Analysis of children’s movements could play a role in identifying autism spectrum disorder (ASD) early, suggests a small study conducted by researchers at the Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD). Using deep learning, artificial intelligence methods to analyze the path of arm movement (trajectories) during an object-sorting task, NICHD researchers were able to distinguish a high proportion of children with ASD from typically developing children and from neurotypical adults.
The study was conducted by Amir Gandjbackhche, Ph.D., of the NICHD Section on Translational Biophotonics, and Anjana Bhat, Ph.D., at the University of Delaware. It appears in Scientific Reports.
Background
ASD is a complex developmental disorder that affects how a person behaves, interacts with others, communicates, and learns. The authors noted that the process of diagnosing ASD in children relies largely on behavioral analysis and focuses heavily on parent-reported social communication difficulties. However, children with ASD often have motor (movement) difficulties, but these typically are not considered for screening or making a diagnosis.
Results
For the current study, researchers compared arm movement trajectories of children diagnosed with ASD to those of typically developing children and healthy adults. They enrolled 27 children with ASD and 15 typically developing children. Children ranged from 6 to 17 years old. On their right wrist, children wore an inertial measurement unit—a device that records the speed and direction of their arm movement. Each child sat at a table opposite one of four adult investigators, who also wore an inertial measurement unit. Two sets of blocks were arranged in a circle and placed in front of the child and the adult. The child and the adults were asked to place the blocks in the box, one by one. Researchers used artificial intelligence software to classify the movements of the children as they completed the session.
Compared to typically developing children, children with ASD had wider and jerkier movements than adults and typically developing children. Children with ASD also took longer to react and to reach the top speed of their arm movement. When they did reach top speed, they moved faster but less fluidly than typically developing children. They also were more likely to overshoot their target before dropping the figure in the box.
Overall, the method distinguished between typically developing children and those with ASD roughly 78 percent of the time. The method distinguished between the adults and typically developing children 98 percent of the time and between adults and children with ASD almost 99 percent of the time.
Significance
The authors concluded that their findings support further study of artificial intelligence to classify children’s movements to assist in the screening and diagnosis of children with ASD.
Reference
Su, WC, et al. Using deep learning to classify developmental differences in reaching and placing movements in children with and without autism spectrum disorder. Scientific Reports (2024).
BACK TO TOP