(Wednesday, 8/1/2018) 1:45 PM - 3:45 PM
Room: Davidson Ballroom A
Purpose: Beam angle optimization for IMPT is a challenging problem which requires the treatment planner to make a decision based on the tradeoffs between different optimization objectives. We introduce a new method for optimizing the IMPT beam orientation based on the multicriterion Pareto surface.
Methods: We developed a multicriterion optimization (MCO) algorithm based on a parallelized differential evolution (DE) algorithm implemented on a multi-node multi-GPU cluster. The MCO algorithm generates a Pareto database of DVH-based objectives, and explores the entire feasible (co-planar and non-coplanar) beam angle space for a given number of fields. For every set of beam angles, the dose calculation was performed using in-house Monte Carlo code, and the proton beam spots intensities were optimized using a GPU-accelerated iterative method. The Pareto-based beam angle optimization algorithm was tested on a brain tumor case with three fields. A set of DVH-based objectives was used to gauge the performance of the plans. The DVH objectives were the 98% target volume coverage and the dose to the organs-at-risk at the 50% and 1% volume levels. A comparison with a manually selected set of beam angles to plans on the Pareto surface was performed.
Results: The beam angle optimization algorithm generated the Pareto surface for the brain tumor case. The algorithm was able to identify beam angles which performed better than the manually selected angles. The improvements of the found angles showed an average of 20-30% reduction in dose to organs at risk for the same target coverage. The total optimization computation time was less than 1 hour
Conclusion: The Pareto based beam angle optimization algorithm was used to automatically generate the Pareto database an provide high quality plans with minimal human input.
Keywords
Not Applicable / None Entered.
Taxonomy
Not Applicable / None Entered.
Contact Email
