Skip to content

Eliott Ramoisiaux


Model Independent and Multi-Objective Tuning of Proton Therapy Beamlines

Abstract

The optimization of cyclotron-based proton therapy beamlines challenges the classical approach used in beam optics due to the very strict constraints imposed on beam quality in view of patient treatment, despite the large losses induced by the emittance increase coming from the energy degrader. Consequently, the improvements in proton therapy beam delivery methods are tightly coupled with advances in beam transport and control techniques and require further developments. The Proteus One compact, single-room, proton therapy solution developed by Ion Beam Applications (IBA), faces challenges regarding the beam calibration: the complexity of the physical system prevents the determination of an exact deterministic input to output model, while the tight constraints that the beam characteristics must meet during a treatment lead to a large amount of time required for the machine calibration. This work focuses on the development of an algorithm for the optimization of the beam transport parameters allowing reaching the clinical beam specifications. An optimization method based on a model independent and multi-objective feedback technique, the so-called “Rotation Rate Tuning”, is investigated and tested numerically. This multi-objective algorithm enables the simultaneous optimization of complex constraints specific to proton therapy beamlines, including those for Pencil Beam Scanning (PBS) maps. The algorithm is implemented and applied to progressively more complex scenarios using numerical beam tracking simulations. The use of the algorithm following and automating the Proteus One calibration procedure is first investigated. Then, the algorithm is used to optimize the machine parameters for different energy degrader materials. Finally, the robustness of the method is assessed in the presence of noise. The results show the ability of the proposed algorithm to evolve smoothly in the phase space and to propose correct parameter sets in a reasonable amount of time. Furthermore, its robustness to noise and its adaptation capability to a large number of situations is established. The results are discussed in detail and applications for automated machine calibration are proposed.

Report

Back to top