Purpose: Proton plans are very sensitive to patient anatomical changes due to the nature of proton interaction with tissue. Initial plans tend to distort throughout the treatment fractions due to anatomical changes occurring during treatment. Dose restoration(DR), a fully-automated adaptive strategy which aims to keep the dose distribution constant in every fraction, has proven powerful in lung and pelvic locations. The objective of this work was to test the performance of DR in head and neck(HN) patients.
Methods: Ten clinically planned HN patients underwent between 4 and 6 repeated-CTs (rCTs) during treatment. Robust DR using the same robust optimization parameters as in planning phase (setup-error=3mm, range-error=3%) was performed for all rCTs(n=50). Isodose contours generated from the initial dose and patient-specific minimum and maximum objectives were used in the DR workflow to re-optimize the plan and reproduce the initial dose in every rCT. Robust evaluations (1mm,3%) of non-adapted and of restored plans were compared for each rCT. Differences to clinical plan on the planning CT scan in DVH-metrics (nominal and worst-case values) were analysed for non-adapted and for restored plans.
Results: An improvement in DVH-metrics and robustness was seen for the restored plans compared to non-adapted plans. For non-adapted plans, 28%(14/50) of rCTs did not fulfill the CTV coverage criteria(i.e. D98%>95%Dprescription(Dp)) and therefore needed offline adaptation. By implementing DR, the median of the D98% over the 50 rCTs was increased by 0.61 Gy and 0.17 Gy in the primary tumour CTV(Dp=70Gy) and the nodal target CTV(Dp=54.25Gy), respectively. Because of large anatomical changes and/or inaccurate patient repositioning, 14%(7/50) of rCTs still needed offline adaptation to ensure an optimal treatment.
Conclusion: DR has the potential to reduce offline adaptation rate by 50%. However, a minority of rCTs(14%) still need to be visually inspected and fully-automation might not be possible for these HN patients.
Not Applicable / None Entered.