Publikationsrepositorium - Helmholtz-Zentrum Dresden-Rossendorf

Introduction
Routinely, a constant relative biological effectiveness (RBE) is used in proton therapy. However, experimental evidence indicates that RBE can vary with dose and linear energy transfer (LET). Here, we introduce a Monte-Carlo simulation method to assess RBE variations from follow-up magnetic resonance (MR) scans and illustrate its applicability for an exemplary brain-tumor patient.
Methods
A proton therapy Monte-Carlo model was setup to simulate dose and dose-averaged LET distributions of treatment fields on computed tomography (CT) scans and also in a water phantom for comparison with quality assurance measurements. A follow-up T1-weighted contrast-enhanced (T1w-CE) MR scan of an astrocytoma (grade 2) patient, acquired 18 months after radiochemotherapy (DRBE=1.1=60Gy), was used to delineate post-treatment image-change regions and rigidly co-registered to the planning CT. A multivariable logistic normal tissue complication model (NTCP) was built by voxel-wise correlating image changes with simulated dose and LET distributions. Tolerance doses TD10 (resulting in 10% probability of image change in a voxel) as function of LET were obtained from the NTCP model.
Results
Measured and Monte-Carlo-predicted dose agreed within 2%. Changes on follow-up T1w-CE-MR images correlated with increasing simulated dose and LET (Fig.1). The NTCP model revealed a significant LET effect (p<0.0001) on the image change probability resulting in a decrease of modeled TD10 values from 70.3Gy to 50.6Gy for 1keV/μm and 6keV/μm, respectively, indicating a varying RBE (Fig.2).
Conclusion
The correlation of post-treatment MR image changes with Monte-Carlo simulations allows for testing the hypothesis of a variable proton RBE. Currently, this method is systematically evaluated and may, after validation in independent patient cohorts, reduce biological uncertainty in proton therapy.