Autism is a common neurodevelopmental disorder in children. The current diagnosis and treatment largely depend on clinical experts’ qualitative behavioral observations of children, which lack objective, quantitative behavioral evaluation metrics. This approach is highly subjective and lack standardized and personalized intervention protocols. With the gradual maturity of AI technology, machine-assisted diagnosis and treatment centered on children offer a promising solution to the challenges of autism diagnosis and treatment. This chapter reviews research on AI-assisted autism diagnosis and treatment system, along with multimodal behavior analysis. It introduces the early auxiliary screening of autism based on a non-contact visual system and early intervention cases utilizing immersive virtual systems. The goal is to leverage multi-sensor data to capture diverse scenarios and objective, quantitative pathological information about autistic children. These insights aim to provide quantitative indicators for personalized interventions, establish standardized protocols, and design personalized intervention programs. By advancing child-centered AI-assisted autism diagnosis and treatment, this approach aspires to significantly improve the welfare of autistic patients and their families. It aligns closely with the principles of a human-centered AI approach.