Related Experiment Video
Updated: Dec 11, 2025
Early Pathological and Magnetic Resonance Detection of Cerebral Injury Using a Rat Model of Neonatal Hypoxic Ischemic Encephalopathy
Published on: October 28, 2022
Neural Networks to Predict Radiographic Brain Injury in Pediatric Patients Treated with Extracorporeal Membrane
Neel Shah1, Abdelaziz Farhat2, Jefferson Tweed3
1Department of Pediatrics, Division of Pediatric Critical Care, Washington University School of Medicine, St. Louis, MO 63110, USA.
Insights
Neural networks can predict brain injury in pediatric patients on Extracorporporeal Membrane Oxygenation (ECMO). Physiological data was key, guiding future research for better patient outcomes.
Area of Science:
- Pediatric critical care medicine
- Neurocritical care
- Biomedical engineering
Background:
- Brain injury is a major complication for pediatric patients requiring Extracorporeal Membrane Oxygenation (ECMO).
- Early prediction of brain injury is crucial for timely intervention and improved outcomes in this vulnerable population.
Purpose of the Study:
- To develop and validate a neural network model for predicting radiographic brain injury in pediatric ECMO patients.
- To identify key variables contributing to brain injury prediction for future research.
Main Methods:
- Retrospective analysis of data from 174 pediatric ECMO patients.
- Collection of 35 variables including physiological, perfusion, acid-base, vasoactive, coagulation, and ECMO parameters.
- Development of a neural network model to predict moderate to severe brain injury (confirmed by CT/MRI), compared against logistic regression and clinician judgment.
Main Results:
- The neural network achieved an Area Under the Curve (AUC) of 0.76, with 73% sensitivity and 80% specificity.
- The neural network significantly outperformed logistic regression (62% sensitivity, 61% specificity) and clinician judgment (39% sensitivity, 69% specificity).
- Feature analysis indicated physiological data contributed most significantly to the prediction model, while coagulation factors had a minor role.
Conclusions:
- Neural networks offer a promising tool for predicting brain injury in pediatric ECMO patients.
- Physiological parameters are critical predictors, warranting further investigation.
- This predictive model can guide future research to mitigate brain injury risks during ECMO support.
Abstract:
Brain injury is a significant source of morbidity and mortality for pediatric patients treated with Extracorporeal Membrane Oxygenation (ECMO). Our objective was to utilize neural networks to predict radiographic evidence of brain injury in pediatric ECMO-supported patients and identify specific variables that can be explored for future research. Data from 174 ECMO-supported patients were collected up to 24 h prior to, and for the duration of, the ECMO course. Thirty-five variables were collected, including physiological data, markers of end-organ perfusion, acid-base homeostasis, vasoactive infusions, markers of coagulation, and ECMO-machine factors. The primary outcome was the presence of radiologic evidence of moderate to severe brain injury as established by brain CT or MRI. This information was analyzed by a neural network, and results were compared to a logistic regression model as well as clinician judgement. The neural network model was able to predict brain injury with an Area Under the Curve (AUC) of 0.76, 73% sensitivity, and 80% specificity. Logistic regression had 62% sensitivity and 61% specificity. Clinician judgment had 39% sensitivity and 69% specificity. Sequential feature group masking demonstrated a relatively greater contribution of physiological data and minor contribution of coagulation factors to the model's performance. These findings lay the foundation for further areas of research directions.
