Login / Signup

Development and evaluation of dose calculation algorithm with a combination of Monte Carlo and point-kernel methods for boron neutron capture therapy.

Mai NojiriTakushi TakataNaonori HuYoshinori SakuraiMinoru SuzukiHiroki Tanaka
Published in: Biomedical physics & engineering express (2023)
We developed a 'hybrid algorithm' that combines the Monte Carlo (MC) and point-kernel methods for fast dose calculation in boron neutron capture therapy. The objectives of this study were to experimentally verify the hybrid algorithm and to verify the calculation accuracy and time of a 'complementary approach' adopting both the hybrid algorithm and the full-energy MC method. In the latter verification, the results were compared with those obtained using the full-energy MC method alone. In the hybrid algorithm, the moderation process of neutrons is simulated using only the MC method, and the thermalization process is modeled as a kernel. The thermal neutron fluxes calculated using only this algorithm were compared with those measured in a cubic phantom. In addition, a complementary approach was used for dose calculation in a geometry simulating the head region, and its computation time and accuracy were verified. The experimental verification indicated that the thermal neutron fluxes calculated using only the hybrid algorithm reproduced the measured values at depths exceeding a few centimeters, whereas they overestimated those at shallower depths. Compared with the calculation using only the full-energy MC method, the complementary approach reduced the computation time by approximately half, maintaining nearly same accuracy. When focusing on the calculation only using the hybrid algorithm only for the boron dose attributed to the reaction of thermal neutrons, the computation time was expected to reduce by 95% compared with the calculation using only the full-energy MC method. In conclusion, modeling the thermalization process as a kernel was effective for reducing the computation time.
Keyphrases
  • monte carlo
  • machine learning
  • deep learning
  • neural network
  • magnetic resonance imaging
  • computed tomography
  • stem cells
  • smoking cessation