Author: Laura Buchanan, Samantha G. Hedrick, Stephen L. Mahan, Isabella Pfeiffer, Chester R. Ramsey, Taylor Ransom 👨🔬
Affiliation: Thompson Proton Center, University of Tennessee 🌍
Purpose: Proton pencil beam scanning can reduce normal tissue dose but is highly sensitive to setup, anatomical changes, and range variations. These uncertainties may compromise target coverage and organs at risk sparing. Robust optimization addresses these uncertainties during planning, but robustness plan evaluation often relies on subjective analysis of dose metrics. This study utilizes a large-scale database of perturbed scenarios to establish benchmarks for robust plan evaluation.
Methods: A Python script was developed to calculate perturbed dose distributions in RayStation for shifts, rotations, and range uncertainties. Perturbed DVHs were prospectively saved in a database for population-based analysis and were evaluated for moderately hypofractionated prostate cancer patients without nodal involvement treated to 70 Gy in 28 fractions. Robustness was evaluated for the prostate clinical target volume (CTVp), seminal vesicles (CTVsv), rectum, and bladder. Perturbations included ±3 mm shifts in three axes, ±3-degree rotations, and ±3.5% density uncertainties. DVHs from all scenarios were aggregated to create population-based tolerance levels based on a 68% confidence interval and action levels at a 95% confidence interval.
Results: A total of 348 perturbations for the CTVp, 179 for the CTVsv, 322 for the rectum, and 28 for the bladder were analyzed. The acceptable band (1σ) for the CTVp was 1% of the prescription dose, while the action level (2σ) was 2% between D1 and D95. Of the 348 CTVp perturbations, 81 exceeded 1σ and 23 exceeded 2σ, primarily due to density uncertainties and roll rotations. The CTVsv bands were wider (3–8% for 1σ), reflecting its proximity to OARs. Among 179 perturbations, 40 exceeded 1σ and 16 exceeded 2σ.
Conclusion: This study demonstrates the utility of population-based analysis in evaluating the robustness of proton plans. By analyzing a large dataset of perturbed DVHs, tolerance and action levels based on statistical thresholds can be established.