Project Details

Enhancing Behavioral Understandings and Interventions in Children with Autism Spectrum Disorder using Artificial Intelligence

Yashar Kiarashi, PhD

Summary

Abstract: Children with autism spectrum disorder (ASD) face significant challenges impacting their behavior and quality of life, with up to 94% exhibiting high-risk behaviors like aggression, elopement, and self-injury. A critical gap in current research is the lack of practical predictions for high-risk behaviors in ASD, especially for individuals with profound autism who have higher support needs and poorer health outcomes. We hypothesize that sleep quality and patterns predict challenging behaviors in ASD (Hypothesis 1), to model these patterns to enhance intervention strategies (Aim 1). We also hypothesize that medication and significant life events influence ASD behavioral patterns (Hypothesis 2), aiming to analyze these correlations for better management strategies (Aim 2). Lastly, we propose integrating data on sleep quality, life events, and past behaviors to predict high-risk events (Hypothesis 3), aiming to construct a comprehensive predictive model (Aim 3). This research utilizes innovative off-body sleep sensing and behavioral event records within a privacy-preserving framework. Our current dataset includes 14 children, with plans to expand to 50 participants. The primary outcome will be an AI-driven multimodal time series predictor, culminating in an open-source algorithm for predicting challenging behaviors and high-risk events, enabling proactive intervention tools for caregivers. This model aims to significantly improve care for individuals with ASD, providing global, proactive interventions to minimize high-risk behaviors and reducing caregiver injuries, staffing shortages, and workers' compensation claims. The Thrasher grant will support expanding this research, facilitating grant applications to scale our work for behavioral quantification in neurodiverse populations. BACKGROUND: Children with autism spectrum disorder face challenges affecting their behavior and quality of life, with up to 94% exhibiting high-risk behaviors that can cause injury or even death, such as aggression, elopement, and self-injury. GAP: Individuals with profound autism are underrepresented in research despite having higher support needs, poorer health outcomes, and decreased quality of life. This underrepresentation results in a lack of practical, real-world predictions for high-risk behaviors in autism spectrum disorder. HYPOTHESIS: Sleep quality and patterns are predictive of challenging behaviors in autism, and integrating data on sleep quality, major life events, and historical challenging behaviors can effectively predict high-risk events in autism. METHODS: This research utilizes records of sleep captured through an innovative off-body sensing system, along with medical incidents and behavioral events, all within a privacy-preserving framework to assess the likelihood of high-risk behavior using explainable AI techniques. RESULTS: Pending. IMPACT: The development of this model for challenging behaviors and medical conditions may significantly improve care for individuals with autism. Through developing an open-source tool, the project aims to enable global, proactive interventions that minimize high-risk behaviors, thereby reducing caregiver injuries, and staffing shortages. Website Link: https://yashkia.github.io