Clinical evaluation of interfractional variations for whole breast radiotherapy using 3-dimensional surface imaging

Transcription

1 Practical Radiation Oncology (2013) 3, Original Report Clinical evaluation of interfractional variations for whole breast radiotherapy using 3-dimensional surface imaging Amish P. Shah PhD, Tomas Dvorak MD, Michael S. Curry MS, Daniel J. Buchholz MD, Sanford L. Meeks PhD Department of Radiation Oncology, MD Anderson Cancer Center Orlando, Orlando, Florida Received 30 November 2011; revised 1 March 2012; accepted 7 March 2012 Abstract Purpose: To evaluate the impact of 3-dimensional (3D) surface imaging on daily patient setup for breast radiotherapy. Materials and Methods: Fifty patients undergoing treatment for whole breast radiotherapy were setup daily using an AlignRT system (VisionRT, London, UK) for 3D surface-based alignment. Daily alignments were performed against a reference surface topogram and shifts from skin marks were recorded daily. This investigation evaluated the following: (1) the performance of the surfacebased imaging system for daily breast alignment; (2) the absolute displacements between setup with skin marks and setup with the surface-based imaging system; and (3) the dosimetric effect of daily alignments with skin marks versus surface-based alignments. Results: Displacements from 1258 treatment fractions were analyzed. Sixty percent of those fractions (749) were reviewed against MV portal imaging in order to assess the performance of the AlignRT system. Daily setup errors were given as absolute displacements, comparing setup marks against shifts determined using the surface-based imaging system. Averaged over all patients, the mean displacements were 4.1 ± 2.6 mm, 2.7 ± 1.4 mm, and 2.6 ± 1.2 mm in the anteroposterior (AP), superoinferior (S/I), and left-right (L/R) directions, respectively. Furthermore, the standard deviation of the random error (σ) was 3.2 mm, 2.2 mm, and 2.2 mm in the A/P, S/I, and L/R directions, respectively. Conclusions: Daily alignment with 3D surface imaging was found to be valuable for reducing setup errors when comparing with patient alignment from skin marks. The result of the surfacebased alignments specifically showed that alignment with skin marks was noticeably poor in the anteroposterior directions. The overall dosimetric effect of the interfractional variations was small, but these variations showed a potential for increased dose deposition to both the heart and lung tissues. Although these interfractional variations would not negatively affect the quality of patient care for whole breast radiotherapy, it may require an increase in PTV margin, especially in cases of partial breast irradiation American Society for Radiation Oncology. Published by Elsevier Inc. All rights reserved. Sources of support: This work was supported in part by a grant from Women Playing for T.I.M.E. (Technology, Immediate Diagnosis, Mammography and Education). Conflicts of interest: None. Corresponding author. Department of Radiation Oncology, MD Anderson Cancer Center Orlando, 1400 S. Orange Ave, Orlando, FL address: Amish.Shah@orlandohealth.com (A.P. Shah) /$ see front matter 2013 American Society for Radiation Oncology. Published by Elsevier Inc. All rights reserved.

2 Practical Radiation Oncology: January-March 2013 Introduction With an increasing number of studies comparing standard fractionation versus hypofractionated whole breast (WBRT) radiotherapy techniques, 1-3 increasing focus has been placed on image-guided radiation therapy (IGRT) to account for interfraction motion in patients treated for breast cancer. In comparison with the numerous IGRT techniques used in partial breast irradiation (PBI), the historical standard for patient setup with WBRT has been alignment with lasers and skin marks and weekly electronic portal image verifications. 4,5 Our study aims to evaluate whether standard-fractionated WBRT could be improved with better alignment techniques or are our current conventional alignment techniques sufficient for whole breast radiotherapy. With respect to standardfractionated WBRT, treatment of the target volume is less sensitive to small errors in patient alignment in comparison with PBI, where the target volume is limited to the lumpectomy resection site plus an appropriate margin. 1,6 However, with WBRT, alignment errors still should be reduced to avoid toxicity from excessive irradiation primarily to normal lung and cardiac tissues. 2 The need for greater accuracy with WBRT and PBI has contributed to technological advances in reducing interfractional variations with patient setup for breast radiotherapy. Numerous investigations have reported on various methods for 3D surface imaging in radiation therapy This investigation focuses on one of those IGRT advancements, 3D surface-based imaging with the AlignRT system (VisionRT, London, UK). The details of this system have been previously described in greater Interfractional variations during WBRT 17 depths. 6, More specifically, several publications have reported on the use of the AlignRT system for partial breast irradiation. 6,9,12,14,15 Bert et al 12 conducted one of the early investigations on the clinical performance of the AlignRT system for 9 PBI patients over 53 fractions. Our investigation is similar to the study by Bert et al in that we evaluate the patient alignment procedure with surface imaging against lasers and skin marks. Gierga et al 6 further quantified the performance of 3D surface imaging for PBI through evaluations of target registration errors on 12 patients using implanted breast clips as their ground truth for patient alignment. More recently, Chang et al 14 evaluated surface alignments on 23 PBI patients using kv X-rays and the centroid of implanted clips as their reference standard. In order to measure the interfraction motion and understand the dosimetric effects on our conventional technique for WBRT patient alignment at our institution, this investigation reviewed alignments with WBRT patients using 3D surface imaging versus our current technique of alignment with lasers and skin marks and megavoltage (MV) portal imaging. We examined daily impact of 3D surface-based imaging by using clinician review of MV portal imaging after surface-based alignments as our ground truth for 1258 treatment fractions. The aim of this investigation was to evaluate the following: (1) the daily clinical performance of the AlignRT system for breast alignment; (2) the absolute displacements between setup with lasers and skin marks and setup with the 3D surface-based imaging system; and (3) the dosimetric effect of daily alignments with skin marks versus 3D surface-based alignments for WBRT patients. Figure 1 Examples of AlignRT reference and verification surface images: (A) reference (pink) and verification surface (green) daily registration and (B) the registered region of interest within 2 mm of the reference surface after registration.

3 18 A.P. Shah et al Practical Radiation Oncology: January-March 2013 Materials and methods Protocol Fifty WBRT patients were treated using our institutional standard for external beam radiotherapy. An institutional review board-approved protocol was designed to retrospectively compare interfractional variations between patient setup from skin marks and lasers with weekly portal image verification, and 3D surface-based alignment methods. Daily patient setup was performed by aligning skin marks with room lasers, and position verification was performed using electronic portal imaging. During the initial visit, patients were set up to tattoos and shifted accordingly to the treatment isocenter. Verification portal images were taken to ensure accurate alignment to the treatment isocenter and new daily setup skin marks were drawn on the patient. Immediately following setup to the treatment isocenter, the AlignRT system was used to evaluate the external body contour from the treatment planning computed tomography (CT) to the actual patient setup and capture a new reference image for daily surfacebased alignments. A real-time evaluation of the external surface from the CT contour was used to assess arm and chin placement before a new AlignRT reference topogram was taken. Following capture of a reference surface, the region of interest for future daily alignments was contoured for each patient. The region of interest was defined as the surface region around the involved breast, with the pendulous portion of the involved breast removed on the surface topogram (Fig 1). In order to provide ample time to ensure setup accuracy by the therapists, the portal image verifications and AlignRT reference images were simulated during the first visit (film day, 30-minute time slot), which was prior to first treatment fraction (patient's second visit). During each subsequent visit (10-minute time slot), initial positioning was performed using lasers and skin marks; the target isocenter was then aligned to the machine isocenter using the AlignRT system. The therapists were instructed to record the calculated shifts from the skin marks based on the AlignRT system and notify the physicist and the physician if these shifts were greater than 1 cm in any one direction. While shifts were calculated in all 6 degrees of freedom, the therapists were instructed to evaluate patient pitch, roll, and yaw, but only perform translational couch shifts given by the AlignRT system. If Figure 2 Schematic showing the sequence of events for patient setup for whole breast radiotherapy when using the AlignRT system. MV, megavoltage; ROI, regions of interest; RT, radiotherapy.

4 Practical Radiation Oncology: January-March 2013 on evaluation of the rotational corrections the therapists noticed large rotations (N2 degrees), they were instructed to re-setup the patient before proceeding with another surface verification image and the resultant translational shifts. Following couch shifts, single exposure portal image verification was then performed to evaluate the surface-based alignment. Therapists were then instructed to perform shifts (if greater than 3 mm in any 1 direction) based on portal imaging and document these shifts for physicist review. Thus, through intercomparison of patient setup using the surface-based imaging system to the MV portal imaging system, the setup accuracy of the system was determined to be at worst within 3 mm. This investigation decided on the arbitrary 3-mm action level for the therapists based on a previous study by Gierga et al, 6 in which they reported the target registration errors for gated surface-based imaging as approximately 3 mm. The sequence of events for patient setup for whole breast radiotherapy, when using the AlignRT system, is outlined in Fig 2. Portal image verification (following surfacebased alignment) was performed as part of our quality management program for the first 5 fractions for every patient and twice weekly for the remaining fractions. It is also important to note that both the daily surface image as well as the portal image verifications were taken when the patient's breathing was approximately at end-exhalation based on visual determination by the therapists. This was done to improve reproducibility in patient alignment. Data analysis The surface-based alignments were compared with patient skin-marks alignment for 1258 fractions, and verified with portal imaging for 60% of those fractions. We calculated the mean and standard deviation for each direction, anteroposterior (A/P), lateral (left-right [L/R]), and superoinferior (S/I), including the nature (sign) of the displacement as well as the absolute values for the AlignRT alignment shifts. To evaluate the dosimetric consequence of the observed interfraction motion we performed a dose recalculation using the daily offsets from skin-mark alignment to the surface-based alignment for each patient. Dose-volume histograms (DVHs) were calculated for the delivered dose to the heart and lung tissues as well as the target volume. The target volume was defined as the volume of the prescription isodose line from the original treatment plan. To perform this retrospective dosimetric analysis, we shifted the isocenter in the original treatment plan based on the daily reported offsets and recomputed the dose in the treatment planning system for each fraction. Subsequently, we summed the delivered dose across all fractions in the MIM Maestro 5.2 software package (MIM Software Inc, Cleveland, OH) and compared it with the planned dose for each patient. The CT anatomy was treated as a rigid body, meaning that our dose recalculation technique assumed that the relationship between the whole breast target volume and the patient did not deform throughout the course of treatment. Last, the comparison of DVHs was only done to report the dosimetric effect had the delivered treatment been based solely on alignment to skin marks. To further evaluate the average performance for the left-sided WBRT patients, several dosimetric indicators, such as D95 and D30 for the breast volume, as well as mean dose for the heart, were reported. Finally, we calculated clinical target volume (CTV) margins by distinguishing between random and systematic error with the measured displacements. 16 These planning target volume (PTV) margins were calculated with the goal to include all the absolute shifts from skin marks as if there were no portal image verification done. This was done to gain a sense of the required margins if treatment was solely based on skin marks and trusting that the surface matching system was the ground truth. The PTV margins necessary to account for these errors were determined using the formula by Stroom et al 17,M=2 Σ σ, to ensure that 95% of the dose is delivered to 99% of the CTV. 16 Results Interfractional variations during WBRT 19 3D displacements: Comparing setup marks versus surface alignment Across our patient cohort, offsets from all 50 patients demonstrate some level of daily systematic and random shifts. Figure 3 shows the initial offsets for 2 patients over a full course of whole breast radiotherapy using the surface topography system versus daily laser and skin mark patient alignment. Figure 3A shows an example of only moderate daily offsets for one patient, while Fig 3B shows extensive daily offsets for another patient. Figure 4 shows histograms representing the difference between conventional patient positioning and the surface-based alignments for all 50 patients and 1258 fractions. More than 50% of the initial displacements were greater than 6 mm in 3D vector length, with more than 85% greater than 3 mm in 3D vector length (Fig 4D). The histograms in Fig 4 show a possible systematic component for all shifts for all patients. However, there is also significant variation from average positions with large ranges in the magnitudes of the shifts as evidenced by the spread of the histograms. For instance, Fig 4 displays the tendency of initial offsets to be greater in the A/P direction in comparison with either the L/R or S/I direction. More than 27% of the initial offsets were greater than 6 mm in the A/P direction; however, in the other 2 directions (L/R and S/I) only approximately 11% of the initial offsets were greater than 6 mm (Fig 4A).

5 20 A.P. Shah et al Practical Radiation Oncology: January-March 2013 A 15 Initial Isocenter Target Offset Ant/Post (mm) Sup/INF (mm) Left/Right (mm) 3D Vector 10 Isocenter Offset (mm) B Fraction # Initial Isocenter Target Offset Ant/Post (mm) Sup/INF (mm) Left/Right (mm) 3D Vector 10 Isocenter Offset (mm) Fraction # Figure 3 Initial target offsets for each fraction: left/right ( ), sup/inf (+), ant/post ( ), and 3-dimensional (3D) vector length ( ) for (A) 1 patient (No. 7) with moderate AlignRT offsets from skin marks and (B) 1 patient (No. 32) with excessive AlignRT offsets from skin marks. Table 1 gives the mean setup errors relative to skin marks and lasers for all 50 patients. Averaged over all patients, the mean absolute displacements from skin marks were 4.1 ± 2.6 mm, 2.7 ± 1.4 mm, and 2.6 ± 1.2 mm in the A/P, S/I, and L/R directions, respectively. The mean displacements (accounting for the direction of the displacements for all alignments) were 0.1 ± 4.7 mm, 0.2 ± 2.6 mm, and 0.6 ± 2.6 mm in the A/P, S/I, and L/R directions, respectively. Furthermore, Table 2 provides the systematic error and random error for all patients aligned with 3D surface alignment following patient setup with laser and skin marks. The standard deviation of the random error (σ) was 3.2 mm, 2.2 mm, and 2.2 mm in the A/P, S/I, and L/R directions, respectively. Performance of 3D surface alignments versus verification portal imaging A total of 749 comparisons were made between surface-based alignments and MV portal images throughout this study (average per patient: 15 ± 3 MV ports). During 104 of 749 fractions (14%), the treatment couch had to be shifted following initial 3D surface alignment and the subsequent portal verification film. These shifts

6 Practical Radiation Oncology: January-March 2013 Interfractional variations during WBRT 21 Figure 4 Histograms showing percent of fractions binned by size of initial target offsets for all patients and all fractions: (A) ant/post; (B) sup/inf; (C) left/right; (D) 3-dimensional (3D) vector length. were due to disagreement with patient setup during portal image verification that exceeded our 3-mm threshold used by the therapists. Furthermore, approximately 37% of these shifts occurred within the first 5 days of treatment (38 of 104 shifts). Dosimetric impact of laser-skin mark alignments Figure 5 displays the comparison of DVHs for planned versus delivered dose for only 3 patients with left-sided breast cancer. Although we recalculated dose for all patients, we found that the average patient did not display a large variation in the DVH from his or her daily shifts. For left-sided WBRT patients (n = 26), the average D95 and D30 for the volume of prescription isodose line were both approximately 0.5% lower when setup was solely based on skin marks. However, the average mean dose to the heart increased by 10.6% (1.63 ± 0.75 Gy vs 1.80 ± 0.67 Gy) when alignment was based solely on skin marks. The small difference in the average performance could be due to the fact that the average of all alignments, including the sign of the displacement, clearly showed the A/P, S/I, and L/R directions being approximately 0. For example, 2 different patients' corresponding daily offsets are displayed in Fig 3. These 2 patients are drastically different in their daily Table 1 Mean setup errors detected by the surface-based imaging system relative to alignment based on skin marks and lasers for all patients (n = 50) No. of treatment fractions Vertical (mm) Longitudinal (mm) Lateral (mm) Average displacement (including nature of displacements) Average displacement (absolute value) Maximum individual average (absolute value) Minimum individual average (absolute value) Average values given as mean of individual means. Maximum and minimum values given as individual means of each patient. 3D, 3-dimensional. 3D vector (mm)

7 22 A.P. Shah et al Practical Radiation Oncology: January-March 2013 Table 2 Systematic error, random error, and planning target volume margins for all patients when alignment was solely based on skin marks Vertical (mm) Longitudinal (mm) Lateral (mm) 3D vector (mm) Systematic Random PTV margin Values determined based on 3-dimensional surface alignment as the true reference standard. The standard deviation of errors is given. PTV, planning target volume; 3D, 3-dimensional. offsets. Yet, Fig 5A shows that the cumulative DVHs (for the patient with only moderate daily offsets) are similar to the example comparison of DVHs in Fig 5B, for a patient (Fig 3B) with extensive daily offsets. Figure 5C displays a worst-case scenario in which the recorded daily offsets placed the isocenter of the tangential beams in closer proximity to the heart and lung tissues while extensively undercovering the PTV. The DVHs in Fig 5C were calculated from the offsets recorded from a right-sided breast cancer patient, while the treatment plan was from a second patient with left-sided breast cancer, in order to show the hypothetical worst-case scenario for a specific patient (not the average). Additionally, only patients with left-sided breast cancer are shown because of the increased risk of cardiac tissue irradiation with the use of tangential beams for left-sided WBRT. Determination of PTV margins with conventional setups Table 2 gives the standard deviation of errors and the calculated PTV margins for all patients when aligned to skin marks and lasers, and determined using 3D surface alignment. In the case where PTV margin would need to be considered using a traditional setup, such as in PBI, the margin should be 7.3 mm in the A/P direction and approximately 4.0 mm in the S/I and L/R directions. Discussion The dosimetric impact from using IGRT to reduce interfraction motion affecting whole breast radiotherapy has not been well investigated. In this WBRT study, our observed interfraction variations of 0.5 to 1.0 cm are, to some extent, tolerable. However, setup errors of this order can potentially lead to significant reductions in clinical target volume coverage as well as increases in tolerance doses to organs at risk. Quantitative evaluation The initial offsets from skin marks showed an approximately 4 mm mean absolute displacement in the A/P direction. However, the average of all alignments including the sign of the displacement clearly showed a canceling out of the absolute 4-mm offset result, with the average displacement in the A/P, S/I, and L/R directions being approximately 0. This finding gives us confidence in our conventional setup process without the AlignRT system in place. However, the individual absolute average of 4 mm and individual displacements as large as 15 mm may be a cause for concern, especially in terms of dosimetric consequence for breast coverage and cardiac or lung toxicities. Our surfacebased alignments specifically showed that with each patient, skin-mark alignments were noticeably poorer in the A/P direction in comparison with the S/I and L/R directions (Fig 4). With whole breast irradiation, increased A/P precision in treatment delivery potentially minimizes dose to normal lung and cardiac tissues. This result indicates that surface-based alignment is beneficial for WBRT. Figure 5 shows that although we did not see a major impact on the lung and heart DVHs for 2 types of our left-sided breast patients, the potential for increases in heart dose may be real (Fig 5C). For patients planned using the conventional WBRT approach, the observed accuracy within 10 mm with conventional patient setup may be sufficient due to the entire breast volume being irradiated. 18 However, for patients planned using CTV and PTV volumes, the clinical implication for conventional positioning with skin marks would require a CTV-to-PTV margin of at least 7 mm, excluding additional margin for respiratory motion. This could be considerably reduced with the use of any type of available IGRT techniques. We find that the surface-based imaging system would provide an accurate method to reduce the PTV margins, in order to reduce the irradiated volume of normal tissue. System performance Portal-image verifications were taken during 60% of the total alignments in order to establish the validity of the daily surface-based alignments to within 3 mm. Agreement between the digitally reconstructed radiograph and the portal image, following alignment, provided us with a means of evaluating the surface-based alignments with a

8 Practical Radiation Oncology: January-March 2013 Interfractional variations during WBRT 23 Figure 5 Planned and delivered dose-volume histograms of heart, lung, and volume of prescription isodose line from the treatment plan for (A) patient (No.7) with moderate AlignRT offsets from skin marks; (B) patient (No. 32) with excessive AlignRT offsets from skin marks (data displayed in Fig 3); and (C) left-sided breast cancer patient's treatment (Tx) plan, combined with a separate patient's daily offsets in order to display a possible worst-case scenario from daily laser and skin mark alignments with large systematic error.

9 24 A.P. Shah et al Practical Radiation Oncology: January-March 2013 ground truth. Approximately 14% of portal images were rejected following review by the therapists or clinician, and the treatment couch had to be shifted before treatment delivery. Often these shifts occurred within the first 5 days of treatment. Potential physiological changes (ie, slower respiration rate and patient relaxation) in patients may occur between the time the reference surface topogram was taken and their daily fractions, as they continue forward with treatment. We find that repeat portal imaging is still necessary within the first 5 days of treatment in order to verify accuracy of the initial reference surface topogram. It has been reported that the use of MV portal images as ground truth for patient alignment may not be the best method for evaluating the surface-imaging system. 6 These investigators found that the most accurate method for patient alignment was kv imaging of implanted clips. However, with WBRT, implanted clips or fiducials were unnecessary for this evaluation of the surface-based imaging system. Furthermore, clinically we found no issue with utilizing the MV portal image to verify the surface-based alignments to within 3 mm as this is currently our standard in evaluating the daily skin-mark alignments in our institution. Qualitative evaluation Improved patient setup procedures are not only valuable for reducing normal tissue toxicity but also for reducing larger scale misalignments that occur with setups based solely on skin marks. Our institution's current standard of care for WBRT relies on patient alignment, which is only verified once every 5 treatment fractions via MV portal imaging. Clinicians have to rely on radiation therapists to properly align patients on a daily basis. A single rejection of a portal image for 1week could imply that the patient was not properly aligned for the last 4 days the patient was treated. Further, the potential for even larger alignment errors are always a possibility without the availability of additional image-guidance to aid in patient setup. In our investigation we found that the therapists were more cognizant of patient alignment to skin marks, with the knowledge that the surface-based imaging system was following their initial alignments with possible new shift information. Arm placement and chin position were heavily scrutinized by the therapists on a daily basis for each patient with the AlignRT system. Surface-based imaging systems provide visual feedback and easily allow therapists to evaluate arm position and chin placement, unlike conventional WBRT setups where evaluation must be done on visual inspection. We found that arm placement was critical in patient alignment even with the patient on an immobilization device with arm support. 12 A patient with a misaligned arm can still be aligned to his or her skin marks, which could result in large interfraction variation from the intended treatment position. Conclusions Daily alignment with 3D surface imaging was found to be valuable for reducing setup errors when compared to patient alignment using skin marks and weekly MV portal imaging. Surface-based alignments specifically showed that alignment with skin marks were noticeably poor in the A/P direction in comparison with the S/I and L/R directions. Considerable random and systematic displacements were found from daily setup on treatment marks alone. While the dosimetric effects of these interfractional variations have been found to be small, we have shown that there is a real possibility for undercoverage of the target volume as well as large increases in dose to the organs at risk in whole breast radiotherapy. References 1. Vicini FA, Kestin LL, Goldstein NS. Defining the clinical target volume for patients with early-stage breast cancer treated with lumpectomy and accelerated partial breast irradiation: a pathologic analysis. Int J Radiat Oncol Biol Phys. 2004;60: Njeh CF, Saunders MW, Langton CM. Accelerated partial breast irradiation using external beam conformal radiation therapy: a review. Crit Rev Oncol Hematol. 2012;81: Lehman M, Hickey B. The less than whole breast radiotherapy approach. Breast. 2010;19: Lirette A, Pouliot J, Aubin M, Larochelle M. The role of electronic portal imaging in tangential breast irradiation: a prospective study. Radiother Oncol. 1995;37: Fein DA, McGee KP, Schultheiss TE, Fowble BL, Hanks GE. Intraand interfractional reproducibility of tangential breast fields: a prospective on-line portal imaging study. Int J Radiat Oncol Biol Phys. 1996;34: Gierga DP, Riboldi M, Turcotte JC, et al. Comparison of target registration errors for multiple image-guided techniques in accelerated partial breast irradiation. Int J Radiat Oncol Biol Phys. 2008;70: Djajaputra D, Li S. Real-time 3D surface-image-guided beam setup in radiotherapy of breast cancer. Med Phys. 2005;32: Liu D, Li S. Accurate calibration of a stereo-vision system in imageguided radiotherapy. Med Phys. 2006;33: Riboldi M, Gierga DP, Chen GT, Baroni G. Accuracy in breast shape alignment with 3D surface fitting algorithms. Med Phys. 2009;36: Spadea MF, Baroni G, Riboldi M, et al. Patient set-up verification by infrared optical localization and body surface sensing in breast radiation therapy. Radiother Oncol. 2006;79: Bert C, Metheany KG, Doppke K, Chen GT. A phantom evaluation of a stereo-vision surface imaging system for radiotherapy patient setup. Med Phys. 2005;32: Bert C, Metheany KG, Doppke KP, Taghian AG, Powell SN, Chen GT. Clinical experience with a 3D surface patient setup system for alignment of partial-breast irradiation patients. Int J Radiat Oncol Biol Phys. 2006;64:

10 Practical Radiation Oncology: January-March Wooten HO, Klein EE, Gokhroo G, Santanam L. A monthly quality assurance procedure for 3D surface imaging. J Appl Clin Med Phys. 2010;12: Chang AJ, Zhao H, Wahab SH, et al. Video surface image guidance for external beam partial breast irradiation. Practical Radiat Oncol. 2012;2: Hasan Y, Kim L, Martinez A, Vicini F, Yan D. Image guidance in external beam accelerated partial breast irradiation: comparison of surrogates for the lumpectomy cavity. Int J Radiat Oncol Biol Phys. 2008;70: Interfractional variations during WBRT van Herk M. Errors and margins in radiotherapy. Semin Radiat Oncol. 2004;14: Stroom JC, de Boer HC, Huizenga H, Visser AG. Inclusion of geometrical uncertainties in radiotherapy treatment planning by means of coverage probability. Int J Radiat Oncol Biol Phys. 1999;43: Morrow NV, Stepaniak C, White J, Wilson JF, Li XA. Intra- and interfractional variations for prone breast irradiation: an indication for image-guided radiotherapy. Int J Radiat Oncol Biol Phys. 2007;69: