KEY POINTS
- The study tested a new treatment-planning strategy in two representative proton cases, one prostate and one nasopharyngeal cancer plan, using a research version of RayStation. The prostate plan prescribed 40 Gy(RBE) in 20 fractions, while the head-and-neck plan prescribed 69.96 Gy(RBE) with robust IMPT optimization.
- Candidate spots were selected using a multistep workflow that required them to stop inside the target, remain away from air cavities and target boundaries, minimize normal-tissue exposure, and avoid excessive overlap with other boosted spots. Between 1 and 5 spots per beam were then reoptimized.
- Two boosting strategies were tested. One assigned a fixed number of 2×10⁸ to 8×10⁸ protons per selected spot; the second used a reward-based objective that allowed the optimizer to increase spot weight above a 2×10⁸-proton lower bound while balancing standard planning objectives.
- Plan quality remained broadly acceptable until approximately 7×10⁸ protons per boosted spot, although the threshold depended on anatomy and the number of boosted spots. Target coverage was restored after optimization by rescaling plans so that target D98 matched the original plan.
- Higher boosting increasingly affected LET-related metrics before conventional dose metrics. In the head-and-neck case, the maximum observed brainstem differences across fixed-boost plans were approximately 3% for D2, 21% for LET2 and 5% for LET-weighted D2; the reward-based approach reduced these maxima to roughly 2%, 15% and 2%, respectively.
- Simulated prompt-gamma statistics improved markedly. Increasing selected spots to the 2×10⁸-proton detectability threshold reduced profile noise by 76.6% ± 9.7% in the prostate case and 86.6% ± 3.5% in the head-and-neck case. At 8×10⁸ protons, noise fell by 88.1% ± 4.9% and 93.3% ± 1.7%, respectively.
- The study remains a proof of concept: only two patient geometries were analyzed, prompt-gamma signals were simulated rather than clinically measured, treatment-delivery time was not quantified, and no actual range deviation or adaptive-treatment decision was prospectively tested.
CLINICAL TAKEAWAY
Instead of treating range verification as something added after proton planning, this approach designs the plan to generate a stronger verification signal from the outset. The physics is promising for adaptive proton therapy, but clinical usefulness will depend on experimental prompt-gamma measurements, broader patient testing and demonstration that improved signal quality actually changes range-verification or replanning decisions.