25793
Patterns of Visual Engagement Identify Distinct Subgroups of School-Age Children with ASD

Thursday, May 11, 2017: 12:00 PM-1:40 PM
Golden Gate Ballroom (Marriott Marquis Hotel)
J. R. Yurkovic1, S. Gillespie2, W. Jones3, A. Klin4 and S. Shultz5, (1)Psychological and Brain Sciences, Indiana University - Bloomington, Bloomington, IN, (2)Emory University School of Medicine, Atlanta, GA, (3)Marcus Autism Center, Children's Healthcare of Atlanta and Emory University School of Medicine, Atlanta, GA, (4)Marcus Autism Center, Children's Healthcare of Atlanta & Emory University School of Medicine, Atlanta, GA, (5)Marcus Autism Center, Children's Healthcare of Atlanta, Emory University, Atlanta, GA
Background: The vast heterogeneity in Autism Spectrum Disorder (ASD) is an obstacle to advancements in identifying and treating causes of the disorder. Eye-tracking measures of atypical visual engagement with the social world—a quantitative metric that captures a defining symptom of the condition – provides a promising means for deriving more homogeneous subgroups (Rice et al., 2012; Campbell et al., 2014). Parsing heterogeneity in ASD by measuring visual scanning during dynamic social scenes may contribute to the identification of intermediate phenotypes for genetics research, and to the development of interventions optimized for individual children.

Objectives: To examine: (1) whether subgroups of children with ASD can be reliably identified based on patterns of variability in social visual engagement; and (2) whether the subgroups differ on standardized measures of social disability.

Methods: A heterogeneous sample of children with ASD (n=178, age=10.51(3.19)) watched age-appropriate, socially-relevant videos while eye-tracking data were collected. Percent fixation on eyes, mouth, body, and object regions was calculated for each child. The Hopkins index was used to assess whether clusters are reliably identifiable by variability in visual scanning. Unsupervised statistical learning methods, including Principal Components Analysis (PCA) and hierarchical clustering, were utilized to visualize and identify clusters of different visual fixation patterns. Analyses were performed on scaled data and inter-observation distances were calculated via Euclidean distances. A three-cluster solution was achieved through hierarchical clustering of fixations to eyes, mouth, body, and object in each participant pair.

Results: The Hopkins Index indicated the clustering tendency as appropriate (0.16<0.5). PCA and hierarchical clustering analyses identified three clusters of children with ASD (Figures 1a & 2b). As expected, ANOVA and post-hoc Tukey analyses revealed significantly different fixation patterns between clusters (Figures1a,2b), with Cluster 1 fixating more on eyes, Cluster 2 on mouths, and Cluster 3 on objects. Children in Cluster 1 had lower Vineland communication scores (p=0.051) compared to Cluster 2, suggesting that higher eye-looking may not be associated with greater adaptive skills in this subgroup of children. Consistent with previous reports that higher mouth fixation is associated with lower social disability (Rice et al., 2012) children in Cluster 2 had significantly lower ADOS symptom severity (p=0.051) than those in Cluster 3, and higher Vineland communication scores (p=0.051) than those in Cluster 1 (Figure 2b). Finally, children in Cluster 3 had higher ADOS symptom severity scores compared to those in Clusters 1 (p=0.098) and 2 (p=0.051), suggesting that high levels of object fixation are associated with greater social disability. Cluster 3 also had a significantly higher proportion of males (p=0.024) than Clusters 1 and 2 (Figure 2a).

Conclusions: Results demonstrate that variability in visual engagement during viewing of dynamic social scenes can be used to identify more homogeneous subgroups of children with ASD. These subgroups displayed distinct patterns of visual attention, and varied by gender and measures of social ability. Future analyses will examine whether the social adaptive value of visual scanning patterns vary for different subgroups of children with ASD, a critical step towards creating interventions optimized to the individual’s learning style.