Related Experiment Video
Updated: May 6, 2026
05:30
Controlled Cortical Impact Model for Traumatic Brain Injury
Published on: August 5, 2014
28.5K
AI-based models to predict decompensation on traumatic brain injury patients
Ricardo Ribeiro1, Inês Neves2, Hélder P Oliveira1
1INESC TEC - Institute for Systems and Computer Engineering, Technology and Science, Porto, Portugal; FCUP - Faculty of Sciences of the University of Porto, Porto, Portugal.
Computers in Biology and Medicine
|January 9, 2025
Summary
Artificial Intelligence (AI) models can predict patient decompensation in Traumatic Brain Injury (TBI) cases using Electronic Health Records (EHR). This approach aids clinical decisions and improves care for TBI patients in the ICU.
Area of Science:
- Medical Informatics
- Neurology
- Artificial Intelligence
Background:
- Traumatic Brain Injury (TBI) poses significant mortality risks, reaching 30-40% in severe cases.
- Effective clinical decision-making for TBI complications remains a challenge.
Purpose of the Study:
- To develop and evaluate AI models for predicting TBI patient decompensation.
- To enhance patient care and clinical decision-making using data-driven approaches.
Main Methods:
- Utilized sequential data from Electronic Health Records (EHR) of 2261 TBI patients from the MIMIC-III dataset.
- Employed Logistic Regression (LR), Long-short term memory (LSTM), and Transformers architectures for prediction.
- Explored feature sets, missing data imputation, and data imbalance techniques.
Main Results:
- LSTM models achieved high performance, with AUROC scores of 0.918 (original features) and 0.929 (added features with class weights).
- The study demonstrated the viability of using EHR time series data for predictive modeling.
Conclusions:
- LSTM models using EHR time series data are effective for predicting TBI patient decompensation.
- This predictive capability can serve as an early indicator for necessary clinical interventions.
