Planning of beam intensity modulation using an advanced 3D dose calculation algorithm and a simulated annealing method

Purpose: The aim of this work was to develop a fast inverse planning algori thm that will calculate optimum beam intensity distributions and beam shape s, and to incorporate the algorithm into a three-dimensional CT planning sy stem. Method: The algorithm is based on the technique of simulated annealing and produces beam intensity distributions that could in principle be implemente d clinically, either by the use of compensators or dynamic multileaf collim ation. Dose distributions are calculated using a voxel beam model based on a spherical co-ordinate system, and transformations are given allowing the dose to be determined at any point within the patient. The dose calculation algorithm calculates primary and scattered dose separately from a knowledg e of tissue/air ratios and differential scatter/air ratios, and both are co rrected for the presence of heterogeneities in three dimensions. Specific a ttention is given to the execution time of the algorithm, and the methods d eveloped allow satisfactory results to be achieved in calculation times whi ch are sufficiently fast to be used interactively in the planning system. S everal objective functions have been developed and can be selected in a sim ple manner by the user. In general, these attempt to achieve a uniform dose within the target while limiting the dose to organs at risk, either by upp er dose limits or by specifying constraints on their dose volume histograms . Results: The beam intensity distributions produced from the optimization ha ve been used automatically by the forward planning system to produce three- dimensional dose distributions, and the results obtained in a number of cli nical situations are presented. Conclusions: The inverse planning algorithm developed has been successfully incorporated into a three-dimensional planning system and is capable of pr oducing beam intensity modulated distributions for clinical implementation. The execution time of the algorithm is sufficiently fast to be used as an optimization tool in an interactive forward planning system. (C) 1999 Elsev ier Science Ireland Ltd. All rights reserved.