GRID-ENABLED TREATMENT PLANNING FOR PROTON THERAPY USING MONTE CARLO SIMULATIONS

Area of Science:

• Medical Physics
• Computational Science
• Distributed Computing

Background:

• Grid computing offers a collaborative approach for computationally intensive tasks.
• Traditional single-computer or small-cluster methods are insufficient for complex problems like radiation transport calculations.
• Monte Carlo (MC) simulations are crucial for accurate proton radiotherapy planning.

Purpose of the Study:

• To adapt a Monte Carlo transport code for proton radiotherapy to leverage grid computing.
• To demonstrate the potential of grid computing in drastically reducing simulation runtimes.
• To establish a proof-of-concept Medical Grid for distributed radiotherapy simulations.

Main Methods:

• Extended an existing MC transport code to incorporate grid computing techniques.
• Utilized the GEANT4 simulation environment for proton transport simulations.
• Conducted experiments on a Medical Grid connecting Texas Tech University and Rice University.

Main Results:

• Achieved approximately linear simulation speedup using grid computing.
• Identified parallel runtime variations and communication overhead as factors affecting efficiency.
• Demonstrated that 3x10^5 to 5x10^5 proton events per processor core yield 65-83% efficiency.
• Extrapolated that ~1000 processor cores could reduce an 18.3-day simulation to approximately 1 hour.

Conclusions:

• Grid computing is a promising technology for accelerating MC simulations in proton radiotherapy.
• Distributed computing via a Medical Grid can significantly reduce treatment planning time.
• The findings support the feasibility of using large-scale grid infrastructure for clinical applications.