Related Experiment Video
Updated: Jan 13, 2026
14:08
Automated Midline Shift and Intracranial Pressure Estimation based on Brain CT Images
Published on: April 13, 2013
43.4K
Radiomics-based machine learning model for predicting secondary decompressive craniectomy in TBI patients after
Tiange Chen1,2, Ganzhi Liu1,2, Ziyuan Liu1,2
1Department of Neurosurgery, Xiangya Hospital, Central South University, No. 87 Xiangya Rd, Changsha, Hunan, 410008, China.
Chinese Neurosurgical Journal
|January 8, 2026
Summary
Machine learning models using radiomics can predict the need for secondary decompressive craniectomy in traumatic brain injury patients. Integrating radiomic features with clinical data significantly improved prediction accuracy, aiding early intervention.
Area of Science:
- Neurosurgery
- Medical Imaging
- Artificial Intelligence
Background:
- Secondary decompressive craniectomy (DC) is a critical intervention for managing elevated intracranial pressure after traumatic brain injury (TBI).
- Early identification of patients requiring DC is crucial for timely intervention and improved outcomes.
- Current methods for predicting secondary DC may not be sufficiently accurate.
Purpose of the Study:
- To develop and validate machine learning-based predictive models using radiomics to assess the likelihood of secondary DC in TBI patients.
- To compare the performance of models based on demographic/clinical data, radiomic features, and a combination of both.
Main Methods:
- Radiomic features were extracted from pre-evacuation CT scans of 65 TBI patients.
- Patients were divided into training (70%) and testing (30%) cohorts.
- Various machine learning algorithms were employed to build predictive models, including randomForest and cforest.
Main Results:
- Models using only demographic and clinical data showed poor predictive performance (AUC < 0.5).
- A radiomics-only model (randomForest) achieved an AUC of 0.83 in the test cohort.
- A multiomic model combining radiomic, demographic, and clinical features (cforest) demonstrated superior performance with an AUC of 0.86 in the test cohort.
Conclusions:
- Radiomics-based models can effectively predict the need for secondary DC in TBI patients.
- Integrating radiomic features with clinical data further enhances predictive accuracy.
- These models hold potential for identifying high-risk patients, enabling early intervention and preventing neurological deterioration.
