Sci-Thur AM: Planning - 11: The impact of distributed calculation framework settings on plan calculation time.

Area of Science:

• Medical Physics
• Radiation Oncology
• Computational Science

Background:

• Advanced radiation therapy techniques like Volumetric Modulated Arc Therapy (VMAT) and electron Monte Carlo (eMC) require significant computational resources for treatment plan calculations.
• Distributed computing frameworks allow parallel processing across multiple workstations to accelerate these demanding calculations.

Purpose of the Study:

• To investigate the impact of Distributed Calculation Framework (DCF) settings on treatment plan calculation times within the Eclipse treatment planning system.
• To identify optimal DCF configurations for VMAT and eMC plans considering network and workstation hardware limitations.

Main Methods:

• The study utilized the Eclipse treatment planning system with its Distributed Calculation Framework (DCF) on a network of 20 workstations.
• Various DCF settings, including control point and Monte Carlo field parallelization factors, and local servant numbers, were tested.
• Calculation times for VMAT and eMC plans were measured and analyzed in relation to DCF parameter adjustments.

Main Results:

• For VMAT plans, increasing the control point parallelization factor initially reduced calculation time but led to increased times and potential network issues at higher levels due to data transfer overhead.
• For eMC plans, calculation time decreased monotonically with increased Monte Carlo field parallelization, with negligible data transfer impact.
• Increasing local servant numbers reduced data sending time but significantly increased overall calculation time for both VMAT and eMC plans.

Conclusions:

• Optimal DCF settings are dependent on the specific treatment planning task (VMAT vs. eMC), available computational resources, and network infrastructure.
• Careful tuning of parallelization factors and local servant numbers is necessary to balance calculation speed and network load.
• The findings provide guidance for radiotherapy facilities to optimize their distributed computing environments for faster treatment planning.