Medical Dosimetry 39 (2014) 74–78
Medical Dosimetry journal homepage: www.meddos.org
Evaluation of overall setup accuracy and adequate setup margins in pelvic image-guided radiotherapy: Comparison of the male and female patients Marko Laaksomaa, R.T.T.,* Mika Kapanen, Ph.D.,*† Tapio Tulijoki, M.D,* Seppo Peltola, Ph. Lic.,*† Simo Hyödynmaa, Ph.D.,*† and Pirkko-Liisa Kellokumpu-Lehtinen, M.D., Ph.D.* *
Department of Oncology, Tampere University Hospital (TAUH), Tampere, Finland; and †Department of Medical Physics, Tampere University Hospital (TAUH), Tampere, Finland
A R T I C L E I N F O
A B S T R A C T
Article history: Received 15 March 2013 Accepted 17 September 2013
We evaluated adequate setup margins for the radiotherapy (RT) of pelvic tumors based on overall position errors of bony landmarks. We also estimated the difference in setup accuracy between the male and female patients. Finally, we compared the patient rotation for 2 immobilization devices. The study cohort included consecutive 64 male and 64 female patients. Altogether, 1794 orthogonal setup images were analyzed. Observer-related deviation in image matching and the effect of patient rotation were explicitly determined. Overall systematic and random errors were calculated in 3 orthogonal directions. Anisotropic setup margins were evaluated based on residual errors after weekly image guidance. The van Herk formula was used to calculate the margins. Overall, 100 patients were immobilized with a housemade device. The patient rotation was compared against 28 patients immobilized with CIVCO's Kneefix and Feetfix. We found that the usually applied isotropic setup margin of 8 mm covered all the uncertainties related to patient setup for most RT treatments of the pelvis. However, margins of even 10.3 mm were needed for the female patients with very large pelvic target volumes centered either in the symphysis or in the sacrum containing both of these structures. This was because the effect of rotation (p r 0.02) and the observer variation in image matching (p r 0.04) were significantly larger for the female patients than for the male patients. Even with daily image guidance, the required margins remained larger for the women. Patient rotations were largest about the lateral axes. The difference between the required margins was only 1 mm for the 2 immobilization devices. The largest component of overall systematic position error came from patient rotation. This emphasizes the need for rotation correction. Overall, larger position errors and setup margins were observed for the female patients with pelvic cancer than for the male patients. & 2014 American Association of Medical Dosimetrists.
Keywords: Radiotherapy Pelvis Setup errors Setup margins
Introduction The anatomy of the male and female pelvis is different. In the male pelvis, the bones and muscular volume are larger than those of the female pelvis. The female pelvic area tends to accumulate more fat. The shape of the pelvis is also different. In addition, the male skin is thicker. The different anatomic properties between the genders may require different considerations for patient setup in radiotherapy (RT). Haslam et al.1 have reported that setup accuracy is independent of patient weight, height, and age and it is not possible to estimate
Reprint requests to: Marko Laaksomaa, R.T.T., Department of Oncology, Tampere University Hospital (TAUH), PO BOX 2000 (Teiskontie 35), Tampere FI 33521, Finland. Tel.: þ358 331 169 623; fax: þ358 331 163 001. E-mail: marko.laaksomaa@pshp.fi
setup accuracy based on these factors. However, they have not investigated the difference between the setup accuracy for the men and the women. To the best of our knowledge, no comprehensive studies exist on that topic. In clinical practice, we have noticed that more image guidance has been needed to confirm the patient setup in pelvic RT for women than that needed for men. However, it is common to assume that equal setup margins can be applied because of the same image-guidance procedure and immobilization. Errors from different sources, such as observer-related errors, have not been investigated comprehensively in recent studies using kV imaging. It might be useful to know to what extent the uncertainty in patient setup is related to translation, rotation, and observerrelated factors. The purpose of this study was to estimate adequate setup margins for image-guided RT (IGRT) based on bony landmarks in
0958-3947/$ – see front matter Copyright Ó 2014 American Association of Medical Dosimetrists http://dx.doi.org/10.1016/j.meddos.2013.09.009
M. Laaksomaa et al. / Medical Dosimetry 39 (2014) 74–78
the pelvis. We evaluated whether an isotropic setup margin of 8 mm is adequate when considering weekly IGRT protocol, combination of patient rotation and deformation, and observer-related errors. We evaluated setup accuracy for both the male and female patients. We analyzed orthogonal x-ray images as they are widely used for frequent (routine) setup verification. As onboard 3dimensional (3D) verification imaging (such as cone beam computed tomography, CBCT) may be performed less frequently, setup margins should be confirmed suitable for the 2D imaging. Based on bony landmarks, 2D kV and CBCT alignments have been reported to correlate highly, but slightly different margins may be needed.2 Methods and Materials
75
of the 3 first treatment fractions and 5 mm in weekly imaging. We analyzed the acquired onboard images retrospectively in offline review according to the presented IGRT protocol. We estimated setup errors also without IGRT and with image guidance performed only in the first 3 treatment fractions to demonstrate transfer errors between the treatment-planning CT and a treatment unit.
Estimation of setup errors Reference treatment level was defined in the middle of PTV (MID-PTV). The MID-PTV point was located usually within ⫾ 1 cm from the midpoint of the pubic symphysis and the sacrum (Fig. 1). Setup errors were determined separately for the group M (n ¼ 33) and the group F (n ¼ 44) and for both the groups together. The number of the analyzed kV images was 816 and 480 for the men and women, respectively. Directions are expressed as anterior-posterior (AP or vertical), superior-inferior (SI or longitudinal), and LAT (lateral).
Patient rotation errors
Patient groups The group males (M) consists of consecutive patients with rectal (n ¼ 25) and prostate cancer (n ¼ 25) and the group females (F) of consecutive patients with gynecologic (n ¼ 27) and rectal (n ¼ 23) cancer. All the patients have a large planned target volume (PTV) because of the lymph node involvement as shown in Fig. 1. The average age for the group F and M was 68 and 71 years, respectively. Both groups were immobilized with a knee support that has been made in our department several years ago (device 1). The feet are tied with a stasis. We compared our fixation device to a commercial one used in our satellite unit in Lahti. This device is the combination of CIVCO's Kneefix and Feetfix (device 2), where both knees and feet are fixed into a supporting cushion. Both the devices are presented in Fig. 2. The Lahti group consists of 14 male and 14 female patients (18 patients with rectal cancer, 8 patients with gynecologic cancer, and 2 patients with urinary bladder cancer). Computed tomography (CT) imaging for treatment planning was done at 120 kVp with either Philips Brilliance Big Bore (Philips Medical Systems, Eindhoven, the Netherlands) or Toshiba Aquilion LB (Toshiba Medical System, Tokyo, Japan) using a slice thickness of 3 mm. The patients were treated with the intensity-modulated radiation therapy technique using 6- and 18-MV photon beams of Clinac 2300 iX (Varian Medical Systems, Palo Alto, CA). Image guidance was carried out using orthogonal kV images acquired with an onboard imaging system at 75 kV with 10 to 16 mAs for the anterior images and at 105 to 120 kV with 80 to 126 mAs for the lateral images.
The effect of pelvis rotation was investigated with 50 male and 50 female patients with total number of images being 652 and 600, respectively. This was done by determining systematic and random variations in the distance between the pubic symphysis and the sacrum as seen in Fig. 1, as they are the clearest landmarks and usually placed at the extreme edges of the pelvic targets. Also, deformation of bony structures may contribute to these results. However, the pelvis is a quite rigid object and the rotations play the most important role for these results. The extent of rotation was double checked by rematching the images with a time gap of 2 months.
Estimation of observer errors and comparison of the 2 fixation devices A group of 20 experienced radiation therapists evaluated the images and their image matches were compared with the reference MID-PTV match. Systematic (∑) and random (s) errors were determined for the differences. Potential benefit of device 2 in the reduction of the rotation errors was investigated based on data obtained from our satellite clinic, comprehending 28 patients with 378 images. Unfortunately, no larger data were available consistent with our IGRT protocol. Therefore, the devices were compared only for combined groups of the male and female patients.
Estimation of overall errors and treatment margins Investigated image-guidance protocols Our image-guidance protocol used an online correction with a fixed 5-mm action level for translational couch (patient setup) corrections in all 3 orthogonal directions. Imaging was performed in the first 3 treatment fractions and weekly thereafter. If the action level was exceeded, the imaging was repeated in the next fraction. An action level of zero was applied for couch vertical based on the average
The residual setup error of the treatment isocenter after weekly imaging, the rotation of the sacrum and the pubic symphysis, and the observer error were combined to obtain the overall setup errors. All the error components were added in quadrature.3 Systematic (∑) and random (s) components were handled separately. The total errors were used to calculate anisotropic setup margins using the van Herk formula m ¼ 2.5Σ þ 0.7s.3
Fig. 1. Regions of interest in the sacrum and the pubic symphysis (white boxes) used to estimate the pelvis rotation. Typical PTV covers both of these structures in (A) anterior and (B) lateral reference images. The translational position errors caused by the rotation about anterior axis (yaw) and lateral axis (pitch) were determined from images (A) and (B), respectively. (Color version of figure is available online.)
76
M. Laaksomaa et al. / Medical Dosimetry 39 (2014) 74–78
Fig. 2. (A) The house-made fixation device 1. Patient has a knee support and the feet are tied with a stasis. The device is not fixed to the treatment couch. (B) The device 2 is the combination of CIVCO's Kneefix and Feetfix, where both the knees and feet are fixed by supporting cushions attached to the couch. (Color version of figure is available online.) Statistical testing The F test was used to measure significance of the differences in systematic errors between the groups (test for equality of variance). The Wilcoxon rank sum test was used for the random errors (test for equality of means). A p r 0.05 was considered statistically significant.
Results Estimation of setup errors Overall setup errors by excluding the patient rotation are given in Table 1. Without the image guidance, an adequate margin in vertical direction was larger than 1 cm in both the groups caused by different bending of the CT and treatment unit couches. Mean overall error was 4 mm anteriorly. This means that it is important to correct vertical direction without tolerance because transfer errors between CT and treatment unit are large, but are often slightly below the tolerance of 5 mm. In the group F, mean overall error was 1.5 mm to the left. When position corrections were done based only on the images of the first 3 fractions, the average overall errors were near zero in both the groups. In the group F, relatively larger margins are needed in longitudinal and lateral directions. The excess of the tolerance limit of 5 mm in the group F was distributed as 11%, 19%, and 24% in AP, SI, and LAT directions, respectively. The corresponding values for the group M were 6%, 11%, and 10%. The low percentage for the vertical direction is because of zero action level applied based on the first 3 images.
With the weekly image guidance, the margins can be approximately reduced to half when compared with no image guidance at all. For the female group, it was impossible to make adequate setup corrections based on only the first 3 fractions and more frequent image guidance was needed. For the male group, the need of weekly imaging was smaller and the required margins were below 6 mm after correction based on the first 3 images. In the group M, the percentage of image-guided fractions was 37% for the patients with prostate cancer and 45% for the patients with rectal cancer. In the group F, both the patients with rectal cancer and the patients with gynecologic cancer, were imaged in 60% of the fractions. We noticed a time trend for setup errors in the AP direction in 10% of the investigated patients. For these patients, couch vertical was gradually raised during the treatment. Group average of overall errors was near zero, suggesting that there is no global systematic shift. Rotation errors The errors related to patient rotation are given in Table 2. It can be seen that systematic position errors due to rotation are larger than those caused by variation of isocenter position after weekly IGRT, given in Table 1. All the differences between the F and M groups are significant in all directions. The errors caused by the rotation of the pelvis were relatively larger in the group F than in the group M. Systematic rotation errors exceeding the tolerance of 5 mm were 4% in the group M and 6% in the group F (in 1 or more direction). A tighter level of 4 mm was exceeded in 8% and 22% in the M and F groups, respectively. With device 2, lateral random
Table 1 Position variation of treatment isocenter alone and the corresponding setup margins with the weekly image-guidance protocol (device 1). Errors for position correction based only on the first 3 fractions are given in parenthesis Women, n ¼ 33 AP ∑ error (mm) p-Value for ∑ s error (mm) p-Value for s Margin (mm) Fractions imaged (%)
1.2 0.72 1.9 0.52 4.3
Men, n ¼ 44 SI
(1.6) (0.05) (2.4) (0.01) (5.6)
1.3 0.35 2.1 0.01 4.8 60
LAT (2.1) (0.06) (2.8) (o 0.001) (7.2)
1.1 0.87 1.9 0.27 4.1
(2.2) (0.02) (3.3) (0.004) (7.8)
AP
SI
LAT
1.0 (1.2)
1.2 (1.5)
1.0 (1.5)
1.8 (1.8)
1.8 (1.8)
1.7 (2.0)
3.9 (4.3)
4.1 (5.1) 41
3.7 (5.1)
M. Laaksomaa et al. / Medical Dosimetry 39 (2014) 74–78
77
Table 2 Variations in the distance between the pubic symphysis and the sacrum in the MID-PTV match, and comparison of 2 different fixation devices. The results provide translational position shifts caused by the pelvis rotation
s (mm)
∑ (mm)
Women (device 1)* Men (device 1)* p-Value for the F and M groups F þ M, device 1† F þ M, device 2‡ p-Value for device 1 and 2
AP
SI
2.1 1.7 0.01 1.9 1.8 0.65
2.5 2.0 0.03 2.3 2.3 0.88
LAT
AP
1.8 1.1 o 0.001 1.5 1.7 0.16
2.1 1.4 0.02 1.9 1.7 0.11
SI
LAT
2.3 1.7 o 0.001 1.9 2.0 0.44
2.0 1.6 0.001 1.8 1.4 o 0.001
Device 1 has a knee support and stasis in feet, whereas device 2 has knee and feet support cushions. n
† ‡
n ¼ 50. n ¼ 100. n ¼ 28.
errors can be slightly reduced because of better fixation of the feet. The rotation errors were consistent within 0.4 mm in the rematch and the same parameters remained statistically significant or insignificant. Observer errors Observer errors were relatively small; still we found significant differences for the group F and M in SI and LAT directions, having an influence on the total setup margins (Table 3). Total setup errors and setup margins Total setup errors and resulting setup margins are listed in Table 4. When setup uncertainties based on bony landmarks were accounted for, the adequate overall setup margins were within 7 to 10 mm depending on target position in both the groups. The largest difference in the margins of 2.3 mm between the groups existed in the longitudinal direction. By removing the effect of rotation, the margins could be reduced approximately by 3 mm (range from 1.4 to 3.9 mm).
Discussion We found larger pelvic setup uncertainties for the female patients when compared with the male patients with the same IGRT protocol. The tattoo marks used for patient positioning were observed to shift in a different direction between these patient groups. This may be because of different skin types and amount of subcutaneous fat. The patient rotation was also larger in the female group. This may be because of different pelvic anatomy and fat content between the women and men. Our results suggest that these differences have an effect on daily setup and cause natural variation in the location of the setup marks relative to the bony structure. The setup errors obtained excluding the patient rotations are consistent with the literature studies.4,5 The random errors are higher than the systematic ones. Both error types can be
remarkably decreased by increasing the frequency of the image guidance.6 We observed shallow time trends in the vertical direction, but they were below the action level of 5 mm. This observation was consistent with recent reports.7,8 Our results indicate that the female pelvic region has larger systematic and random rotations than that of the male's. Rotation was observed both in the first treatment fraction and during the treatment course. The contribution of the patient rotation was remarkable as it required a 3-mm expansion to the margins, resulting in much greater irradiated volumes. Thus, our results suggest that couch corrections for both translation and rotation (6D) may be beneficial as has been recently suggested.4 Because the systematic errors were as wide as the random errors, it may be possible to partly correct the rotation based on offline image matching. In contrast, the random rotation about the lateral axis (causing mainly longitudinal random error) was large in the F group, which may render offline corrections inadequate. Because of this type of rotation, the bony landmarks shifted toward each other on an average by 1 mm in the anterior images for both the groups. The rotation of the pelvis about the coronal axis was smaller in both the patient groups. Even though the differences between the groups were statistically significant, they were still small. It may be difficult to remove all such small errors in practice. At least small deformation and rotation errors exist always between the pubic symphysis and the sacrum in the RT of the pelvis. Selection of the correct treatment alignment (isocentre) based on the images is not unique and there is variation between observers. Observer-related errors are usually not investigated or they are assumed to be negligible. If image matching is done only to the pubic symphysis, position error of the sacrum remains large. This explains mostly the obtained systematic observer errors. The observer errors were still relatively small in this study and Table 4 Total setup errors and adequate setup margins for 2 categories of PTV sizes estimated for the weekly image-guidance protocol based on bony landmarks (device 1). The errors include position shifts because of variation in isocenter position, pelvis rotation, and observer variation in image matching Women, n ¼ 50
Table 3 Observer-related errors in MID-PTV image matching (device 1) Women, n ¼ 25
∑ error (mm) p-Value for ∑ s error (mm) p-Value for s
∑ error (mm)*
Men, n ¼ 25
s error (mm)*
AP
SI
LAT
AP
SI
LAT
1.3 0.36 1.7 0.16
1.3 0.04 1.5 0.11
0.8 0.03 1.1 0.15
1.1
0.8
0.5
1.4
1.1
0.9
Margins (mm)* ∑ error (mm)† s error (mm)† Margins (mm)† n
†
Men, n ¼ 50
AP
SI
LAT
AP
SI
LAT
2.0 2.7 7.0 2.7 3.2 9.0
2.2 2.8 7.5 3.2 3.4 10.3
1.6 2.4 5.8 2.3 2.7 7.6
1.7 2.4 6.0 2.3 2.8 7.7
1.7 2.2 6.0 2.5 2.8 7.6
1.2 2.1 4.7 1.6 2.3 5.9
Results for PTV centered between the symphysis and the sacrum. Results for a very large PTV centered in the symphysis or the sacrum.
78
M. Laaksomaa et al. / Medical Dosimetry 39 (2014) 74–78
consistent with the literature. Lewis et al.9 have reported 1.1-mm (root-mean-square) deviation between the observers, using MV image guidance. We demonstrated that systematic observer errors are significantly different for the female and male patients in the SI and LAT directions. The error in the longitudinal direction has the largest contribution to the required margins. The results in Tables 3 and 4 implicate that larger observer errors are related to larger rotation errors. We expect that observer-related errors can be decreased by more accurate rules for image matching for large pelvic targets treated usually with intensity-modulated radiation therapy. We recommend the MID-PTV match to minimize overall setup error, unless otherwise considered by the responsible oncologist. It should be confirmed that the images cover the whole PTV area or at least the middle part of it and that the PTV structure is visible in the images. The current results were based on the bony landmarks in the 2D images. Because these images visualize bone projections, uncertainty in image matching may be larger than in tomographic 3D imaging, such as CBCT. In this sense, our results provide upper limits for the setup margins. Internal organ motion in pelvis is reported to be large. Internal motion occurs in a complex way not always correlating with the bony landmarks. Organs such as the uterine fundus may need even a 4-cm margin to properly cover its motion.10 Internal organs may also have systematic shifts during treatment period and, therefore, application of CBCT or magnetic resonance imaging during a treatment course is recommended in the literature.10-12 Internal tissue motion is out of the scope of our investigation and must be included separately to the margins. However, frequent image guidance is usually based on planar kV or MV images, supporting the feasibility of our current results. To investigate the effect of surgery, we calculated the errors merely for patients with rectal cancer before operation. There were 25 such patients in both the groups. The errors were equal within 0.3 mm with the results obtained for the whole groups, suggesting only a minor potential effect of the surgery. Despite the fact that the fraction of surgical patients is smaller in the male group, the results on the overall reflect normal clinical patient groups, supporting the feasibility of our results. Image guidance based on orthogonal 2D kV images is an effective way to control setup accuracy. It was effective enough to cover the isotropic margin of 8 mm for typical pelvic RT (such as treatment of rectum, cervix, endometrium, and prostate carcinoma with lymph node involvement) when accounting for the patient rotation and observer-related variation. The uncertainty related to the fixation device itself is not prominent because increase in margin caused by the patient rotation was small in the male group. We obtained very similar setup errors for the investigated devices consistently with the literature.13 This implies that it is not optimal to spend resources on the improvement of the immobilization device. Instead, it might be beneficial to harmonize the patient fixation and image-matching procedures. Because the commonly applied isotropic margin of 8 mm proved insufficient only for very large PTVs (centered in the symphysis or in the sacrum and encompassing both of these structures), even when daily IGRT is applied, it is appropriate to find a threshold for systematic position errors to render this margin sufficient. By accounting for the determined random errors, a threshold of 6 mm for systematic errors seems appropriate (estimated by using the van Herk formula) and should be applied for all the bony landmarks with the large target volume. The threshold could be used to screen patients needing immediate correction of patient posture or fixation. If this is not possible, the clinical importance of the large systematic position error should be evaluated, e.g., using 3D imaging. When necessary, adaptive replanning should be performed. We introduced this procedure in our clinic based on these results. We found routine daily IGRT
not necessary and we retained our weekly IGRT protocol because it suggests more frequent IGRT when tolerances are exceeded. As a conclusion, the adaptation of the threshold, the implementation of 6D couch corrections, and harmonization of the image-matching procedure seem to be the best improvements for patient setup accuracy in pelvic RT.
Conclusions We conclude that setup margins of 8 mm together with our weekly IGRT protocol are sufficient for the male patients with pelvic cancer and for the female patients with typical target volumes centered between the symphysis and the sacrum. However, for the female patients with large target volumes centered either in the symphysis or in the sacrum containing both of these structures, larger setup margins up to 10.3 mm should be applied with the weekly IGRT protocol, whereas margins up to 8.7 mm are sufficient with daily IGRT. This was because of larger patient rotation and observer variation in image matching observed for the female patients. For the female patients with such large targets, the margins of 8 mm can be used only when systematic position errors exceeding 6 mm are corrected for all the bony landmarks within the target, e.g., by correcting patient posture or fixation. If this is not possible, adaptive replanning should be considered. The results are based on matching of orthogonal 2D xray images to the bony landmarks in the middle of the target volume.
Acknowledgments This study was supported by the Pirkanmaa Hospital District, Elna Kaarina Savolainen fund (grant no. R12536). References 1. Haslam, J.J.; Lujan, A.E.; Mundt, A.J.; et al. Setup errors in patients treated with intensity-modulated whole pelvic radiation therapy for gynecological malignancies. Med. Dosim. 30:36–42; 2005. 2. Heng, L.; Zhu, X.R.; Zhang, L.; et al. Comparison of 2D radiographic images and 3D cone beam computed tomography for positioning head-and-neck radiotherapy patients. Int. J. Radiat. Oncol. Biol. Phys. 71:916–25; 2008. 3. International Commission on Radiation Units and Measurements (ICRU). ICRU Report 62. Prescribing, recording, and reporting photon beam therapy. Supplement to ICRU Report 50. ICRU, Bethesda, MD; 1999. 4. Ahmad, R.; Hoogeman, M.S.; Quint, S.; et al. Residual setup errors caused by rotation and non-rigid motion in prone-treated cervical cancer patients after online CBCT image-guidance. Radiother. Oncol. 103:322–6; 2012. 5. Stroom, J.C.; Olofsen-van Acht, M.J.J.; Quint, S. On-line set-up corrections during radiotherapy of patients with gynecological tumors. Radiat. Oncol. 6:499–506; 2000. 6. Rudat, V.; Hammoud, M.; Pillay, L.; et al. Impact of the frequency of online verifications on the patient set-up accuracy and set-up margins. Radiat. Oncol. 6:101; 2011. 7. El-Gayed, A.A.; Bel, A.; Vijlbrief, R.; et al. Time trend of patient setup deviations during pelvic irradiation using electronic portal imaging. Radiother. Oncol. 26:162–71; 1993. 8. Cranmer-Sargison, G.; Kundapur, V. Using kV-kV and CBCT imaging to evaluate rectal cancer patient position when treated prone on a newly available belly board. Med. Dosim. 37:117–21; 2012. 9. Lewis, D.G.; Ryan, K.R.; Smith, C.W. Observer variability when evaluating patient movement from electronic portal images of pelvic radiotherapy fields. Radiother. Oncol. 74:275–81; 2005. 10. Stewart, J.; Lim, K.; Kelly, V. Automated weekly replanning for intensitymodulated radiotherapy of cervix cancer. Radiat. Oncol. 78:350–8; 2010. 11. Brierley, J.D.; Dawson, L.A.; Sampson, E.; et al. Rectal motion in patients receiving preoperative radiotherapy for carcinoma of the rectum. Radiat. Oncol. 80:97–102; 2011. 12. Chan, P.; Dinniwell, R.; Haider, M.A.; et al. Inter- and intrafractional tumor and organ movement in patients with cervical cancer undergoing radiotherapy: A cinematic-MRI point-of-interest study. Int. J. Radiat. Oncol. Biol. Phys. 70: 1507–15; 2008. 13. Mitine, C.; Hoornaert, M.T.; Dutreix, A.; Beauduin, M. Radiotherapy of pelvic malignancies: Impact of two types of rigid immobilisation devices on localisation errors. Radiother. Oncol. 52:19–27; 1999.
Medical Dosimetry journal homepage: www.meddos.org
Evaluation of overall setup accuracy and adequate setup margins in pelvic image-guided radiotherapy: Comparison of the male and female patients Marko Laaksomaa, R.T.T.,* Mika Kapanen, Ph.D.,*† Tapio Tulijoki, M.D,* Seppo Peltola, Ph. Lic.,*† Simo Hyödynmaa, Ph.D.,*† and Pirkko-Liisa Kellokumpu-Lehtinen, M.D., Ph.D.* *
Department of Oncology, Tampere University Hospital (TAUH), Tampere, Finland; and †Department of Medical Physics, Tampere University Hospital (TAUH), Tampere, Finland
A R T I C L E I N F O
A B S T R A C T
Article history: Received 15 March 2013 Accepted 17 September 2013
We evaluated adequate setup margins for the radiotherapy (RT) of pelvic tumors based on overall position errors of bony landmarks. We also estimated the difference in setup accuracy between the male and female patients. Finally, we compared the patient rotation for 2 immobilization devices. The study cohort included consecutive 64 male and 64 female patients. Altogether, 1794 orthogonal setup images were analyzed. Observer-related deviation in image matching and the effect of patient rotation were explicitly determined. Overall systematic and random errors were calculated in 3 orthogonal directions. Anisotropic setup margins were evaluated based on residual errors after weekly image guidance. The van Herk formula was used to calculate the margins. Overall, 100 patients were immobilized with a housemade device. The patient rotation was compared against 28 patients immobilized with CIVCO's Kneefix and Feetfix. We found that the usually applied isotropic setup margin of 8 mm covered all the uncertainties related to patient setup for most RT treatments of the pelvis. However, margins of even 10.3 mm were needed for the female patients with very large pelvic target volumes centered either in the symphysis or in the sacrum containing both of these structures. This was because the effect of rotation (p r 0.02) and the observer variation in image matching (p r 0.04) were significantly larger for the female patients than for the male patients. Even with daily image guidance, the required margins remained larger for the women. Patient rotations were largest about the lateral axes. The difference between the required margins was only 1 mm for the 2 immobilization devices. The largest component of overall systematic position error came from patient rotation. This emphasizes the need for rotation correction. Overall, larger position errors and setup margins were observed for the female patients with pelvic cancer than for the male patients. & 2014 American Association of Medical Dosimetrists.
Keywords: Radiotherapy Pelvis Setup errors Setup margins
Introduction The anatomy of the male and female pelvis is different. In the male pelvis, the bones and muscular volume are larger than those of the female pelvis. The female pelvic area tends to accumulate more fat. The shape of the pelvis is also different. In addition, the male skin is thicker. The different anatomic properties between the genders may require different considerations for patient setup in radiotherapy (RT). Haslam et al.1 have reported that setup accuracy is independent of patient weight, height, and age and it is not possible to estimate
Reprint requests to: Marko Laaksomaa, R.T.T., Department of Oncology, Tampere University Hospital (TAUH), PO BOX 2000 (Teiskontie 35), Tampere FI 33521, Finland. Tel.: þ358 331 169 623; fax: þ358 331 163 001. E-mail: marko.laaksomaa@pshp.fi
setup accuracy based on these factors. However, they have not investigated the difference between the setup accuracy for the men and the women. To the best of our knowledge, no comprehensive studies exist on that topic. In clinical practice, we have noticed that more image guidance has been needed to confirm the patient setup in pelvic RT for women than that needed for men. However, it is common to assume that equal setup margins can be applied because of the same image-guidance procedure and immobilization. Errors from different sources, such as observer-related errors, have not been investigated comprehensively in recent studies using kV imaging. It might be useful to know to what extent the uncertainty in patient setup is related to translation, rotation, and observerrelated factors. The purpose of this study was to estimate adequate setup margins for image-guided RT (IGRT) based on bony landmarks in
0958-3947/$ – see front matter Copyright Ó 2014 American Association of Medical Dosimetrists http://dx.doi.org/10.1016/j.meddos.2013.09.009
M. Laaksomaa et al. / Medical Dosimetry 39 (2014) 74–78
the pelvis. We evaluated whether an isotropic setup margin of 8 mm is adequate when considering weekly IGRT protocol, combination of patient rotation and deformation, and observer-related errors. We evaluated setup accuracy for both the male and female patients. We analyzed orthogonal x-ray images as they are widely used for frequent (routine) setup verification. As onboard 3dimensional (3D) verification imaging (such as cone beam computed tomography, CBCT) may be performed less frequently, setup margins should be confirmed suitable for the 2D imaging. Based on bony landmarks, 2D kV and CBCT alignments have been reported to correlate highly, but slightly different margins may be needed.2 Methods and Materials
75
of the 3 first treatment fractions and 5 mm in weekly imaging. We analyzed the acquired onboard images retrospectively in offline review according to the presented IGRT protocol. We estimated setup errors also without IGRT and with image guidance performed only in the first 3 treatment fractions to demonstrate transfer errors between the treatment-planning CT and a treatment unit.
Estimation of setup errors Reference treatment level was defined in the middle of PTV (MID-PTV). The MID-PTV point was located usually within ⫾ 1 cm from the midpoint of the pubic symphysis and the sacrum (Fig. 1). Setup errors were determined separately for the group M (n ¼ 33) and the group F (n ¼ 44) and for both the groups together. The number of the analyzed kV images was 816 and 480 for the men and women, respectively. Directions are expressed as anterior-posterior (AP or vertical), superior-inferior (SI or longitudinal), and LAT (lateral).
Patient rotation errors
Patient groups The group males (M) consists of consecutive patients with rectal (n ¼ 25) and prostate cancer (n ¼ 25) and the group females (F) of consecutive patients with gynecologic (n ¼ 27) and rectal (n ¼ 23) cancer. All the patients have a large planned target volume (PTV) because of the lymph node involvement as shown in Fig. 1. The average age for the group F and M was 68 and 71 years, respectively. Both groups were immobilized with a knee support that has been made in our department several years ago (device 1). The feet are tied with a stasis. We compared our fixation device to a commercial one used in our satellite unit in Lahti. This device is the combination of CIVCO's Kneefix and Feetfix (device 2), where both knees and feet are fixed into a supporting cushion. Both the devices are presented in Fig. 2. The Lahti group consists of 14 male and 14 female patients (18 patients with rectal cancer, 8 patients with gynecologic cancer, and 2 patients with urinary bladder cancer). Computed tomography (CT) imaging for treatment planning was done at 120 kVp with either Philips Brilliance Big Bore (Philips Medical Systems, Eindhoven, the Netherlands) or Toshiba Aquilion LB (Toshiba Medical System, Tokyo, Japan) using a slice thickness of 3 mm. The patients were treated with the intensity-modulated radiation therapy technique using 6- and 18-MV photon beams of Clinac 2300 iX (Varian Medical Systems, Palo Alto, CA). Image guidance was carried out using orthogonal kV images acquired with an onboard imaging system at 75 kV with 10 to 16 mAs for the anterior images and at 105 to 120 kV with 80 to 126 mAs for the lateral images.
The effect of pelvis rotation was investigated with 50 male and 50 female patients with total number of images being 652 and 600, respectively. This was done by determining systematic and random variations in the distance between the pubic symphysis and the sacrum as seen in Fig. 1, as they are the clearest landmarks and usually placed at the extreme edges of the pelvic targets. Also, deformation of bony structures may contribute to these results. However, the pelvis is a quite rigid object and the rotations play the most important role for these results. The extent of rotation was double checked by rematching the images with a time gap of 2 months.
Estimation of observer errors and comparison of the 2 fixation devices A group of 20 experienced radiation therapists evaluated the images and their image matches were compared with the reference MID-PTV match. Systematic (∑) and random (s) errors were determined for the differences. Potential benefit of device 2 in the reduction of the rotation errors was investigated based on data obtained from our satellite clinic, comprehending 28 patients with 378 images. Unfortunately, no larger data were available consistent with our IGRT protocol. Therefore, the devices were compared only for combined groups of the male and female patients.
Estimation of overall errors and treatment margins Investigated image-guidance protocols Our image-guidance protocol used an online correction with a fixed 5-mm action level for translational couch (patient setup) corrections in all 3 orthogonal directions. Imaging was performed in the first 3 treatment fractions and weekly thereafter. If the action level was exceeded, the imaging was repeated in the next fraction. An action level of zero was applied for couch vertical based on the average
The residual setup error of the treatment isocenter after weekly imaging, the rotation of the sacrum and the pubic symphysis, and the observer error were combined to obtain the overall setup errors. All the error components were added in quadrature.3 Systematic (∑) and random (s) components were handled separately. The total errors were used to calculate anisotropic setup margins using the van Herk formula m ¼ 2.5Σ þ 0.7s.3
Fig. 1. Regions of interest in the sacrum and the pubic symphysis (white boxes) used to estimate the pelvis rotation. Typical PTV covers both of these structures in (A) anterior and (B) lateral reference images. The translational position errors caused by the rotation about anterior axis (yaw) and lateral axis (pitch) were determined from images (A) and (B), respectively. (Color version of figure is available online.)
76
M. Laaksomaa et al. / Medical Dosimetry 39 (2014) 74–78
Fig. 2. (A) The house-made fixation device 1. Patient has a knee support and the feet are tied with a stasis. The device is not fixed to the treatment couch. (B) The device 2 is the combination of CIVCO's Kneefix and Feetfix, where both the knees and feet are fixed by supporting cushions attached to the couch. (Color version of figure is available online.) Statistical testing The F test was used to measure significance of the differences in systematic errors between the groups (test for equality of variance). The Wilcoxon rank sum test was used for the random errors (test for equality of means). A p r 0.05 was considered statistically significant.
Results Estimation of setup errors Overall setup errors by excluding the patient rotation are given in Table 1. Without the image guidance, an adequate margin in vertical direction was larger than 1 cm in both the groups caused by different bending of the CT and treatment unit couches. Mean overall error was 4 mm anteriorly. This means that it is important to correct vertical direction without tolerance because transfer errors between CT and treatment unit are large, but are often slightly below the tolerance of 5 mm. In the group F, mean overall error was 1.5 mm to the left. When position corrections were done based only on the images of the first 3 fractions, the average overall errors were near zero in both the groups. In the group F, relatively larger margins are needed in longitudinal and lateral directions. The excess of the tolerance limit of 5 mm in the group F was distributed as 11%, 19%, and 24% in AP, SI, and LAT directions, respectively. The corresponding values for the group M were 6%, 11%, and 10%. The low percentage for the vertical direction is because of zero action level applied based on the first 3 images.
With the weekly image guidance, the margins can be approximately reduced to half when compared with no image guidance at all. For the female group, it was impossible to make adequate setup corrections based on only the first 3 fractions and more frequent image guidance was needed. For the male group, the need of weekly imaging was smaller and the required margins were below 6 mm after correction based on the first 3 images. In the group M, the percentage of image-guided fractions was 37% for the patients with prostate cancer and 45% for the patients with rectal cancer. In the group F, both the patients with rectal cancer and the patients with gynecologic cancer, were imaged in 60% of the fractions. We noticed a time trend for setup errors in the AP direction in 10% of the investigated patients. For these patients, couch vertical was gradually raised during the treatment. Group average of overall errors was near zero, suggesting that there is no global systematic shift. Rotation errors The errors related to patient rotation are given in Table 2. It can be seen that systematic position errors due to rotation are larger than those caused by variation of isocenter position after weekly IGRT, given in Table 1. All the differences between the F and M groups are significant in all directions. The errors caused by the rotation of the pelvis were relatively larger in the group F than in the group M. Systematic rotation errors exceeding the tolerance of 5 mm were 4% in the group M and 6% in the group F (in 1 or more direction). A tighter level of 4 mm was exceeded in 8% and 22% in the M and F groups, respectively. With device 2, lateral random
Table 1 Position variation of treatment isocenter alone and the corresponding setup margins with the weekly image-guidance protocol (device 1). Errors for position correction based only on the first 3 fractions are given in parenthesis Women, n ¼ 33 AP ∑ error (mm) p-Value for ∑ s error (mm) p-Value for s Margin (mm) Fractions imaged (%)
1.2 0.72 1.9 0.52 4.3
Men, n ¼ 44 SI
(1.6) (0.05) (2.4) (0.01) (5.6)
1.3 0.35 2.1 0.01 4.8 60
LAT (2.1) (0.06) (2.8) (o 0.001) (7.2)
1.1 0.87 1.9 0.27 4.1
(2.2) (0.02) (3.3) (0.004) (7.8)
AP
SI
LAT
1.0 (1.2)
1.2 (1.5)
1.0 (1.5)
1.8 (1.8)
1.8 (1.8)
1.7 (2.0)
3.9 (4.3)
4.1 (5.1) 41
3.7 (5.1)
M. Laaksomaa et al. / Medical Dosimetry 39 (2014) 74–78
77
Table 2 Variations in the distance between the pubic symphysis and the sacrum in the MID-PTV match, and comparison of 2 different fixation devices. The results provide translational position shifts caused by the pelvis rotation
s (mm)
∑ (mm)
Women (device 1)* Men (device 1)* p-Value for the F and M groups F þ M, device 1† F þ M, device 2‡ p-Value for device 1 and 2
AP
SI
2.1 1.7 0.01 1.9 1.8 0.65
2.5 2.0 0.03 2.3 2.3 0.88
LAT
AP
1.8 1.1 o 0.001 1.5 1.7 0.16
2.1 1.4 0.02 1.9 1.7 0.11
SI
LAT
2.3 1.7 o 0.001 1.9 2.0 0.44
2.0 1.6 0.001 1.8 1.4 o 0.001
Device 1 has a knee support and stasis in feet, whereas device 2 has knee and feet support cushions. n
† ‡
n ¼ 50. n ¼ 100. n ¼ 28.
errors can be slightly reduced because of better fixation of the feet. The rotation errors were consistent within 0.4 mm in the rematch and the same parameters remained statistically significant or insignificant. Observer errors Observer errors were relatively small; still we found significant differences for the group F and M in SI and LAT directions, having an influence on the total setup margins (Table 3). Total setup errors and setup margins Total setup errors and resulting setup margins are listed in Table 4. When setup uncertainties based on bony landmarks were accounted for, the adequate overall setup margins were within 7 to 10 mm depending on target position in both the groups. The largest difference in the margins of 2.3 mm between the groups existed in the longitudinal direction. By removing the effect of rotation, the margins could be reduced approximately by 3 mm (range from 1.4 to 3.9 mm).
Discussion We found larger pelvic setup uncertainties for the female patients when compared with the male patients with the same IGRT protocol. The tattoo marks used for patient positioning were observed to shift in a different direction between these patient groups. This may be because of different skin types and amount of subcutaneous fat. The patient rotation was also larger in the female group. This may be because of different pelvic anatomy and fat content between the women and men. Our results suggest that these differences have an effect on daily setup and cause natural variation in the location of the setup marks relative to the bony structure. The setup errors obtained excluding the patient rotations are consistent with the literature studies.4,5 The random errors are higher than the systematic ones. Both error types can be
remarkably decreased by increasing the frequency of the image guidance.6 We observed shallow time trends in the vertical direction, but they were below the action level of 5 mm. This observation was consistent with recent reports.7,8 Our results indicate that the female pelvic region has larger systematic and random rotations than that of the male's. Rotation was observed both in the first treatment fraction and during the treatment course. The contribution of the patient rotation was remarkable as it required a 3-mm expansion to the margins, resulting in much greater irradiated volumes. Thus, our results suggest that couch corrections for both translation and rotation (6D) may be beneficial as has been recently suggested.4 Because the systematic errors were as wide as the random errors, it may be possible to partly correct the rotation based on offline image matching. In contrast, the random rotation about the lateral axis (causing mainly longitudinal random error) was large in the F group, which may render offline corrections inadequate. Because of this type of rotation, the bony landmarks shifted toward each other on an average by 1 mm in the anterior images for both the groups. The rotation of the pelvis about the coronal axis was smaller in both the patient groups. Even though the differences between the groups were statistically significant, they were still small. It may be difficult to remove all such small errors in practice. At least small deformation and rotation errors exist always between the pubic symphysis and the sacrum in the RT of the pelvis. Selection of the correct treatment alignment (isocentre) based on the images is not unique and there is variation between observers. Observer-related errors are usually not investigated or they are assumed to be negligible. If image matching is done only to the pubic symphysis, position error of the sacrum remains large. This explains mostly the obtained systematic observer errors. The observer errors were still relatively small in this study and Table 4 Total setup errors and adequate setup margins for 2 categories of PTV sizes estimated for the weekly image-guidance protocol based on bony landmarks (device 1). The errors include position shifts because of variation in isocenter position, pelvis rotation, and observer variation in image matching Women, n ¼ 50
Table 3 Observer-related errors in MID-PTV image matching (device 1) Women, n ¼ 25
∑ error (mm) p-Value for ∑ s error (mm) p-Value for s
∑ error (mm)*
Men, n ¼ 25
s error (mm)*
AP
SI
LAT
AP
SI
LAT
1.3 0.36 1.7 0.16
1.3 0.04 1.5 0.11
0.8 0.03 1.1 0.15
1.1
0.8
0.5
1.4
1.1
0.9
Margins (mm)* ∑ error (mm)† s error (mm)† Margins (mm)† n
†
Men, n ¼ 50
AP
SI
LAT
AP
SI
LAT
2.0 2.7 7.0 2.7 3.2 9.0
2.2 2.8 7.5 3.2 3.4 10.3
1.6 2.4 5.8 2.3 2.7 7.6
1.7 2.4 6.0 2.3 2.8 7.7
1.7 2.2 6.0 2.5 2.8 7.6
1.2 2.1 4.7 1.6 2.3 5.9
Results for PTV centered between the symphysis and the sacrum. Results for a very large PTV centered in the symphysis or the sacrum.
78
M. Laaksomaa et al. / Medical Dosimetry 39 (2014) 74–78
consistent with the literature. Lewis et al.9 have reported 1.1-mm (root-mean-square) deviation between the observers, using MV image guidance. We demonstrated that systematic observer errors are significantly different for the female and male patients in the SI and LAT directions. The error in the longitudinal direction has the largest contribution to the required margins. The results in Tables 3 and 4 implicate that larger observer errors are related to larger rotation errors. We expect that observer-related errors can be decreased by more accurate rules for image matching for large pelvic targets treated usually with intensity-modulated radiation therapy. We recommend the MID-PTV match to minimize overall setup error, unless otherwise considered by the responsible oncologist. It should be confirmed that the images cover the whole PTV area or at least the middle part of it and that the PTV structure is visible in the images. The current results were based on the bony landmarks in the 2D images. Because these images visualize bone projections, uncertainty in image matching may be larger than in tomographic 3D imaging, such as CBCT. In this sense, our results provide upper limits for the setup margins. Internal organ motion in pelvis is reported to be large. Internal motion occurs in a complex way not always correlating with the bony landmarks. Organs such as the uterine fundus may need even a 4-cm margin to properly cover its motion.10 Internal organs may also have systematic shifts during treatment period and, therefore, application of CBCT or magnetic resonance imaging during a treatment course is recommended in the literature.10-12 Internal tissue motion is out of the scope of our investigation and must be included separately to the margins. However, frequent image guidance is usually based on planar kV or MV images, supporting the feasibility of our current results. To investigate the effect of surgery, we calculated the errors merely for patients with rectal cancer before operation. There were 25 such patients in both the groups. The errors were equal within 0.3 mm with the results obtained for the whole groups, suggesting only a minor potential effect of the surgery. Despite the fact that the fraction of surgical patients is smaller in the male group, the results on the overall reflect normal clinical patient groups, supporting the feasibility of our results. Image guidance based on orthogonal 2D kV images is an effective way to control setup accuracy. It was effective enough to cover the isotropic margin of 8 mm for typical pelvic RT (such as treatment of rectum, cervix, endometrium, and prostate carcinoma with lymph node involvement) when accounting for the patient rotation and observer-related variation. The uncertainty related to the fixation device itself is not prominent because increase in margin caused by the patient rotation was small in the male group. We obtained very similar setup errors for the investigated devices consistently with the literature.13 This implies that it is not optimal to spend resources on the improvement of the immobilization device. Instead, it might be beneficial to harmonize the patient fixation and image-matching procedures. Because the commonly applied isotropic margin of 8 mm proved insufficient only for very large PTVs (centered in the symphysis or in the sacrum and encompassing both of these structures), even when daily IGRT is applied, it is appropriate to find a threshold for systematic position errors to render this margin sufficient. By accounting for the determined random errors, a threshold of 6 mm for systematic errors seems appropriate (estimated by using the van Herk formula) and should be applied for all the bony landmarks with the large target volume. The threshold could be used to screen patients needing immediate correction of patient posture or fixation. If this is not possible, the clinical importance of the large systematic position error should be evaluated, e.g., using 3D imaging. When necessary, adaptive replanning should be performed. We introduced this procedure in our clinic based on these results. We found routine daily IGRT
not necessary and we retained our weekly IGRT protocol because it suggests more frequent IGRT when tolerances are exceeded. As a conclusion, the adaptation of the threshold, the implementation of 6D couch corrections, and harmonization of the image-matching procedure seem to be the best improvements for patient setup accuracy in pelvic RT.
Conclusions We conclude that setup margins of 8 mm together with our weekly IGRT protocol are sufficient for the male patients with pelvic cancer and for the female patients with typical target volumes centered between the symphysis and the sacrum. However, for the female patients with large target volumes centered either in the symphysis or in the sacrum containing both of these structures, larger setup margins up to 10.3 mm should be applied with the weekly IGRT protocol, whereas margins up to 8.7 mm are sufficient with daily IGRT. This was because of larger patient rotation and observer variation in image matching observed for the female patients. For the female patients with such large targets, the margins of 8 mm can be used only when systematic position errors exceeding 6 mm are corrected for all the bony landmarks within the target, e.g., by correcting patient posture or fixation. If this is not possible, adaptive replanning should be considered. The results are based on matching of orthogonal 2D xray images to the bony landmarks in the middle of the target volume.
Acknowledgments This study was supported by the Pirkanmaa Hospital District, Elna Kaarina Savolainen fund (grant no. R12536). References 1. Haslam, J.J.; Lujan, A.E.; Mundt, A.J.; et al. Setup errors in patients treated with intensity-modulated whole pelvic radiation therapy for gynecological malignancies. Med. Dosim. 30:36–42; 2005. 2. Heng, L.; Zhu, X.R.; Zhang, L.; et al. Comparison of 2D radiographic images and 3D cone beam computed tomography for positioning head-and-neck radiotherapy patients. Int. J. Radiat. Oncol. Biol. Phys. 71:916–25; 2008. 3. International Commission on Radiation Units and Measurements (ICRU). ICRU Report 62. Prescribing, recording, and reporting photon beam therapy. Supplement to ICRU Report 50. ICRU, Bethesda, MD; 1999. 4. Ahmad, R.; Hoogeman, M.S.; Quint, S.; et al. Residual setup errors caused by rotation and non-rigid motion in prone-treated cervical cancer patients after online CBCT image-guidance. Radiother. Oncol. 103:322–6; 2012. 5. Stroom, J.C.; Olofsen-van Acht, M.J.J.; Quint, S. On-line set-up corrections during radiotherapy of patients with gynecological tumors. Radiat. Oncol. 6:499–506; 2000. 6. Rudat, V.; Hammoud, M.; Pillay, L.; et al. Impact of the frequency of online verifications on the patient set-up accuracy and set-up margins. Radiat. Oncol. 6:101; 2011. 7. El-Gayed, A.A.; Bel, A.; Vijlbrief, R.; et al. Time trend of patient setup deviations during pelvic irradiation using electronic portal imaging. Radiother. Oncol. 26:162–71; 1993. 8. Cranmer-Sargison, G.; Kundapur, V. Using kV-kV and CBCT imaging to evaluate rectal cancer patient position when treated prone on a newly available belly board. Med. Dosim. 37:117–21; 2012. 9. Lewis, D.G.; Ryan, K.R.; Smith, C.W. Observer variability when evaluating patient movement from electronic portal images of pelvic radiotherapy fields. Radiother. Oncol. 74:275–81; 2005. 10. Stewart, J.; Lim, K.; Kelly, V. Automated weekly replanning for intensitymodulated radiotherapy of cervix cancer. Radiat. Oncol. 78:350–8; 2010. 11. Brierley, J.D.; Dawson, L.A.; Sampson, E.; et al. Rectal motion in patients receiving preoperative radiotherapy for carcinoma of the rectum. Radiat. Oncol. 80:97–102; 2011. 12. Chan, P.; Dinniwell, R.; Haider, M.A.; et al. Inter- and intrafractional tumor and organ movement in patients with cervical cancer undergoing radiotherapy: A cinematic-MRI point-of-interest study. Int. J. Radiat. Oncol. Biol. Phys. 70: 1507–15; 2008. 13. Mitine, C.; Hoornaert, M.T.; Dutreix, A.; Beauduin, M. Radiotherapy of pelvic malignancies: Impact of two types of rigid immobilisation devices on localisation errors. Radiother. Oncol. 52:19–27; 1999.
