A new machine learning tool could predict cases of cerebral palsy in young infants with brain damage, according to a July 11 international study published in JAMA Network Open.
The study looked at more than 500 infants with a high risk of perinatal brain damage at hospitals across the U.S., Belgium, India and Norway using video recordings of their movements. Of these videos, 75 percent went through a deep learning algorithm analysis for predicting cerebral palsy and 25 percent of them went through an external validation process using medical professionals.
The results revealed that the machine learning algorithm had a sensitivity of 71.4 percent, meaning it predicted true cases at that rate. It also had a specificity of 94.1 percent, indicating that the model correctly predicts almost all negative cases. The authors concluded that "this study's findings suggest that deep learning–based assessments could support early detection of CP in infants at high risk."