2.1 Patient population

This study included eight anonymized postmastectomy breast cases as in our previous study, who received a radiation treatment on tomotherapy with the tissue expander in place. Three of the eight cases received a left-sided breast irradiation. Each patient received 25 fractions to a total dose of 50 Gy, resulting in a total of 200 fractions, 193 of which were included in this study (seven fractions were not available due to un-archiving issues).

2.2 Treatment planning

A clinical plan for each of the three techniques (3DCRT, VMAT, and helical tomotherapy) was created for each of the eight cases, for a total of 24 plans. The following is a brief summary of the treatment planning per technique.

For 3DCRT and VMAT plans, the commercial treatment planning system (TPS) Monaco 5.11 (Elekta AB, Sweden) was used to generate the plans. The beam model used was a 6 MV photon beam from an Elekta Synergy linear accelerator. The tangential 3DCRT plans consisted of two main half-blocked parallel opposed fields that were open anteriorly for plan robustness against breathing motion during the treatment and against swelling of the breasts over the 5 weeks of treatment. Multiple smaller subfields were added to achieve a homogenous dose distribution. No wedges or bolus were used. The 3DCRT plans were calculated using both the collapsed cone (CC) and Monte Carlo (MC) dose calculation engines in Monaco. The MC calculations were performed to 1% uncertainty per control point, corresponding to an approximate uncertainty of 0.7% per plan.

The VMAT plans consisted of a single 230° arc with the isocenter located approximately in the center of the planning target volume (PTV). The PTV was defined as the 5-mm expansion of the clinical extent of the chest wall, including the expander implant and its contents. The PTV was limited in the skin anteriorly and the ribs posteriorly. The target objectives were D90% > 50.0 Gy and D2% < 55.0 Gy. For the OARs, the dose constraints for the ipsilateral lung were Dmean < 16 Gy, V5Gy < 50%, V20Gy < 25%, and V40Gy < 10%. For the heart, the constraints were Dmean < 8 Gy, V5Gy < 50%, V25Gy < 5%, and V30Gy < 2.5%. For the VMAT plans, the MC-based dose calculation algorithm was used. Similar to 3DCRT, the MC dose calculations were performed to 1% uncertainty per control point.

For helical tomotherapy, the original clinical plans were used since this cohort of patients was initially treated using this technique. The helical tomotherapy plans were calculated using the tomotherapy planning station, version 5.1.1.6. (Accuray, CA, USA) which implements convolution/superposition (CS) dose calculation engine.15

2.3 Modeling the measured inter-fractional variations

The daily acquired MVCT images on tomotherapy have reduced artefacts around the metal port which allowed a more accurate localization of the port. The planned adaptive module of tomotherapy was used to load the co-registered daily MVCT and the treatment CT, representing the true position of the patient relative to the planned position in each treatment fraction. In fractions where the metal port on the fused view of the MVCT and treatment planning CT did not align, the distance between the center of metal port on the two fused images was measured in the three cardinal directions. This distance is referred to herein as port positional error. From the previous study,14 the measured errors were generally small, with 87% of positional deviations smaller than 5 mm. However, the errors in the lateral, vertical, and longitudinal directions ranged from −17 to 11 mm, −10.8 to 7.0 mm and −8.0 to 7.0 mm, respectively.

Immobilization and positioning of patients for breast treatments at our center is identical in tangential 3DCRT, VMAT, and tomotherapy. However, tomotherapy is the only technique of the three that acquires daily MVCT images with reduced artefacts around the metal port. Thus, for the purpose of the current study, the inter-fractional positional errors of the metal port that were measured for the cohort of patients treated on tomotherapy were assumed to apply to the tangential 3DCRT and VMAT treatment plans created for the same patients.

Our approach to assessing the dosimetric effect of the measured positional errors is by artificially modeling a shift in the position of the metal port in the TPS and comparing the resulted dose distribution against the originally planned dose. From the viewpoint of the metal port, the observed positional errors can be the caused by internal movement of the port, variations in daily patient setup, or a combination of the two. For modeling purposes, these causes were divided into two classes of error. The first class, referred to as internal port error (IPE), is the port displacement relative to the internal anatomy of the patient, caused by anatomical changes and/or the migration of the whole tissue expander. The second class, referred to as patient registration error (PRE), is the displacement of the whole patient relative to the treatment beam, caused by minor necessary compromises in the position of the patient during patient setup. These compromises are clinical judgments that are made every day when aligning the patient on the treatment couch before daily treatment.

To model IPE, the structure of the metal port was isolated in the treatment planning CT and artificially shifted by the magnitude of the measured daily displacement using an in-house software developed for the purpose of this study. The original structure of the metal port contoured by the treatment planners during image segmentation was used. In slices with heavy metal artefacts in the treatment planning CT, the breast tissue around the metal port was overridden with a density value corresponding to the average breast density in artefact-free slices. For the metal port, the nominal densities of the metal components of the port and shell were assigned. Voxels that were inside the titanium shell and the magnet were assigned Hounsfield unit values of 3926 and 10248, corresponding to physical densities of 4.0 g/cm3 and 8.0 g/cm3, respectively. An example of a corrected CT slice is shown in Figure 1. This was repeated for all fractions per patient, resulting in a separate CT set representing a given fraction, where the metal port is shifted from its original position by the measured error in that fraction. The data quality assurance features of Monaco and tomotherapy were used to calculate the tangential 3DCRT, VMAT, and tomotherapy plans on the modified CT sets of all fractions for all patients.

Computed Tomography (CT) slice of a representative patient before panel (a) and after panel (b) density overrides of the metal port and the tissue surrounding it. The maroon and blue contours are the magnetic core and titanium shell of the metal port, respectively. The area enclosed by the green contour is the region of interest (ROI), defined as a 5-mm expansion around the saline implant. The red contour is the clinically defined planning target volume (PTV). The silhouette of the metal port in panel (b) demonstrates the artificially shifted metal port used for modeling internal port errors

To model PRE, the metal port and the surrounding breast tissue in the original treatment planning CT were assigned density values as described above, with no displacements applied to the metal port. Within the data quality assurance feature of Monaco and tomotherapy, the corrected CT set of each patient was shifted relative to the planned photon fluence by the measured port displacement error of a given fraction, and the original plan of each respective technique was recalculated.

For all eight breast cases, the dosimetric effects of IPE and PRE in each of the three treatment techniques (tangential 3DCRT, VMAT, and helical tomotherapy) were evaluated for two scenarios. The first scenario is the cumulative effect of the daily measured error, called in the results “daily variable” error, for which the calculated dose distributions for all fractions per patient were summed to a cumulative dose distribution. This scenario represents the true course of treatment. The second scenario is the cumulative effect of a systematic realistic large error in port position, called in the results “systematic” error. The magnitude of the error was derived from the largest positional error measured during metal port registration, and the same magnitude of error was simulated in all patients. The systematic error represents a change in port position after the treatment planning CT was acquired that persisted throughout the course of the treatment. For this scenario, the dose distribution of the plan with the simulated large error in port position was multiplied by the number of fractions to yield the equivalent of a cumulative dose distribution. For all three treatment techniques, the cumulative dose distributions were compared with the originally planned dose, that is, the clinical plan calculated on a density-corrected CT set where neither the position of the metal port nor the position of the whole patient was changed. The flowchart in Figure 2 summarizes all modeled scenarios in the study.

Flow chart summarizing the modeled scenarios of port positional errors included in the study. Internal port error (IPE) is displacement of the metal port relative to the internal anatomy of the patient. Patient registration error (PRE) is an error in the position of the port caused by a displacement of the whole patient relative to the radiation beam

In addition, the effect of ignoring the metal port during planning was compared between the three techniques for all patients. To model this, each plan was calculated on a corrected CT scan where the metal port was overridden with tissue-equivalent density. The resulting dose distributions were compared with the originally planned dose, where the density of the metal port was included.

2.4 Robustness analysis

To quantify and compare the robustness of the three treatment techniques, point dose differences and dose volume histogram (DVH) parameters for a clinically meaningful region of interest (ROI) and for relevant OARs were calculated for all patients for the three treatment techniques. The ROI was defined as a 5-mm expansion around the implant, provided that it is within the PTV (rationale below). The percent volume of the ROI receiving 100% of the dose (V100%Rx), the percent volume of the ipsilateral lung receiving 20 Gy (V20Gy), and the percent volume of the heart receiving 5 Gy (V5Gy) were evaluated. The analysis of the OARs was limited to the slices where the ROI was present.

The ROI used for robustness analysis was introduced because the clinically defined PTV contains the large non-biological temporary implant, which can mask local changes to the dose distribution in clinically relevant areas. The expansion of the ROI was truncated to the limit of the clinical PTV. The ROI was defined in the slices where the metal port was present, including the farthest slices that the metal port migrated to during simulation. The newly defined ROI contains tissue directly abutting the implant, which is the more probable location of local recurrence.16