Intrinsic spatial heterogeneity in inverse planning and its significant role in defining the universe of IMRT solution space

Abstract

Listen

Translate article

Voxels within a target or a sensitive structure volume are generally not equivalent in achieving their dosimetric goals in IMRT planning. Depending on the patients geometry, beam modality and field configuration, some regions may have better chance to meet the prescription than others, and vise versa. In this work we model the intrinsic spatial heterogeneity by introducing the concept of voxel-dependent dosimetric capability. We then propose an approach to compensate for the adverse effect of the heterogeneity by purposely modulating the importance of the voxels based on the a priori dosimetric capability information of the system and demonstrate that the technique provides an effective way to substantially improve the sub-optimal performance of the existing inverse planning algorithms. We quantify the degree for a voxel to achieve its dosimetric goal by introducing the concept of dosimetric capability for each voxel in a target or sensitive structure and describe a Cimmino algorithm for numerically computing the capability distribution on a case specific basis. The capability of a voxel represents a priori dosimetric knowledge of the system and is useful to construct an objective function with customized non-uniform importance factors. Generally speaking, the final optimal dose distribution depends not only on the prescription and the values of the structure specific importance factors, but also on the dosimetric capability distribution in an implicit fashion. In some extreme cases, a region in a target may never meet the prescribed dose without seriously deteriorating the doses in other areas. Conversely, the prescription in a region may be easily met without violating the tolerance of any sensitive structure. The intra-organ tradeoff was modulated purposely using a heuristic model with consideration of the intrinsic heterogeneity of the dosimetric capability of the system. The formalism is applied to three clinical cases (thoracic, prostate and head and neck) to illustrate the technical details of the new inverse planning technique and to show its advantage in improving IMRT dose distributions. The dosimetric capability of a voxel has been proved to be a useful measure of relative easiness for a point in the target or sensitive structure to achieve its dosimetric goal. An inverse planning framework with customized non-uniform importance factors has been established. It has been shown that the incorporation of the capability information permits us to construct physically and clinically more meaningful objective function and greatly enlarge the universe of solution space. The purposed modulation of spatial penalty distribution has been shown to be advantageous over the conventional inverse planning technique with structurally uniform importance factors, leading to significantly improved IMRT treatment plans that would otherwise unattainable. For all three cases, we found that the doses to the sensitive structures can be reduced almost uniformly by more than 20% (normalized the maximum dose of the corresponding structure) while keeping the same target dose coverage. Conversely, the target dose can often be escalated by more than 10% while keeping the radiation toxicity at its current IMRT level. The concept of the a priori dosimetric capability distribution of the system sheds new insight into inverse planning problem and allows us to better interpret a patients IMRT dose distribution. The tactic assumption in current inverse planning that all points within a structure are equivalent represents only one of the numerous penalty schemes (generally, not an optimal one) and leads to sub-optimal IMRT treatment plans. IMRT dose distributions can be improved substantially by incorporating a priori dosimetric capability information into the inverse planning process