Related Experiment Video
Updated: Apr 2, 2026
10:25
Brain Infarct Segmentation and Registration on MRI or CT for Lesion-symptom Mapping
Published on: September 25, 2019
49.7K
Voxel-based model predicts proton-induced brain necrosis using LET and ventricular distance
1Department of Radiation Oncology, Ruijin Hospital, Shanghai Jiao Tong University, School of Medicine, Shanghai, China; Shanghai Key Laboratory of Proton-therapy, Shanghai, China.
Summary
This study developed a predictive model for radiation-induced brain necrosis after proton therapy. Variable relative biological effectiveness (vRBE) improved prediction accuracy, highlighting its importance for risk mitigation.
Area of Science:
- Radiation Oncology
- Medical Physics
- Neuro-oncology
Background:
- Radiation-induced brain necrosis (BN) is a serious complication after proton therapy for skull base tumors.
- Accurate prediction of BN is crucial for treatment planning and patient safety.
- Current models may not fully capture the biological effects of proton therapy, especially concerning variable relative biological effectiveness (vRBE).
Purpose of the Study:
- To develop and validate a predictive model for radiation-induced brain necrosis (BN) following intensity modulated proton therapy (IMPT).
- To incorporate variable relative biological effectiveness (vRBE), dose, linear energy transfer (LET), and proximity to ventricles into the predictive model.
- To compare the predictive accuracy of vRBE-based dose calculations versus constant RBE (cRBE) for BN.
Main Methods:
- A voxel-level support vector machine (SVM) model was developed using dosimetric parameters (dose, LET, distance to ventricles) from four patients with radiation-induced BN.
- Necrotic regions were delineated on follow-up MRI scans.
- The model's performance was evaluated using Area Under the Curve (AUC), and vRBE values in necrotic regions were compared to non-necrotic regions.
Main Results:
- The SVM model achieved high accuracy in predicting BN (AUC = 0.965).
- Necrotic voxels were found to cluster near ventricles (distance < 10 mm).
- Significantly higher vRBE values were observed in necrotic regions compared to non-necrotic regions (mean 1.2 vs. 1.15, p < 0.01).
- vRBE-based dose calculations identified high-dose volumes (≥0.1 cc) exceeding 90 GyRBE in necrotic areas, which were not apparent with cRBE.
Conclusions:
- Variable relative biological effectiveness (vRBE) is a critical factor in predicting radiation-induced brain necrosis (BN) after proton therapy, especially in periventricular zones.
- Constant RBE (cRBE) may underestimate the risk of BN.
- Volume-based dose constraints (e.g., D0.1cc < 90 GyRBE) are recommended for risk mitigation in BN prediction.
- Further validation in larger patient cohorts is warranted.
