Quality assurance of VMAT on flattened and flattening filter‐free accelerators using a high spatial resolution detector

Abstract Purpose This study investigated the use of high spatial resolution solid‐state detectors (DUO and Octa) combined with an inclinometer for machine‐based quality assurance (QA) of Volumetric Modulated Arc Therapy (VMAT) with flattened and flattening filter‐free beams. Method The proposed system was inserted in the accessory tray of the gantry head of a Varian 21iX Clinac and a Truebeam linear accelerator. Mutual dependence of the dose rate (DR) and gantry speed (GS) was assessed using the standard Varian customer acceptance plan (CAP). The multi‐leaf collimator (MLC) leaf speed was evaluated under static gantry conditions in directions parallel and orthogonal to gravity as well as under dynamic gantry conditions. Measurements were compared to machine log files. Results DR and GS as a function of gantry angle were reconstructed using the DUO/inclinometer and in agreement to within 1% with the machine log files in the sectors of constant DR and GS. The MLC leaf speeds agreed with the nominal speeds and those extracted from the machine log files to within 0.03 cm s−1. The effect of gravity on the leaf motion was only observed when the leaves traveled faster than the nominal maximum velocity stated by the vendor. Under dynamic gantry conditions, MLC leaf speeds ranging between 0.33 and 1.42 cm s−1 were evaluated. Comparing the average MLC leaf speeds with the machine log files found differences between 0.9% and 5.7%, with the largest discrepancy occurring under conditions of fastest leaf velocity, lowest DR and lowest detector signal. Conclusions The investigation on the use of solid‐state detectors in combination with an inclinometer has demonstrated the capability to provide efficient and independent verification of DR, GS, and MLC leaf speed during dynamic VMAT delivery. Good agreement with machine log files suggests the detector/inclinometer system is a useful tool for machine‐specific VMAT QA.

Commercial amorphous-silicon EPIDs have a pixel size of 0.392 mm 2 × 0.392 mm 2 and spatial resolution of 0.784 mm. 12,13 Recently, an EPID-based system 14 has been tested for VMAT QA. This system showed good agreement with a plan for all dynamic parameters including MLC motion, however, the readout of gantry angle information extracted from the On-Board Imaging system may not be considered as machine-independent. In addition, it has been reported that EPIDs have a non-linear response to low MU 15,16 in IMRT and VMAT deliveries and have shown discrepancies with respect to the expected dose in small MU fields. 17 An alternative method for VMAT QA is the use of machine log files. [18][19][20] During dynamic deliveries, the log files record cumulative MU index and positional information (gantry and MLCs) from machine encoders. 21 This information requires prior validation with an external device and is only retrieved once delivery is completed. Therefore, machine log files do not offer real-time linac-independent analysis. In this work, we report on the use of a high spatial resolution (0.2 mm) solid-state detector (DUO), operated in transmission mode which records the current generated in each pixel by the flux of photons and combined to an inclinometer to provide an independent measurement of the angular position of the gantry. The two datasets are synchronized with the sync pulse of the linac allowing for highly detailed timing information available to perform independent machine-specific QA for VMAT based on the CoP. The system is capable of assessing the DR and the GS as a function of gantry angle in both flattened and unflattened beams. Dynamic MLC movement is a major component of VMAT delivery. In fact, MLC leaf speed has been proven to have a greater impact on the accuracy of VMAT delivery in comparison to DR and GS, 22,23 and was found to be the main contributor to inaccuracies in MLC positioning. 6,18,24 Therefore, the DUO was replaced by another high spatial resolution array detector named Octa that allowed simultaneous speed assessment of multiple MLC leaves during arc deliveries. Both the DUO and Octa have a sensitive volume of 0.032 mm 2 which is substantially smaller than that of EPIDs. The DUO and Octa's submillimeter spatial resolution is essential in the precise evaluation of MLC-defined fields that are commonly used in VMAT plans while the millisecond time resolution allows finite and time-resolved evaluation of the DR profiles and GS. That in combination with the easy setup and simple calibration procedure provides a device with the capability to simplify the application of the CoP and its use in clinical practice over existing commercial designs. 25

2.A | Proposed QA instrumentation
The DUO [ Fig. 1(a)] is a monolithic silicon detector, consisting of 505 sensitive volumes arranged in two orthogonal linear arrays. Each volume has an area of 0.04 mm 2 × 0.8 mm 2 and the five central volumes intersecting the arrays are 0.18 mm 2 × 0.18 mm 2 in size. The sensitive volumes are equally spaced with a center-to-center distance of 0.2 mm giving the detector overall dimensions of 52 mm 2 × 52 mm 2 . The DUO has been characterized for machine-specific QA in small radiation fields produced by megavoltage-flattened beams during in-phantom studies. 28,29 To allow the speed evaluation of multiple MLC leaves, the DUO detector was replaced by another solid-state monolithic sensor named Octa. 30,31 The Octa displayed in Fig. 1  The detector was stacked between two 5 mm thick PMMA slabs with a recess in the slab on top of its active area and covered with an aluminum film to shield it from external light and electromagnetic noise. The detector was synchronized to a digital inclinometer by a fast data acquisition system 32 based on a Field Programmable Gate Array. The inclinometer was an ADIS16209 from Texas Instruments (TI -Nexville US) with a bi-directional accuracy of 0.1°and a resolution of 0.025°. 33 The inclinometer was attached to the linac head and calibrated against the linac gantry position encoder at 0°(IEC scale). The detector assembly was fixed to a custom mechanical adapter attached to a Varian accessory tray and lodged into the designated tray slot at a source-to-detector distance of approximately 60 cm (Fig. 2). This setup placed the central sensitive volumes of the detector perpendicular to the incident radiation beam during gantry rotation to eliminate any angular dependence of the detector's response. The vertical array was aligned with respect to the linac central axis using the smallest available rectangular radiation fields and Vernier micro-positioners that shift the detector with respect to the beam in the superior-inferior and lateral directions.

2.B | Detector calibration
The linearity of the DUO's response to irradiations that ranged from 2 to 1000 MU was investigated at fixed DRs of 300, 400 and 600 MU min −1 delivered on the 21iX with 6 MV flattened beams and 1200 MU min −1 delivered on the Truebeam with 10 MV unflattened beams and an MLC-defined field size of 10 cm 2 × 10 cm 2 at isocenter.
The calibration factors that convert the charge collected by the detector to delivered MU were obtained from the slope of the linear fit.

2.C | Mutual dependence of dose rate and gantry speed
The mutual dependence of DR and GS during VMAT deliveries was investigated with the use of the Customer Acceptance Plan (CAP).
The CAP test is a standard plan provided by Varian at acceptance across all centers and has been reported in the literature. 9 The plan is designed to demonstrate the linac's dynamic performance through the full range of motions within the manufacturer's specifica-

2.E | MLC leaf speed
The Octa detector has two linear arrays oriented at ±45°with respect to the superior-inferior axis of the MLCs. This geometrical arrangement can be exploited to provide simultaneous speed evaluation of multiple leaves. Figure 3 shows In order to evaluate the distance covered by the leaves, a correction factor was used to correct for the projection of the leaves on the detector plane with respect to isocenter.
The leaf speed was verified under static and dynamic gantry con-

3.B.1 | Varian Clinac 21iX
The plan delivers four distinct DR and GS combination sectors during the arc delivery and these are repeated on either side of the 0°g antry position (Fig. 5). The DUO measured all four combinations as well as changes in DR and GS between control points.  The effect of vibrations on the response of the inclinometer can be reduced using a low-pass filter. The inclinometer provides a few tools to achieve this, however, the command settings required to activate these functions have not been used and would entail further development of the electronics' readout.
Deviations in the speed were calculated as percentage differences between the Octa and the trajectory log files data.  the two parameters to achieve the desired dose delivery during VMAT.
Variations in the recorded DRs between the proposed system and the machine log files tended to be at the transition between control points. In Figs  On the Truebeam, the linac exhibited a better compliance with the maximum speed limit. Instead of attempting to achieve the nominal leaf motion of 2.80 cm s −1 , the Truebeam modulated the DR from 400 to 355 MU min −1 in order for the MLCs to run at the specified maximum speed of 2.50 cm s −1 . That is, a reduction in the DR accompanied the reduction in the MLC leaf speed in order for the correct dose to be delivered during that interval. Since the leaves did not move at a speed that exceeded their mechanical limits, the effect of the gravity on the movement of the MLCs was therefore negligible.
The MLC leaf speed assessed under dynamic gantry conditions showed agreement with the trajectory log files data with percentage differences that ranged between 0.9% and 5.7% (Table 1). The error in the MLC leaf speed was found greater in the arc sectors of lowest DR and highest MLC leaf velocity. This may be due to the corresponding low detector signal in these sectors.
The reconstruction of DR and GS as a function of gantry angle using the proposed system has allowed independent verification of VMAT delivery components with comparison to the machine log files

| CONCLUSION
This study investigated the use of solid-state array detectors combined with an inclinometer to perform machine-specific VMAT QA