MR-guided breast radiotherapy: feasibility and magnetic-field impact on skin dose

Ten BCT patients treated at the RT Department of the UMC Utrecht in 2011 or 2012 were enrolled in our treatment-planning study. The median age of the women was 63 years (range: 39–72 years). One patient had breast cancer on both sides and for this study both tumours were considered separately. In total, six of the tumours were right-sided and five were located in the left breast. The majority of the excised tumours were stage T1c (7), while the rest were stage T1b (2) or stage T2 (2). No patients had tumour-positive lymph nodes. The median volume of the ipsilateral breast was 1003 cc (range: 705–2373 cc).

The magnetic-field-induced effects were studied for both WBI and APBI. In WBI a dose of 42.56 Gy was prescribed to the breast, delivered in 16 fractions of 2.66 Gy. The first WBI technique considered was the classical tangential two-field set-up using IMRT (WBI-2). The beam angles in WBI-2 were determined such that the two beam edges inside each patient aligned, which was done for each patient individually. In order to investigate the influence of the magnetic field on multiple beam directions, also a seven-field IMRT technique was used for WBI (WBI-7) in which the beam orientations were chosen individually for each patient. For APBI, target prescription was 10 fractions of 3.85 Gy, to a total dose of 38.5 Gy, according to the 2008 IRMA study protocol of the European Organization for Research and Treatment of Cancer Radiation Oncology Group (EORTC ROG). A seven-field IMRT technique with individually determined beam angles was chosen to for the APBI approach (APBI-7).

All patients underwent a planning CT with 3 mm slice thickness, at three weeks (median: 21 days, range: 14–50 days) after surgery, in supine RT position. The palpable breast tissue was indicated with a copper wire. For WBI, the clinical target volume for the whole breast (CTV_WBI) was delineated according to the RT Oncology Group (RTOG) Breast Cancer Contouring Atlas. A 5 mm margin was applied to the CTV_WBI to generate the planning target volume (PTV_WBI), excluding the first 5 mm under the surface (figure 2). The median PTV_WBI was 741 cc (range: 517–2028 cc). Delineations of the target volumes in APBI were performed according to the 2008 IRMA study protocol of EORTC ROG. The gross tumour volume (GTV_APBI) was contoured using preoperative diagnostic imaging, preoperative MRI in treatment position, surgical clips, and possible postoperative seroma. A 15 mm margin in all directions was applied to the GTV_APBI to generate the clinical target volume (CTV_APBI), while excluding the chest wall and skin. The PTV_APBI was delineated as the CTV_APBI with a 5 mm margin, while excluding the skin (figure 2). The median PTV_APBI was 148 cc (range: 88–248 cc). For both WBI and APBI, the delineated OARs were the heart, lungs, contralateral breast, body (comprising all unspecified tissue) and the skin (figure 2). The skin considered for the analysis of the impact of the ERE was defined to be the first 5 mm under the surface of the ipsilateral breast. It was further extended up to 35 mm in anterior, posterior, medial, and lateral directions relative to the PTV_WBI in order to include skin tissue expected to be irradiated in WBI-2 (figure 2). An extension of 15 mm was made in the caudal and cranial directions to include skin tissue receiving scattered dose. The median volume of the skin was 318 cc (range: 256–474 cc).

Zoom In Zoom Out Reset image size

Specially developed MRL treatment planning software (MRLTP) was used to generate IMRT plans while taking into account the presence of a magnetic field. MRLTP is a combination of GPUMCD—which is a graphics processing unit (GPU)-oriented Monte Carlo dose calculation algorithm as described by Hissoiny et al (2011)—and the central processing unit (CPU)-based fast inverse dose optimization (FIDO) as described by Goldman et al (2009). A more detailed description of this system can be found in Bol et al (2012). FIDO has the distinct advantage that the optimization process is executed within a short time frame. The MRL's characteristics are incorporated in the MRLTP. The 6 MV photon beam can rotate in full 360° around the patient with maximum field sizes of 24 cm in caudocranial direction and 56 cm in anterior–posterior direction. As a consequence, the isocentre is fixed, at 14 cm above the treatment table. The isocentre-source distance is 142.7 cm. With MRLTP, plans can be generated at any magnetic-field strength required. IMRT plans were generated for 0 T and 1.5 T, which is the field strength in the MRL. Also, plans were calculated at the intermediate value of 0.35 T. For each IMRT technique (WBI-2, WBI-7, APBI-7) a class solution of optimization objectives was developed at 0 T, which was then applied for all patients. These sets of objectives were unaltered when applied at higher field strengths. The plans were based on the fully optimized fluence maps with a 5 mm grid resolution and the acquired dose distributions consisted of 2 × 2 × 2 mm³ voxels.

For comparison of the plans, dose–volume histograms (DVHs) for all VOIs and corresponding dose parameters were calculated. For the PTV the D_95%—the percentage of the PTV receiving at least 95% of the prescribed dose—to describe PTV coverage, and the D_107%, describing the amount of overdose, were calculated. The starting point of the plan comparisons was a similar D_95% and D_107%.

Several parameters were compared for the OARs. The mean lung dose (MLD) over two lungs was computed, which was to be a maximum of 7 Gy. For the ipsilateral lung, V_{10 Gy}—the fraction of the volume receiving 10 Gy or more—V_{20 Gy} and V_{30 Gy} were derived for the WBI plans while for APBI the V_{5 Gy}, V_{10 Gy} and V_{20 Gy} were found. Similar parameters were calculated for the heart. Moreover, the D_2cc—maximum dose encompassing 2 cc of the structure—was calculated for the heart as an indication of the maximum dose. The range 30–40 Gy is considered to be relevant for the skin dose, so the mean dose and parameters V_{35 Gy} and D_{2 cc} were calculated for the skin. For the V_{35 Gy}, box-whisker plots are derived to determine how the values at different magnetic-field strengths are distributed. Also, the D_{2 cc} of the unspecified tissue (body) and the mean dose of the contralateral breast were acquired, both of which were to be kept low. The statistical significance of differences between all dosimetric parameters was tested by performing a paired student's t-test. A value of p < 0.05 was considered to represent a statistically significant result.

Additionally, dose-difference maps for all plans were derived, which depict for the same patient the dose differences per voxel relative to the situation at 0 T, i.e. dose values at 0.35 T versus 0 T and at 1.5 T versus 0 T. This allows further analysis of the spatial distribution of any dose differences caused by the magnetic field.

IMRT plans were successfully generated using MRLTP. The time required to calculate the Monte Carlo beamlets varied with size of the PTV, the number of beam angles and the magnetic-field strength (i.e. number of electron steps). For this, a computer with 4 GB RAM, a 2.27 GHz CPU, and a GeForce GX 580 GPU was used. Beamlet calculations for a WBI-2 plan took 15 min at 0 T which went up to 25 min at 1.5 T, while a typical WBI-7 set-up took 40 min for 0 T and up to an hour for a 1.5 T plan. Calculating beamlets for APBI-7 typically took 20 min for a 0 T plan, up to 25 min for a plan at 1.5 T.

For each of the three RT techniques, a class solution of objectives was employed for the optimization of all plans. Once a suitable template of objective parameters was found for one patient, an automated script was activated to apply it for optimizing the plans of all other patients. Some individual plans required additional optimization. With the use of a 32 GB RAM, 3.40 GHz, eight-core CPU, the optimization time of FIDO for a WBI-2 plan or an APBI-7 plan was typically in the order of a few seconds, while a WBI-7 plan took only half a minute to optimize.

At the three different magnetic-field strengths 0 T, 0.35 T and 1.5 T, similar target coverage is achieved. Furthermore, no target overdose is observed (table 1).

Relative to the situation at 0 T, the mean skin dose in WBI-2 is raised by 9.5% (p < 0.01) and 12.5% (p < 0.01) at 0.35 T and 1.5 T, respectively (table 1). This dose increase is clearly visible in a typical DVH and dose-difference maps as shown in figures 3(a) and 4(a), (b), respectively. It can also be observed in figures 4(a), (b) that the increase in skin dose is especially manifested at the most superficial voxels of the skin. The most significant change in dose distribution in the skin is found in the range of around 30–40 Gy (figure 3(a)). Further analysis shows that the ERE causes the skin volume receiving more than 35 Gy (V_{35 Gy}) at both non-zero magnetic-field strengths to increase considerably. This can be observed in the box-whisker plot in figure 5(a), where the whiskers of the boxes of V_{35 Gy} at both 0.35 T and 1.5 T do not overlap with those at 0 T, indicating significant differences (p < 0.01). Also, the maximum skin dose (D_{2 cc}) is slightly raised, with an increase of 2.3 Gy at 1.5 T compared to no magnetic field (table 1).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The MLD in WBI-2 is low (∼3.0 Gy) and unaffected by the magnetic field (table 1). This is also true for the V_{10 Gy} of the ipsilateral lung. However, it receives more high dose at increasing field strength as indicated by the V_{30 Gy} going up by 2.2% (p < 0.01) at 1.5 T relative to 0 T. This high-dose volume can also be observed in the dose-difference maps of figures 4(a), (b). The heart dose is unaffected by the presence of the magnetic field in WBI-2. Both the V_{25 Gy} and D_{2 cc} of the heart remain constant. This is also true for the contralateral-breast mean dose, which remains very low. The unspecified tissue receives a higher dose locally as indicated by the D_{2 cc} which is on average 1.2 Gy more (p = 0.02) at 1.5 T than at 0 T (table 1).

Similar target coverage is achieved for all patients in WBI-7 at all field strengths and, again, no target overdoses are observed (table 1).

Although the mean skin dose in WBI-7 is lower than in WBI-2, it is raised by the ERE. This augmentation, relative to the 0 T situation, is on average 8.2% (p < 0.01) and 6.8% (p < 0.01) at 0.35 T and 1.5 T, respectively (table 1). The fact that this increase is less severe than in WBI-2 can be seen in the typical DVH example in figure 3(b) and the dose-difference maps of figures 4(c), (d). The principle change in skin dose is in the range of around 30–40 Gy, which was also observed for WBI-2. Indeed, V_{35 Gy} is increased considerably (p < 0.01) at 0.35 T and at 1.5 T. This can also be seen in the box-whisker plot in figure 5(b), where the interquartile ranges (height of the boxes) of V_{35 Gy}, at both 0.35 T and at 1.5 T, do not overlap with those at 0 T, indicating that there is a significant difference between the medians. It nevertheless shows a smaller increase compared to the tangential technique. The D_{2 cc} of the skin is similar to that in WBI-2 and is in presence of a magnetic field raised by a similar amount (table 1).

The MLD in WBI-7 is higher than in WBI-2 and is increased with respectively 0.6 Gy and 0.7 Gy at 0.35 T and 1.5 T (table 1). The volumes of the ipsilateral lung receiving at least 10 Gy and 20 Gy, respectively, are larger than in WBI-2 and remain almost constant. However, the V_{30 Gy} is smaller relative to WBI-2, and is only enlarged from on average 3.5% (p = 0.17) at 0 T to 3.9% (p = 0.11) at 1.5 T. Due to the beam orientation in WBI-7, the V_{25 Gy} and D_{2 cc} for the heart and mean dose of the contralateral breast are higher than for WBI-2, but they remain constant with increasing magnetic-field strengths. Also, the maximum dose indicated by D_{2 cc} of the unspecified tissue is increased with on average only 0.2 Gy more (p = 0.02) at 1.5 T than at 0 T (table 1). This effect is smaller relative to WBI-2 (figures 4(c), (d)).

For the APBI-7 plans, similar dose coverage is achieved for all magnetic-field values. Furthermore, no overdose is observed for the three situations (table 2).

The impact of the ERE on the skin dose is less prominent in APBI than in WBI. The mean skin dose in APBI-7 is low when compared to the WBI techniques and is only slightly raised when a magnetic field is present (table 2). This can also be observed in the typical DVH in figure 3(c) and is also reflected in the dose-difference maps of figures 4(e), (f). Since the skin dose is already observed to be low, the V_{35 Gy} for the skin in APBI-7 is low and does not increase considerably (figure 5(c)). This can be clearly seen by the boxes of V_{35 Gy} at 0.35 T and at 1.5 T both largely overlapping with the box at 0 T (p = 0.13 and p < 0.01, respectively). Although the increase at 1.5 T is significant, it does not reflect a clinically relevant difference since the absolute change is very small. The maximum skin dose (D_{2 cc}) is lower when compared to WBI and remains unchanged at non-zero magnetic-field strengths (table 2).

The MLD in APBI-7 is low (∼2 Gy) and remains almost constant for non-zero field strengths. Furthermore, there are no observed significant differences (p > 0.05) in the presence of a magnetic field in the V_{5 Gy} and V_{10 Gy} of the ipsilateral lung, while also the V_{20 Gy} remains unchanged. The V_{5 Gy}, V_{10 Gy}, and D_{2 cc} of the heart in APBI-7 remain constant relative to 0 T at both 0.35 T and 1.5 T (p > 0.05 for all parameter differences). The heart V_25 Gy stays zero in APBI-7 (not given in table 2). Relative to WBI, the mean dose of the contralateral breast remains low, whereas the unspecified tissue D_{2 cc} seems to be raised slightly. The maximum raise is 0.5 Gy at 1.5 T compared to 0 T, however, this is statistically insignificant (p = 0.17). Small increases are visible in the dose-difference maps in figures 4(e), (f).