Prone Breast Radiotherapy on the MR‐Linac: A Technical Note on Skin‐Couch Contact Dose Effects Authors: Kerryn Brown, Drew Smith, Kristie Matthews, Jessica Lye, Leah McDermott, Michael Chao, Farshad Foroudi, Caroline Wright Published: 2026 ABSTRACT ABSTRACTAccelerated partial breast irradiation (APBI) delivers a higher dose to a smaller breast volume with equivalent tumour control to whole breast irradiation and improved cosmesis. The MR‐Linac offers superior soft tissue visualisation and adaptive planning, but limited bore size may result in skin‐couch contact, potentially affecting skin dose. This retrospective pilot study evaluated the effect of MR‐Linac couch contact on skin dose. Three patients with left‐sided breast cancer positioned prone with unavoidable skin‐couch contact were included. MR images were acquired using a 1.5 T MR simulator and contoured for target and organs‐at‐risk. Step‐and‐shoot IMRT plans (30Gy in 5 fractions) were optimised and recalculated with and without magnetic field presence. Dose to the whole skin rind (Skin_Rind) and couch‐contact region (Skin_Rind_Ant) were compared. All plans met target and OAR constraints. Mean Skin_Rind dose was consistently higher with magnetic field presence, while dose to Skin_Rind_Ant was consistently lower. These findings contrast with conventional linac data, suggesting magnetic field and interface effects may alter skin dose deposition. Background Accelerated partial breast irradiation (APBI) is a technique where a higher fractionated dose is delivered across fewer fractions to a smaller, targeted area of the breast. APBI for early‐stage breast cancer provides equivalent local recurrence rates to whole breast irradiation [1, 2, 3, 4, 5]. With fewer travel barriers for patients [4, 6, 7]. The magnetic resonance linear accelerator (MR‐Linac), an integrated linear accelerator and magnetic resonance imaging (MRI) scanner, provides superior soft tissue imaging compared to computed tomography (CT), enhancing APBI accuracy, especially in patients with limited postoperative markers. The MR‐Linac enables real‐time visualisation of target volume and surrounding anatomy, allowing higher radiation doses with reduced margins through daily adaptation. This approach has been shown to reduce toxicities across a range of body sites [8, 9, 10]. Simulation and treatment in the prone position is preferred for APBI in our department [11, 12, 13, 14, 15]. The prone position allows breast tissue to fall away from the chest wall, lungs, and heart, increasing distance between the target and these organs‐as‐risk (OAR)s. This particularly benefits patients with large/pendulous breasts, where prone positioning minimises breast folds and improves dose homogeneity [12, 16, 17, 18, 19, 20]. Additionally, respiratory motion may be reduced [13, 21, 22]. The diameter of the MR‐Linac bore and fixed placement of the anterior imaging coil limit patient positioning and may prevent adequate elevation to avoid skin‐couch contact in some patients [23]. In our experience, this has occurred in approximately 15%–20% of simulated patients. Skin‐couch contact is concerning as it may increase skin dose due to absorbed scattered radiation from the treatment couch, as demonstrated on conventional linacs [24, 25, 26, 27, 28, 29]. Several studies have raised concerns regarding skin dose for breast treatments on the MR‐Linac [11, 30, 31] van Heijst et al. [32] reported increasing skin dose with stronger magnetic fields, with mean (standard deviation (SD)) of 5.2 Gy (2.1), 5.6 Gy (2.4) and 5.8 Gy (2.4) for 0 T, 0.35 T and 1.5 T, respectively. Similarly, Raaijmakers et al. [33] reported higher skin dose with stronger magnetic fields and large oblique beam angles. Increasing beam number has been proposed to mitigate this effect [11, 31, 32]. However, others recommend avoiding beam entry through stabilisation equipment and high‐density couch structures due to dose uncertainty [34, 35]. Depending on target location, this may limit beam entry points. Skin toxicity is a clinically relevant concern in high dose per fraction breast irradiation and may negatively affect patient quality‐of‐life, including physical pain or discomfort, psychological distress and interference with daily activities, making skin‐couch contact during prone APBI on the MR‐Linac an important consideration. Additional MR‐Linac specific phenomena include the electron return effect (ERE) [30, 35, 36] and electron stream effect (ESE) [37, 38, 39]. ERE occurs when secondary electrons are redirected by the magnetic field back into tissue. This increases dose at tissue‐air interfaces and draws the dmax closer to the skin compared to a conventional linac [30]. ESE sees these electrons forced along the direction of B0, depositing dose if contacting the chin, forearms or abdomen. These phenomena alter superficial dose distributions compared to conventional linacs. The report of AAPM Task Group 176 reviewed studies investigating couch effects on skin dosimetry [40]. All values were based on conventional linacs so did not consider magnetic field effects. To date, no studies have investigated skin‐couch contact for prone breast treatments on the MR‐Linac. Given the increasing use of MR‐Linacs for APBI, this study aims to evaluate the dose effects of skin‐couch contact and provide preliminary data to inform future research. Methods and Materials Patient Selection Ethics approval was granted from Austin Health Human Research Ethics Committee (HREC/108373) for this retrospective observational study of female adults with prior RT for primary breast cancer. Three patients were randomly selected for analysis. Imaging and Planning MR images were obtained as per standard‐of‐care using a 1.5 T Ingenia Ambition X MR‐Simulator (Philips, Amsterdam, Netherlands). Patients were positioned prone on a custom MRI‐safe prone breast board with both arms abducted. This allowed the target breast to fall away from the body whilst maintaining the position of the contralateral breast and other OARs. Within MIM Maestro 7.1.6 (GE, Boston, Massachusetts), gross tumour volumes (GTVs) were contoured by a Radiation Oncologist (RO). Normal tissue contouring was completed by a Radiation Therapist using an automated contour workflow developed in‐house for MR images. No manual edits to the skin were required. Skin‐rind structures were generated using a standardised 0.5 cm contraction of the outermost skin for consistency. A whole‐body skin‐rind (Skin_Rind) and a skin‐rind cropped to the couch contact region (Skin_Rind_Ant) were created Figure 1. A clinical target volume (CTV) was created by a 1.0 cm expansion of the GTV, cropped to muscle and skin‐rind. A planning target volume (PTV) was created by a 0.5 cm expansion of the CTV, cropped to lung and skin‐rind. Schematic of pendulous breast in contact with treatment couch. Skin_Rind is the first 5 mm of outermost skin. Skin_Rind_Ant is the Skin_Rind contour cropped to the area in contact with the couch. Approximate GTV locations for patients 1, 2 and 3 shown as solid, dotted and dashed circles, respectively. Datasets were transferred to Monaco 5.51 Treatment Planning System (TPS) (Elekta, Stockholm, Sweden) for planning. As MR intensity values do not correspond to electron density (ED), population‐based densities were applied to anatomical structures, consistent with Jonsson et al. [41] Tissue types, densities and layering are listed in Table 1. An ‘External’ contour including the patient, and excluding couch and coil structures, was used to aid dose calculation at the skin couch contact area. TABLE 1List of tissue types and MR‐Linac couch structures, and their associated densities.LayerContoured structuresAssigned density (g/cm3)LayerMR‐Linac couch structuresAssigned density (g/cm3)1Clip a 1.50011Z1—Bridge1.6742High‐density tissue a 1.10012Z2a—Bridge1.6743Seroma a 1.10013Z2b—Bridge0.0694Lungs0.20014Z3—Bridge0.3175Heart1.00015Z4—Couch Support1.6746Chestwall1.10016Z5—Hard Plate1.8027Sternum1.10017Z6—Couch Support0.8998Muscle1.00018Z7—Couch Support0.9909Skin0.90019Z8—Mattress0.10410Prone breast board0.14020Z9—Ant Coil base0.18524External0.01021Z10—Couch Support0.14122Z11—Ant Coil0.016 b 23Z12—Post Coil0.068 b a As required. b Average value used. List of tissue types and MR‐Linac couch structures, and their associated densities. As required. Average value used. Step‐and‐shoot intensity modulated radiation therapy (IMRT) plans with an Elekta Unity MR‐Linac (Elekta, Stockholm, Sweden) beam model were created for each patient (MRL1.5T). Plans used a Monte Carlo algorithm, 0.2 cm grid spacing and 1% statistical uncertainty. Due to the MR‐Linac fixed isocentre, each plan was aligned to the superior/inferior centre of its respective PTV, with the left/right and anterior/posterior position remaining unchanged. Prescription dose was 30 Gy in 5 fractions. For all three patients, an eight‐beam template was applied. One patient required three beam modifications, < 5° each. Each beam avoided lateral high‐density couch support structures and maintained a clearance of 1 cm from the breast‐board. The skin‐rind contour was included as an optimisation structure within the plan template to ensure a dose of 28.5 Gy was not exceeded. Each plan met departmental dosimetric criteria. To separate magnetic field influences from the dose effect of skin‐couch contact, each optimised plan was recalculated using the MR‐Linac beam model with identical parameters but without the magnetic field present (MRL0T). All other planning parameters remained the same. To ensure robust planning procedures and to remove any potential confounding bias, the above process was repeated for each patient, this time optimising plans using the MR‐Linac beam model with no magnetic field first (Conv0T), then recalculating using the MR‐Linac beam model with magnetic field present (Conv1.5T). Plan quality, including target coverage and OAR dose, was assessed as per departmental protocol. Statistical Analysis Descriptive statistics summarised trends across patients. Paired t‐tests and linear mixed effects modelling were used to examine dose differences and to assess optimisation order. Formal testing of distributional assumptions was not feasible. Given the small sample size (n = 3), all analyses were exploratory, and p‐values were only reported for completeness and should not be interpreted as confirmatory evidence. Statistical analysis was conducted using Stata version 19.0 (StataCorp, College Station, Texas). Results Three patients with left‐sided, grade 1–2 breast cancer, aged 54, 62 and 57 years, were included. Tumours were located in the upper outer quadrant for two patients (2 o'clock, 11 cm from nipple and 3 o'clock, 7 cm from nipple) and lower outer quadrant for one patient (4 o'clock, 7 cm from nipple). Tumour bed sizes ranged between 2.00–4.92 cc, CTVs 29.65–41.10 cc, and PTVs 62.83–80.83 cc. Ipsilateral breast volumes ranged between 1571.18–1641.07 cc. All six optimised plans—two for each patient—met dosimetric and homogeneity criteria for clinical plan acceptance as per departmental protocol. Doses to PTV and OARs were largely unchanged between plans with magnetic field presence and without magnetic field presence; however, differences were observed for the skin rind structures. The differences are visually demonstrated in Figure 2, with dose distributions and DVHs showing increased dose to Skin_Rind and reduced dose to Skin_Rind_Ant with magnetic field presence. The mean (SD) dose to Skin_Rind for plans with the presence of the magnetic field was consistently higher, both when optimising with magnetic field presence (MRL1.5T: 24.74 Gy (2.81 Gy) vs. MRL0T: 21.89 Gy (2.94 Gy), p = 0.012) and without magnetic field presence (Conv0T: 23.04 Gy (3.49 Gy), p = 0.004 vs. Conv1.5T: 25.67 Gy (3.78 Gy)). In contrast, Skin_Rind_Ant dose was consistently lower for plans with magnetic field presence, both when optimising MRL1.5T: 6.80 Gy (1.21 Gy) vs. MRL0T: 8.13 Gy (1.49 Gy), p = 0.051, and Conv0T: 8.14 Gy (1.79 Gy), p = 0.022 vs. Conv1.5T: 6.76 Gy (1.46 Gy). Individual patient D0.035cm3 values for Skin_Rind_Ant and Skin_Rind, according to magnetic field presence and optimisation order, are shown in Figure 3. (a) Dose differences with and without magnetic field presence (MRL1.5T—MRL0T) for patient 1, 2, and 3. (b) Dose Volume Histograms for each patient displaying PTV, Skin_Rind_Ant and Skin_Rind structures (MRL1.5T plans = solid lines, MRL0T = dashed lines). Increased dose to Skin_Rind and reduced dose to Skin_Rind_Ant are observed with magnetic field presence. Maximum dose (D0.035cm3) to Skin_Rind_Ant (left) and Skin_Rind (right) for each patient, comparing plans with and without magnetic field presence. Results are shown for both optimisation orders. Across all patients, magnetic field presence is associated with reduced dose to Skin_Rind_Ant and increased dose to Skin_Rind. When considering the influence of the magnetic field and optimising order in a linear mixed effects model (Table 2), a higher dose was observed for Skin_Rind within the presence of the magnetic field and lower Skin_Rind_Ant dose after adjusting for the optimising order. The optimising order had no apparent effect for the Skin_Rind_Ant dose, while optimising with the magnetic field presence first appeared to lower the Skin_Rind dose. This pattern is consistent with magnetic field‐induced interface effects. TABLE 2Values of linear mixed effects models for Skin_Rind and Skin_Rind_Ant.VariablesCoefficient95% CI p Skin_RindMagnetNo magnetic presence0With magnetic presence2.7431.199 to 4.288< 0.001Optimising OrderNo magnetic presence0With magnetic presence−1.041−2.585 to 0.5040.187Skin_Rind_AntMagnetNo magnetic presence0With magnetic presence−1.357−1.912 to −0.801< 0.001Optimising OrderNo magnetic presence0With magnetic presence0.010−0.546 to 0.5650.973 Values of linear mixed effects models for Skin_Rind and Skin_Rind_Ant. Discussion This retrospective study reports skin dose effects for patients with breast cancer eligible for prone treatment on our department's MR‐Linac, with unavoidable skin‐couch contact. The results showed that regardless of optimisation order, Skin_Rind was consistently higher, and Skin_Rind_Ant was consistently lower with a magnetic field presence. At present, we aim to position patients for APBI prone without skin‐couch contact, increasing distance from the couch where possible. If this is unachievable, the patients are assigned supine treatment. Whilst the magnitude of absolute dose changes in this study were small, 2.33–3.49 Gy, and remain within departmental tolerance levels, the clinical relevance of these differences should be interpreted in the context of hypo‐fractionated breast irradiation. Large dose per fraction and limited fraction number may increase the sensitivity of skin toxicity endpoints to even the most modest dose perturbations. Adhering to skin dose criteria during adaptive planning may mitigate this issue, although caution is required if patient position deviates during treatment. In the context of hypo‐fractionated regimens, these changes in superficial dose may still be clinically relevant given the spatial heterogeneity of skin dose distribution and the potential for localised surface hotspots, which may not be fully captured by departmental tolerance metrics. From a planning perspective, this may be addressed through optimisation to skin structures, careful beam arrangement, and minimising skin‐couch contact where achievable. The preliminary data presented in this study suggest a potential skin‐dose sparing effect associated with skin‐couch contact; however, analysis using a larger population would be valuable to better inform any possible change to clinical practice. The observed increase in Skin_Rind dose with magnetic field presence is consistent with ERE, where secondary electrons are redirected by the magnetic field back into tissue, increasing superficial dose deposition. This likely explains the higher dose in the Skin_Rind structure observed across all patients. Conversely, the reduction in dose to Skin_Rind_Ant at the skin‐couch interface may reflect the absence of a tissue‐air boundary at this interface, reducing the magnitude of ERE. This may limit ERE to the patient surface, contributing to the lower Skin_Rind_Ant dose. Whilst the couch mattress has only a small density difference to air, the presence of the structure may intercept some of the secondary electrons, thereby reducing superficial dose. These findings suggest magnetic field‐induced interface effects may partially explain the differential dose patterns observed between contact and non‐contact regions. ESE was not expected to influence the Skin_Rind_Ant region as ESE primarily affects tissue surfaces aligned with the magnetic field direction. Couch composition is also a factor Modern couches on conventional linacs are constructed of carbon fibre and polymeric foam with density equivalent to air. Studies have shown the carbon fibre material acts as a bolus structure which increases anterior skin dose, and therefore maintaining an airgap of 2–5 cm between skin and couch is recommended to minimise this dose increase [40, 42]. Maintaining an airgap on a conventional linac is more feasible than on the MR‐Linac, due to the limited physical space. The MRI‐safe materials used in MR‐Linac couch construction are a complex combination of glass fibre, various polymers, and materials needed for imaging coils Figure 4. Whilst the carbon fibre of a conventional linac couch causes a bolus effect whereby increasing anterior skin dose, this study has shown a decrease in skin dose where there is skin‐couch contact on the MR‐Linac. Additionally, secondary electrons may deposit their energy within the low‐density material of the mattress rather than returning to the skin surface. The magnetic field may also alter electron trajectories within the couch material. Secondary electrons generated within the low‐density components may experience ERE within the material rather than at the tissue‐air interface, potentially contributing to the observed reduction in Skin_Rind_Ant dose. Whilst tumour position and beam arrangement may have had some influence on individual data points, the consistent direction of the effect across all three patients suggests a potentially reproducible dosimetric effect. It should also be noted that the findings are based on TPS calculations. Although Monte Carlo algorithms have shown to provide accurate dose modelling, [43, 44, 45, 46] validating surface dosimetry with phantom measurements and film may strengthen future studies. Cross section of the lateral components of the Elekta Unity MR‐Linac couch. 1Glass fibre, 2Glass fibre covered foam, 3Polyethylene, 4High density polyurethane, 5Foam, 6Foam, Circuit board, Adhesive, Copper traces. Further methodological considerations exist. Densities of patient tissue types were selected based on established values used within our department; however, it is noted that in wider clinical practice, selection of tissue density values varies between centres. Some centres limit density options to just tissue, air, and bone; however, other centres use individual densities. The latter may report dose with increased accuracy [47] and further investigation is recommended to evaluate any clinical significance between the two approaches specifically for breast treatments. However, this issue is beyond the scope of this study. Finally, the cohort consisted of three patients with left‐sided, centrally placed tumours. Whilst all patients had similar target locations in this study, in clinical practice locations vary. Alternative tumour locations may alter skin‐couch interaction, and dose effects may differ accordingly. Evaluation of varied tumour positions warrants further investigation. The small patient number was chosen to provide initial insight; however, this is a key limitation of this study. These results are preliminary, should not be generalised, and validation of the study with a larger cohort is warranted. Conclusion This study highlights the effect of couch contact within the magnetic environment of a MR‐Linac on skin‐dose in prone APBI compared to that observed on conventional linacs. Further investigation using a larger number of patients with a broad range of tumour locations will be undertaken in future research to assess overall trends. Funding This study was conducted with the financial support of a research grant from the Victorian Medical Radiation Practitioner's Education Trust [VMRPET] And the Australian Government's Research Training Program [RTP]. The funding sources did not have input into the design of the study protocol, data collection, analysis, interpretation or writing of the manuscript. Ethics Statement Ethics approval was granted by the Austin Health Human Research and Ethics Committee (HREC), reference HREC/108373/Austin‐2024. Consent The authors have nothing to report. Conflicts of Interest The authors declare no conflicts of interest.