Predictive Smart-Spill Scheduling: Minimizing Dead Time In Synchrotron-Based Spot Scanning Via Look-Ahead Energy Management
Abstract
Purpose
Synchrotron-based proton spot scanning includes non-productive dead time from layer switching and spill change/reinjection. In dose-driven continuous scanning, spill changes are commonly triggered only after beam exhaustion, causing sequential overheads. We propose a predictive Smart-Spill scheduling strategy that initiates spill changes earlier when they can be overlapped with unavoidable dead time, reducing total delivery duration.
Methods
Smart-Spill was implemented in a high-fidelity control-system simulator (“Digital Twin”) that reproduces the timing handshake between the Treatment Planning System and Delivery Control System. The Digital Twin models layer-switch latency, spill change and synchrotron reinjection, and pulse steering (PST) off/on events associated with break spots. At each layer transition, a cost-based search evaluates candidate proactive spill-change points, trading absorbed dead time against the risk of requiring an additional spill cycle later in the layer. To ensure non-degrading behavior, optimization is applied only when (beam wasted/max beam) < (time saved/spill-change time).
Results
Five clinical cases were simulated (Prostate SBRT, Head & Neck, Lung, Breast, and Partial Breast Irradiation), using total delivery time (beam time plus modeled dead time) summed across all fields. Delivery time decreased in four of five cases: Prostate SBRT (80.23 to 79.98 s; 0.25 s, 0.31%), Head & Neck (125.90 to 125.65 s; 0.25 s, 0.20%), Breast (244.19 to 241.69 s; 2.50 s, 1.02%), and Partial Breast Irradiation (35.29 to 34.79 s; 0.50 s, 1.42%). The Lung case showed no change, consistent with the safeguard rejecting unfavorable candidates in low-MU-per-layer plans. Mean improvement across all cases was 0.59% (0.70 s).
Conclusion
Predictive Smart-Spill scheduling provides a plan-dependent reduction in synchrotron spot-scanning dead time, with safeguards to prevent increases in delivery time.