Optimization of importance factors in inverse planning

Inverse treatment planning starts with a treatment objective and obtains th e solution by optimizing an objective function. The clinical objectives are usually multifaceted and potentially incompatible with one another. A set of importance factors is often incorporated in the objective function to pa rametrize trade-off strategies and to prioritize the dose conformality in d ifferent anatomical structures. Whereas the general formalism remains the s ame, different sets of importance factors characterize plans of obviously d ifferent flavour and thus critically determine the final plan. Up to now, t he determination of these parameters has been a 'guessing' game based on em pirical knowledge because the final dose distribution depends on the parame ters in a complex and implicit way. The influence of these parameters is no t known until the plan optimization is completed. In order to compromise pr operly the conflicting requirements of the target and sensitive structures, the parameters are usually adjusted through a trial-and-error process. In this paper, a method to estimate these parameters computationally is propos ed and an iterative computer algorithm is described to determine these para meters numerically. The treatment plan selection is done in two steps. Firs t, a set of importance factors are chosen and the corresponding beam parame ters (e.g. beam profiles) are optimized under the guidance of a quadratic o bjective function using an iterative algorithm reported earlier. The 'optim al' plan is then evaluated by an additional scoring function. The importanc e factors in the objective function are accordingly adjusted to improve the ranking of the plan. For every change in the importance factors, the beam parameters need to be re-optimized. This process continues in an iterative fashion until the scoring function is saturated. The algorithm was applied to two clinical cases and the results demonstrated that it has the potentia l to improve significantly the existing method of inverse planning. It was noticed that near the final solution the plan became insensitive to small v ariations of the importance factors.