Researchers have developed an algorithm that uses machine learning techniques to automatically track an infant’s wiggles and squirms, translating the motion of a child to a simplified stick person, researchers reported March 26 in Acta Paediatrica.
This sort of information can be helpful in detecting patterns of motion that occur with certain disorders, including cerebral palsy, and could allow doctors to come to a diagnosis and intervene earlier in a child’s development.
To test if their method captured enough information about the children’s patterns of motion, the scientists used their algorithm to capture the movements of 21 babies between eight and 17 weeks old, including seven who were diagnosed with cerebral palsy later. They provided the stick-figure videos alone to doctors who were then able to accurately diagnose the children 95 percent of the time based on abnormal movements. In a proof-of-concept test, the algorithm automatically analyzed the motion capture videos to reveal clear differences between groups of children with typical or atypical motion.
The goal is to broaden access to screenings outside of hospital settings as machine learning techniques can provide doctors with useful information, even from home videos. “The ultimate aim is to find methods that will make it possible to provide high and even quality infant healthcare everywhere in the world,” says paper coauthor Sampsa Vanhatalo of the University of Helsinki in a statement.
V. Marchi et al., “Automated pose estimation captures key aspects of General Movements at 8-17 weeks from conventional videos,” Acta Paediatrica, doi:10.1111/apa.14781, 2019.