Efficiency enhancements of a Monte Carlo beamlet based treatment planning process: implementation and parameter study

Area of Science:

• Medical Physics
• Radiation Oncology
• Computational Science

Background:

• Intensity-modulated radiation therapy (IMRT) planning using Monte Carlo (MC) simulations is computationally intensive.
• Existing treatment planning processes (TPP) face significant computational burdens, especially with MC dose calculations.
• Optimizing computational efficiency is crucial for widespread adoption of advanced TPP techniques.

Purpose of the Study:

• To enhance the computational efficiency of a fully MC-based TPP for various photon and electron beam techniques.
• To investigate the impact of implementing parallelization, sparse data formats, and voxel merging on computational time and plan quality.
• To identify optimal parameter settings for balancing computational efficiency and treatment plan quality.

Main Methods:

• Developed a framework for efficient parallel MC beamlet calculation across CPU cores.
• Implemented user-defined parameters for statistical uncertainty, sparse dose threshold, and voxel merging distances.
• Evaluated trade-offs between computational efficiency and plan quality using clinical and academic cases across different treatment techniques.

Main Results:

• Optimized parameters (5% photon/15% electron uncertainty, 0.1% sparse threshold, 1-2 cm voxel merging) reduced computation times by 58-96% for beamlet calculation and optimization.
• Achieved significant computational speed-up with only minor degradation in overall plan quality, except for a slight decrease in Organ at Risk (OAR) sparing.
• Demonstrated that reduced statistical uncertainty and data reduction techniques yield substantial efficiency gains.

Conclusions:

• Implemented methods effectively improve the computational efficiency of MC-based TPP.
• The optimized approach offers a viable trade-off between computational speed and treatment plan quality.
• This work facilitates faster and more efficient MC-based treatment planning.