Development and evaluation of robotic detection technology for assessing autism

Abstract

Objective: An objective and standardized assessment for assessing autism is needed. This study aimed to develop and validate robotic detection technology for assessing autism. The robot HUMANE, installed with computer vision and linked with face and motion recognition technology, autonomously detected atypical eye gaze and repetitive motor movements, two of the features of autism, while narrating stories. It autonomously prompted the child if they did not establish eye gaze with the robot or produced motor movements for five seconds continuously. Method: The study involved 119 children aged between three and six years old (M=4.53, SD=1.89; 38 females) and included children confirmed or not confirmed with autism. They all received the Autism Diagnostic Observation Schedule—second edition (ADOS‑2), the standard diagnostic tool for autism. HUMANE’s detection performance – the number of robot prompts and the cumulative duration of inattentiveness/improper posture – was then evaluated against the calibrated severity score of ADOS-2. Results: Our results showed that the average sensitivity and specificity of the detection reached 0.80, the Diagnostic Odds Ratio was beyond 30, and the AUC was.85. Discussion: These results indicate that the robotic detection technology of atypical eye gaze and repetitive motor movements can contribute to the diagnostic process to identify the presence or absence of autism.