Proof-of-concept: Novel CBCT-based adaptive robust optimization in sinonasal cancer proton therapy
Nadine Vatterodt,
Denmark
PD-0243
Abstract
Proof-of-concept: Novel CBCT-based adaptive robust optimization in sinonasal cancer proton therapy
Authors: Nadine Vatterodt1,2, Raul Argota-Perez3, Maja Bendtsen Sharma4, Anne Ivalu Sander Holm4, Ulrik Vindelev Elstrøm1, Kenneth Jensen1, Stine Sofia Korreman1,2,4
1Aarhus University Hospital, Danish Center for Particle Therapy, Aarhus, Denmark; 2Aarhus University, Department of Clinical Medicine, Aarhus, Denmark; 3Herlev and Gentofte Hospital, Department of Oncology, Herlev, Denmark; 4Aarhus University Hospital, Department of Oncology, Aarhus, Denmark
Show Affiliations
Hide Affiliations
Purpose or Objective
Anatomical variations in proton therapy for sinonasal cancer patients can potentially deteriorate treatment quality. Previous studies including robustness towards mucus variations has shown promising results, however using simple PTV-based optimization with artificially created nasal cavity filling scenarios. In this study, we investigated novel strategies combining adaptation and multi-image robust optimization by using actual anatomical variations from daily CBCTs.
Material and Methods
A retrospective study on five sinonasal cancer patients was performed in RayStation 9A. For each patient, synthetic CTs (synCTs) were derived from daily CBCTs in MIM Maestro. In addition, two CTs with filled and cleared cavities were created by overwriting the CT numbers in the planning scans (using CERR software).
Plans for four IMPT robustness schemes were generated as illustrated in Fig 1a:
1. Conventional robust optimized plans (cRO)
2. Initial full anatomical robust optimized plans including the two artificial scenarios (fRO)
3. Adaptation after the first week with individualized fRO plans using the first five fractions’ synCTs (afRO)
4. Weekly adaptation with individualized fRO plans including the previous five fractions’ synCTs (wafRO)
For all plans, robust optimization and evaluation for geometrical uncertainties were performed assuming ±2mm setup and ±3.5% range uncertainty. All plans were approved according to passing criteria in Fig. 1b. Results were evaluated by comparison of target coverage (V95) and maximum dose to OAR (D0.03cc) for fraction doses recalculated on the daily synCTs and accumulated dose by deformable mapping of fraction doses to the planning CT.
Results
All approaches showed sufficient robustness in accumulated target coverage (Fig. 2a). Adaptive approaches using daily images showed comparable or better robustness than conventional plans and recovered two cases of underdosage for single fractions in the conventional plans. Largest improvements in accumulated and daily target coverage were obtained for weekly adaptation.
The cost in maximum OAR dose in comparison to the gain in target coverage was highly dependent on structure and patient. This is demonstrated in Fig. 2b for structures that violated clinical constraints in total or daily doses for the adaptive approaches. Optimization using artificial scenarios compared unfavorably with any other plans.
Conclusion
Using prior daily images as uncertainty scenarios for plan adaptation with multi-image robust optimized plans improved target coverage and performed better than with artificial cavity filling scenarios. The strategy using weekly adaptation was superior to any other strategy. As conventional plans for the investigated patient group were already quite robust with respect to target coverage, doses to OAR may take priority. However, adaptive CBCT-based robust optimization could be considered for cases with severe anatomical changes during the treatment course.
