RSNA 2013 

Abstract Archives of the RSNA, 2013


SSE24-02

An Evolutionary Algorithm for the Optimization of Parameters in Radiation Beam Profile Modeling for the CyberKnife X-band Linear Accelerator

Scientific Formal (Paper) Presentations

Presented on December 2, 2013
Presented as part of SSE24: Physics (Image-guided Radiation Therapy I)

Participants

Matthew Roy Witten PhD, Presenter: Nothing to Disclose
Owen Clancey PhD, Abstract Co-Author: Nothing to Disclose

ABSTRACT

Purpose/Objective(s): To optimize parameters for the modeling of the beam profile of the CyberKnife X-band linear accelerator, and to investigate the dependencies of the parameters on depth and collimator size.Materials/Methods: The measured data for the CyberKnife linear accelerator were acquired using the M3000-RS scanning system (Advanced Radiation Measurements, Inc., Port Saint Lucie, FL) and two 60012 p-type stereotactic dosimtery diodes (PTW, Inc., Freiburg, DE). The scanning system was used to precisely and accurately position the diodes to acquire both depth-dose data (the variation in radiation dose as the beam penetrates the surface of the water and interacts deeper and deeper in it), as well as beam profile data (the variation in dose in the plane perpendicular to the direction of travel of the beam). The beam profile shape is a consequence of the circular collimators, which range in diameter from 5 mm to 60 mm. The dose is at the maximum at the central axis of the beam. The dose is then relatively constant until the beam meets the physical edge of the collimator. Near the edge, the dose precipitously decreases, from nearly 100% of the maximum just inside the field, through 50% at the field edge, and finally to approximately 1-3% far from the field edge, where the dose is primarily attributable to scatter, as there is no primary irradiation outside the collimator. Accurately modeling the penumbra can be difficult, as the shape of the curve used to model the penumbra needs to be finely tuned to reflect the reality of the physical decrease in dose. The model used to calculate the beam profile assumed that there were three Gaussian scattering kernels which contributed to the dose, and the convolution of the kernels in the dose calculation formalism produced an analytical form of the off-center ratio involving the sum of error functions. Associated with each scattering kernel is an amplitude and standard deviation. The amplitudes and standard deviations associated with the scatter kernels were optimized with an evolutionary algorithm, where each individual was represented by a sextuple of real numbers. No a priori assumptions were made regarding the functional dependence of the scatter kernel amplitudes and standard deviations upon these variables.Results: The optimization was able to produce an acceptable agreement, to within 2%, with measurement for the off-center ratio (OCR) values for full range of collimator sizes and at all depths (15 mm, 50 mm, 100 mm, 200 mm, 300 mm). There were depth- and collimator size-dependencies of the scatter kernel amplitudes, and the standard deviations possessed a dependence upon depth.Conclusions: An evolutionary algorithm can be successfully used to optimize scatter kernel parameters in the modeling of the beam profile of the CyberKnife x-band linear accelerator.

Cite This Abstract

Witten, M, Clancey, O, An Evolutionary Algorithm for the Optimization of Parameters in Radiation Beam Profile Modeling for the CyberKnife X-band Linear Accelerator.  Radiological Society of North America 2013 Scientific Assembly and Annual Meeting, December 1 - December 6, 2013 ,Chicago IL. http://archive.rsna.org/2013/13042695.html