Kilovoltage cone-beam CT: Comparative dose and image quality evaluations in partial

and full-angle scan protocols

Sangroh Kim, Sua Yoo, Fang-Fang Yin, Ehsan Samei, and Terry Yoshizumi

Citation: Medical Physics 37, 3648 (2010); doi: 10.1118/1.3438478

View online: http://dx.doi.org/10.1118/1.3438478
View Table of Contents: http://scitation.aip.org/content/aapm/journal/medphys/37/7?ver=pdfcov
Published by the American Association of Physicists in Medicine

Articles you may be interested in

Investigation into image quality and dose for different patient geometries with multiple cone-beam CT systems
Med. Phys. 41, 031908 (2014); 10.1118/1.4865788

A quality assurance framework for the fully automated and objective evaluation of image quality in cone-beam
computed tomography
Med. Phys. 41, 031901 (2014); 10.1118/1.4863507

Technical assessment of a cone-beam CT scanner for otolaryngology imaging: Image quality, dose, and
technique protocols
Med. Phys. 39, 4932 (2012); 10.1118/1.4736805

SUGGI57: Dose and Image Quality Evaluation for Partial and FullAngle Kilovoltage ConeBeam CT Protocols
Med. Phys. 37, 3114 (2010); 10.1118/1.3468090

Intraoperative cone-beam CT for guidance of head and neck surgery: Assessment of dose and image quality
using a C-arm prototype
Med. Phys. 33, 3767 (2006); 10.1118/1.2349687
Kilovoltage cone-beam CT: Comparative dose and image quality
evaluations in partial and full-angle scan protocols
Sangroh Kim
Medical Physics Graduate Program, Duke University Medical Center, Durham, North Carolina 27710
Sua Yoo and Fang-Fang Yin
Medical Physics Graduate Program, Duke University Medical Center, Durham, North Carolina 27710
and Department of Radiation Oncology, Duke University Medical Center, Durham, North Carolina 27710
Ehsan Samei
Department of Radiology, Duke University Medical Center, Durham, North Carolina 27710; Carl Ravin
Advanced Imaging Laboratories, Duke University Medical Center, Durham, North Carolina 27710;
and Duke Radiation Dosimetry Laboratory, Duke University Medical Center, Durham, North Carolina 27710
Terry Yoshizumia兲
Medical Physics Graduate Program, Duke University Medical Center, Durham, North Carolina 27710;
Department of Radiation Oncology, Duke University Medical Center, Durham, North Carolina 27710;
Department of Radiology, Duke University Medical Center, Durham, North Carolina 27710;
and Duke Radiation Dosimetry Laboratory, Duke University Medical Center, Durham, NC 27710
共Received 5 November 2009; revised 8 May 2010; accepted for publication 10 May 2010;
published 21 June 2010兲
Purpose: To assess imaging dose of partial and full-angle kilovoltage CBCT scan protocols and to
evaluate image quality for each protocol.
Methods: The authors obtained the CT dose index 共CTDI兲 of the kilovoltage CBCT protocols in an
on-board imager by ion chamber 共IC兲 measurements and Monte Carlo 共MC兲 simulations. A total of
six new CBCT scan protocols were evaluated: Standard-dose head 共100 kVp, 151 mA s, partial-
angle兲, low-dose head 共100 kVp, 75 mA s, partial-angle兲, high-quality head 共100 kVp, 754 mA s,
partial-angle兲, pelvis 共125 kVp, 706 mA s, full-angle兲, pelvis spotlight 共125 kVp, 752 mA s, partial-
angle兲, and low-dose thorax 共110 kVp, 271 mA s, full-angle兲. Using the point dose method, various
CTDI values were calculated by 共1兲 the conventional weighted CTDI 共CTDIw兲 calculation and 共2兲
Bakalyar’s method 共CTDIwb兲. The MC simulations were performed to obtain the CTDIw and
CTDIwb, as well as from 共3兲 central slice averaging 共CTDI2D兲 and 共4兲 volume averaging 共CTDI3D兲
techniques. The CTDI values of the new protocols were compared to those of the old protocols
共full-angle CBCT protocols兲. Image quality of the new protocols was evaluated following the
CBCT image quality assurance 共QA兲 protocol 关S. Yoo et al., “A quality assurance program for the
on-board imager®,” Med. Phys. 33共11兲, 4431–4447 共2006兲兴 testing Hounsfield unit 共HU兲 linearity,
spatial linearity/resolution, contrast resolution, and HU uniformity.
Results: The CTDIw were found as 6.0, 3.2, 29.0, 25.4, 23.8, and 7.7 mGy for the new protocols,
respectively. The CTDIw and CTDIwb differed within +3% between IC measurements and MC
simulations. Method 共2兲 results were within ⫾12% of method 共1兲. In MC simulations, the CTDIw
and CTDIwb were comparable to the CTDI2D and CTDI3D with the differences ranging from ⫺4.3%
to 20.6%. The CTDI3D were smallest among all the CTDI values. CTDIw of the new protocols were
found as ⬃14 times lower for standard head scan and 1.8 times lower for standard body scan than
the old protocols, respectively. In the image quality QA tests, all the protocols except low-dose head
and low-dose thorax protocols were within the tolerance in the HU verification test. The HU value
for the two protocols was always higher than the nominal value. All the protocols passed the spatial
linearity/resolution and HU uniformity tests. In the contrast resolution test, only high-quality head
and pelvis scan protocols were within the tolerance. In addition, crescent effect was found in the
partial-angle scan protocols.
Conclusions: The authors found that CTDIw of the new CBCT protocols has been significantly
reduced compared to the old protocols with acceptable image quality. The CTDIw values in the
point dose method were close to the volume averaging method within 9%–21% for all the CBCT
scan protocols. The Bakalyar’s method produced more accurate dose estimation within 14%. The
HU inaccuracy from low-dose head and low-dose thorax protocols can render incorrect dose results
in the treatment planning system. When high soft-tissue contrast data are desired, high-quality head
or pelvis scan protocol is recommended depending on the imaging area. The point dose method
can be applicable to estimate CBCT dose with reasonable accuracy in the clinical
environment. © 2010 American Association of Physicists in Medicine. 关DOI: 10.1118/1.3438478兴

3648 Med. Phys. 37 „7…, July 2010 0094-2405/2010/37„7…/3648/12/$30.00 © 2010 Am. Assoc. Phys. Med. 3648
3649 Kim et al.: Kilovoltage cone-beam CT: Dose and image quality evaluation
Key words: partial-angle scan, cone-beam CT, CTDI, point dose method, BEAMrnc, Monte Carlo
I. INTRODUCTION
Cone-beam CT 共CBCT兲 has become an essential tool to lo-
calize the treatment target in image guided radiation therapy
共IGRT兲. With relatively higher soft-tissue contrast compared
to megavoltage CBCT and three-dimensional 共3D兲 imaging
capabilities, on-board kilovoltage CBCT systems enable
more accurate patient setup in the treatment room.1–4 The
positioning deviations can be corrected immediately after
comparing 3D anatomical information between planning CT
and on-board CBCT when the patient is on the treatment
table. However, the use of the CBCT imaging should be
carefully justified by its clinical needs and radiation dose,
since the imaging volume is much larger than the treatment
volume and additional imaging dose will be added to healthy
tissues.5 Thus, accurate assessment of the imaging dose to
the healthy tissues/organs from the CBCT scan is crucial in
managing the cumulative dose resulting from the fraction-
ated treatments.
There have been several efforts performed to evaluate the
radiation dose from kilovoltage CBCT scans. Islam et al.6
investigated patient dose from the x-ray volume imaging
共XVI兲 system. They found that the maximum dose for a body
phantom 共30 cm diameter cylinder兲 was in the range from
1.8 to 2.3 cGy for 120 kVp and from 2.8 to 3.5 cGy for 140
kVp beams. For a head phantom 共16 cm diameter cylinder兲,
the maximum dose values ranged from 1.5 to 2.0 cGy for
100 kVp and from 2.6 to 3.4 cGy for 120 kVp beams. Note
that these CBCT scans were performed with relatively
smaller tube-current-time-product 共660 mA s兲 compared to
those of an On-Board imager® 共OBI兲 system. The mean
tube-current-time-product of the OBI system for the full ro-
tation scan was 1340 mA s. Wen et al.7 measured patient
dose from their OBI system using thermoluminescent dosim-
eters 共TLDs兲 and a RANDO® pelvis phantom 共The Phantom
Laboratory, Salem, NY兲. They found that the surface doses at
the anterior and posterior locations ranged from 3 to 6 cGy
and the largest doses were 6–11 cGy at femoral head and
neck. Song et al.8 reported the weighted CT dose index
共CTDIw兲 of the CBCT scan for both XVI and OBI systems.
They found that the CTDIw values from their OBI system
were 8.3 cGy for a head scan and 5.4 cGy for a body scan in
standard-dose mode. Ding et al.9 found that radiation doses
to radiosensitive organs can be ⬃300 cGy in total for a
whole treatment procedure if CBCT scans are performed at
every treatment. The situation parallels to the cumulative
dose effects of the repeated scans in multidetector CT
共MDCT兲. It should be noted that all the above results are
produced from full-angle 共360° rotation兲 CBCT scan
protocols.
In early 2008, Varian Medical Systems10 released the ad-
vanced imaging system 共OBI V1.4兲 using new cone-beam
CT scan protocols, with an intention to reduce the imaging
dose and time while maintaining similar CBCT image qual-
ity of the previous scan protocols 共OBI V1.3兲. There are six
Medical Physics, Vol. 37, No. 7, July 2010
3649
new CBCT scan protocols included in the advanced imaging
system based on the scan locations and image qualities. With
the reductions in peak x-ray tube voltage 共kVp兲, tube current
共mA兲, and number of projections with rotation angles, the
CBCT imaging dose from the new protocols was expected to
decrease significantly. In addition, ⬃200° partial-angle irra-
diation technique was introduced in the new protocols which
further reduced the imaging dose. This angle is approxi-
mately equivalent to 关180° +fan angle兴, the minimum angle
to achieve a perfect image reconstruction. Hereafter, we
refer this 200° CBCT scan as “partial-angle scan 关180°
+fan angle兴” and the 360° CBCT scan as “full-angle scan.”
By extending the concept of the conventional CT dose
index 共CTDI兲, Dixon11 proposed a point dose method to
evaluate the radiation dose of the MDCT scanner. The point
dose method measures a single point dose at the center of the
phantom while scanning the entire phantom with sufficient
length. Conceptually, this point dose has been used to repre-
sent the equilibrium dose that is fundamentally equivalent to
the CTDI of the CT scan protocol. We have investigated the
applicability of the point dose method to CBCT.
To the best of our knowledge, no accepted dose metric
currently exists for CBCT. The present work is a natural
extension of previously published works of CTDI in MDCT.
In particular, we adopted the point dose method in CBCT
geometry to estimate the CTDI of the new CBCT protocols.
Further, this paper presents new CTDI data, not previously
available, pertaining to the partial-angle scan in CBCT.
The purpose of this study is, therefore, to evaluate the
radiation doses via various CTDI calculation methods along
with the image quality analysis in the new CBCT protocols.
We measured the CTDI for the new scan protocols of the
OBI system using Dixon’s method11 with an ion chamber
共IC兲 and results were compared to the old scan protocols
共standard-dose head and body modes兲. We also performed
Monte Carlo 共MC兲 simulations using BEAMnrc/EGSnrc
code12,13 to estimate the CTDI values. In addition, we as-
sessed the image quality parameters for the new protocols
using the CBCT image quality assurance 共QA兲 protocol.14
Five parameters were tested: Hounsfield unit 共HU兲 linearity,
spatial linearity, spatial resolution, contrast resolution, and
HU uniformity.
II. METHOD AND MATERIALS
II.A. Point dose measurements
OBI V1.4 system 共Varian Medical Systems, Palo Alto,
CA兲 consists of a kilovoltage x-ray source and a kilovoltage
amorphous silicon detector orthogonally mounted on the lin-
ear accelerator.10 All the CBCT scans in this study were per-
formed using this OBI system. A total of six new CBCT scan
protocols were employed: 共1兲 Standard-dose head, 共2兲 low-
dose head, 共3兲 high-quality head, 共4兲 pelvis, 共5兲 pelvis spot-
light, and 共6兲 low-dose thorax. Table I provides the scan
parameters for the six scan protocols. Note that all the head
3650 Kim et al.: Kilovoltage cone-beam CT: Dose and image quality evaluation 3650

TABLE I. Detailed scan parameters and CTDIw values for new and old CBCT scan protocols.

New scan protocol Old scan protocol

Standard-dose Low-dose High-quality Pelvis Low-dose Standard Standard

head head head Pelvis spotlight thorax head body

Peak voltage 共kVp兲 100 100 100 125 125 110 125 125
Tube current 共mA兲 20 10 80 80 80 20 80 80
Exposure time 共ms兲 20 20 25 13 25 20 25 25
Rotation range 共deg兲 200 200 200 360 200 360 360 360
Number of projections 377 376 377 679 376 677 670 670
Exposure 共mA s兲 150.8 75.2 754 706.2 752 270.8 1340 1340
Fan type Full-fan Full-fan Full-fan Half-fan Full-fan Half-fan Full-fan Half-fan
Default pixel matrix 384⫻ 384 384⫻ 384 384⫻ 384 384⫻ 384 384⫻ 384 384⫻ 384 384⫻ 384 384⫻ 384
Slice thickness 2.5 2.5 2.5 2.5 2.5 2.5 2.5 2.5

protocols 共standard-dose, low-dose, and high-quality兲 and
the pelvis spotlight protocol use the “partial-angle scan”
technique while the pelvis and low-dose thorax protocols use
“full-angle scan” technique.
We measured the point doses with a 0.18 cm3 ion cham-
ber 共Model: 10⫻ 5-0.18, Radcal Corporation, Monrovia,
CA兲 at the center of the head and body CTDI phantoms,
which were constructed of polymethyl methacrylate
共PMMA兲 with a density of ␳ = 1.19 g / cm3. The head phan-
tom has a diameter of 16 cm and the body phantom has a
diameter of 32 cm. Both phantoms have the same longitudi-
nal length of 15.2 cm. The head phantom was scanned with
standard-dose head, low-dose head, and high-quality head
scan modes, and the body phantom was scanned with pelvis,
pelvis spotlight, and low-dose thorax scan modes. Note that
the head phantom was generally scanned with “full-fan”
mode, a typical cone-beam scanning method using a “full-
bowtie” filter, whereas “half-fan” mode was used with a
“half-bowtie” filter for the body phantom. In the half-fan
mode, the kilovoltage amorphous silicon detector was shifted
by 14.8 cm laterally so that the opening of the blades 共colli-
mators兲 is automatically adjusted corresponding to the loca-
tion of the detector; the x-ray beam partially covers the por-
tion of the body phantom 共see Fig. 2兲.
For each CBCT scan, the IC was sequentially placed at
the midpoint of the cylindrical cavities located in the center
and four peripheral positions 关Fig. 1共a兲 for head scan and
FIG. 1. Experimental setups for new CBCT scan protocols. 共a兲 Standard-
dose head, low-dose head, and high-quality head scans with a head CTDI
phantom 共16 cm in diameter兲; 共b兲 pelvis, pelvis spotlight, and low-dose
thorax scans with a body CTDI phantom 共32 cm in diameter兲. Note that both
CTDI phantoms are in the length of 15.2 cm.
Fig. 1共b兲 for body scan兴. Five cavity positions were labeled
as 3, 6, 9, 12 o’clock, and center positions. Thus, five CBCT
scans produced one set of scan data. When one cavity was
used in the measurements, the others were filled in with four
PMMA rods to avoid the influence of the cavity on the mea-
surements. For comparison purposes, old CBCT scan proto-
cols 共standard head and body scan兲 were also employed in
this study. The scan parameters of the old scan protocols are
also shown in Table I. For each scan mode, several sets of
scans were performed to check the systematic stability of the
IC measurements. The f-factor of 0.873 was used to convert
the measured exposures to the absorbed doses to the air.
These point doses were considered as the CTDI values, as-
suming that dose equilibrium was achieved as described in
Dixon’s point dose method.11
II.B. Monte Carlo simulations
The CTDI values of the new CBCT scan protocols were
also calculated based on simulations performed by using the
BEAMnrc/EGSnrc MC system.12,13 CBCT beam outputs
were characterized by the peak tube voltages and filtrations
in the MC simulations; all the six scan protocols were cat-
egorized into four CBCT beam output modes: 100 kVp full-
bowtie, 125 kVp full and half-bowtie, and 110 kVp half-
bowtie. These four CBCT modes were simulated in the
BEAMnrc system. Thus, a total of four corresponding phase-
space files were acquired from the BEAMnrc simulations. As
dose is linearly proportional to tube-current-time-product
共mA s兲, all the MC results were normalized by mA s. It may
be noted that the pelvis mode scan parameters in the new
protocol were the same with those of the old protocol, with
the exception of the tube-current-time-product. Thus, the
phase-space file of the old scan protocol was reused for the
pelvis scan simulation of the new protocol.
A Varian G242 x-ray tube model 共Varian Medical Sys-
tems, Palo Alto, CA兲 was built in the BEAMnrc/EGSnrc
system using the XTUBE component module 共CM兲 with a
pure tungsten target, CONSTAK CM for the tube exit window,
JAWS CM for the blades 共i.e., collimators兲, and SLABS CM for
the window of the OBI system. For both full-bowtie and
half-bowtie filters, we employed JAWS CM. The detailed tube

Medical Physics, Vol. 37, No. 7, July 2010

3651 Kim et al.: Kilovoltage cone-beam CT: Dose and image quality evaluation
FIG. 2. Varian OBI x-ray tube geometries for each full-fan mode and half-
fan mode. Full-fan mode uses full-bowtie filter and half-bowtie is used for
half-fan mode. Note that the opening of the blades 共collimators兲 is different
in each mode.
geometries for full-fan and half-fan modes are shown in Fig.
2. The MC x-ray tube models were validated by comparison
to the x-ray spectrum, half value layer, and lateral/depth dose
profiles in the previous studies.15,16
Mainegra-Hing and Kawrakow17 have reported the ineffi-
ciency of x-ray tube MC simulations without variance reduc-
tion techniques due to the small number of bremsstrahlung
photons produced from low energy incident electrons. To
increase the simulation efficiency, we used the variance re-
duction technique known as directional bremsstrahlung split-
ting with a splitting number of 10 000. We also enabled the
following low energy physics options for accurate simulation
referred from their study: Electron impact ionization, bound
Compton scattering, photoelectron angular sampling, Ray-
leigh scattering, atomic relaxation, and simple bremsstrah-
lung angular sampling. For the bremsstrahlung cross-
sections, NIST data18,19 were used in all the simulations and
XCOM data20 were used for photoabsorption and Rayleigh
scattering cross-sections. We also used an electron splitting
method to prevent high-weighted electrons from interacting
with other materials in the tube component. Both electrons
and photons were tracked to the threshold energy of 1 keV.
To create the material data for the 1 keV cutoff energies, we
reproduced all the relevant material data by using PEGS4 user
code.21
After the BEAMnrc simulations were completed, phase-
space files from the BEAMnrc were used to run DOSX-
YZnrc user code22 in order to calculate absorbed doses in the
phantoms. We used a source type 8 共phase-space source from
multiple directions兲 to simulate the rotational irradiation of
the CBCT scans.
Since the number of projections for each CBCT scan was
slightly different even in the same protocol, we introduced a
normalization factor 共NF兲 in order to account for the varia-
tion; NF is defined by the ratio of the normalized average ion
chamber dose, D̄IC cal
/ NPIC to the normalized average MC
dose, D̄MC / NPMC for several calibration points, where D̄IC
cal cal
cal
and D̄MC represent the average doses of the five point doses
and NPIC and NPMC are the number of projections for ion
chamber measurements and MC simulations, respectively.
Note that this normalization factor concept is similar to that
in a previous study,23 except for the number of projection
Medical Physics, Vol. 37, No. 7, July 2010
3651
divisions. With the normalization factor, absorbed doses
were calculated from MC simulations using the following
equation:
DMC_norm = NF ⫻ DMC_phantom ,
where
NF = 共D̄IC /D̄MC兲 ⫻ 共NPMC/NPIC兲,
cal cal
DMC_norm = normalized MC dose to the measurements
and
DMC_phantom = MC dose before the normalization.
To determine the exact phantom geometry for the MC
simulations, the phantoms were scanned with a GE
Lightspeed® RT 共GE Healthcare, Milwaukee, WI兲 scanner.
Then, CTCREATE user code22 was used to create a MC phan-
tom by converting the HU into the physical density based on
the CT calibration curve used for the CT scanner. Converted
voxel dimension was 2 ⫻ 2 ⫻ 2 mm3. DOSXYZ_SHOW user
code24 was used to find the axis locations of the individual
cavities in the CT phantoms to extract the point dose values
from the 3D dose files.
II.C. CTDI estimations with various methods
Leitz et al.25 proposed a method to estimate the average
CTDI values for a single slice CT scan in CT phantoms.
Assuming two premises, 共1兲 dose outside the nominal slice
thickness is zero and 共2兲 a linear dose decrease between the
peripheral and center positions, they derived the average
dose equation by integrating the CTDI in the phantoms. This
concept is equivalent to the current weighted CTDI 共CTDIw兲,
which is well known as follows:
CTDIw = WcenterCTDIcenter + WperipheriesCTDIperipheries ,
where
Wcenter = weighting factor for CTDIcenter ,
Wperipheries = weighting factor for CTDIperipheries ,
CTDIcenter = a point dose at a central axis,
and
CTDIperipheries = an average point dose at peripheries.
They found that the CTDIw equation with Wcenter = 1 / 3
and Wperipheries = 2 / 3 produces the average dose in the phan-
tom.
Recently, Bakalyar26 investigated the accuracy of the nu-
merical coefficients in this CTDIw method. On the basis of
six premises: 共1兲 A homogeneous cylinder phantom, 共2兲 full
lateral coverage of the CT beam to the phantom, 共3兲 smooth
variation in the CT beam density, 共4兲 uniform 360° irradia-
tion, 共5兲 monotonic dose delivery rate, and 共6兲 smooth
change in the bowtie filter thickness, he found that the nu-
merical coefficients Wcenter = 1 / 3 and Wperipheries = 2 / 3 are not
best fits because the discontinuity of the dose gradient at the
3652 Kim et al.: Kilovoltage cone-beam CT: Dose and image quality evaluation
center position is not physically real in the CT dosimetry.
From the second order 共parabolic兲 Taylor expansion, he
found the numerical coefficients Wcenter = 1 / 2 and Wperipheries
= 1 / 2 produce a more accurate fit in averaging the volumetric
CT doses.
In this study, four calculation methods were used to esti-
mate the CTDI values: 共1兲 Conventional CTDIw calculation
共Wcenter = 1 / 3 and Wperipheries = 2 / 3兲, 共2兲 Bakalyar’s method
共CTDIwb: Wcenter = 1 / 2 and Wperipheries = 1 / 2兲, 共3兲 averaging
the absorbed doses within a center slice 共CTDI2D兲, and 共4兲
averaging the absorbed doses through entire phantom slices
共CTDI3D兲. Note that methods 共3兲 and 共4兲 numerically aver-
age the dose distribution in a slice and a volume, respec-
tively, without using CTDIw equations. Methods 共1兲 and 共2兲
were applied to the ion chamber data, and all four methods
were applied to the MC data. To perform methods 共3兲 and
共4兲, 3D dose distribution data from MC simulations were
imported into MATLAB system 共Mathworks, Natick, MA兲 and
analyzed to calculate the average slice and volume doses. All
the results were cross-compared to each other. In the analy-
sis, we assume that the CTDIw of the method 共4兲 is the most
accurate CTDIw estimate because it is the numerically aver-
aged value directly calculated from the 3D dose distribution
in the phantoms; that is, there are no assumptions employed
in method 共4兲, such as the fitting and dose equilibrium used
in methods 共1兲 and 共2兲. Note that the results of the method
共3兲 are not expected as accurate as those of the method 共4兲
because the dose also decreases along the longitudinal
direction.
II.D. Image quality analysis
We performed the CBCT image quality QA protocol to
evaluate the image quality of the new protocols following the
work of Yoo et al.14 At the present time, no standard CBCT
image quality QA protocol exists. We employed the Catphan
504 phantom 共The Phantom Laboratory, Salem, NY兲 for all
the image quality tests. The Catphan 504 phantom consists of
four different modules and each module serves its own pur-
pose in the QA test. The CT images of each module are
shown in Figs. 3共a兲–3共e兲. The phantom was hanged at the
end of the couch with the aid of its wooden case as a coun-
terbalance and was aligned to the treatment isocenter. Six
CBCT scan protocols were evaluated and each image was
assessed for the five QA criteria: HU linearity, spatial linear-
ity, spatial resolution, contrast resolution, and HU unifor-
mity. All the image tests were performed in an Eclipse treat-
ment planning workstation 共Varian Medical Systems, Palo
Alto, CA兲.
共a兲 HU verification. The CTP 404 module contains seven
materials which have different densities. The materials
and corresponding HU values can be referred from the
manual of the phantom. We selected a region of interest
共ROI兲 in the designated material section using the
“Area profile” tool to measure the mean HU values.
The ROI size was about 0.7⫻ 0.7 cm2 and the accep-
tance criteria was ⫾40 HU 关see Fig. 3共a兲兴.
共b兲 Spatial linearity. There are four rods of 3 mm diameter
Medical Physics, Vol. 37, No. 7, July 2010
3652
formulating a 5 cm square in the CTP 404 module. The
distance between two adjacent rods was calculated us-
ing “Distance” tool in the Eclipse system. The toler-
ance range was ⫾0.5 cm 关see Fig. 3共b兲兴.
共c兲 Spatial resolution. The CTP 528 module contains 21
line pair/cm high resolution test gauge and w ⫾ e deter-
mined the least discernable bar group in the test. The
acceptance tolerance was 6 lp/cm 关see Fig. 3共c兲兴.
共d兲 Contrast resolution. The CTP 515 module contains
three groups of nine supraslice disks with diameters of
2–15 mm and subject contrast of 0.3%, 0.5%, and
1.0%. We visually inspected the image to determine the
lowest contrast disk visible. The recommended toler-
ance was 1% fourth disk 关see Fig. 3共d兲兴.
共e兲 HU uniformity. The CTP 486 module consists of a
single uniform disk of 20 cm diameter. We measured
mean HU for five selected ROI regions: Center, top,
left, right, and bottom. The size of ROI was about
1.0⫻ 1.0 cm2 and the acceptance criterion was ⫾40
HU 关see Fig. 3共e兲兴.
III. RESULTS
Table II presents the CTDI values obtained from IC mea-
surements and MC simulations and CTDI differences be-
tween each calculation method. All the CTDI values
共CTDIw, CTDIwb, CTDI2D, and CTDI3D兲 of the new CBCT
protocols ranged from 3 to 29 mGy for both IC measure-
ments and MC simulations, while those of the old protocols
ranged from 39 to 85 mGy. The CTDI of the new protocols
were found to be lower by about a factor of 14 for a
standard-dose head scan mode and 1.8 times for standard
body scan mode 共pelvis兲 than the old head and body proto-
cols, respectively. Note that the CTDIw per 100 mA s calcu-
lated from method 共1兲 are similar across same kVp and fil-
tration to 7%.
As can be seen in Table II, the differences between IC
measurements and MC simulations were ⬍2% for method
共1兲 and ⬍3% for method 共2兲. The differences between meth-
ods 共1兲 共conventional CTDIw method兲 and 共2兲 共Bakalyar’s
CTDIw method兲 were less than 2% for the head scans, while
they were about 8%–12% for the body scans. The differences
between point dose method of IC measurements 共CTDIw and
CTDIwb兲 and central slice averaging method 共CTDI2D兲 were
found as ⬍4% for head scans while ⬍6% for body scans.
Large differences 共9%–21%兲 were found between the results
of method 共1兲 共conventional CTDIw method兲, 共2兲 共Bakalyar’s
CTDIw method兲, 共3兲 central slice averaging method, and that
of method 共4兲 共entire volume averaging method兲.
Two-dimensional 共2D兲 dose distributions of each CBCT
scan protocol were visualized in the MATLAB system as
shown in Fig. 4. The differences of dose distributions are
noted between partial-angle and full-angle CBCT scan pro-
tocols, i.e., the partial-angle CBCT scan protocols show
larger dose along the scanning angles than in the nonirradia-
tion angles, while full-angle CBCT scan protocols produce
uniform angular dose distribution. The phantom size effect in
3653 Kim et al.: Kilovoltage cone-beam CT: Dose and image quality evaluation
the dose distribution of the short scan can be also noticed
between the head and body protocols; pelvis spotlight proto-
col delivers more crescent-shaped dose distribution 关Fig.
4共c兲兴, while head protocols show more half-moon dose dis-
tribution 关Fig. 4共a兲兴.
Table III shows the results of image quality QA tests for
new and old scan protocols. In the HU verification test, all
the protocols were within the tolerance 共⫾40 HU兲 except
low-dose head and low-dose thorax scan protocols; low-dose
head failed in Teflon and Delin and low-dose thorax failed
Medical Physics, Vol. 37, No. 7, July 2010
3653
FIG. 3. CT images of the CATPHAN
504 phantom for the image quality
tests of 共a兲 HU verification 共module:
CTP 404兲, 共b兲 spatial linearity 共CTP
404兲, 共c兲 spatial resolution 共CTP 528兲,
共d兲 contrast resolution 共CTP 515兲, and
共e兲 HU uniformity 共CTP 486兲.
every material except air. As can be seen in Fig. 5, the mea-
sured HU values of the low-dose head scan start to deviate at
the nominal value of ⬃340 HU and the deviation becomes
more significant in the higher HU region. The low-dose tho-
rax protocol shows same pattern; the HU values increase at
the nominal HU of ⫺1000 and the deviation becomes larger
as the HU increases. All the tests of the spatial linearity,
spatial resolution, and HU uniformity met the manufacturer
suggested tolerances. In the contrast resolution test, only
high-quality head and pelvis scan protocols were within the
3654 Kim et al.: Kilovoltage cone-beam CT: Dose and image quality evaluation 3654

TABLE II. CTDI values and comparisons between the CTDI values acquired from method 共1兲–共4兲. The CTDI differences were calculated by following
equation: 共CTDIfront − CTDIrear兲 / CTDIrear ⫻ 100. Note that the abbreviations IC and MC represent ion chamber measurements and Monte Carlo simulations.

New scan protocols Old scan protocols

Standard-dose Low-dose High-quality Pelvis Low-dose Standard Standard

CTDI values head head head Pelvis spotlight thorax head body

CTDIw IC 共mGy/100mA s兲 4.00 4.24 3.84 3.60 3.16 2.83 6.27 3.41
IC 共mGy兲 6.03 3.19 28.96 25.44 23.78 7.68 83.97 45.75
MC 共mGy兲 5.99 3.16 28.78 25.09 23.53 7.58 83.50 45.22
CTDIwb IC 共mGy兲 6.08 3.24 29.27 23.44 21.97 6.99 84.94 42.03
MC 共mGy兲 6.01 3.17 28.85 22.66 21.40 6.77 83.87 40.84
CTDI2D MC 共mGy兲 5.93 3.13 28.48 24.34 22.40 7.30 83.65 43.88
CTDI3D MC 共mGy兲 5.36 2.83 25.74 21.52 19.72 6.43 75.68 38.80

Comparison between each CTDI value

CTDIw共IC vs MC兲 共%兲 +0.67 +0.95 +0.63 +1.39 +1.06 +1.32 +0.56 +1.17
CTDIwb 共IC vs MC兲 共%兲 +1.16 2.21 +1.46 +3.44 +2.66 +3.25 +1.28 +2.91
CTDIw vs CTDIwb 共IC兲 共%兲 ⫺0.82 ⫺1.54 ⫺1.06 +8.53 +8.24 +9.87 ⫺1.14 +8.85
CTDIw vs CTDIwb 共MC兲 共%兲 ⫺0.33 ⫺0.32 ⫺0.24 +10.72 +9.95 +11.96 ⫺0.44 +10.72
CTDIw vs CTDI2D 共IC兲 共%兲 +1.69 +1.92 +1.69 +4.52 +6.16 +5.21 +0.38 +4.26
CTDIwb vs CTDI2D 共IC兲 共%兲 +2.53 +3.51 +2.77 ⫺3.70 ⫺1.92 ⫺4.25 +1.54 ⫺4.22
CTDIw vs CTDI3D 共IC兲 共%兲 +12.50 +12.72 +12.51 +18.22 +20.59 +19.44 +10.95 +17.91
CTDIwb vs CTDI3D 共IC兲 共%兲 +13.43 +14.49 +13.71 +8.92 +11.41 +8.71 +12.24 +8.32
CTDI2D vs CTDI3D 共%兲 +10.63 +10.60 +10.64 +13.10 +13.59 +13.53 +10.53 +13.09

acceptance level 共disk 4 in supraslice 1.0% group兲. In addi-
tion, a crescent effect was found especially at the interface of
two different materials in the partial-angle scan protocols
shown in Fig. 6共a兲.
IV. DISCUSSION
Since the introduction of kV CBCT imaging technique to
radiation oncology, significant efforts6–9 had been made to
accurately quantify CBCT imaging dose in order to minimize
it. Theoretically, one can perfectly reconstruct a 3D image of
a fan-beam CT scan with a partial-angle scan which acquires
the projections of 180° plus fan angle. With the similar fan
geometry, this concept may also be applicable to the CBCT
scan. The partial-angle CBCT scan protocols has been em-
ployed in the new OBI system 共V1.4兲 in order to achieve
lower dose while maintaining similar CBCT image quality.
To the best of our knowledge, this paper presents the first
dosimetric data for the partial-angle CBCT scan protocols
where the dose distribution is angularly inhomogeneous. In
addition, no suitable technology exists to accurately measure
the whole nonuniform 3D dose distribution for kV energy
range at the present time. Radiochromic film or a large num-
ber of TLDs may be used to acquire the 2D dose distribution;
however, the latter, in particular, would be highly labor in-
tensive and time consuming. To overcome these limitations,
we employed the MC method.
In this study, we have estimated the CTDI for both partial
and full-angle CBCT scan protocols by using the point dose
method and numerical averaging methods. MC simulations
were also performed to benchmark the accuracy of CTDI of
the point dose measurements. As shown in Table II, MC
simulations were validated by the comparison of CTDIw and
CTDIwb values. As expected, the CTDI values of the head
protocols were linearly proportional to the tube-current-time-
product 共mA s兲; standard-dose head protocol produced twice
higher dose than low-dose head protocol and five times
lower dose than high-quality head protocol did. In the body
protocols, pelvis protocol produced higher dose than pelvis
spotlight protocol did in spite of its lower mA s. This is
caused by the fan-type differences directly related to the
shape of the bowtie filters as well as irradiation pattern; a
full-bowtie has ⬃2.8 cm thick aluminum filter in both left
and right off-axial locations, while a half-bowtie has the
thick part of the filter only in right off-axis location 共see Fig.
2兲. The irradiation patterns between full-fan and half-fan
modes are also different as described in Sec. II. These dif-
ferences between full-fan and half-fan modes cause different
attenuations of the off-axial beam which consequently
brought the nonlinearity in CTDI. Note that CTDI values of
pelvis protocol 共new兲 are proportionally smaller than those
of the standard body protocol 共old兲 with smaller mA s.
We noticed that the differences between the conventional
CTDI method 共CTDIw兲 and Bakalyar’s CTDI method
共CTDIwb兲 for head scans were relatively smaller than those
for body scan. To analyze this effect, we acquired lateral
profiles from the 3D dose data of head and body scans 共old
protocols兲 at center, middle, and edge location in the CT
phantoms. As can be seen in Figs. 7共a兲 and 7共b兲, the profiles
from the head scan are relatively linear, while those of body
scan are in parabolic shapes. In a linear curve such as the
head profile, there will be no large differences between linear
and parabolic fittings, while large differences could be ex-
pected in a parabolic curve 共body profile兲 because linear fit-
ting is inappropriate for parabolic functions.

Medical Physics, Vol. 37, No. 7, July 2010

3655 Kim et al.: Kilovoltage cone-beam CT: Dose and image quality evaluation 3655

FIG. 4. Visualization of the 2D dose distribution for each new CBCT scan protocol: 共a兲 Head, 共b兲 pelvis, 共c兲 pelvis spot light, and 共d兲 low-dose thorax scans.
The start and end points for the partial-angle scan was marked in Fig. 4共a兲. Note that the color scale is in the unit of mGy.

We found that the CTDI differences between the Bakal-
yar’s CTDI method 共CTDIwb兲 and central slice averaging
method 共CTDI2D兲 are smaller than those between the con-
ventional CTDI method 共CTDIw兲 and central slice averaging
method 共CTDI2D兲 for the body scans 共see Table II兲, whereas
the opposite results were found for the head scans. This can
be explained by the previously described fitting effect for
different size phantoms. The lateral profile of the head scan
is relatively flat 共linear兲, while that of the body scan shows a
parabolic shape for the center profiles 关see Fig. 7共b兲兴; thus,
one can expect that linear fitting 共fit method A兲 will produce
smaller differences for the head scans, while parabolic fitting
共fit method B兲 will produce smaller errors for the body scan.
To validate this observation, we performed linear and para-
bolic fittings to the lateral profiles of the center slice for both
head and body scans using a MATLAB function, POLYFIT.
Then, we estimated the dose profile integral 共DPI兲 using true
values and fitted curves to evaluate the deviations of the DPI
values. In the results, the DPI deviation of the linear fit was
11.1% and that of the parabolic fit was 11.4% for the head
scan. The deviation of the linear fit was 24.7% and that of
parabolic fit was 16.3% for the body scan. Thus, we verified
that linear and parabolic fits do not produce significant dif-
ferences for the head scan, whereas parabolic fitting pro-
duces more accurate dose estimation than linear fitting for
the body scan.
Although we expected that the central slice averaging
method 共CTDI2D兲 yields a better dose estimation to entire
volume averaging method 共CTDI3D兲 than the Bakalyar’s
CTDI method 共CTDIwb兲 does, opposite results were found in
the body scan data; the differences between CTDI2D vs
CTDI3D was larger than those between CTDIwb of IC vs
CTDI3D 共or CTDIwb of MC vs CTDI3D兲 for the body scan,
while those between CTDI2D vs CTDI3D was smaller for the
head scan. This can be also explained with the fitting effect;
as can be seen in Fig. 7共b兲, the parabolic fitting curve is
lower than the true lateral curve, while the linear fitting curve
is higher than the true lateral curve. Thus, one can expect
that the parabolic fitting 共fit method B兲 will underestimate
the dose, while the linear fitting 共fit method A兲 will overes-
timate it for the body scan. This trend can be seen in Table II;
the differences of CTDIwb of IC vs CTDI2D are all negative,
while CTDIw of IC vs CTDI2D are all positive. In summary,
because the Bakalyar’s CTDI method 共CTDIwb兲 underesti-

Medical Physics, Vol. 37, No. 7, July 2010

3656 Kim et al.: Kilovoltage cone-beam CT: Dose and image quality evaluation 3656

TABLE III. Results of image quality QA tests for new and old CBCT scan protocols. Note that all the QA data of the old protocols are referenced from Ref.
14. The HU differences were calculated by following equation: 共HUmeasured − HUnominal兲. Note that the abbreviation n.HU represents nominal HU.

New scan protocols Old scan protocols

Standard-dose Low-dose High-quality Pelvis Low-dose Standard Standard

QA parameters head head head Pelvis spotlight thorax head body

HU Material n.HU Image quality analysis

verification Air ⫺1000 −991⫾ 13.8 −998⫾ 3.8 −996⫾ 5.7 −994⫾ 16.0 −999⫾ 2.2 −980⫾ 22.9 −1022⫾ 1.8 −1023⫾ 1.9
Teflon 990 996⫾ 37.9 1135⫾ 41.8 1001⫾ 17 985⫾ 20.3 1019⫾ 16.4 1155⫾ 19.1 970⫾ 7.3 964⫾ 20.4
Delin 340 352⫾ 43.4 395⫾ 48.2 344⫾ 17.2 336⫾ 10.7 365⫾ 17.0 465⫾ 12.7 349⫾ 13.4 343⫾ 18.0
Acrylic 120 123⫾ 33.6 133⫾ 35.8 121⫾ 16.6 116⫾ 10.6 123⫾ 14.3 211⫾ 12.8 128⫾ 5.8 128⫾ 13.9
Polystylene ⫺35 −45⫾ 38.9 −51⫾ 39.3 −41⫾ 14.8 −43⫾ 11.0 −39⫾ 13.1 35⫾ 13.3 −39⫾ 5.2 −44⫾ 13.8
LDPE ⫺100 −99⫾ 32.9 −110⫾ 40.6 −100⫾ 13.8 −98⫾ 8.4 −99⫾ 12.4 −19⫾ 13.4 −94⫾ 3.6 −99⫾ 11.0
PMP ⫺200 −192⫾ 34.9 −215⫾ 30.4 −193⫾ 14.3 −187⫾ 10.3 −192⫾ 11.5 −115⫾ 13.1 −189⫾ 8.2 −193⫾ 10.4
HU Air Unit: %, +9 +2 +4 +6 +1 +20 ⫺22 ⫺23
difference Teflon tolerance: ⫾40 +6 +145 +11 ⫺5 +29 +165 ⫺20 ⫺26
Delin HU +12 +55 +4 ⫺4 +25 +125 +9 3
Acrylic +3 +13 +1 ⫺4 +3 +91 +8 8
Polystylene −10 ⫺16 ⫺6 ⫺8 ⫺4 +70 ⫺4 ⫺9
LDPE +1 ⫺10 0 +2 +1 +81 +6 +1
PMP +8 ⫺15 +7 +13 +8 +85 +11 +7
Spatial linearity 共cm兲 4.99⫾ 0.01 4.99⫾ 0.02 4.99⫾ 0.02 5.00⫾ 0.01 5.01⫾ 0.01 4.98⫾ 0.02 4.97–5.01 4.97–5.01
Spatial resolution 共group兲 8 8 9 6 8 6 8 6
Contrast resolution 共disk兲 0 0 4 5 3 1 5 4
HU ROI #1 共HU兲 12⫾ 32.6 14⫾ 31.4 12⫾ 15.2 −9 ⫾ 7.5 13⫾ 12.2 74⫾ 8.5 12⫾ 8.3 14⫾ 48.1
uniformity ROI #2 共HU兲 −6 ⫾ 32.1 −12⫾ 34.1 −6 ⫾ 13.6 2 ⫾ 8.7 −6 ⫾ 12.2 82⫾ 11.8 −2 ⫾ 4.5 −15⫾ 20.9
ROI #3 共HU兲 −3 ⫾ 32.4 −9 ⫾ 36.0 −4 ⫾ 14.4 4 ⫾ 8.7 0 ⫾ 10.8 80⫾ 11.8 −1 ⫾ 3.8 −19⫾ 20.0
ROI #4 共HU兲 −12⫾ 33.1 −19⫾ 31.7 −11⫾ 15.4 −4 ⫾ 7.8 −11⫾ 12.3 80⫾ 11.9 −3 ⫾ 5.1 −20⫾ 18.4
ROI #5 共HU兲 −4 ⫾ 34.6 −9 ⫾ 34.3 −1 ⫾ 15.2 −1 ⫾ 9.1 −1 ⫾ 12.4 80⫾ 12.7 −7 ⫾ 5.1 −20⫾ 12.7
Mean ⫺2.6 ⫺7 ⫺2 ⫺1.6 ⫺1 79.2 ⫺0.2 ⫺12
Standard deviation 8.9 12.4 8.6 5.1 9.0 3.0 7.2 14.7

mates the dose compared to central slice averaging method
共CTDI2D兲, results from the Bakalyar’s CTDI method
共CTDIwb兲 are closer to results from the entire volume aver-
FIG. 5. HU linearity test for the new CBCT protocols. Legends are as fol-
lows: CBSDH: Standard-dose head; CBLDH: Low-dose head; CBHQH:
High-quality head; CBPEL: Pelvis; CBPSL: Pelvis spotlight; CBLDT: Low-
dose thorax; CBFFold: Old full-fan; and CBHFold: Old half-fan. The
negCriteria and posCriteria represent the acceptance criteria of ⫾40 HU.
aging method 共CTDI3D兲, which is theoretically assumed to
represent a most accurate model among the methods.
We noted large CTDI differences between the results of
methods 共1兲–共3兲 and that of method 共4兲, as shown in Table II.
This can be explained by the nature of the calculation meth-
ods. Methods 共1兲–共3兲 use the point dose or plane dose in the
central axial plane of the phantoms, while method 共4兲 nu-
merically averages the entire volumetric dose distribution.
The differences are due to the variation in the longitudinal
dose distribution. As can be seen in Fig. 8, the dose in the
longitudinal direction for both head and body scans de-
creases with distance from the central plane; this effect is not
considered in methods 共1兲–共3兲. We also found that CTDIwb of
IC is slightly inferior to CTDIw of IC for the head phantom
but remarkably better for the body phantom compared to
CTDI3D. Again, this is mainly due to the fitting discrepancies
previously described.
Therefore, when IC measurements are chosen to estimate
CBCT dose, such as in the clinical environment, the use of
method 共2兲 is recommended to estimate the CBCT dose more
accurately than method 共1兲, especially for the body scans. In
addition, it should be noted that methods 共1兲 and 共2兲 always
overestimate the CTDI values compared to method 共4兲 and
the amount of radiation dose is as small as in the mGy level.
Thus, the use of methods 共1兲 and 共2兲 would be still accept-
able from the patient safety viewpoint.

Medical Physics, Vol. 37, No. 7, July 2010

3657 Kim et al.: Kilovoltage cone-beam CT: Dose and image quality evaluation
FIG. 6. CBCT images for the HU uniformity test obtained from 共a兲 CBHQH
共partial-angle scan兲 and 共b兲 CBPEL 共full-angle scan兲 protocols. Note the
crescent effect at the interface of the two different density materials in the
Fig. 6共a兲.
The image qualities of the new CBCT protocols were not
comparable to the old CBCT protocols as expected, e.g.,
low-dose head and low-dose thorax could not pass the accep-
tance criteria of HU verification and contrast resolution tests
and most of the new protocols could not pass the contrast
resolution test except high-quality head and pelvis protocols.
Note that all the old protocols had been passed every test of
the QA protocol which referred from the paper of Yoo et al.13
From our results, we conclude that low x-ray intensity
共mA s兲 and fan type are the main causes of HU deviation;
low-dose head and low-dose thorax protocols failed the HU
verification test which both used the low mA s. However, it
should be mentioned that the standard-dose head protocol
did not fail the HU verification test though its mA s was
lower than that of low-dose thorax protocol. Thus, one would
expect that the HU values could be affected from the com-
bination of both low x-ray statistics and half-fan scan geom-
etry. Due to the HU inaccuracy, CBCT images from low-
dose head and low-dose thorax protocols are inadequate for
the purpose of dose calculations unless corrections are ap-
plied. Systematic inaccuracies can be expected and ac-
Medical Physics, Vol. 37, No. 7, July 2010
3657
FIG. 7. 共a兲 Cross-sectional profiles obtained from standard head scan and 共b兲
from standard body scan. The solid circle, open square, and open circle
denote the profiles acquired at center, middle, and edge location in the CT
phantoms.
counted for. In addition, it should be noticed that only high
mA s protocols such as high-quality head and pelvis proto-
cols passed the contrast resolution test. Thus, when high soft-
tissue contrast images are needed, the use of high-quality
head or pelvis protocol is recommended depending on the
imaging area. It should be noted that the image quality of all
the new scan protocols was still clinically acceptable for the
localization of the treatment target in IGRT.
As seen in Table III, it can be found that high-dose pro-
tocols 共high-quality head, pelvis, and pelvis spotlight兲 pro-
duced smaller standard deviations in HU verification and
uniformity tests; larger photon statistics of higher dose pro-
tocols reduced the noise levels of the CBCT images obtained
from higher dose protocols. It is interesting to note that low-
dose thorax protocol passed the HU uniformity test with
small standard deviation 共3.0兲 though it provided different
HU values 共80 HU兲 compared to other protocols 共⫺3 HU兲.
Thus, every QA test should be performed to accurately
evaluate a certain CBCT protocol. It should be noted, how-
ever, that the main point of the OBI system is positional
accuracy which cannot be captured by simple image charac-
3658 Kim et al.: Kilovoltage cone-beam CT: Dose and image quality evaluation
FIG. 8. Longitudinal profiles acquired from standard head and standard body
scan. Note the dose drops at the edge location.
terization such as HU accuracy or uniformity tests. Further
research is needed to define a task-specific metric of image
quality for the OBI system.
We found that only partial-angle CBCT scan protocols
produce a crescent effect at the interface of two different
density materials and the effect was more significant with
high mA s; high-quality head protocol produce more signifi-
cant crescent effect than low and standard-dose head proto-
cols. However, within the homogeneous region, the effect
was not significant. More through investigation will be nec-
essary for this effect which is beyond the scope of this study.
V. CONCLUSION
We have evaluated the CTDI of new CBCT protocols
based on the point dose and numerical averaging methods
using IC measurements and MC simulations, and results
were compared to old protocols. We found that CTDI of the
new CBCT protocols has been significantly reduced com-
pared to the old protocols. Thus, the use of new CBCT pro-
tocols is recommended in order to reduce the patient dose as
long as the image quality meets the purpose of the CBCT
scan. The point dose calculation method showed the CTDIw
values close to the volume averaging method of MC simula-
tions within 9%–21% for both partial and full-angle CBCT
scan protocols. In addition, the Bakalyar’s method 共CTDIwb兲
was found to provide more accurate dose estimation within
14%. Because dose deviation of ⬃10% may be normal in
diagnostic x ray,11,27,28 the point dose method may be used to
estimate the CTDIw in the clinical CBCT environment. No
significant deviations were found between the CTDIw equa-
tion calculations 关methods 共1兲 and 共2兲兴 and numerical aver-
aging methods 关methods 共3兲 and 共4兲兴 both partial and full-
angle CBCT scan protocols. Thus, the CTDIw equations
关methods 共1兲 and 共2兲兴 may be used to estimate the CTDIw
with reasonable accuracy in the partial-angle CBCT scan
protocols. The HU inaccuracy from low-dose head and low-
dose thorax protocols can render incorrect dose results in the
treatment planning system. When high soft-tissue contrast
Medical Physics, Vol. 37, No. 7, July 2010
3658
data are needed, high-quality head protocol or pelvis scan
protocol is recommended depending on the imaging area.
ACKNOWLEDGMENT
The authors thank Robert E. Reiman of Duke Radiation
Safety Division for the editorial assistance.
a兲
Electronic mail: yoshi003@mc.duke.edu
1
C. Thilmann, S. Nill, T. Tucking, A. Hoss, B. Hesse, L. Dietrich, R.
Bendl, B. Rhein, P. Haring, C. Thieke, U. Oelfke, J. Debus, and P. Huber,
“Correction of patient positioning errors based on in-line cone beam CTs:
Clinical implementation and first experiences,” Radiat. Oncol. 24, 1–16
共2006兲.
2
R. B. Case, J. J. Sonke, D. J. Moseley, J. Kim, K. K. Brock, and L. A.
Dawson, “Inter- and intrafraction variability in liver position in non-
breath-hold stereotactic body radiotherapy,” Int. J. Radiat. Oncol., Biol.,
Phys. 75共1兲, 302–308 共2009兲.
3
F. F. Yin, Z. Wang, S. Yoo, Q. J. Wu, J. Kirkpatrick, N. Larrier, J. Meyer,
C. G. Willett, and L. B. Marks, “Integration of cone-beam CT in stereo-
tactic body radiation therapy,” Technol. Cancer Res. Treat. 7共2兲, 133–139
共2008兲.
4
T. Fatunase, Z. Wang, S. Yoo, J. L. Hubbs, R. G. Prosnitz, F. F. Yin, and
L. B. Marks, “Assessment of the residual error in soft tissue setup in
patients undergoing partial breast irradiation: Results of a prospective
study using cone-beam computed tomography,” Int. J. Radiat. Oncol.,
Biol., Phys. 70共4兲, 1025–1034 共2008兲.
5
M. J. Murphy et al., “The management of imaging dose during image-
guided radiotherapy: Report of the AAPM Task Group 75,” Med. Phys.
34, 4041–4063 共2007兲.
6
M. K. Islam, T. G. Purdie, B. D. Norrlinger, H. Alasti, D. J. Moseley, M.
B. Sharpe, J. H. Siewerdsen, and D. A. Jaffray, “Patient dose from kilo-
voltage cone beam computed tomography imaging in radiation therapy,”
Med. Phys. 33共6兲, 1573–1582 共2006兲.
7
N. Wen, H. Guan, R. Hammoud, D. Pradhan, T. Nurushev, S. Li, and B.
Movsas, “Dose delivered from Varian’s CBCT to patients receiving IMRT
for prostate cancer,” Phys. Med. Biol. 52共8兲, 2267–2276 共2007兲.
8
W. Y. Song, S. Kamath, S. Ozawa, S. Al Ani, A. Chvetsov, N. Bhandare,
J. R. Palta, C. Liu, and J. G. Li, “A dose comparison study between XVI®
and OBI® CBCT systems,” Med. Phys. 35共2兲, 480–486 共2008兲.
9
G. X. Ding and C. W. Coffey, “Radiation dose from kilovoltage cone
beam computed tomography in an image-guided radiotherapy procedure,”
Int. J. Radiat. Oncol., Biol., Phys. 73共2兲, 610–617 共2009兲.
10
Varian Medical Systems, On-Board Imager Advanced Imaging 共1.4兲, Cus-
tomer Release Note, P/N KC 1402 共Varian Medical Systems, April 2008兲.
11
R. L. Dixon, “A new look at CT dose measurement: Beyond CTDI,” Med.
Phys. 30共6兲, 1272–1280 共2003兲.
12
D. W. O. Rogers, B. A. Faddegon, G. X. Ding, C. M. Ma, J. We, and T.
R. Mackie, “BEAM: A Monte Carlo code to simulate radiotherapy treat-
ment units,” Med. Phys. 22共5兲, 503–524 共1995兲.
13
D. W. O. Rogers, B. Walters, and I. Kawrakow, “BEAMnrc user manual,”
NRCC Report No. PIRS-0509共A兲, rev. K 共NRCC, Ottawa, Canada, 2004兲.
14
S. Yoo, G. Kim, R. Hammoud, E. Elder, T. Pawlicki, H. Guan, T. Fox, G.
Luxton, F. Yin, and P. Munro, “A quality assurance program for the on-
board imager®,” Med. Phys. 33共11兲, 4431–4447 共2006兲.
15
S. Kim, T. Yoshizumi, G. Toncheva, S. Yoo, F. Yin, and D. Frush, “Esti-
mation of computed tomography dose index in cone beam computed to-
mography: MOSFET measurements and Monte Carlo simulations,”
Health Phys. 98共5兲, 683–691 共2010兲.
16
S. Kim, T. Yoshizumi, G. Toncheva, F. Yin, and D. Frush, “A Monte
Carlo dose estimation method using a voxelized phantom for pediatric
17
CBCT,” Health Phys. 97共1兲, S56–S57 共2009兲.
E. Mainegra-Hing and I. Kawrakow, “Efficient x-ray tube simulations,”
Med. Phys. 33共8兲, 2683–2690 共2006兲.
18
S. M. Seltzer and M. J. Berger, “Bremsstrahlung spectra from electron
interactions with screened atomic nuclei and orbital electrons,” Nucl. In-
strum. Methods Phys. Res. B 12, 95–134 共1985兲.
19
S. M. Seltzer and M. J. Berger, “Bremsstrahlung energy spectra from
electrons with kinetic energy from 1 keV to 10 GeV incident on screened
nuclei and orbital electrons of neutral atoms with Z = 1 – 100,” At. Data
Nucl. Data Tables 35, 345–418 共1986兲.
3659 Kim et al.: Kilovoltage cone-beam CT: Dose and image quality evaluation
20
M. J. Berger and J. H. Hubbell, “XCOM: Photon cross sections on a
personal computer,” NIST Report No. NBSIR87–3597 共NIST, Gaithers-
burg, MD, 1987兲.
21
W. R. Nelson, H. Hirayama, and D. W. O. Rogers, “The EGS4 code
system,” SLAC Technical Report No. SLAC-265 共Stanford Linear Accel-
erator Center, 1985兲.
22
B. Walters, I. Kawarakow, and D. W. O. Rogers, “DOSXYZnrc users
manual,” NRCC Report No. PIRS-794, rev. B 共NRCC, Ottawa, Canada,
2004兲.
23
P. Downes, R. Jarvis, E. Radu, I. Kawrakow, and E. Spezi, “Monte Carlo
simulation and patient dosimetry for a kilovoltage cone-beam CT unit,”
Med. Phys. 36, 4156–4167 共2009兲.
Medical Physics, Vol. 37, No. 7, July 2010
3659
24
I. Kawrakow, “The dose visualization tool DOSXYZ_show,” NRCC Re-
port No. PIRS-624 共NRCC, Ottawa, Canada, 2006兲.
25
W. Leitz, B. Axelsson, and G. Szendro, “Computed tomography dose
assessment—A practical approach,” Radiat. Prot. Dosim. 57, 377–380
共1995兲.
26
D. M. Bakalyar, “A critical look at the numerical coefficients in CTDI-
VOL,” Med. Phys. 33共6兲, 2003 共2006兲.
27
J. M. Boone, “The trouble with CTDI100,” Med. Phys. 34, 1364–1371
共2007兲.
28
S. Mori, M. Endo, K. Nishizawa, T. Tsunoo, T. Aoyama, H. Fujiwara, and
K. Murase, “Enlarged longitudinal dose profiles in cone-beam CT and the
need for modified dosimetry,” Med. Phys. 32共4兲, 1061–1069 共2005兲.