A dosimetry study of post‐mastectomy radiation therapy with AeroForm tissue expander

Abstract Purpose To evaluate the dosimetric effects of the AeroFormTM (AirXanpders®, Palo Alto, CA) tissue expander in‐situ for breast cancer patients receiving post‐mastectomy radiation therapy. Methods and Materials A film phantom (P1) was constructed by placing the metallic canister of the AeroForm on a solid water phantom with EBT3 films at five depths ranging from 2.6 mm to 66.2 mm. A breast phantom (P2), a three‐dimensional printed tissue‐equivalent breast with fully expanded AeroForm in‐situ, was placed on a thorax phantom. A total of 21 optical luminescent dosimeters (OLSDs) were placed on the anterior skin–gas interface and the posterior chest wall–metal interface of the AeroForm. Both phantoms were imaged with a 16‐bit computed tomography scanner with orthopedic metal artifact reduction. P1 was irradiated with an open field utilizing 6 MV and 15 MV photon beams at 0°, 90°, and 270°. P2 was irradiated using a volumetric modulated arc therapy plan with a 6 MV photon beam and a tangential plan with a 15 MV photon beam. All doses were calculated using Eclipse (Varian, Palo Alto, CA) with AAA and AcurosXB (AXB) algorithms. Results The average dose differences between film measurements and AXB in the region adjacent to the canister in P1 were within 3.1% for 15 MV and 0.9% for 6 MV. Local dose differences over 10% were also observed. In the chest wall region of P2, the median dose of OLSDs in percentage of prescription dose were 108.4% (range 95.4%–113.0%) for the 15MV tangential plan and 110.4% (range 99.1%–113.8%) for the 6MV volumetric modulated arc therapy plan. In the skin–gas interface, the median dose of the OLSDs were 102.3% (range 92.7%–107.7%) for the 15 MV plan and 108.2% (range 97.8–113.5%) for the 6 MV plan. Measured doses were, in general, higher than calculated doses with AXB calculations. The AAA dose algorithms produced results with slightly larger discrepancies between measurements compared with AXB. Conclusions The AeroForm creates significant dose uncertainties in the chest wall–metal interface. The AcurosXB dose calculation algorithm is recommended for more accurate calculations. If possible, post‐mastectomy radiation therapy should be delivered after the permanent implant is in place.


| INTRODUCTION
Post-mastectomy radiation therapy (PMRT) to the chest wall and locoregional nodes has been shown to reduce the incidence of locoregional and distant recurrence and to improve overall survival in patients with node-positive breast cancer. [1][2][3] At our institution, we have a well-established planning protocol for patients with McGhan style tissue expander in situ. If the chest wall is irradiated with tangential beams, 15 MV (15X) photon beams are used for better transmission through the expander's metal (compared with 6 MV photon tangents) and 1cm bolus is used to provide adequate skin dose. 4 If Volumetric Modulated Arc Therapy (VMAT) techniques is used, 6 MV (6X) photon arcs with a 0.3 cm bolus are used. 5 A new tissue expander system, the AeroForm™ (AirXpanders ® , Palo Alto, CA), has recently become available to patients. [6][7][8][9] Aeroform expander consists of an implantable silicone tissue expander containing a metal reservoir of compressed carbon dioxide. The metallic reservoir causes a significant attenuation and inhomogeneous doses around the expander in PMRT 9-12 with 60 Co and 6MV. Moni 10 has shown that the treatment planning system (TPS) can have significant dose error (>5%) at the chest wall-expander interface. The computed tomography (CT) high-Z artifact associated with the device creates additional dosimetric challenges. Shah 13 suggests that further studies on the potential increase in toxicity, underdosing, and tissue delineation challenges associated with this gas expander in situ during radiotherapy is warranted.
In this study, we investigate the feasibility of our current clinical planning practice for patients with tissue expander in situ which was not addressed in published studies. We focus on the dosimetry of the AeroForm in the chest wall-metal and skin-gas interfaces to develop treatment planning guidelines for patients who are receiving PMRT with the AeroForm expander. The dosimetry of skin-gas interface of the Aeroform, which would have impacted on the skin dose of the patient, has also not been addressed by other studies.
The impact of different assignment of Hounsfield units (HU) and materials of the Aeroform on 16-bit CT scans was also investigated.
An intact AeroForm expander was inflated to full capacity (400 cm 3 ). Three-dimensional (3D) printing technology was used to create a breast phantom with TangoPlus FLX930 tissue-equivalent material (Stratasys, Eden Prairie, MN) with the expander in situ. The 3D printed "breast" was taped to a plastic phantom as the breast phantom (P2) to simulate a post-mastectomy patient (Fig. 2). The anterior part of the 3D breast in P2 is 0.5 cm thick, mimicking the patient's skin.
Multiple optically stimulated luminescence dosimeters (OSLD) (LANDAUER, Glenwood IL) were placed throughout this 3D printed breast to measure doses in the chest wall-metal and skin-gas interfaces. In all, 12 OLSDs were used to generate a calibration curve from 100 to 700 cGy. Another 50 OSLDs were used in the clinical study. To improve the dosimetric accuracy, each OSLD, i, was pre-irradiated with a calibration dose, D c , of 200cGy and was read out as D c,i . Chip-specific sensitivity factor, SF i, , was obtained as follows: A total of eight chips were taken from the clinical chips as reference chips and were grouped into two equal groups. Each group was irradiated with the prescribed energy and dose, D p , of the clinical plan. The average of the net dose, defined by subtracting D c,i from the read-out, deposited on each chip, D p,i , was used to generate the correction factor, CF, as follows: The clinical dose of each OSLD, D i , was determined by correcting the net clinical dose, D o,i , as: The clinical OSLDs were placed on the interfaces of the breast phantom. A set of nine OLSDs were placed in the chest wall--AeroForm ( Fig. 3a) and AeroForm-skin interfaces underneath the 0.5 cm 3D printed skin ( Fig. 3b) in P2.  The plans without re-assignments were labeled as AAA-16 bits with AAA and AXB-16 bits with AXB calculations. The plans with HU reassignment to the canister (titanium alloy, HU = 4800), air inside the AeroForm (Air, HU = −1000), and artifacts in soft tissue (tissue, HU = 100) were recalculated and labeled as AAA-4800 and AXB-4800. To simulate the HU saturation in 12-bit CT, the pixel with HU greater than 3071 was re-assigned to 3071 and were labeled as AAA-12 bit and AXB-12 bit for AAA and AXB calculations.

2.C | Measurements and analysis
All measurements were delivered on a Varian TrueBeam with M120 MLC. The three-channel calibration protocol, 17 18 CBCTs were used to setup the breast phantom before delivery of each plan. OLSDs were contoured on the CBCT images and fused with simulation CTs to reduce dosimetric uncertainty. Figure 4 illustrates the film results. At 2.6 mm downstream from the canister, with a 6X anterior-posterior (AP) open field, the average dose differences within the ROI, <ΔD ROI > for AAA and AXB were 1.4% and 3.8%, respectively. Local dose differences amounting to AE14% and AE 8% for AAA and AXB, respectively, were observed for 6X. The <ΔD ROI > was within AE 3.4% for both energies and algorithms. In a composite beam arrangement of AP plus bilateral fields mimicking C2, the <ΔD ROI > was 3.9% and 2.0% for AAA and AXB, respectively ( Fig. 4b and c). The local dose errors were within 10%

| RESULTS
for AAA and 6% for AXB. The <ΔD ROI > of the AP open field for 15X were −1.1% and −3.1% for AAA and AXB, respectively. However, the local errors were over 10% for both algorithms. The analysis also revealed significant dose variation from films behind the region of the canister comprising high Z materials. In the case of 15X, over 20% local differences were observed for both algorithms.
A pair of bilateral tangential fields, modeling C1, were measured and had a <ΔD ROI > of −0.6 and −2.6 for AAA and AXB at the same depth ( Fig. 4d and e) (Fig. 5a). Although both algorithms showed similar <ΔD ROI >, AAA showed more local variation than AXB. The <ΔD ROI > of C1 were within 1.4% and 2.6% for AAA and AXB, respectively (Fig. 5b).
The <SF i > was found to be 0.967 (σ = 2.9%). The CF for 6x and 15x were found to be 0.98 and 0.97, respectively. Table 1  | 37 limited 12-bit dynamic range also increased TPS calculation uncertainties.
Although the <ΔD ROI > were modest, significant local dosimetric differences (over 20%) were found in P1 phantoms within 5 mm of the canister interfaces. This can be attributed to the inaccurate forward scatter modeling of the high Z materials by both algorithms.
However, this study has shown the local dosimetric errors can be moderated with the current clinical beam arrangement. 5,16 The results from P2, approximately correspond to the results at 5 mm in P1, showed that AAA has higher dosimetric uncertainty than AXB. Significant dose uncertainty (~16.6% locally) at chest wall-metal interface can result in potential underestimation of heart dose. This is perhaps partly attributed to the inaccurate forward and back scatter of the dose calculation algorithms near high Z materials.
On the skin and the skin-gas interface, both algorithms showed reasonable dose agreement with measurements (5%-7%). These uncertainties should be incorporated into the clinical planning process.

ACKNOWLEDGMENTS
This work was partially supported by the MSK Cancer Center Support Grant/Core Grant (P30 CA008748). We would like to thank AirXanpders ® (Palo Alto, CA) for supply of devices for this research.