Last vertex splitting: a new retroactive Monte Carlo splitting technique applied to LINAC out-of-field dose...

Area of Science:

• Medical Physics
• Computational Physics
• Radiotherapy Physics

Background:

• Monte Carlo (MC) simulations are crucial for accurate dose calculations in radiotherapy.
• Simulations in high-attenuating media, like LINAC collimators, are computationally intensive.
• Efficient out-of-field (OOF) dose prediction is vital for treatment planning and safety.

Purpose of the Study:

• To introduce a novel variance reduction technique, Last Vertex Splitting (LVS), to accelerate MC simulations.
• To improve the efficiency of out-of-field (OOF) dose calculations in high-attenuating media.
• To facilitate the generation of large MC datasets for deep learning model training.

Main Methods:

• Developed and implemented the Last Vertex Splitting (LVS) variance reduction technique.
• Integrated LVS with a hybrid track length estimator (hTLE) for enhanced performance.
• Evaluated the computational speed-up and residual bias of the LVS method in MC simulations.

Main Results:

• LVS significantly reduces computation time in MC simulations for high-attenuating media.
• The method achieves an efficiency speed-up of up to ×4.5, and ×10 when combined with hTLE.
• Residual bias introduced by LVS remains below one percent, ensuring accuracy.
• Accelerated MC simulations enable rapid generation of datasets for deep learning.

Conclusions:

• Last Vertex Splitting (LVS) is an effective technique for accelerating MC simulations in radiotherapy.
• The LVS method optimizes photon tracking and interaction modeling for faster OOF dose calculations.
• This approach has broad applicability in radiotherapy, radiation shielding, and other particle transport simulations.