Developing a tri-hybrid algorithm to predict patient x-ray scatter into planar imaging detectors for therapeutic photon beams
Loading...
Date
2020
Authors
Guo, Kaiming
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
In vivo dosimetry via transmission imaging of the therapy beam can verify the intended treatment plan was delivered to the patient. However, the EPID (electronic portal imaging device) transmission images are contaminated with patient-generated scattered photons. If this component can be accurately estimated, its effect can be removed and therefore the resulting in vivo patient dose estimate will be more accurate. This thesis presents the development of a ‘tri-hybrid’ (TH) algorithm to provide accurate estimates of patient-generated photon scatter at the EPID.
The TH method combines three approaches: 1) Analytical methods to solve exactly for singly-scattered photon fluence. 2) For multiply scattered photon fluence, a modified hybrid Monte Carlo (MC) simulation method was applied, using only a few thousand histories. From each second and higher-order interaction site in the MC simulation, energy fluence entering all pixels of the EPID scoring plane was calculated using analytical methods. 3) For the bremsstrahlung and positron annihilation component, a convolution/superposition approach was employed using pre-generated pencil beam scatter kernels superposed on the incident fluence.
Since no experimental measurement method is available to directly confirm the separate subcomponents of photon scatter, a Monte Carlo simulation tool was developed to separately score them. This tool was used as the ‘gold standard’ for the development and validation of the TH method.
The TH-predicted total patient-scattered photon fluence entering the EPID, as well its energy spectra, are compared with full Monte Carlo simulation (EGSnrc) for validation. A variety of phantoms are tested, including simple slab and anthropomorphic CT, as well as monoenergetic and polyenergetic beams with different field sizes.
For these tests, the proposed TH method was demonstrated to be in good agreement with full Monte Carlo simulation, generally within 1%. Parameters of the TH method were optimized to maintain an accuracy of <2% while improving execution speed. The optimized TH method takes as little as ~70 seconds to execute on a single (non-parallel) CPU, while full MC simulations took over 30 hours. It is concluded that this patient-generated scattered photon fluence prediction algorithm is relatively fast and accurate and is suitable for implementation into clinical in vivo dosimetry approaches.
Description
Keywords
Monte Carlo simulation, scatter estimation, in vivo dosimetry, EPID dosimetry