Eye-Tracking Features As Diagnostic Markers of Autism Spectrum Disorder, Symptom Severity, and Change over Time

Background: Research in eye-tracking has indicated that individuals with autism spectrum disorder (ASD) may differ from typically developing (TD) individuals in relation to allocation of attention to social and non-social stimuli. Though not conclusive, there is evidence of reduced attention to faces, eyes, and mouths, and increased attention to bodies and non-social elements in ASD. However, large-scale replication of these results is warranted, as is a deeper exploration of underlying relationships between eye tracking measures and clinically-relevant features of autism as well as applicability of these measures across the lifespan.

Objectives: To examine the discriminant validity (ASD vs TD) of different eye-tracking paradigms including dynamic, naturalistic, and static stimuli reported in literature broadly across development (childhood to adult); and to examine the relationship between ASD symptoms in ASD and performance on eye tracking tasks.

Methods: 127 ASD participants between six years and adulthood (mean [SD] age: 14.6 [7.91]) were monitored for attention to screen and pre-identified regions of interest using the Tobii X2-30 Eyetracker. A range of stimuli was presented, including Dynamic videos (Plesa-Skwerer, Chu, Brukilaachio, & Tager-Flusberg, 2016), Biological Motion, and a Visual Exploration Task. Stimuli were viewed at three time points during the course of the observational study. Parent reports on behavior rating scales were obtained at the same time points (0, 4, 8 weeks). Also, 41 TD individuals between six years and adulthood (mean [SD] age= 16.2 [13.18]) were presented with the same stimuli at baseline (0 weeks) only.

Results: Multiple eye-tracking stimuli discriminated between ASD and TD groups. Statistically significant findings included Biological Motion: Time spent looking at biological motion (n=108/38;[ASD/TD] d= -0.93, p<0.001); Dynamic Videos: Time spent looking at screen overall (n=82/38 d=-0.47; p=0.015),time spent looking at faces (n=82/38d= -0.72; p=0.001); and Visual Exploration Task: ASD participants explored less images per total time while looking at the screen (n=108/41); d= -0.63 p<0.001).

Many features across eye-tracking stimuli also correlated with increased ASD symptomology. Statistically significant findings included: reduced attention to screens during videos (n=52; r= -0.50, p<0.001); reduced attention to object of joint attention (moving toy) (n=81; r=0.34, p<0.001);

Finally, there were statistically significant correlations between change in biosensors from baseline with change in reported symptoms; Dynamic videos; change in attention to faces during the opening segment of videos statistically significantly correlated with change in ASD symptoms between baseline and endpoint (n=41; r= -0.42, p<0.01); visual exploration task; change from baseline in amount of time spent looking at images of high autism interest correlated with change ASD symptoms (n=75, r=0.37 p<0.001).

Conclusions: It was possible to discriminate between ASD and TD groups across a broad developmental range on eye-tracking measures that have previously been reported in the literature. For some of these features, there was also a statistically significant correlation with behavior rating scales. Finally, change in eye-tracking features over time was significantly correlated with reported change in behaviors during the same period. Taken together, these findings demonstrate the potential utility of eye-tracking measures as meaningful biomarkers and sensitive outcomes for use in clinical trials.