While TOPAS provides accurate dose calculations, typical simulations of patient treatments take several hours, too slow for intensity modulated proton therapy (IMPT) treatment planning, extensive 4D studies or adaptive treatment strategies. Fast Monte Carlo systems are necessary for a simulation within seconds or minutes. For this purpose, we have developed a GPU based fast Monte Carlo tool, gPMC, for proton therapy in collaboration with UT Southwestern, which focuses on dose and LET calculations only.

The speed-up produced by using GPU architectures for MC simulations comes from the nature of the problem to study and the design of the hardware. Protons (and other particles) in proton therapy simulations of dose in the patient can be considered independent of each other. This allows the massive parallelization of the problem. A large number of particles may be simulated at the same time as GPU threads, drastically reducing the total simulation time from hours to seconds/minutes.

  • IMPT treatment planning studies: inverse treatment planning optimization can be understood as the process of deciding what particular combination of weighted inputs best fulfills a certain set of conditions. As such, the process relies on the creation of a set of inputs to be optimized. Often times, the set of inputs represents physical dose and are created with analytical dose calculation (ADC) algorithms. However, these analytical methods lack the accuracy to simulate media with high heterogeneities and/or low density. More importantly, they lack the flexibility of MC simulations to score virtually any quantity. As of August 2017, no analytical algorithm has been published to precisely calculate LET distributions. We have employed gPMC to generate the set of inputs by simulating each spot of a non-optimized spot map individually when analytical calculations are not applicable. Specific applications in which we have employed this are: plans for 4D treatment planning of non-small cell lung cancer, LET re-optimization and end-of-track re-optimization.
  • 4D studies: The simulation of the complete breathing period of a patient is very time consuming. It usually consists of the simulation of a plan on a 4DCT (usually 10 phases). This is roughly equivalent to simulating 10 individual patients, depending on the desired statistical precision. ADC are not recommended on lung due to the bone/soft tissue interfaces on the chest wall and the big multiple Coulomb scattering effect in the lung tissue. GPU MC is therefore the best choice for these studies.
  • Adaptive proton therapy:  IMPT is very sensitive to treatment geometry changes. patient mispositioning, weight loss or deformations of the soft tissues of a patient distorts the dose distribution that a plan was initially designed to deliver. GPU MC calculations allows the fast and accurate validity assessment of a plan in the current treatment geometry obtained from some imaging technique, such as cone-beam CT (CBCT). GPU MC would therefore be a corner stone of an accurate adaptive proton therapy protocol.

Key MGH personnel involved

  • Harald Paganetti, PhD
  • Jan Schuemann, PhD  
  • Jungwook Shin, PhD
  • Pablo Botas


  • Steve Jiang (UT Southwestern, UTSW)
  • Xun Jia (UT Southwestern, UTSW)
  • Nan Qin (UT Southwestern, UTSW)


  1. Jia X; Schuemann J; Paganetti H and Jiang SB: GPU-based fast Monte Carlo dose calculation for proton therapy. Physics in Medicine and Biology 2012 57: 7783-7798
  2. Qin N, Botas P, Giantsoudi D, Schuemann J, Tian Z, Jiang SB, Paganetti H, Jia X. Recent developments and comprehensive evaluations of a GPU-based Monte Carlo package for proton therapy. Phys Med Biol. 2016; 61(20):7347-7362. PMID: 27694712.
  3. Unkelbach J, Botas P, Giantsoudi D, Gorissen BL, Paganetti H. Reoptimization of Intensity Modulated Proton Therapy Plans Based on Linear Energy Transfer. Int J Radiation Oncology Biol Phys. 2016; 96(5):1097-1106. PMID: 27869082.