Pipeline for Assisting Diagnosis of Children with Autism Spectrum Disorder via Automated Method for Classifying Repetitive Behaviors

Abstract

We present automated methods to collect and provide motion data via sensors that can be used to assist the clinical decision-making process when diagnosing autism spectrum disorder (ASD) in children. The current limitation of assessment tools for diagnosing ASD in children is the subjectivity of the observer. We present a pipeline for automating the process of data collection when children interact with toys, specifically focusing on the repetitive behavior commonly portrayed among children with ASD. We used data that mimicked repetitive behaviors identified in different studies. Existing research and assessment of the diagnosis of ASD in children indicates the need for having an automated and quantifiable approach that can provide more than mere observation. It is important to note, however, that this can only aid existing methods of diagnosing autism, and its clinical relevance is to be further evaluated. It is critical for ASD to be detected at the early stages of the life of children. It has been supported that intervention for children with ASD is more effective in the younger age group and is beneficial for the long-term prognosis of children. Therefore, automated technology will potentially help increase the efficiency and objectivity of observation-based diagnostic procedures. We discuss how using simple technologies such as an Arduino board, could bring efficiency and objective analysis to the observation procedure in the diagnosis of ASD.