Comparison of MLC error sensitivity of various commercial devices for VMAT pre‐treatment quality assurance

Abstract The purpose of this study was to compare the MLC error sensitivity of various measurement devices for VMAT pre‐treatment quality assurance (QA). This study used four QA devices (Scandidos Delta4, PTW 2D‐array, iRT systems IQM, and PTW Farmer chamber). Nine retrospective VMAT plans were used and nine MLC error plans were generated for all nine original VMAT plans. The IQM and Farmer chamber were evaluated using the cumulative signal difference between the baseline and error‐induced measurements. In addition, to investigate the sensitivity of the Delta4 device and the 2D‐array, global gamma analysis (1%/1, 2%/2, and 3%/3 mm), dose difference (1%, 2%, and 3%) were used between the baseline and error‐induced measurements. Some deviations of the MLC error sensitivity for the evaluation metrics and MLC error ranges were observed. For the two ionization devices, the sensitivity of the IQM was significantly better than that of the Farmer chamber (P < 0.01) while both devices had good linearly correlation between the cumulative signal difference and the magnitude of MLC errors. The pass rates decreased as the magnitude of the MLC error increased for both Delta4 and 2D‐array. However, the small MLC error for small aperture sizes, such as for lung SBRT, could not be detected using the loosest gamma criteria (3%/3 mm). Our results indicate that DD could be more useful than gamma analysis for daily MLC QA, and that a large‐area ionization chamber has a greater advantage for detecting systematic MLC error because of the large sensitive volume, while the other devices could not detect this error for some cases with a small range of MLC error.

that MLC error was linearly correlated with the generalized equivalent uniform dose for all error types. Therefore, quality assurance (QA) is needed to address MLC positioning errors.
Several studies have attempted to examine this issue. One approach to MLC QA is to analyze the machine's logfile to perform MLC QA, 2,3 and this approach can be used with data from a dynamic delivery (MLC position, Gantry Angle, and Monitor Unit) with higher sampling rate (10-50 ms). In addition, logfile QA can be performed using three-dimensional dose reconstruction based on patient geometry, and several recent reports have described this technique. [4][5][6] The second MLC QA method is based on measurements that are obtained using various radiation detectors, which can involve an electric portal imaging device (EPID), 7 a film or a two-dimensional (2D) diode (or chamber) array, 8 or a 3D diode array. 9 A gamma analysis was developed by Low et al., 10 which combined the dose difference (DD) and distance to agreement (DTA). Although 2D gamma analysis has been used in clinical situations, 3D gamma analysis has been used for volume dose analysis, such as comparing phantom and patient doses. 11 Although the gamma pass rate is usually used in clinical situations, it is not an absolute evaluation index, and Hussein et al.
demonstrated that different device and software combinations exhibited variable agreement with the predicted values for the same pass rate criteria. 12 Furthermore, Nelms et al. reported that gamma pass rates were not strongly correlated with dose errors in anatomical regions of interest. 13 Therefore, to predict the real patient dose, the measurement dose reconstruction method has been developed. 14 Although there is increasing awareness of the importance of irradiation QA, the importance of accurately monitoring MLC movement/position remains unchanged, as it affects the accuracy of dose prescription. Some studies have investigated the MLC error sensitivity of different QA devices. 8,15 However, there remains insufficient evidence regarding the advantages or disadvantages of each QA device in the clinical setting. Furthermore, a large area ionization chamber has not been investigated. Therefore, this study aimed to compare the MLC error sensitivity of various measurement devices for VMAT pre-treatment QA.

2.A | QA devices
This study used four QA devices (Delta4: ScandiDos, Uppsala, Sweden; 2D-array seven29: PTW, Freiburg, Germany; IQM: iRT Systems GmbH, Koblenz, Germany; Farmer chamber: PTW 30013). Figure 1 shows the phantom setup for each device. The Delta4 device is a cylindrical PMMA phantom that surrounds two crossing orthogonal planes with 1,069 p-Si diodes. The diodes are disc-shaped, have a volume of 0.04 mm 3 , and are placed at 5-mm intervals in the central areas (6 cm 9 6 cm) and at 10-mm intervals in the outer areas (up to 20 cm 9 20 cm). The 2D-array is equipped with 729 equally spaced ionization chambers, with a center-to-center distance of 1 cm and covering an active area of 27 cm 9 27 cm. Each chamber has a size of 5.0 9 5.0 9 5.0 mm. An octagon-shaped phantom (Octavius Phantom) with a central cavity was used to insert the 2D ion chamber array.
An integral quality monitoring system (IQM) is a large-area ionization chamber that was introduced by Islam et al. 16 The IQM is usually mounted on the linac gantry head, has aluminum components, and provides a sensitive area of 22 cm 9 22 cm. The chamber can monitor a radiation field that projects to a size of approximately 34 cm 9 34 cm at the isocenter. The sensitive volume of the ion chamber was approximately 530 cm 3 . To produce greater spatial sensitivity, the optimized inclined detection plane was used, and additionally technical details have been described in a F I G . 1. Phantom setup for each device ((a)-(d)). A Farmer chamber was used by inserting to the solid phantom (a). The IQM was used by mounting on the linac gantry head (b). The 2D-array (c) was used by inserting to an octagon-shaped phantom (Octavius Phantom) with a central cavity.
previous report. 16 The 0.6-cc Farmer chamber was used by inserting a solid water phantom (30 9 30 9 20 cm) to a depth of 10 cm.

2.B | MLC error plans
Nine retrospective VMAT plans were used: three head and neck boost plans (30 Gy/15 fractions), three lung stereotactic body radiation therapy (SBRT) plans (50 Gy/4 fractions), and three prostate plans (76 Gy/ 38 fractions). All plans were generated as half arcs, full arcs, or two full arcs using the Pinnacle3 treatment planning system (Philips Radiation Oncology Systems, Madison, WI, USA) and the SmartArc optimization algorithm. The calculations were performed with a dose calculation grid of 2 mm and control points of 2°. The irradiation was performed using a 6-MV photon beam that was generated by a Synergy linier accelerator (Elekta Oncology Systems, UK) and an Agility gantry head, which has 160 multi-leaf collimators (width: 5 mm).
To investigate the MLC error sensitivities, nine MLC error plans were generated for all nine original VMAT plans. First, we investi-

2.C | Evaluation of MLC error sensitivity
The MLC error sensitivity for each device was evaluated by comparing the doses of the original VMAT plan to that of the corresponding error plan. To investigate the sensitivity of the Delta4 device and the 2D-array, global gamma analysis (1%/1, 2%/2, and 3%/3 mm), DD (1%, 2%, and 3%) were used. The analyses were performed using 20% of the maximum dose as the low dose threshold, and pass rates (%) were calculated for all analyses. The IQM and Farmer chamber were evaluated using the cumulative signal difference (S diff ) which was calculated using the following equation: In this equation, S baseline and S error are the cumulative signals of the baseline plan and the MLC error plan, respectively. The Wilcoxon signed-rank test was used to compare IQM versus the Farmer chamber and the 2D-array versus the Delta4 device. All analyses were performed using JMP Pro software (version 11; SAS Institute Inc., NC).  the 2D-array using the three criteria (1%/1, 2%/2, and 3%/3 mm).

| RESULTS
The pass rates decreased as the magnitude of the MLC error increased for both devices. In particular, the gamma analysis using the strictest criteria (1%/1 mm) had better sensitivity than the loosest criteria (3%/3 mm). Furthermore, the small MLC error for small aperture sizes, such as for lung SBRT, could not be detected using the loosest gamma criteria (3%/3 mm). Tables 2 and 3 shows the gamma pass rates and DD between the baseline and error plans for the 2D-array and Delta4 device. The DD provided better sensitivity than the gamma analyses, although there were no significant differences between the two devices for almost all cases.

| DISCUSSION
The accuracy of MLC position and movement affects dose prescription accuracy, and MLC QA is needed for complicated treatments, such as intensity-modulated radiation therapy. Therefore, AAPM TG-142 report mentioned that the leaf position repeatability should be within AE1 mm. 17  2D-gamma and 3D-gamma analyses, and concluded that 3D-gamma analysis provided up to 2.9% more pixels passing than 2D-gamma analysis. 11 Therefore, we used 2D-gamma analysis because it is more sensitive than 3D-gamma analysis. Based on the results from Fig. 4 and Table 2, there were no significant differences between two devices (Delta4 and 2D-array) for all criteria. It is possible that their course detector resolution affected the pass rates, which could enhance the sensitivities of these devices to MLC error, especially using the loosest criteria (ie, 3%/3 mm). In addition, for the lung SBRT plans, the conventional gamma criteria (2%/2 mm or 1%/ 1 mm) could not detect small MLC errors, based on a magnitude of AE0.25 mm (Fig. 4). Thus, our results suggest that gamma analysis using loose criteria might be not suitable for detecting relatively small F I G . 4. The gamma pass rates between the baseline and MLC error plans for the Delta4 device and the 2D-array using the three criteria (1%/1, 2%/2, and 3%/3 mm). The results of Delta4 ((a)-(c)) and 2D-array ((d)-(f)) are represented.
MLC errors when 2D gamma analysis was used, although it may be more sensitive to large MLC errors. However, DD had a better sensitivity than gamma analysis. In addition, the DD of Delta4 had significantly better sensitivity than that of the 2D-array for some error patterns, as shown in Table 3
T A B L E 2 Gamma pass rate and dose difference between a baseline and the nine error plans for a 2D-array and Delta4 for all nine plans. Three criteria of gamma pass rate (1%/1, 2%/2, and 3%/3 mm) are represented.