Evaluation of the truebeam machine performance check (MPC): mechanical and collimation checks

Abstract Machine performance check (MPC) is an automated and integrated image‐based tool for verification of beam and geometric performance of the TrueBeam linac. The aims of the study were to evaluate the performance of the MPC geometric tests relevant to beam collimation (MLC and jaws) and mechanical systems (gantry and collimator). Evaluation was performed by comparing MPC to QA tests performed routinely in the department over a 4‐month period. The MPC MLC tests were compared to an in‐house analysis of the Picket Fence test. The jaw positions were compared against an in‐house EPID‐based method, against the traditional light field and graph paper technique and against the Daily QA3 device. The MPC collimator and gantry were compared against spirit level and the collimator further compared to Picket Fence analysis. In all cases, the results from the routine QA procedure were presented in a form directly comparable to MPC to allow a like‐to‐like comparison. The sensitivity of MPC was also tested by deliberately miscalibrating the appropriate linac parameter. The MPC MLC was found to agree with Picket Fence to within 0.3 mm and the MPC jaw check agreed with in‐house EPID measurements within 0.2 mm. All MPC parameters were found to be accurately sensitive to deliberately introduced calibration errors. For the tests evaluated, MPC appears to be suitable as a daily QA check device.

gantry were compared against spirit level and the collimator further compared to Picket Fence analysis. In all cases, the results from the routine QA procedure were presented in a form directly comparable to MPC to allow a like-to-like comparison.
The sensitivity of MPC was also tested by deliberately miscalibrating the appropriate linac parameter. The MPC MLC was found to agree with Picket Fence to within 0.3 mm and the MPC jaw check agreed with in-house EPID measurements within 0.2 mm. All MPC parameters were found to be accurately sensitive to deliberately introduced calibration errors. For the tests evaluated, MPC appears to be suitable as a daily QA check device. It is the aim of this study to compare the MPC mechanical geometric checks against standard QA tests to provide the reader with a sense of how the MPC checks might compare to their standard QA tests. The study was performed over a longer period (4 months) than the Clivio study and provides an assessment of the MPC stability and sensitivity to drift of the linac systems being tested. Sensitivity is further examined by the use of deliberate changes to the MLC centerline offset and gap, offset of the collimator calibration and offset and change in calibration span of the gantry calibration. The study also attempts to provide standard QA results in a form that is directly comparable to the equivalent MPC test.

2.A | Materials
All measurements in this study were performed on a single Varian TrueBeam 2.0 STx linac fitted with an aS1200 EPID and 6 degree of freedom couch. The aS1200 EPID utilizes a 43 9 43 cm 2 panel with backscatter absorber plate between the detection panel and positioning arm. The detector matrix is 1280 9 1280 with a smaller 1190 9 1190 pixel region employed for Dosimetry (Integrated) imaging mode providing a 0.23 mm resolution when EPID is at 150 cm source to detector distance (SDD) as is used for the MPC geometric tests.

2.A.1 | MPC geometric checks
The MPC geometric tests utilize a series of kV and 6 MV images of the IsoCal phantom situated in a specific bracket on the IGRT couch top to assess: treatment/radiation isocenter size, coincidence of MV and kV isocenters, accuracy of collimator and gantry angles, accuracy of jaw and MLC leaf positions, and accuracy of couch positioning including pitch and roll. All measurements are highly automated and the user is simply required to setup the IsoCal phantom and bracket onto the treatment couch at position H2 and to beam-on. For the geometric tests, the system makes all required motions automatically and beams on when all is in position. Images are automatically analyzed at the TrueBeam console and results are presented with a pass/fail criteria applied. Functionality for presenting trends in results is also available in the package.

2.B.1 | Repeatability
Short-term repeatability of the MPC geometric tests was evaluated by taking five successive measurements and calculating the SD.

2.B.2 | Jaw position evaluation
The MPC check of jaw positioning is performed using an 18 9 18 cm field. Jaw edges are detected on the EPID and the result is calculated as the distance between the measured jaw edge and the center of rotation of the MLC, which is determined from a series of collimator rotated MLC defined fields. As such, the mea- positioning. The QA3 is a 2D array of ionization chambers and diodes. Following alignment to crosshairs or lasers, data are acquired from a single 20 9 20 cm 2 field at 100 cm source to surface distance (SSD) and are compared to baseline. The X and Y size and shift parameters refer to the position and size of the radiation field relative to the laser/crosshair used to setup the device. The edge of the radiation field is detected using an array of diodes positioned across the beam penumbra. Analysis of the radiation field size and shift parameters in combination allows assessment of each individual jaw position and these were then compared against MPC.
The traditional method of measuring jaw positions relative to the crosshairs using the light field and graph paper was also performed for comparison to MPC. For jaw positioning, MPC uses a tolerance of AE 1 mm for the X jaws and AE 2 mm for the Y jaws. Having two different tolerances is likely due to mechanical reasons and the fact BARNES AND GREER | 57 that the Y jaws are situated further from isocenter. The departmental monthly QA test uses a AE 1 mm tolerance and TG-142 recommends monthly AE 2 mm for symmetric jaws, which is the tolerance used for the Daily QA3 check.

2.B.3 | MLC position evaluation
The MPC MLC test utilizes a static MLC "comb" pattern whereby alternating leaves are set at either 5 mm or 35 mm. The leaf positions are measured using EPID and the position of each leaf is determined relative to the collimator rotation axis determined from a series of collimator rotated MLC fields. MPC reports both the mean and maximum measured offset for each MLC bank. As such, the  The traditional method the department uses to test gantry angle readouts is to use a calibrated spirit level on the collimator faceplate at the cardinal gantry angles. Tolerance is AE 0.5°. For comparison to MPC gantry absolute, the difference from nominal gantry angle at G0 was compared, while for the gantry relative comparison, the maximum variation in measured angle from the expected 90°for two adjacent cardinal angles is compared.

2.B.6 | Sensitivity of MPC to linac miscalibration
The MPC gantry, collimator, MLC, and jaw positioning checks were each tested for sensitivity to miscalibration of the relevant linac parameter. Firstly, the collimator was deliberately miscalibrated using the digital spirit level. Successive offsets in the collimator rotation calibration were induced based upon the spirit level to the order of AE 0.5°. MPC was run premiscalibration, after each miscalibration and finally at the end when the calibration was returned to optimal. The measured changes in MPC were compared to the expected from the spirit level.
The sensitivity of the MPC gantry tests was examined by inducing miscalibration to the gantry angle in two ways. Firstly, a systematic offset of 0.5°was introduced into the calibration alternately in both directions and secondly the span of the calibration was miscalibrated by 2°(0.55%) alternately smaller and larger. MPC was performed before and after each miscalibration and measured changes in MPC gantry absolute and gantry relative were compared to the expected from the spirit level. The sensitivity of the MPC jaw positioning was investigated using deliberate miscalibration of jaw positions based upon the traditional method using the light field and graph paper. The jaws were systematically adjusted by AE 2 mm for the Y jaws and AE 1 mm for the X jaws and recalibrated thus. The measured change in MPC and in-house EPID QA were recorded.

3.A | Repeatability
The results of Table 1 show how repeatable each of the MPC geometric tests were across five successive measurements. The MLC results of Table 1 are for the maximum and mean offsets across each MLC bank. MPC also reports results for each MLC leaf individually and the spread in repeatability across the leaves for each bank is presented in histogram form in Fig. 1.
The repeatability results of Table 1 show that for all MPC tests, the methods are repeatable to within 0.1 mm or 0.01°for all parameters at 1 SD. The results of Fig. 1 show that the repeatability of the individual MLC leaf positions is within 0.04 mm at 1 SD and that bank A is systematically more repeatable than bank B by approximately 0.01 mm.

3.B | MLC positioning
The results of   3.C | Jaw positioning show that the MPC measurement is within AE 0.2 mm of optimal for all measurements. As there is no obvious trend in the data for either method with any of the jaws then a mean and SD is meaningful.
These data are presented in Table 3. The measured difference between the in-house EPID method and the MPC measurement on the same day is presented in Fig. 5. Figure 5 shows that for all measurements, the in-house EPID method agreed with MPC to within 0.2 mm. Analysis using the t-test shows that MPC is in statistical agreement with the in-house EPID method for the Y jaws, but not for the X jaws (Y1: P = 0.08, Y2: P = 0.23, X1: P << 0.01, X2: As there is no obvious trend in the data, the mean and SD data for the QA3 were included in Table 3.  Over the period of the study, the jaw positions were also tested using the traditional method using the light field and graph paper placed at isocenter. The method allows only 1 mm measurement resolution and this was insufficient to make any meaningful comparison with MPC.

3.C.1 | MPC jaw position sensitivity to miscalibration
The results of Table 4 show that both the MPC and in-house EPID measured changes in jaw positioning are not always as expected.
Differences from expected up to 0.81 mm were recorded for MPC and up to 0.68 mm were recorded for the In-house EPID. However, agreement between MPC and in-house EPID was always within 0.13 mm.

3.D | Collimator
The results of Fig. 7 show the MPC collimator rotation offset and the Picket Fence total skew parameters were always within 0.12°of nominal. The mechanical QA method is limited by the collimator readout being limited to tenth of a degree resolution so over the     Fig. 9. Figure 9 demonstrates that the MLC bank skew contributes significantly to the measurement.

3.D.1 | MPC collimator sensitivity to miscalibration
The results of Table 5 show that for an offset in the collimator rotation calibration of the order of magnitude comparable to the MPC tolerance (AE 0.5°), the measured change in MPC agrees with the measured change on the spirit level to within 0.07°.

3.E | Gantry
The results of Fig. 10 show that from the beginning of May until the July 23, both the MPC relative and absolute measurements were relatively stable. In this period, MPC relative gantry measured a mean of 0.07 AE 0.07°(1 SD) and the MPC gantry absolute measured a mean of À0.17 AE 0.07°(1 SD). At July 23, there was an overnight jump in the gantry absolute results, which was thereafter stable with a mean of 0.07 AE 0.01 degrees (1 SD). This is a statistically significant change (t-test: P << 0.01).
During the periods before and after the jump observed on the July 23 in the gantry absolute data, there was no statistical difference in the gantry relative measurement according to the t-test (P = 0.06). In this period, the gantry relative had a mean of 0.08 AE 0.01°(1 SD). At August 6, the gantry readouts were recalibrated at which point the gantry absolute results returned to a mean of À0.16 AE 0.02°(1 SD), which is statistically equivalent to before the jump on the July 23 (t-test: P = 0.12). In the same period after the gantry readout calibration, the gantry relative measurement changed to a mean of 0.01 AE 0.04°(1 SD) and was no longer in statistical agreement to before the recalibration (t-test: P << 0.01). In this period, the results are seen to oscillate about zero, which is a known behavior (Varian MPC user guide p39 7 ).
The relatively coarse resolution of 0.1°for the mechanical QA gantry measurements makes meaningful comparison with MPC difficult. However, Fig. 10  absolute measurements jumps the mechanical measurement also jumps to the order of 3.5 times the measurement resolution and then returns back after the gantry recalibration on August 6. The mechanical gantry relative measurement agreed with MPC gantry relative to within twice measurement resolution over the entire measurement period.
3.E.1 | MPC gantry sensitivity to miscalibration Table 6 shows that when the gantry angle is miscalibrated by an offset up to 0.5°, the MPC gantry absolute measure is in agreement with the digital spirit level to within 0.05°and that the MPC gantry relative measure is insensitive to such a miscalibration. Table 7 shows that for a 2°miscalibration in the gantry calibration span, there is insignificant change in the gantry absolute measure. However, the gantry relative measure changed from initial by 0.79°and 0.88°, respectively.

4.A | Repeatability
The repeatability results of Table 1 are well inside the tolerances for all tests indicating that the tolerances are meaningful in that recorded fails are distinguishable from day to day variation. Figure 1 indicates that the positioning of MLC leaves in Bank A are in general more repeatable than those in Bank B. However, the difference in the means of the two banks of 0.01 mm is insignificant.

4.B | MLC Positioning
The measurement resolution of the EPID at 150 cm SDD of 0.23 mm is well within the tolerances for the MLC and jaw positioning of 1 mm. As such, the EPID resolution is considered suf- The results of Table 2

4.C | Jaw positioning
The results in Table 3   Over the measurement period of this study, no method indicated a fail in jaw position including the graph paper method.

4.D | Collimator
The statistical agreement between the MPC and Picket Fence colli-

4.E | Gantry
The results of Fig. 10 show that the MPC gantry absolute measure agrees well with spirit level measurements of gantry angle at G0. Figure 10 and Table 6 suggest that the MPC gantry absolute result rather than the gantry relative result is sensitive to an offset miscalibration of the gantry angle encoder. On the July 23, there was an overnight jump in the MPC gantry absolute readout by 0.24°. Upon investigation, it was found that the day before the jump service engineers were investigating an SF 6 gas leak on the linac in the vicinity of the primary gantry encoder. It is surmised that the encoder must have been accidentally bumped or some other way inadvertently adjusted during the service. The MPC gantry relative measure was relatively insensitive to this change.
On the TrueBeam system, the gantry encoder is calibrated using two measurements at G180 degrees (gantry head down) with the gantry approaching from opposite sides. The first miscalibration performed in this experiment was a systematic offset and the span of the encoder was not altered. This may explain why the gantry absolute measure was sensitive to the miscalibration while the gantry relative was not. The gantry absolute measure compares the angle of the beam axis at G0 to the axis of couch vertical motion. If the gantry encoder is miscalibrated, then the beam axis at G0 will not be vertical and hence the system will be sensitive to the miscalibration. As potential further work to this study, the next step for evaluation of MPC could be a multicenter study to evaluate the variation in MPC performance across multiple linacs. As part of such a study, the

| CONCLUSION
The MPC MLC mean and maximum offset, jaw positioning, collimator rotation, and gantry absolute and relative parameters have been evaluated for short-term repeatability, sensitivity to induced changes, and compared against the methods currently in use in the department's linac QA program over a period of 4 months. The MPC MLC tests have all been shown to be accurate and sensitive within clinical significance and to exceed the recommendations of TG-142 for daily linac QA. The results indicate that the MPC tests covered in this study could be used for linac daily QA as long as the system is properly commissioned, its limitations understood and the individual department satisfied as to its utility. A robust ongoing QA program for MPC is also required.

ACKNOWLEDG MENTS
The authors thank the Calvary Mater hospital Newcastle Radiation therapists for their MPC data collection. We would also particularly like to thank the head Biomedical engineer of the radiotherapy department Calvary Mater hospital Newcastle for his insights into linac operation and his assistance in safely miscalibrating the linac for the sensitivity tests.

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