Premium
An algorithm for systematic selection of beam directions for IMRT
Author(s) -
Gaede S.,
Wong E.,
Rasmussen H.
Publication year - 2004
Publication title -
medical physics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.473
H-Index - 180
eISSN - 2473-4209
pISSN - 0094-2405
DOI - 10.1118/1.1636572
Subject(s) - beam (structure) , laser beam quality , beam diameter , computer science , set (abstract data type) , optics , physics , medical physics , algorithm , laser beams , laser , programming language
Selection of the number of beams and their directions can be an important problem in radiation therapy, especially when a tumor surrounds a critical organ or is surrounded by multiple critical organs. Beam directions, in this sense, are chosen to not only avoid critical organs, but also to achieve better target dose uniformity. In intensity‐modulated radiation therapy (IMRT), optimization of beam directions is further complicated due to the dependence of one beam direction on its corresponding beamlet intensities and the beamlet intensities of all other beam directions. The result is an excessively enlarged search space, even when the number of beams is small (two to three). Until now, only a handful of publications exist regarding beam direction optimization in IMRT. Here, we report a new systematic approach that determines a suitable number of “more optimal” beam directions without optimizing a complicated objective function or resorting to brute force. We start by assuming that beam directions chosen for an N ‐beam plan are candidates for beam directions in the search for an ( N + 1 ) ‐beam plan. Knowing that beam directions in an N ‐beam plan are not always the best choices for the ( N + 1 ) ‐beam plan, we introduce into the beam direction selection process an analysis of the beamlet weights of every beam direction set sampled. If the relative weights of any particular beam compared to other beams are insignificant and hence have no significant effect on the quality of the treatment plan, then we eliminate this beam from the plan. The algorithm terminates basically when the relative weights of the last beam compared to other beams are insignificant or the replacement of an eliminated beam does not improve the plan. This concept was applied to three two‐dimensional phantoms and each plan was compared to a standard equally spaced IMRT plan in terms of dose distributions, dose–volume histograms, and objective function values. The results show improvements in both target dose uniformity and critical organ sparing often with a fewer number of beams than standard equally spaced beam plans.