Practical implications for the quality assurance of modulated radiation therapy techniques using point detector arrays

Abstract Purpose Linac parameters potentially influencing the delivery quality of IMRT and VMAT plans are investigated with respect to threshold ranges, consequently to be considered in a linac based quality assurance procedure. Three commercially available 2D arrays are used to further investigate the influence of the measurement device. Methods Using three commercially available 2D arrays (Mx: MatriXXevolution, Oc: Octavius1500, Mc: MapCHECK2), simple static measurements, measurements for MLC characterization and dynamic interplay of gantry movement, MLC movement and variable dose rate were performed. The results were evaluated with respect to each single array as well as among each other. Results Simple static measurements showed different array responses to dose, dose rate and profile homogeneity and revealed instabilities in dose delivery and profile shape during linac ramp up. Using the sweeping gap test, all arrays were able to detect small leaf misalignments down to ±0.1 mm, but this test also demonstrated up to 15% dose deviation due to profile instabilities and fast accelerating leaves during linac ramp up. Tests including gantry rotation showed different stability of gantry mounts for each array. Including gantry movement and dose rate variability, differences compared to static delivery were smaller compared to dose differences when simultaneously controling interplay of gantry movement, leaf movement and dose rate variability. Conclusion Linac based QA is feasible with the tested commercially available 2D arrays. Limitations of each array and the linac ramp up characteristics should be carefully considered during individual plan generation and regularly checked in linac QA. Especially the dose and dose profile during linac ramp up should be checked regularly, as well as MLC positioning accuracy using a sweeping gap test. Additionally, dynamic interplay tests including various gantry rotation speeds and angles, various leaf speeds and various dose rates should be included.

During the last years, measurement devices have largely improved with regard to resolution 1-3 and needs for rotational techniques like VMAT. 4 In addition, improvements in tests for the correct beam model parameters were proposed. [5][6][7][8] Nevertheless, the evaluation of treatment plan QA is still cumbersome as shown by ongoing discussions about the widely used gamma-evaluation method itself, the optimal evaluation parameters and the comparison between different measurement devices. [9][10][11][12][13][14][15][16][17][18] Bypassing the measurement, linac log files, which monitor each single parameter of the linac, can be used to reconstruct the delivered dose. [19][20][21] As linac log files monitor surrogate parameters for absolute linac values, miscalibrations could stay undetected without tight linac QA. Furthermore, the sampling rate might be too low in some cases to reveal errors in fast changing parameters like leaf acceleration, which was shown to have a large impact on good agreement between calculation and delivery. 22,23 Besides plan-individual measurements and evaluation of linaclog files, delivery subsystems such as MLC movement or dose rate stability can be analyzed separately in a linac based QA approach.
This means that the complexity of especially VMAT is split into its single significant components and that measurements are conducted without devices that are developed especially for the rotational needs in VMAT. Instead, known devices (e.g. 2D arrays) are used. During commissioning, this simplification points out general linac parameters that are potentially subject to influence the delivery quality of IMRT and VMAT and thus should be respected throughout the treatment plan generation. As a consequence, delivery mismatch with respect to the TPS calculation can be modeled and quantified, thereby establishing action levels. With 2D array measurements, we show how the splitting of complexity into significant parameters, that may influence IMRT and VMAT delivery quality, enables effective and efficient regular QA of delivery subsystems. This can provide information about the actual linac conditions, which can be used to ensure good treatment plan delivery.
Thus, we propose an efficient, fast and meaningful linac based QA approach without extensive, cumbersome and time-consuming plan individual QA by decomposing the complexity of IMRT and VMAT and evaluating the delivery performance of the linac. As different linacs may behave differently and guidelines may require department specific limits, our work focuses on the workflow of identifying linac parameters limiting the delivery quality and separating them from restrictions of measurement devices. Three different 2D arrays are used as an example for how to separate restrictions of the measurement device from the limitations of the linac delivery, which can be translated to other measurement devices. Eventually, the limiting parameters and their safe ranges need to be carefully considered during beam modeling and could be implemented to the TPS in order to guarantee IMRT and VMAT plans with safe delivery parameters.

2.A | Complexity levels in IMRT and VMAT QA
We break down the complexity of modulated radiation therapy techniques into three categories to investigate threshold ranges that may potentially limit plan delivery quality. These comprise linac parameters which should be included into plan generation as well as measurement shortcomings using 2D arrays: 1. Simple static QA tests (Table 1A): Treatment planning systems assume linearity of dose and dose rate as well as independence of the dose profiles from the delivered dose and dose rate. Furthermore, dose output with respect to the field size (output factors) is measured once and assumed constant over time. With simple static tests, these assumptions are verified with respect to the delivery accuracy of the linac as well as with respect to each array's capability of measuring these assumptions. As IMRT and VMAT plans may introduce several (differently shaped and steep) dose gradient areas, the resolution capability of the different arrays is verified by measuring the penumbra of a static field, investigating each array's potential of correct dose measurements. Therefore, these simple static tests may reveal general threshold ranges for each tested parameter, which may degrade the delivery and measurement quality of IMRT and VMAT plans. (Table 1B): MLC positioning is one of the most crucial concerns in radiation therapy, especially in modulated techniques, and a reliable test using either measurement method is required for regular MLC QA. As either the picket fence or the sweeping gap test will be used for this regular QA, we explore the power of identifying leaf mispositioning in the range of   (Table 1C). The maximal leaf speed is set at the manufacturer's specification to 6.0 cm/s. The minimal speed is assumed to be 0.1 cm/s after having investigated several clinical VMAT plans. The maximal gantry rotation speed is 6°/s by regulation. A reasonable minimal gantry rotation speed is assumed to be at 1°/s for these studies. The maximal dose rate of the used linac was around T A B L E 1 Three level QA: simple static QA tests (A) serve as inspection of the measurement limits of the used three detector arrays. MLC parameter measurements (B) explore the different assumptions for TPS modeling. Dynamic interplay measurements (C) ensure the correct dose delivery at gantry speed, leaf speed and dose rate limits.

2.B | Measurement devices and setup
Similar consistent measurements for the three commercially available  Table 2.
All measurements were carried out using a 6 MV photon beam of an Elekta Synergy linac equipped with an Agility TM multi-leaf-collimator (MLC, leaf width: 5 mm). Before each measurement session, the arrays were dose calibrated at 5 cm depth.

3.A.5 | Measurement of output factors
For field sizes ≥2 9 2 cm² no difference between the arrays and the according base data measurements (Farmer-type ion chamber for field sizes larger than 5 9 5 cm², pinpoint detector for smaller field sizes) was found. Due to the volume averaging of the ion chambers, Mx and Oc measure 76% and 13% less dose compared to Mc and the according base data measurement for 1 9 1 cm².

3.A.6 | Dose gradient measurement
As a surrogate for dose gradients, the penumbra of a 10 cm 9 10 cm field was used (Fig. 3). more pronounced compared to the ion chamber arrays (Fig. 4).   or leaf gap only, the deviation between the arrays is ≤1%. In the region with transmission only, the deviation is ≤2% (Fig. 6) . 7).

| DISCUSSION
During previous years, commercially available measurement devices for IMRT and VMAT QA have been tested for general performance and plan-specific QA abilities 2,3,10,24,25 . Dose, dose rate and field size dependence as were described by these authors are supported by our data, even though the non-linearity for very small doses in the linac ramp up was either not studied before or described to a different reference and also field sizes below 2 9 2 cm² we not studied.
This study therefore contributes to more detailed understanding of array behavior in regions of small dose, dose rate, and field size,  Fig. 1(b)] and scaled to the isocenter; Oc and Mc were setup isocentric and shifted [ Fig. 1(a)]. F I G . 7. Sweeping gap test with a 2 cm wide gap swept over 20 cm across the field with 50MU from left to right (orange) and right to left (green) for Mx (left), Oc (middle) and Mc (right). The dose is calculated relative to the dose at the isocenter of a 20 9 20 cm² field. Red arrows indicate regions with up to 15% dose deviation in the starting region of the leaves; blue arrows indicate up to 5% overdose due to decreased profile homogeneity durung the linac ramp up at low dose rates. cylindrical phantom). Therefore, either method, which compares the calculated dose distribution against the measured, will give different results for different arrays. Additionally, using one array, different plans having different characteristics with regard to e.g. the amount of small dose segments may also differ in the optimal parameters to set for the evaluation. Consequently, the found array characteristics have to be kept in mind during plan individual QA.
Searching for optimal the QA method, two main different ways are possible from the view of the authors: one way is perusing the plan individual QA; the other way is working towards linac based QA.
In general, plan-individual QA focuses the challenge of finding the correct pass criteria. On the one hand, using the popular method of gamma-evaluation and subsequent pass rate analysis, 26 one can argue about the correct distance-to-agreement and dose criteria or pass rate, that has to be used. 2,9,10,12,13,15,18 On the other hand, one can question this method in general, 27,28 as there might be a lack of correlation between the results of the gamma-analysis and the clinical implication. 9,11,13,14,27,[29][30][31][32] Furthermore, the optimal criteria might depend on the measurement device, 11,[16][17][18]  This tolerance may depend on the used array, the exact composition of dose distribution as well as the used linac. However, increasing tolerances will also mask important shortcomings of the plan and its delivery. Therefore, this procedure will not only be cumbersome but also misleading with regard to identify important delivery errors.
Instead of introducing location specific tolerances in plan QA, one could include the found characteristics of the measurement device and linac in the treatment planning system, for example by lookup tables. As this would imply unchanged characteristics of the linac during its life time, the QA procedure might get even stricter.
Generating plans that respect shortcomings of the linac delivery as well as the shortcoming of used plan QA tool, e.g. by setting quite high constraints to the minimal dose per segment, may degrade the plan as the necessary degrees of freedom of IMRT and VMAT cannot be fully utilized.
Therefore, tight linac QA, which guarantees the TPS assumption within requested limits, may give enough confidence about the correct dose delivery of the plans. One part is finding the limiting factors that will degrade plan delivery and thereby change the dose distribution clinical significantly. According to the presented results, this would especially include the delivery conditions at linac ramp up and low dose rates, since larger deviations (e.g. from dose and dose rate linearity) were found at these conditions. Furthermore, dynamic techniques may not deliver dose as calculated, if leaves have to be driven at large speed and the dose rate has to change rapidly at the same time. Consequently, a second part toward linac QA consists of decomposing IMRT plans with respect to factors that may limit delivery quality and implement them in the treatment planning process and to predict its clinical influence for the specific plan. Eventually, this would lead to a plan specific constraint in the TPS, which monitors the potential amount of delivery outside of the given limitations. The quality of plan delivery will then depend on the range of essential linac parameters and the amount of dose delivery close to or outside the given limits. While the amount of dose delivered close to or outside of linac limits may depend on the specific requirements of the plan, the range of factors limiting delivery quality has to be kept small by meaningful tests, which are fast and easy to handle.
The tests presented here will be suitable to not only detect clockwise gantry rotation no gantry rotation counterclockwise gantry rotation F I G . 8. Sweeping gap test including gantry rotation (green) compared to the same sweeping gap without gantry rotation (orange) (field D, Table 1 The presented results also show the differences in several 2D arrays, which give rise to the assumption that the used 2D arrays are suitable for constancy test in the context of linac based QA, but might lack accuracy in initial linac commissioning. This is because the comparison of the results between different arrays may give different assumptions of the linac's condition and if actions need to be taken or not. Therefore, initial linac commissioning should be done using the known measurement devices, like single point detectors, which are well understood and known to have only small deviations in important linac characterization (e.g. small dose measurements).
The results of this study concerning the linac characteristicsespecially during the ramp upmay be prone to a single misadjusted linac. Therefore, main results were verified on two matched linacs and showed the same results. Still the instabilities during ramp up might be limited to the construction of the specific type of linac used and should be verified for each single linac. Consequently, if limitations are found, these should be respected using either QA method (plan individual or linac based QA). The presented measurements could easily be translated to other measurement devices; the implementation procedure for a linac based QA approach would not differ.

| CONCLUSION
Based on measurements with three different commercially available 2D arrays, we suggest linac based QA tests and how to derive threshold ranges for linac parameters, potentially influencing the delivery quality of IMRT and VMAT plans, which should be carefully considered during beam modeling and individual plan generation.
These threshold ranges include restricting small MU per segment for step-and-shoot IMRT as well as fast traveling leaves at rapidly changing dose rates and small dose rates for dynamic IMRT techniques (dynamic sliding window and VMAT) due to nonlinearity of the beam homogeneity, especially during linac ramp up. Furthermore, the calibration point for the dose rate should not be at the maximal possible dose rate but around the most often used value of delivered IMRT and VMAT plans.
These parameters with their threshold ranges should be checked regularly together with the MLC positioning accuracy using a sweeping gap test as well as dynamic interplay tests that include various gantry rotation speeds and angles, various leaf speeds and various dose rates.
All used arrays are suitable for the suggested tests, even though small fields, steep dose gradients, low doses and beam profile homogeneity will be measured differently. Therefore, each array has its own limitations that need to be considered during plan individual QA as well as linac based QA. These limitations restrict the comparability of results measured with the different arrays with respect to plan individual QA as well as linac based QA, which makes the arrays potentially more suitable for constancy test, in general.

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