Robust optimization for intensity-modulated proton therapy with soft spot sensitivity regularization.

Area of Science:

• Medical Physics
• Radiation Oncology
• Computational Biology

Background:

• Proton dose distribution accuracy is challenged by range and patient positioning uncertainties.
• Current worst-case scenario optimization methods for proton robustness are computationally intensive.

Purpose of the Study:

• To develop a novel, computationally efficient intensity-modulated proton therapy (IMPT) optimization method.
• To integrate dose fidelity with a sensitivity term accounting for range and positioning uncertainties.

Main Methods:

• A new optimization framework was formulated with dose fidelity and robustness terms.
• Spot sensitivity to perturbations was calculated using spatial gradients of dose distributions.
• The fast iterative shrinkage-thresholding algorithm was employed for optimization.
• The sensitivity-regularized (SenR) method was tested on skull base and head-and-neck cancer patients, compared to conventional (Conv) and worst-case (WC) methods.

Main Results:

• SenR methods improved CTV dose coverage and plan robustness compared to the conventional method under both range and setup uncertainties.
• CTV-based SenR demonstrated superior organ-at-risk (OAR) sparing compared to the WC approach.
• The SenR optimization runtime was significantly shorter (eight times) than the WC method.

Conclusions:

• A computationally efficient, robust IMPT optimization method (SenR) was developed.
• SenR integrates dose fidelity with spot sensitivity to perturbations, offering flexibility and trade-offs.
• The method is resilient to uncertainties and its fast computation time makes it suitable for clinical IMPT planning.