Premium
Validation of GPU ‐accelerated superposition–convolution dose computations for the Small Animal Radiation Research Platform
Author(s) -
Cho Nathan,
Tsiamas Panagiotis,
Velarde Esteban,
Tryggestad Erik,
Jacques Robert,
Berbeco Ross,
McNutt Todd,
Kazanzides Peter,
Wong John
Publication year - 2018
Publication title -
medical physics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.473
H-Index - 180
eISSN - 2473-4209
pISSN - 0094-2405
DOI - 10.1002/mp.12862
Subject(s) - monte carlo method , convolution (computer science) , computation , dosimetry , superposition principle , imaging phantom , materials science , beam (structure) , physics , computer science , computational physics , computational science , optics , nuclear medicine , algorithm , mathematics , artificial intelligence , artificial neural network , medicine , statistics , quantum mechanics
Purpose The Small Animal Radiation Research Platform ( SARRP ) has been developed for conformal microirradiation with on‐board cone beam CT ( CBCT ) guidance. The graphics processing unit ( GPU )‐accelerated Superposition–Convolution ( SC ) method for dose computation has been integrated into the treatment planning system ( TPS ) for SARRP . This paper describes the validation of the SC method for the kilovoltage energy by comparing with EBT 2 film measurements and Monte Carlo ( MC ) simulations. Methods MC data were simulated by EGS nrc code with 3 × 10 8 –1.5 × 10 9 histories, while 21 photon energy bins were used to model the 220 kV p x‐rays in the SC method. Various types of phantoms including plastic water, cork, graphite, and aluminum were used to encompass the range of densities of mouse organs. For the comparison, percentage depth dose ( PDD ) of SC , MC , and film measurements were analyzed. Cross beam (x,y) dosimetric profiles of SC and film measurements are also presented. Correction factors ( CF z) to convert SC to MC dose‐to‐medium are derived from the SC and MC simulations in homogeneous phantoms of aluminum and graphite to improve the estimation. Results The SC method produces dose values that are within 5% of film measurements and MC simulations in the flat regions of the profile. The dose is less accurate at the edges, due to factors such as geometric uncertainties of film placement and difference in dose calculation grids. Conclusion The GPU ‐accelerated Superposition–Convolution dose computation method was successfully validated with EBT 2 film measurements and MC calculations. The SC method offers much faster computation speed than MC and provides calculations of both dose‐to‐water in medium and dose‐to‐medium in medium.