Author: Robbie Beckert, Weiren Liu, Thomas R. Mazur, Allen Mo, Stephanie Perkins, Hailei Zhang, Tianyu Zhao 👨🔬
Affiliation: Washington University in St. Louis School of Medicine, University of South Florida, WashU Medicine 🌍
Purpose: Online adaptation may mitigate uncertainties in proton therapy arising from interfractional anatomical changes. While robust optimization accounts for setup and range uncertainties during plan optimization, this study aims to assess the extent to which the robustness of proton plans can be improved, especially in cases where anatomical change is greater than what can be modeled on robust optimization.
Methods: In this in-silico study, an online adaptive proton therapy workflow was developed for a proton machine equipped with CT-on-rails (CToR) as on-board imaging and applied to nine previously treated patients with recurrent rectal tumors. Prescription doses ranged from 25 to 40 Gy over 5 fractions. Significant overlap existed between the planning target volume (PTV) and organs at risk (OARs) with priority given to constraints for OARs such as the small bowel, large bowel, and bladder over target coverage. PTVOpt , which was PTV cropped from OARs by 5mm, has prescription goal of V95%Rx>95%. Initial plans created on simulation CT served as templates for adaptive planning. The optimization templates ensured OAR constraints were met across three range uncertainty scenarios, prioritizing PTVOpt coverage only in the nominal scenario. Adaptive plans were generated for each fraction using these templates. Robustness evaluation with ±3% range uncertainty was performed for both initial and adaptive plans on CToRs, assessing dose metrics for OARs and PTVOpt.
Results: Initial plans exhibited significant variation (average variation 3.6%, maximum13.4%) in PTVOpt coverage (V95%Rx) in evaluation on treatment fraction CTs. In contrast, adaptive plans demonstrated markedly reduced variation (average variation 1.4%, maximum 4.9%), indicating improved robustness of PTVOpt coverage with the adaptive treatment workflow.
Conclusion: This study reveals that in scenarios when significant overlap between target and OARs exists and when OAR constraints take precedence, online adaptive proton therapy can substantially improve the robustness of target coverage.