A gradient inverse planning algorithm with dose-volume constraints

Abstract

An inverse planning algorithm for determining the intensity-modulated beams that will most closely generate a desired dose distribution is presented. The algorithm is three-dimensional and does not explicitly depend on beam energies and modalities. It allows a single prescription dose or a window of acceptable doses to be specified for the target, with additional constraints to account for under- or over-dosing. For the protection of organs at risk, it provides maximum-dose and dose-volume constraints. The latter apply to the entire volume of the organ exposed to the corresponding dose levels. Several levels of each type of constraint, with varying penalty weights, may be specified for each organ. The objective function that serves as the measure of the goodness of the solution is of the least-squares type and is minimized using conjugate gradient methods. Typical clinical cases involving 40,000 points and 4000 rays to be determined require about 10 min of CPU time on a DEC AlphaStation. Results are presented for two clinical sites, prostate and lung. The optimization algorithm yielded plans that featured higher target dose homogeneity, compared with the human planner's plan, while selectively sparing more of the normal organs at the desired dose regions.