Smoothing intensity-modulated beam profiles to improve the efficiency of delivery

Intensity-modulated beam profiles are generated by an inverse planning or optimization algorithm, a process that, being computationally complex and intensive, is inherently susceptible to noise and numerical artifacts. These artifacts make delivery of the beams more difficult, oftentimes for little, if any, observable improvement in the dose distributions. In this work we examine two approaches for smoothing the beam profiles. The first approach is to smooth the beam profiles subsequent to each iteration in the optimization process (method A). The second approach is to include a term within the objective function that specifies the smoothness of the profiles as an optimization criterion (method B). The two methods were applied to a phantom study as well as three clinical sites: paraspinal, nasopharynx, and prostate. For the paraspinal and nasopharynx cases, which have critical organs with low tolerance doses in close proximity, method B produced sharper dose gradients, better target dose homogeneity, and more critical organ sparing. In the less demanding prostate case, the two methods give similar results. In addition, method B is more efficient during optimization, requiring fewer iterations, but less efficient during DMLC delivery, requiring a longer beam-on time.