Program Information
Mixed-Integer Optimization-Based Aperture Sequencing for Deliverable 4D IMRT in Lung SBRT
M Hamzeei1*, A Modiri1 , A Hagan1 , N Kazemzadeh1 , A Sawant1 , (1) University of Maryland, School of Medicine, Baltimore, MD
Presentations
TU-D-108-1 (Tuesday, August 1, 2017) 11:00 AM - 12:15 PM Room: 108
Purpose: We have previously developed a particle swarm optimization (PSO)-based 4D-IMRT technique that explicitly accounts for and optimizes across all respiratory phases in order to achieve significantly higher dose-sparing of critical organs. A major challenge is delivering the large number of apertures in a clinically feasible manner (e.g., a 10-phase 4DCT, a 10-beam plan, with 166 apertures per beam, would involve 16,600 apertures). We address this problem by developing a mixed-integer optimization technique for aperture sequencing that aims to maximize the delivery efficiency by accounting for leaf velocity, aperture weights and time-per-respiratory phase.
Methods: 4D IMRT plans were generated for five lung SBRT patients using our in-house PSO engine. For each aperture, delivery time per aperture and the transition times from that aperture to all other apertures were calculated by considering the maximum MLC leaf velocity (3.5cm/s for Varian MLC). Two key criteria were considered in problem formulation: (i) each control point could not be visited multiple times in a respiratory cycle, (ii) in each respiratory phase, only its associated apertures could be delivered. Our model could handle sparsity of apertures, as well as piecewise delivery of apertures whose required delivery time was longer than the phase duration in one cycle. We compared our proposed method with a greedy method which only considered neighboring apertures for the following steps in the sequence.
Results: Our optimization-based sequencing yielded deliverable sequences of apertures and maximized the delivery efficiency. Average number of respiratory cycles required for delivery was reduced by 19%; i.e., 76.6 respiration cycles (ranging in 28-127 cycles for the studied 5 patients), compared to the greedy method.
Conclusion: Efficient delivery of large number of sparse control points introduces a challenge in 4D-IMRT treatment planning and delivery. We demonstrate an aperture sequencing optimization technique that results in clinically-feasible time scales for delivery.
Funding Support, Disclosures, and Conflict of Interest: This work was supported by NIH R01 CA169102 grant.
Contact Email: