Efficacy of a digital intervention on visual scan pattern for faces in school-aged children with autism

Integrating scalable digital interventions into routine care may expand access to support and reduce the clinical burden for children with autism spectrum disorder (ASD). To evaluate this potential, we first used eye-tracking to identify distinct visual scanning patterns using a Hidden Markov Model approach between school-aged children with ASD and typically developing (TD) controls. This analysis revealed more random visual scan patterns and a reduced preference for social versus nonsocial stimuli in the ASD group. We then conducted a randomized clinical trial assessing a targeted, app-based emotion recognition intervention. Children with ASD were assigned to either two months of emotion recognition training(n = 25, 6.88 ± 1.31 years)or an active control (memory training, n = 25, 7.05 ± 1.54 years), with a six-month follow-up. Importantly, compared to the active control group, the emotion training group showed significantly greater improvements in clinical symptoms, cognitive performance, and eye-tracking measures of face processing, social preference, and joint attention. Mechanistically, the intervention promoted more normative visual scanning, including increased attention to eyes and a greater preference for social stimuli. These findings demonstrate that a targeted digital intervention can modify core visual-behavioral mechanisms and improve symptoms in ASD. This not only underscores the centrality of emotion processing in therapeutic design but also demonstrates the potential of scalable digital tools to augment care of autistic children. TRIAL REGISTRATION ClinicalTrial.gov Identifier: NCT06421272