Download Inclusion of the dose from kilovoltage cone beam CT in the radiation

Inclusion of the dose from kilovoltage cone beam CT in the radiation
therapy treatment plans
Parham Alaeia兲
University of Minnesota, Minneapolis, Minnesota 55455
George Ding
Vanderbilt University, Nashville, Tennessee 37232
Huaiqun Guan
Good Samaritan Health System, Kearney, Nebraska 68847
共Received 18 May 2009; revised 13 November 2009; accepted for publication 15 November 2009;
published 10 December 2009兲
Purpose: Cone beam CT is increasingly being used for daily patient positioning verification during
radiation therapy treatments. The daily use of CBCT could lead to accumulated patient doses higher
than the older technique of weekly portal imaging. There have been several studies focusing on
measurement or calculation of the patient dose from CBCT recently.
Methods: This study investigates the feasibility of configuring a kV x-ray source in a commercial
treatment planning system to calculate the dose to patient resulting from an IGRT procedure. The
method proposed in this article can be used to calculate dose from CBCT imaging procedure and
include that in the patient treatment plans.
Results: The kilovoltage beam generated by the CBCT imager has been modeled using the planning system. The modeled profiles agree with the measured ones to within 5%. The modeled beam
was used to calculate dose to phantom in the pelvic region and the calculations were compared to
TLD measurements. The agreement between calculated and measured doses ranges from 0% to
19% in soft tissue with larger variations observed near and within the bone.
Conclusions: The modeling of the beam produces reasonable results and the dose calculation
comparisons indicate the potential for computing kilovoltage CBCT doses using a treatment planning system. Further improvements in the dose calculation algorithm are necessary, especially for
dose calculations in and near the bone. © 2010 American Association of Physicists in Medicine.
关DOI: 10.1118/1.3271582兴
Key words: kilovoltage cone beam CT, treatment planning
I. INTRODUCTION
Cone beam CT 共CBCT兲, is a relatively new imaging modality commonly installed on linear accelerators 共LINACs兲 and
used for image guided radiation therapy 共IGRT兲. The radiation source used to obtain CT images is either the megavoltage beam produced by the LINAC itself or an additional
kilovoltage 共kV兲 x-ray tube/image receptor installed perpendicularly to the megavoltage beam axis.
In general, daily use of CBCT adds to the radiation dose
given to patients to a greater amount than that given by the
traditional weekly portal imaging. For example, Wen et al.1
measured the cumulative kV CBCT dose in pelvic bones to
be ⬃400 cGy during the treatment of prostate in a total of
42 fractions. Ding et al.2 reported the dose resulting from a
single fraction kV CBCT acquisition being as high as 25 cGy
in cranial bones. The AAPM Task Group 75 report3 addresses the issue of imaging dose from CBCT and makes
recommendations on reducing, as well as estimating, the
dose to patient.
There have been several papers on measuring the dose
from CT in phantom and on patient using various dosimeters
or by Monte Carlo methods.1,2,4–8 Inclusion of this dose in
the treatment planning process is the subject of additional
244
Med. Phys. 37 „1…, January 2010
investigations. In the case of megavoltage CBCT imaging,
since the radiation beam used for imaging is the same as the
one used for treatment planning, the addition of the imaging
dose to the treatment plan is a relatively straightforward process. For example, Miften et al.9 incorporated the daily dose
from megavoltage CT in the IMRT treatment planning optimization.
In the case of kilovoltage CT imaging, however, the inclusion of the imaging dose requires defining the kilovoltage
beam characteristics, adding the beam data, and commissioning the planning system for this purpose. This work is the
first attempt in modeling a kilovoltage CBCT beam in a commercial treatment planning system.
Including the imaging dose at the time of planning would
allow for a better prediction of the total dose to tumor and
critical organs since this dose can be accounted for at the
time of optimization for IMRT planning, or can simply be
added to the treatment plan in the case of conventional planning. For kilovoltage beams, there is also the consideration
of additional dose to bone, which may affect the dose distribution within the patient to a greater degree. Accounting for
the dose to patient resulting from image guidance procedures
in the stage of treatment planning can also provide choices
0094-2405/2010/37„1…/244/5/$30.00
© 2010 Am. Assoc. Phys. Med.
244
245
Alaei, Ding, and Guan: CBCT dose in treatment planning
FIG. 1. The image of half-bowtie filter 共left兲 and the profile of the wedge
mimicking it in Pinnacle 共right兲. Note that the wedge profile orientation is
reverse of the bowtie filter. The profile was entered in this orientation due to
the planning system limitation and all the fields were given an 180° collimator rotation to match the filter orientation. The wedge is assigned a density of 7.8 g/cc.
for clinicians to make an informed decision regarding the
risk and benefits of additional radiation exposure.
This investigation focuses on modeling the kilovoltage
beam from Varian’s on-board imager 共OBI兲 system installed
on the Varian linear accelerators in a commercial treatment
planning system.
II. MATERIALS AND METHODS
II.A. Beam modeling
We have modeled the kV CBCT beam from a Varian OBI
system on a Trilogy linear accelerator 共Varian Medical Systems, Palo Alto, CA兲 in the Philips PINNACLE treatment planning system v8.0 共Philips Medical Systems, Milpitas, CA兲.
The OBI system has been extensively described
elsewhere.1,10 In brief, the system consists of an x-ray tube
capable of producing x-ray beams with the peak energy
range of 40–125 kV and an image receptor at 140–170 cm
from the source with a 100 cm isocenter. The CBCT images
can be acquired in full-fan or half-fan modes. Additional
aluminum filters called the bowtie are used during acquisition to improve image quality. In the full-fan mode, the full
bowtie filter is usually used while in the half-fan mode the
half-bowtie filter is used. This study concentrates on using
the 125 kVp beam and a half-fan beam acquisition with a
half-bowtie, but it can be extended to include other beam
qualities and acquisition modes.
In order to model a beam in the kilovoltage energy range,
the PINNACLE system’s capabilities have been extended with
the addition of monoenergetic energy deposition kernels in
the range of 20–110 keV, details of which have been explained previously.11,12
The x-ray beam modeled here has technical settings of
125 kVp and 80 mA, and acquired at half-fan mode with a
half-bowtie filter. The blade settings are at X1 = 8.3 cm, X2
= 24.9 cm, Y1 = Y2 = 11.8 cm. In order to model the halfbowtie filter, a wedge mimicking the filter has been added to
the beam 共Fig. 1兲. The beam spectrum was generated by
Monte Carlo2,10 and used as the baseline spectrum for modeling.
The beam data used for modeling consist of the depth
dose curve, three profiles parallel to the bowtie 共wedge兲 direction at depths of 1, 5, and 10 cm, and one profile in the
Medical Physics, Vol. 37, No. 1, January 2010
245
perpendicular direction at a depth of 5 cm. The data were
generated using Monte Carlo and verified by
measurement.2,10 The beam outputs for a specific CBCT scan
determined by measured values in phantom were entered in
the system as cGy/MU, with 1 MU being equivalent to 1
min. The choice of time would make it possible to enter
acquisition times for a particular imaging study in lieu of
monitor units which are meaningless in CBCT imaging.
Due to the shape of the cross profiles and the bowtie
共wedge兲 filter, automodeling routines of the planning system
were not usable, thus the beam modeling was performed
manually and through an iterative process. Various factors,
including the beam spectrum, electron contamination parameters, effective source size, and the shape of the filter were
modified in order to obtain an acceptable model, i.e., obtaining the best fit for all profiles with the minimum difference
between measured and modeled values. For example, the
filter shape and density was changed several times and the
one depicted in Fig. 1 is the finally accepted, and not the
initially designed, one.
II.B. Dose calculations
The modeled kV CBCT beam was used to compute the
dose from the CBCT procedure on a Rando body phantom
共The Phantom Laboratory, Salem, NY兲. The goal here was to
calculate the dose in the phantom measured by Wen et al.1 in
which they utilized thermoluminescent dosimeters 共TLDs兲
for dose measurements. So, the image data set acquired in
that study was used to compute the doses and the same blade
settings and acquisition mode was reproduced in the planning system. Depending on treatment planning 共based on
CT兲 or adaptive radiotherapy planning 共ART planning, based
on CBCT兲, either a conventional CT-to-density curve or a
CBCT-to-density curve could be used to correct for inhomogeneities. It is also possible to calibrate kV CBCT so that a
single curve can be used for both. In this study, we calculated
the KV dose based on planning CT and therefore a conventional CT-to-density curve was used.
A typical CBCT image acquisition on OBI system consists of hundreds of projections over an arc of 200° or more.
This particular acquisition consists of 660 projections over a
370° arc. Reproducing this number of beams in a treatment
planning system is time consuming and prohibitive in terms
of system usage. However, these discrete projections can be
estimated in the planning system as arcs. PINNACLE system
computes arcs as a multitude of stationary beams, one every
5° by default. The 5° increment, however, can be changed.
So, comparison was made between using 5° and 1° increments and the resultant dose distributions were virtually
identical, so the 5° increment was used to speed up calculations. In order to reproduce a 370° rotation for the current
study, three arcs were created in PINNACLE. The choice of the
arcs was to reproduce the measurements done by Wen,1 including the overlap on the left side of the phantom and accounting for variable gantry speeds at the start and end positions. The beam arrangements and weightings used are
shown in Table I. This beam arrangement constitutes calcu-
246
Alaei, Ding, and Guan: CBCT dose in treatment planning
TABLE I. The beam arrangement and weighting used for dose calculations,
the angles conform to Varian IEC-1217 scale.
Arc 1
Arc 2
Arc 3
Start angle
Stop angle
Beam weighting
共%兲
94
0
86
359
86
94
65.5
21.5
13.0
lation of 75 stationary beams to mimic the 370° arc rotation.
The calculated doses were then compared to measured values
by examining the point doses at the locations of TLD measurements.
III. RESULTS AND DISCUSSION
III.A. Beam modeling
The resultant modeled depth dose and cross profiles are
shown in Figs. 2共a兲–2共e兲. In these figures, solid lines represent measured profiles as reported by Ding10 and dashed
lines represent computed ones. The half-bowtie filter wedge
is always present in the beam for both modeling and calculations. As seen in the figure, there are reasonable agreements between the modeled and measured profiles. The percent error for depth dose curves as reported by the planning
system for all points except surface is 4% with the surface
point being 5%. The percent error is calculated as
共computed-measured兲/max depth dose. The same quantity
for all cross profiles is better than 3% except for regions of
steep dose drop-off and out-of-the-field areas. This value is
defined as 共computed-measured兲/central axis dose.
FIG. 2. The results of modeling the 125 kVp CBCT beam in Pinnacle treatment planning system and comparison with measured data: 共a兲 Depth dose
curve, 共b兲 X cross profile at 1 cm depth, 共c兲 X cross profile at 5 cm depth, 共d兲
X cross profile at 10 cm depth, and 共e兲 Y cross profile at 5 cm depth. The X
profiles are in the direction of the bowtie filter and the Y profile is in the
perpendicular direction. The field size 共blade setting兲 used for these profiles
are: X1 = 24.9, X2 = 8.3, Y1 = Y2 = 11.8 cm. The solid lines represent measured profiles and dashed lines represent computed ones.
Medical Physics, Vol. 37, No. 1, January 2010
246
TABLE II. Comparisons of measured and computed point doses in soft tissue
areas of Rando phantom. Measured doses are from Ref. 1 and have a maximum uncertainty of 10%. Computed doses are from the beam arrangement
indicated in Table I. Point numbers refer to those in Fig. 3.
Point
Measured dose
共cGy兲
Computed dose
共cGy兲
% difference
1
2
3
4
5
6
7
8
9
11
13
16
17
18
19
22
23
24
4.3
4.0
3.3
3.1
2.9
2.7
3.2
2.7
3.7
2.9
2.1
2.8
2.9
2.8
2.8
4.1
4.5
4.2
3.8
3.8
3.5
3.3
3.0
2.8
2.7
3.1
3.4
2.4
2.4
2.7
2.7
2.8
2.8
3.3
3.7
4.0
⫺11.63
⫺5.00
6.06
6.45
3.45
3.70
⫺15.63
14.81
⫺8.11
⫺17.24
14.29
⫺3.57
⫺6.90
0.00
0.00
⫺19.51
⫺17.78
⫺4.76
III.B. Dose calculations
Dose calculations using the three-arc beam arrangement
共Table I兲 indicate a difference in the range of 0% to 19%
between planned and measured doses for points within the
soft tissue portion of phantom. These results are tabulated in
Table II. Point index numbers correspond to TLD locations
indicated on Fig. 3. The calculated dose around each point
was investigated and there is essentially no dose gradient
within the approximate area occupied by each TLD chip.
One should note that although percentage differences of up
to 19% are observed, the absolute dose differences are in the
order of 0.8 cGy or less. Since the range of doses measured/
FIG. 3. The locations of the TLDs as indicated in the CT image of the Rando
phantom and the isodose distribution generated by calculating the dose from
the 125 kVp CBCT beam using the beam arrangement listed in Table I. This
figure corresponds to Fig. 3a of Ref. 1.
247
Alaei, Ding, and Guan: CBCT dose in treatment planning
TABLE III. Comparisons of measured and computed point doses in and near
the bony areas of Rando phantom. “Measured” bone doses are from Ref. 1
and were actually calculated in that work and have a maximum uncertainty
of 15%. Computed doses are from the beam arrangement indicated in Table
I. Point numbers refer to those in Fig. 3.
Point
Measured dose
共cGy兲
Computed dose
共cGy兲
% difference
10
12
14
15
20
21
4.7
3.2
6.7
6.3
10.2
9.1
3.7
2.5
2.4
2.3
2.9
2.9
⫺21.28
⫺21.88
⫺64.18
⫺63.49
⫺71.57
⫺68.13
calculated is between 2 and 5 cGy, small absolute dose differences translate to large percentage differences.
In the areas near and inside the bone, larger differences,
up to 70%, are observed 共Table III兲. The convolution/
superposition algorithm employed in PINNACLE is based on
the energy deposition kernels generated in water13 and relies
on density scaling theorem14 to scale the dose in materials of
varying densities but the same atomic number. This approach
works well in the megavoltage energy range but does not in
the kilovoltage energies. To examine this further, the plan
was recomputed ignoring inhomogeneity corrections. The resultant uncorrected bone doses were virtually identical 共either the same value or within 0.1 cGy兲.
Assuming the average energy of the 125 kVp beam to be
45 keV, the mass energy absorption coefficient for bone is
about five times that of soft tissue.15 Therefore, it is not
possible to obtain accurate doses near and inside the bone
using this algorithm. In previous work,12,16 the accuracy of
dose calculations in this energy range was evaluated and it
was shown that calculations in high atomic number materials
such as bone suffer from major inaccuracies. A proposed
work-around to account for increased bone absorption using
modified CT-to-density table12 would work for simple geometry but proved inadequate in this study for points near and
inside the bone using humanoid CT data.
III.C. Correction of bone dose
In the absence of more accurate dose calculation algorithms for kilovoltage beams, a manual postprocessing
method could be employed to correct for the dose. Table IV
TABLE IV. Dose to points in bone corrected by the ratios of bone/water mass
energy absorption coefficients in Table X of TG-61 共Ref. 17兲, based on a
HVL of 5.4 mm for 125 kVp beam with bowtie filter.
247
is the result of such postprocessing. The “corrected dose”
values in the table are obtained by multiplying the PINNACLE
“computed dose” values by the ratio of bone/water mass energy absorption coefficients obtained from Table X of AAPM
Task Group 61 report,17 assuming a 5.4 mm HVL for the 125
kVp beam measured before.10 As seen in the table, this results in improved agreement between measured and calculated values, although there is still a difference of up to 30%.
It should be noted that the “measured” bone dose values
were not in fact measured but rather calculated,1 hence there
are large uncertainties 共up to 15%兲 associated with them.
It should also be noted that this postprocessing method
assumes a constant energy for the beam even though the
energy of the kilovoltage beam changes with depth. In addition the phantom thickness varies as the beam rotates around
it so the beam energy at any one point depends on the thickness it traverses the medium from all directions and is not a
constant value. Further improvement in bone dose calculations requires a new algorithm such as the one proposed by
Ding.18
IV. CONCLUSIONS AND DISCUSSIONS
This study demonstrates the feasibility of modeling a kV
x-ray beam used for IGRT in a commercial treatment planning system. The kV beam was successfully modeled and its
accuracy was evaluated. The results obtained from modeling
the kV CBCT beam and calculating dose from this imaging
modality in the treatment planning can provide for an easy
estimation of imaging dose for all patients as part of their
treatment plans. This will lead to better accounting of imaging dose prior to the commencement of treatment which will
prevent potential overdosing of sensitive organs.
More work is needed to improve the accuracy of dose
calculation by better accounting for variable gantry speeds
and different filters used. Potential improvements in the
modeling could reduce the uncertainties in dose calculation
in soft tissue. The planning system’s ability to predict dose to
lung will also need to be evaluated further. But the major
limitation of the system remains in its inability of predicting
the dose in and near bony structures. This limitation can only
be overcome by introducing newer algorithms capable of
accounting for atomic number changes.
ACKNOWLEDGMENTS
The authors wish to thank Dr. T.R. Mackie for discussions
involving the convolution/superposition algorithm. The authors also wish to thank Philips Medical Systems for providing equipment support for this project.
a兲
Electronic mail: [email protected]
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, 2267–2276 共2007兲.
2
G. X. Ding, D. M. Duggan, and C. W. Coffey, “Accurate patient dosimetry of kilovoltage cone-beam CT in radiation therapy,” Med. Phys. 35,
1135–1144 共2008兲.
3
M. J. Murphy, J. Balter, S. Balter, I. J. Das, S. B. Jiang, C.-M. Ma, G. H.
Olivera, R. F. Rodebaugh, K. J. Ruchala, H. Shirato, and F. F. Yin, “The
management of imaging dose during image-guided radiotherapy: Report
1
Point
Computed dose
共cGy兲
Corrected dose
共cGy兲
% difference 共after correction兲
14
15
20
21
2.4
2.3
2.9
2.9
8.47
8.11
11.42
11.05
26.47
28.65
11.97
21.46
Medical Physics, Vol. 37, No. 1, January 2010
248
Alaei, Ding, and Guan: CBCT dose in treatment planning
of the AAPM Task Group 75,” Med. Phys. 34, 4041–4063 共2007兲.
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 kilovoltage cone beam computed tomography imaging in radiation therapy,”
Med. Phys. 33, 1573–1582 共2006兲.
5
O. Gayou, D. S. Parda, M. Johnson, and M. Miften, “Patient dose and
image quality from mega-voltage cone beam computed tomography imaging,” Med. Phys. 34, 499–506 共2007兲.
6
A. Amer, T. Marchant, J. Sykes, J. Czajka, and C. Moore, “Imaging doses
from the Elekta Synergy x-ray cone beam CT,” Br. J. Radiol. 80, 476–
482 共2007兲.
7
M. W. K. Kan, L. H. T. Leung, W. Wong, and N. Lam, “Radiation dose
from cone beam computed tomography for image-guided radiation
therapy,” Int. J. Radiat. Oncol., Biol., Phys. 70, 272–279 共2008兲.
8
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, 610–617 共2009兲.
9
M. Miften, O. Gayou, B. Reitz, and R. Fuhrer, “IMRT planning and
delivery incorporating daily dose from mega-voltage cone-beam computed tomography imaging,” Med. Phys. 34, 3760–3767 共2007兲.
10
G. X. Ding, D. M. Duggan, and C. W. Coffey, “Characteristics of kilovoltage x-ray beams for cone-beam computed tomography in radiation
therapy,” Phys. Med. Biol. 52, 1595–1615 共2007兲.
4
Medical Physics, Vol. 37, No. 1, January 2010
248
11
P. Alaei, B. J. Gerbi, and R. A. Geise, “Generation and use of photon
energy deposition kernels for diagnostic quality x rays,” Med. Phys. 26,
1687–1697 共1999兲.
12
P. Alaei, B. J. Gerbi, and R. A. Geise, “Evaluation of a model-based
treatment planning system for dose computations in the kilovoltage energy range,” Med. Phys. 27, 2821–2826 共2000兲.
13
T. R. Mackie, A. F. Bielajew, D. W. O. Rogers, and J. J. Battista, “Generation of photon energy deposition kernels using the EGS Monte Carlo
code,” Phys. Med. Biol. 33, 1–20 共1988兲.
14
J. E. O’Connor, “The variation of scattered x-rays with density in an
irradiated body,” Phys. Med. Biol. 1, 352–369 共1957兲.
15
J. H. Hubbell, “Photon mass attenuation and energy-absorption coefficients from 1 keV to 20 MeV,” Int. J. Appl. Radiat. Isot. 33, 1269–1290
共1982兲.
16
P. Alaei, B. J. Gerbi, and R. A. Geise, “Lung dose calculations at kilovoltage x-ray energies using a model-based treatment planning system,”
Med. Phys. 28, 194–198 共2001兲.
17
C.-M. Ma, C. W. Coffey, L. A. DeWerd, C. Liu, R. Nath, S. M. Seltzer,
and J. P. Seuntjens, “AAPM protocol for 40–300 kV x-ray beam dosimetry in radiotherapy and radiobiology,” Med. Phys. 28, 868–893 共2001兲.
18
G. X. Ding, J. M. Pawlowski, and C. W. Coffey, “A correction-based dose
calculation algorithm for kilovoltage x rays,” Med. Phys. 35, 5312–5316
共2008兲.