Room: AAPM ePoster Library
Purpose: Monte Carlo (MC) radiation transport methods are known to be the most precise method for radiation dose calculation. However, the use of phase space files based source terms in our previous studies related to the ARCHER code do not satisfy the needs of a clinical quality assurance (QA) software. This paper reports the integration of virtual source modeling for a GPU-based MC software ARCHER QA for independent radiotherapy dose checking.
Methods: The virtual source is composed of three analytical sources representing the primary photon, scattering photon, and electron contamination, as originally specified by H. Liu et al. The Gaussian distribution is selected as the basis function of the analytical sources, together with extra corrections for the off-axis effect. Jaws and multi-leaf collimators (MLCs) are simply treated as weighted fluence maps in this model. The parameters of the three analytical sources are obtained from both MC simulations of the clinical treatment head and the measurements of the water phantom. The phase space information at the scoring plane (Phsp2) behind jaws and MLCs is generated on the fly and is used as the source terms in subsequent MC dose calculations in a 3D patient CT phantom.
Results: A virtual source model has developed for Varian TrueBeam linac, Elekta machine, and Unity MRIgRT machine. The reconstructed Phsp2 files have been found to agree within 1% with the reference phase space files. Clinical cases are tested and the results show both high speed (GPU card: NVIDIA Titan V) and gamma passing rate (3%/3mm): Breast (13s, 95.2%), Lung (14s, 97.6%) and prostate (30s, 94.7%).
Conclusion: The approach of replacing the large and cumbersome pre-obtained phase space files with the virtual source modeling improves the computing efficiency and machine diversity, making it possible to take advantage of accurate MC methods in independent dosimetry QA.