WE‐F‐105‐03: Development of GPMC V2.0, a GPU‐Based Monte Carlo Dose Calculation Package for Proton Radiotherapy

X. Jia, J. Schuemann, H. Paganetti, S. Jiang

Research output: Contribution to journalArticle

2 Citations (Scopus)

Abstract

Purpose: Monte Carlo (MC) simulation is considered the most accurate method for proton therapy dose calculations. To improve its efficiency, we have previously developed a MC proton dose calculation package, gPMC, on a graphics‐processing unit (GPU) platform. This abstract reports our recent progress towards developing a new version gPMC v2.0, which aims at improving its clinical usability and evaluating it using proton therapy treatment plans. Methods: gPMC v2.0 supports loading source particles from a phase space file. Source particles are first placed to different energy bins according to their energies. Simulation is conducted to transport particles in each bin sequentially. This enforces that particles transported at the same time have similar energies and hence similar computational thread lengths, ensuring computational efficiency in parallel processing. Realistic treatment geometry is supported. A multi‐dose‐counter technique is developed, where multiple dose counters are allocated. Each dose deposition event is randomly assigned to one of the counters. This alleviates memory‐writing conflict problem, when multiple threads attempts to deposit dose to the same voxel. The final dose distribution and statistical uncertainty are obtained based on results at those counters. gPMC v2.0 also supports scoring dose‐to‐water. We have tested the new version in head‐and‐neck patient cases. Results: Computation time is ∼5 seconds for a typical patient case, where 2.1 million protons in a phase space file are simulated. It takes additional ∼8 seconds to load the phase space file. The dose distribution agrees well with that from TOPAS/Geant4. Inside a region of interest defined by the 50% iso‐dose line, gamma passing rate is ∼98% (3mm/3%) and ∼95% (2mm/2%) compared with TOPAS/Geant4 Result. The average relative uncertainty is 1.1%. Conclusion: gPMC has achieved accuracy and efficiency well beyond what was previously accomplished. It supports the recalculation of proton therapy treatment plans, making it feasible for routine clinical applications. This work is supported in part by the University of California Lab Fees Research Program and in part by R01 CA140735.

Original languageEnglish (US)
Pages (from-to)498
Number of pages1
JournalMedical Physics
Volume40
Issue number6
DOIs
StatePublished - 2013

Fingerprint

Proton Therapy
Protons
Radiotherapy
Uncertainty
Statistical Distributions
Fees and Charges
Therapeutics
Research

ASJC Scopus subject areas

  • Biophysics
  • Radiology Nuclear Medicine and imaging

Cite this

WE‐F‐105‐03 : Development of GPMC V2.0, a GPU‐Based Monte Carlo Dose Calculation Package for Proton Radiotherapy. / Jia, X.; Schuemann, J.; Paganetti, H.; Jiang, S.

In: Medical Physics, Vol. 40, No. 6, 2013, p. 498.

Research output: Contribution to journalArticle

@article{2ad1a7e7a98643ec80d5a825698e6d64,
title = "WE‐F‐105‐03: Development of GPMC V2.0, a GPU‐Based Monte Carlo Dose Calculation Package for Proton Radiotherapy",
abstract = "Purpose: Monte Carlo (MC) simulation is considered the most accurate method for proton therapy dose calculations. To improve its efficiency, we have previously developed a MC proton dose calculation package, gPMC, on a graphics‐processing unit (GPU) platform. This abstract reports our recent progress towards developing a new version gPMC v2.0, which aims at improving its clinical usability and evaluating it using proton therapy treatment plans. Methods: gPMC v2.0 supports loading source particles from a phase space file. Source particles are first placed to different energy bins according to their energies. Simulation is conducted to transport particles in each bin sequentially. This enforces that particles transported at the same time have similar energies and hence similar computational thread lengths, ensuring computational efficiency in parallel processing. Realistic treatment geometry is supported. A multi‐dose‐counter technique is developed, where multiple dose counters are allocated. Each dose deposition event is randomly assigned to one of the counters. This alleviates memory‐writing conflict problem, when multiple threads attempts to deposit dose to the same voxel. The final dose distribution and statistical uncertainty are obtained based on results at those counters. gPMC v2.0 also supports scoring dose‐to‐water. We have tested the new version in head‐and‐neck patient cases. Results: Computation time is ∼5 seconds for a typical patient case, where 2.1 million protons in a phase space file are simulated. It takes additional ∼8 seconds to load the phase space file. The dose distribution agrees well with that from TOPAS/Geant4. Inside a region of interest defined by the 50{\%} iso‐dose line, gamma passing rate is ∼98{\%} (3mm/3{\%}) and ∼95{\%} (2mm/2{\%}) compared with TOPAS/Geant4 Result. The average relative uncertainty is 1.1{\%}. Conclusion: gPMC has achieved accuracy and efficiency well beyond what was previously accomplished. It supports the recalculation of proton therapy treatment plans, making it feasible for routine clinical applications. This work is supported in part by the University of California Lab Fees Research Program and in part by R01 CA140735.",
author = "X. Jia and J. Schuemann and H. Paganetti and S. Jiang",
year = "2013",
doi = "10.1118/1.4815619",
language = "English (US)",
volume = "40",
pages = "498",
journal = "Medical Physics",
issn = "0094-2405",
publisher = "AAPM - American Association of Physicists in Medicine",
number = "6",

}

TY - JOUR

T1 - WE‐F‐105‐03

T2 - Development of GPMC V2.0, a GPU‐Based Monte Carlo Dose Calculation Package for Proton Radiotherapy

AU - Jia, X.

AU - Schuemann, J.

AU - Paganetti, H.

AU - Jiang, S.

PY - 2013

Y1 - 2013

N2 - Purpose: Monte Carlo (MC) simulation is considered the most accurate method for proton therapy dose calculations. To improve its efficiency, we have previously developed a MC proton dose calculation package, gPMC, on a graphics‐processing unit (GPU) platform. This abstract reports our recent progress towards developing a new version gPMC v2.0, which aims at improving its clinical usability and evaluating it using proton therapy treatment plans. Methods: gPMC v2.0 supports loading source particles from a phase space file. Source particles are first placed to different energy bins according to their energies. Simulation is conducted to transport particles in each bin sequentially. This enforces that particles transported at the same time have similar energies and hence similar computational thread lengths, ensuring computational efficiency in parallel processing. Realistic treatment geometry is supported. A multi‐dose‐counter technique is developed, where multiple dose counters are allocated. Each dose deposition event is randomly assigned to one of the counters. This alleviates memory‐writing conflict problem, when multiple threads attempts to deposit dose to the same voxel. The final dose distribution and statistical uncertainty are obtained based on results at those counters. gPMC v2.0 also supports scoring dose‐to‐water. We have tested the new version in head‐and‐neck patient cases. Results: Computation time is ∼5 seconds for a typical patient case, where 2.1 million protons in a phase space file are simulated. It takes additional ∼8 seconds to load the phase space file. The dose distribution agrees well with that from TOPAS/Geant4. Inside a region of interest defined by the 50% iso‐dose line, gamma passing rate is ∼98% (3mm/3%) and ∼95% (2mm/2%) compared with TOPAS/Geant4 Result. The average relative uncertainty is 1.1%. Conclusion: gPMC has achieved accuracy and efficiency well beyond what was previously accomplished. It supports the recalculation of proton therapy treatment plans, making it feasible for routine clinical applications. This work is supported in part by the University of California Lab Fees Research Program and in part by R01 CA140735.

AB - Purpose: Monte Carlo (MC) simulation is considered the most accurate method for proton therapy dose calculations. To improve its efficiency, we have previously developed a MC proton dose calculation package, gPMC, on a graphics‐processing unit (GPU) platform. This abstract reports our recent progress towards developing a new version gPMC v2.0, which aims at improving its clinical usability and evaluating it using proton therapy treatment plans. Methods: gPMC v2.0 supports loading source particles from a phase space file. Source particles are first placed to different energy bins according to their energies. Simulation is conducted to transport particles in each bin sequentially. This enforces that particles transported at the same time have similar energies and hence similar computational thread lengths, ensuring computational efficiency in parallel processing. Realistic treatment geometry is supported. A multi‐dose‐counter technique is developed, where multiple dose counters are allocated. Each dose deposition event is randomly assigned to one of the counters. This alleviates memory‐writing conflict problem, when multiple threads attempts to deposit dose to the same voxel. The final dose distribution and statistical uncertainty are obtained based on results at those counters. gPMC v2.0 also supports scoring dose‐to‐water. We have tested the new version in head‐and‐neck patient cases. Results: Computation time is ∼5 seconds for a typical patient case, where 2.1 million protons in a phase space file are simulated. It takes additional ∼8 seconds to load the phase space file. The dose distribution agrees well with that from TOPAS/Geant4. Inside a region of interest defined by the 50% iso‐dose line, gamma passing rate is ∼98% (3mm/3%) and ∼95% (2mm/2%) compared with TOPAS/Geant4 Result. The average relative uncertainty is 1.1%. Conclusion: gPMC has achieved accuracy and efficiency well beyond what was previously accomplished. It supports the recalculation of proton therapy treatment plans, making it feasible for routine clinical applications. This work is supported in part by the University of California Lab Fees Research Program and in part by R01 CA140735.

UR - http://www.scopus.com/inward/record.url?scp=84987745766&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84987745766&partnerID=8YFLogxK

U2 - 10.1118/1.4815619

DO - 10.1118/1.4815619

M3 - Article

AN - SCOPUS:84987745766

VL - 40

SP - 498

JO - Medical Physics

JF - Medical Physics

SN - 0094-2405

IS - 6

ER -