A 3-D Learning Environment for Infants and Toddlers at-Risk for ASD: Can Technology Improve Early Social Communication Vulnerabilities

Background: Early detection and intervention is critical to optimal treatment for children with ASD. [1, 2]. Technology-assisted intervention approaches are being considered due to their potential flexibility, controllability, duplicability, and overall lower costs when compared to traditional human dominant intervention [3-5]. However, very few systems have been designed for infants and toddlers. In the current work we present a proof-of-concept system designed to enhance early social orienting, specifically enhancing early response to name within a three dimensional technological learning environment. 

Objectives: We designed a closed-loop, computer-assisted intervention system for potential future use with infants and toddlers at risk for ASD.  The system incorporated a series of target monitors, a non-contact eye-tracking system, and a series of audio and visual cues within a reinforcement hierarchy designed to improve children’s social orienting within this technological environment.  The system was designed to operate in closed-loop fashion (i.e., automatically responded to estimated child gaze to activate components of the hierarchy) in real time.  The ultimate goal of the current system was to create a technological learning environment that systematically enhances a child’s ability to respond to caregiver’s attempts to garner attention by calling his/her name from a variety of locations.  

Methods: The name prompting system encompassed a series of coordinated target monitors within a room.  These monitors would initially display a name press (i.e., recording of individual calling child’s name) and activate an internal prompt hierarchy based on estimated child gaze in real time.  If the child did not look at the target monitor when name was called, then coordinated visual/audio attention attractors appeared based estimated gaze in order to cue for orientation to correct target.  When the participant looked at the target monitor, then social and non-social reward videos and animations were displayed to reinforce this orientation.  The attention tracking system computed orientation in real-time via vision driven head pose estimation algorithms.  This data was fed back into system to further advance the experimental protocol.  In order to test system feasibility 7 typically developing infants and toddlers (all under 2) participated in a coordinated series of 10 trials within the system.  Trials were interspersed with introductory, transition, and conclusion activities to familiarize children with the system. 

Results: All the participants tolerated the system well with all infants and toddlers completing the entire experiment.  The participants’ performance was well distributed across the prompting levels with in excess of 98.57% of all trials ultimately resulting in correct orientation.  Preliminary data also suggest that the attention attractor was capable of shaping participants’ attention to the target position. 

Conclusions: This work provides preliminary support for the feasibility for employing technologies with very young children potentially at risk for ASD.  The development of robust technological systems capable of addressing meaningful core early social communication vulnerabilities prior to the full manifestation of ASD impairments may represent new paradigms for intervention science.