Clinical implementation of AXB from AAA for breast: Plan quality and subvolume analysis

Abstract Purpose Two dose calculation algorithms are available in Varian Eclipse software: Anisotropic Analytical Algorithm (AAA) and Acuros External Beam (AXB). Many Varian Eclipse‐based centers have access to AXB; however, a thorough understanding of how it will affect plan characteristics and, subsequently, clinical practice is necessary prior to implementation. We characterized the difference in breast plan quality between AXB and AAA for dissemination to clinicians during implementation. Methods Locoregional irradiation plans were created with AAA for 30 breast cancer patients with a prescription dose of 50 Gy to the breast and 45 Gy to the regional node, in 25 fractions. The internal mammary chain (IMCCTV) nodes were covered by 80% of the breast dose. AXB, both dose‐to‐water and dose‐to‐medium reporting, was used to recalculate plans while maintaining constant monitor units. Target coverage and organ‐at‐risk doses were compared between the two algorithms using dose–volume parameters. An analysis to assess location‐specific changes was performed by dividing the breast into nine subvolumes in the superior–inferior and left–right directions. Results There were minimal differences found between the AXB and AAA calculated plans. The median difference between AXB and AAA for breastCTV V 95%, was <2.5%. For IMCCTV, the median differences V 95%, and V 80% were <5% and 0%, respectively; indicating IMCCTV coverage only decreased when marginally covered. Mean superficial dose increased by a median of 3.2 Gy. In the subvolume analysis, the medial subvolumes were “hotter” when recalculated with AXB and the lateral subvolumes “cooler” with AXB; however, all differences were within 2 Gy. Conclusion We observed minimal difference in magnitude and spatial distribution of dose when comparing the two algorithms. The largest observable differences occurred in superficial dose regions. Therefore, clinical implementation of AXB from AAA for breast radiotherapy is not expected to result in changes in clinical practice for prescribing or planning breast radiotherapy.

There are currently two dose calculation algorithms available in Varian Eclipse software (Varian Medical Systems, Palo Alto, USA): Anisotropic Analytical Algorithm (AAA) and Acuros External Beam (AXB).
AXB was introduced in 2010 to address limitations of AAA in inhomogeneous regions, including overestimation of dose in the lung, [1][2][3][4] and the over/under estimation of dose beyond low/high density materials. 5,6 While many centers with Varian Eclipse software have access to AXB, implementation of a new dose calculation algorithm requires a thorough understanding of how this change will affect plan quality and, subsequently, clinical practice. This paper characterizes differences in dose-volume parameters and plan quality when implementing AXB from AAA for breast cancer radiotherapy planning.

1.B | Background
AXB is based on solving the Linear Boltzmann Transport Equation numerically, and calculates dose to the medium instead of the dose to water, resulting in more accurate calculation for inhomogeneous tissues than AAA. In breast radiotherapy, there is a large range of tissue densities in the irradiated volume (breast, lung, bone, air). The irradiation of internal mammary chain (IMC) nodes using a modified wide tangent field is growing in popularity, 7 and this technique results in a larger volume of lung included in the field than standard tangents. This may introduce larger uncertainties in calculated dose due to inhomogeneous interfaces.
Fogliata et al. characterized the performance of AXB for simple two-tangent field plans in different breast tissue types (adipose vs. ductal), and for lung inside and outside the radiation field. 8 This investigation determined that AXB could differentiate between breast tissue types and suggested that this may improve accuracy of dose calculation for patient treatments. This study also showed that for lung dose calculations the largest difference between AXB and AAA was for deep inspiration breath hold (DIBH) with consistent overestimation by AAA. 8 Our institution currently uses AAA for breast radiotherapy planning, and this study was designed to inform the clinical implementation of AXB for four-field breast planning in our clinic. Specifically, we aim to determine if a clinically relevant difference exists between the two calculation algorithms in this context. For this purpose, we define a clinically relevant difference as a difference in the appearance of the AXB plans (through evaluation of isodose lines or DVH curves) when compared to AAA plans, that may be noticeable or concerning to a physician reviewing the plan. The largest differences between dose calculations by AXB and AAA occur in patients planned during DIBH as lung density is reduced through this maneuver. 8 9 Specifically, the IMC CTV (IMC CTV ) was defined as a 5 mm expansion around the medial vascular structure from the first to third interspaces. Treatment plans were developed on this dataset using clinical field borders, and the breast was contoured for dose evaluation purposes only. The superficial region was defined as a 5 mm margin deep to the external body contour bordering on, but excluding, the breast. This represents the dose in the superficial region but is not anatomically representative of the skin. 10 A summary of patient characteristics can be found in Table 1. The prescription dose for the whole breast was 50 Gy in 25 fractions and the mid-axillary dose was 45 Gy in 25 fractions. The IMC CTV was covered by 80% of the tangent dose.

2.B | Separation across beam path
The beam path length along the 2500 cGy isodose line ranged from 15.1 to 32.5 cm. The profiles of the total plan dose, as well as the CT profile along the line, for AXB and AAA are provided in Fig. 1.

2.C | Treatment planning and dose calculation algorithms
Forward-planned right breast field-in-field plans were created for each patient by the same certified dosimetrist. All patients were

2.D | Plan Comparison and Evaluation Metrics
Dose-volume histogram (DVH) parameters were retrieved using an in-house MATLAB R2016a (The MathWorks, Inc., Natick, MA) Graphical User Interface (GUI) with a bin size of 0.1 cGy. DVH parameters: breast V 95% , V 105%, and hotspot, defined as dose to 1 cm 3 (D 1 cc ); IMC CTV V 80%; heart mean dose (D mean ) and maximum point dose (D max ); ipsilateral lung V 20 Gy and V 5 Gy ; and superficial region mean dose (D mean ) and superficial region dose to 2 cm 3 (D 2 cc ). Wilcoxon Signed-Rank tests were performed between AAA and both AXB-DM and AXB-DW with a significance level of 0.05.
To elucidate overall dosimetric trends between the two algorithms, point-wise median dose-volume histograms (mDVH) were computed for IMC CTV , superficial region, and ipsilateral lung by finding the population median value in each DVH bin.
The homogeneity of the dose to the ipsilateral breast was assessed using the Homogeneity Index (HI), given by the following equation.
To examine possible regional trends, the breast was subdivided into nine subvolumes ( found in Table 2. The mean dose (D mean ) and hot spot (D 1 cc ) was compared between the subvolumes.
To examine the effects of patient size on the difference between AXB and AAA, dose and HU profiles were collected 1 cm anterior to the 2500 cGy isodose line, which was used as an analog for the field edge. This was chosen as it represents the path of largest separation through the patient.
This study was determined to be of minimal risk and consistent with a quality improvement project using the Alberta Research Ethics Community Consensus Initiative (ARECCI) screening tool provided by the Heath Research Ethics Board of Alberta and did not require further ethics board approval. 11 3 | RESULTS

3.A | Targets and OARs
Minimal differences were observed across all patients between AAA and AXB radiotherapy plans (Table 3). Figure 3(a)

3.B | Subvolume analysis
The results of the breast subvolume analysis are provided in Fig. 5.
There was a statistically significant trend for AXB calculated plans to be slightly hotter medially and slightly cooler laterally, but all median differences were within AE1 Gy for AXB-DM and within AE1.5 Gy for AXB-DW. All patient-specific differences in quadrant mean dose for both AXB-DM and AXB-DW were within AE2 Gy.
A hotspot analysis was performed as part of the subvolume analysis. For most patients, the hotspot was in the superior central or superior lateral subvolumes. We observed no overall trends in hotspot location between AAA and AXB.

| DISCUSSION
When introducing a new dose calculation algorithm into routine patient care, it is important to have a thorough understanding of the expected differences between the old and new plans, so that clinical practice can be adjusted accordingly. This need was previously identified in the radiation oncology community while moving from 2D to 3D planning and subsequently led to the introduction of dose inhomogeneity corrections. 12 AXB (version 10 and 11) has been extensively validated in heterogeneous phantoms against Monte Carlo (MC) and AAA calculations. 5,13,14 In all cases, AXB exhibited closer agreement with MC calculations than AAA and the differences were generally less than 5%. The results of these studies support the hypothesis that AXB produces more accurate dose calculations in heterogeneous materials and has led to follow-up clinical application studies into the implications of using AXB in lung treatment planning. 13,[15][16][17][18] The irradiation volume in whole breast radiotherapy is quite inhomogeneous with severe contour changes. The dosimetric impact of using AXB instead of AAA has been studied in standard two tangent fields with the purpose of assessing the behavior of the algorithm in different breast tissues and lung densities. 8 Our aim was to evaluate the expected plan differences between AAA and AXB in a locoregional DIBH scenario, where the density changes are greatest, with the purpose of implementing AXB for breast planning into clinical practice.
We investigated the difference between the two AXB dose reporting modes, dose-to-water and dose-to-medium, compared to AAA. Larger magnitude differences were observed for dose-to-water compared to AAA than dose-to-medium in the breast DVH metrics.
This trend is consistent with findings of Zifodoya et al. 19 While some parameters were statistically significant, the difference between AXB Similar results were found for heart, IMC CTV , and ipsilateral lung parameters. Thus, for commonly evaluated target volumes and OAR dose constraints and DVHs, noticeable differences between AAA and AXB are not expected during clinical implementation. The choice of dose-to-medium or dose-to-water continues to be a debate in the community. 20,21 The advantage of dose-to-water is that most clinical experience, including calibration and outcome data, are based on dose-to-water dose reporting. Dose-to-medium has the advantage that is the most consistent with Monte Carlo. 22,23 The largest difference between AAA and both AXB reporting modes was mean superficial dose. This is to be expected because AXB handles interfaces and inhomogeneities more accurately than AAA. 13,24 AAA has been shown to be agree more closely with MC calculations than older algorithms, 25  sue, particularly when the beam was delivered at a wide angle as is the case in breast radiotherapy. 27 Since AXB has been shown to agree more closely with MC calculations in the buildup region, 13 it is consistent that it should estimate the mean dose in the superficial 5 mm to be hotter. The 2-cc hotspot did not change with AXB-DW.
In the subvolume analysis, there was a small but measurable trend toward higher mean dose to medial subvolumes and lower mean dose to lateral subvolumes when dose was calculated using AXB as compared to AAA. This may be a result of a combination of factors including geometric location of the prescription point in original plans, volume averaging in the subvolumes, modulation, and path through lung along chest wall interface. The contribution of each factor cannot be completely separated. As can be seen in Table 2, the lateral segments of the breast are on average larger than the medial and central segments. This could contribute to volume averaging resulting in the lower mean dose recorded in the lateral segments. Figure 1 These results may not extend to wedged field planning, especially physical wedges, where the skin dose from scatter is higher. As well, the impact on partial breast planning has not been explored.
Fogliata et al. 8 investigated the dosimetric implication of separately analyzing different tissues in the breast using AXB and AAA. This study found that AXB calculates higher dose in glandular breast tissue, which has a similar density to muscle, and does not calculate as large a difference for adipose tissue, which combined result in an overall dose to the breast that is not clinically relevant. Our results are consistent with these findings.
If large differences exist between two dose calculation algorithms, questions may arise about the impact of switching to a new dose calculation algorithm on clinical practice. For example, in lung, AXB has been shown to calculate lower dose to the edge of the tumor than AAA, which may result in physicians changing practice in order to cover the tumor to the same level as when using AAA. 13 This behavior can be problematic when all outcomes and toxicity data are based on doses calculated using the older algorithm. Apart from superficial dose, we found no clinically relevant differences between the breast plans calculated with AAA and AXB. Upon clinical implementation of AXB for breast, clinicians should be made aware that higher superficial doses are a result of the new dose calculation algorithm.

| CONCLUSIONS
In this study, the differences in dose-volume parameters and plan quality were evaluated in AXB and AAA for breast cancer. Marginal differences were found for target volume coverage and homogeneity and, aside from skin mean dose, organ at risk dose-volume parameters. As the dosimetric differences between the two algorithms are small, we conclude that using AXB for breast treatment planning would have a minimal impact on clinical practice.

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