Modeling of the metallic port in breast tissue expanders for photon radiotherapy

Abstract The purpose of this study was to model the metallic port in breast tissue expanders and to improve the accuracy of dose calculations in a commercial photon treatment planning system (TPS). The density of the model was determined by comparing TPS calculations and ion chamber (IC) measurements. The model was further validated and compared with two widely used clinical models by using a simplified anthropomorphic phantom and thermoluminescent dosimeters (TLD) measurements. Dose perturbations and target coverage for a single postmastectomy radiotherapy (PMRT) patient were also evaluated. The dimensions of the metallic port model were determined to be 1.75 cm in diameter and 5 mm in thickness. The density of the port was adjusted to be 7.5 g/cm3 which minimized the differences between IC measurements and TPS calculations. Using the simplified anthropomorphic phantom, we found the TPS calculated point doses based on the new model were in agreement with TLD measurements within 5.0% and were more accurate than doses calculated based on the clinical models. Based on the photon treatment plans for a real patient, we found that the metallic port has a negligible dosimetric impact on chest wall, while the port introduced significant dose shadow in skin area. The current clinical port models either overestimate or underestimate the attenuation from the metallic port, and the dose perturbation depends on the plan and the model in a complex way. TPS calculations based on our model of the metallic port showed good agreement with measurements for all cases. This new model could improve the accuracy of dose calculations for PMRT patients who have temporary tissue expanders implanted during radiotherapy and could potentially reduce the risk of complications after the treatment.


| INTRODUCTION
More and more postmastectomy patients have immediate breast reconstructions mainly for cosmetic reasons. [1][2][3][4][5][6][7] A temporary tissue expander, which usually includes a high-density magnetic injection port, offers many advantages including relative simplicity, low morbidity, and good aesthetic results, over other types of breast reconstructions. 3,6 American Society of Plastic Surgeons reported that there were 74,694 breast reconstructions using tissue expander/implant in 2014 in the US and the number kept increasing. 8 Many patients will receive postmastectomy radiotherapy (PMRT) with the temporary tissue expander present, while the tissue expander could negatively impact the effectiveness of PMRT and increase the risk of complications. 9,10 Some physicians felt the reconstructions challenged their ability to deliver effective radiotherapy. 6 It has also been reported that failures in the breast reconstruction and complication rates were significantly higher for patients who received PMRT with the temporary tissue expanders than patients who received PMRT with permanent implants 11 or autologous tissue reconstruction. 12,13 Previous literature about the effect of the tissue expander containing the high-density metallic injection port on dose distribution was conflicting and controversial: Moni et al. 14  Monte Carlo (MC) simulation and found 7% to 13% dose reduction with the expander in place for two 6 MV tangential photon beams, and around 6% dose reduction for 18 MV beams; Chen et al. 18 reported that tissue expanders with metal ports will increase dose heterogeneity and reduce dose coverage significantly for patients treated with 6 MV or 18 MV opposed tangent photon fields; Sharabi et al. reported in an abstract 19 5%~20% dose attenuation due to the metallic port and claimed a nondeformable implant model of the port was created based on manufacture specifications, but they did not show the details of the model or validate the dosimetric accuracy of the model; Trombetta et al. 20 initially reported that no significant change in dose distributions was found for an opposed pair of 6 MV photon beams delivered to a breast phantom containing a metallic port, but later drew a conflicting conclusion in a separate paper 21 that the metallic port must be taken into account in the dose calculations; Strang et al. 22 used TLD measurements and concluded that radiation doses around the tissue expander were unaltered; Srivastava et al. 23 conducted measurements in a water phantom using a small ion chamber (IC) and concluded that dose perturbation caused by metallic port in photon beams was 5%~20% and this perturbation could not be predicted by TPS; Zabihzadeh et al. 24 used MC simulation and found a dose enhancement about 15% in front of the port and a dose reduction of about 10% at 5 cm under the port; Gee et al. 25 used radiochromic films as in vivo dosimeter and found an average 7% dose reduction to skin surface in a sample of PMRT patients with the tissue expander present during radiotherapy.
One of the reasons for these contradictory findings is that most of the current TPSs are not calibrated or validated for the highdensity metallic port in the tissue expander. Considering TPS is an essential step among the whole radiotherapy procedure, it is critical to calibrate it to generate accurate treatment plans for patients with tissue expanders. The modeling of the metallic port in the current TPS may not be able to accurately calculate the dosimetry impact introduced by the high-density materials, particularly in the areas near these materials. 16,18 Thompson and Morgan 15 manually assigned a bulk density of 7.9 g/cm 3 to the implant in TMS TPS (Nucletron, Veenendaal, the Netherlands) and concluded this kind of modeling was ineffective; Chen et al. 18 used a series of phantom and film measurements to find an electron density relative to water of 11.8 for the metallic port in an open field photon beam in Eclipse TPS (Varian Medical Systems, Palo Alto, CA, USA). However, this value did not yield good agreement between measurements and TPS calculations for both 6 and 18 MV photon beams; Trombetta et al. 20 assigned the density of the metallic port in the Eclipse TPS to be 5 g/cm 3 which is the highest available value in the system. Except for one abstract 19 in which the details and accuracy of the model were not provided, none of the previous studies created and validated a model of the metallic port with generic dimensions and density that can be applied to different beam energies, neither did they investigate the impact of the metallic port on any advanced techniques like intensity-modulated radiotherapy (IMRT) or volumetric-modulated arc therapy (VMAT), while these techniques have been used to treat postmastectomy patients. 26 In this study, we aim to model the metallic port in a commercial  The measurement setup is shown in Fig. 2: the isocenter was located at 15 cm depth and the source-to-surface (SSD) was 85 cm; the metallic port was put on the surface of a solid water slab and the distance from the water surface to the slab surface was 5 cm. The IC was placed directly under the metallic port to measure the dose attenuation through the metallic port and the position of the IC varied from 7 to 15 cm under the water surface. To confirm if the IC was exactly aligned with the metallic port for the parallel setup, the location of the metallic port was slightly adjusted laterally until the IC reading reached the minimum. The measured doses at depths were compared with dose calculations by collapsed cone convolution (CCC) algorithm in Pinnacle TPS with a dose grid of 1 9 1 9 1 mm 3 .
Although there are more accurate dose algorithms 27 that can be used to reduce metal artifact and calculate dose around high-density heterogeneities, the CCC algorithm was chosen because it is the default and also the most accurate dose algorithm used by Pinnacle and the purpose of this study was to calibrate the Pinnacle TPS to accurately calculate dose around the metallic port. The density in the TPS model of the metallic port was adjusted with an extended CT conversion table in Pinnacle until the calculation results agreed with the measurements. However, the attenuation from the metallic port model is determined not only by the density of the model but also the dimensions (both diameter and thickness) of the model in TPS. For example, increasing the diameter of the metallic port model in the TPS will yield more attenuation in the parallel direction [ Fig. 2 while the effect in the perpendicular direction would not be significant. Also, increasing the density of the model will yield increased dose close to the metallic port because of increased scatter, while doses at deeper points will decrease due to increased attenuation. Therefore, modeling of the metallic port in the TPS requires fine adjustments of the combination of density, diameter, and thickness.
In this study, we aimed to find the best model that yields the best agreement between all TPS calculations and measurements. As a result, the dimension of the metallic port model may be different from the film measurement result.
Contouring the metallic port may have large uncertainties because of the presence of artifacts induced by the metallic port in the kilovoltage (kV) CT images, and most patients only have kV CT images available. In some clinic, part of the artifact is included in the metallic port contouring which can cause possible dose error. In this study, we compared our new model with two widely used clinical models ( Fig. 3): in clinical model #1, the titanium shell and magnetic disk were contoured by the dosimetrist based on the physical dimensions, CT images, and "bone" window/level. The densities of the Titanium shell and magnetic disk were overridden by their nominal densities, and surrounding artifacts were overridden as water; in clinical model #2, the metallic port was contoured based on CT images only with "bone" window/level and some of the artifacts were therefore included. The surrounding artifacts were overridden as water and the contoured metallic port utilized the assigned default density converted from the CT number; in our new model, the location of the disk and its tilted angles in transverse and sagittal plane were identified from the CT images using "bone" window/level. The geometrical information of the metallic port, such as the location of the center of the disk and tilted angles in lateral and sagittal plane on F I G . 1. Image of a typical metallic injection port taken out of a breast tissue expander (ALLERGAN, Santa Barbara, CA). the CT images, was transferred to an in-house MATLAB (Mathworks, version 7.9, Natick, MA, USA) code to model the disk three-dimensionally. We then generated the contour of the metallic port using the in-house code and used the new contour to update the file containing the structure set in Pinnacle. The density of the metallic port was overridden by the value determined from IC measurement described previously. The surrounding artifacts were overridden as water.  To minimize daily variation due to decay after irradiation, TLD reading were actually started after 2 days so that signal fading after irradiation could be ignored.

|
The possible TLD uncertainties include fading, dose-response nonlinearity, energy response corrections, and system sensitivity.
Among these uncertainties, the fading effect was basically negligible because TLD reading were started after 2 days so that signal decay after irradiation could be ignored; energy response corrections were not necessary because we used the same beam energy for both measurements and calibrations; the dose-response nonlinearity was baseline condition (no disk) was created with the metallic port and the tissue expander overridden as water to simulate a homogeneous breast without a tissue expander. Doses at two points close to the skin but at depths deeper than 1 cm and doses at two points on the chest wall were calculated (Fig. 5). A "reduced-PTV" was defined for the DVH evaluation purpose: the original PTV is narrowed down to the slices containing the metallic port in the CT images and the tissue expander plus the metallic port were excluded because we were only interested in the dose delivered to the patient's tissue.

| RESULTS & DISCUSSION
The film measurement results for parallel and perpendicular setups are shown in Fig. 6. It was found that the radiological diameter of the metallic port was 1.75 cm and the thickness was 2.5 mm, and these were smaller than the nominal diameter (2.1 cm) and thickness in the film image. This was expected, since the effective thickness of the Ti wall was less than 0.2 g/cm 2 (= 4.5 g/cm 3 9 0.04 cm), which can be ignored without noticeable change in the calculated doses.
Using the determined dimensions from the film measurements, we found that the calculated doses in the TPS did not match the IC measurements with the perpendicular setup when the density was adjusted for the parallel setup measurements. To achieve the best agreement for both setups, the thickness and the diameter of the metallic port were decided to be 5 mm and 1.75 cm, and the density of the port was decided to be 7.5 g/cm 3 . The IC measurements and calculated doses are shown in Fig. 7 The calculated and measured doses using the simplified anthropomorphic phantoms are shown in Table 1  Point doses within the PMRT patient are listed in Table 2. For all plans and models, the dosimetric impact of the metallic port on chest F I G . 5. Photon dose calculation points in a PMRT patient with a tissue expander. The blue contour represents PTV, and the purple color wash represents reduced-PTV (PTV minus tissue expander) on this slice.