DVH analysis using a transmission detector and model‐based dose verification system as a comprehensive pretreatment QA tool for VMAT plans: Clinical experience and results

Abstract Purpose Dose volume histogram (DVH)‐based analysis is utilized as a pretreatment quality assurance tool to determine clinical relevance from measured dose which is difficult in conventional gamma‐based analysis. In this study, we report our clinical experience with an ionization‐based transmission detector and model‐based verification system, using DVH analysis, as a comprehensive pretreatment QA tool for complex volumetric modulated arc therapy plans. Methods and Materials Seventy‐three subsequent treatment plans categorized into four clinical sites (Head and Neck, Thorax, Abdomen, and Pelvis) were evaluated. The average dose (Dmean) and dose received by 1% (D1) of the planning target volumes (PTVs) and organs at risks (OARs) calculated using the treatment planning system (TPS) were compared to a computed (model‐based) and reconstructed dose, from the measured fluence, using DVH analysis. The correlation between gamma (3% 3 mm) and DVH‐based analysis for targets was evaluated. Furthermore, confidence and action limits for detector and verification systems were established. Results Linear regression confirmed an excellent correlation between TPS planned and computed dose using a model‐based verification system (r 2 = 1). The average percentage difference between TPS calculated and reconstructed dose for PTVs achieved using DVH analysis for each site is as follows: Head and Neck — 0.57 ± 2.8% (Dmean) and 2.6 ± 2.7% (D1), Abdomen — 0.19 ± 2.8% and 1.64 ± 2.2%, Thorax — 0.24 ± 2.1% and 3.12 ± 2.8%, Pelvis 0.37 ± 2.4% and 1.16 ± 2.3%, respectively. The average percentage of passed gamma values achieved was above 95% for all cases. However, no correlation was observed between gamma passing rates and DVH difference (%) for PTVs (r 2 = 0.11). The results demonstrate a confidence limit of 5% (Dmean and D1) for PTVs using DVH analysis for both computed and reconstructed dose distribution. Conclusion DVH analysis of treatment plan using a model‐based verification system and transmission detector provided useful information on clinical relevance for all cases and could be used as a comprehensive pretreatment patient‐specific QA tool.


| INTRODUCTION
A radical development in radiotherapy treatment planning techniques combined with advanced imaging and delivery systems with various degrees of freedom allows the delivery of high doses with steep dose gradients aimed at the target while sparing surrounding normal tissues. 1,2 Volumetric modulated arc therapy (VMAT) and stereotactic body radiotherapy (SBRT) techniques are routinely used in clinical practice to treat complex targets in various treatment sites. 2,3 Due to reduced safety margin and high doses delivered in short fraction, any potential errors in planning and delivery would lead to serious consequences for patients. 1 The dose delivery to the target is influenced by uncertainties in the planning (complexity of plans) and delivery (design of the multileaf collimators) systems. 4 Hence, each treatment plan created using complex techniques requires a comprehensive patient-specific pretreatment quality assurance (QA) procedure to verify the dose calculation generated in the treatment planning system (TPS) and delivery system such as linear accelerator.
Pretreatment QA methods based on films, ionization chamber or scintillation detectors, portal dosimetry, Monte Carlo and log file analysis has been published and proven useful for patient-specific pretreatment QA, but each method has its weaknesses as well. Traditionally, independent monitor unit calculation softwares are utilized to verify the TPS dose calculation. To verify the beam delivery, 2D detector arrays equipped with ionization chambers or semiconductor detectors are commonly used and play a major role to ensure that an IMRT treatment plan is accurately delivered. [5][6][7] Conventional pretreatment QA includes delivering the patient plan to a standard phantom and comparing the measured and calculated 3D dose distribution using gamma analysis with different passing criteria. 5 The gamma index is calculated by combining the percentage dose difference and distance to an agreement for each of the pixels within the region of interest. 6 While gamma analysis based on measurements using different detectors provides a valuable understanding of whether the linear accelerator is operating as planned, it does not provide any correlation indicating a decrease in clinical metric with increasing and decreasing passing rate nor predict the clinical impact. 4,6 Furthermore, the gamma analysis has limited accuracy in the regions of steep dose gradients. Portal dosimetry and log file analysis are also used for pretreatment dose verification. Nevertheless, it is often difficult to quantify and interpret the results in terms of dose to targets and organs at risk (OAR) using the traditional QA methods. [6][7][8][9] To overcome these limitations, the incorporation of dose volume histogram (DVH) information within the QA procedure, in addition to gamma passing rates, is required to provide a comprehensive patient-specific pretreatment QA. 10 This would offer an insight into the relevance of observed differences between measured and the TPS planned dose to the target and surrounding normal structures.
Furthermore, dose verification at planning level combined with verification of delivery system is also required which in turn would result in a more comprehensive QA methodology as it consists of an independent dose calculation to verify TPS and verification of delivery systems for detrimental dose differences in target and OAR. 10 Previous studies have reported the utilization of Compass system (IBA Dosimetry, Schwarzenbruck, Germany) along with 2D ionization chamber array (MatriXX, IBA Dosimetry) for pretreatment verification. [10][11][12] The studies have reported a good agreement between the Compass computation and reconstructed dose in VMAT plans for dose calculated with Monaco (Elekta Inc., St Louis, MO, USA) and Eclipse (Varian Medical Systems Finland Oy, Helsinki, Finland) treatment planning system. 11,12 Furthermore, the accuracy of the reconstructed 3D dose distributions obtained using the Compass system and Dolphin detector has been evaluated. 13,14 However, no studies to date have reported the local confidence limits and action limits for targets and OARs in the utilization of model-based comprehensive patient-specific pretreatment QA for all clinical sites, with a suf-

2.B | Dose objective and constraints
The individual plan was generated to achieve better dose conformity with steep dose gradients and good target coverage while MOHAMED YOOSUF ET AL.
| 81 maintaining the constraints to critical organs as recommended in QUANTEC and RTOG protocols. 15,16 The OARs evaluated in this study included: Head and Neckbrainstem, mandible, oral cavity, larynx, spinal cord, eyes, optic nerves, optic chiasm, lens, cochlea, and parotid; Thoraxesophagus, lung, spinal cord, and heart; Abdomenbowel, kidney, liver, and spinal cord; Pelvisrectum, femoral head, bladder, and bowel. All plans were accepted and clinically approved for treatment by a Consultant Physician. Further details on the Compass system are published elsewhere. 1,14,17 The dose engine implemented in Compass uses a collapsed cone convolution/superposition (CC) algorithm. 18 The DICOM files of each treatment plan (CT image sets, RT structures, RT plans, and RT doses) were exported to the Compass verification system.

2.C | Compass verification system
The grid size used in Compass for dose computation and reconstruction is similar to the TPS plan (2.5 mm). The spacing of the detector is 5 mm for a field size of up to 140 mm 2 × 140 mm 2 which projects to approximately 8 mm in isocenter distance when the source-to-detector distance is 600 mm, whereas 5-10 mm in the remaining area. 13 Previous studies have investigated the validation and error detection capability of the Compass verification system and the Dolphin detector in detail. 1,14,19,20

2.E | Evaluation metrics
A conventional global gamma analysis was performed for all cases by normalizing both calculated and measured to the maximum absolute dose from TPS. A distance to agreement (DTA) of 3 mm and dose difference of 3% with a 10% lower dose limit threshold was applied for all cases to exclude the clinically irrelevant dose levels. A passing percentage of 95% with gamma values ≤1 were applied for all cases. 21,22 For independent model-based TPS verification, the dose calculation generated in TPS was compared to dose computed (DC) by Compass verification system using the CC algorithm. Secondly, for measurement-based pretreatment QA, TPS calculated dose was compared to reconstructed dose (RD) generated directly on patient anatomy based on fluence measurement using Dolphin transmission detector.
The DVH-based indices: the average dose (D mean ) and dose received by 1% (D 1 ) of the target volumes and OARs for all cases calculated using TPS was compared to Compass dose computation (DC) and reconstructed doses (RD). The results were statistically evaluated and local confidence limits were derived utilizing the concept confidence limit of |mean|+1.96σ and successively an action limit was established that account for the deviation in quality measures which requires clinical intervention. 4 Furthermore, the correlation between gamma pass rate (3% 3 mm criteria) and mean difference, attained between planned and measured dose, using DVH analysis for PTVs was studied. The Pearson correlation coefficient was used to calculate the correlation between calculated and measured dose distribution. Furthermore, TPS achieved dose constraints for OARs in each case were compared to the reconstructed dose measured using the Dolphin detector with the Compass verification system. Uncertainty analysis was studied for a transmission detector and Compass verification system. The overall uncertainty was calculated as the square root of the sum of the squares of all the listed uncertainties.

| RESULTS
The DVH parameters (D mean and D 1 ), averaged on 73 cases, with corresponding Pearson's correlation coefficient value for target volumes and OARs between TPS calculated and Compass computed/reconstructed doses are presented in Tables 1 and 2. The mean planned dose (D mean and D 1 ) using the Monte Carlo algorithm, for target volumes (PTVs) and OARs for all cases, resulted in good agreement with Compass computed dose calculated using CC algorithm (r 2 = 1). This ensured an independent dose verification of TPS calculation.
As shown in Table 1, a good correlation was observed between TPS planned dose (mean) and Compass reconstructed dose (using Dolphin detector) for target volumes. Figure 1 illustrates the comparison of measured and calculated dose to PTVs for all cases. A marginally reduced PTV coverage was observed for Head and Neck cases from the reconstructed doses to those planned in TPS (r 2 = 0.98).
Likewise, the reconstructed dose for individual OARs correlated well with TPS planned dose for all sites except small structure in Head and Neck cases and heart in thorax. The dose measured using Dolphin detector in Head and Neck cases was found to be slightly higher than TPS planned doses but mostly within the institution accepted tolerances. It was observed more so for OARs that were small in size and adjacent to or within PTVs. A similar trend was observed for D 1 of targets and OARs between planned and measured dose using the Dolphin detector as presented in Table 2.
The overall confidence limit for all sites was determined to be 5% (D mean and D 1 ) for evaluating targets (PTV) using DVH analysis.
The maximum difference observed for each case between the TPS calculated dose to the PTVs (D mean and D 1 ) and the computed/reconstructed dose, using Compass and Dolphin detector, was within the confidence limit of 5%. Likewise, a confidence limit of 5% (D mean and D 1 ) were determined for OARs of all sites (Table 1) except for a few in Head and Neck cases, which are small in volumes or received very low dose (e.g., eye, lens, cochlea, parotid, optic nerves). A small deviation in dose to these structures resulted in large percentage differences. The action limit for targets and majority of the OARs were determined to be 7%.

3.A | Evaluation of uncertainty
In this work, as the comparison of TPS planned and calculated dose distribution using Dolphin detector was considered the end result, standard uncertainty (a combination of type A and type B) was estimated as shown in Table 3. The square root of the sum of squares of all uncertainties was used to calculate the overall uncertainty and T A B L E 1 Comparison of TPS calculated dose (mean) for target volumes and OARs to reconstructed (measured using Dolphin detector) and Compass computed dose (CC algorithm).

| DISCUSSION
Pretreatment patient-specific QA is widely used as the core component of most QA programs that involves complex treatment planning and delivery to combat the errors related to planning and delivery system like linear accelerator. 4 The patient-specific QA based on 2D array detectors is a clinically proven method for VMAT and stereotactic dose delivery. The conventional patient-specific QA methods using gamma pass rates might provide acceptable passing rates but limited in terms of clinical impact and outcomes. 7 This work aimed to report our clinical experience of a transmission-based detector (Dolphin) and Compass verification system as a comprehensive patient-specific pretreatment QA tool. The advantage of utilizing the Compass verification system is that it can act as an independent secondary TPS verification tool utilizing the CC algorithm for dose computation. Furthermore, patient-specific measurements are performed inside the patient's anatomy as opposed to other QA tools in which the doses are calculated in the phantom. 23 The absolute dose and geometric calibration for Dolphin detector are not needed for every measurement due to low dependency of ionization chambers within the detector and it also provides a robust QA which includes the daily variation of absolute dose of the linear accelerator. 17,24,25 In general, the DVH-based dose evaluation provides a quantitative analysis between TPS planned and computed/reconstructed The evaluation of D 1 is emphasized as the constraints for the majority of the OARs are fixed at a dose to smaller volumes. The confidence limit used in this study was determined based on the mean difference between the measured and expected values. The discrepancies between the average dose to PTVs and majority of the OARs (D mean and D 1 ) calculated by TPS and the Compass dose computation/reconstruction were well within the confidence limit of 5%.
Large disagreement for OARs in Head and Neck cases was observed for small structures that were close to or within targets but were within the tolerance thresholds. This is attributed to detector resolution, limitation in dose calculation by CC algorithm in high-density regions, and the electron contamination resulting from collimators and flattening filters as the detector is placed at 60 cm SSD. 2,23,26 We suggest an institutional-based protocol in evaluating these struc-

4.A | Limitations
The accuracy of ionization detector, when compared to QA systems like film dosimetry, is subject to uncertainties due to volume averaging, geometrical resolution, and self-attenuation which lead to concern about their sensitivity. Furthermore, as the Dolphin detector is attached on the head of linear accelerator, errors related to gantry, collimator, and table rotation cannot be detected. To mitigate the detector resolution limitation, a Monte Carlo generated response function for each ion chamber is applied in Compass. 23 Furthermore, online measurement using the Dolphin detector is not within the scope of this study. Moreover, the study is limited to treatment planning and delivery system from a single institution. This warrants a multi-institutional analysis utilizing different treatment planning and delivery systems.
DVH-based analysis, for targets and OARs, using the Dolphin detector and Compass verification system has been demonstrated as a comprehensive tool for patient-specific pretreatment QA. It has been showed that DVH-based analysis provides a better interpretation of the dose distribution within the targets and OARs in case the involvement of a physician is needed for any action before the patient treatment. Furthermore, local confidence limits and action limits based on DVH differences in the PTVs and OARs, for dose computed and reconstructed using Dolphin detector, on the patient anatomy established for routine patient-specific pretreatment QA.

| CONCLUSION S
DVH analysis of complex treatment plan using a model-based verification system (Compass) and Dolphin transmission detector provided useful information on clinical relevance for all cases and could be used as a comprehensive pretreatment patient-specific QA tool.
Local confidence and action limits based on the average dose difference in PTVs and OARs were established for clinical QA.

ACKNOWLEDG MENT
The authors have received no financial support/funding for, or have any other form of commercial interest in this study, which have been undertaken as part of the regular employment. There are no conflicts of interest. No ethics approval was required for this study. No copyright material has been used that requires approval for it to be published in this manuscript.

CONFLI CT OF INTEREST
The authors declare that there is no conflict of interest regarding the publication of this article.
T A B L E 3 Estimated standard uncertainty for Dolphin detector and Compass verification system.