Program Information
Accelerating Sliding Window VMAT by Open-Out/close-in Field Delivery Techniques
D Craft1*, (1) Massachusetts General Hospital, Boston, MA
Presentations
TH-AB-BRB-6 (Thursday, July 16, 2015) 7:30 AM - 9:30 AM Room: Ballroom B
Purpose:
To describe how to make sliding window VMAT delivery more efficient by customizing the delivery of each angular sector based on the shape of the fluence maps.
Methods:
VMAT is a highly non-convex optimization problem. We decompose the problem by first considering an IMRT problem with 30 equispaced beams. This problem is solved with smoothing terms and the resulting fluence maps are analyzed for delivery considerations. Rows of the fluence maps that are bimodal are to be delivered with a leaf sweep across the field, both leaves moving in same direction. Unimodal rows can be delivered more efficiently with leaves either closing inward or opening outward. More classifications of this type are done, as well as re-optimizations of the 30 beam IMRT problem with penalties to achieve the desired fluence map restrictions. Mixed integer programming is used for some leaf trajectory calculations and to decide which sectors will get non-sliding window delivery.
Results:
By iterating between fluence map classification and optimization with penalties which support the classifications, VMAT plans of similar quality to full sliding window plans can be produced with improved delivery time. For a fluence row that is unimodal and “steep enough” (shallow unimodal fields require unrealistically fast leaf speeds), delivery time using leaf close-in/open-out can be more than twice as fast as sliding window delivery. For a prostate case, by classifying 50% of the fluence maps as unimodal deliverable, we are able to reduce the treatment time from 3.5 minutes to 2.6 minutes with a negligible change in plan quality.
Conclusion:
By choosing the delivery style of fluence maps based on their shapes, rather than imposing that every map gets a full sliding window delivery, we can retain the high-quality of sliding window style VMAT while gaining delivery efficiencies of non-sliding window VMAT.
Funding Support, Disclosures, and Conflict of Interest: David Craft is partially supported by a research collaboration with RaySearch Laboratories.
Contact Email: