An artificial intelligence algorithm capable of signaling early signs of neurodevelopment disorders in infants has been created by researchers in Finland and Italy. By analyzing conventional videos of infants, the algorithm can create “skeleton” videos, which depict a child’s movement in the form of a stick figure. The research could help in early detection of neurodevelopment disorders, such as cerebral palsy.
“Medical doctors have shown that observing special features in the spontaneous movements of infants may be the most accurate way to predict later development of cerebral palsy,” Sampsa Vanhatalo, a neurophysiologist at the University of Helsinki who led the study, told Digital Trends. “However, such visual analysis of infant movements by experts is always subjective, and it requires substantial training. Here, we showed for the first time that it is possible to extract infant movements from conventional video recordings. That is, we make skeleton videos, at a very high accuracy.”
The algorithm works by scanning a conventional video of an infant to detect certain poses and movement patterns. The algorithm uses a “pose estimation method” to generate a stick-man depiction of the infant. These movement patterns can then be analyzed to detect normal or unusual movements.
Children are typically diagnosed with cerebral palsy between the ages of six months and two years. However, early detection would allow doctors to begin to provide therapeutic interventions to alleviate the impact of the condition. A system that can help doctors detect early signs of the condition could offer children a jump start on treatment.
“A pose estimation method of this kind is like a Rosetta stone, which opens the world to myriad of A.I. solutions for advanced assessments, diagnostics, and monitoring of spontaneous infant behavior,” Vanhatalo said. “The first application would be to develop a diagnostic classifier of infant movements to be used in screening of at-risk infants that are not able to reach specialist attention. Indeed, most infants in this world live in areas or in conditions beyond the immediate reach of pertinent medical expertise.”
Vanhatalo partnered with researchers from the University of Pisa and Neuro Event Labs, a company that specializes in A.I.-based video analysis for medical purposes.
A paper detailing the research was published this month in the journal Acta Pediatrica.