as well as across cognitive profile subgroups of children with ASD based on full-scale IQ and the discrepancy between verbal and nonverbal IQ.

Results: Preliminary analyses suggest that deviations from normative time-varying visual salience are more robust classifiers of ASD than summary visual fixation measures and that the degree of deviation is related to level of social disability. Ongoing analyses are examining how an individual’s cognitive profile may modulate the relationship between time-varying visual salience and social functioning.

Conclusions: The present study proposes that a measure of time-varying visual salience may be a reliable and useful diagnostic marker for autism. In addition, results on differences in dynamic visual scanning patterns in cognitive profile subgroups of children with ASD not only suggest differences in etiologies, but may also support targeted interventions tailored to an individual’s specific learning style.