Encrypted login | home

Program Information

Efficient Monte Carlo Track Structure Simulations for Nanodosimetric Applications

J Ramos-Mendez

J Ramos-Mendez1*, J Schuemann2 , A McNamara2 , J Perl3 , K Held2 , H Paganetti2 , B Faddegon1 , (1) University of California San Francisco, San Francisco, CA, (2) Massachusetts General Hospital, Boston, MA, (3) Stanford Linear Accelerator Center, Menlo Park, CA


SU-K-205-7 (Sunday, July 30, 2017) 4:00 PM - 6:00 PM Room: 205

Purpose: To reduce computation time of the Geant4-DNA Monte Carlo track structure code for the estimation of radiation induced damage to DNA.

Methods: In this study a double strand break (DSB) occurs when 2 or more single strand breaks (SSBs) are on opposite DNA strands and separated by up to 10 base pairs. Using particle splitting for variance reduction produces more tracks within the same history. Proper counting of DSB frequency requires keeping the SSBs from different split particles separate. Flagged uniform particle splitting was implemented in TOPAS-nBio (which wraps and extends the Geant4/Geant4-DNA Monte Carlo toolkit) to enhance the electrons generated in the first ionization event of secondary electrons by assigning an identification number, inherited by all of its progeny. This flag was used later to reclassify the split particles in the estimation of: 1) the mean M₁ of the ionization cluster size distributions; 2) the single strand breaks (SSB) or DSB classified with DBSCAN; and 3) the SSB and DSB classified with a geometry-based algorithm. For these three endpoints the accuracy and computational efficiency relative to simulations without any splitting techniques were evaluated for protons and carbon ions, for energies ranging from 0.5-20 MeV and 1-20 MeV/u, respectively.

Results: The relative computational efficiency increased up to a factor of about 70 for protons and 50 for carbon ions depending on the complexity of the geometries used to score M₁, SSB or DSB. The efficiency gain was largest when scoring M₁. In all cases, the accuracy was within 2%, i.e. within 1 standard deviation of the statistical uncertainty, compared to the reference simulations.

Conclusion: It was shown that commonly-used variance reduction techniques of condensed history Monte Carlo codes can be used in track structure codes, resulting in a significant improvement of the computational efficiency without compromising accuracy.

Contact Email: