3065 - Machine-Specific Delivery Sequence Optimization for Spot-Scanning Proton Arc Therapy Using a Compact Superconducting Synchrocyclotron
Presenter(s)
P. Liu1,2, and X. Ding1; 1Department of Radiation Oncology, Corewell Health William Beaumont University Hospital, Royal Oak, MI, 2Medical Physics,Wayne State University, Detroit, MI
Purpose/Objective(s): Proton arc therapy (PAT) improves dose conformity and normal tissue sparing compare to intensity-modulated proton therapy (IMPT) but faces a "static-to-dynamic" gap. Traditional planning discretizes arcs into static control points, often ignoring continuous gantry rotation and machine-specific delivery constraints. This mismatch can cause significant dosimetric deviations, especially in high-precision treatments like stereotactic radiosurgery (SRS), where unintended dose escalation to adjacent normal tissue or target underdosage can occur. This study evaluates a new machine-specific dynamic delivery sequencing optimization frameworks designed to bridge this gap for both synchrocyclotron and cyclotron-based systems.
Materials/Methods: A dynamic sequencing optimization framework was developed to convert nominal static PAT plans into temporally resolved dynamic arc plans. The process integrated machine-specific timing models—accounting for energy layer switching, spot switching, and spill times—with gantry rotation constraints on angular velocity and acceleration. For synchrocyclotron systems, spots were decomposed into discrete "bursts" to model pulsed delivery, while cyclotron plans utilized continuous spot delivery modeling. Static control points were resampled into dynamic sub-control points at a fixed temporal resolution, followed by a spot-weight fine-tuning optimization that adjusted MU while holding delivery timing and gantry motion fixed. The algorithm was validated using five multi-metastatic brain SRS cases, comparing the nominal plans against reconstructed delivered doses from virtual machine logfiles.
Results: Across the five cases, both frameworks preserved nominal plan quality with no statistically significant differences in target coverage or normal brain sparing (V12 and V8). However, delivery accuracy significantly improved: for the synchrocyclotron system, the median absolute ?D98 for the total target decreased from 38 cGy (range: 8–173 cGy) to 8 cGy (range: 6–16 cGy), while the median ?D98 for the worst-metastasis dropped from 163 cGy to 14 cGy. For the cyclotron system, the median absolute ?D98 for the total target was significantly reduced from 48 cGy (range: 3–179 cGy) to 15 cGy (range: 4–35 cGy) (p=0.031), and the median ?D98 for the worst-metastasis decreased from 176 cGy to 7 cGy (p=0.031). Treatment efficiency remained stable, as there were no significant increases in total spots, energy layers, or dynamic delivery times for either system (synchrocyclotron: p=0.063; cyclotron: p=0.438).
Conclusion: Delivery-aware sequencing optimization is critical for the clinical translation of SPArc. By embedding machine-specific mechanical constraints directly into the optimization, these frameworks mitigate delivery-induced dose perturbations without compromising efficiency or plan quality. This approach provides a solution for future PAT implementation across different proton therapy delivery system.