Comparison of DVH‐based plan verification methods for VMAT: ArcCHECK‐3DVH system and dynalog‐based dose reconstruction

Abstract The purpose of this study was comparing dose‐volume histogram (DVH)‐based plan verification methods for volumetric modulated arc therapy (VMAT) pretreatment QA. We evaluated two 3D dose reconstruction systems: ArcCHECK‐3DVH system (Sun Nuclear corp.) and Varian dynalog‐based dose reconstruction (DBDR) system, developed in‐house. Fifteen prostate cancer patients (67.6 Gy/26 Fr), four head and neck cancer patient (66 Gy/33 Fr), and four esophagus cancer patients (60 Gy/30 Fr) treated with VMAT were studied. First, ArcCHECK measurement was performed on all plans; simultaneously, the Varian dynalog data sets that contained the actual delivered parameters (leaf positions, gantry angles, and cumulative MUs) were acquired from the Linac control system. Thereafter, the delivered 3D patient dose was reconstructed by 3DVH software (two different calculating modes were used: High Sensitivity (3DVH‐HS) and Normal Sensitivity (3DVH‐NS)) and in‐house DBDR system. We evaluated the differences between the TPS‐calculated dose and the reconstructed dose using 3D gamma passing rates and DVH dose index analysis. The average 3D gamma passing rates (3%/3 mm) between the TPS‐calculated dose and the reconstructed dose were 99.1 ± 0.6%, 99.7 ± 0.3%, and 100.0 ± 0.1% for 3DVH–HS, 3DVH–NS, and DBDR, respectively. For the prostate cases, the average differences between the TPS‐calculated dose and reconstructed dose in the PTV mean dose were 1.52 ± 0.50%, −0.14 ± 0.55%, and −0.03 ± 0.07% for 3DVH–HS, 3DVH–NS, and DBDR, respectively. For the head and neck and esophagus cases, the dose difference to the TPS‐calculated dose caused by an effect of heterogeneity was more apparent under the 3DVH dose reconstruction than the DBDR. Although with some residual dose reconstruction errors, these dose reconstruction methods can be clinically used as effective tools for DVH‐based QA for VMAT delivery.


| INTRODUCTION
Recently, volumetric modulated arc radiotherapy (VMAT) has become a routine technique in many facilities. Although this technique improves the conformity of dose distribution to PTV and reduces the impact on OARs, its use in a complex dose distribution with a sharp gradient necessitates patient-specific quality assurance (QA). The most frequently employed method for QA has been comparison of the calculated and measured doses in a phantom. Particularly, the point dose on an ion chamber and planar dose distribution on a film are usually measured. In general, gamma analysis has been used to compare measured and calculated dose distributions in a commercial radiation treatment planning system (TPS). 1 However, these conventional patient-specific QA procedures are very time consuming for the clinical staff. In addition, some previous studies showed that gamma analysis cannot directly predict the actual patient dose. 2,3 To tackle these problems, some independent dose reconstruction methods have been proposed to evaluate patient dose-volume histogram (DVH) and dose index for VMAT pretreatment QA. One of the methods is called measurement-guided dose reconstruction (MGDR), a system which is commercially provided as MatriXX-COM-PASS (IBA Dosimetry, Schwarzenbruck, Germany), Delta4 anatomy (ScandiDos, Inc., Ashland, VA, USA), and ArcCHECK-3DVH (Sun Nuclear Corporation, Melbourne, FL, USA). The ArcCHECK-3DVH system delivers a 3D patient dose that can be reconstructed using 3DVH software from the original TPS plan and ArcCHECK measurement data. The reconstructed dose from this system could be compared with the TPS-calculated dose using 3D gamma and DVH dose index analyses. The accuracy of the system has already been investigated in several studies. [4][5][6] Another method is machine log file-based dose reconstruction.
Some have reported the use of log files generated by multi-leaf collimator (MLC) controller as a tool for DVH-based dose verification for patient-specific QA of intensity-modulated radiotherapy (IMRT). 7 In VMAT, on the other hand, machine log files that contain other dynamic parameters (gantry angles and cumulative MU) are generated from a Linac control system. Some studies demonstrated dose reconstruction methods that modify TPS plan data using delivered parameters from these log files and recalculate by TPS dose calculation algorithm. 8

2.B | ArcCHECK-3DVH system
The ArcCHECK-3DVH system (Sun Nuclear Corporation, Melbourne, FL, USA) was commercially available tool for DVH-based QA. ArcCHECK was cylindrical 3D diode array, which contained 1386 diodes (detector sixe: 0.8 9 0.8 mm 2 ) in a helical arrangement at intervals of 10 mm and with diameter of 21 cm. To reconstruct an "actual" 3D patient dose from the measured ArcCHECK data, we used 3DVH software ver. 3.2 that had an internal calculation engine, which was called ArcCHECK planned dose perturbation (ACPDP). To perform ACPDP, the following data set were pre-  The 3DVH-HS dose morphing is recommended for detecting even very small deviations from ideal behavior.

2.C | In-house dynalog-based dose reconstruction system
In VMAT, two sets of Varian dynalog were generated. One was beam delivery dynalog, which was created by the Linac console and contained information delivered from the dynamic beam (e.g., the actual cumulative dose delivered (MU) versus the actual gantry angle); these parameters were only recorded for each control point. The other log was the MLC dynalog, which was created by the MLC controller. The file was separately generated and acquired every 50 ms for the MLC banks A and B. Details of the MLC dynalog have been described elsewhere. 13 In this study, dynalog-based dose reconstruction (DBDR) was

2.D | Validation of the in-house DBDR system
Before using the in-house DBDR system for patient-specific QA, the system was validated by a method similar to the one used by Juan et al. to check for programming errors. 7 For a baseline plan (Singlearc prostate VMAT, 2.6 Gy/1 Fr), nine MLC error plans were generated (Table 1). To measure the absolute dose at the center of the phantom, these plans were delivered by a 15-MV X-ray beam of Varian 23EX with a 120 millennium MLC to an ArcCHECK phantom with a customized acrylic plug that was holding a 0.6-cc PTW 30013 Farmer ionization chamber. Absolute isocenter dose was used to evaluate the accuracy of the in-house DBDR system.

2.E | Workflow and Analysis of DVH-based QA
A schematic design of this study is shown in Fig. 1. First, ArcCHECK QA plans were created from the original plans for all patients. Second, ArcCHECK measurement (ArcCHECK was calibrated with 200 MU with a 10 9 10 cm 2 field size at a gantry angle 0°before plan irradiation) was performed on all plans; simultaneously, the Varian dynalog data sets that contained the actual delivered parameters (leaf positions, gantry angles, and cumulative MUs) were acquired from the Linac control system. Thereafter, the delivered 3D patient dose was reconstructed by 3DVH software and in-house DBDR system. We evaluated the differences between the TPS-calculated dose and the reconstructed dose using whole body 3D gamma passing rates (3%/ 3 mm, 2%/2 mm, and 1%/1 mm with global normalization, threshold 10%) and DVH dose index analysis. For the DVH analysis of the prostate case, PTV doses (mean dose, D95% and maximum dose) and rectum wall and bladder wall dose (mean dose, V35 and V55) were evaluated. For the whole neck case, PTV doses (D50% and maximum dose), brainstem and spinal cord doses (maximum dose), and parotids dose (mean dose and maximum dose) were evaluated. For the esophagus case, PTV doses (D50% and maximum dose), spinal cord dose (maximum dose), and lungs doses (mean dose and V20) were evaluated. In addition, we calculated the clinically effective confidence limit values for each DVH dose index using the following eq. (1) Confidence limit ¼ jDD mean j þ 1:96 DD SD (1) where DD mean was the average dose differences between the TPS-calculated dose and the reconstructed dose and DD SD was the standard deviation. All 3D analyses were performed in 3DVH software ver.3.2.

3.A | Validation of the in-house DBDR System
The results of validation of the in-house DBDR system are summarized in

| DISCUSSION
In this study, before comparing two DVH-based QA methods that were the ArcCHECK-3DVH system and the in-house DBDR system, we validated the in-house DBDR system using MLC error plans, showing that the system was a well-developed DVH-based QA tool.
Thereafter, we compared these methods using 3D analysis, showing that the ArcCHECK-3DVH system had some differences from the in-house DBDR system.
The accuracy of the ArcCHECK-3DVH system has been vali- reported that 3DVH-NS was better than 3DVH-HS in terms of dose reconstruction accuracy. 6 These findings were consistent with our results in the present study. Particularly, compared with 3DVH-HS, the 3DVH-NS had a dose distribution that was in good agreement with the TPS-calculated dose distribution (Fig. 2). In addition, 3DVH methods use the ACPDP model parameters optimized in other facilities beforehand. Therefore, for DVH dose index analysis, the systematic errors for a specific DVH parameter were observed in 3DVH methods than DBDR (Fig. 3). In addition, we evaluated the method under the heterogeneous treatment sites (Figs. 5 and 6). The 3DVH reconstructed dose is calculated by dose ratio map between ACPDP and TPS based on the homogeneity cylindrical phantom. 12 Therefore, the effect of heterogeneity is not considered under the dose reconstruction, resulting in a change of the dose distribution especially in the heterogeneous region such as paranasal sinus and lungs.
T A B L E 3 DVH dose index analysis between TPS-calculated dose and each reconstructed dose (average AE SD, %), and confidence limits for all DVH parameters calculated by the eq. (1) in the text. head and neck, and 4 esophagus VMAT patients. We calculated the confidence limit for each DVH-based QA metrics (Table 3). In terms of tolerance for DVH-based patient-specific QA, Visser et al. suggested that action levels may clearly distinguish the role of the medical physicist and radiation oncologist during the QA procedure. 14 Our results indicate that these confidence limits may be used by medical physicists.

| CONCLUSION
The two DVH-based QA methods that we evaluated in this study had different dose reconstruction accuracies. Although with some residual dose reconstruction errors, these two methods can be clinically used as effective tools for DVH-based QA for VMAT.

CONF LICT OF I NTEREST
There is no conflict of interest with regard to this manuscript.