Abstract
Autism is a serious threat to an individual's physical and mental health. Early screening, diagnosis, and intervention can effectively reduce the level of deficits in individuals with autism. However, traditional methods of screening, diagnosis, and intervention rely on the professionalism of psychiatrists and require a great deal of time and effort, resulting in a large proportion of individuals with autism being diagnosed after the age of 6. Artificial intelligence (AI) combined with machine learning is being used to improve the efficiency of early screening, diagnosis, and intervention of autism in young children. This review aims to summarize AI-assisted methods for early screening, diagnosis, and intervention of autism in young children (infants, toddlers, and preschoolers). To achieve early screening and diagnosis of autism in young children, AI methods have built predictive models to improve the automation of early behavioral diagnosis, analyzed brain imaging and genetic data to break the age barrier for diagnosis, and established intelligent screening systems for early mass screening. For early intervention of autism in young children, AI methods built intelligent education systems to optimize the teaching and learning environment and provide individualized interventions, constructed intelligent monitoring systems for dynamic tracking, and created intelligent support systems to provide continuous support and meet the diverse needs of young children with autism. As AI continues to develop, further research is needed to build a large and shared database on autism, to generalize and migrate the effects of AI interventions, and to improve the appearance and performance of AI-powered robots, to reduce failure rates and costs of AI technologies.