AbstractID: 9051 Title: Algorithm and performance of a clinical beam-angle
optimization system
An IMRT beam-angle optimization (BAO) system optimizes two sets of variables: the beam angles and, for each set of
beam angles, the beam profiles. Two main hurdles preventing an IMRT BAO system from being used in a routine clinical
setting are (1) the substantial amount of time required to perform a single IMRT beam-profile optimization and (2) the
huge size of search space involved in the beam-angle optimization. We have investigated the performance and potential of
an efficient BAO system that overcomes these two impediments. In this BAO system, successive sets of beam angles are
selected from a set of predefined beams using the fast simulated annealing algorithm. This global search method is capable
of finding the global or near-global minimum efficiently. For each set of beam angles, an IMRT beam-profile optimization
is performed. The IMRT optimization is accelerated by using a fast dose calculation method that utilizes a precomputed
dose kernel. This technique allows the IMRT optimization to be performed more than two orders of magnitude faster than a
similar optimization that uses a convolution dose calculation engine. We demonstrate the potential of this system on a few
clinical cases with different complexities including prostate, lung, and head-and-neck cancers. The performance benchmark
and techniques of accelerating the program will be reported. Preliminary results show that this system is fast and practical.
In its current implementation, the BAO for a typical prostate case can be completed in about one hour on a modest
workstation.