Related Experiment Video
Updated: Jul 26, 2025
07:21
Systems Analysis of the Neuroinflammatory and Hemodynamic Response to Traumatic Brain Injury
Published on: May 27, 2022
3.2K
Predicting traumatic brain injury outcomes using a posterior dominant rhythm
Nathaniel A Cleri1, Jordan R Saadon1, Xuwen Zheng1
11Department of Neurosurgery, Renaissance School of Medicine at Stony Brook University, Stony Brook, New York.
Journal of Neurosurgery
|June 17, 2023
Summary
A posterior dominant rhythm (PDR) on electroencephalography (EEG) in severe traumatic brain injury (sTBI) patients predicts better outcomes. A novel machine learning model accurately forecasts recovery, aiding clinical decisions.
Area of Science:
- Neuroscience
- Clinical Neurology
- Medical Technology
Background:
- Predicting outcomes for severe traumatic brain injury (sTBI) is complex, with current models lacking individual patient applicability.
- Identifying reliable metrics for predicting recovery post-sTBI is crucial for patient management.
Purpose of the Study:
- To determine if a posterior dominant rhythm (PDR) on electroencephalography (EEG) is associated with positive outcomes in sTBI patients.
- To develop a machine learning (ML)-based model for accurate prediction of consciousness recovery after sTBI.
Main Methods:
- A retrospective analysis of 195 intubated sTBI patients (Glasgow Coma Scale ≤ 8) with EEG recordings within 30 days of injury.
- Patients were categorized into PDR-positive (n=51) and PDR-negative (n=144) cohorts to compare outcomes including survival, command following, and Glasgow Outcome Scale-Extended (GOS-E) scores.
- An ML model (AutoScore) was developed to predict in-hospital survival and command following recovery using clinical, radiographic, and EEG variables.
Main Results:
- The PDR-positive cohort showed significantly higher rates of in-hospital survival (84.3% vs 63.9%), recovery of command following (76.5% vs 53.5%), and discharge GOS-E scores (3.00 vs 2.39).
- The ML model, incorporating PDR and clinical variables (age, BMI, blood pressure, pupil reactivity, glucose, hemoglobin), demonstrated strong predictive accuracy for in-hospital survival (AUC 0.815) and command following recovery (AUC 0.700).
- The ML model outperformed existing models like MRC-CRASH and IMPACT in predicting patient outcomes.
Conclusions:
- The presence of a PDR on EEG is a significant predictor of favorable outcomes in sTBI patients.
- The developed ML-based prognostic model offers high accuracy and clinical utility for predicting sTBI recovery and guiding patient management.
- This model can assist clinicians in decision-making and family counseling following severe traumatic brain injuries.
