Related Experiment Video
Updated: Oct 10, 2026
Automated Midline Shift and Intracranial Pressure Estimation based on Brain CT Images
Published on: April 13, 2013
Baseline CT radiomics for early prediction of post-traumatic hydrocephalus in traumatic brain injury: development and
Chunwang Wu1, Kaijie Yu1, Xujian Shui1
1Department of Neurosurgery, Affiliated Jinhua Hospital, Zhejiang University School of Medicine, Jinhua, Zhejiang, China.
Background:
Post-traumatic hydrocephalus (PTH) impairs recovery after traumatic brain injury (TBI), yet reliable early risk stratification based solely on baseline assessment remains limited. We aimed to develop and temporally validate a preoperative CT-based radiomics model for early prediction of PTH and to assess the incremental value of clinical predictors.
Methods:
In this retrospective single-center cohort, 341 consecutive adults undergoing neurosurgical treatment for TBI were included. Patients treated in 2021-2023 (n = 289) formed the training cohort, and those treated in 2024 (n = 52) served as a temporally independent validation cohort. A total of 11,998 radiomics features were extracted from 14 automatically segmented intracranial regions on the last preoperative CT. Feature selection was performed using mutual information filtering and LASSO regression. Machine learning classifiers were benchmarked, and a clinical-radiomics fusion model was constructed using stacking. Model performance was evaluated using discrimination, calibration, and decision curve analysis.
Results:
PTH occurred in 27.3% of the training cohort and 23.1% of the validation cohort. In temporal validation, the clinical, radiomics, and fusion models achieved AUCs of 0.769 (95% CI, 0.608-0.906), 0.783 (95% CI, 0.604-0.929), and 0.865 (95% CI, 0.735-0.958), respectively. The fusion model also demonstrated acceptable calibration and greater net benefit across clinically relevant thresholds.
Conclusion:
Baseline preoperative CT radiomics enables early risk stratification of PTH after TBI. Integration with clinical predictors provides incremental predictive value under temporal validation and may support individualized surveillance.
