Towards real-time photon Monte Carlo dose calculation in the cloud

Area of Science:

• Medical Physics
• Computational Science
• Radiotherapy

Background:

• Monte Carlo (MC) dose calculations are crucial for accurate radiotherapy planning but are computationally intensive.
• Existing solutions using GPUs or CPU clusters are costly and have scalability limitations.
• Near real-time MC application in clinics is hindered by long runtimes.

Purpose of the Study:

• To develop and validate a cloud-based MC framework for high-scalability, accurate photon dose calculations.
• To provide a cost-effective and accessible alternative to hardware-based MC solutions.
• To enable near real-time dose computation for clinical and research applications.

Main Methods:

• Developed a client-server software solution for cloud-based MC simulations on a virtual supercomputer.
• Implemented dynamic resource provisioning and advanced encryption for secure, on-the-fly data handling.
• Integrated the client application with a treatment planning system for seamless workflow automation.

Main Results:

• Achieved near real-time dose computation with excellent linear scaling for high-resolution datasets.
• Demonstrated absolute runtimes of 1.1 to 10.9 seconds for clinical prostate and liver cases (up to 1% uncertainty).
• Runtimes included data transfer, scheduling, and synchronization overhead.

Conclusions:

• Cloud-based MC simulations offer a fast, affordable, and accessible alternative for accurate dose calculations.
• The proposed framework overcomes the hardware cost and scalability barriers of traditional GPU/CPU solutions.
• Enables near real-time MC dose computation, advancing radiotherapy planning and research.