Dosimetric effects of intrafractional isocenter variation during deep inspiration breath‐hold for breast cancer patients using surface‐guided radiotherapy

Abstract The aim of this study was to investigate potential dose reductions to the heart, left anterior descending coronary artery (LAD), and ipsilateral lung for left‐sided breast cancer using visually guided deep inspiration breath‐hold (DIBH) with the optical surface scanning system Catalyst™, and how these potential dosimetric benefits are affected by intrafractional motion in between breath holds. For both DIBH and free breathing (FB), treatment plans were created for 20 tangential and 20 locoregional left‐sided breast cancer patients. During DIBH treatment, beam‐on was triggered by a region of interest on the xiphoid process using a 3 mm gating window. Using a novel nonrigid algorithm, the Catalyst™ system allows for simultaneous real‐time tracking of the isocenter position, which was used to calculate the intrafractional DIBH isocenter reproducibility. The 50% and 90% cumulative probabilities and maximum values of the intrafractional DIBH isocenter reproducibility were calculated and to obtain the dosimetric effect isocenter shifts corresponding to these values were performed in the treatment planning system. For both tangential and locoregional treatment, the dose to the heart, LAD and ipsilateral lung was significantly reduced for DIBH compared to FB. The intrafractional DIBH isocenter reproducibility was very good for the majority of the treatment sessions, with median values of approximately 1 mm in all three translational directions. However, for a few treatment sessions, intrafractional DIBH isocenter reproducibility of up to 5 mm was observed, which resulted in large dosimetric effects on the target volume and organs at risk. Hence, it is of importance to set tolerance levels on the intrafractional isocenter motion and not only perform DIBH based on the xiphoid process.


| INTRODUCTION
Adjuvant radiotherapy for breast cancer reduces the risk of locoregional recurrence as well as breast cancer death. 1,2 However, some radiation is inevitably delivered to normal tissue, such as the heart and lungs, which has been shown to increase the risk of cardiovascular and pulmonary disease. [3][4][5][6][7][8][9] Darby et al. 5 have shown that the relative risk of ischemic heart disease increases with 7.4% per Gy increased mean heart dose, with no apparent threshold. This relationship was recently validated by van den Bogaard et al. 8 for more modern radiotherapy techniques. Also, a higher incidence of coronary artery disease has been observed for the left anterior descending coronary artery (LAD) for left-sided compared to right-sided breast radiotherapy. 10 This could possibly reduce the survival benefit of breast cancer radiotherapy.
Since a large proportion of the breast cancer patients are cured from their disease and hence become long-term survivors, with the 5-year survival being approximately 90%, 11 it is important to reduce the late side-effects as much as possible. Therefore, there has been much focus in the last years in breast cancer radiotherapy to develop treatment techniques that reduce the dose to normal tissues, such as treatment during deep inspiration, prone patient positioning, intensity modulated radiotherapy, proton radiotherapy, and partial breast radiotherapy. 12 Treatment during deep inspiration has been shown to decrease the cardiopulmonary doses without compromising target coverage, due to increased spatial distance between the organs at risk and the target as well as decreased lung density. [13][14][15][16] Treatment in deep inspiration breath-hold (DIBH) requires patient compliance and the use of visual guidance has been shown to improve intrafractional reproducibility of the inspiration level. 17,18 Several techniques for tracking the breathing motion have been introduced in radiotherapy, such as measuring the motion extent of external markers or the pressure in a belt or the variation in air flow.
The latest techniques involve optical surface (OS) scanning systems such as the Sentinel TM and Catalyst TM (C-rad Positioning AB, Uppsala, Sweden) or AlignRT (VisionRT, London, UK). The systems project light onto the patients' skin surface and reconstruct a three-dimensional surface of the patient. The OS system detects the patients' position and movements and is used to trigger the beam for treatment delivery in DIBH. 19 Several studies have evaluated patient setup accuracy during DIBH for left-sided breast cancer for 3D surface matching, [20][21][22] but few have investigated any dosimetric effects due to potential positioning deviations. For instance, Tang et al. 23 evaluated the dosimetric impact of motion during DIBH for an iterative closest point (ICP)-based algorithm for surface matching with the AlignRT system, using a AE 3 mm and AE3°tolerance for translational and rotational differences. They reported very small (<1 mm) breath-hold motion and the dosimetric consequences were found to be small. However, only the rigid motion of the surface was investigated.
By using a novel nonrigid registration algorithm as well as finite element simulation of underlying tissues, the real-time isocenter position can be determined from the surface motion. Thus, the optical surface scanning system (Catalyst TM ) can not only track the surface but also the real-time isocenter position. 24,25 Several studies have shown dosimetric benefits of the DIBH treatment technique for breast cancer patients [13][14][15] and several studies have shown increased accuracy in patient positioning using optical surface scanning. [20][21][22] However, to the best of our knowledge, no investigation of potential dosimetric effects due to intrafractional isocenter motion during DIBH has been carried out and are thus highly desirable.
In this study, audio and visual guidance were used for the patients to achieve reproducible DIBHs. The surface over the xiphoid process worked as the surrogate for the target position for beam triggering during both CT imaging and treatment. During DIBH at the treatment machine, the Catalyst TM system was used for triggering the beam when the xiphoid process entered the gating window and simultaneously tracked the isocenter position. The intrafractional DIBH isocenter reproducibility in between breath holds was investigated and the subsequent dosimetric effects evaluated for both tangential and locoregional treatment of left-sided breast cancer.
The aim of this study was to investigate potential dose reductions to organs at risk (OARs) using DIBH and optical surface scanning, and further evaluate how any dosimetric benefits are affected by possible intrafractional isocenter motion in between breath holds.

2.A | Ethical consideration and consent
The use of the radiotherapy database for retrospective research has been approved by the Regional Ethical Review Board in Lund (No. 2013/742).

2.B | Patient selection
A total of 40 patients receiving radiotherapy for left-sided breast cancer in DIBH were enrolled in this study, 20 patients received tangential treatment after breast-conserving surgery and 20 patients received locoregional treatment after either breast-conserving surgery or mastectomy. The patients started treatment between September 2015 and August 2016. The median age was 59 yr (range: 45-77 yr) for the group receiving tangential treatment and 46 yr (35-85 yr) for the group receiving locoregional treatment.

2.C | Computed tomography simulations and treatment planning
All patients underwent supine computed tomography (CT) in separate scans for free breathing (FB) and DIBH. Images with a slice thickness of 3 mm were acquired using a Siemens Somatom definition AS plus (Siemens Medical Solutions, Erlangen, Germany). The prospective DIBH study was performed with the Sentinel TM system (C-rad Positioning AB, Uppsala, Sweden) using laser (k = 635-690 nm) to create a reference surface of the patient in FB and record the breathing motion. 26 The patients were scanned with the Sentinel TM system in FB and a region of interest (ROI) was defined on the surface of the skin above xiphoid process in a shape of a circle with a diameter of 2 cm.
Motion in the vertical direction in the ROI was registered as the breathing signal. Since the registered motion was in the vertical direction only, the rather flat and stable surface on sternum was chosen.
The position of the end-expiration for FB level, that is, the baseline, was automatically tracked and the amplitude was individually set for each patient at the maximum comfortable and reproducible breath hold level. The gating window was 3 mm for all patients included.
Visual guidance with video goggles was used to help the patients keep the inspiration level in the gating window during the CT acquisition.
To avoid abdominal breathing or so-called fake breathing, that is, arching the back, the patients were asked to breathe with their chest during a short training session of a few breath holds using visual feedback prior to CT imaging. For patients receiving locoregional treatment after mastectomy, a bolus (Superflab, Mick Radio-Nuclear Instruments, Inc. An Eckert & Ziegler BEBIG Company) was placed on the thoracic wall over the operation scar with a 3 cm margin at CT.

2.C.1 | Structure delineation
All structures were delineated in both the DIBH and FB CT sets by the same radiation oncologist, to reduce the interobserver variability. For tangential breast irradiation after breast-conserving surgery, the PTV was defined as the clinical limits of the remaining breast including all glandular tissue. For cases receiving locoregional treatment after breast-conserving surgery, ipsilateral axillary lymph nodes level II-III and lymph nodes in the supra-and infraclavicular fossa were included in the PTV. Thus, the internal mammary nodes were not included. The CTV-T was delineated as the tumor's position in the breast with at least 10 mm margin, approximately equivalent to a quadrant of the breast. After mastectomy, the PTV was defined as the part of the thoracic wall where the breast had been located (visualized on CT scans by markers), including ipsilateral lymph node stations as above. No CTV-T was delineated for these patients. For all patients, the PTV was cropped 5 mm from the skin surface.
The OARs delineated were the heart, LAD, and the ipsilateral lung. The lung was automatically delineated using the segmentation wizard in the treatment planning system (TPS, Eclipse, version 13.6.30, Varian Medical Systems, Palo Alto, CA, USA) and then manually verified. The heart was defined as the entire myocardium including the large vessels up to the departure of the coronary arteries from aorta ascendens. LAD was delineated with a 6 mm diameter from the vessels departure from aorta as far as it could be visualized, often to the middle of the heart. The heart and LAD were delineated manually and all OARs were delineated without margins.

2.C.2 | Treatment planning
The treatment plans were created in the Eclipse TPS for an Elekta Synergy (Elekta AB, Stockholm, Sweden) and the dose was calculated using the anisotropic analytic algorithm (AAA, version 10.0.28 and 13.6.23). The prescribed dose was 50 Gy in 25 fractions, normalized to the PTV mean dose. All treatment plans were made by one dosimetrist, for the plans to be comparable between DIBH and FB. The main goal when creating the treatment plans was to fulfill the constraints presented in Table 1, based on national guidelines for 2014-2016 from the Swedish Breast Cancer Group (www.swebcg.se). At the same time, the OAR doses were kept as low as possible.
Essentially identical three-dimensional conformal radiotherapy (3D-CRT) isocentric treatment plans were created for both DIBH and FB, where only minor differences in the beam arrangement were allowed to achieve comparable target coverage. For tangential treatment, two tangential opposing fields were used to irradiate the breast. If needed for dose homogenization, wedges and/or supplementary fields were used. The energy 6 MV was used for all fields. field. Additionally, a 10 MV PA field shielding the lung was added.
For tangential treatment, the isocenter was placed in the center of the breast, and for locoregional treatment, the isocenter was placed in the junction between the tangential and AP/PA fields.

2.D | Dosimetric comparison between DIBH and FB treatment plans
For the comparison between the DIBH and FB treatment plans, the mean dose to the heart (D mean,heart ), LAD (D mean,LAD ), and ipsilateral lung (D mean,lung ), the dose received by 2% of the volume for the heart (D 2%,heart ) and LAD (D 2%,LAD ), the volume receiving 20 Gy for the ipsilateral lung (V 20Gy,lung ) and the dose received by 98% of the PTV volume (D 98%,PTV ) were retrieved from the TPS. Two-sided paired Wilcoxon tests were carried out to investigate if the differences between DIBH and FB were statistically significant, using a significance level of 0.05.

2.E | The DIBH treatment workflow using
Catalyst TM When the treatment plan was finished, the plan isocenter, treatment fields, and plan UID were exported in DICOM format from the TPS to the Catalyst TM system. To assess the treatment isocenter for patient positioning in FB, the reference surface from the Sentinel TM T A B L E 1 Constraints aimed to be fulfilled when creating the treatment plans.

Structure
Constraint Two of the three algorithms provided by the Catalyst TM system with calculations for beam triggering for DIBH treatments were used in this study. 24 The first algorithm was calculating the separation in z-direction between the REF (Treat in DIBH) and the live surfaces in an area predefined by the ROI. When the respiratory signal recorded in the ROI was within the gating window, the treatment beam was triggered [ Fig. 1 DVHs, the D mean,heart , D mean,LAD, D mean,lung , D 2%,heart , D 2%,LAD , V 20Gy, lung , and D 98%,PTV were obtained and the minimum and maximum values for each patient and probability level were considered for the motion-induced dose effect evaluation.

3.A | Dosimetric comparison between DIBH and FB treatment plans
Overall, the dose to the heart, LAD, and ipsilateral lung was reduced for all dose levels with comparable target coverage for DIBH compared to FB for both tangential and locoregional treatment [ Fig. (2)].
The heart and LAD mean doses and D 2% were reduced for essentially all patients using DIBH and the mean lung dose and V 20Gy were reduced for the majority of the patients [ Fig. (3)].

3.B | Assessment of the intrafractional DIBH isocenter reproducibility
The cumulative probability of having an intrafractional DIBH isocenter reproducibility less or equal to a certain value was calculated in the lat, long, and vert directions and are presented for the two patient groups in Fig. 4. For tangential treatment, 50%/90% cumulative probabilities of having intrafractional DIBH isocenter reproducibility less or equal to 1.4/3.2, 1.1/3.1, and 0.9/2.1 mm in the lat, long, and vert directions, respectively, were observed (Fig. 4). The corresponding values for locoregional treatment were 0.6/1.8, 0.9/2.3, and 0.7/2.0 mm.

3.C | Analysis of dosimetric effects induced by intrafractional DIBH isocenter reproducibility
The dosimetric effect of the intrafractional DIBH isocenter reproducibility is presented in Table 3 and Fig. 5 for tangential treatment and in Table 4 and Fig. 6

| DISCUSSION
This study showed that using the Catalyst TM system for DIBH treatments with visual guidance significantly reduces both the mean dose to the heart and LAD and high dose volumes, for both tangential and locoregional treatment (Fig. 2). Also, the dose to the ipsilateral lung could be reduced. This may reduce the risk of long-term showing a significant reduction of the mean heart and LAD dose using DIBH. The relative reduction in the mean dose was between 38% and 65% for the heart and between 31% and 71% for LAD.
The relative reductions in the mean heart dose observed in our study (44% and 36%) were in the lower part or slightly below the range presented by Smyth et al. This could, however, be explained by the fact that the absolute mean heart dose in both FB and DIBH were lower in our study compared to all studies included in the review. For the mean LAD doses, the relative reductions observed in our study (70% and 57%) were within or slightly larger than the range presented by Smyth et al. 14 However, also the mean LAD doses presented in our study were lower than the smallest values presented by Smyth et al. 14 In a large systematic review of cardiac doses by Taylor et al., 13 it was shown that when the internal mammary node was not included in the target, the average mean dose to the heart could be reduced from 3.8 Gy in FB to 1.3 Gy using DIBH.
Also, compared to that review, our study generally demonstrated lower mean heart doses. In the review by Smyth et al., 14   F I G . 4. Cumulative probability of the intrafractional DIBH isocenter reproducibility for the tangential treatment (a) and locoregional treatment (b). The intrafractional DIBH isocenter reproducibility in lat, long, and vert directions corresponding to the cumulative probability of 50%, 90%, and maximum value are marked with blue arrows.
Comparing the results of this study and our previous study investing the benefits of enhanced inspiration gating (EIG), 15 lower relative reductions of the doses to the heart and LAD could generally be observed (except for the LAD D 2% ), which is likely because overall higher absolute doses were observed in the previous study.
The reason for this may be differences in the delineations of the structures and the creation of the treatment plans, since these tasks were carried out by different physicians and dosimetrists in the two studies. A significant reduction in the lung dose for tangential treatment was observed for DIBH in this study, which was not seen for EIG in our previous study. This was also observed by Damkjaer et al. 28 comparing DIBH and EIG, and is probably due to the higher breathing amplitude achieved using DIBH.
In a study by Chung et al., 29 32 patients underwent cardiac SPECT-CT before and after left-sided breast cancer radiotherapy, where no part of the heart was allowed inside the treatment beams.
No perfusion defects were observed, which has been seen in previous studies where parts of the heart were located inside the treatment fields. 30 This may indicate that it is the inclusion of the heart in the primary beam that is of concern. It is therefore of importance to remove the entire heart from the primary beam, shown to be possible using DIBH. In this study, the number of patients with the heart completely outside the treatment fields was increased from 4 for FB to 16 for DIBH for tangential treatment and from 0 for FB to 9 for DIBH for locoregional treatment.
In this study, we have shown that it is possible to reduce OAR doses using the Catalyst TM system for DIBH treatments with visual guidance, but when using this technique, it is of utmost importance not to introduce motion-induced uncertainties during the treatment delivery. We have, therefore, assessed and estimated the dosimetric effect of intrafractional DIBH isocenter reproducibility for the two patient groups, using real-time tracking of the isocenter position during the treatment delivery. The intrafractional DIBH isocenter reproducibility was found to be very good for the majority of the treatment sessions observed in this study, with a typical median value around 1 mm (Fig. 4). These results are in the same order as reported previously from similar studies, showing discrepancies of approximately 2 mm. 20,22 In these studies, however, the surface was used as a surrogate during DIBH, and hence, the isocenter position was not investigated. However, for a few occasional treatment sessions in this study, the intrafractional DIBH isocenter reproducibility was found to be approximately 5 mm, which resulted in large effects on the target coverage and OARs doses (Tables 3   and 4, Figs. 5 and 6). However, reduced OAR doses were maintained compared to FB in most cases, with some exceptions observed for the maximum isocenter shifts. There is also motion during FB, but this has not been taken into account in this study.
Despite only allowing beam-on within a 3 mm gating window based on the movement of the xiphoid process, larger differences in the isocenter position between two different DIBHs were observed, in either of the three translational directions, for 16 patients and 26 treatment sessions in total. This implies that the motion of the target volume differs from the xiphoid process, used to trigger the T A B L E 3 The dosimetric effect of the intrafractional DIBH isocenter reproducibility for tangential treatment. The median values and ranges for the minimum and maximum values of the various dosimetric parameters for the three cumulative probability levels: 50%, 90%, and maximum (M) are presented for the isocenter-shifted plans.  10.5%/2.1% of the treatment sessions for tangential treatments (Fig. 4).
Worse intrafractional DIBH isocenter reproducibility was observed for tangential treatment compared to locoregional treatment, probably due to the different positioning of the isocenter. For tangential treatment, the isocenter was positioned in the center of the breast, whereas for the locoregional treatment, the isocenter was positioned in the junction between the breast and the AP/PA fields. The breast is more deformable, while the junction between breast and the AP/PA fields is a more rigid structure, and therefore, the two isocenter positions move differently relative to the xiphoid process. However, the breast tissue is also a part of the PTV for the locoregional treatment. The time between the two DIBHs could potentially also be a reason for the difference in reproducibility.
However, this was found to be similar for the two patient groups as For some patients, the heart and LAD were well out of the treatment fields in the original DIBH plan. For these cases, the applied isocenter shifts did not bring the heart and LAD into the treatment fields and hence only small dosimetric effects were observed. Similarly, for some patients, a large part of the heart and LAD was already inside the treatment fields in the original DIBH plan and the applied isocenter shifts did not bring the heart and LAD out of the treatment fields. The largest dosimetric effects were observed for the patients where the treatment field edges were contiguous with the edge of the heart and LAD, since for these patients, the applied isocenter shifts would either bring the heart and LAD into or out of the treatment fields. This effect was most pronounced for D 2% , since this represents the near-maximum dose. In Figs for the median D mean,LAD , and 161% to 107% for the median D 2%, LAD (Table 4). Also, the minimum deviations in the median D 98%,PTV could be reduced from 8% to 2% and from 4% to 1% for tangential and locoregional treatment, respectively (Tables 3 and 4).
One limitation of this study is that the dosimetric effect of the intrafractional DIBH isocenter reproducibility was estimated using isocenter shifts in the TPS. When performing isocenter shifts a rigid motion is assumed, that is, the whole patient moves the F I G . 6. The minimum and maximum values of D mean,heart (a), D 2%,heart (b), D mean,LAD (c), D 2%,LAD (d), D mean,lung (e), V 20Gy,lung (f), and D 98%,PTV (g) for the isocenter-shifted DIBH plans versus the original DIBH plans. The results are presented for each individual patient receiving locoregional treatment and for all three cumulative probability levels (50%, 90%, and maximum). The lines are for illustration purpose only, and represent where the dosimetric parameters are equal for the isocenter-shifted and original DIBH plans.
corresponding isocenter shift and the distance between the heart and target volume is thus kept constant. This is actually not true since the distance between the heart and target volume changes with breathing. Using a deformable patient model would probably improve the accuracy of these calculations. But since the isocenter shifts were rather small (in the order of a few millimeter), we believe our calculations still gives a reasonable approximation of the dosimetric effect.
The analysis of this study was population based, using the cumulative probability of the intrafractional DIBH isocenter reproducibility to simulate the dosimetric effects for each patient plan.
Hence, the dosimetric effect of each patient's individual isocenter reproducibility was not simulated. The maximum intrafractional DIBH isocenter reproducibility represents the worst-case scenario for the entire population and deviations of this magnitude were only observed for a few percent of the DIBHs. The results in this study are based on the assumption that the isocenter reproducibility was the same for every DIBH, which of course is not the case.
Slightly different isocenter positions will be obtained for each DIBH throughout the treatment, resulting in a small blurring of the dose distribution. The total delivered dose to the patient would, therefore, most likely differ less from the planned dose than the result presented in this study. The 50% cumulative probability level would probably be a more realistic representation of the dosimetric effects caused by the intrafractional DIBH isocenter reproducibility for an entire treatment.
Overall, using the xiphoid process as a surrogate for the breast tissue during DIBH was found to be reproducible (Fig. 4). Gierga et al. 22 reported that 22% of the DIBHs were out of a 5 mm tolerance using the breast surface to trigger the beam when a rigid match algorithm and audio coaching were used. If using a 5 mm tolerance in either lat, long, or vert direction in this study, only 2% of the DIBH would be out of tolerance for locoregional treatments and 1% for tangential treatments. This implies that using the xiphoid process as a surrogate for the breast tissue improved the intrafractional DIBH isocenter reproducibility compared to using the breast surface.
And according to this study, the intrafractional DIBH isocenter reproducibility could be improved even further by introducing tolerances on the isocenter position.

| CONCLUSION
Deep inspiration breath-hold (DIBH) treatments for breast cancer radiotherapy, using the optical surface scanning system Catalyst TM including visual guidance, reduces the absorbed doses to the heart, LAD, and ipsilateral lung in accordance to previous studies, which may reduce the risk of long-term cardiovascular and pulmonary mortality and morbidity.
Excellent intrafractional DIBH isocenter reproducibility was observed for the majority of the treatment sessions for both tangential and locoregional treatment. However, values of the intrafractional DIBH isocenter reproducibility up to approximately 5 mm were seen for some treatment sessions, which resulted in large dosimetric effects, primarily for the OARs. Hence, it is of importance to set tolerance levels on the intrafractional isocenter motion and not only perform DIBH based on the motion of the bony anatomy of the xiphoid process.

CONFLI CT OF INTEREST
All authors approved the final manuscript, and declared that they have no potential conflicts of interest to this work.

DECLARATION OF INTEREST
Our research is partly financially supported by C-RAD AB, Uppsala, Sweden, however, the authors alone are responsible for the content, analyses, and writing.