Direct Leaf Trajectory Optimization for Volumetric Modulated Arc Therapy with Sliding Window Delivery
D Papp*, J Unkelbach, Massachusetts General Hospital, Boston, MAMO-A-137-8 Monday 8:00AM - 9:55AM Room: 137
Purpose: We present a new single-step algorithm for volumetric modulated arc therapy (VMAT) planning. The algorithm is inspired by the observation that IMRT plans with a large number of beams (e.g., N=20) yield near-optimal treatment plans. However, an arc sequencing step to convert an IMRT plan into a deliverable VMAT plan compromises quality or delivery efficiency. To avoid this caveat, we devise a single-step method to directly optimize the MLC leaf trajectories.
Methods: In our approach, a full 360-degree arc is divided into N arc segments in which the leaves move unidirectionally, i.e. an intensity-modulated field is delivered over each arc segment using a sliding window technique. We formulate the VMAT optimization problem in terms of the leaf trajectories over each arc segment, including all machine and time constraints. Assuming constant dose-influence matrices in each arc segment and constant gantry speed, this optimization problem is convex. Thus, we can reliably determine the optimal VMAT plan under these restrictions. An extension of this basic model is devised that accounts for a varying dose influence matrix within each arc segment.
Results: The algorithm allows us to find highly conformal treatment plans, deliverable in a short time. Allowing for enough delivery time, the algorithm is guaranteed to match the ideal IMRT plan. For prostate cases, the plans are observed to closely approximate the ideal plan with 3-4 minutes delivery time. Head-and-neck results are similarly promising.
Conclusion: The results indicate that high quality single-arc treatments can be delivered with 20 arc segments if sufficient time is allowed for modulation in each segment. Our current implementation is customized towards finding optimal VMAT plans for constant gantry speed and dose rate, which is desirable in practice. In the future, the model can be extended to allow variable dose rate and more general leaf trajectories.
Funding Support, Disclosures, and Conflict of Interest: Research supported by Philips Radiation Oncology Systems.