Optimization of proton therapy plans with respect to biological and physical dose distributions

Abstract

Introduction: Proton therapy is a radiation treatment method growing around the world. This is mainly due to the protons ability to deposit dose more conformal compared to conventional photon therapy. Protons also differ in terms of biological effect compared to photons for the same physical dose. To account for this increased relative biological effectiveness (RBE), a constant RBE of 1.1 is applied in clinical proton therapy treatment planning, i.e. approximately 10% lower physical dose is given to the tumor if proton therapy is used. It is however known that the RBE is not constant, and is dependent on e.g. the linear energy transfer (LET), physical dose and tissue type. By using biological optimization the tumor may get a homogenous biological dose distribution, and prevent over- and under dosage to healthy tissue and the tumor volume. Variable RBE models can be used to optimize a treatment plan with respect to RBE-weighted dose, but these are not yet available in commercial treatment planning systems. The aim of this study was to implement a method for optimization of treatment plans with respect to both biological and physical dose, and further use this to analyze the differences in physical and biological dose distributions depending on the optimization strategies applied. Methods: The FLUKA Monte Carlo code was used together with a prototype optimization software to calculate and optimize RBE weighted dose distribution for proton treatment plans. Treatment plan information from the TPS, such as beam energies and positions, was exported and translated to fit the format of the optimizer. The mathematical formulation of three RBE models were also included in the biological dose calculation and optimization; the Rørvik model, the Unkelbach model and the Wedenberg model. These models vary in which parameters they are based on, and therefore provide a good basis for comparison. The different treatment plans consisted of a water phantom with a cubic planning target volume (PTV), three plans for a water phantom with an L-shaped PTV and a small organ at risk (OAR), and a clinical patient plan. The results from the optimization were verified by running a FLUKA simulation with the original plan, optimizing the result, and then a final FLUKA simulation to verify the dose distribution. The physical dose distributions from the RBE models were compared and a single field optimization of a patient plan was also performed to analyze the dose distribution. Results: The results from the optimizer showed a homogenous RBE-weighted dose distribution for the different treatment plans and RBE-models. The PTV for all plans received the prescribed dose when the optimized result was verified. Promising results was also achieved from the patient plan where the optimizer provided a relatively even dose distribution although with some discrepancies. The calculations of physical dose distribution from the variable RBE-optimized plans showed in general lower physical dose to the PTV, compared to plans optimized with a constant RBE of 1.1. Physical dose distribution from the plan optimized with respect to the Wedenberg model was also observed to be lower than plans optimized with respect to both the Unkelbach- and Rørvik model. The latter two models showed similar physical dose distributions. Conclusions: A method for the use of a prototype optimization algorithm was integrated into an existing Monte Carlo based dose calculation framework. The implementation was applied for different RBE models, where the physical dose distributions from the RBE models were compared. Applying variable RBE-models in treatment planning may lead to lower physical dose to the target, thus preventing underand overdosage to the tumor and surrounding tissue, respectively. Differences between the variable RBE models were also seen, indicating that more research is needed before clinical application.