Dose-guided radiation therapy with megavoltage cone-beam CT
J CHEN, PhD, O MORIN, BSc, M AUBIN, Eng-MSc, M K BUCCI, MD, C F CHUANG, PhD and J POULIOT, PhD
UCSF Comprehensive Cancer Center, Department of Radiation Oncology, University of California
San Francisco, 1600 Divisadero Street, Suite H1031, San Francisco, CA 94143, USA
ABSTRACT. Recent advances in fractionated external beam radiation therapy have
increased our ability to deliver radiation doses that conform more tightly to the tumour
volume. The steeper dose gradients delivered in these treatments make it increasingly
important to set precisely the positions of the patient and the internal organs. For this
reason, considerable research now focuses on methods using three-dimensional images
of the patient on the treatment table to adapt either the patient position or the
treatment plan, to account for variable organ locations. In this article, we briefly review
the different adaptive methods being explored and discuss a proposed dose-guided
radiation therapy strategy that adapts the treatment for future fractions to
compensate for dosimetric errors from past fractions. The main component of this
strategy is a procedure to reconstruct the dose delivered to the patient based on
treatment-time portal images and pre-treatment megavoltage cone-beam computed
tomography (MV CBCT) images of the patient. We describe the work to date performed
to develop our dose reconstruction procedure, including the implementation of a MV
CBCT system for clinical use, experiments performed to calibrate MV CBCT for electron
density and to use the calibrated MV CBCT for dose calculations, and the dosimetric
calibration of the portal imager. We also present an example of a reconstructed patient
dose using a preliminary reconstruction program and discuss the technical challenges
that remain to full implementation of dose reconstruction and dose-guided therapy.
Received 30 June 2005
Revised 8 August 2005
Accepted 7 September
’ 2006 The British Institute of
The rationale for adaptive radiation therapy
and dose-guided radiation therapy
Recent advances in fractionated external beam radia-
tion therapy, such as three-dimensional conformal and
intensity-modulated radiation therapy (IMRT), have
increased our ability to deliver radiation doses that
conform more tightly to the tumour volume. Clinical
studies and simulations indicate that these more con-
formal, higher dose treatments can decrease both the
spread of disease and normal tissue complications [1–5].
Increasing use of functional imaging will also motivate
further complexity in radiation treatment plans to
include concurrent boosts in regions of high cancerous
growth [6, 7]. As these dose distributions conform more
tightly to the patient anatomy, dose gradients necessarily
become steeper inside the irradiated volume. Using
IMRT, a dose gradient of 10% mm21can be achieved
easily. Thus, it is increasingly important to set precisely
the positions of the patient and the internal organs.
Currently, external markers and patient immobilizing
masks and casts are used to reproduce the skeletal
position of the patient with about 3 mm accuracy over
several weeks of treatment . However, the effective-
ness of these alignment and immobilization techniques
are limited by changes in the internal organ locations
relative to bony and external markers. For example, the
prostate can shift up to 1 cm relative to the pelvic bones
due to variations in rectal/bladder filling. During the
course of head and neck cancer treatment, the tumour
can shrink and the patient can lose significant weight,
resulting in dosimetric errors as large as 40% [9, 10]. For
this reason, imaging tools in the treatment room and
methods of adapting treatments to match the patient
anatomy on the treatment table are the keys to realising
the full benefit of conformal therapy.
For many decades, imaging inside the treatment room
has played a role in verifying radiation therapy
treatment. Portal images, projection images of the patient
using the treatment aperture, are used to confirm the
patient position and verify coverage of the tumour. The
use of radiographic film for portal imaging has limited
the frequency of this verification due to the required time
and dose to the patient. However, recent implementation
of electronic portal imaging devices (EPIDs) allows a
digital image to be acquired in a few seconds with low
doses. This has allowed the use of daily portal imaging to
visualize and adjust the patient position before each
treatment. For example, using implanted gold markers to
locate the prostate, daily portal imaging has been used to
position the prostate with 1–2 mm accuracy [11–13]. The
use of portal imaging to adjust patient position before
treatment is limited, however, because soft tissue cannot
be visualized without implanted markers and the full
three-dimensional (3D) geometry is obscured by the
projection onto a two-dimensional (2D) plane. Therefore,
considerable research now focuses on developing three-
dimensional imaging of the patient on the treatment
table. Several systems have been developed including
(1) a ‘‘CT on rails’’ system, requiring an additional
diagnostic CT machine in the treatment room ; (2) a
kilovoltage cone-beam CT (kV CBCT) system, consisting
This research was supported by Siemens Oncology Care Systems.
The British Journal of Radiology, 79 (2006), S87–S98
The British Journal of Radiology, Special Issue 2006 S87
of an additional kV X-ray source and detector attached to
the treatment gantry [15,16] (these systems are described
more fully in this issue in papers by Thieke et al and
Moore et al, respectively); (3) a megavoltage cone-beam
CT (MV CBCT) system using the pre-existing treatment
machine and EPID for imaging [17–19]; (4) a MV CT
system, using the pre-existing treatment machine with
an attached arc of detectors ; and (5) a tomotherapy
system, replacing the traditional treatment machine
(beam) with a CT ring and a MV beam source [21–23].
These imaging systems continue to improve and recent
results indicate that 1–2% soft-tissue contrast resolution
is possible [15, 17, 18, 21] as well as accurate localization
of various tumours [14, 16, 19, 20, 22, 23].
In the above examples of image-guided radiation
therapy (IGRT), treatment room imaging modalities are
used to translate and rotate the patient to better match
the patient position used for treatment planning.
Another potentially more powerful use of these images
is to modify the delivered treatment fields to account for
the variable patient position. This type of adaptive
radiation therapy could adjust for the changing relative
positions of the internal organs and the changing shape
of the organs. This is particularly important for organs
that move significantly during the course of treatment.
For these sites, techniques under current development
include gated treatments (halting irradiation when the
target is out of a certain acceptable region) [24–27] or
target trackingduring irradiation
designed mobile linear accelerators [28, 29]. For some
sites, however, the most important anatomical changes
occur between treatment fractions. In this case, a pre-
treatment image may be used to adjust the treatment
fields immediately before irradiation [30, 31]. Another
possibility is to determine patient-specific anatomical
variation using images from the first week of treatment
and to tailor the treatment plan for future fractions to
account for the individual’s variation [32–34]. Finally, if
the dose that was delivered in previous fractions can be
estimated, the treatment plan for future fractions may be
re-optimized to compensate for dosimetric errors .
This dose-guided therapy could correct for both errors
due to patient anatomical changes as well as machine
delivery errors, thus providing the most accurate dose
schemes are depicted in Figure 1.
The development of dosimetric verification
Currently, few methods are used to track the dose
delivered during treatment. Standard techniques involve
measuring doses on the patient surface using diodes or
thermoluminescent dosemeters. However, these techni-
ques provide only point dose measurements, and the
time and effort to place the dosemeters on the patient
and processthe data
Consequently, few institutions use these methods reg-
ularly for treatment verification. A new implantable
MOSFET dosemeter has also been developed . This
dosemeter directly measures the dose in critical internal
structures, but again provides only a point measurement
and is an invasive technique with limited application.
What is needed to verify conformal therapies is an
automated method to reconstruct the full 3D dose
Several researchers have suggested methods to recon-
struct the delivered patient dose during treatment. Most
methods propose using on-board EPIDs to quickly and
easily acquire a two-dimensional array of digitized X-ray
measurements in a precisely positioned plane in the
treatment exit beam. A few formulae have been derived
to estimate the dose to the exit surface, midplane, or centre
point of the patient based solely on EPID measurements
[37–40]. To find a 3D patient dose distribution, however,
requires additional information about the patient position
and attenuation of the beam. For breast treatments, a
simple patient contour may give sufficient information
. However, in general, information on tissue inhomo-
geneity is also necessary. Several years ago, it was
suggested that the planning CT could be used for this
purpose [42, 43], but this method would fail to detect
dosimetric errors produced by the variable patient and
organ positions and shapes. The 3D imaging modalities
that are being developed for IGRT provide an obvious
for reconstructing dose. Currently, there is active devel-
opment of dose reconstruction procedures for tomother-
apy systems, and 3% accuracy in low-gradient regions has
been demonstrated . A pilot study using MV CBCT on
a traditional treatment machine also found good relative
agreement with measurements, but a systematic absolute
limit their clinicaluse.
Figure 1. A general view of adap-
tive radiation therapy. The large
grey arrow represents the conven-
tional flow of treatment, and the
small arrows indicate the possible
points of feedback into the process.
J Chen, O Morin, M Aubin et al
S88The British Journal of Radiology, Special Issue 2006
Dose-guided radiation therapy using MV CBCT
and treatment-time portal images
In 2003 , we began developing a procedure to
reconstruct the dose delivered to the patient based on
treatment-time portal images and pre-treatment MV
CBCT. Our procedure follows the steps described below
and depicted in Figure 2.
Step 1A: Prior to treatment, with the patient in the
treatment setup position, acquire a MV CBCT image.
This image can be used to align the patient as closely as
possible to the planned position and also provides the
photon attenuation information necessary to reconstruct
the delivered dose.
Step 1B: Convert the MV CBCT image to effective
photon attenuation coefficient. Generally, this can be
accomplished by calibrating the MV CBCT system using
a calibration phantom composed of materials with
known electron densities. However, imaging artefacts
in the MV CBCT image may need to be corrected to
improve the calibration accuracy.
Step 2A: During the treatment, acquire portal images
of the treatment beam as it exits the patient. This portal
image is acquired using the same EPID used for the
Step 2B: Convert the portal images to a 2D map of
treatment beam energy fluence. The acquired portal image
signal is a convolution of the energy fluence incident on
the detector with the detector response to radiation.
Moreover, the energy fluence consists of both the primary
beam and radiation scattered from the patient. To use the
portal image for dose calculations, the primary energy
fluence must be derived from the portal image.
Step 3: Back-project the energy fluence measured at the
for the 1/r2falloff of radiation from a point source and
attenuation through the patient. This calculation is easily
accomplished if the position of the detector plane relative
to the patient and source is accurately known.
Step 4: Calculate the 3D dose distribution delivered to
the patient using a dose calculation engine. This type of
Figure 2. Overview of proposed
dose reconstruction procedure using
MV CBCT imaging and treatment-
time portal imaging.
Dose-guided radiation therapy with MV CBCT
The British Journal of Radiology, Special Issue 2006S89
dose calculation is the same as that performed for
treatment planning purposes, and all the techniques that
have been developed for treatment planning may be
The reconstruction procedure described above pro-
vides an estimate of the 3D dose distribution deposited
in the patient as represented by the MV CBCT. Several
uses of the reconstructed dose distribution to guide
future treatments can be envisaged. Scenario 1: The most
basic use of the reconstructed dose is to provide a
dosimetric verification that the treatment delivery gen-
erally provides the desired dose distribution and that no
gross errors exist. This verification could be performed
during the first treatment and repeated weekly through-
out treatment. This simple approach would effectively
reduce gross dosimetric errors, but would not otherwise
increase the precision of the delivered dose. Scenario 2: If
the patient dose is reconstructed for the first week of
treatment, the variation in the delivered dose may also
be evaluated. If the MV CBCT for each treatment is
contoured to delineate the various important structures,
the variation in dosimetric indices, such as the maximum
dose to sensitive normal structures or the dose to 95% of
the tumour volume, can be calculated. General systema-
tic trends such as the under or over dosing of particular
extremities of a structure may also be detected by
examining the dose distributions over the first week.
Based on this information, the treatment plan can be
modified, for example, to increase or decrease margins of
the tumour in particular directions. In this manner, the
treatment plan can be tailored to each individual patient.
Scenario 3: Finally, a complete dose-guided therapy
system would be able to integrate the dose over previous
fractions. This would require the ability to deform the
daily MV CBCT images to map identical points in
the patient before the integral dose is calculated . The
cumulative dose distribution can be used to adjust
the treatment plan to compensate for deviations from
the desired distribution, thus improving the accuracy
and conformality of the overall treatment.
The dose reconstruction procedure and the dose-
guided therapy described above continue to be devel-
oped and researched. This article summarizes the work
to date and comments on the remaining challenges. First,
we present a description of a MV CBCT system that has
been implemented on a linear accelerator for clinical use.
We then describe experiments performed to calibrate the
MV CBCT for electron density and to use the calibrated
MV CBCT for dose calculations. We also briefly describe
the dosimetric calibration of an EPID for dose recon-
struction. Finally, we present an example of a recon-
structed patient dose using a preliminary reconstruction
program and discuss the technical challenges that
remain to full implementation of dose reconstruction
and dose-guided therapy.
MV cone-beam CT imaging
MV cone-beam CT imaging is a 3D reconstruction
procedure similar to conventional CT. A series of
projection measurements, in this case 2D portal images,
are acquired at many angles around the patient. The
image reconstructed is a 3D image without slice artefacts.
In the radiation oncology context, the imaging beam is
produced by the conventional linear accelerator used for
treatment, and the projection images are detected using
on-board EPIDs. The imaging photons, therefore, are
primarily in the mega-electron volt energy range. In this
configuration, the patient can be positioned once on the
treatment table and need not be repositioned between
imaging and treatment.
As the linear accelerator gantry and the EPID rotate
about the patient, the EPID and beam source positions
will shift from their ideal isocentric locations due to
sagging of the mechanical supports. To correct for this
effect, we perform a geometric calibration of the system,
illustrated in Figure 3 [48, 49]. This calibration provides a
unique relationship between the position of a voxel in
the reconstruction volume and a pixel on the detector
plane for each angle. Because the EPID used for imaging
is also used to detect the exit beam fluence, the same
calibration information can be employed during the dose
reconstruction procedure to back-project the energy
Figure 3. Depiction of the geometric calibration of the
linear accelerator/electronic portal imaging device (EPID)
system for cone beam CT (CBCT) imaging and for dose
reconstruction. The result of the calibration is a set of
projection matrices (P) that map a point in space (RXYZ) to the
projected point on the detector plane (Ruv).
J Chen, O Morin, M Aubin et al
S90 The British Journal of Radiology, Special Issue 2006
fluence through the MV CBCT volume. This prevents
any possibility of misregistration between the EPID
measurements and the MV CBCT volume.
The MV CBCT system installed in our clinic has been
previously described . Briefly, it consists of an
amorphous-silicon flat panel EPID integrated with a
clinical linear accelerator. The total exposure of the CBCT
acquisition can be varied from 1 to 60 monitor units. Upon
patient selection, a reference CT is automatically loaded
intothe software.Thelinearacceleratorgantry thenrotates
in a continuous 200˚arc acquiring images at 1˚increments.
This acquisition procedure lasts about 45 s. The image
reconstruction starts immediately after the acquisition of
the first portal image, and a 25662566256 reconstruction
volume is completed in 110 s. The software automatically
registers the MV CBCT with the reference CT and
calculates table shifts for patient alignment.
To date, 38 patient MV CBCT images have been
acquired in our clinic. All patients have given informed
consent, and the patient image acquisitions are per-
formed in accordance with the institutional review
board’s ethical standards. Depending on the frequency
of the acquisitions, the dose used for MV CBCT ranges
from approximately 1.5 cGy to 12 cGy delivered at the
point of rotation (the isocentre). The dose at the entrance
surface of the arc reaches about 160% of the isocentre
dose for an imaged pelvis and 133% for the head and
neck region. The dose at the exit surface falls to about
66% of the isocentre dose for a pelvis and 55% for the
head and neck region. Figure 4 presents four MV CBCT
images acquired weekly on the same patient to study
tumour evolution. At each new acquisition, the dose was
lowered. The last CBCT of the series was acquired with
approximately 2.9 cGy delivered at the isocentre, still
presenting enough soft-tissue information to assess the
tumour size and perform patient alignment.
Three-dimensional imaging of the patient in the
treatment position exposes the difficulties created by
distortion of patient anatomy. Figure 5 displays the
fusion of a MV CBCT image (grey) with the planning
CT (colour). In this case, a physician has manually
registered the two sets of images by aligning the base of
the skull. A considerable shift, up to 6 mm, can be
observed in the positions of the spinal cord between the
two image sets. This misplacement of the spinal cord
could not be corrected by translating or rotating the MV
CBCT image relative to the CT as it was caused by an
increase in the arching of the patient’s neck. Although
several fractions would be needed to assess if this
misplacement occurs regularly, the new anatomy, as
depicted by the MV CBCT image, could be used to
study the dosimetric impact of the patient’s anatomical
MV CBCT calibration for dose calculation
To use the MV CBCT image in a dose reconstruction
program, the signal from each voxel must be converted
to effective photon attenuation coefficient for the beam
spectrum (Step 1B of our dose reconstruction procedure).
To perform this conversion, the MV CBCT system can be
calibrated using a CT calibration phantom (CIRS Model
062, Norfolk, VA) with tissue-equivalent inserts, as is
currently done with kV CT. A table is formed mapping
CT signal intensity to electron or physical density which
can then be converted to photon attenuation coefficient
for a known beam spectrum. Figure 6 shows the results
of performing this simple calibration on our MV CBCT
system using the following inserts of relative electron
density with respect to water: lung inhale (0.190), lung
exhale (0.489), adipose (0.952), breast (0.976), water (1),
muscle (1.043), liver (1.052), trabecular bone (1.117) and
dense bone (1.512). The relationship between MV CBCT
signal and electron density is linear. These results are
similar to previous work with MV fan-beam CT
performed on a tomotherapy unit at 6 MV .
Although the above calibration works well for the
narrow CT calibration phantom, the MV CBCT images of
extended objects exhibit cupping artefacts due to the
influence of scattered radiation reaching the EPID.
Figure 7 illustrates this cupping effect on the MV CBCT
of a large cylinder of water. If uncorrected, this cupping
artefact will also appear in the image converted to
photon attenuation coefficient, leading to errors in the
calculated dose. However, a simulation study using the
large cylinder of water pictured in Figure 7 indicates that
the dosimetric errors in a homogeneous medium
produced by such severe cupping artefacts remain
relatively small, approximately 4% for a single open
field . This suggests that a crude correction of the
cupping artefact in MV CBCT images may be sufficient
to obtain acceptable dosimetric accuracy. To test this
hypothesis, the MV CBCT of a water cylinder was used
to model the spatial dependence of the cupping artefact.
A spatially dependent correction function was derived
from this cupping model. This correction function was
then applied to the MV CBCT of an anthropomorphic
head phantom as a rough correction for the cupping
artefact in the image. After conversion to density using
the MV CBCT calibration curve, this image was imported
into a commercial treatment planning system (Philips
Pinnacle, Bothell, WA). The dose calculated using the
MV CBCT compared well with the dose calculated using
a kV CT of the same phantom. Using a gamma index
comparison with a 3% dose and 3 mm distance-to-
agreement criterion, 98% of calculated dose points fell
within the acceptance criteria.
The above example demonstrates the potential of
using MV CBCT images for dose calculations. Besides
using these images for dose reconstruction, using patient
MV CBCT images in the treatment planning system, as
performed on the head phantom described above, would
also provide a useful verification. The MV CBCT
provides a more accurate representation of the patient
on the treatment table. Applying the treatment plan to
the MV CBCT would provide a first estimate of the dose
delivered to the patient during treatment. The effects of
modified patient position or anatomy could be evalu-
ated. However, the beam delivery itself could not be
verified without a full dose reconstruction based on
measurements of the treatment beam.
Calibration of EPIDs for exit-plane dose
Besides the patient photon attenuation data, the other
necessary piece of information for dose reconstruction is
Dose-guided radiation therapy with MV CBCT
The British Journal of Radiology, Special Issue 2006S91
the treatment beam energy fluence derived from the
treatment-time portal images (Step 2 of our dose
reconstruction procedure). An intermediate step to
determining the energy fluence is to convert the EPID
image to a measurable form of dose, in our case the dose
in water measured in the detector plane and at a depth of
1.5 cm . The advantage of first calibrating the EPID
against dose in water is that it can be accomplished by
experiments since the dose in a water phantom is easily
measured. The calibration can then be validated by
measurements as well. Moreover, the dose in water can
be more easily converted to energy fluence due to the
great number of water dose deposition models and
algorithms that have already been developed.
To translate the EPID signal to dose in water, we
employ convolution models of dose deposition. The
lateral spread of the dose in the EPID and in the water is
described by empirically derived kernels. Because the
EPID consists of millions of individual pixels, the dose
deposited in each pixel is also multiplied by a spatially
dependent sensitivity factor that accounts for inhomo-
geneity in the detector response. Finally, comparisons of
Figure 4. Examples of megavoltage cone beam CT (MV CBCT) images at different exposure levels, from 2.9 cGy to 10 cGy.
J Chen, O Morin, M Aubin et al
S92 The British Journal of Radiology, Special Issue 2006
EPID and ion chamber measurements are used to form
conversion tables that translate between the EPID signal
and dose in water.
To test the calibration procedure, EPID images of the
exit beam were acquired through a Rando anthropo-
morphic head phantom (The Phantom Laboratory,
Salem, NY). The calibrated EPID images were compared
(Scanditronix-Wellho ¨fer CC13, Bartlett, TN) scanned in
a water tank (Scanditronix-Wellho ¨fer blue phantom,
using an ion chamber
Bartlett, TN). Figure 8 shows a comparison between the
measured dose at a depth of 1.5 cm of water and the
calibrated EPID signal for a 10 cm square open field.
The EPID signal matches the measured dose to within
2% (2 standard deviations) for the in-field regions
(excluding the penumbra).
A dose reconstruction program
Utilizing some of the work described above, we
performed a preliminary version of the dose reconstruc-
tion procedure on the treatment of a head and neck
patient in our clinic. A MV CBCT image was acquired of
the patient set up on the table as for treatment (Step 1A).
The same day, portal images were acquired (Step 2A)
during the patient’s normal course of treatment (6 MV
beam, 2 opposed lateral wedged fields and an anterior–
inferior oblique open field). To utilize the MV CBCT
image in the dose reconstruction program, it must first
be converted to effective photon attenuation coefficient
(Step 1B). For this test case, the MV CBCT was converted
to attenuation coefficient using a spatially dependent
calibration that utilizes the kV CT patient image as a
reference. This allowed us to reduce the effects of the MV
CBCT calibration on the reconstructed dose, thus high-
lighting the dosimetric impact of the remaining steps of
To convert the portal images to energy fluence (Step
2B), the portal images were first converted to equivalent
dose in water using the calibration procedure described
above. To infer the energy fluence at the detector plane
from the equivalent dose in water, we used an in-house
dose calculation program that predicts the dose at a
depth of 1.5 cm of water given the energy fluence at the
water surface. This energy fluence is then iteratively
corrected until the predicted dose matches the measured
dose. To calculate the dose in water, we used convolu-
tion kernels published in the literature , derived
using Monte Carlo calculations and assuming a 6 MV
spectrum. The energy fluence that is derived using this
method is composed of both primary beam as well as
radiation scattered from the patient. For this study, the
contribution of the scattered radiation was neglected.
The two remaining steps to the dose reconstruction
process are (Step 3) the back-projection of the energy
fluence measured at the detector plane through the
CBCT of the patient and (Step 4) the calculation of the 3D
dose distribution delivered to the patient using a dose
calculation engine. To perform the back-projection, we
utilized the geometric information obtained during
calibration of the MV CBCT imaging system (depicted
in Figure 3). The geometric calibration of the system
yields a set of projection matrices that map a point in
space to a pixel in the detector plane. The projection
matrix for each angle accurately accounts for all
geometric factors such as sag in the detector or gantry,
detector rotation, or variation in the detector to source
distance. These projection matrices were used to back-
project the energy fluence from the detector plane
through the CBCT volume while correcting for 1/r2
fall-off and the attenuation of each intersected voxel.
The final step of the reconstruction procedure is to
calculate the dose deposited in the patient from the
Figure 5. Registration of a patient megavoltage cone beam
CT (MV CBCT) (grey) with the kV CT (colour) used for
treatment planning. A large difference in the arching of the
neck causes a considerable deviation in the spinal cord
Figure 6. Megavoltage cone beam CT (MV CBCT) intensity as
a function of electron density for tissue-equivalent inserts in
a CT calibration phantom (pictured in above left).
Dose-guided radiation therapy with MV CBCT
The British Journal of Radiology, Special Issue 2006S93
energy fluence and the attenuation coefficient for each
voxel. The total energy released in each voxel that
interacts with the beam is proportional to the energy
fluence multiplied by the attenuation coefficient. The
spatial distribution of the deposited energy can then be
described using a kernel. The kernels we used for this
Figure 7. Radial (top row) and axial (bottom row) profiles through the megavoltage cone beam CT (MV CBCT) images of a large
cylinder filled with water. The unmodified CBCT (left) exhibits a large cupping artefact as a result of scattered radiation reaching
the electronic portal imaging device (EPID). Using a simple 3D cupping model effectively reduces the artefact (right). The radial
and axial slices of the MV CBCT images (insets) are displayed using the same windowing level.
Figure 8. Comparisons of measured dose profiles (line) in water and calibrated electronic portal imaging device (EPID) profiles
(circle with dot) for a 10 cm square field through a Rando head phantom.
J Chen, O Morin, M Aubin et al
S94 The British Journal of Radiology, Special Issue 2006
purpose were the same kernels used to determine the
energy fluence at the detector plane from the equivalent
dose in water. The application of the kernels to calculate
the dose was performed using in-house software utiliz-
ing the collapsed-cone superposition method . In this
method, the energy deposition calculation is only
performed along a set of rays emanating from each
Figure 9 shows the comparison between the planned
dose distribution found using the patient kV CT image
and a commercial treatment planning system (Philips
Pinnacle, Bothell, WA) and the reconstructed dose
distributions found using the MV CBCT, the treatment-
time portal images, and the in-house dose reconstruction
program. There are some qualitative similarities, but also
some marked differences. The reconstructed dose dis-
tribution appears to be approximately 10% higher than
the dose predicted by the planning system. It is likely
that this is in part due to an increase in the portal image
signal from the scattered radiation that was not corrected
in this preliminary version of the dose reconstruction.
There also appears to be a slight difference in the
alignment of the beams detected by the portal images.
The doses from the treatment planning system suggest a
slight gap between the opposed lateral fields and the
anterior field. In contrast, the reconstructed dose
distribution has a high dose band at the intersection of
the fields. Without further verification, it is not clear
whether this slight difference in field alignment was a
real event detected using the treatment-time portal
images. Other possible causes for the differences in the
two dose distributions include differences in the dose
calculation engines, differences in patient position or
anatomy in the two images, as well as persistent cupping
artefacts in the MV CBCT.
As the above example demonstrates, much research
remains to be done to increase the dosimetric accuracy of
our dose reconstruction program. Currently, we continue
to work toward simple but effective techniques to reduce
cupping artefacts in the MV CBCT images and to
calibrate the MV CBCT for photon attenuation coeffi-
cient. We also continue to refine our EPID dosimetric
Figure 9. Comparisons between planned isodose contours calculated using the patient kV CT image and a commercial
treatment planning system (left) and reconstructed isodose contours calculated using the megavoltage cone beam CT (MV
CBCT), the treatment-time portal images, and an in-house dose reconstruction program (right).
Dose-guided radiation therapy with MV CBCT
The British Journal of Radiology, Special Issue 2006S95
calibration models described above and to improve the
conversion of the EPID signal to primary energy fluence.
One of the remaining challenges is to implement a
correction for the scatter contribution in the portal
images. Portal image scatter correction has been inves-
tigated by other researchers, and some good results have
been reported using a scatter-to-primary ratio model and
Monte Carlo-based scatter kernels [54–56]. Finally, once
the individual steps of the dose reconstruction procedure
have been optimized, the dosimetric accuracy of the full
procedure will need to be determined using dose
measurements in phantoms. As discussed below, the
dosimetric accuracy achieved will affect the clinical
application of the dose reconstruction procedure.
Future directions in dose-guided therapy
This article has summarized the work performed as
well as the challenges remaining to develop a dose
reconstruction procedure based on MV CBCT images of
the patient on the treatment table and treatment-time
portal images. As described earlier, the ability to
reconstruct the delivered patient dose opens up the
possibility of adapting the patient treatment plan to
improve dose delivery. The accuracy of the dose
reconstruction procedure and the availability of image
processing tools will affect how treatment may be guided
using this new dose information. Our initial goal is to
achieve 5% accuracy for the reconstructed patient dose.
With this level of accuracy, gross dosimetric errors,
which have been demonstrated to be as high as 40% in
cases of considerable patient weight loss , could be
detected and corrected. Implementation of more com-
plex dose-guidance strategies, such as scenarios 2 and 3
discussed earlier, will require increased dosimetric
accuracy as well as the ability to precisely locate the
dose distribution in terms of critical structures. It is here
that the rapidly advancing field of 3D image processing
will play a key role. Tools such as automated segmenta-
tion and 3D deformable registration increase our ability
to determine under or over dosed regions as well as track
the cumulative dose to various organs in the patient.
By focusing on the key parameter determining radia-
tion treatment outcomes, dose verification and dose-
guided therapy have the potential to considerably
improve the treatment of cancer. Moreover, they offer
the opportunity to increase our understanding of
treatment effectiveness, improving our knowledge of
the radiation doses and distributions that lead to the
control of cancer or the injury of normal structures.
Although this level of precision has long been a goal in
radiation oncology, the continuing advances in imaging
technology and in imaging processing may soon make
this goal attainable.
The authors would like to acknowledge the following
persons for their valuable contributions, enlightening
discussions and active participation on the acquisition of
the clinical cone-beam images. At UCSF, Albert Chan,
Chris Malfatti, Amy Gillis, Ping Xia, Lynn Verhey. And
at Siemens OCS, Ali Bani-Hashemi. This research was
supported by Siemens Oncology Care Systems (OCS).
One of the authors (OM) wishes to acknowledge a
doctoral scholarship from NSERC-Canada.
1. Jacob R, Hanlon AL, Horwitz EM, Movsas B, Uzzo RG,
Pollack A. The relationship of increasing radiotherapy dose
to reduced distant metastases and mortality in men with
prostate cancer. Cancer 2004;100:538–43.
2. Lee N, Xia P, Fischbein NJ, Akazawa P, Akazawa C, Quivey
JM. Intensity-modulated radiation therapy for head-and-
neck cancer: the UCSF experience focusing on target
3. Kuban D, Pollack A, Huang E, Levy L, Dong L, Starkschall
G, et al. Hazards of dose escalation in prostate cancer
radiotherapy. Int J Radiat Oncol Biol Phys 2003;57:1260–8.
4. Hurkmans CW, Cho BC, Damen E, Zijp L, Mijnheer BJ.
Reduction of cardiac and lung complication probabilities
after breast irradiation using conformal radiotherapy with
or without intensitymodulation.
5. Tubiana M, Eschwege F. Conformal radiotherapy and
6. Pouliot J, Kim Y, Lessard E, Hsu IC, Vigneron DB,
Kurhanewicz J. Inverse planning for HDR prostate brachy-
therapy used to boost dominant intraprostatic lesions
defined by magnetic resonance spectroscopy imaging. Int
J Radiat Oncol Biol Phys 2004;59:1196–207.
7. Wang JZ, Li XA. Impact of tumor repopulation on radio-
8. Verhey L. Patient immobilization for 3-D RTP and virtual
simulation. In: Purdy JA, Starkschall G, editors. A practical
guide to 3-D planning and conformal radiation therapy.
9. Barker JL Jr, Garden AS, Ang KK, O’Daniel JC, Wang H,
Court LE, et al. Quantification of volumetric and geometric
changes occurring during fractionated radiotherapy for
head-and-neck cancer using an integrated CT/linear accel-
erator system. Int J Radiat Oncol Biol Phys 2004;59:960–70.
10. Hansen EK, Xia P, Quivey J, Weinberg V, Bucci MK. The
roles of repeat CT imaging and re-planning during the
course of IMRT for patients with head and neck cancer
(abstract). ASTRO 46th Annual Meeting; 2004 October 3-7;
Atlanta, GA. Int J Radiat Oncol Biol Phys 2004;60:S159.
11. Millender LE, Aubin M, Pouliot J, Shinohara K, Roach M
3rd. Daily electronic portal imaging for morbidly obese men
undergoing radiotherapy for localized prostate cancer. Int J
Radiat Oncol Biol Phys 2004;59:6–10.
12. Chung PW, Haycocks T, Brown T, Cambridge Z, Kelly V,
Alasti H, et al. On-line aSi portal imaging of implanted
fiducial markers for the reduction of interfraction error
during conformal radiotherapy of prostate carcinoma. Int J
Radiat Oncol Biol Phys 2004;60:329–34.
13. Pouliot J, Aubin M, Langen KM, Liu YM, Pickett B,
Shinohara K, et al. (Non)-migration of radio-opaque
markers used for on-line localization of the prostate with
an electronic portal imaging device. Int J Radiat Oncol Biol
14. Uematsu M, Sonderegger M, Shioda A, Tahara K, Fukui T,
Hama Y, et al. Daily positioning accuracy of frameless
stereotactic radiation therapy with a fusion of computed
tomography and linear accelerator (focal) unit: evaluation
of z-axis with a z-marker. Radiother Oncol 1999;50:337–9.
J Radiat OncolBiolPhys
J Chen, O Morin, M Aubin et al
S96The British Journal of Radiology, Special Issue 2006
15. Jaffray DA, Siewerdsen JH, Wong JW, Martinez AA. Flat-
panel cone-beam computed tomography for image-guided
16. Le ´tourneau D, Martinez AA, Lockman D, Yan D, Vargas C,
Ivaldi G, et al. Assessment of residual error for online cone-
beam CT-guided treatment of prostate cancer patients. Int J
Radiat Oncol Biol Phys 2005;62:1239–46.
17. Seppi EJ, Munro P, Johnsen SW, Shapiro EG, Tognina C,
Jones D, et al. Megavoltage cone-beam computed tomo-
graphy using a high-efficiency image receptor. Int J Radiat
Oncol Biol Phys 2003;55:793–803.
18. Ghelmansarai FA, Bani-Hashemi A, Pouliot J, Calderon E,
Hernandez P, Mitschke M, et al. Soft tissue visualization
using a highly efficient megavoltage cone beam CT imaging
system. In: Flynn MJ, editor. Medical imaging 2005: physics
of medical imaging. Proceedings of SPIE Medical Imaging:
Physics of Medical Imaging; 2005 February 13–15; San
Diego, CA. Bellingham, WA: SPIE Press, 2005:159–70.
19. Pouliot J, Bani-Hashemi A, Chen J, Svatos M, Ghelmansarai
F, Mitschke M, et al. Low-dose megavoltage cone-beam CT
for radiation therapy. Int J Radiat Oncol Biol Phys
20. Nakagawa K, Aoki Y, Tago M, Ohtomo K. MV CT assisted
stereotactic radiosurgery for thoracic tumors. Int J Radiat
Oncol Biol Phys 2000;48:449–57.
21. Ruchala KJ, Olivera GH, Schloesser EA, Mackie TR.
Megavoltage CT on a tomotherapy system. Phys Med Biol
22. Mackie TR, Kapatoes J, Ruchala K, Lu W, Wu C, Olivera G,
et al. Image guidance for precise conformal radiotherapy.
Int J Radiat Oncol Biol Phys 2003;56:89–105.
23. Langen KM, Zhang Y, Andrews RD, Hurley ME, Meeks SL,
Poole DO, et al. Initial experience with megavoltage (MV)
CT guidance for daily prostate alignments. Int J Radiat
Oncol Biol Phys 2005;62:1517–24.
24. Mageras GS, Yorke E. Deep inspiration breath hold and
respiratory gating strategies for reducing organ motion in
radiation treatment. Semin Radiat Oncol 2004;14:65–75.
25. Berson AM, Emery R, Rodriguez L, Richards GM, Ng T,
Sanghavi S, et al. Clinical experience using respiratory
gated radiation therapy: comparison of free-breathing and
breath-hold techniques. Int J Radiat Oncol Biol Phys
26. Petersch B, Bogner J, Dieckmann K, Potter R, Georg D.
Automatic real-time surveillance of eye position and gating
for stereotactic radiotherapy of uveal melanoma. Med Phys
27. Shirato H, Shimizu S, Kitamura K, Nishioka T, Kagei K,
Hashimoto S, et al. Four-dimensional treatment planning
and fluoroscopic real-time tumor tracking radiotherapy for
moving tumor. Int J Radiat Oncol Biol Phys 2000;48:435–42.
28. Murphy MJ. Tracking moving organs in real time. Semin
Radiat Oncol 2004;14:91–100.
29. Schweikard A, Shiomi H, Adler J. Respiration tracking in
radiosurgery. Med Phys 2004;31:2738–41.
30. Court LE, Dong L, Lee AK, Cheung R, Bonnen MD,
O’Daniel J, et al. An automatic CT-guided adaptive
radiation therapy technique by online modification of
multileaf collimator leaf positions for prostate cancer. Int J
Radiat Oncol Biol Phys 2005;62:154–63.
31. Wu C, Jeraj R, Lu W, Mackie TR. Fast treatment plan
modification with an over-relaxed Cimmino algorithm.
Med Phys 2004;31:191–200.
32. Brabbins D, Martinez A, Yan D, Lockman D, Wallace M,
Gustafson G, et al. A dose-escalation trial with the adaptive
radiotherapy process as a delivery system in localized
prostate cancer: analysis of chronic toxicity. Int J Radiat
Oncol Biol Phys 2005;61:400–8.
33. Rehbinder H, Forsgren C, Lof J. Adaptive radiation therapy
for compensation of errors in patient setup and treatment
delivery. Med Phys 2004;31:3363–71.
34. van Herk M. Errors and margins in radiotherapy. Semin
Radiat Oncol 2004;14:52–64.
35. Wu C, Jeraj R, Olivera GH, Mackie TR. Re-optimization in
adaptive radiotherapy. Phys Med Biol 2002;47:3181–95.
36. Scarantino CW, Ruslander DM, Rini CJ, Mann GG, Nagle
HT, Black RD. An implantable radiation dosimeter for use
inexternal beam radiation
37. Kirby MC, Williams PC. The use of an electronic portal
imaging device for exit dosimetry and quality control
38. Boellaard R, van Herk M, Uiterwaal H, Mijnheer B. Two-
dimensional exit dosimetry using a liquid-filled electronic
portal imaging device and a convolution model. Radiother
39. Boellaard R, Essers M, van Herk M, Mijnheer BJ. New
method to obtain the midplane dose using portal in vivo
dosimetry. Int J Radiat Oncol Biol Phys 1998;41:465–74.
40. Chang J, Mageras GS, Chui CS, Ling CC, Lutz W. Relative
profile and dose verification of intensity-modulated radia-
tion therapy. Int J Radiat Oncol Biol Phys 2000;47:231–40.
41. Louwe RJW, Damen EMF, van Herk M, Minken AWH,
Torzsok O, Mijnheer BJ. Three-dimensional dose recon-
struction of breast cancer treatment using portal imaging.
Med Phys 2003;30:2376–89.
42. McNutt TR, Mackie TR, Reckwerdt P, Paliwal BR. Modeling
dose distributions from portal dose images using the
43. Hansen VN, Evans PM, Swindell W. The application of
transit dosimetry to precision radiotherapy. Med Phys
44. Kapatoes JM, Olivera GH, Balog JP, Keller H, Reckwerdt PJ,
Mackie TR. On the accuracy and effectiveness of dose
45. Partridge M, Ebert M, Hesse B-M. IMRT verification by
three-dimensional dose reconstruction from portal beam
measurements. Med Phys 2002;29:1847–58.
46. Pouliot J, Xia P, Aubin M, Verhey L, Bani-Hashemi A,
Ghelmansarai F, et al. Low-dose megavoltage cone-beam
CT for dose-guided radiation therapy (abstract). ASTRO
45th Annual Meeting; 2003 October 19-23; Salt Lake City,
UT. Int J Radiat Oncol Biol Phys 2003;57:S183–4.
47. Schaly B, Kempe JA, Bauman GS, Battista JJ, van Dyk J.
Tracking the dose distribution in radiation therapy by
48. Wiesent K, Barth K, Navab N, Durlak P, Brunner T, Schutz
O, et al. Enhanced 3-D reconstruction algorithm for C-arm
systems suitable for interventional procedures. IEEE Trans
Med Imaging 2000;19:391–403.
49. Morin O, Chen J, Aubin M, Pouliot J. Evaluation of the
mechanical stability of a megavoltage imaging system using
a new flat panel positioner. In: Flynn MJ, editor. Medical
Imaging 2005: Physics of Medical Imaging. Proceedings of
SPIE Medical Imaging: Physics of Medical Imaging; 2005
February 13-15; San Diego, CA. Bellingham, WA: SPIE
50. Ruchala KJ, Olivera GH, Schloesser EA, Hinderer R, Mackie
TR. Calibration of a tomotherapeutic MVCT system. Phys
Med Biol 2000;45:N27–36.
51. Chen J, Morin O, Chen H, Aubin M, Pouliot J. The effect of
MV cone-beam CT cupping artifacts on dose calculation
accuracy (abstract). AAPM 47th Annual Meeting; 2005 July
24-28; Seattle, WA. Med Phys 2005;32:1936.
therapy. Med Phys
Dose-guided radiation therapy with MV CBCT
The British Journal of Radiology, Special Issue 2006 S97
52. Chen J, Chuang C, Morin O, Aubin M, Pouliot J. Calibration Download full-text
of an amorphous-silicon flat panel detector for absolute
dosimetry (abstract). AAPM 46th Annual Meeting; 2004
July 25-29; Pittsburgh, PA. Med Phys 2004;31:1831 and
Chen J, Chuang C, Morin O, Aubin M, Pouliot J. Calibration
of an amorphous silicon flat panel portal imager for exit-
beam dosimetry. [Submitted to Med Phys 2005].
53. Ahnesjo ¨ A. Collapsed cone convolution of radiant energy
for photon dose calculation in heterogeneous media. Med
54. Hansen VN, Swindell W, Evans PM. Extraction of primary
signal from EPIDs using only forward convolution. Med
55. Partridge M, Evans PM. The practical implementation of a
scatter model for portal imaging at 10 MV. Phys Med Biol
56. Spies L, Partridge M, Groh BA, Bortfeld T. An iterative
algorithm for reconstructing incident beam distributions
from transmission measurements using electronic portal
imaging. Phys Med Biol 2001;46:N203–11.
J Chen, O Morin, M Aubin et al
S98 The British Journal of Radiology, Special Issue 2006