Moving GPU-OpenCL-based Monte Carlo dose calculation toward clinical use: Automatic beam commissioning and source...

Area of Science:

• Medical Physics
• Computational Biology
• Radiotherapy

Background:

• The goMC GPU-based Monte Carlo (MC) dose engine was previously developed on the OpenCL platform.
• Advancements are needed to adapt MC dose engines for routine clinical use in radiation therapy.

Purpose of the Study:

• To enhance the goMC dose engine for clinical applications by improving its commissioning process and dose calculation efficiency.
• To validate the accuracy and speed of the improved goMC engine for treatment plan dose calculations.

Main Methods:

• Adapted an automatic beam commissioning approach using optimization to minimize dose discrepancies between calculations and measurements.
• Developed an efficient source particle sampling strategy using pre-generated fluence maps to bias particle sampling.
• Validated the goMC engine's accuracy using 3D gamma index tests on patient treatment plans.

Main Results:

• Successfully commissioned six Varian TrueBeam linac photon beam models (6-18 MV, including FFF beams).
• The new sampling strategy reduced the number of source particles needed, accelerating dose calculation by a factor of ~2.8 (e.g., 150.9 s to 51.5 s for a VMAT case).
• Achieved high accuracy with passing rates of 99.82% (2%/2 mm) and 95.71% (1%/1 mm) compared to a commercial treatment planning system.

Conclusions:

• The auto-commissioning approach and efficient source sampling strategy are effective and feasible for fast and accurate MC dose calculations in radiotherapy.
• The improvements position goMC as a viable tool for clinical treatment planning, offering speed and precision.