AlignRT® and Catalyst™ in whole‐breast radiotherapy with DIBH: Is IGRT still needed?

Abstract Purpose Surface guided radiotherapy (SGRT) is reported as a feasible setup technique for whole‐breast radiotherapy in deep inspiration breath hold (DIBH), but position errors of bony structures related to deeper parts of the target are not fully known. The aim of this study was to estimate patient setup accuracy and margins obtained with two different SGRT workflows with and without daily kV‐ and/or MV‐based image guidance (IGRT). Methods A total of 50 breast cancer patients were treated in DIBH, using SGRT for the patient setup, and IGRT for isocenter corrections. The patients were treated at two different departments, one using AlignRT® (25 patients) and the other using Catalyst™ (25 patients). Inter‐fractional position errors were analyzed retrospectively in orthogonal and tangential setup images, and analyzed with and without IGRT. Results In the orthogonal kV‐kV images, the systematic residual errors of the bony structures were ≤ 3 mm in both groups with SGRT‐only. When fine‐adjusted by daily IGRT, the errors decreased to ≤ 2 mm; except for the shoulder joint. The residual errors of the ribs in tangential images were between 1 and 2 mm with both workflows. The heart planning margins were between 3 and 7 mm. Conclusions The frequency of IGRT may be considerably reduced with a well‐planned SGRT‐workflow for whole‐breast DIBH with residual errors ≤ 3 mm. This accuracy can be further improved with an IGRT scheme.

with various techniques such as voluntary BH, optical tracking of the chest wall, and spirometric methods.
Conventional patient setup relies on tattoo marks and lasers.
With such setup, isocenter variation in DIBH breast RT is up to 4.4 mm and daily MV-or kV-based image guidance (IGRT) is recommended. 3,4 However, surface guided radiation therapy (SGRT) has recently commenced to enhance or substitute the tattoo-based setup, possibly improving the setup accuracy and the reproducibility of the DIBH position. [5][6][7][8] SGRT monitors the patient′s surface optically and compares it to the planned reference surface. 5,9 AlignRT ® (VisionRT, London, Great Britain) and Sentinel/Catalyst™ (C-Rad, Uppsala, Sweden) are the current two most used systems for SGRT.
With both AlignRT ® and Catalyst™, sub-millimeter accuracy has been reported in phantom studies. 10,11 In patients, the accuracy is influenced by several additional factors, such as the fixation method, workflow at the CT, BH training, the absolute BH level (BHL), setup procedures, selection of region of interest (ROI) for the SGRT, tissue deformations such as swelling, protocols for IGRT, and tolerances for the SGRT and IGRT. Intra-fractional setup variation is small for lymph node negative (N0) breast treatments with tangential field arrangement. 12,13 However, unsuccessful patient setup is a major contributor to residual errors in DIBH. 3 The advantage of SGRT over conventional tattoo-and laser-based setup is the large surface for patient positioning compared to a few isolated marks. Furthermore, both the FB and BH surfaces are available for patient positioning.
However, the longitudinal accuracy is limited due to flatness of the body surface in this direction. During treatment, the intra-fractional movement of the patient surface and/or isocenter can be monitored in six-dimensional (6D) and radiation is allowed only when the patient is within the given thresholds. 14,15 The heart position in DIBH is reported to have daily variability up to ± 10 mm. 16 The advantage of DIBH over FB in reducing the dose to the heart depends on BHL reproducibility during treatment. 17 In breast-only (N0) RT using DIBH, tangential images can be considered as the most important estimator for the accuracy of the actual treatment delivery if tangential beams are used. 18 In previous studies, good inter-fractional accuracy has been achieved in N0 DIBH with margins of 5-6 mm. 4,13,19 However, it is known that the errors of patient posture and isocenter in lateral (LAT) and anteriorposterior (AP) directions can be underestimated in tangential images. 20 Patient rotation may increase the heart dose, arm position lower than planned may increase the dose to the shoulder joint, 4 or the heart dose and irradiated lung volume may increase if BHL is too shallow 17,21all of which are poorly visualized in tangential images.
The purpose of this retrospective study was to evaluate the inter-fractional setup accuracy of N0 SGRT-guided DIBH patients with and without IGRT to determine if daily IGRT is still needed with SGRT. The accuracy was studied in terms of isocenter error and residual errors of bony landmarks for the workflows of two RT centers using SGRT. Residual setup errors and required margins for PTV and organs at risk (OAR) with and without IGRT were calculated.  A, C). The slice thickness was 3 mm. In both groups the midline tattoo was set on the sternum, and the lateral tattoos dorsally out of the soft breast tissue. Group A used an additional tattoo 20 cm caudally from the mid tattoo to control for patient yaw. Patients in Group A were guided with the RPM™ system (Varian Medical systems, Palo Alto, USA) with a patient monitor for visual DIBH guidance. The patients in Group C were guided with the Sentinel™ (C-Rad, Uppsala, Sweden) system with wireless goggles. The gating points at the CT were two dot markers on the sternum in Group A and a virtual point on the processus xiphoideus in Group C. The gating window was 3 mm (±1.5 mm) in both groups. In Group C only a DIBH-CT was acquired, while in Group A an additional FB scan was acquired for 20 of the patients.

2.A | Patients
The patients were trained on BH technique on the CT-couch before the DIBH-imaging. The patients in Group A were advised to inhale as much as comfortably possible and hold their breath for at least 20 s. The patients in Group C were instructed for partial DIBH based on national recommendations. 22 Thus, the vertical change between FB and DIBH was not larger than 16.5 mm in Group C. 22 The CT time scheduled for the DIBH patients, including training, was 60 min at both centers.

2.C | Treatment
Patients were treated to 42.56 Gy in 16 fractions (n = 46) or 50 Gy in 25 fractions (n = 4) with field-in-field technique consisting of two or three tangential 6-MV fields on Varian TrueBeam treatment machines (Group A, C). The PTV was contoured in both units based on ESTRO guidelines. 23 Daily orthogonal kV and/or tangential MV imaging was used for all patients. The treatment started 10-12 days (Group A) or 3-6 days (Group C) after the CT.

2.D | SGRT
With both AlignRT ® and Catalyst™, the position of the patient was monitored with three cameras. The optical system compared the actual patient surface with the reference surface from the planning CT or from a reference surface acquired optically in the treatment room. Initial positioning was aligned with tattoos and thereafter finetuned with SGRT. DIBH guidance was performed with SGRT.

2.D.1 | Group A
The ROI was determined to include bony structures of the chest wall but to mostly exclude the soft tissue of the breast to diminish the effect of soft tissue swelling or deformation [ Fig. 2(a)]. However, the medial part of the soft tissue with curvature shaping was included for the aid of better optical localization of the longitudinal direction.
An additional ROI was used for arm positioning [ Fig. 2(b)]. In both FB and BH setup 1 mm/degree accuracy was aimed for, and correct (<1 mm) isocenter according to AlignRT ® was achieved with manual couch shifts. During treatment, BH surface tolerances for real time deltas (RTD) were set to ± 3 mm for translations, except for the AP direction where it was set to ± 2 mm, resulting in a 4-mm BHL gating window. The tolerance for the magnitude was set to 4 mm, and for the rotations to ±3 (yaw, roll and pitch). The initial FB surface originated from the CT (n = 20) or from the first treatment (n = 5).
The initial FB surface was replaced to achieve better patient position in cases where systematic errors were seen in the IGRT at the first three fractions. For the BH instead, the initial surface was primarily used. New BH surface was recorded only if systematic errors in the isocenter were larger than 5 mm. The vertical couch value was acquired after the first three fractions based on IGRT to standardize the location of the spine.

2.D.2 | Group C
A box-shaped ROI was used which included the whole chest wall

2.E | IGRT protocol
The matching points used for patient positioning and evaluation are presented in Fig. 3. The BHL was measured between Th6-7 and sternum [ Fig. 3(a)].

2.E.1 | Group A
In the first three fractions, both LAT and AP kV and tangential MV setup images were acquired for all patients in BH. Tangential images only were acquired from the fourth fraction onwards if tolerances were not exceeded on the first three fractions in IGRT verification.
However, at least eight additional AP + LAT images were acquired during the treatment for all patients. Thus, at least 11 AP + LAT images were acquired during the treatment course.
In the LAT images, the AP and CC directions were first matched based on vertebrae (Fig. 3). For the sternum, the location was determined using AlignRT ® irrespective of the couch VRT. The BHL was thus based on kV-based vertebrae match and optical-tracking-based sternum position, and correct BHL was verified with another LAT image after couch shifts. The couch VRT was acquired after the first three fractions, diminishing the need for two LAT images on the next fractions. LAT was matched on the middle of the ribs in the AP images.
The tolerances were 10 mm for patient yaw rotation (measured in the vertebrae between Th1 and Th10), ±5 mm for isocenter in all directions, and ±3 and ±4 mm for BHL in AP and LNG, respectively.
The action limit for the shoulder joint was 8 mm if it was shifted toward PTV and into the field.

2.E.2 | Group C
Daily AP + LAT kV setup images were acquired in BH, and the isocenter error was corrected before each fraction. Tangential images were acquired on the first three fractions only. The action limit for the displacement error was ±4 mm in BHL. Action limits for the shoulder joint were not applied. The AP direction was matched on the sternum. The CC direction was matched to a compromise between the sternum and the middle of the ribs, and the LAT direction was matched on the middle of the ribs.

2.F | Errors in isocenter and bony structures
The residual isocenter errors of SGRT were evaluated as the couch  (7) and soft tissue (8) in the tangential image (c).
deviation. 25,26 The orthogonal margins were intended for volumetric modulated arc therapy (VMAT), and the tangential skin margins were estimations for the sufficiency of skin flash in VMAT treatments.

2.G | Errors and margins for OAR
The Stroom and Heijmen's margin recipe (1.6Σ + 0.2σ) 27 was used for calculating the planning margin for the heart and shoulder joint.
The OAR margins for the heart were based on the heart shadow in portal imaging, where AP and LAT were combined. LNG was not evaluated due to uncertainties in image analysis. The OAR margins for the shoulder joint were based on the AP image.

2.H | Statistical analysis
Statistical tests were used to evaluate the differences between the workflows. The Mann-Whitney U test was used to test for equality of means. The two-tailed F-test was applied to test for equality of variances. A P ≤ 0.05 was considered statistically significant.

3.A | Errors in isocenter and bony structures
The residual errors of the SGRT-guided isocenter are presented in Table 1. The calculated PTV margins are presented in Table 2.
With SGRT-based setup, IGRT showed that 10% of the isocenter errors exceeded a 5-mm limit in the CC direction in the compromise between the ribs and sternum in both groups. The corresponding percentages for the ribs in the LAT direction were 5% and 8%; and for the sternum in the AP direction 6% and 2% in Group A and Group C, respectively. The systematic errors were below 3 mm with SGRT-only.
The residual errors reduced when switching from SGRT-only to daily IGRT (Table 3). With daily IGRT, the systematic errors were below 2 mm in both groups, but the random CC-error of the sternum in Group A was larger than 2 mm (Table 3). For both groups, the differences in the systematic and random errors of the vertebrae were not significant in the AP direction between using SGRT-only and daily IGRT. In Group A, the random residual error of the sternum did not improve in CC nor AP direction. In Group C, the systematic CC-error of the sternum did not improve. All other residual errors were significantly lower (P < 0.05) with SGRT + IGRT than with SGRT only.
Tangential images were taken as confirmation after AP + LAT images (n = 368, including both Group A and C), or as the only imaging (n = 102, Group A). In the tangential images (Table 1), the systematic residual errors in both AP/LAT and CC directions were less than 2 mm in both groups. In none of the investigated 470 tangential images the 5-mm limit was exceeded in the CC direction, nor the 4-mm limit in the AP/LAT direction. In Group A, during the 102 fractions with tangential images with no prior AP/LAT images, in only 1% the 5-and 4-mm tolerances were exceeded.

3.B | Errors and margins for OAR
The mean amplitude of the BHL was 18.4 mm (range from 10.8 to 34.2 mm) in Group A and 11.5 mm (range from 9.5 to 16.5 mm) in T A B L E 1 Errors of the isocenter, breath hold level (BHL), patient yaw rotation, and shoulder joint after patient alignment with optical surface monitoring. Image direction anterior-posterior (AP) or lateral (LAT) is in parenthesis for each structure. Patient yaw is shown as the displacement errors between Th1 and Th10. The residual errors in tangential images are reported after isocenter corrections. Errors are presented as systematic error Σ + random error σ in mm. | 101 Group C. In portal imaging, the heart distance from the field border changed from the planned 3.4 ± 7.2 mm to actual treatment values of 4.9 ± 8.1 mm in Group A; and from 7.2 ± 4.2 mm to 7.8 ± 4.5 mm in Group C. The OAR margins for the heart and shoulder joint are presented in Table 2.

| DISCUSSION
The setup accuracy in image-guided radiation therapy (IGRT) is well studied, achieving 4.3-8.4 mm margins in portal images (2D). 3,4,14,17 However, surface guidance may enhance the reproducibility of patient setup, and potentially decrease the need for image guidance. In this study, the systematic isocenter accuracy was ≤2 mm in all directions with a SGRT + IGRT workflow. The systematic isocenter errors of up to 2 mm in all three directions were acceptable with SGRT only, a clear improvement compared with up to 4 mm using conventional laser setup. 4 The clinical importance of systematic residual errors reducing from 3 mm with SGRT-only to 2 mm with the combination of SGRT and IGRT may be questionable; however, also the random errors were reduced with the addition of IGRT.
In previous RPM™ studies using tattoo-based setup, random isocenter error of up to 4.4 mm has been reported in the CC direction and daily IGRT is needed. 4 The most used landmarks for AP/ LAT matches are the ribs and sternum. 4 In the current work, SGRT with either workflow led to smaller errors in these landmarks than the tattoo-based errors. In the AP-direction, the discrepancy was small between SGRT and IGRT. With SGRT-only the 5-mm limit was exceeded in 10% of the fractions in the CC direction. The flat shape of the chest wall with limited curvature in the CC direction is challenging for accurate SGRT localization. Furthermore, the cranial breathing movement may not always reproduce similarly in the soft tissue. 28 Daily or at least weekly image guidance, based on patientspecific consideration, is therefore still needed.
Couch corrections in 6D could decrease the errors in yaw, roll, and pitch. 29 There were some patients in both groups where SGRT detected pitch, but the users were unable to correct the error in the FB setup. In Group C there was more tilt in the fixation device than in Group A. This may force the patient vertebrae towards the couch, thus better reproducing the patient setup. The yaw rotation errors in both groups were comparable with the tattoo-based setup using additional tattoo for yaw, 4 as both SGRT and tattoo setups result in 2-mm systematic and random errors. 4 As the SGRT setup was prepared with tattoo-based setup in both Group A and C, it may be concluded that SGRT did not improve yaw compared to the laser setup.
The BHL was reproducible in both groups in the AP direction, but the random AP and CC errors were larger in Group A (BHL ± 2 mm) than in Group C (BHL ± 1.5 mm). The difference (0.5 mm in either direction) was probably too small to outweigh other sources of error. The random errors are more likely to arise from unsuccessful patient setup and inter-fractional changes in the breathing technique. Also the required heart planning margins of Group A were larger than those of Group C, which is likely due to patient cheating Significant difference in random errors between the groups (Mann-Whitney P < 0.05).
AlignRT ® due to differences in the user interface of the current soft-  17 and in Group C, the errors were even smaller (AP 1.7 ± 1.4 mm, CC 1.6 ± 1.7 mm). Because the systematic residual errors in the BHL were ± 2 mm, clinically significant changes of lung volume were not likely to exist in either group.
For VMAT, PTV margins 3.7-4.9 mm have been previously published with cone beam CT images (3D) 19 The 3D margins in this study were similar: 3-6 mm with SGRT only, and 3-5 mm with SGRT+IGRT. For soft tissue deformations on the skin contour, the use of 5-mm skin flash with 8-mm optimization bolus has been suggested. 30 In this study, the soft tissue margins of 5 mm in the combined AP-LAT direction in the tangential images were within the optimization bolus thickness.
In tangential images, Schönecker et al. 31 found 1.6-mm deviation in the distance from the field edge to the ribs in portal images. Also in the present study 1-2 mm random deviation was found (Table 1). In the AP/LAT direction the systematic errors of 1.0 and 1.5 mm in Group A and Group C, respectively, were associated with respective BHL errors of 1.4 and 1.7 mm. In the CC direction the BHL errors of 2.5 and 1.6 mm in both groups were associated with only 0.6-mm errors in the tangential images, indicating that errors in pitch or changes in the breathing technique are poorly visualized in tangential images.
Imaging the LAT direction for BHL verification cannot be therefore fully omitted and this may be further emphasized when treating lymph-node positive breast cancer with higher heart and lung doses.
The risk of shoulder mobility disturbances or pain increases with increasing radiation doses to the shoulder joint. 32 The systematic position error was at the same level in both groups leading to OAR margins of up to 6 mm. In Group A, an additional arm-ROI was manually changed on the monitor during patient setup because simultaneous following of two ROIs was not possible with AlignRT ® . The

| CONCLUSION
After imaging at the first fractions, Surface Guided Radiation Therapy (SGRT) without IGRT is feasible for DIBH treatments of the breast with systematic isocenter accuracy within 3 mm. If the SGRT workflow is combined with daily imaging the accuracy can be decreased to 2 mm, and random errors can be reduced. The BHL was accurate within 2 mm, however heart planning margin of up to 3-7 mm may be needed due to errors in pitch and CC movement.

ACKNOWLEDG MENTS
This study was supported by Futurum -The Academy of Health and Care, Region Jönköping County, Sweden.

CONFLI CT OF INTEREST
The authors declare none.