Machine Learning Models for the Prediction of Blunt Cerebrovascular Injury in Children


Previous studies have demonstrated the inadequacies of the Denver and Memphis model when predicting blunt cerebrovascular injury (BCVI) in pediatric patients. A recent study featuring PISC Senior Scholar Michael Nance, MD, introduces a pediatric prediction model for BCVI. The model is informed by data from the National Trauma Databank from 2007 to 2015. Test and training datasets were used to build a random forest model predicting BCVI. The Denver and Memphis models variables were re-validated to test the success of the pediatric prediction model. Results indicate that the new pediatric model is significantly more accurate than the Denver and Memphis model. The new pediatric model accounts for 94.4% of BCVI patients within the National Trauma Databank, while the Denver and Memphis model only correctly identifies 13.4% of the total BCVI patients in the National Trauma Databank. The team concludes that “the prediction model developed in this study is able to better identify pediatric patients who should be screened with further imaging to identify BCVI.”


Go to Article

Back to News