Author: Grant Evans, Maxwell Arthur Kassel, Charles Shang, Stephen Shang 👨🔬
Affiliation: South Florida Proton Therapy Institute, SFPRF 🌍
Purpose: Proton pencil beam scanning therapy is particularly sensitive to field translational shifts and beam range variations, which can degradation of dose distribution and compromise the treatment. To mitigate these uncertainties, treatment plans are optimized for robustness, ensuring acceptable dose distributions under varying conditions. While current robustness evaluations focus on discrete perturbation scenarios, a continuous metric for assessing dose coverage loss remains an unmet need. This study introduces a novel approach using the area under the curve (AUC) difference between perturbation and baseline DVH curves to quantitatively assess dose plan robustness to the target.
Methods: Dose perturbations to the target were generated for various failure types, including translational and range shifts, and their effects were quantified using AUC differences between perturbed and original DVH curves. A binary image processing method was applied to extract CTV contours for analysis. AUC differences were computed across multiple failure thresholds, integrating information from DVH curve deviations. A box-and-whisker analysis was performed to compare AUC differences across failure categories, with statistical significance assessed using t-tests and interaction p-values.
Results: The AUC difference metric successfully captured multiple types of coverage loss, including high-dose reduction, shoulder degradation, and low-dose deficits. A combined analysis of all perturbations demonstrated increasing AUC differences with larger perturbations, highlighting the sensitivity of this metric to dose variation. Statistical testing revealed significant differences in AUC between failure categories, with a p-value of 0.001 between the first two failure types and 0.053 between the second and third.
Conclusion: AUC difference provides a continuous, quantitative measure of plan robustness, integrating dose loss across the entire DVH curve. This approach enables a more comprehensive assessment of treatment plan stability compared to discrete perturbation evaluations. Future work will explore the application of this method to real-time CBCT-guided setup decisions in proton therapy, potentially improving adaptive treatment strategies.