Feasibility and Preliminary Efficacy of Gazefinder Eye-Tracking Technology for Use with Infants Showing Early Signs of Autism

Background: Identification and diagnosis of autism currently relies on the appraisal of overt behavioural difficulties. Manufactured by JVCKENWOOD Corporation, Gazefinder is a bespoke eye-tracker which may support accurate and efficient autism identification across the lifespan. Gazefinder includes an infra-red eye-tracker, short 2-minute stimulus sequence, and automated visualisation of results and data extraction. Stimulus trials target referential attention to pointed-at objects (vs. distractors), and preferential attention to eyes (vs. mouth), people (vs. geometry) and biological motion (upright vs. inverted point-light displays). Previous studies – all conducted in Japan – suggest Gazefinder differentiates people with and without autism in childhood (Fujisawa et al., 2014) and adolescence/adulthood (Fujioka et al., 2016). The short stimulus sequence is also well tolerated by young, typically developing children (Nishizato et al., 2017). However, the efficacy of Gazefinder to support early identification of autism in infants has yet to be evaluated.

Objectives: To trial the feasible use of Gazefinder – by an independent team, in a Western country –including evaluating preliminary efficacy for supporting the early identification of autism.

Methods: Participants were 54 infants (75% boys) aged 9-16 months (M=12, SD=2) who were eligible for the study on the basis of showing early behavioural signs of autism according to the Social Attention and Communication Surveillance (SACS) protocol (Barbaro and Dissanayake, 2010). Specifically, we recruited infants showing ≥3 (of 5) key markers on SACS 12-month checklist (i.e., absent/atypical eye-contact, response to name, imitation, pointing, and other gestures – for which Barbaro et al. (2018) have reported 72% positive predictive value for later autism diagnosis. We assessed infants with Gazefinder, the Autism Observation Scale for Infants (AOSI; Bryson et al., 2008) and the Mullen Scales of Early Learning (MSEL; Mullen, 1995).

Results: Gazefinder captured tracking data ranging from 0% (n=3; no data due to examiner error) to >99% (M=72, SD=28), including rates >80% for half of the sample. Among the 51 infants with some tracking data, referential attention to pointed-at objects (vs. distractors) and preferential attention – toward eyes (vs. mouth) of a human face, and to people (vs. geometry) – were as expected, suggesting ecological validity of the stimulus sequence. Variation in attention to pointed-at objects was associated with MSEL verbal skills (r=.33, p=.020). Further, greater attention to geometry (vs. people) was associated with poorer verbal skills (r= -.39, p=.008) and with more behavioural signs of autism (AOSI; r= -.42, p=.002) suggesting potential utility for Gazefinder to aide early identification.

Conclusions: Tracking rates captured by Gazefinder exceeded the standard often accepted by researchers conducting eye-tracking studies with infants at high-risk for autism using other technologies (i.e., ~30%). The large majority of infants tolerated the Gazefinder procedure very well. Researchers more familiar with behavioural assessments found the apparatus easy to use. Initial evidence suggests that some gaze data automatically exportable from Gazefinder may be associated with concurrent behavioural signs of autism and language skill among 9-16-month-old infants referred with early social-communication atypicalities. Gazefinder may facilitate the early identification and diagnosis of autism in clinical settings, warranting a larger scale trial.