Triggered plan adaptation using multi-image optimization for improved robustness in head-and-neck cancer proton...

Anatomical variations in head-and-neck (HNC) proton therapy may degrade target coverage and organ-at-risk (OAR) sparing. This study introduces a triggered robust adaptation strategy utilizing multi-image optimization to progressively enhance plan robustness guided by observed dose deviations for targets and OARs.

Five oropharyngeal cancer patients treated with proton therapy were retrospectively analyzed. Synthetic computed tomography (CT) scans were generated from daily cone-beam CTs for dose recalculation. Four strategies were compared: no adaptation (NA), triggered adaptation without anatomical robustness (TA), and triggered robust adaptation using single (TRA-S) or multiple (TRA-M) triggered fraction images. Adaptation was triggered by exceeded thresholds in dose-volume metric deviations. Treatment strategies were evaluated by comparing target coverage, OAR dose, integral dose and number of threshold violations.

TRA-S and TRA-M consistently improved target coverage up to +0.76 Gy in median D_98% vs. NA, with fewer threshold violations (5 and 4) compared to NA and TA (both 8). Larger unfavorable dose deviations were reduced, with OAR doses generally comparable to NA. TA maintained OAR doses closer to the planned values, while not improving target coverage in all cases. Integral dose increased with TRA strategies up to 3.7 Gy·L compared to NA. Differences between TRA-S and TRA-M were generally small.

Triggered robust adaptation balanced target coverage and OAR sparing, while requiring fewer adaptations compared to triggered adaptation without anatomical robustness. It offers a potential pathway for implementing anatomical robust optimization in HNC proton therapy without relying on predicted images or additional CT scans.