A fast robust optimizer for intensity modulated proton therapy using GPU.

Area of Science:

• Medical Physics
• Radiation Oncology
• Computational Biology

Background:

• Robust optimization enhances treatment plan stability in intensity modulated proton therapy (IMPT).
• Current IMPT optimization algorithms are often time-consuming, limiting clinical application.
• There is a need for faster, robust optimization methods in proton therapy.

Purpose of the Study:

• To develop and evaluate a fast robust optimization tool for IMPT utilizing GPU parallel computing.
• To assess the clinical feasibility and efficiency of the new tool compared to existing methods.
• To improve the reliability and speed of proton therapy treatment planning.

Main Methods:

• A novel robust optimization model based on nine boundary dose distributions considering range and setup uncertainties.
• Calculation of boundary influence matrices using an in-house finite size pencil beam dose engine.
• Optimization using the conjugate gradient method tuned for CPU+GPU heterogeneous platforms.

Main Results:

• The proposed method demonstrates robustness against range and setup uncertainties in IMPT.
• Clinical cases (head and neck, lung, prostate) show high target dose uniformity and better OAR protection.
• Significant reduction in execution time for head and neck and prostate cases compared to Varian Eclipse.

Conclusions:

• The developed fast robust optimizer improves the reliability of proton therapy planning.
• GPU acceleration makes robust optimization faster and potentially more clinically applicable.
• The tool offers a faster alternative for robust IMPT, enhancing safety and efficiency.