MORGANTOWN: Researchers at West Virginia University are turning to artificial intelligence to devise a way to properly diagnosis autism spectrum disorder.
Characterising an autism patient’s behaviour can be challenging, so researchers are using behaviour-tracking technology and phenotyping to better understand and identify autism spectrum disorder.
Phenotyping is the characterisation of a behaviour or trait, and in this study, researchers will be looking at autism patients’ behaviours and traits.
Supported by a US$500,000 (RM2.09mil) award from the National Science Foundation, Xin Li, professor in the Lane Department of Computer Science and Electrical Engineering, and Shuo Wang, an adjunct assistant professor, will conduct the research using imaging and data science.
“This project is important because it aims at filling an important gap in our existing knowledge about ASD," Li said. “Improved understanding of autism phenotyping is expected to help with not only more accurate diagnosis, but also more personalised intervention for ASD patients."
One of the biggest challenges facing autism research is that there is not one form of autism, but many subtypes. Each person with autism can have unique strengths and challenges, which also make it difficult to identify the specific traits associated with this disorder, genetically or behaviorally, according to Li.
Li said that there is currently no consensus about the standard of behaviour characterisation for humans yet, but animal models have used the three phenotypes of abnormal social interactions, communication deficits and repetitive behaviour to consider the standards of behaviour.
“We expect to identify similar phenotypes for ASD patients as the first step,” Li said.
This project will assess ASD using behaviour-imaging data, such as eye-tracking and audio and video with neuroimaging data.
According to Li, neuroimaging data are a direct measurement of brain activities and behaviour imaging data are the consequence of brain activities.
“Integrating these two multimodal data represents a natural strategy for understanding the relationship between brain activities and behavioral patterns,” Li said.
Artificial Intelligence will identify the traits associated with ASD using both neuroimaging and behavioral imaging, Li said.
According to Li, ASD is a neurodevelopmental disorder that affects one out of 54 children in the US, and their study could also help with early detection in young children.
The earlier children with ASD get intensive intervention, the better their developmental outcomes, he said.
“Currently, the average age of a child when she or he receives an ASD diagnosis in the US is four years old,” Li said. “However, about half the parents of children with ASD report that they suspected a problem before their child was one year of age. This has been known as the ‘detection gap’. Many research teams including us are working on reducing this gap.” – Times West Virginian, Fairmont/Tribune News Service