PITCH, ROLL, AND YAW VARIATIONS IN PATIENT POSITIONING
ADEEL KAISER, M.D.,* TIMOTHY E. SCHULTHEISS, PH.D.,†‡JEFFREY Y. C. WONG, M.D.,†
DAVID D. SMITH, PH.D.,§CHUNHUI HAN, PH.D.,†NAYANA L. VORA, M.D.,†
RICHARD D. PEZNER, M.D.,†YI-JEN CHEN, M.D.,†AND ERIC H. RADANY, M.D.†
*Department of Radiation Oncology, University of California Irvine, Orange, CA; Departments of†Radiation Oncology and
‡Radiation Physics, City of Hope National Medical Center, Duarte, CA; and§Division of Information Sciences,
City of Hope National Medical Center, Duarte, CA
Purpose: To use pretreatment megavoltage-computed tomography (MVCT) scans to evaluate positioning vari-
ations in pitch, roll, and yaw for patients treated with helical tomotherapy.
Methods and Materials: Twenty prostate and 15 head-and-neck cancer patients were selected. Pretreatment
MVCT scans were performed before every treatment fraction and automatically registered to planning kilo-
voltage CT (KVCT) scans by bony landmarks. Image registration data were used to adjust patient setups before
treatment. Corrections for pitch, roll, and yaw were recorded after bone registration, and data from fractions 1–5
and 16–20 were used to analyze mean rotational corrections.
Results: For prostate patients, the means and standard deviations (in degrees) for pitch, roll, and yaw corrections
were ?0.60 ? 1.42, 0.66 ? 1.22, and ?0.33 ? 0.83. In head-and-neck patients, the means and standard deviations
(in degrees) were ?0.24 ? 1.19, ?0.12 ? 1.53, and 0.25 ? 1.42 for pitch, roll, and yaw, respectively. No significant
difference in rotational variations was observed between Weeks 1 and 4 of treatment. Head-and-neck patients
had significantly smaller pitch variation, but significantly larger yaw variation, than prostate patients. No
difference was found in roll corrections between the two groups. Overall, 96.6% of the rotational corrections
were less than 4°.
Conclusions: The initial rotational setup errors for prostate and head-and-neck patients were all small in magnitude,
statistically significant, but did not vary considerably during the course of radiotherapy. The data are relevant to
couch hardware design for correcting rotational setup variations. There should be no theoretical difference between
these data and data collected using cone beam KVCT on conventional linacs. © 2006 Elsevier Inc.
Prostate cancer, Head-and-neck cancer, Setup variations, megavoltage-computed tomography, Image-guided
Conformal radiotherapy techniques for cancer treatment
have become prevalent over the last 15 years. They enable
the delivery of radiation with increasing precision to target
volumes. Such techniques include three-dimensional con-
formal radiation therapy (3D-CRT) and intensity-modulated
radiotherapy (IMRT), including tomotherapy. The enhanced
precision afforded by these methods allows for improved
sparing of normal tissues through a reduction in treatment
margins. Tighter margins, however, necessitate the use of
techniques to ensure proper target localization. As a result,
different systems have been developed to minimize setup
variations and improve target localization through image-
guided radiotherapy (IGRT). In the case of prostate cancer,
portal imaging, transabdominal ultrasound, periodic com-
puted tomography (CT) examination, and fiduciary markers
have all been used for IGRT (1–6).
Tomotherapy is a form of IMRT treatment using a helical
radiation delivery system. It deploys IGRT through com-
parison of daily pretreatment mega-voltage CT (MVCT)
scans with a kilo-voltage CT scan performed at the time of
simulation for treatment planning. The tomotherapy system
includes software for automatic registration of either soft-
tissue or bone structure from MVCT to planning CT. Al-
though both are available, we have used automatic bone
registration only. Translational deviations in the superior-
inferior (SI), anteroposterior (AP), and medial-lateral (ML)
coordinates are determined during the registrations process.
SI and AP translational errors can be corrected automati-
cally by computerized couch adjustments, but adjustments
in the ML direction must be made manually. Rotational
Reprint requests to: Timothy E. Schultheiss, Ph.D., Department
of Radiation Physics, City of Hope National Medical Center, 1500
Duarte Rd., Duarte, CA 91010. Tel: (626) 301-8247; Fax: (626)
930-5334; E-mail: email@example.com
Funded in part by a research grant from TomoTherapy, Inc.
Received March 16, 2006, and in revised form May 24, 2006.
Accepted for publication May 24, 2006.
Int. J. Radiation Oncology Biol. Phys., Vol. 66, No. 3, pp. 949–955, 2006
Copyright © 2006 Elsevier Inc.
Printed in the USA. All rights reserved
0360-3016/06/$–see front matter
variations about the SI axis (roll), AP axis (yaw), and ML
axis (pitch) are also recorded by the software. Roll varia-
tions are automatically corrected by the gantry at the time of
treatment, whereas pitch and yaw corrections must be made
manually (7). After bony alignment is achieved, patient
positioning can be further manipulated to match soft-tissue
structures or contoured regions on the KVCT and MVCT.
The patient is then treated when positioning is fully opti-
Image-guided radiotherapy therapy systems such as tomo-
therapy strive to increase the accuracy of patient positioning
before treatment. As described previously, in the case of to-
motherapy, many of the setup deviations are corrected auto-
matically. However, errors in pitch, yaw, and lateral directions
must be adjusted for manually. This can be an error-prone
process, sometimes requiring reimaging. Therefore, some
groups have tried to develop algorithms for pitch and yaw
corrections for helical tomotherapy using computerized trans-
lational couch movements that would eliminate the need for
on any linear accelerator, has been to develop devices that
permit couch movement with 6° of freedom. A University of
Michigan group has developed a device that is mounted on
treatment couches, and allows pitch and roll corrections to
within 0.03° (9, 10). Similarly, the Medical Intelligence Co.
has developed the HexaPOD robotic treatment couch, which
can correct for pitch, roll, or yaw displacements up to ?3°.
Several studies have examined the magnitude of rota-
tional setup errors in reference to bony landmarks. With
respect to pelvic tumor sites, many of these investigations
employed portal image registration techniques to detect
differences in patient positioning (11–14). Other studies
used serial CT scans to detect setup variations (15, 16). In
many of these studies, imaging was performed at a fre-
quency no greater than once per week. To our knowledge no
study, in any tumor site, has examined such errors using
daily, automatic pretreatment image registration techniques
such as those afforded by tomotherapy. Therefore, in No-
vember 2005, we initiated a retrospective evaluation of all
prostate and head-and-neck (H&N) cancer patients treated
at the City of Hope National Medical Center since the start
of the tomotherapy IGRT program. We analyzed daily
MVCT images, which had been registered to treatment
planning CT scans by bony landmarks. This radiographic
comparison of anatomic structures did not take organ mo-
tion into account because daily variations in the soft-tissue
targets were adjusted for manually after the automatic bone
registration process. Two patient populations were selected
to compare setup variations of geographically distinct tumor
sites treated using two different immobilization techniques.
METHODS AND MATERIALS
Prostate (n ? 20) and H&N cancer (n ? 15) patients treated
with tomotherapy were selected for this study. Patients who were
undergoing palliative radiotherapy, postprostatectomy radiation, or
who required multiple treatment plans during the course of radio-
therapy were excluded. With the exception of these exclusion
criteria, all H&N and prostate patients treated using tomotherapy
since it became available at the City of Hope Medical Center were
included in the analysis. No other inclusion criteria were used.
Before beginning radiation treatment, each patient had a KVCT
scan for radiotherapy planning. H&N patients were scanned head
first in the supine position. Conventional thermoplastic masks
(Med-Tec Inc., Orange City, IA) were used for immobilization
with three radiopaque markers placed on the mask (two laterally
and one centrally) for initial setup localization. After the first
fraction, H&N patients were subsequently set up through laser
alignment at the isocenter. Prostate cancer patients were scanned
feet-first using Vac-Lok devices (Med-Tec Inc.) for immobiliza-
tion of the lower extremities. Prostate cancer patients were simu-
lated with a full bladder and had three small tattoos placed on the
lower torso (right and left iliac crests and between the pubic bone
and umbilicus) to serve as reference points for daily treatment
setups. Spiral KVCT scans for both groups of patients were per-
formed using a 3 ? 3 mm stacked axial slice technique with a pitch
of 1.7. KVCT image resolution was 512 ? 512 pixels. The studies
were transferred for contouring only to the Voxel Q treatment
planning system (Philips Medical System, Eindhoven, The Nether-
lands). After gross tumor volume (GTV), clinical target volume
(CTV), planning target volume (PTV), and critical organ structures
were all contoured, the data were sent to the tomotherapy planning
station using the DICOM-RT (The Digital Imaging and Communi-
cations in Medicine-RT) protocol. Image resolution was downgraded
to 256 ? 256 pixels during the transfer process. This set of reference
images was used for automatic registration with MVCT scans.
All patients in this study had received at least 20 fractions of
radiation at the time of analysis. Before each treatment, patients
were immobilized in the supine position on the treatment table
with either thermoplastic face masks or Vac-Lok devices. They
were then aligned with room lasers using radiopaque markers on
face masks or tattoos on the lower torso. All patients then received
an MVCT scan of approximately 3.5 MV. The bone density
structures of the MVCT images were automatically registered to
the corresponding structures in the simulation KVCT scan. Slice
thicknesses for MVCT scans were either 4 mm or 6 mm with
corresponding pitches of two or three, respectively. These settings
correspond to “normal” and “coarse” modes as seen by the oper-
ator. “Fine” mode, which has a slice thickness of 2 mm and a pitch
of one, was not used in this study. Furthermore, we have not found a
noticeable difference in image quality among the three modes, and
94% of the scans reported in this study used the “coarse” mode. The
scan length encompassed at least the entire gross tumor volume.
After the MVCT scan was registered to the KVCT, the couch
was automatically shifted to correct for any deviations in the SI
and AP directions. Rotations about the SI axis (roll) were adjusted
by delaying or advancing the leaf sequence by the time equivalent
of the roll adjustment. Setup deviations in the lateral coordinates
required manual adjustment using the vernier on the table top.
Pitch and yaw deviations were adjustable only by repositioning the
patient. Additional manual shifts were also made after comparing
soft tissue positions in the KVCT and MVCT scans. All manual
image comparisons were made by overlaying KVCT and MVCT
images on the tomotherapy work station, and then examining them
In this study, we collected results of the image registration for
20 prostate and 15 H&N cancer patients using images from Week
1 (fractions 1–5) and Week 4 (fractions 16–20) of treatment. Thus
a total of 350 MVCT scans were reviewed. There were no missing
950I. J. Radiation Oncology ● Biology ● PhysicsVolume 66, Number 3, 2006
scans. Rotational deviations between daily MVCT scans and cor-
responding treatment planning KVCT scans were recorded. The
total scan length for each MVCT scan was also recorded.
We analyzed the pitch, roll, and yaw rotational variations in
both prostate and H&N patients. Statistical analysis included sum-
mary statistics, comparison of Week 1 and Week 4 data, and
comparisons of H&N to prostate results. Comparisons included
t-test for independent samples, homogeneity of variance, and cor-
relations among variables.
Table 1 shows the means, standard deviations, and 95%
confidence intervals of pitch, roll, and yaw corrections for
all patients. Table 2 illustrates the percentage of patients
requiring rotational corrections of either ?4° or ?3° by
patient group. Average MVCT scan lengths and standard
deviations in the SI direction for H&N and prostate cancer
patients were 92 ? 45 mm (range, 30–234 mm) and 71 ?
40 mm (range, 24–228 mm), respectively. For prostate
patients, the mean rotational variations were less than ?0.7°
in all cases, but were significantly different from zero.
Within this group, the means and standard deviations (in
degrees) for pitch, roll, and yaw corrections were ?0.60 ?
1.42, 0.66 ? 1.22, and ?0.33 ? 0.83. For H&N cases, the
means were less than ?0.25°, and the pitch and yaw cor-
rections were significantly different from zero. The means
and standard deviations (in degrees) for pitch, roll, and yaw
variations among H&N patients were ?0.24 ? 1.19, ?0.12
? 1.53, and 0.25 ? 1.42, respectively. The greatest rota-
tional variation was seen for yaw in prostate cases with 94%
of patients within ?3°.
An analysis of possible correlations between different
angle variations is shown in Table 3. A small, but statisti-
cally significant, correlation was seen for roll and yaw in
H&N patients (Pearson correlation 0.295, p ? 0.01). For
prostate patients, correlations were demonstrated between
yaw and roll (p ? 0.01) as well as yaw and pitch (p ?
0.05). All rotational correlations for prostate patients had
absolute values ?0.245. Such a small value of the cor-
relation coefficient is not clinically predictive, and exam-
ination of the data yielded no anatomic basis for the
Data addressing the change in magnitude of rotational
corrections by week of treatment are shown in Table 4. The
greatest difference between the standard deviations of angle
variations from Weeks 1 and 4 was seen for roll corrections
in H&N patients (1.72° vs. 1.31° for Weeks 1 and 4,
respectively). However, using the Student’s t-test, no statis-
tical significance was achieved for any comparison of angle
variations between Week 1 and Week 4 of radiotherapy.
To assess whether there was significant variability be-
tween prostate and H&N setups with regard to pitch, roll,
and yaw, we applied the Brown-Forsythe test for homoge-
neity of variance (17). Lacking a systematic error, the
means should all trend toward zero, and a test of the
difference in means may not be informative, especially with
regard to dispersion of the data. However, testing whether
the variances are different gets to the very nature of the
distribution of setup errors. We found that the variation in
pitch (as measured by the standard deviation or variance)
was significantly smaller for H&N patients, whereas yaw
variation was significantly larger. No difference was found
in the variation of roll.
In comparison to prior reports of setup variations in the
literature, this study used daily MVCT scans with automatic
registration to planning KVCT scans by bone anatomy. This
allowed for accurate angular measurements without ob-
server bias. In addition, this is also one of only a few studies
to have examined initial setup variations before any correc-
tions (13). Therefore, the actual setup deviations at the time
of treatment certainly should be smaller than those reported
in our analysis. The data from this study are relevant to
hardware design for correcting rotational setup variations.
There should be no theoretical difference between these
data and data collected using cone beam KVCT on conven-
Tables 5 and 6 compare the measured setup variations in
this study to results reported by other groups. At least eight
prior studies have examined rotational setup differences in
patients with pelvic tumor sites using portal imaging and CT
scan techniques (11–16, 18, 19). The same differences were
also examined in H&N patients using either portal imaging
or an optically guided patient localization system (20, 21).
In studies involving pelvic radiotherapy, the standard devi-
ations (in degrees) for pitch, roll, and yaw corrections
Table 1. Mean angle variations (in degrees) among all patients
Angle Disease siteMean
Head and neck
Head and neck
Head and neck
Abbreviation: MVCT ? megavoltage computed tomography.
n ? 200 registered MVCT scans for prostate patients.
n ? 150 registered MVCT scans for head-and-neck patients.
Table 2. Percentage of rotational corrections less than 4° or 3°
AngleMagnitude (degrees)ProstateHead and neck
951Pitch, roll, and yaw variations in patient positioning ● A. KAISER et al.
ranged from 1.27–3.4, 0.8–1.75, and 0.6–1.6, respectively
(11–13, 15, 16, 19). In general, pitch variations were greater
than yaw or roll deviations for patients treated to the pelvis.
The rotational variations in H&N patients in this study were
similar to those reported by the group from Memorial Sloan
Kettering Cancer Center when their data were adjusted to
reflect the standard deviation of the distribution rather than
the reported standard error of the means of the deviations.
Standard deviations (in degrees) for pitch, roll, and yaw
corrections in this study, vs. those from the Memorial Sloan
Kettering Cancer Center study, were 1.19 vs. 1.3, 1.53 vs. 0.9,
and 1.42 vs. 1.0, respectively (20). These values were less than
those reported from University of Wisconsin. The University
of Wisconsin group used an optical tracking system for local-
ization, which revealed standard deviations (in degrees) of 2.3,
3.2, and 1.6 for pitch, roll, and yaw, respectively (21).
Unlike prior studies, our investigation also compared
rotational errors in patients with tumors at two geographi-
cally distinct sites. The current analysis demonstrated sta-
tistically significant differences in rotational corrections for
pitch and yaw between prostate and H&N patients. The
standard deviations of the rotational corrections (in degrees)
for prostate vs. H&N patients were 1.42 vs. 1.19, 1.22 vs.
1.53, and 0.83 vs. 1.42 for pitch, roll, and yaw, respectively.
Although differences between pitch and yaw corrections in
the two patient groups were statistically significant, the
magnitudes of these corrections were consistently small
(less than 4° for 98% of cases). These small deviations
maybe explained by the use of different immobilization
techniques in prostate and H&N patients.
One of the paramount objectives of IGRT is to achieve
patient treatment positioning that is as close as possible to
the planning position. In addition, it is necessary to under-
stand the distribution of deviations from the planning posi-
tion so that errors that may go uncorrected are included in
the margins that define the PTV. To correct setup errors that
are encountered in IGRT, the most accurate methods are
those that do not rely on manual repositioning of the patient,
unless of course, the patient’s body is deformed relative to
its original position. Most manufacturers are now address-
ing the problem of automatic linear patient shifts after
pretreatment imaging. Tomotherapy also provides an auto-
matic shift in roll angle. For linacs, yaw can be addressed by
shifts in table angle, assuming sufficiently accurate table
control, and roll can be adjusted by modifying the gantry
angle. Thus only corrections for pitch remain to be inte-
grated into the patient setup procedure. Because 96.6% of
the angular variations found in this study were less than 4°,
it is unlikely that repositioning the patient manually can
yield a sufficiently accurate setup.
Overall, Table 6 shows that the standard deviations of
pitch, roll, and yaw do not show a consistent difference among
each other or between prostate and H&N groups. Although
larger values do exist, a standard deviation of about 1.5° seems
to be a reasonable upper bound for all angles. Thus new couch
tops that have ?3° rotation for all angles will be able to adjust
for most, but not all angular variations.
The idea that angular setup variations are as important as
linear variations should not be in dispute. If roll is 3° and a
15-cm target is centered on the origin of the coordinate
system, then the resultant positional variation at the edge of
the target is about 4 mm. This is certainly not small in terms
of corrections for image guided treatments.
Another potential concern on the part of some may be
that with increasingly sophisticated corrections for setup
Table 3. Pearson coefficient correlations between rotational variations and scan lengths
Disease siteVariableRoll YawScan length
–0.017 (NS) –0.154 (p ? 0.05)
0.239 (p ? 0.01)
–0.295 (p ? 0.01)
0.239 (p ? 0.01)
–0.295 (p ? 0.01)
Head and neck
Abbreviation: NS ? nonsignificant.
p values are noted in parentheses.
Table 4. Angle variations in week 1 vs. week 4 of treatment
Patient group Angle Week
Head and neckPitch
Abbreviation: MCVT ? megavoltage computed tomography.
n ? 100 registered MVCT scans per week for prostate patients.
n ? 75 registered MVCT scans per week for head-and-neck
952I. J. Radiation Oncology ● Biology ● PhysicsVolume 66, Number 3, 2006
Table 5. Rotational setup variations in studies involving pelvic or head-and-neck radiotherapy
Number of registered
Manual vs. automatic
techniqueImage registration time points
Pelvic disease sites
Crook et al. (11)DRRs74 Manual Gold seed and bone
Bone landmarks points
One DRR for boost phase
Daily EPID images. PFs were
performed every 5–7 days
Planning CT followed by one
CT scan during Weeks 2,
4, and 6
Weekly PFs before or after
Weekly PF in 3 of 6 fields
Hunt et al. (12) EPID images and PFs 76 (EPID)
Van Herk et al. (16)CT scans AutomaticChamfer matching with
bone and organ
Bone landmarks points
Schewe et al. (13)Digitized PFs 188Manual Pre and Post
Hanley et al. (14)Digitized PFs NG*Manual ? automaticChamfer matching using
Bone landmark points
Antolak et al. (15)DRRs from CT Scans51Manual ? automatic3 CT scans performed at
Daily PF with EPID until
approved setup, then
Weekly PF using EPID
Mubata et al. (18) EPID images compared
with DRR or plain
EPID images compared
with DRRs or CT data
NGManual ? automaticBone landmarks points
Remeijer et al. (19)279 Manual ? automaticBone landmarks points Post
Current study200 AutomaticMVCT to KVCT volume
registration with bone
Fractions 1–5 and 16–20 Pre
Head-and-neck disease sites
Hunt et al. (20)
Hong et al. (21)
Bone landmarks points
Point localization with
fiducial arrays on
maxillary bite plates
MVCT to KVCT volume
registration with bone
Daily port films
Daily optical alignment using
external fiducial arrays in
reference to isocenter
Fractions 1–5 and 16–20
Current studyMVCT scans150 Automatic
Abbreviations: EPID ? electronic portal imaging device; DRR ? digitally reconstructed radiograph; CT ? computed tomography; PF ? manual port films; MVCT ? megavoltage CT
scan; KVCT ? kilovoltage CT scan; NG ? not given.
* A total of 1239 portal images were obtained, but the number of image sets per fraction was not stated.
†This is based on an estimate of 10 fractions per patient given by the authors of the study.
positional variations, therapists may be less vigilant under
the assumption that all errors will be corrected. We have
seen no evidence for this. Our observation is that providing
therapists more effective tools for achieving accurate patient
setup results in their putting increasing effort in the task.
Finally, we want to emphasize that to apply the margin
reductions possible with image-guided technology, every
treatment should be image guided. This technology is not
used to check that the patient was in the correct position
during treatment. It is used to adjust the patient’s position
before treatment. The protocol for making position adjust-
ments in prostate treatments at our institution is to examine
visually the MVCT relative to the KVCT and make final
manual adjustments relative to those automatically calcu-
lated based on bone anatomy. These final adjustments are
based on the imaged position of the soft tissue target. Unless
the soft-tissue target is imaged daily, and the patient posi-
tioned accordingly, the margins must be substantially in-
creased. Otherwise, the target may be missed and the ad-
vantage of IGRT lost.
1. Alasti H, Petric MP, Catton CN, et al. Portal imaging
for evaluation of daily on-line setup errors and off-line
organ motion during conformal irradiation of carcinoma of
the prostate. Int J Radiat Oncol Biol Phys 2001;49:869–
2. Kitamura K, Shirato H, Seppenwoolde Y, et al. Three-dimen-
sional intrafractional movement of prostate measured during
real-time tumor-tracking radiotherapy in supine and prone
treatment positions. Int J Radiat Oncol Biol Phys 2002;53:
3. Mechalakos JG, Mageras GS, Zelefsky MJ, et al. Time trends
in organ position and volume in patients receiving prostate
three-dimensional conformal radiotherapy. Radiother Oncol
4. Pouliot J, Aubin M, Langen KM, et al. (Non)-migration of
radiopaque markers used for on-line localization of the pros-
tate with an electronic portal imaging device. Int J Radiat
Oncol Biol Phys 2003;56:862–866.
5. Serago CF, Chungbin SJ, Buskirk SJ, et al. Initial experience
with ultrasound localization for positioning prostate cancer
patients for external beam radiotherapy. Int J Radiat Oncol
Biol Phys 2002;53:1130–1138.
6. Shirato H, Harada T, Harabayashi T, et al. Feasibility of
insertion/implantation of 2.0-mm-diameter gold internal fidu-
cial markers for precise setup and real-time tumor tracking in
radiotherapy. Int J Radiat Oncol Biol Phys 2003;56:240–247.
7. Mackie TR, Olivera GH, Kapatoes JM, et al. Helical tomo-
therapy. In: Palta J, Mackie TR, editors. Intensity-modulated
radiation therapy. The state of the art. AAPM Summer School
Proceedings. Colorado Springs, CO: Medical Physics Publish-
ing; 2003. pp. 247–284.
8. Boswell SA, Jeraj R, Ruchala KJ, et al. A novel method to
correct for pitch and yaw patient setup errors in helical tomo-
therapy. Med Phys 2005;32:1630–1639.
9. Litzenberg DW, Balter JM, Hornick DC, et al. A mathemat-
ical model for correcting patient setup errors using a tilt and
roll device. Med Phys 1999;26:2586–2588.
10. Hornick DC, Litzenberg DW, Lam KL, et al. A tilt and roll
device for automated correction of rotational setup errors. Med
Table 6. Results of studies in Table 5
Rotational deviations in degrees
Pelvic disease sites
Crook et al. (11)
Hunt et al. (12)
NG 1.63 (1.55)
Van Herk et al. (16)*
Schewe et al. (13)
Hanley et al. (14)
Antolak et al. (15)
Mubata et al. (18)
Remeijer et al. (19)
?2 (in 95% of cases with immobilization)
Head-and-neck disease sites
Hunt et al. (20)
Hong et al. (21)
Abbreviations: EPID ? electronic portal imaging device; PF ? port films; NG ? not given.
Standard deviations are shown in parentheses.
* Reported standard deviations multiplied by square root of 2 to obtain the standard deviation of setup error.
†Hunt et al. reported standard error of the means of the deviations rather than the standard deviation of the distribution under the
assumption that all patients had the same number of images. The variance of the distribution is simply the average of variances for the
individual patients. The standard deviation is the square root of the variance. These values are shown in parentheses.
954I. J. Radiation Oncology ● Biology ● PhysicsVolume 66, Number 3, 2006
11. Crook JM, Raymond Y, Salhani D, et al. Prostate motion Download full-text
during standard radiotherapy as assessed by fiducial markers.
Radiother Oncol 1995;37:35–42.
12. Hunt MA, Schultheiss TE, Desobry GE, et al. An evaluation
of setup uncertainties for patients treated to pelvic sites. Int J
Radiat Oncol Biol Phys 1995;32:227–233.
13. Schewe JE, Balter JM, Lam KL, et al. Measurement of patient
setup errors using port films and a computer-aided graphical
alignment tool. Med Dosim 1996;21:97–104.
14. Hanley J, Lumley MA, Mageras GS, et al. Measurement of
patient positioning errors in three-dimensional conformal ra-
diotherapy of the prostate. Int J Radiat Oncol Biol Phys
15. Antolak JA, Rosen II, Childress CH, et al. Prostate target
volume variations during a course of radiotherapy. Int J Ra-
diat Oncol Biol Phys 1998;42:661–672.
16. van Herk M, Bruce A, Kroes AP, et al. Quantification of organ
motion during conformal radiotherapy of the prostate by three
dimensional image registration. Int J Radiat Oncol Biol Phys
17. Brown MB, Forsythe AB. Robust tests for the equality of
variances. J Amer Stat Assoc 1974;69:364–367.
18. Mubata CD, Bidmead AM, Ellingham LM, et al. Portal im-
aging protocol for radical dose-escalated radiotherapy treat-
ment of prostate cancer. Int J Radiat Oncol Biol Phys 1998;
19. Remeijer P, Geerlof E, Ploeger L, et al. 3-D portal image
analysis in clinical practice: An evaluation of 2-D and 3-D
analysis techniques as applied to 30 prostate cancer patients.
Int J Radiat Oncol Biol Phys 2000;46:1281–1290.
20. Hunt MA, Kutcher GJ, Burman C, et al. The effect of setup
uncertainties on the treatment of nasopharynx cancer. Int J
Radiat Oncol Biol Phys 1993;27:437–447.
21. Hong TS, Tome WA, Chappell RJ, et al. The impact of daily
setup variations on head-and-neck intensity-modulated radia-
tion therapy. Int J Radiat Oncol Biol Phys 2005;61:779–788.
955 Pitch, roll, and yaw variations in patient positioning ● A. KAISER et al.