Magnetic field dose effects on different radiation beam geometries for hypofractionated partial breast irradiation

Abstract Purpose Hypofractionated partial breast irradiation (HPBI) involves treatment to the breast tumor using high doses per fraction. Recent advances in MRI‐Linac solutions have potential in being applied to HPBI due to gains in the soft tissue contrast of MRI; however, there are potentially deleterious effects of the magnetic field on the dose distribution. The purpose of this work is to determine the effects of the magnetic field on the dose distribution for HPBI tumors using a tangential beam arrangement (TAN), 5‐beam intensity‐modulated radiation therapy (IMRT), and volumetric modulated arc therapy (VMAT). Methods Five patients who have received HPBI were selected with two patients having bilateral disease resulting in a total of two tumors in this study. Six planning configurations were created using a treatment planning system capable of modeling magnetic field dose effects: TAN, IMRT and VMAT beam geometries, each of these optimized with and without a transverse magnetic field of 1.5 T. Results The heart and lung doses were not statistically significant when comparing plan configurations. The magnetic field had a demonstrated effect on skin dose: for VMAT plans, the skin (defined to a depth of 3 mm) D1cc was elevated by +11% and the V30 by +146%; for IMRT plans, the skin D1cc was increased by +18% and the V30 by +149%. Increasing the number of beam angles (e.g., going from IMRT to VMAT) with the magnetic field on reduced the skin dose. Conclusion The impact of a magnetic field on HPBI dose distributions was analyzed. The heart and lung doses had clinically negligible effects caused by the magnetic field. The magnetic field increases the skin dose; however, this can be mitigated by increasing the number of beam angles.


| INTRODUCTION
Hypofractionated partial breast irradiation (HPBI) has been proposed as an ablative procedure to replace surgical resection 1,2 and also for the preoperative/postoperative adjuvant settings. 3,4 As an ablative procedure, HPBI has been proposed to reduce the treatment burden for those patients who cannot tolerate surgery. As an adjuvant treatment, HPBI aims to reduce the long, protracted radiation therapy (RT) schedule for standard breast conservation therapy (BCT) into a shorter, hypofractionated regimen targeted toward the postoperative bed where the cancer most often recurs. 5,6 The minimum treatment burden for low-risk breast cancer patients is the hypofractionated approach-as similar strategies are employed regularly and successfully for the brain and lung tumor sites, this is an attractive clinical solution. As data are accumulating for HPBI for these clinical settings, a variety of techniques have been developed to administer HPBI, such as interstitial high-dose-rate (HDR) brachytherapy, permanent seed implants, 7 intraoperative RT (for surgical candidates), and inversely planned external beam RT. 8 An exciting development in radiation oncology that shows promise for HPBI is the emergence of MRI-guided external beam solutions. 9 Several ambitious efforts are in progress to integrate MRI into a realtime or near-real-time solution for tumor targeting during or immediately prior to radiation delivery. These efforts have produced a number of hybrid MRI-RT solutions such as the Cross Cancer Institute MRI-Linac prototypes at the University of Alberta, 10 the Australian MRI-Linac concept headed by Paul Keall's research group, 11 the first commercialized Cobalt-60/MRI solution from the company ViewRay 12 and the pre-commercial Elekta (Elekta AB, Stockholm, Sweden) MRI-Linac. 13 Our cancer center is preparing to install a clinical prototype of the Elekta MRI-Linac. We anticipate that image-guided, hypofractionated partial breast irradiation will be a good candidate for the MRI-Linac's capabilities and thus have interest in developing treatment processes to that end. MRI can visualize breast tumors better than CT-based imaging 14,15 potentially affording a great improvement in daily geographical matching and tumor delineation than with current state-ofthe-art cone-beam CT (CBCT). Also, with exquisite soft tissue contrast, MRI can lend itself to adaptive radiation therapy (ART), where the MRI guidance can be used for online visualization and contouring of the tumor as it changes shape and size throughout treatment-particularly of interest in a highly deformable organ such as the breast. Finally, MRI-Linac solutions are attractive for HPBI due to the ability for these advanced platforms to cut down on internal motion margins using MLC tracking or exception gating using guidance from real-time MRI images or navigator signals. The MRI-Linac has been suggested by many to be ideal for image-guided ablative radiation therapy, a treatment paradigm that is expected to become increasingly prevalent in radiation oncology. 9,13,16 For low-risk breast cancer patients, the aforementioned technological advantages given by an MRI-Linac may pave the way for enabling high precision ablative techniques to be employed for intact breast tumors.
The Elekta MRI-Linac consists of a 1.5 T closed bore magnet with a linac rotating circumferentially about the imaging system and delivering beam through the cryostat. Since the magnetic field is always on, the electrons liberated by the x-ray photons are perturbed by the ever-present magnetic force. One of the consequences of this is the "electron return effect" (ERE), where electrons liberated at tissue-air and tissue-lung interfaces curl back on themselves due to the Lorentz force, depositing larger doses in tissue at these interfaces. 17 This can potentially lead to unwanted elevated doses at the skin or other high-to-low density interfaces.
The objective of this work is to determine the effects of the magnetic field for HPBI treatment geometries using 3 different treatment beam arrangements: 2-3 beam tangential arrangement (TAN), 5-beam intensity-modulated RT (IMRT) and volumetric modulated arc therapy (VMAT). By determining the magnetic field dose effects with these beam geometries, this work can provide good guidance for beam geometry selection in clinical scenarios.
For these comparisons, treatment plans were generated for all beam geometries and optimized with and without the magnetic field B 0 = 1.5 T. Throughout this paper, the following abbreviations will be used: TAN, IMRT and VMAT labels indicate plans with no magnetic field effects; TANB0, IMRTB0 and VMATB0 labels indicate plans using the named beam geometries and with magnetic field effects i.e., B 0 = 1.5 T. The plans were created with the same isocoverage of the PTV with the comparisons made via the organs-at-risk (OARs), which were the skin, heart, and lungs. The hypothesis was that the skin would be increasingly affected by the ERE with increasing number of beam angles and that the heart and lungs would have both increased maximum doses (due to ERE at high-to-low tissue density interfaces) and mean doses (due to more low dose wash).

2.A | Patient selection
This study consists of patients from our hypofractionated partial breast irradiation program who are approved for retrospective analysis by our institutional research ethics board. These are breast cancer patients who did not undergo surgery due to metastatic disease or severe medical comorbidities, the intent being local control and reducing symptom burden. Patients with tumors close to the skin were preferentially selected. Five patients were selected for this study. Two of the patients had bilateral disease, but for this study both tumors were considered separately. Monte Carlo dose calculation algorithm can account for the magnetic field effects on the dose deposition by the radiation beam, which for the MRI-Linac has an energy of 7 MV. We used Monaco for this study as it is possible to simulate and characterize the magnetic field effects using different treatment planning methodologies.
Patients were positioned supine, with the ipsilateral arm raised overhead (for patients with bilateral tumors, both arms were raised overhead). The GTV was contoured on the planning CT with an isotropic margin of 1 cm around the GTV to form the PTV. This PTV margin is likely much larger than required for an MRI-Linac; however, we used this margin as it was used clinically for our patients.
The heart and lungs were previously contoured on the clinically delivered plan. There were two skin contours that were evaluated: Skin3mm and Skin5mm, which were defined as the volumes 3 and was used in conjunction with the GPUMCD algorithm which uses a Monte Carlo approach for dose calculation. One key distinction of GPUMCD is that the dosimetric effects of the magnetic field are calculated in the plan optimization stage and thus deleterious effects such as the ERE can be partially mitigated by inverse planning. 13,17 The beam geometries were carefully controlled such that the TAN, IMRT, VMAT, TANB0, IMRTB0, and VMATB0 configurations can be compared. The TAN and TANB0 geometries were 180°parallel opposed (POP) beams arranged to encompass the PTV whilst minimally passing through normal lung tissue. A 3 rd beam was added (entering from the anterior-lateral oblique direction) if it did not pass directly through the heart and was required to produce reasonable coveragethis was the case for Tumors 1, 3, 4, and 6. Clearly the TAN or TANB0 plans will be far less conformal to the target than the equivalent IMRT and VMAT plans; however, this was done because these will be instructive in understanding magnetic dose effects with very low numbers of beam angles. The five IMRT (and IMRTB0) geometry beams were placed within the span of the TAN POP beams at equal gantry spacings 45°a part, with all beams entering from the anterior oblique direction. Since the MRI-Linac couch cannot rotate relative to the treatment plane, noncoplanar beams were not considered in this study. The VMAT (and VMATB0) arc was also placed within the span of the TAN POP beams, with the single arcing beam also coming from the anterior oblique direction (see Fig. 1 for an example of all beam geometries with Tumor 5, with B 0 turned on). Each tumor, for all three beam geometries and with B 0 on and off, were optimized with identical inverse planning objectives and isocoverage of the PTV (i.e., 99% of the PTV was covered by the 95% prescription isodose). The hot spots within the target were controlled by attempting to keep the V44 < 1%, although breaching this constraint slightly was not a cause for plan rejection (which is also according to our clinical practice). The PTV was evaluated with a modi-   Table 3.
As expected, the mean heart dose parameters were statistically significant between the TAN datasets and the IMRT & VMAT datasets, regardless of whether the magnetic field was on or off-clearly due to the fact that the TAN and TANB0 beam geometries avoided internal OARs almost completely. The differences between VMAT and IMRT plans, with B 0 on or off, are not statistically significant for the heart and lung mean doses. The results of the hypothesis testing for these dose parameters can be visualized more quantitatively in box-whisker plots shown in Figs. 3(b) and 4(b). The max heart dose generally does not have significant differences across most of the configuration pairings, which can be seen graphically in Fig. 3(a).
More interesting is that the differences in the max lung doses have some notable differences when comparing B 0 -on and B 0 -off for both the VMAT and IMRT geometries in the box-whisker plot in Fig. 4(a), although they are not strictly statistically significant differences. This is most likely due to the electron return effect on the lung-tissue interface in the complexly modulated VMATB0 and IMRTB0 plans.  The heart and lung doses were not affected significantly by the magnetic field when using the IMRT or VMAT geometries. The max lung dose (which is a parameter not often looked at as lung is a parallel organ) appears to be slightly increased by the magnetic field for IMRT and VMAT, though in this study was not shown to be strictly a significant difference. The elevated max lung dose in the presence of a magnetic field may be due to the lung-tissue interface at the chestwall-this is particularly prominent in the TANB0 plans,   20 It is worth noting that the mean heart doses reported here are higher than will eventually be implemented in an MRI-Linac because the large PTV margin (1 cm) used in this study will likely be reduced in the future (though this margin reduction is currently unknown for an MRI-Linac).
Skin dose may very well be an important clinical constraint in dose optimization and prescription of HPBI. One study that demonstrates this is the Canadian RAPID study-trials were stopped due to adverse skin toxicity and poor cosmesis in their accelerated partial breast irradiation study arm with a prescription dose of 38.5 Gy in 10 fractions b.i.d. 21 Strategies to reduce unwanted skin dose for HPBI treated by the MRI-Linac are thus desirable. For both skin contours (Skin3mm and Skin5mm), there were large differences apparent due to the B 0 field. Moreover, the skin doses decreased T A B L E 3 Hypothesis testing using one-way ANOVA for various combinations of the six planning configurations in this study, i.e., plans with B 0 on (VMATB0, IMRTB0, TANB0) and plans with B 0 off (VMAT, IMRT, TAN). The bolded entries are those with P < 0.10 and thus are considered statistically significant differences between the data sets.
P values resulting from one-way ANOVA Data set 1 ?  VMATB0  IMRTB0  TANB0  VMATB0  IMRTB0  VMATB0  VMAT  IMRT  VMAT   Data set 2 ?  VMAT  IMRT  TAN  IMRTB0  TANB0  TANB0  IMRT  TAN Max heart dose (a) and mean heart dose (b) box-whisker plots for all six plan configurations. The first and third quartiles are indicated by the ends of the box, with the line in the middle indicating the median. The "whiskers" display the maximum and minimum of the data. No outliers were considered in these plots (these settings apply for all box-whisker plots). with increasing number of beam angles, especially with the B 0 field turned on (i.e., VMATB0 skin doses were lower than IMRTB0 skin doses, which in turn were lower than the TANB0 plans). These differences are quantified in Table 4. So, one can say that in the presence of a magnetic field, an IMRT plan will have on average an 18% increase in the D1cc dose and a 149% increase in the V30 compared to an IMRT plan with no magnetic field. One strategy to reduce the deleterious effects of the magnetic field could be to use more IMRT beam angles or even a VMAT configuration-if a half arc VMAT geometry is used,   The reason for this setup is because any electrons generated above the patient will be swept away longitudinally by the transverse magnetic field, so that with B 0 on there will be effectively zero EC. To our knowledge, this is the first study to compare HPBI plans with different beam geometries and also compare plans with the 1.5 T magnetic field turned on and off. In particular, the comparison between IMRT and VMAT and the magnetic field dose effects on both geometries is novel in the context of HPBI. The work by van Heijst et al. from Utrecht examined magnetic field effects on 7-beam IMRT plans for HPBI; this present study further explores these themes by observing that skin dose is significantly impacted not only by the magnetic field but also varies with depth and varies when increasing the number of beam angles such as with a VMAT implementation.

| CONCLUSION
The impact of a 1.5 T magnetic field transversely placed to the radiation beam on HPBI dose distributions was analyzed using TAN, IMRT and VMAT beam geometries. The heart and lung doses are minimally impacted by the presence of the magnetic field, with the exception of the max lung dose which may be attributed to the ERE.
The magnetic field increases the skin dose; however, the skin dose decreases with increasing number of beam angles. The data suggest that the effects of the ERE and the beam geometry are more impactful on skin dose if evaluated at shallower depths, as we analyzed the skin dose from two different depths: 0-3 mm and 0-5 mm. We expect that there would be lower ERE impact on plans KIM ET AL.

| 69
where the majority of the beam angles causes the beam to enter into and nearby the tumor, since ERE is maximized at the exit point.

ACKNOWLEDGMENTS
The authors wish to thank Elekta AB, Stockholm, Sweden for providing the treatment planning system for this study and specifically Spencer Marshall (Elekta) for useful technical guidance.

CONF LICTS OF INTEREST
The treatment planning system was provided by Elekta AB, Stockholm, Sweden.