The open archive for STFC research publications

Full Record Details

Persistent URL http://purl.org/net/epubs/work/12273478
Record Status Checked
Record Id 12273478
Title Fast Monte Carlo proton treatment plan validation in the Google Cloud
Abstract Monte Carlo validation of proton therapy treatment plans is one method to improve the quality of delivered irradiation by reducing the dose calculation uncertainty, but its use is hampered by its typically slow speed in comparison to the overall planning and treatment workflow. Most attempts at fast Monte Carlo calculation have hitherto used locally-based clusters or graphical processing units, usually combined with a simplified implementation of the proton transport. Here we present the use of cloud computing to perform rapid Monte Carlo dose estimation using the comprehensive physics model implemented within GEANT4. Mock proton treatment plans in a head-and-neck phantom were validated using 1600 Google Cloud CPUs. The plan consisted of 16 million voxels each approximately 1 × 1 × 2 mm in size, and 10 million primary particles were tracked, giving a dose uncertainty of 1% within the high-dose region. The validation was completed in just 6 minutes and cost $11.20 (£7.17); reduced execution time is possible with more CPUs, bringing real-time validation within reach. Use of commercial cloud-based infrastructure can eliminate much of the local computing cost for this application, and on-demand scalability is expected to allow decreased calculation turnaround time without incurring greater cost. Security concerns are likely to be solvable, and costs are likely to reduce. We conclude that cloud computing is an efficient and cost effective way of simulating radiation transport in a patient, making it attractive as a treatment planning or plan validation tool for any radiotherapy modality.
Organisation CI , STFC
Funding Information
Related Research Object(s):
Licence Information:
Language English (EN)
Type Details URI(s) Local file(s) Year
Journal Article Phys Med Biol (2015). GCE_paper_final.pdf 2015