TU‐FF‐A1‐04: Determining Parameters for a Multiple‐Source Model of a Linear Accelerator Using Optimization Techniques

Abstract

Purpose: To determine the parameters of a multiple‐source model for an arbitrary linear accelerator using optimization methods. Method and Materials: A multiple‐source model describing the energy fluence output of a linear accelerator was developed in this study. A point source modeled radiation from the target, a finite‐size source all extra‐focal radiation, and an electron source contaminant particles. The parameters determined were the mean energy curve (for off‐axis softening), intensity profile curve (for non‐uniform photon energy fluence), electron source values, extra‐focal source size, energy, and intensity. The parameters were optimized by minimizing the gamma error between the dose calculation results and the beam data measurements by applying a non‐linear optimization technique not requiring gradient information. The dose was calculated by an algorithm based on superposition/convolution of Monte Carlo determined scatter kernels. The beam data measurements required were depth dose curves, lateral profiles, and diagonal profiles for multiple field sizes. The model requires minimal data about the internal dimensions and construction of the accelerator head. Results: The method was applied to 231 realistic data sets of varying quality and consistency for Elekta, Siemens and Varian accelerators. The gamma error (1%, 3 mm) for an average optimized model was lower than 1.0 for 98% of the measurement points. Typical duration of the optimization to derive the model parameters was 5–15 minutes. In cases where the measurements contained inconsistencies, the resulting gamma errors were significant, which indicates that the method could be useful in quality assurance of measurement data. Conclusion: This study demonstrated that the parameters for a multiple‐source model can be determined in an efficient and stable manner using optimization methods. The model is applicable to an arbitrary accelerator and has clinically acceptable accuracy and execution time. Conflict of Interest: This work was supported by Varian Medical Systems.

Report of the AAPM Task Group No. 105: Issues associated with clinical implementation of Monte Carlo‐based photon and electron external beam treatment planning

“…Fluence distributions in measurement-driven models may be developed starting with analytical models whose parameters are optimized based on minimization of the differences between calculations and measurements. 100,125,[127][128][129][130][131] Information for the model can be deconvolved from measured data. 100,127 Similar methods have been reported for beammodel specification using conventional dose calculation algorithms with good agreement established between calculations and measurements in water phantoms.…”

Development of a Monte Carlo multiple source model for inclusion in a dose calculation auditing tool

“…This calculation method has the advantage of being independent of the complexities within the treatment head . Parameters to the analytical models describing the multiple sources are derived by minimizing the difference between simulated and measured data . Differences in accelerator design between manufacturers lead to dosimetric differences that necessitate separate Monte Carlo models for each manufacturer in order to achieve suitable model accuracy.…”

“…15 Parameters to the analytical models describing the multiple sources are derived by minimizing the difference between simulated and measured data. [21][22][23][24][25] Differences in accelerator design between manufacturers lead to dosimetric differences that necessitate separate Monte Carlo models for each manufacturer in order to achieve suitable model accuracy. The developed multiple source models for Elekta 6 MV and 10 MV beams will be integrated with previously developed Varian 6 MV and 10 MV models 26 and eventually Varian FFF 6 MV and FFF 10 MV models into a dose calculation tool that will be capable of accurately calculating doses for over 90% of all beams monitored by IROC-H.…”

A modified simulated annealing algorithm for parameter determination for a hybrid virtual model

“…While reproduced PDDs show a very good agreement with measurements, the match in energy spectra against independently measured spectra or those calculated using Monte Carlo and incorporating the actual treatment head, is not so good. Issues that must be addressed before using a virtual source model are the optimization methods that are to be used in order to define the physical characteristics of the sources (energy spectrum, position, shape in the case of extended sources, etc), and the set of measurements necessary to determine the model (Siljamki et al 2005). Several prescriptions have been published on how to do this (see for example Jiang et al (2000)) in the framework of Monte Carlo TPSs commissioning.…”

“…For a sufficiently large set of measurements there should be only one configuration of virtual sources that reproduces all of them. Recently proposed virtual source models have considered central axis PDDs together with horizontal and oblique measurements in order to define virtual source models (Siljamki et al 2005). From a practical point of view, it is of course desirable to keep the set of measurements necessary to determine the source model to a minimum.…”