Related Experiment Video
Updated: Jun 20, 2026
14:08
Automated Midline Shift and Intracranial Pressure Estimation based on Brain CT Images
Published on: April 13, 2013
Early prediction of progressive cerebral contusion using a deep transfer learning-enhanced multimodal nomogram
Wanqin Yang1, Shanshan Qian1, Ping Zhao2
1Department of Emergency, The Third People's Hospital of Hefei, Hefei Third Clinical College of Anhui Medical University, Hefei, China.
Biomedizinische Technik. Biomedical Engineering
|June 19, 2026
Summary
A new multimodal fusion model accurately predicts early progressive cerebral contusion (PCC) using clinical data and deep transfer learning. This approach offers improved clinical decision-making for patients with brain injuries.
Area of Science:
- Neuroimaging
- Artificial Intelligence in Medicine
- Clinical Decision Support
Background:
- Progressive cerebral contusion (PCC) poses a significant challenge in neurotrauma management.
- Accurate early prediction of PCC is crucial for timely and effective treatment decisions.
Purpose of the Study:
- To develop and validate a multimodal fusion model for early prediction of progressive cerebral contusion (PCC).
- To integrate deep transfer learning (DTL) features with clinical variables for enhanced predictive accuracy.
Main Methods:
- A retrospective cohort of 196 cerebral contusion patients was analyzed.
- A multimodal model combined ResNet-50 DTL features with clinical data.
- Model performance was evaluated using ROC curves, calibration plots, and decision curve analysis.
Main Results:
- The integrated nomogram achieved high predictive performance (AUC 0.999 training, 0.972 validation).
- The multimodal model outperformed standalone DTL and clinical models.
- Grad-CAM visualization confirmed accurate localization of brain lesions.
Conclusions:
- The developed multimodal fusion model demonstrates excellent predictive performance for early PCC.
- This model offers significant clinical value and supports precise treatment decisions in neurotrauma care.