Incorporation of Non-Uniform Segment Sampling Into Fluence-Map Based VMAT Planning
H Kim*, R Li, L Xing, Department of Radiation Oncology,Stanford University, Stanford, CASU-E-T-641 Sunday 3:00PM - 6:00PM Room: Exhibit Hall
Purpose: VMAT has been widely used due to enhanced delivery efficiency with high angular coverage. However, general VMAT optimized by single aperture at the control points cannot meet sufficient intensity modulation in some directions. To improve the plan quality over the limit, this work proposes to use the non-uniform segments sampling for VMAT planning by fluence-map based optimization approach.
Methods: Reweighted L1-minimization can optimize the fluence-map for VMAT as it successfully eliminates unnecessary information by further simplifying the fluence-map. To increase the connectivity of adjacent fluence-maps for VMAT planning, this study considers the similarity of neighboring fluence-maps by combining the simple penalized term with the reweighted L1-minmization. Importantly, instead of taking one segment uniformly, the non-uniform segments sampling that takes additional segment at certain directions is introduced to enhance the plan quality for VMAT plans. To choose the appropriate directions giving great benefits in plan quality from the additional segment, we sequentially quantified the respective cost at a specific field by adding extra segment to the field. The extra segments were assigned to 6 to 8 field directions with the lowest costs, which were redistributed into adjacent fields and linked by linear interpolation for VMAT delivery. Prostate patient data was employed to evaluate and compare the uniform/non-uniform segment sampled VMAT in plan quality and treatment time.
Results: Introducing the additional term into optimization effectively enhanced the similarity of adjacent fluence-maps, which were used to form uniform/non-uniform segments sampled VMAT plans. Significantly, non-uniform sampling VMAT formed by assigning 8 segments to the 8 chosen fields has better dose conformity to the target (0.8245 to 0.8409) than uniform sampled VMAT plan at the small expense of the estimated treatment time (63s to 65s).
Conclusion: Fluence-map based VMAT planning incorporated by non-uniform segments sampling can achieve efficient VMAT plan with enhanced plan quality.