Trimmer sequencing time minimization during dynamically collimated proton therapy using a colony of cooperating agents

Abstract

The dynamic collimation system (DCS) can be combined with pencil beam scanning proton therapy to deliver highly conformal treatment plans with unique collimation at each energy layer. This energy layer-specific collimation is accomplished through the synchronized motion of four trimmer blades that intercept the proton beam near the target boundary in the beam’s eye view. However, the corresponding treatment deliveries come at the cost of additional treatment time since the translational speed of the trimmer is slower than the scanning speed of the proton pencil beam. In an attempt to minimize the additional trimmer sequencing time of each field while still maintaining a high degree of conformity, a novel process utilizing ant colony optimization (ACO) methods was created to determine the most efficient route of trimmer sequencing and beamlet scanning patterns for a collective set of collimated proton beamlets. The ACO process was integrated within an in-house treatment planning system optimizer to determine the beam scanning and DCS trimmer sequencing patterns and compared against an analytical approximation of the trimmer sequencing time should a contour-like scanning approach be assumed instead. Due to the stochastic nature of ACO, parameters where determined so that they could ensure good convergence and an efficient optimization of trimmer sequencing that was faster than an analytical contour-like trimmer sequencing. The optimization process was tested using a set of three intracranial treatment plans which were planned using a custom research treatment planning system and were successfully optimized to reduce the additional trimmer sequencing time to approximately 60 s per treatment field while maintaining a high degree of target conformity. Thus, the novel use of ACO techniques within a treatment planning algorithm has been demonstrated to effectively determine collimation sequencing patterns for a DCS in order to minimize the additional treatment time required for trimmer movement during treatment.