Film‐based dose validation of Monte Carlo algorithm for Cyberknife system with a CIRS thorax phantom

Abstract Monte Carlo (MC) simulation, as the most accurate dose calculation algorithm, is available in the MultiPlan treatment planning system for Cyberknife. The main purpose of this work was to perform experiments to thoroughly investigate the accuracy of the MC dose calculation algorithm. Besides the basic MC beam commissioning, two test scenarios were designed. First, single beam tests were performed with a solid water phantom to verify the MC source model in simple geometry. Then, a lung treatment plan on a CIRS thorax phantom was created to mimic the clinical patient treatment. The plan was optimized and calculated using ray tracing (RT) algorithm and then recalculated using MC algorithm. Measurements were performed in both a homogeneous phantom and a heterogeneous phantom (CIRS). Ion‐chamber and radiochromic film were used to obtain absolute point dose and dose distributions. Ion‐chamber results showed that the differences between measured and MC calculated dose were within 3% for all tests. On the film measurements, MC calculation results showed good agreements with the measured dose for all single beam tests. As for the lung case, the gamma passing rate between measured and MC calculated dose was 98.31% and 97.28% for homogeneous and heterogeneous situation, respectively, using 3%/2 mm criteria. However, RT algorithm failed with the passing rate of 79.25% (3%/2 mm) for heterogeneous situation. These results demonstrated that MC dose calculation algorithm in the Multiplan system is accurate enough for patient dose calculation. It is strongly recommended to use MC algorithm in heterogeneous media.

For Cyberknife (Accuary Inc., Sunnyvale, CA, USA), the TPS (Multiplan V4.6, Accuray Inc.) incorporated two dose algorithms: ray tracing (RT) and Monte Carlo (MC). The RT dose algorithm utilizes the beam data measured in a water tank and the tissue heterogeneity is corrected only along the direction of the primary photons. Heterogeneity effects on scatter dose are not considered in this algorithm.
The MC algorithm simulates radiation interactions with tissues and takes into account the lateral electronic disequilibrium. It is considered to be the most accurate algorithm especially for heterogeneous tissues. 3 4 But this commissioning was performed for their own dual-source model built previously. 5 Dechambre et al.
once proposed a novel method for commissioning the MC dose calculation algorithm of Multiplan using a cylindrical 3D-array with variable density inserts. 6 However, the diode spacing of these arrays (0.5-1.0 cm) is generally too large for the small field of Cyberknife, thus may not provide sufficient information for dose verification.
Currently, radiochromic film measurement is widely accepted for CyberKnife quality assurance (QA). 7,8 In 2008, Wilcox et al. used EBT film to measure dose in heterogeneous slab phantoms for single beams of Cyberknife and compared to dose calculated with both ray tracing and Monte Carlo algorithms. 9 The limitation is that geometry of the slab phantom is obviously much simpler than real patient. Therefore, the effect of geometric heterogeneity on dose was not fully accounted and the relationship between dose distribution and organ location could not be explored. Many studies have been reported to perform dose validation with anthropomorphic phantoms for linear accelerators, 10 However, film measurements of the two studies were only focused on PTV region. Thus, dose delivered to normal tissues was not evaluated whereas this is one of the important factors need to be considered in radiotherapy. To address these issues, a thorough experimental validation for MC dose calculation of Cyberknife is needed. But to our knowledge, no detailed report of such work has been published so far.
In this study, a comprehensive validation procedure of MC algorithm for Cyberknife was established. First, the MC source model established in Multiplan system was validated through a series of measured data required by the commissioning procedure 16,17 together with single beam tests performed in a solid water phantom.
Then, a lung case test was carried out to investigate the accuracy of MC algorithm in heterogeneous situation, with a commercially available CIRS thorax phantom (CIRS Inc., Norfolk, VA, USA). Therefore, the main purpose of this work was to thoroughly investigate the

2.A.2 | Measurement data required
As mentioned earlier, a set of measurement data are required as an input to the beam commissioning procedure. These data include central percent dose depth (PDD), OCR, TPR, and OF. PDD measurements were performed in water for the largest collimator (60 mm) at 800 mm source-to-surface distance (SSD). PTW 60017 diode detector was used for the measurements. TPR and OCR beam data were measured for all collimators at a constant source-to-axis distance (SAD) of 800 mm. The TPR data were normalized to the value at the depth of 15 mm. For the OCR measurements, beam data were calculated from sets of orthogonal scans using the IBA scanning system. Different from the commissioning for ray tracing, OF measurement for MC commissioning was performed in air rather than water.
Birdcage with diode detector was used for this measurement.
For quantitative analysis, the root mean square (RMS) of the difference between measured and MC calculated OCR/TPR was used for all collimators. | 143 was based on four fiducials within the phantom. Initial calculation was run using RT algorithm, and then, recalculated with MC algorithm. The measured results were compared with calculated data.

2.B | Single beam tests
Before film measurements, a dose calibration was made range from 0.5 to 10 Gy using the 60 mm collimator. The films were placed perpendicular to the central beam axis. All films irradiated in this work were scanned 24 h after exposure using an Epson scanner.
Then a calibration curve was obtained using a film analysis software RIT 113 (Radiological Imaging Technology Inc.).

2.C.1 | CT images acquisition
For the lung case, a CIRS thorax phantom 002LFC was used to evaluate the accuracy of MC dose calculation in heterogeneous situation ( Fig. 1). This phantom is elliptical in shape (30 cm wide 9 30 cm long 9 20 cm thick) to represent average patient size. Three kinds of materials are contained in this phantom to emulate different tissues in human body, including soft tissue, lung, and bone. For the purpose of target localization, four fiducials were implanted into the phantom in a noncoplanar way. The phantom was scanned using Philips Brilliance CT Big Bore with the same protocol for patients (tube voltage: 120 kVp, tube current: 395 mAs, slice thickness: 1 mm). Then the reconstructed CT images were imported into Multiplan system.

2.C.2 | Treatment planning
A treatment plan was created based on the CIRS phantom images to mimic a clinical case of lung cancer. As shown in Fig. 1, a cylinder with diameter of 3 cm and length of 5.5 cm was drawn as PTV inside the right lung. Bilateral lung and spinal cord were contoured as organs at risk (OARs). To compare point dose from calculation and measurement, active volume of the chamber was also contoured. Fiducial tracking was selected as the tracking method, which uses the gold markers preimplanted in or around the target as reference for tracking. 12.5 and 30 mm fixed collimators were used in this plan. And the prescription was 48 Gy in 4 fractions according to RTOG 0915. 18 The plan was calculated and optimized using RT algorithm. Then it was recalculated using MC algorithm in high resolution with uncertainty of 1%. Thus, the beam sets and monitor units of the two algorithms for each case were identical.

2.C.3 | Measurement in homogeneous phantom
Before performing measurements in the heterogeneous CIRS phantom, tests were carried out with homogeneous phantom to investigate to what extent the calculation accuracy can achieve for MC algorithm when no heterogeneity issue involved. QA plans were created by registering the original lung plan to a solid water phantom CT images. The QA plans were calculated using MC and RT algorithm, respectively, and compared with measured results. Since the single fraction dose (12 Gy) was beyond the films calibration range (0.5-10 Gy), the prescription dose for the QA plans was scaled to 7 Gy. The phantom setup was same as the single beam test described in Section 2.B. The film was placed in the coronal plane for this measurement, which was different from the CIRS phantom configuration.

2.C.4 | Measurement in heterogeneous phantom
A CIRS phantom was used to present heterogeneity in clinical situations. This phantom includes different rod locations that enable chambers or other dosimeters to be positioned. About half of the phantom is sliced into 1 cm interval, and films can be placed between these layers to measure planar dose. In this study, an ionchamber PTW 31016 was inserted into the right lung and EBT3 film was placed at the axial plane next to the chamber. The treatment plan created in Section 2.C.2 was delivered to the CIRS phantom after rescaled to 7 Gy for one fraction, to ensure that the delivered dose is within the film calibration range.
In this work, gamma-index method was used for the quantitative analysis of planar dose with a low dose threshold of 10%. It combines both the local percentage dose difference (LDD) and the distance to agreement (DTA) criteria, denoted by LDD%/DTA mm. According to AAPM TG 135, a 2%/2 mm criterion was used for the lung case test, requiring a passing rate above 90%. 19 The results were also evaluated using a 3%/2 mm criterion to take into account the dose errors that may be caused by film analysis. For single beam tests, a 3%/1 mm criterion was chosen, considering the tracking accuracy of Cyberknife and higher dose deviation that is likely to occur in the penumbra region.

3.A | Commissioning results
The RMS of TPR is 0.6% averaged for all collimators. And the RMS of OCR is 0.5%. The distance to agreement (DTA) at the OCR penumbra region is within 0.

3.B | Results for single beam tests
As shown in Table 1, the differences between MC calculated and measured chamber dose are within 3%. The gamma passing rates using 3%/1 mm criteria are also summarized in Table 1. Passing rates were superior to 94% for the first four collimators whereas the value dropped to 84.55% for the largest collimator (60 mm). However, the passing rate for 60 mm collimator was improved to 95.33% if using a 3%/2 mm criterion. Table 2 shows the results for lung treatment plan delivered to a homogeneous solid water phantom. MC and RT dose calculation results were compared with the measured data, respectively. As can be seen, the differences between calculated and measured chamber dose were within 3% for both algorithms. For planar dose, the gamma passing rate for MC algorithm was up to 92.51% and 98.31% using 2%/2 and 3%/2 mm criteria, respectively. RT dose calculation was a little less accurate, especially using a stricter criterion (2%/2 mm). The gamma index maps for MC and RT algorithm were shown in Fig. 3. In both maps, the largest gamma value was found in the same location, which was corresponded to the area near the metal stem of the chamber. Discrepancy can also be observed in the medium-low dose region.

3.C.2 | Heterogeneous phantom
The isodose line of 10% which crossed the left lung was much more dispersed for RT algorithm than that for MC algorithm.

| CONCLUSION
In this work, MC dose calculation was validated through experiments performed with ion-chamber and radiochromic film. The results proved that MC algorithm is accurate enough both in homogenous and heterogeneous situations. In contrast, significant dose discrepancy was observed between the RT calculated and measured results when low-density heterogeneity was present. Therefore, MC algorithm is recommended for dose calculation in heterogeneous media, such as lung tumor.

ACKNOWLEDGMENTS
This work was supported by National Natural Science Foundation of China (Grant number 81071237, 81372420).

CONF LICT OF I NTEREST
The authors declare no conflict of interest.