This photo taken with a mobile phone shows a painting created by an autistic child in Shenyang, northeast China's Liaoning Province, Feb. 27, 2025. [Photo/Xinhua]

GUANGZHOU -- In the classroom at an autism rehabilitation center in the southern Chinese city of Shenzhen, a three-year-old boy tried to utter a faint "i" sound while gazing into a mirror.

"Now you try. Make the sound. Too loud now. Softer," said the therapist, Yu Canyang, softly. "There you go! You're learning so fast -- great job!" It marked a hard-won breakthrough in the speech development of Haohao (pseudonym).

Holding a tongue depressor in one hand, Yu promptly recorded it on a tablet. At the end of the appointment, an AI system swiftly generated the boy's progress report and a personalized home training plan was sent to the boy's parents.

This AI-enhanced moment has offered a glimpse into the profound transformation underway in China's autism rehabilitation sector.

UNMET DEMANDS

The classrooms at Dami & Xiaomi Autism Rehabilitation Center replicate a kindergarten setting with soft, bright colors -- all carefully designed to help children with autism develop the skills needed for eventual integration into mainstream education through years of specialized training.

"For some three-year-olds still unable to speak," noted Yu. "The first word may demand years of patient therapy." Data from the business information platform TianYancha has shown only 654 entries related to "autism rehabilitation centers." Dami & Xiaomi is one of them.

Unlike most medical disorders, autism cannot be directly diagnosed through biochemical tests, imaging scans, or even genetic analyses. It requires diagnosis through direct behavioral observation, making screening and intervention highly individualized.

A report has shown that China has only approximately 100,000 rehabilitation therapists, with about 20,000 registered, and each therapist can only handle four children per week. By contrast, the country's autism community exceeds 10 million individuals, including over 3 million children aged 14 and under.

Many kids with social interaction deficits are missing out on crucial early intervention and lacking the one-on-one and long-term care they need.

"Analyzing intervention outcomes across large populations -- whether effective or not -- can optimize treatment recommendations for autistic children with similar behavioral profiles," said Zou Xiaobing, chief physician from the child behavioral development center under the Third Affiliated Hospital of Sun Yat-sen University.

"Integrating large AI models into autism rehabilitation is an attempt that could expand our understanding, enhance intervention quality, and deliver more personalized support at every stage," Zou added.

AI FOR AUTISTIC KIDS

In November 2024, Dami & Xiaomi, partnered with Amazon Web Services (AWS), launched RICE AI, an AI-powered solution for kids with autism spectrum disorder, language impairments, and learning disabilities.

This AI application integrates foundation models, professional expertise, and rehabilitation big data to provide comprehensive support in assessment, therapeutic intervention, and home-based training.

Since its establishment in 2014, this rehabilitation center has prioritized data collection. Now, over 2,000 supervisors and therapists systematically record children's performance, tracking their task completion and assistance levels in each session to generate behavioral tag data.

The platform has housed over 80 million tagged behavioral data points to fuel RICE AI, significantly simplifying diagnostic processes.

Leveraging big data analytics, the system generates next-phase intervention strategies while slashing assessment report time from 3 hours to under 30 minutes.

"This efficiency relieves supervisors and therapists from repetitive tasks, allowing them to dedicate more energy to child engagement and strategic planning," said Wang Zhanwei, vice president of Dami & Xiaomi.

The platform can further automate the creation of customized picture books for home-based autism rehabilitation. Unlike normal children, autistic learners tend to demand rigid visual consistency, such as identical character counts across scenes. "If a family of four leaves home, they must return intact," explained Wang.

However, AI hallucinations may generate nonsensical visuals, which could confuse autistic children. "Only by embedding this domain-specific knowledge into foundation models can we have viable materials," noted Wang.

RICE AI slashes picture book production time from two months to just 10 minutes while enabling personalized customization. This breakthrough bridges institutional therapy and home-based intervention and reduces the family burden of autism care.

PARADIGM SHIFT

China's autism rehabilitation sector is quickly embracing AI transformation. The Beijing-based facilities ALSOLIFE and ING Care are exploring digital therapeutics for autism.

Also, specialized AI models such as "Starlight," developed by a Chinese Academy of Sciences team, and therapeutic picture books generated by Alibaba's AI tool Tongyi, are reshaping the intervention paradigms.

This year, Dami & Xiaomi plans to integrate its existing AI applications into a unified platform for industry-wide sharing among rehabilitation professionals. "We are set to open-source our data and capabilities to empower more rehabilitation centers and better support affected families," said Wang.

Families of children with autism have already reaped the rewards of AI-aided rehabilitation approaches.

"When my son suddenly called out 'Mama,' I burst into tears, even though I know our road to recovery is still a long haul," said Haohao's mother, her voice trembling with emotion.