Automated proton treatment planning with robust optimization using constrained hierarchical optimization

Area of Science:

• Medical Physics
• Radiation Oncology
• Computational Biology

Background:

• Robust optimization in proton therapy planning aims to mitigate uncertainties in treatment delivery.
• Existing methods like stochastic and worst-case optimization present limitations in balancing plan quality and robustness.
• A need exists for a method that offers a tunable level of robustness, bridging the gap between these extremes.

Purpose of the Study:

• To develop a fully automated method for generating high-quality, robust proton treatment plans.
• To introduce a novel robust optimization approach using the p-norm function to control robustness levels.
• To evaluate the proposed method's performance against existing stochastic, worst-case, and non-robust approaches.

Main Methods:

• The Expedited Constrained Hierarchical Optimization (ECHO) algorithm was extended for proton therapy.
• A p-norm function was integrated to combine objective functions from 13 scenarios (setup and range uncertainties).
• The p-norm parameter ($$) allows intuitive control over the trade-off between robustness and plan quality.

Main Results:

• The automated method successfully generated robust proton plans for all evaluated cases (head-and-neck patients and phantom).
• Robust plans exhibited narrower dose-volume histogram (DVH) bands and met all hard constraints across all scenarios.
• The p-norm approach with intermediate $$ values improved median objective function values by 15% with only a 3% degradation in the worst-case scenario, compared to the worst-case approach.

Conclusions:

• A novel automated treatment planning approach for proton therapy has been successfully developed.
• The p-norm based method provides adjustable robustness, dose-volume constraint enforcement, and improved plan quality.
• This technique allows clinicians to tailor the level of robustness to specific clinical priorities.