The effect of imaging modality (magnetic resonance imaging vs. computed tomography) and patient position (supine vs. prone) on target and organ at risk doses in partial breast irradiation

Abstract Introduction Conventionally computed tomography (CT) has been used to delineate target volumes in radiotherapy; however, magnetic resonance imaging (MRI) is being continually integrated into clinical practice; therefore, the investigation into targets derived from MRI is warranted. The purpose of this study was to evaluate the impact of imaging modality (MRI vs. CT) and patient positioning (supine vs. prone) on planning target volumes (PTVs) and organs at risk (OARs) for partial breast irradiation (PBI). Methods A retrospective data set, of 35 patients, was accessed where each patient had undergone MRI and CT imaging for tangential whole breast radiotherapy in both the supine and prone position. PTVs were defined from seroma cavity (SC) volumes delineated on each respective image, resulting in 4 PTVs per patient. PBI plans were generated with 6MV external beam radiotherapy (EBRT) using the TROG 06.02 protocol guidelines. A prescription of 38.5Gy in 10 fractions was used for all cases. The impact analysis of imaging modality and patient positioning included dose to PTVs, and OARs based on agreed criteria. Statistical analysis was conducted though Mann–Whitey U, Fisher’s exact and chi‐squared testing (P < 0.005). Results Twenty‐four patients were eligible for imaging analysis. However, positioning analysis could only be investigated on 19 of these data sets. No statistically significant difference was found in OAR doses based on imaging modality. Supine patient position resulted in lower contralateral breast dose (0.10Gy ± 0.35 vs. 0.33Gy ± 0.78, p = 0.011). Prone positioning resulted in a lower dose to ipsilateral lung volumes (10.85Gy ± 11.37 vs. 3.41Gy ± 3.93, P = <0.001). Conclusions PBI plans with PTVs derived from MRI exhibited no clinically significant differences when compared to plans created from CT in relation to plan compliance and OAR dose. Patient position requires careful consideration regardless of imaging modality chosen. Although there was no proven superiority of MRI derived target volumes, it indicates that MRI could be considered for PBI target delineation.


Abstract
Introduction: Conventionally computed tomography (CT) has been used to delineate target volumes in radiotherapy; however, magnetic resonance imaging (MRI) is being continually integrated into clinical practice; therefore, the investigation into targets derived from MRI is warranted. The purpose of this study was to evaluate the impact of imaging modality (MRI vs. CT) and patient positioning (supine vs. prone) on planning target volumes (PTVs) and organs at risk (OARs) for partial breast irradiation (PBI). Methods: A retrospective data set, of 35 patients, was accessed where each patient had undergone MRI and CT imaging for tangential whole breast radiotherapy in both the supine and prone position. PTVs were defined from seroma cavity (SC) volumes delineated on each respective image, resulting in 4 PTVs per patient. PBI plans were generated with 6MV external beam radiotherapy (EBRT) using the TROG 06.02 protocol guidelines. A prescription of 38.5Gy in 10 fractions was used for all cases. The impact analysis of imaging modality and patient positioning included dose to PTVs, and OARs based on agreed criteria. Statistical analysis was conducted though Mann-Whitey U, Fisher's exact and chi-squared testing (P < 0.005). Results: Twenty-four patients were eligible for imaging analysis. However, positioning analysis could only be investigated on 19 of these data sets. No statistically significant difference was found in OAR doses based on imaging modality. Supine patient position resulted in lower contralateral breast dose (0.10Gy AE 0.35 vs. 0.33Gy AE 0.78, p = 0.011). Prone positioning resulted in a lower dose to ipsilateral lung volumes (10.85Gy AE 11.37 vs. 3.41Gy AE 3.93, P = <0.001). Conclusions: PBI plans with PTVs derived from MRI exhibited no clinically significant

Introduction
The rationale for adjuvant radiotherapy is to destroy potential microscopic disease foci in the residual breast after breast conserving surgery (BCS). 1 Increasing cancer survivorship means more women are living longer after initial diagnosis and are at risk of long-term side effects of radiation treatment. 1,2 These may include cardiac disease, 3 pulmonary fibrosis, 4 rib fractures 2 and/or the potential for radiation induced secondary malignancies. 2 There has been increasing interest in delivering radiation to the site of the tumour bed alone, known as partial breast irradiation (PBI) for patients with 'lowrisk', lymph node-negative breast cancer. 5,6 This technique delivers therapeutic dose to the seroma with a margin, as opposed to irradiation of the entire ipsilateral breast, as seen in whole breast irradiation (WBI). 7,8 A smaller treatment volume allows hypofractionation of external beam radiation therapy (EBRT) and aims to reduce dose to organs at risk (OARs), such as the contralateral breast, lungs and heart. 9 Clinical trial evidence 10-12 has demonstrated non-inferiority, in terms of local relapse and equivalent or fewer late normal tissue complications. This applied to a select subset of patients (age > 50, 0-3 involved axillary nodes, grade 1-3 disease and with tumours ≤ 3 cm). This is supported by similar patient selection criteria in additional reviews and longterm trial updates on PBI delivery. 13,14 The use of PBI is demonstrated further by additional trials 13,14 and the development of protocol and consensus guidelines within the United States 5 and the United Kingdom. 6 Target volumes have been typically derived from computed tomography (CT). The use of magnetic resonance imaging (MRI) for defining radiotherapy treatment volumes has gained increased interest for breast cancer due to superior resolution between glandular breast and adipose tissue 15 as the result of intrinsic soft tissue visualisation properties. 16,17 Supine position is most common for breast radiotherapy; 7 however, most breast MRI is performed with the patient in the prone position, to achieve the best signal to noise ratio. 18 Reported dosimetric benefits of the prone position in breast radiotherapy include improved dose distribution in women with large pendulous breasts, 19 due to increased distance between targets and OARs, resulting in reduction of dose to healthy tissues such as the ipsilateral lung. 19,20 Use of this positioning method varies across clinical centres and countries. 7,21 The feasibility of MRI integration into breast cancer radiotherapy planning and treatment has been demonstrated through recent investigations (with the patient in both the supine and prone position). 22,23 Furthermore, seroma cavities are smaller when derived from MRI, compared to CT due to clearer seroma definition. 24 For these reasons, integrating MRI into the radiotherapy planning pathway may result in a smaller region of the breast receiving radiation.
At the time of project commencement, to our knowledge, there was no published data that had evaluated the dosimetric implications of chosen image modality (MRI vs. CT) and patient positioning on target and OAR doses in PBI planning. Therefore, the primary aim investigated if PTV and OAR doses differed when PBI planning is based on target volumes derived from MRI when compared to CT. A secondary aim investigated if patient positioning (supine vs. prone) had an effect on the OAR differences due to imaging modality chosen for target delineation in PBI.

Methods and Materials
Following ethics approval (South Western Sydney Local Health District Human Research Ethics Committee (HREC number: HREC/16/LPOOL/603)), a retrospective cohort of patient data collected from previous investigations, was accessed. 24

Context
The cohort of 35 patients received adjuvant radiotherapy after BCS between July 2012 and November 2014. Each patient underwent MRI and CT scans in supine and prone positions, resulting in 4 image data sets per patient. Supine positioning included the utilisation of a vac bag with flat wingboard and arms over head. Prone position used a 16-channel Sentinelle Breast MRI System for both CT and MRI scans. Further details are provided in previous work, including patient exclusion based on body habitus. 24 CT scans were performed on Phillips CT scanner (Phillips Healthcare, The Netherlands), and MRI was conducted on a MAGNETOM Skyra 3T (Siemens Medical Systems, Erlangen, Germany). Patients were positioned under a coil bridge with a surface coil for the supine MRI scans and using the Breast MRI System for the prone MRI Scans. 22 25,26 of 11 different SC contours (delineated by nine radiation oncologists and two radiologists). Subsequently, supine and prone MRI data sets were rigidly registered to the corresponding CT, to allow radiation therapy planning. The determination of these volumes and the image registration technique has been described in previous work. 24,27 The supine and prone CT image sets each containing CT and MRI determined SC volumes and OAR volumes as determined from CT were used for this study. Figure 1 depicts a typical CT image set and contour information that was available for this study.

Patient selection
From the initial cohort of 35 patients described above, the following criteria were used to determine eligibility for this study; age ≥ 50 years at time of diagnosis and pathology ≤ 3.0 cm (length as stated on histology report). This was based on international consensus statements 5,6 and published clinical trial data [10][11][12] . Laterality, topography (breast quadrant as per ICD-03 topography site notes), 28 age at diagnosis and size of breast were recorded in the patient notes. Other factors typically considered for PBI EBRT eligibility such as histology were not considered as this was primarily a planning investigation, and these characteristics should not affect geometry or dosimetry.

Radiotherapy treatment planning
TROG 06.02 guidelines 29 were followed for the radiotherapy treatment planning aspects of this study. For each data set, clinical target volumes (CTV) and planning target volumes (PTV) were generated from SC volumes. Target expansions and OAR structures used in this investigation are outlined in Table 1.
This planning study utilised a 3 field non-coplanar beam arrangement, as recommended by the TROG 06.02 study protocol. 30 The direction of beam entry varied with patient position. In the supine position, this technique consisted of one lateral tangential beam, one superior medial beam and one inferior medial beam. The lateral beam was placed as close to horizontal as possible without exiting through the contralateral breast. The couch position was angled on the two medial beams to deliver dose from the superior and inferior directions, respectively. In the prone position, the beam arrangement consisted of two lateral beams and one medial beam. A limited range of gantry angles were available for the noncoplanar beams to avoid potential gantry and couch collision. An example of the beam arrangement for a patient with right-sided SC, in the supine and prone position, is demonstrated in Fig. 2.
Plans were generated using 6MV EBRT with wedges for dose optimisation. A prescription of 38.5Gy in 10 fractions was used for all cases. 29,30 All planning calculations were performed on CT data set utilising a 0.25 dose grid and the collapsed cone convolution (CCC) calculation algorithm within Pinnacle (version 9.10, Phillips, Netherlands). 31 Manual optimisations of each plan were performed with the intent to meet the TROG 06.02 trial compliance criteria and are outlined in Table 1. Plans were categorised as compliant when all dose constraints in Table 1 were met. The category of minor violation was only applicable for planning target volume evaluation (PTV EVAL) structures. A plan was categorised as non-compliant if at least one OAR objective was not met. Random plans, that is not all plans, were reviewed for quality assurance (QA), where one of two experienced radiation therapists randomly chose and selected plans to assess based on quality, clinical deliverability and protocol compliance. Additional QA checks were conducted for difficult/non-compliant cases.

Data collection and analysis
Plan compliance or non-compliance was based on criteria outlined in Table 1 using scorecard features in Pinnacle. Additionally, Pinnacle was also used to record the volume of the contoured ipsilateral breast volume and PTV. These values were used to generate the PBI volume ratio, which was a ratio between ipsilateral breast volume to PTV EVAL volume. Digital imaging and communication in medicine (DICOM) files containing; structure, plan and dose data of completed plans were exported to MIM Maestro (MIM Software Inc., OH, USA) (MIM). Plans were assessed on further dose metrics outlined in Table 2, this assessment criteria was based on dose metrics outlined in previous investigations to ensure robustness of plan design. [10][11][12]30 Dose metric data were exported to Microsoft Excel (Microsoft Excel 2019, Microsoft Corporation, USA) for review. Organised data were then analysed using IBM SPSS Statistics Version 26.
To assess the impact of image modality (MRI vs. CT) for target delineation, non-parametric testing was employed. Mann-Whitney U tests were completed to show if there was significant variations in doses received by OARs between groups (MRI vs. CT). This method of analysis was also used to analyse dose discrepancies in individual OARs between different positioning methods (supine vs. prone). To determine if other factors such as topography, laterality, breast volume or seroma size influenced plan compliance, categorical analysis was performed. Separate chi-squared tests were performed between the MRI and CT group to test for a relationship in resulting plan compliance. This method of testing was also used to determine if an association existed between supine and compliant plan or prone and compliant plans. Chi-squared tests were also used to test for an association PB CTV: partial breast clinical target volume, SC: Seroma Cavity, PTV EVAL: planning target volume used for evaluation, DVH: dose volume histogram, IpsiLung: ipsilateral lung, ContraLung: contralateral lung, IpsiBreast: ipsilateral breast, IpsiBreastPRV: ipsilateral breast planning reference volume contour excluding the PTV structure, ContraBreast: contralateral breast. † OARs did not have minor violation criteria. Any plan with OAR dose exceeding described metrics was classed as non-compliant. between PBI breast volume ratio vs. plan compliance, seroma size vs. plan compliance and laterality vs. plan compliance. Fisher's exact test and chi-squared tests were performed to assess for a relationship between ranked ipsilateral breast volume (<500 ml, 501-1000 ml and >1000 ml) and plan compliance. To determine if topography influenced plan compliance, specifically volumes located in the upper inner quadrant (UIQ), a Fisher's exact test was used. A P value of < 0.05 was used to denote significance for all tests.

Results
After PBI patient selection criteria was applied to the initial cohort, 24 patients were deemed eligible for this study. Table 3 displays the patient characteristics, which includes seroma laterality, tumour position and age at diagnosis. Each patient had 4 SC volumes resulting in 96 possible plans. However, from the PBI eligible cohort, a further four patients prone data sets were omitted from the positioning arm due to inadequate alignment of the patient's sternum on the prone breast board. This resulted in inconsistent and incorrect positioning of the contralateral breast and has been documented in previous publications 21,27 . A further patient's prone data sets were removed due to the absence of a prone STAPLE SC volumes. In total, there were 86 data sets (24 patients) included in the analysis of MRI vs. CT and 76 data sets (19 patients) included in the supine vs. prone comparison. Mann Whitney U testing of PTV EVAL did not display a statistical significance for size discrepancies between modalities (MRI 144 cc 3 AE 86 and CT 173 cc 3 AE 96 P = 0.051).

Plan compliance
Plan compliance rates are displayed in Table 4. MRI supine plans resulted in the highest plan compliance rate, of 88%, CT supine had a plan compliance rate of 83%. CT prone had an 84% plan compliance rate and MRI prone had a plan compliance rate of 79%. However, in an analysis of all 86 plans there was no statistical association between image modality and plan compliance; both CT and MRI resulted in 36 compliant and 7 noncompliant plans each. There was also an equal number of compliant plans when positioning was compared, with both positions resulting in 31 compliant plans and 7 noncompliant each.
In all non-compliant plans, the laterality of tumour was left sided and heart dose limitations were unable to be met. Other correlations for non-compliance included observations of small ipsilateral breast volume and tumour located in UIQ. In a comparison of compliant and non-compliant plans, ipsilateral breasts < 500 ml resulted in 72% compliance (23/32 plans), 501-1000 ml resulted in 87% compliance (34/39 plans) and > 1000 ml  Factors with no proven significance include seroma volume when categorised based on size groupings of < 3cm and < 2cm, breast volume ratio (as displayed in Table 4), middle range breasts, where breast volume measures between 501 -1000ml (5/14 non-compliant plans) and large breast size, classified as volume larger than 1000 ml (0/14 non-compliant plans).

Dose metrics
Resulting mean dose metrics for target structures and OARs for all patients are displayed in Table 5. Contralateral lung metrics and the volume of PTV receiving 120% of prescribed dose (PTV D 120% ) were not assessed due to prevalence of zero values. There were no statistically significant differences found between any of the dose metrics when comparing MRI to CT. Statistically significant differences were found between supine and prone positions for the volumes of contralateral breast that received 3% of the prescribed dose (p = 0.011) and volume of the ipsilateral lung that received 30% of the prescribed dose (P < 0.001).

Discussion
At the time of analysis, no previous publications have, to our knowledge, combined MRI derived target volumes (vs. CT) in PBI planning. Imaging modality did not appear to significantly impact OAR doses or the rate of compliance in this cohort, and PTV EVAL sizes did not vary enough to display statistical significance across competing modalities. However, previous investigations using the same patient data sets have demonstrated that STAPLE SC volumes were smaller based on MRI delineation compared to CT. 24 This suggests that potential nuances in size when targets are based on differing imaging modalities may be non-consequential when PTV expansions are applied. As seroma volumes were expanded to PTVs and clipped 5 mm within the patient contour, this study found no significant difference to OAR doses if target volumes are derived from MRI or CT. Although this study has not shown any dosimetric difference in a comparison of MRI and CT derived target volumes, it has shown that the integration of MRI derived target volumes may be unlikely to affect planning dosimetry, resulting in the modality being able to be utilised for its soft tissue visualisation properties, 18,24 without impacting current clinical practice. However, the impact of modality on target expansion in PBI planning should be continued to be investigated.
In this cohort, OAR dosimetry was more favourable for the contralateral breast in the supine position and more favourable for the ipsilateral lung in the prone position. This supports the existing literature that reports that prone positioning is able to reduce overall mean lung dose 19 and statistical significance found in other investigations where dose to the lungs is investigated based on the patients method of positioning. 4,21 The dose to the contralateral breast may have been influenced by issues with patient positioning within the available number of retrospective data sets. Issues with contralateral breast position and  reproducibility in the prone position is supported by previous investigations. 21,27 Unlike the results from our investigation, which showed no significance between dose to the heart and positioning, significant benefit has been found in reduction of heart dose in the prone position, 19 suggesting that further investigations into the prone position and reductions to heart dose should be considered. However, there is conflicting data in regard to the best positioning method for heart sparing, as some investigations have found higher doses when using the prone position. 21,27 Within this group of patients, the PBI volume ratio data were collected for analysis, but not used as the basis of exclusion. TROG 06.02 credentialing publications have demonstrated feasibility of selecting patients for PBI based on the PBI breast volume ratio. 32 This ratio was set at 0.25; however, other publications have not considered for PBI breast volume ratio in patient selection. 8,11,12,14,19 Ultimately, there was no significance between ratio below 0.25 and compliance or conversely ratio larger than 0.25 and non-compliance in this cohort. This indicates that patients should not be excluded for PBI on the basis of PBI breast volume ratio. Patients may also still be eligible for PBI if the seroma is larger than 2 cm, which has been used as selection criteria in recent clinical trials 14,30,32 and < 3 cm as per international consensus statements and guidelines. 5,6,13 Our study reported a 21% incidence of UIQ targets for the imaging study and 26% for the positioning component, which is higher than the 16% reported elsewhere. 33 This increased proportion of patients with tumours in the UIQ likely influenced the incidence of heart dose and non-compliance in this cohort, due to geographical proximity, supporting the findings from this publication that demonstrated an association between UIQ location and non-compliance. There was also a proven association between smaller breast size (< 500 ml) and non-compliance, which is a measure seldom reported in PBI literature, other than credentialing papers for a PBI breast volume ratio. 13,14,30 Therefore, it may be beneficial to consider topography and breast size in patient selection for PBI. However, further investigation on the basis of patient selection or exclusion based on this criterion is warranted.
Due to study design, there was no minor violation metrics for dose to OARs. This was chosen to limit dose to OARs as much as possible and align with previous publications. 30 However, the RTOG 0319 10,11 trial allowed differing dose limitations depending on tumour laterality, specifically for the heart, with a small amount of their cohort resulting in non-compliance due to dose to the heart, 7% in minor violation and 2% in major violation. 11 RTOG 0319 10,11 study design differs to more recent publications, such as TROG 06.02 30 , where dose constraints were not impacted by laterality of tumour. This may have impacted the cohort within our investigation. Every plan that was non-compliant had been impacted by heart limitations and high proportion of UIQ tumours. Strict dose constraints are necessary in limiting potential radiation induced side effects, such as  3 However, using these constraints for left sided tumours resulted in high non-compliance within this cohort, which has not been in the same magnitude in previous publications. 11 This suggests when using stricter constraints PBI may not be feasible for all patients who otherwise meet selection criteria when employing a 3DCRT technique. Using inverse planning techniques such as intensity modulated radiation therapy (IMRT) and volume modulated radiation therapy (VMAT) could be explored as alternative methods of treatment delivery, if heart dose constraints were unable to be met using 3DCRT techniques.
This investigation was limited by the number of retrospective patient data sets available for examination (n = 35) and PBI eligibility criteria in image modality comparisons (n = 24) and in positioning (n = 19). However, the results from this small sample size may be beneficial in the ongoing investigation of MRI integration into radiotherapy practice, as some of the experiences learnt may be preserved to further equivalence studies.
Additionally, patients with a large body habitus had been removed from the initial cohort due to physical limitations of the MRI bore size. Therefore, these results may not apply to patients with a larger body habitus, for which prone positioning may be the most beneficial. 20 To our knowledge, no previous publications have combined MRI derived target volumes (vs. CT) and differences in positioning methods (supine vs. prone) for PBI. Notably, positioning seems to influence dose to OARs. This investigation reports that contralateral breast doses are reduced in the supine position, ipsilateral lung is reduced in the prone position, and there was no significance in dose delivered to the heart in either position.

Conclusion
In respect to our primary aims, PBI plans with PTVs generated from MRI indicated no statistical difference with respect to PTVs and OAR doses or plan compliance and OAR doses were not lowered when PBI planning was based on target volumes derived from MRI. However, statistical significance was found with different methods of positioning, where dose delivered to contralateral breast was lower in the supine positioning and ipsilateral lung doses were reduced in the prone position. Although there was no proven superiority of MRI derived target volumes, the integration of the modality remains the topic of intense interest with recent investigations into breast radiotherapy planning, single dose PBI treatment and the ongoing development of breast MRI imaging sequences. 17,18,22,23 However, more research in a clinical setting is needed to validate the modalities integration.