Usability of detecting delivery errors during treatment of prostate VMAT with a gantry‐mounted transmission detector

Abstract Volumetric‐modulated arc therapy (VMAT) requires highly accurate control of multileaf collimator (MLC) movement, rotation speed of linear accelerator gantry, and monitor units during irradiation. Pretreatment validation and monitoring of these factors during irradiation are necessary for appropriate VMAT treatment. Recently, a gantry mounted transmission detector “Delta4 Discover® (D4D)” was developed to detect errors in delivering doses and dose distribution immediately after treatment. In this study, the performance of D4D was evaluated. Simulation plans, in which the MLC position was displaced by 0.5, 1.0, 1.5, 2.0, 2.5, and 3.0 mm from the clinically used original plans, were created for ten patients who received VMAT treatment for prostate cancer. Dose deviation (DD), distance‐to‐agreement (DTA), and gamma index analysis (GA) for each plan were evaluated by D4D. These results were compared to the results (DD, DTA and GA) measured by Delta4 Phantom + (D4P). We compared the deviations between the planned and measured values of the MLC stop positions A‐side and B‐side in five clinical cases of prostate VMAT during treatment and measured the GA values. For D4D, when the acceptable errors for DD, DTA, and GA were determined to be ≤3%, ≤2 mm, and ≤3%/2 mm, respectively, the minimum detectable errors in the MLC position were 2.0, 1.5, and 1.5 mm based on DD, DTA, and GA respectively. The corresponding minimum detectable MLC position errors were 2.0, 1.0, and 1.5 mm, respectively, for D4P. The deviation between the planned and measured position of MLC stopping point of prostate VMAT during treatment was stable at an average of −0.09 ± 0.05 mm, and all GA values were above 99.86%. In terms of delivering doses and dose distribution of VMAT, error detectability of D4D was comparable to that of D4P. The transmission‐type detector “D4D” is thus suitable for detecting delivery errors during irradiation.


| INTRODUCTION
Currently, volumetric-modulated arc therapy (VMAT), is being widely performed. 1,2 VMAT requires highly accurate control of multileaf collimator (MLC), rotation speed of linear accelerator gantry, and delivering monitor units during irradiation. Even small errors in these factors would lead to significant accidents in highly precise radiotherapy such as VMAT; therefore, patient-specific pretreatment verification of MLC movement and gantry rotation speed are necessary for VMAT. 3,4 In addition, there exists a possibility of failure of the linear accelerator or its control system while the patient is being treated. Some serious accidents have already occurred in clinical radiotherapy. [5][6][7][8][9] In one of these accidents, MLC opened incorrectly during intensity modulated radiotherapy for head and neck cancer, and the patient was seriously injured. 9 Owing to a hang-up while using the treatment planning system(TPS), the MLC control point data, which should have been present in the treatment planning data, were absent. 9 The accident occurred when a large number of monitor units were delivered to patients without the MLC control point data. 9 Therefore, prior patient-specific verification is imperative to prevent such an accident. 3,4 However, according to a report from the Netherlands in 2010, dose errors could not be detected even after patient-specific preverification, and they were finally detected by in vivo dosimetry (IVD) using an electronic portal imaging device (EPID) during treatment. 10,11 Therefore, not only pretreatment verification but also monitoring of delivering doses and dose distribution during irradiation are necessary for an efficient performance of the VMAT. 12 In recent years, the importance of dose verification during irradiation has been recognized. 10,11 Accordingly, EPID-based IVD is expected to be used for dose verification during irradiation. 13 In traditional IVD, thermoluminescent dosimeter and diode detector placed on the surface of the patient's body have been used for dose verification during irradiation. 13 However, these dosimeters do not seem to be suitable for IVD of VMAT that requires complex dose distribution and steep dose gradient. 13 Furthermore, real-time dose monitoring is impossible for IVD using EPID because IVD using EPID requires a recalculation of the doses monitored during treatment.
For the precise detection of delivery errors during irradiation in VMAT, a gantry-mounted transmission detector was developed.
Gantry-mounted transmission detectors allow monitoring of delivering doses and dose distribution during irradiation immediately after treatment. Errors can be automatically detected immediately after treatment by comparing the dose monitored by the transmitting detector with the planned dose. Therefore, these systems have an advantage in the detection of delivery errors during irradiation dose monitoring and have been applied clinically. 14,15 The purpose of this research is to investigate the possibility of detecting errors during treatment by performing basic experiments that create an error plan for the stop position of MLC using a gantry-mounted transmission detector. This study was conducted based on the assumption that a systematic error occurred owing to the installation position error of the leaf offset during MLC adjustment. In this study, we compared the error detection capabilities of the gantry-mounted transmission detector used for monitoring during treatment and a three-dimensional detector used for pretreatment patient quality assurance (QA). If the error detection capabilities of both the detectors are verified to be comparable, the gantry-mounted transmission detectors can be used independently and will be useful in clinical practice.

2.A | Material
In this study, we used the Delta 4 Discover (D4D) system (ScandiDos AB, Uppsala, Sweden) as the monitoring system to detect errors in dose and MLC position during treatment as well as delivery errors during irradiation. This system consists of Delta 4 Phantom + (D4P) (ScandiDos AB, Uppsala, Sweden) used for patient-specific preverification and the D4D gantry-mounted transmission detector.
The linear accelerator used was a TrueBeam (Millennium 120 MLC 5 mm leaf) (Varian Medical Systems, California, USA), and 10X energy was applied. To ensure the output stability, the output coefficient of variation of the output measurement of weekly QA during the data collection period was calculated. The coefficient of variation of the output dose was found to be stable at 0.19% in the weekly QA dose control using the ionization chamber. The calculated dose of TPS was obtained using Eclipse version 11.0 (Varian Medical Systems, California, USA) based on the anisotropic analysis algorithm.

2.A.1 | Delta 4 discover
The very thin, disk-shaped main unit with a diameter of 790 mm and a thickness of approximately 22 mm measured from the front of the collimator, was designed to fit inside the TrueBeam laser guard. The D4D consisted of 4,040 p-type diode detectors with a diameter of 1.0 mm and a thickness of 0.1 mm. Regarding the detector arrangement, the elements were separated by approximately 1.5 mm, in the X direction (parallel direction along the MLC trajectory) and by 3.0 mm approximately in the Y direction (direction perpendicular to the MLC trajectory). This arrangement is equivalent to 2.5 mm and 5.0 mm intervals in the X and Y directions, respectively, converted to the isocenter plane; the detector covered 195 × 250 mm on the isocenter. The overall thickness, including the detector cover, was 5.5 cm, and the source-to-detector distance was 604 mm. In addition, as a transmission type detector, it has a small effect on the beam transmission and surface dose. 16

2.C | Method
First, the D4P was setup at the isocenter using the treatment room laser. Next, 100MU irradiation was performed from a gantry angle of 0°and 90°in a 10 cm × 10 cm irradiation field, and the profile was confirmed using the software provided with D4P. Based on the profile calculated by TPS, the TrueBeam couch was moved in 0.1 mm increments to a position where they matched best, and the D4P was re-setup. Figure 2 shows the procedure of D4D measurement. As D4D is a fluence measuring transmission detector, the dose is measured by synthesizing the dose as measured by D4P. 18

2.C.1 | Evaluation of reproducibility
As D4D is a gantry-mounted transmission detector, the measurement value may vary depending on the mounting position. First, the reproducibility of the measurement due to the accuracy of the detector's attachment to the gantry is evaluated. D4D then obtains the measurement results based on the reference measurements collected at the same time as D4P. Thereby, we evaluate the reproducibility of the D4P setup.
Evaluation of D4D reproducibility in the same setup (D4D mounting reproducibility) The D4D mounting accuracy reproducibility was evaluated for one prostate cancer treatment. First, D4P was set up to measure the dose distribution of the original plan and six simulated plans. Next, as shown in Figure 2, the reference measurements were performed using D4P and D4D. Further, using D4D only, we have measured the dose distribution of the original plan and the six types of simulated plans. The measurements in this study were performed first, F I G . 2. Procedure of D4D measurement. Combine D4P and D4D to measure the reference measurement using the "synthetic mode". Next, the fluences are converted into doses by directly measuring using D4D. * Reference measurement of this process should be performed for all dose verification cases.
with D4P alone; then, D4D was mounted and measured. This is defined as one set of measurement. D4D was dismounted and mounted before each set of measurements. To measure the D4D mounting repeatability, the mount-dismount process was repeated ten times.

Evaluation of reproducibility of D4P setup (Reproducibility of reference measurement)
To evaluate the reproducibility of the D4P setup, we have used the original plan for one prostate treatment used in the evaluation of D4D reproducibility, and six types of simulated plans. D4P and D4D were used to re-setup ten times on another day instead of the same day, and the dose distribution was measured.

2.C.2 | Evaluation of VMAT for ten prostate cancer patients
The setup of ten patients' prostate treatment plans was measured similar to the evaluation of the reproducibility of the D4D setup.
The process in Figure. 2 was repeated for each case and the dose distribution of D4P and D4D was measured once.

2.C.3 | Evaluation criteria
To evaluate the methods of 2.C.1-2.C.2, dose deviation (DD), distance-to-agreement (DTA), and gamma index analysis (GA) were used for D4P and D4D reproducibility. The evaluation criteria observed were DD at 3%, DTA at 2 mm, GA at 3 %/2 mm, with the threshold set at 10%. 4 The variation was evaluated at a 95% confidence interval (1.96 standard deviations). The GA was performed with a global normalization in the absolute doses. For the evaluation of VMAT for ten prostate cancer patients, we have performed a statistical test for significant differences between the original and simulated plans, using Welch's t-test. A p-value of 0.05 was used; after performing Bonferroni correction considering multiple comparisons, a p-value less than 0.007 was considered statistically significant.

2.D | Variation in MLC stop position and GA of prostate VMAT clinical data using D4D
Rangel et al. reported a less significant impact on delivery errors owing to random errors compared to systematic errors in MLC. 19 Therefore, the evaluation of the basic error detection capability in this study was performed under the assumption that systematic errors occurred due to leaf offset placement errors during MLC adjustment. However, it is also important to investigate the MLC random errors. Therefore, we compared the deviations between the  and D4D was 69.8% and 77.6% at 3.0 mm, respectively, and the D4D minimum pass-ratio was 7.8% higher than that for D4P. The average values of the pass-ratio variation of the original plan and the MLC position error simulation plan were ± 0.8% for D4P and ± 1.3% for D4D. Evidently, the variation was smaller for D4P compared to that for D4D. Figure 3b shows that, when the MLC position error exceeds 0.5 mm, the value of D4D becomes higher than of D4P, and the tendency is similar to that of DD. The maximum pass-ratio was 97.7% of the original plan in the case of D4P and 97.5% in the case of MLC position error of 0.5 mm in the case of D4D. The minimum pass-ratio of D4P and D4D was 80.5% and 87.1%, respectively, at 3.0 mm, and the D4D minimum pass-ratio was 6.6% higher than that for D4P.

3.A.2 | DTA
The average values of the pass-ratio variation of the original plan and the MLC position error simulation plan were ± 0.9% for D4P and ± 1.2% for D4D. Evidently, the variation was smaller for D4P compared to that for D4D. | 69 the original plan and the MLC position error simulation plan were ± 0.5% for D4P and ± 0.8% for D4D. Again, the variation was evidently smaller for D4P compared to that for D4D. the minimum pass-ratio for D4D was 6.7% higher than that for D4P.

3.A.3 | GA
The average values of the pass-ratio variation of the original plan and the MLC position error simulation plan were ± 1.2% for D4P and ± 1.5% for D4D. Evidently, the variation for D4P was smaller than that for D4D. Compared to the evaluation of D4D reproducibility in the same setup (D4D mounting reproducibility), the minimum pass-ratio for D4P became higher, the maximum pass-ratio for D4D was 94.6%, and the minimum pass-ratio was smaller at 76.9%. In addition, both D4P and D4D showed higher variability. Figure 4b shows that the value of D4D becomes higher than that of D4P when the MLC position error exceeds 0.5 mm, and the tendency is similar to that of DD. The maximum pass-ratio was 97.7% of the original plan for D4P and 97.6% for MLC position error of 0.5 mm in the case of D4D. The minimum pass-ratios for D4P and D4D were 80.6% and 87.2%, respectively, at 3.0 mm, and the minimum pass-ratio for D4D was 6.6% higher than that for D4P. The average values of the pass-ratio variation of the original plan and the MLC position error simulation plan were ±0.8% for D4P and ±0.9% for D4D. The variation for D4P was smaller than that for D4D.

3.B.2 | DTA
Compared to the evaluation of D4D reproducibility in the same F I G . 3. Evaluation of D4D reproducibility in the same setup (D4D mounting reproducibility). (a) DD: the horizontal axis shows the D4P pass-ratio, and the vertical axis shows the D4D pass-ratio. The line connecting the values where D4P and D4D are equal is defined as the reference line. (b) DTA: the horizontal axis shows the D4P passratio, and the vertical axis shows the D4D pass-ratio. (c) GA: the horizontal axis shows the D4P pass-ratio, and the vertical axis shows the D4D pass-ratio. setup (D4D mounting reproducibility), the variability of both D4P and D4D was small. Figure 4c shows that the value of D4D exceeds that of D4P when the MLC position error exceeds 1.0 mm, similar to DD and DTA. The maximum pass-ratio for D4P and D4D was 100.0% when the original plan and the MLC position error were 0.5 mm. The minimum pass-ratios for D4P and D4D were 85.6% and 89.3%, respectively, at 3.0 mm, and the minimum pass-ratio for D4D was 3.7% higher than that for D4P.

3.B.3 | GA
The average values of the pass-ratio variation of the original plan and the MLC position error simulation plan were ± 0.4% for D4P and ± 0.8% for D4D. Similarly, the variation for D4P was evidently smaller than that for D4D. Compared to the evaluation of D4D reproducibility in the same setup (D4D mounting reproducibility), the minimum passratio of D4D became smaller. Additionally, the variation was almost the same for both D4P and D4D.

3.C | Evaluation of VMAT for ten prostate cancer patients
The DD, DTA, and GA pass-ratios of evaluation of VMAT for ten prostate cancer patients are shown in Figure 5. These pass-ratios tend to be smaller than the evaluation of D4D reproducibility in the same setup (D4D mounting reproducibility) and the evaluation of the reproducibility of D4P setup (Reproducibility of reference measurement). Additionally, the variability was higher for both D4P and D4D. The differences from the evaluation of D4D reproducibility in the same setup (D4D mounting reproducibility) and evaluation of the reproducibility of D4P setup (Reproducibility of reference measurement) are shown below. Figure 5a shows that the pass-ratio of D4D DD exceeds that of D4P when the MLC position error exceeds 1.5 mm, and this tendency increases as the MLC position error increases. When the MLC position error was 1.0 mm, the maximum pass-ratios for D4P and D4D were 94.5 and 93.2% respectively. The minimum pass-ratios of D4P and D4D were 68.6% and 74.2% at 3.0 mm, respectively, and the minimum pass-ratio for D4D was 5.6% higher than that for D4P.

3.C.1 | DD
The average values of the pass-ratio variation of the original plan F I G . 4. Evaluation of reproducibility of D4P setup (Reproducibility of reference measurement). (a) DD: the horizontal axis shows the D4P pass-ratio, and the vertical axis shows the D4D pass-ratio. The line connecting the values where D4P and D4D are equal is defined as the reference line. (b) DTA: the horizontal axis shows the D4P pass-ratio, and the vertical axis shows the D4D pass-ratio. (c) GA: the horizontal axis shows the D4P pass-ratio, and the vertical axis shows the D4D pass-ratio. and the MLC position error simulation plan were ±5.0% for D4P and ±4.9% for D4D. The variation for D4D was smaller than that for D4P. The maximum and minimum pass-ratio D4P and D4D were smaller than the evaluation of D4D reproducibility in the same setup (D4D mounting reproducibility) and evaluation of reproducibility of D4P setup (Reproducibility of reference measurement). In addition, both D4P and D4D showed higher variability. Figure 5b shows that the value for D4D was higher than that for D4P when the MLC position error exceeded 0.5 mm, and the tendency was similar to that of DD. The maximum pass-ratio was 96.9% of the original plan for D4P and 96.8% for MLC position error of 0.5 mm for D4D. The minimum pass-ratios for D4P and D4D

3.C.2 | DTA
were 83.0% and 85.9%, respectively, at 3.0 mm, and the D4D minimum pass-ratio was 2.9% higher than that for D4P. The average values of the pass-ratio variation of the original plan and the MLC position error simulation plan were ±3.4% for D4P and ±3.0% for D4D. The variation for D4D was smaller than that for D4P. The minimum pass-ratio for D4P was greater than evaluation of D4D reproducibility in the same setup (D4D mounting reproducibility) and evaluation of the reproducibility of D4P setup (Reproducibility of reference measurement). Figure 5c shows that the value for D4D was higher than that for D4P when the MLC position error exceeded 1.0 mm, similar to DD and DTA. The maximum pass-ratio for D4P and D4D was 100.0% when the original plan and the MLC position error were 0.5 mm.

3.C.3 | GA
The minimum pass-ratios for D4P and D4D were 82.9% and 87.1%, respectively, at 3.0 mm, and the minimum pass-ratio for D4D was 4.2% higher than that for D4P. The average values of the pass-ratio variation of the original plan and the MLC position error simulation plan were ±3.9% for D4P and ± 3.2% for D4D. Similarly, the variation for D4D was evidently smaller than that for D4P. The maximum pass-ratios were the same for both D4P and D4D. However, variation was observed at 0.5 mm.

3.C.4 | Statistical significance test
We performed a statistical significance test to evaluate the differences between the simulated plan and the original D4D plan using F I G . 5. Evaluation of VMAT for ten prostate cancer patients. (a) DD: the horizontal axis shows the D4P pass-ratio, and the vertical axis shows the D4D pass-ratio. The line connecting the values where D4P and D4D are equal is defined as the reference line. (b) DTA: the horizontal axis shows the D4P pass-ratio, and the vertical axis shows the D4D pass-ratio. (c) GA: the horizontal axis shows the D4P pass-ratio, and the vertical axis shows the D4D pass-ratio.
Welch's t-test on DD, DTA, and GA.

3.D | Variation in MLC Stop Position and GA of Prostate VMAT Clinical Data Using D4D
According to the data listed in Table 2

| DISCUSSION
In this study, we evaluated the ability of the D4D for detecting MLC position error, measuring the dose distribution of the original plan and simulated plan. As EPID in a previous study reportedly detected an MLC error of 3 mm during treatment, we evaluated whether the gantry-mounted transmission detector used in this study had the same order of detection capability. 20,21 First, we assessed the reproducibility of the D4D setup for the same day. To ensure the output stability, the output coefficient of variation of the output measurement of weekly QA during the data collection period was calculated.
The variation coefficient of the output dose of the linear-accelerator device used in this study was found to be stable at 0.19% in the weekly QA dose control using an ionization chamber. The one daily short-term reproducibility output coefficient of variation was 0.01%.
In the previous reports, the short-term reproducibility of ten D4P measurements was 0.1% (1SD) and the long-term stability was 0.5% (1SD). 17   is a more sensitive device. Assuming a 95% pass rate, some plans passed with D4D but failed with D4P at leaf offsets of 1.5 and 2 mm.
The user should determine whether D4D is sensitive enough for general use after evaluating the data based relevant guidelines. The UK guidelines do not specify the procedure to assess the accuracy of the system or the indicators that must be used. 13 Therefore, we believe that this study contains useful information for D4D and D4P users.
Furthermore, MLC position error can be detected by D4D alone, and we can expect its application in the detection of delivery errors during irradiation.
Assuming that D4D is used for the detection of delivery errors during irradiation, it is possible to accurately detect errors of the linear accelerator alone since the output dose distribution emitted from the linear accelerator is directly observed. In contrast, when EPID is used, the dose distribution of radiation penetrated through the patient is measured, the effects of body shape change, and patient setup are included. Therefore, it is challenging to evaluate error detection only for the linear accelerator alone.
A 2010 report from the Netherlands highlighted the importance of dose verification during treatment. 10,11 Among the MLC positional errors that occur during VMAT and IMRT treatment, this study shows that systematic shifts in both MLC banks can be used to detect delivery errors during irradiation at a constant level. In this study, error detection analysis was performed on the prostate as it is considered to have the least intensity modulation among the sites. 4 However, additional studies are required for more detailed detection of errors during irradiation, involving other sites such as head and neck cases.

| CONCLUSION
In this study, we examined the suitability of D4D in detecting MLC position error. It was found that D4D can detect delivery errors during irradiation. D4D has almost the same detection power as D4P and can detect MLC position errors of 1.5 mm or more using DTA or GA. In conclusion, a transmission-type detector can be suitable for the detection of delivery errors during irradiation.

CONFLI CT OF INTEREST
The authors declare that there are no conflicts of interest.

AUTHOR CONTRI BUTIONS
The conceptual design of the study was carried out by H. Honda, M.

D A T A A V A I L A B I L I T Y S T A T E M E N T
The data that support the findings in this study are from cases conducted at our hospital. Also, the database released for other institutions is not subject to ethical review approval. Furthermore, due to privacy and ethical restrictions, data provision is not available upon request. Therefore, data sharing is not applicable.