Assessment of combined use of ArcCheck® detector and portal dosimetry for delivery quality assurance of head and neck and prostate volumetric‐modulated arc therapy

Abstract Purpose To assess the efficiency of combined use of ArcCheck® detector (AC) and portal dosimetry (PDIP) for delivery quality assurance of head and neck and prostate volumetric‐modulated arc therapy. Materials and methods Measurement processes were studied with the Gamma index method according to three analysis protocols. The detection sensitivity to technical errors of each individual or combined measurement processes was studied by inserting collimator, dose and MLC opening error into five head and neck and five prostate initial treatment plans. A total of 220 plans were created and 660 analyses were conducted by comparing measurements to error free planned dose matrix. Results For head and neck localization, collimator errors could be detected from 2° for AC and 3° for PDIP. Dose and MLC errors could be detected from 2% and 0.5 mm for AC and PDIP. Depending on the analysis protocol, the detection sensitivity of total simulated errors ranged from 54% to 88% for AC vs 40% to 74% for PDIP and 58% to 92% for the combined process. For the prostate localization, collimator errors could be detected from 4° for AC while they could not be detected by PDIP. Dose and MLC errors could be detected from 3% and 0.5 mm for AC and PDIP. The detection sensitivity of total simulated errors ranged from 30% to 56% for AC vs 16% to 38% for PDIP and 30% to 58% for combined process. Conclusion The combined use of the two measurement processes did not statistically improve the detectability of technical errors compared to use of single process.


| INTRODUCTION
Volumetric-modulated arc therapy (VMAT) has become widely used, especially for head and neck (H&N) and prostate cancer treatments.
VMAT dose distribution depends on the dose rate modulation, arm movement speed, multi-leaf collimator (MLC) position, and collimator angulation. 1 The large number of degrees of technical freedom allows a significant preservation of organs at risk. [2][3][4][5] However, the complexity of this technique requires verification of the planned dose distribution by performing quality control (QC) treatment plans.
These QCs are performed prior to treatment and consist of checking the concordance between planned dose distribution and real distribution delivered by the linear accelerator. Different measurement systems are available: the electronic portal imaging device (EPID), 2-D or semi 3-D independent detector, or ionization chamber. The dose distribution analysis is generally performed with the Gamma index method with highly versatile analysis protocols regarding the choice of the deviation criterion of dose and acceptable distance, the selection of the pixels to be analyzed, the normalization method, and the Gamma pass rate criterion. [6][7][8][9] Depending on the experience curve acquired by the teams, these different verification methods can be used individually or in combination, which can make the pretreatment verification process very time-consuming. In this context, the purpose of this study was to determine the efficiency of combined use of ArcCheck ® detector (Sun Nuclear, Melbourne, FL, USA) and portal dosimetry with EPID As 1200 ® (Varian Medical System, Palo Alto, CA, USA) for H&N and prostate VMAT delivery quality assurance. The detectability threshold of both measurement processes to different types of technical errors was studied using the Gamma index method according to three analysis protocols. Then, the error detection sensitivity of each individual and combined measurement process was compared.   For both AC and PDIP processes, the treatment plan was computed in the ArcCheck ® or EPID As 1200 ® image to obtain the planned dose or CU matrix (Fig. 1).

2.C | Dose matrix concordance analysis
Planned and delivered dose matrix comparisons were performed using the Gamma index method proposed in 1998 by Low et al. 10 Three combinations of parameters for the Gamma index found in the literature were used and are detailed in Table 1. 11,12 The dose difference criteria (DD) and the distance to agreement criteria (DTA) match, respectively, the difference in dose and distance accepted.
Global and local mode of dose normalization (Mode) were studied.
The threshold pixels criterion (TH) in terms of percentage of the maximum dose for AC and area of the MLC complete irradiated area outline (CIAO) for PDIP was evaluated. Thresholding of all the pixels was made at a dose greater than or equal to 10% (10% Dmax) and 20% (20% Dmax) of the maximum dose for AC. For PDIP, we used our historical threshold criteria MLC CIAO + 1 cm which corresponds to the opening envelop of the MLC incremented by 1 cm.
The passing rate (PR) for the plan to be considered to conform corresponds to the minimum percentage of pixels having a Gamma index less than one (GI <1).

2.D | Assessment of process sensitivity for various potential technical errors
To assess the detection sensitivity of AC and PDIP processes, the following potential technical errors were simulated: • Collimator error (from 1°to 5°, increment of 1°) • Dose error (equal to +2% and +3%) • MLC opening error (equal to +0.5 mm, +1 mm and +2 mm)  performed. The average detection threshold of each type of error was determined for both measurement processes as follows: mean PR for each type and level of error were calculated. If the mean PR was below the protocol tolerance limit studied, then the level and type of error could be detected in agreement with the AAPM report TG-218 which reports the risk of systematic errors when the plan does not pass the tolerance limit. 9 The error detection sensitivity by type and cumulative was also calculated for each individual process (AC or PDIP) (δError individual process ) and for the combined measurement process (AC + PDIP) (δError combined processes ) according to formulae 1 and 2. The error was considered detected when the plan no longer respected the PR. For the combined process, we cumulated all errors detected by at least one of the two systems.
The statistical comparison of error detection sensitivity was performed by Chi-squared test and a P-value <0.05 was considered statistically significant.

| RESULTS
Dose calibration verification for AC and PDIP showed a discrepancy between the calculated and measured dose at the isocenter less than 0.5%. The mean PR of reference plans (ref) and plans with collimator, dose, and MLC errors for AC and PDIP processes according to the three Gamma index protocols and localization are presented in Fig. 2. For H&N localization, collimator errors could be detected from 2°for AC (mean PR ± SD = 92.5% ± 2.9%) and 3°f or PDIP (86.7% ± 6.7%) with 3%/3 mm/Local analysis protocol.
For prostate localization, this rate ranged from 30% to 56% for AC vs 16% to 38% for PDIP and 30% to 58% for combined process ( Table 3).
Regardless of the measurement process, the H&N error detection sensitivity was statistically superior compared to prostate (Pvalue <0.01).
Irrespective of the localization, AC with 3%/3 mm/Local analysis protocol could detect all the MLC errors plans. Moreover, AC allowed a significant improvement of error detection sensitivity compared to PDIP (P-value <0.01). Regardless of the analysis protocol used, the combined measurement process did not statistically improve errors detection sensitivity compared to AC for H&N (Pvalue = 0.26) and prostate (P-value = 0.62). PDIP 3%/3 mm/Local analysis protocol did statistically improve errors detection sensitivity compared to combined process with 3%/3 mm/Global analysis protocol for H&N (P-value = 0.01) and was equivalent for prostate (Pvalue = 0.23). Similarly, AC 3%/3 mm/Local analysis protocol did statistically improve errors detection sensitivity compared to combined process with 3%/3 mm/Global analysis protocol for H&N (P-value <0.001) and prostate (P-value <0.001).
T A B L E 1 Concordance analysis protocol for planned and measured dose matrices. DD is the accepted dose difference. DTA is the distance difference accepted. Mode corresponds to the dose normalization mode. TH corresponds to the thresholding pixel criterion. 10% Dmax and 20% Dmax correspond to a thresholding of all the pixels, with a dose greater than or equal to 10% and 20% of the maximum dose of the plan. MLC CIAO + 1 cm correspond to a threshold of all the pixels included in Complete Irradiated Area Outline (CIAO) of the MLC incremented by 1 cm. PR is the minimum success criterion on the pixel percentage with a Gamma index less than one for the plan to be considered compliant.

| DISCUSSION
VMAT pretreatment delivery quality assurance is an essential step before final approval of the treatment plan. Depending on the verification process protocol, this step can be very time-consuming. Previous studies have evaluated the sensitivity of detection of potential technical errors by different measurement processes and Gamma analysis protocols. [11][12][13][14][15] The goal of this study was to assess the efficiency of combined use of AC and PDIP for H&N and prostate VMAT delivery quality assurance.
Regardless of the localization, the mean detection threshold of errors appeared equivalent between the two measurement processes with the optimal analysis protocol. Only the collimator error could be detected earlier from 2°for the H&N localization and 4°for the prostate localization by the AC (Fig. 2). The results of this study concerning the detection threshold of simulated technical errors seem consistent with the litterature, since Vieillevigne et al. found a detection threshold for a 2%/2 mm/Global analysis protocol with AC from 2°for collimator errors and 0.5 mm for MLC errors. 11  improve the error detection sensitivity. In our study, the combined process increased the total error detection sensitivity of AC from +0% to +4%, depending on analysis protocol and localization. This result is in agreement with Fredh et al. who showed a marginal benefit in the use of multiple detectors. 13 Moreover, none of the processes (AC, PDIP, or combined) allowed systematic detection of all errors. This result is in agreement with the study of Kry et al. which highlighted the low power of pretreatment QC. 16 Finally, the individual use of the AC or the PDIP with an optimal analysis protocol made it possible to obtain an error detection sensitivity equal to or greater than combined measurement process. For example, the individual use of AC or PDIP with 3%/3 mm/Local analysis protocol had an error detection sensitivity equal to or greater than the combined process with 3%/3 mm/Global analysis protocol (Tables 2 and 3).
This study highlights the importance of optimizing the analysis protocols, particularly according to the localization and to the measurement process in order to find the right balance between error detection sensitivity and false positive risk.

| CONCLUSION
This work showed that the combined use of AC and PDIP did not significantly improve the error detection sensitivity compared to use of a single process. None of the measurement processes used individually or in combination allowed systematic detection of all errors. The analysis protocols optimization of each measurement process appeared necessary in order to obtain an optimal error detection sensitivity.