Linac Model Research Articles

TU‐D‐352‐05: The Effect of Magnetic Interference On a Coupled MR‐Linac System: 3D Modeling of RF in a Linac

Purpose: In order to provide real‐time Image Guided Radiotherapy (IGRT), the coupling of a 0.2T bi‐planar Magnetic Resonance Imager (MRI) and medical linear accelerator (linac) is proposed. This integration forces the linac to be within the fringe fields of the bi‐planar magnets causing magnetic interference. Appropriate shielding must be designed and optimized in order to reduce the fringe field magnitudes to a point where a clinically useful radiation beam is produced. Method and Materials: A first step to designing shielding is the full 3D radio‐frequency (rf) modeling of the linac waveguide using the Finite Element Method (FEM). The main accelerating cavities were modeled in 3D using COMSOL Multiphysics and compared to the widely used program Poisson Superfish. The full waveguide was then modeled in 3D and the coupling coefficients were determined. Results: A comparison of linac important parameters (i.e. resonant frequency, power dissipation, stored energy, Q value, shunt impedance and transit time factor) were made between the 3D model in COMSOL and that of Poisson Superfish. All values were found to be within 0.8% where the differences are attributed to differences in meshing between the two programs. A linac model consisting of 5 and one half accelerating cavities with 5 coupling cavities was then generated. A 3D instead of a 2D model was used since it is more accurate in determining the coupling coefficients between the accelerating and coupling cavities. Conclusion: This work is the first step to determining the magnetic shielding required in producing a clinically useful radiation beam from a coupled MR‐Linac system. The extracted parameters from the 3D model were found to be less than 1% different from the benchmarked program Poisson Superfish. This 3D model is however inherently more accurate since the full modeling of coupling between accelerating and coupling cavities has been achieved.

TU‐D‐AUD‐04: Comparison of Three Different Source Models for Head Scatter Facto Calculation

Purpose: to evaluate three different source models for the calculation of head scatter factors. Method and Materials: The three different source models evaluated in this work are (1) The three‐source model (Yang et al) (2) The Gaussian analytic model (Jiang et al) (3) The numerical source model (Chui et al). These source models were developed based on different LINAC models. These source models were implemented and their results reproduced. Each of the source models was then applied to the other's measurement in order to cross‐validate accuracy. The effect of the monitor backscattering on the source model was also investigated. Finally, measurement data from a Varian Trilogy machine was used to evaluate each of the source models. The behavior of each model for different SSD data was also evaluated. Results: Measured output factors exhibit ∼1% difference at largest field sizes. The Gaussian analytic model gives 0.2% accuracy when fitting Yang's data, while the three‐source model has problem at large fields with error >15% when attempted on Jiang's data. So was the numerical model. Inclusion of the monitor backscattering effect into the source models significantly improved the agreement. When applying the models to the Trilogy measurement, the Gaussian analytic model gives the best agreement (<0.3%) while the three‐source model has trouble in asymmetric and large symmetric fields again. The numerical source model didn't fit those fields either. While the other two source models gave consistent results for data measured at different SSD, the numerical source model gave large errors (>0.5%). Conclusion: The Gaussian analytic model gave excellent fit to head scatter factor for a variety of LINAC models. Reasonable accuracy can be achieved if monitor backscatter effect were built into the other two source models.Research sponsored by Sun Nuclear Corp.

MO‐E‐AUD‐04: Automated Linear Accelerator Quality Assurance Using a Commercial Cylindrical Calibration Phantom

Purpose: Precise mechanical operation of a linear accelerator is critical for accurate dose delivery. Available quantitative procedures for the linac mechanical quality assurance (QA) are time consuming and therefore conducted on a relatively infrequent basis. We present a method for evaluating the mechanical performance of a linac based on a series of projection portal images of a prototype cylindrical calibration phantom with embedded markers.Method and Materials: We used non‐linear multiobjective optimization of information extracted from the images to determine a number of geometric parameters of interest. The markers detection included modeling the imager response to radiation beams where significantly non‐uniform background was expected. Results: The average results for the parameters of our geometric linac model were: gantry angle deviation 0.066 ± 0.085° (1 SD), gantry sag 0.026 ± 0.02°, imager in‐plane rotation 0.026 ± 0.055°, roll — 0.082 ± 0.16° and pitch −0.9 ± 0.604°, SDD 1489.7 ± 5 mm, SAD 998 .3 ± 1.7 mm, and the imager shift [− 0.66,3.9]±[0.30,1.6] mm. The results were corrected for the phantom center shift relative to the linac rotational center. The average rotational center was . The average couch height and angle variations were 0.15±0.9 mm and 0.154±0.1°, respectively. The image analysis quality was examined by comparing the detected set of marker coordinates to its simulated counterpart for three regions of the phantom image: central, near the edge and the intermediate region (relative to the central line of the cylinder). The upper limit of the mean difference was less than 0.25 mm with the cumulative mean of 0.146 mm and SD of 0.07 mm. The results of the primary optimization of directly detected marker coordinates virtually coincided with their counterparts based on the simulated coordinates for all the geometric parameters of the model.Conclusion: This procedure is accurate and automated, which allows precise mechanic QA to be performed more frequently.Conflict of Interest: partially supported by Varian Medical Systems.

SU-FF-T-43: A Planar Dose Calculation Algorithm for IMRT Quality Assurance

Purpose: To develop a fast algorithm for independent planar dose verification of IMRT plans delivered with an MLC as part of the IMRT quality assurance procedure. Methods and Materials: A fast empirical IMRT planner dose calculation algorithm was implemented. In-air output factors were used to model the source distribution, which is used to calculate the planar fluence at the detector plane based on the projected source integration on the source plane shaped by the combined positions of the jaws and MLC. The planner dose is obtained by convolving the planner fluence with a kernel, which is modified from the published kernel by fitting to standard field profiles at the specified depth. MLC-introduced dosimetric effects, such as leaf transmission, rounded leaf end, and tongue-and-groove effects were accounted for explicitly by empirical fits to measured data. The algorithm was applied to two linac models with completely different head designs: a Varian 2100 C/D with 120-leaf MLC and an Elekta Synergy S with Beam Modulator (40-leaf MLC with no backup jaws). The calculated planner dose distributions were compared with both the MapCHECK measurements and with calculations using a commercial planning system (Pinnacle3, version 7.6) for several clinical cases. Comparison was done using the MapCHECK software in absolute dose mode with 3% dose error and 3-mm distance-to-agreement criteria. Results: The passing rate for most of the cases is above 95% with most of the failing points in the field edges. The calculation time for each field is about 5 sec using a personal laptop computer and is independent of field size. Conclusions: A fast algorithm has been developed for independent planar dose verification of IMRT plans as part of the IMRT quality assurance procedure. The algorithm was applied to two linac models with excellent results. Research sponsored by Sun Nuclear Corp.

TU-D-224C-09: On the Discrepancies Between Monte Carlo Dose Calculations and Measurements for Varian Megavoltage Photon Beams

Significant discrepancies between Monte Carlo dose calculations and measurements for the Varian 18 MV photon beam with a large field size (40 × 40 cm2) were reported by different investigators. The GEPTS Monte Carlo code is used to investigate these discrepancies and a new geometry model (New Model) of the Varian 2100EX linac is implemented. A misrepresentation of some geometric parameters in previous investigations (Old Model) was suggested by Chibani in his AAPM presentation (2004) and later confirmed by the manufacturer. The entrance and exit radii of the primary collimator are 2‐mm larger than previously thought. The primary collimator is also found to be 2‐mm shorter and 1‐mm closer to the target. The New Model also includes the lead shield between the monitor chamber and the Y jaws. A detailed analysis of the phase space data (before and after the primary collimator, after the flattening filter, and at the phantom surface) shows the effect of the changes introduced in the New Model on the beam characteristics(photons and electrons). The main source of the dose discrepancies between measurements and calculations based on the Old Model is the underestimated electron contamination. The photon and electron fluences at isocenter are 4.7% and 30% larger in the New Model in comparison with the Old Model. The flattening filter and lead shield contribute 50% and 10.1% of the total electron contamination at isocenter, respectively. Dose calculations based on the GEPTS implementation of the New Model for the 18 MV open and filtered (2‐mm Pb foil) fields are in excellent agreement with measurements. Local differences are less than 1.5% for depths larger than 1 cm. Monte Carlo simulation of the ionization chamber response shows that difference between calculated and measured percent depth ionizations can be reduced to less than 3% even on the surface.

SU‐FF‐T‐293: Inter‐Linac and Intra‐Linac Variability of X‐Ray and Electron Dose Distributions

Purpose: To determine the maximum expected variation in fluence, differential in energy and position, with an accuracy of 1% for a single linac model, and eventually, combining with Monte Carlo beam models, to develop gold standard beam model benchmarks. Method and Materials: The 6 and 18 MV beams of the Siemens Primus linear accelerator (linac) were tuned through the manufacturer acceptance range in field flatness by adjusting beam energy. Dose profiles were measured at the low, high and average energy in this range for small, medium and large field sizes to quantify the intra‐linac variations caused by energy adjustment. The 6 and 18 MV x‐ray beams, and 6–21 MeV electron beams on three Primus machines, with the same treatment head, were tuned to give the same central axis depth dose curves to better than 0.5% for the largest fields available: 40×40 cm for photons, and 25×25 cm for electrons. Profiles were measured on these linacs to investigate the inter‐linac variations. Results: For the intra‐linac variations, the energy tuning has the largest effect of 7% at 10 cm on beam penetration for small fields. Scattering in the field reduces this effect for larger fields. Wedged‐field profiles are affected more than the open field, due to the dependence of transmission on energy. For inter‐linac variation, profiles measured from three different machines vary more than 2% due to beam tuning differences, lateral displacement of flatteners or scattering foils, and differences in position of components relative to the source. Conclusion: The range of inter‐linac and intra‐linac variability has been measured on a Primus linac and is clinically significant. This data will be used with Monte Carlo treatment head simulation to calculate gold standard beam model benchmarks for the development of tunable beam models of high accuracy and detail in the dose‐critical components of fluence.

Predictability of electron cone ratios with respect to linac make and model.

In the past, the Radiological Physics Center (RPC) has developed standard sets of photon depth‐dose and wedge‐factor data, specific to the make, model, and wedge design of the linear accelerator (linac). In this paper, the RPC extends the same concept to electron‐cone ratios. Since 1987, the RPC has measured and documented cone‐ratio (CR) values during on‐site dosimetry review visits to institutions participating in National Cancer Institute cooperative clinical trials. Data have been collected for approximately 500 electron beams from a wide spectrum of linac models. The analysis presented in this paper indicates that CR values are predictable to 2% to 3% (two standard deviations) for a given make and model of linac with a few exceptions. The analysis also revealed some other interesting systematics. For some models, such as the Varian Clinac 2500 and the Elekta/Philips SL18, SL20, and SL25, CR values were nearly identical for cone sizes 15 cm×15 cm (or 14 cm×14 cm) and 20 cm×20 cm across the range of available energies. Certain models of the same make of linac, such as the Mevatron MD, KD, and 6700 series models or the Clinac 2100 and 2300 models, exhibited indistinguishable CRs. Irrespective of linac model, two consistent general trends were observed: namely, an increase in CR value with incident beam energy for cone sizes smaller than 10 cm×10 cm and a decrease with energy for cone sizes larger than 10 cm×10 cm. These data are valuable to the RPC as a quality assurance remote‐monitoring tool to identify potential dosimetry errors. The physics community will also find the data useful in several ways: as a redundant check for clinical values in use, to validate the values measured during commissioning of new machines or to ensure consistency of values measured during annual quality assurance procedures.PACS number(s): 87.54.–n, 87.53.–j

Predictability of electron cone ratios with respect to linac make and model

In the past, the Radiological Physics Center (RPC) has developed standard sets of photon depth-dose and wedge-factor data, specific to the make, model, and wedge design of the linear accelerator (linac). In this paper, the RPC extends the same concept to electron-cone ratios. Since 1987, the RPC has measured and documented cone-ratio (CR) values during on-site dosimetry review visits to institutions participating in National Cancer Institute cooperative clinical trials. Data have been collected for approximately 500 electron beams from a wide spectrum of linac models. The analysis presented in this paper indicates that CR values are predictable to 2% to 3% (two standard deviations) for a given make and model of linac with a few exceptions. The analysis also revealed some other interesting systematics. For some models, such as the Varian Clinac 2500 and the Elekta/Philips SL18, SL20, and SL25, CR values were nearly identical for cone sizes (or ) and across the range of available energies. Certain models of the same make of linac, such as the Mevatron MD, KD, and 6700 series models or the Clinac 2100 and 2300 models, exhibited indistinguishable CRs. Irrespective of linac model, two consistent general trends were observed: namely, an increase in CR value with incident beam energy for cone sizes smaller than and a decrease with energy for cone sizes larger than . These data are valuable to the RPC as a quality assurance remote-monitoring tool to identify potential dosimetry errors. The physics community will also find the data useful in several ways: as a redundant check for clinical values in use, to validate the values measured during commissioning of new machines or to ensure consistency of values measured during annual quality assurance procedures. PACS number(s): 87.54.–n, 87.53.–j

Comparison of characteristics of photon and electron beams generated by Philips/Elekta and Varian linear accelerators.

Commissioning data of the common energy of 6 MV photon beams are analyzed for seven Philips/ Elekta linacs of five different models: SL 75/5, SL 15, SL 25, SLi Precise, and SLi. The observed beam quality %dd(10)x of 67.7% +/- 0.3 agrees with the manufacturer's specification of 67.5% +/- 2% but with a much smaller variation, suggesting the possibility of a more accurate beam quality guideline for acceptance commissioning of these linac models. Field size factors are less uniform, especially for the biggest field size of 40 x 40 cm2, but all agree within 1.1% between the five models. These 6 MV photon beams are also shown to be 0.4 MV harder than the Varian 2100C model. For commissioning electron beams generated by the Philips/Elekta models SL 15, SLi Precise and SLi, guidelines can be found from this work for the following parameters: actual field size at 100 SSD (up to +/- 0.6%) (different from manufacturer specified values at 95 SSD), and virtual source distance (up to +/- 0.7%). Significant differences are seen for R50 (up to 3.7 mm), sigma-theta-x, effective source distance, and cone size factor. However, except for R50 where Varian 2100C datasets are not available for comparison, the differences between the different Philips/Elekta models are comparable to those between the same Varian 2100C model linacs measured by us and Watts.

Frequency map analysis of a three-dimensional particle in the core model of a high intensity linac

We consider the dynamical properties of a particle-core model for a uniformly filled triaxial ellipsoid in a periodic lattice of a high intensity linac. The mismatched oscillation modes are analytically computed in the smooth approximation and are compared with the numerical results of a tracking program. The study of the phase space in the mismatched case is performed by the frequency map analysis. In particular, we can analyze the effect of the nonlinear resonances between the envelope modes and the single particle sincrobetatron frequencies. A chaoticity criterion based on the frequency map analysis allows one to compute the stability region around the beam core. An estimate of the transport and its enhancement due to mismatch is provided by tracking orbits at the border of the stability region.

Phase rotation channel design with induction linacs

Two models of the phase rotation channels of the neutrino factory were designed with a solenoid transport system and the induction linac acceleration system. One model has the one stage induction linac channel, and another model has two stages of the induction linac channels and an absorber and drift space in between them. Major figure of merits of the channel are the particle ratio of muons over primary protons, the momentum spread and the muon polarization correlation to the arrival time at the end of the phase rotation channel. The μ/proton ratios were 0.25 and 0.13 for the single stage and the double stage induction linac channels, respectively. The effective muon polarization, 〈helicity2〉, was 0.16 and the muon polarization correlation was 0.23/(100ns) for the single stage induction linac model.

Stabilizing characteristics of post-couplers

A model of Alvarez linac equipped with post-couplers was constructed. The field distribution and the resonant frequency of the accelerating mode were measured to study its properties against perturbations. The experimental results were compared with an analysis using equivalent circuits. Post-couplers have an effect to distort the field distribution as well as to stabilize it depending upon the length of the posts.

Experimental Investigation of Heating Phenomena in Linac Mechanical Interfaces from RF Field Penetration

In a high duty-factor, high-current, drift-tube linear accelerator, a critical interface exists between the drift-tube stem and the tank wall. This interface must provide vacuum integrity and RF continuity, while simultaneously allowing alignment flexibility. Because of past difficulties with RF heating of vacuum bellows and RF joints encountered by others, a paucity of available information, and the high reliability requirement for the Fusion Materials Irradiation Test (FMIT) accelerator, a program was initiated to study the problem. Because RF heating is the common failure mode, an attempt was made to find a correlation between the drift-tubestem/linac-tank interface geometry and RF field penetration from the tank into the interface region. Experiments were performed at 80 MHz on an RF structure designed to simulate the conditions to which a drift-tube stem and vacuum bellows are exposed in a drift-tube linac. Additional testing was performed on a 367-MHz model of the FMIT prototype drift-tube linac. Experimental results, and a method to predict excessive RF heating, is presented. An experimentally tested solution to the problem is discussed.