Accelerated prompt gamma estimation for clinical proton therapy simulations

Area of Science:

• Medical Physics
• Particle Therapy
• Computational Imaging

Background:

• Prompt gammas (PGs) are of interest for particle therapy range verification and dose control.
• Estimating PG yield with traditional Monte Carlo (MC) simulations is slow due to the rarity of PG production.
• Existing methods like pgTLE are limited to analytical phantoms.

Purpose of the Study:

• To develop and validate a novel two-stage variance reduction method, voxelized pgTLE (vpgTLE), for accelerating PG yield estimation in voxelized patient geometries.
• To extend the applicability of PG-based verification methods to realistic, complex treatment plans.

Main Methods:

• Developed a two-stage variance reduction technique (vpgTLE) building upon pgTLE.
• Precomputed PG production probabilities for various materials and energies.
• Simulated primary particle interactions in patient CT data to generate an intermediate PG yield image.
• Used the intermediate PG image as a source for subsequent PG propagation simulations.

Main Results:

• Achieved a gain of approximately 10^3 in simulation speed compared to analog MC for both heterogeneous phantoms and patient CT data.
• Demonstrated agreement with analog MC simulations within 10^-4 with negligible bias.
• Reported voxel-specific gains ranging from 10^2 to 10^4.
• Achieved 2% relative uncertainty in the 90% yield region.

Conclusions:

• vpgTLE effectively extends PG estimation acceleration to voxelized geometries, enabling faster and more accurate range verification in particle therapy.
• The method is generic and compatible with various geometries and beam configurations.
• Memory consumption is a potential concern for large datasets, with trade-offs discussed.