An electronic portal image device (EPID)‐based multiplatform rapid daily LINAC QA tool

Abstract Purpose To develop an efficient and economic daily quality research tool (DQRT) for daily check of multiplatform linear accelerators (LINACs) with flattening filter (FF) and flattening filter‐free (FFF) photon beams by using an Electronic Portal Image Device (EPID). Materials and Methods After EPID calibration, the monitored parameters were analyzed from a 10 cm × 10 cm open and 60° wedge portal images measured by the EPID with 100 MU exposure. Next, the repeatability of the EPID position accuracy, long‐term stability, and linearity between image gray value and exposure were verified. Output and beam quality stability of the 6‐MV FF and FFF beams measured by DQRT with the introduced setup errors of EPID were also surveyed. Besides, some test results obtained by DQRT were compared with those measured by FC65‐G and Matrixx. At last, the tool was evaluated on three LINACs (Synergy, VersaHD, TrueBeam) for 2 months with two popular commercial QA tools as references. Results There are no differences between repeatability tests for all monitored parameters. Image grayscale values obtained by EPID and exposure show good linearity. Either 6 MV FF or FFF photon beam shows minimal impact to the results. The differences between FC65‐G, Matrixx and DQRT results are negligible. Monitor results of the two commercial tools are consistent with the DQRT results collected during the 2‐month period. Conclusion With a shorter time and procedure, the DQRT is useful to daily QA works of LINACs, producing a QA result quality similarly to or more better than the traditional tools and giving richer contents to the QA results. For hospitals with limited QA time window available or lack of funding to purchase commercial QA tools, the proposed DQRT can provide an alternative and economic approach to accomplish the task of daily QA for LINACs.


| INTRODUCTION
Daily Quality Assurance (QA) is the most frequently performed procedure among all the QA procedures on linear accelerator (LINACs) and has a direct impact on the performance of the IMRT procedure.
Existing daily QA methods are usually time-consuming. For metropolitan hospitals in China, one of the major challenges for daily LINAC QA is the limited time window available, since most LINACs are overloaded with cancer patients: Taking the LINACs at our hospital as example, the average number of procedures performed per day is 90-120 per system, compared with 25-50 in Europe and 13-25 in the United States. [1][2][3] Therefore, there is a strong clinical need to develop a more efficient and cost-effective QA solution.
EPID has been introduced to LINAC systems since the early 1980s. 4 It was initially developed for verifying the patient position and was later applied to LINAC QA. [5][6][7][8][9][10][11][12] Compared with films and other QA devices, EPID has two major advantages [13][14][15] : first, since EPID is integrated with the LINAC gantry, the QA procedure can be setup more quickly; second, the EPID data are in a digital form, The purpose of this work is to develop and assess a rapid EPIDbased multiplatform daily LINAC QA tool to meet the imperative clinical need for efficient and cost-effective daily LINAC QA procedure. Note that using EPID for daily LINAC QA is not a new concept, as several previous works have been reported over the past decades, both for photon beams [16][17][18][19] and electron beams. 12,20,21 For example, Clivio et al. 22 and Michael et al. 23  dors. In addition, MPC was developed by Varian and tightly integrated with their LINACs. However, it is often more desirable to have a daily QA tool that is independent of the LINAC vendor to facilitate the intersystem comparisons. 24 In this study, both FF and flattening filter-free (FFF) photon beams are covered, and the robustness of the proposed EPID-based QA method was evaluated across different LINAC vendors and models. To be more specific, a rapid EPID-based daily QA tool entitled DQRT (Daily QA Research Tool) was developed in this work, and its reliability was verified using multiple Elekta (Synergy, VersaHD) and Varian (TrueBeam) LINAC systems. The tool can be used to evaluate dose constancy, beam quality (BQ), flatness (F) and symmetry (S), center of field, and field size accuracy. The proposed method was used to perform daily QA for both the 6-MV FFF and FF photon beams. In addition to the cross-platform and multienergy features mentioned above, the theory of this tool is straightforward and robust which make the tool have a strong reliability. In this study, we first demonstrate the physical principle and parameter calculation process of DQRT, then the stability and accuracy are assessed by comparing with other devices.
At last, the clinical performance of DQRT was compared with Dai-lyQA3 and Beamcheck in different linacs, which showed that the DQRT is more convenient for daily QA of Linacs.

2.A | Calibration of EPID
In order to ensure the accuracy and precision of DQRT, we performed calibration of the EPID system of each LINAC via the following procedures.
For the Elekta LINAC, the iViewGT system was used. The MV detector of the iViewGT system has a fixed source-to-skin distance (SSD) of 160 cm and can only move along the longitudinal and lateral directions. This EPID has been calibrated using the following procedures 25 : First, the mechanical accuracy of EPID was verified.
The longitudinal movement was operated with a handheld controller in order to make sure that the reduced-speed, isocenter pause and longitudinal limit work correctly. Next, the MV detector was moved to the isocenter position, and the mechanical pointer on the longitudinal scale was checked to make sure it points to the isocenter mark within AE 1 mm. For the lateral direction, the same validation step was performed. Then, the offset correction and the first radiation synchronization calibration were auto-executed by the iViewGT system. Next, a gain calibration was performed at the zero degree gantry angle and collimator angle, the maximal dose rate, 26 cm × 26 cm field size, and an exposure dose of 100 MU. After the gain calibration, a bad-pixel map was applied to help correct those pixels known to give inconsistent responses. Finally, a second radiation synchronization calibration was performed in order to make sure that the image has a clear contrast. After these steps were completed, image scaling was executed starting from an exposure image with a field size of 15 cm × 15 cm and an exposure dose of 10 MU. Then the horizontal size of the exposed field was measured used the iViewGT measure tool. After being divided 150 by the horizontal size, the result was set to the "HorizMMPixel" value that stored in the field with a file name of sri.ini. The DF image was acquired without radiation and was averaged over a series of measurements to provide the pixel offsets. The flood field image was acquired by irradiating the detector panel with an open field covering the entire imager. The measurements were averaged over a fixed number of frames to determine the mean difference in individual pixel sensitivity. Dosimetry calibration was applied based on the recommendation from Varian; the beam diagonal profile measured at D max in water for the 40 cm × 40 cm open field was used to scale the off-axis pixel response, and dose normalization is performed to set the calibration unit (CU) of the portal dose equal to the clinical dose unit (cGy). Finally, a pixel correction map generated from the approved DF and flood field images was applied to the TrueBeam system. 26

2.B | Image acquisition and processing
All measurements of the Elekta LINACs were carried out using the EPID in its default position without any additional build-up. The source-to-detector distance (SSD) was kept at 160 cm. The detector signals were read out with a fixed integration time of 250 ms for each image frame. After the exposure, all images were added up over the entire exposure time and integrated into a 64-bit buffer. At last, an arbitrary scaling factor is included to be able to store the signal data in a 16-bit format and a final portal dose image with *.tif format was produced after the standard image processing procedures was performed by iViewGT automatically.
The measurements of TrueBeam were performed in the clinical mode and the integrated images were acquired when the test beams were executed. The SDD was kept at 100 cm without any other additional build-up in all tests. The "PortimageIntegrated" mode designed for dosimetric application was selected to get the integrated image. In this mode, the dose for per frame image keeps constant and the image frame readout time is approximately 110 ms.
The data flow began with the beam being triggered by the Truebeam supervisor module. Next, the MV image acquisition module was triggered by the MV imager and the detector began to acquire the frames. The digitized images were processed and corrected by an XI software based on the stored calibration set in the software.
At last, the image encoded in a standard DICOM RT image was exported from the imaging workstation. 26

2.C | Test and evaluate methods
Based on the recommendations of AAPM TG-142, 27 the DQRT mainly focuses on the stability of output, BQ, F and S, center of field, and field size.
For the first step of the proposed QA procedure, two EPID images were acquired when the gantry angle was at zero: one (I1) with an open field and the other one (I2) with a 60°wedge. Both images were obtained with 10 cm × 10 cm field size and 100 MU.
When the DQRT was first applied to a LINAC, the LINAC was adjusted to its best state, and the values of monitored parameters mentioned above were obtained by DQRT at this time served as the benchmark data. In each work day, I1 and I2 images were reacquired using the same setup mentioned above. The output, BQ, F and S, center of field, and field size of each day were obtained from I1 and the benchmark data need be reacquired. If none of the above has happened, the benchmark data are reacquired each year.
Mean grayscale pixel value of a 10 mm × 10 mm central region in each image obtained by DQRT is defined as μ. The stability of output was evaluated based on μ value of the image I1. The profiles along X and Y axes were calculated from the average grayscale values of ten rows in the field center. Next, the field edges were determined by points whose first derivative values along X or Y profile are maximal and the second derivative values of them are zero.
Then, the field sizes along the X and Y directions were calculated and the intersection of the field diagonal lines was considered as the field center. For the F and S, they were quantified based on the X and Y profiles of I1. Formulas 1 and 2 are the formulas used to calculate F and S. The BQ, which could be reflected by wedge factor W, was determined from μ wedge and μ open using Formula 3.
In the Formulas above, D max , D min , D center represent the maximal dose, minimal dose, central dose, respectively; D left ,D right are the dose of two points which are symmetric about the field central axis.
All of the points above are selected from X or Y profile and within 80% region of the field size. The readout step of the DQRT for the field is illustrated in Fig 1. 2.D | Short-term stability of EPID   3; "S,""T,""V,""O,""W,""X,""Y,""6,"and "F" represent the Synergy, TrueBeam, VersaHD, Open field, Wedge field, left-right (LR) direction, gantry-target (GT) direction, 6 MV beam, 6 FFF MV beam, respectively. For example, VFWX means LR direction coordinate of the beam center of a 6-MV FFF photon field with wedge measured in VersaHD LINAC system. For GT direction, coordinate of TrueBeam, a value 128 is added.
T A B L E 1 Mean offset between the EPID and beam center measured over time.    Fig. 2 and Fig. 3.

3.B | Long-term stability of EPID
The variation of radiation center of each LINAC measurement by the EPID was shown in Table 1 (Fig. 7) and a strong correlation was found between them ( Pearson's correlation coefficient is 0.71). X and Y field size of the open field detected by DQRT are within F I G . 6. Normalized beam quality (a) and output (b) of the TrueBeam 6 MV X ray measured by DQRT and FC65-G during 2o weeks.

3.E | Validate DQRT
T A B L E 2 The adjacent field size variation measured by DQRT and Matrixx two times each. All the unit in this table are mm. for inline (Fig. 8) Fig. 9(c)]. For each test series, the change of beam center position is within three pixels (Fig. 3).

| DISCUSSION
As both Clivio et al. 22 and Sun et al. 17  EPID-based daily QA mainly focused on a particular type of LINAC and FF photon beams. 23  effects. 29 As regardse the effect of FFF photon beam distribution, the sensitivity of DQRT to the position errors has ruled out the concern.
For the repeatability test, one problem should be addressed: the maximum variation of output measured by EPID is 0.10% for the repeatability measurement, which is much larger than the fluctuation of the ionization chamber, such as 0.03% for FC65-G chamber obtained by ten times repeatability measurement. Therefore, the larger fluctuation of EPID may introduce a larger error than the recommendation of TG142. In order to avoid this, a tighter tolerance AE 2.5% is selected as the dose constancy standard in our study.
Another problem revealed by the results is, when the exposure is small, parameters such as field size, F and S fluctuated strongly during the repeatability measurement. For the Synergy LINAC, the fluctuation scope of the parameters detected by DQRT is close to 5% when the exposure is smaller than 10 MU, and the scope decreased to 1% when the exposure increased to 10 MU-80 MU. When the exposure becomes larger than 80 MU, the scope reduces to less than 0.1%, and it can be considered that the measurement stability of each parameter no longer changes with the increase of exposure.
Therefore, in clinical practice, exposure for daily QA using EPID is suggested to be greater than 80 MU. However, too much exposure would have a negative effect on the EPID detectors, so a larger number of MU is not recommended. Considering the above reasons, 100 MU is selected as the exposure in DQRT.
Results of clinical size measurements show that, field sizes of Another issue needs to be addressed is that, for the DailyQA3 and the QABeamCheckerPlus, the definitions of BQ, F and S are different from those used in the DQRT. As a result, the measured values between them have obvious difference, but the general trend of these parameters is consistent [Figs. 7 and 8], which means that the DQRT has the same accuracy and reliability as the DailyQA3 and QABeamCheckerPlus. In DQRT, though the measured values of the parameters are not confirmed with the actual values, they are also used to reflect the trend of corresponding parameters. This could be done because the parameters calculated in DQRT have a stable relationship with the parameters in reality, and because the daily QA in clinical practice focuses primarily on the data stability.
The research suggests that the DQRT could be used for daily QA of electronic beam. For electron BQ measurement, a glass with appropriate thickness could be chosen as a substitution of the wedge, and the glass can be placed on the surface of EPID detector.
Other tests for electron beam are similar to the FF photon beam.
In addition, a region of 40 pixel × 40 pixel located at the field center is selected to measure the output in this study, which is the same with Sun etc. al. 24 For different LINACs, the size of the pixel may be different. Hence, in order to keep the balance between precision and anti-interference, appropriate region size should be chosen for different systems.

| CONCLUSIONS
Several EPID-based QA methods have been proposed, including inhouse tools and commercial products, such as MPC. However, these schemes have some disadvantages, such as generality and independence inadequate. This work demonstrates that the EPID-based tool (DQRT) is capable of daily QA for multiple LINAC systems with and without flattening filter and is suitable for each type of LINAC. At the same time, it can monitor more aspects of the LINAC performance than conventional devices within a shorter period of time with stable and reliable results. In addition, the only additional equipment required for this method is a conventional computer, and thus the proposed method is very cost effective. In summary, DQRT could serve as a low-cost and highly efficient daily QA tool. The DQRT is vital for detecting potential mechanical and dose issues due to the increasing demand of integration and automation of QA works.