We introduce a novel approach to BAO, which decouples the selection of beam angles from fluence optimization. Based on a patient specific score beneficial combinations of treatment beams are selected before fluence optimization. The score compromises a matrix of weighted ratios of dose delivered to the tumor to dose delivered to normal tissue and organs at risk for every target voxel and potential irradiation angle. This matrix is searched by a custom-developed genetic algorithm to identify ideal ensembles of beams.

The feasibility and benefit of our approach is illustrated by means of one example: A standard treatment plan featuring seven equi-spaced coplanar beams is compared to optimized plans featuring five and seven non-coplanar beams for a clinical example. Both optimized plans reduce the mean dose to the brainstem and chiasm by more than 60% while guaranteeing superior or equivalent dose distributions in all other volumes of interest as well as comparable target coverage. The presented BAO process requires only a few minutes of computation time on a state-of-the-art workstation for the selected case.