A novel dynamic robotic moving phantom system for patient‐specific quality assurance in real‐time tumor‐tracking radiotherapy

Abstract In this study, we assess a developed novel dynamic moving phantom system that can reproduce patient three‐dimensional (3D) tumor motion and patient anatomy, and perform patient‐specific quality assurance (QA) of respiratory‐gated radiotherapy using SyncTraX. Three patients with lung cancer were enrolled in a study. 3D printing technology was adopted to obtain individualized lung phantoms using CT images. A water‐equivalent phantom (WEP) with the 3D‐printed plate lung phantom was set at the tip of the robotic arm. The log file that recorded the 3D positions of the lung tumor was used as the input to the dynamic robotic moving phantom. The WEP was driven to track 3D respiratory motion. Respiratory‐gated radiotherapy was performed for driving the WEP. The tracking accuracy was calculated as the differences between the actual and measured positions. For the absolute dose and dose distribution, the differences between the planned and measured doses were calculated. The differences between the planned and measured absolute doses were <1.0% at the isocenter and <4.0% for the lung region. The gamma pass ratios of γ3 mm/3% and γ2 mm/2% under the conditions of gating and no‐gating were 99.9 ± 0.1% and 90.1 ± 8.5%, and 97.5 ± 0.9% and 68.6 ± 17.8%, respectively, for all the patients. Furthermore, for all the patients, the mean ± SD of the root mean square values of the positional error were 0.11 ± 0.04 mm, 0.33 ± 0.04 mm, and 0.20 ± 0.04 mm in the LR, AP, and SI directions, respectively. Finally, we showed that patient‐specific QA of respiratory‐gated radiotherapy using SyncTraX can be performed under realistic conditions using the moving phantom.


| INTRODUCTION
Stereotactic body radiation therapy (SBRT) has been used in clinical practice for a variety of tumor types and anatomical locations. 1 However, when treating tumors, particularly those located in the thoracic or the abdominal regions, tumor motion during respiration results in considerable geometric and dosimetric uncertainties in dose delivery. Conventionally, large internal margins (IMs) are required to fully cover the geometric changes that occur during free breathing; such large IMs may result in toxicity to healthy tissue.
At our institution, a novel system that combines TrueBeam (Varian Medical Systems, Palo Alto, CA) and a real-time tumor-tracking radiotherapy system called SyncTraX (Shimadzu Co., Kyoto, Japan) was installed to manage tumor motion due to respiration. This system consists of two color image intensifiers (I.I.s) and X-ray tubes.
The color fluoroscopic images are acquired simultaneously from two directions. There are three options for selecting the positions of the X-ray tubes and I.I.s 2,3 ; these positions are indicated in Fig. 1. Thus, fiducial markers implanted near a tumor can be observed using fluoroscopy during radiation treatment with noncoplanar beams.
Color fluoroscopic images acquired along two directions facilitate the automatic extraction of the position of a fiducial marker close to a tumor by a template pattern matching technique in order to calculate the three-dimensional (3D) coordinates of the markers. When the tracked fiducial marker comes within several millimeters of its 3D planned position (referred to as the gating box), the megavoltage (MV) treatment beam is turned on. This system uses a spatial gating technique that gates the beam by means of the absolute 3D position of the internal fiducial marker instead of using an external surrogate, such as those used in phase or amplitude gating. The details of the SyncTraX system have been reported elsewhere. [2][3][4] To use current motion management strategies (e.g., breath-holding, respiratory-gated radiotherapy, and dynamic tumor tracking radiotherapy technique) in clinical practice, [5][6][7] a correlation between external markers or sensors and internal tumor motion is required. Many researchers have reported a correlation between external markers and internal tumor motion and have revealed that the maximum variation between the internal tumor motion and the external markers is approximately 10 mm. 8,9 Although a correlation between external markers and internal tumor positions exists, the external markers cannot adequately indicate the internal tumor positions in some patients. As the SyncTraX system uses the internal fiducial marker for respiratory gating, it is effective in reducing the IMs. This results in a lower dose to normal tissues and, consequently, a lower risk of complications. 10 In a preliminary study, our group reported that this system can track the motion of a fiducial marker and control radiation delivery with reasonable accuracy. 2 Flattening filter free (FFF) respiratorygated SBRT of a lung tumor was performed using this system in September 2015. We reported that treatment verification in terms of geometric and positional accuracy was achieved in clinical cases using an electronic portal image device and a log file of SyncTraX. 3 We also reported the imaging dose for real-time tumor monitoring during respiratory-gated radiotherapy for the lung using the Monte Carlo technique, but we could not confirm the dose region that may be susceptible to radiation toxicity for the organ at risk. 4 Moreover, patient-specific quality assurance (QA) before treatment was not established.
Several researchers have reported patient-specific dosimetric QA for respiratory-gated radiotherapy using an external surrogate for commercially moving phantoms. 11,12 Commercially moving phantoms cannot reproduce the complex 3D respiratory motion. For respiratory-gated radiotherapy using SyncTraX, the tracking accuracy of the robotic motion phantom using parallel links was proposed 15 and they subsequently evaluated its positioning accuracy. 16 A four-axis moving phantom for patient specific QA, in which four high precision prismatic actuators are used to control the center-of-gravity of the phantom, was also proposed recently. 17 The proposed moving phantom provides sufficiently high accuracy but can be costly. Use of an industrial robotic manipulator for a dosimetric phantom has also been attempted, 18 and evaluation of its positioning accuracy was performed using the log of the joint actuators of the manipulator. In this case, the pure sinusoidal reference trajectories were evaluated, but the recorded 3D tumor motion of the patient was not tested.
Although some researchers have used a simple water-equivalent phantom and a 2D array to perform dosimetric verification for respiratory-gated radiotherapy, 11,19 such an approach lacks the complex human anatomical information. Recently, 3D printing technology has opened the possibility of customization of a wide variety of applications in the medical field. 20,21 It is capable of producing individualized lungmimicking phantoms and is therefore potentially useful for investigating the accuracy of respiratory-gated radiotherapy using SyncTraX.
This study was conducted with the objective of assessing a developed novel dynamic moving phantom system that can reproduce patient 3D tumor motion and patient anatomy, and performing patient-specific QA of respiratory-gated radiotherapy using SyncTraX.

2.A | Patients and treatment planning
Three patients, who underwent respiratory-gated SBRT with SyncTraX for a lung tumor, were enrolled in a study by the institutional review board. Three or four fiducial markers with diameters of 1.5 mm (FMR-201CR; Olympus Co., Ltd, Tokyo, Japan) were implanted close to the tumor in these patients. The clinical characteristics of the patients are summarized in Table 1    previous study. 20 The patient-specific phantom was generated on the basis of the G-code data using an NJB-300W personal 3D printer (Ninjabot; LCC, Shizuoka, Japan) with a polylactic acid (PLA) filament, which is a fused deposition modeling-based 3D printer. 20 The lung region was filled with wood clay. The two nanoDot OSL dosimeters were set into the lung region and one sheet of Gafchromic film (EBT-XD; Industrial Specialty Products, Wayne, New Jersey) was set in the isocenter plane. Four 3D-printed plate lung phantoms were inserted into the WEP, which was set at the tip of the robotic arm.

2.D | Creation of 3D-printed plate lung phantom
T A B L E 2 Technical specification of robot manipulator. J i (i = 1,2, … ,6) in the table denotes the i-th joint axis.  2.E | Reproduce 3D respiratory motion using dynamic robotic moving phantom The command values for driving each joint were sent from the robot controller to the dynamic robotic moving phantom. Then, the dynamic robotic moving phantom reproduced the 3D respiratory motion of the lung tumor. The driving accuracy of the dynamic robotic moving phantom was <0.5 mm. 22 Finally, driving signals of each joint of the dynamic robotic moving phantom were sent to the robot controller. Then, these signals in the joint coordinate system were transferred into those in the treatment room coordinate system via forward kinematics calculation, and they were recorded in a log file.
Method for creating the 3D-printed plate lung phantom. The phantom was designed for planning CT images using the treatment planning system. Four plate phantoms having dimensions of 16 × 16 × 1 cm 3 were created to include the lung tumor and the fiducial marker used as an internal surrogate. The designed phantom structures were exported as a DICOM-RT structure file. This file was converted into an STL file using 3D Slicer. The STL file was compiled into G-code using Simplify 3D software. The patient-specific phantom was generated on the basis of the G-code data using a 3D printer with a polylactic acid (PLA) filament. The lung region was filled with wood clay. The two nanoDot OSL dosimeters were set into the lung region and one sheet of Gafchromic film was set in the isocenter plane. Four 3D-printed plate lung phantoms were inserted into the water-equivalent phantom, which was set at the tip of the robotic arm.
2.G | Patient-specific dosimetric QA for respiratorygated radiotherapy using SyncTraX For planned absolute dose, the nanoDot OSL dosimeters were contoured using TPS. For each contoured nanoDot OSL dosimeter, the mean ± standard deviation (SD) of the planned absolute dose was calculated using TPS. Figure 5 shows the 3D-printed lung phantom in the isocenter plane, where the numbers denote individual nanoDot OSL measurements at the numbered locations. The nanoDot OSL dosimeters were located in the high-dose region with low-dose gradient while checking the dose distribution using the TPS. The counts acquired from the nanoDot OSL dosimeters irradiated by respiratory-gated radiotherapy with SyncTraX were converted into absolute doses using the calibration curve. To measure the absolute dose using nanoDot OSL dosimeters, the high efficiency mode was used in a controlled setting because the measurement was considered for clinical application 23 . The measured absolute dose (D) using a nanoDot OSL dosimeter was calculated using the following equation; where M is the average of three readings corrected for readout depletion, V is the daily variation factor calculated from the ratio of The absolute dose measurements were performed three times.
The mean ± SD of the absolute dose was calculated for each nano-Dot OSL dosimeter and compared with the planned absolute dose.
For dose distribution measurement, 18 irradiated films (three × gating or no-gating × three patients) were scanned in the same orientation (ES-10000G; Epson Corp., Nagano, Japan) with a resolution of 72 dpi in 48-bit color scale with a 24-h postexposure period. All the films were analyzed using commercially available radiation dosimetry software (DD system, version 10.12; R'Tech Inc., Tokyo, Japan). The density of the irradiated films was converted into the absolute dose distribution using a calibration curve; then, the measured and planned dose distributions were compared for an area receiving more than 30% isodose using the gamma index with dose difference/distance-to-agreement criteria (γ D%/ dmm ) of 3%/3 mm and 2%/2 mm. The dose distribution was normalized to the maximum dose.  3.B | Patient-specific dosimetric QA for respiratory-gated radiotherapy using SyncTraX Table 4 summarizes the measured and planned absolute doses for each nanoDot OSL dosimeter. The location numbers correspond to those in Figure 5. For all the patients, the differences between the planned and measured absolute doses were <1.0% for the nanoDot OSL dosimeter set into the 3D-printed tumor. For the other nano-Dot OSL dosimeters in the 3D-printed lung, the differences between the planned and measured absolute doses were <4.0%.
For the organ at risk, such as the lung, the measured absolute doses were in good agreement with the calculated ones. T A B L E 3 Volume, HU, and mass density for 3D-printed plate lung phantoms of all patients.    3.C | Geometric QA for tracking accuracy of SyncTraX system Table 6   phantom system will also be able to support tumor rotation, such as pitch, roll, and yaw, with 6DOFs (Table 2).

Pt
Furthermore, our developed system comprises an industrial robot. As industrial robots are mass produced, cost reduction can be expected. Therefore, our developed system will be less expensive than other phantom systems.
Jung et al. 30 developed individualized lung phantoms that can closely mimic the lung anatomy of actual patients using 3D printing technology. The individualized lung inserts and QUASAR respiratory motion phantom were combined to verify the accuracy of Cyber-Knife tumor-tracking radiotherapy. We also created 3D-printed plate lung phantoms using patient CT images and 3D printing technology, and combined them with the developed dynamic robotic moving phantom to confirm the absolute dose, dose distribution, and tracking accuracy for respiratory-gated radiotherapy with SyncTraX. In particular, the absolute dose into the 3D-printed tumor could be measured using a nanoDot OSL dosimeter. Our results showed that a lung plate phantom with good similarity to a patient can be manufactured using commercially available 3D printing technology (Fig. 6).
The mass densities of the phantom were similar to those of the patients, and the 3D-printed tumor volumes were nearly consistent with those of the patients. According to AAPM-TG 101, 1 treatmentspecific and patient-specific QA should be established using a moving phantom that simulates respiratory motion. Therefore, we developed a moving phantom that can reproduce the complex 3D respiratory motion and patient anatomy. Then, we established patient-specific QA for respiratory-gated radiotherapy with Sync-TraX.
However, the 3D-printed plate lung phantom does not represent the entire human body. The vertebral structures are not included; therefore, dosimetric QA might be easier to achieve than that in an actual situation. Furthermore, the 3D-printed plate lung phantom could not reproduce the lung deformation due to respiration. In this study, the 3D-printed plate lung phantoms were created at end-exhalation. In clinical practice, respiratory-gated radiotherapy with SyncTraX is performed at end-exhalation. Thus, patient-specific QA was performed in a near-clinical situation.
The positional error perpendicular to the beam axis causes a dosimetric difference compared to the planned dose distribution.
The positional error causes blurring in the dose profiles in the case of no-respiratory-gated radiotherapy under a moving phantom (Fig. 7). On the other hand, respiratory-gated radiotherapy with SyncTraX reduced the blurring effect and the measured dose profiles were consistent with the planned dose profile. Our results indicated that the gamma pass ratios of γ 3 mm/3% and γ 2 mm/2% under the gating condition were 99.9 ± 0.1% and 97.5% ± 0.9%, respectively. These results are comparable with the previous relevant results from CyberKnife tumor-tracking radiotherapy. 30 Mutaf et al. 31 reported that irregular respiratory motion included the baseline shift that introduced critical dosimetric consequences for the target coverage for free breathing. Respiratory-gated radiotherapy using external markers mitigated the dosimetric impact of F I G . 8. Irregular respiratory motion pattern of Patient 2 in SI, LR and AP directions for patient-specific QA.
the irregularity of patient respiratory motion. 32 Breath-hold improved the effect of irregular respiratory motion; however, the reproducibility of breath-hold is very important. 5 Figure 8 shows the irregular respiratory motion pattern of Patient 2 in SI, LR, and AP directions for patient-specific QA. Respiratory-gated radiotherapy with SyncTraX could correct the 3D baseline shift using real-time tumor monitoring. In this study, respiratory-gated radiotherapy with SyncTraX reduced the dosimetric impact of 3D irregular respiratory motion using the developed system (Fig. 7).
According to the European Society of Radiotherapy and Oncology guidelines of SBRT, dosimetric accuracy with a median of 3% at the isocenter (2-5%) in a lung phantom inside the treatment field is required. 33 In this study, the differences between the measured and planned absolute doses into the 3D-printed tumor inside the treatment field and into the 3D-printed lung outside the treatment field were <1.0% and <4.0%, respectively, for clinical cases. In our research, nanoDot OSL dosimeters were used to measure the absolute dose in the 3D-printed plate lung phantom. The nanoDot OSL dosimeters have certain characteristics. Kerns et al. 34 reported that the angular dependence of the nanoDot OSL should be considered to measure the dose. Lehmann et al. 35 reported that the angular dependence of the nanoDot OSL dosimeter could be improved by using multiple coplanar beams for clinical measurement, as the overall measurement uncertainty would be reduced. Our measurement results for multiple coplanar and noncoplanar beams in clinical cases were consistent with the results obtained using the TPS.
Ravkilde et al. 36  Here, the tumor was not tracked using the color fluoroscopic images of SyncTraX. Yamazaki et al. 37 indicated that the fiducial marker-tumor misalignment and initial fiducial marker-tumor distance are related, and when the initial fiducial marker-tumor distances were within 25 mm, the misalignments were within 2.5 mm. In our study, the 3D distances between the tumor and the fiducial marker used as an internal surrogate were within 22 mm for all patients ( Table 1). The geometric relationship between the fiducial markers and the lung tumor could be reproduced using 3D printing technology for all patients. Therefore, as the misalignments between the fiducial marker-tumor were small, the tracking accuracy could be evaluated for clinical cases.
In the future, we plan to perform respiratory-gated intensitymodulated radiotherapy (IMRT) with SyncTraX. The developed phantom system will be useful for performing patient-specific QA for respiratory-gated IMRT. Furthermore, it will be useful for the acceptance, commissioning, and QA of novel motion management technologies, such as CyberKnife, 30 kilovoltage intrafraction monitoring, 36,38 and DMLC tracking. 39 However, this developed phantom system has not yet been implemented clinically. Although there are no problems associated with the accuracy of the developed phantom system, a graphic user interface of the phantom control software must be developed to efficiently introduce it to clinical practice.

| CONCLUSION S
In this study, we developed a novel dynamic robotic moving phantom system that can reproduce patient 3D tumor motion and patient anatomy of the respiratory phase at the time of respiratory gating.
Furthermore, we showed that patient-specific QA of respiratorygated radiotherapy using SyncTraX can be performed under realistic conditions using the moving phantom. Overall, the dosimetric and geometric accuracies were found to be sufficiently high in respiratory-gated radiotherapy with SyncTraX.

CONFLI CT OF INTEREST
The authors have no conflict of interest.