3D in vivo dosimetry using megavoltage cone-beam CT and EPID dosimetry.
ABSTRACT To develop a method that reconstructs, independently of previous (planning) information, the dose delivered to patients by combining in-room imaging with transit dose measurements during treatment.
A megavoltage cone-beam CT scan of the patient anatomy was acquired with the patient in treatment position. During treatment, delivered fields were measured behind the patient with an electronic portal imaging device. The dose information in these images was back-projected through the cone-beam CT scan and used for Monte Carlo simulation of the dose distribution inside the cone-beam CT scan. Validation was performed using various phantoms for conformal and IMRT plans. Clinical applicability is shown for a head-and-neck cancer patient treated with IMRT.
For single IMRT beams and a seven-field IMRT step-and-shoot plan, the dose distribution was reconstructed within 3%/3mm compared with the measured or planned dose. A three-dimensional conformal plan, verified using eight point-dose measurements, resulted in a difference of 1.3 +/- 3.3% (1 SD) compared with the reconstructed dose. For the patient case, planned and reconstructed dose distribution was within 3%/3mm for about 95% of the points within the 20% isodose line. Reconstructed mean dose values, obtained from dose-volume histograms, were within 3% of prescribed values for target volumes and normal tissues.
We present a new method that verifies the dose delivered to a patient by combining in-room imaging with the transit dose measured during treatment. This verification procedure opens possibilities for offline adaptive radiotherapy and dose-guided radiotherapy strategies taking into account the dose distribution delivered during treatment sessions.
- SourceAvailable from: PubMed Central[Show abstract] [Hide abstract]
ABSTRACT: To investigate for prostate cancer patients the comparison of 'in-vivo' measured portal dose images (PDIs) with predictions based on a kilovoltage cone-beam CT scan (CBCT), acquired during the same treatment fraction, as an alternative for pre-treatment verification. For evaluation purposes, predictions were also performed using the patients' planning CTs (pCT). To get reliable CBCT electron densities for PDI predictions, Hounsfield units from the pCT were mapped onto the CBCT, while accounting for non-rigidity in patient anatomy in an approximate way. PDI prediction accuracy was first validated for an anatomical phantom, using IMRT treatment plans of ten prostate cancer patients. Clinical performance was studied using data acquired for 50 prostate cancer patients. For each patient, 4--5 CBCTs were available, resulting in a total of 1413 evaluated images. Measured and predicted PDIs were compared using gamma-analyses with 3% global dose difference and 3 mm distance to agreement as reference criteria. Moreover, the pass rate for automated PDI comparison was assessed. To quantify improvements in IMRT fluence verification accuracy results from multiple fractions were combined by generating a gamma-image with values halfway the minimum and median gamma values, pixel by pixel. For patients, CBCT-based PDI predictions showed a high agreement with measurements, with an average percentage of rejected pixels of 1.41% only. In spite of possible intra-fraction motion and anatomy changes, this was only slightly larger than for phantom measurements (0.86%). For pCT-based predictions, the agreement deteriorated (average percentage of rejected pixels 2.98%), due to an enhanced impact of anatomy variations. For predictions based on CBCT, combination of the first 2 fractions yielded gamma results in close agreement with pre-treatment analyses (average percentage of rejected pixels 0.63% versus 0.35%, percentage of rejected beams 0.6% versus 0%). For the pCT-based approach, only combination of the first 5 fractions resulted in acceptable agreement with pre-treatment results. In-room acquired CBCT scans can be used for high accuracy IMRT fluence verification based on in-vivo measured EPID images. Combination of gamma results for the first 2 fractions can largely compensate for small accuracy reductions, with respect to pre-treatment verification, related to intra-fraction motion and anatomy changes.Radiation Oncology 09/2013; 8(1):211. · 2.11 Impact Factor
- [Show abstract] [Hide abstract]
ABSTRACT: In vivo dosimetry is now widely recommended to avoid major treatment error. Transit dosimetry using portal imagers allows fast and accurate in vivo dose verifications. Several teams have published clinical studies but no recommendation has been proposed to define tolerance levels and validation criteria. This study proposes a simple methodology to assess the overall standard deviation of transit dosimetry and was applied to our transit dosimetry method. In a first step, the uncertainties due to the dose reconstruction method are evaluated. Their estimation is based on a set of geometries, representative of clinical situations for which 45 points of measurement have been defined. In a second step, we studied the variations of our method in clinical situations. During the treatment session of the patient, the dose was reconstructed and the differences between reconstructed dose and prescribed dose were used to define a realistic tolerance level, adapted to the clinical routine. Then, a methodology is proposed to determine if the transit dosimetry method, with the defined tolerance level allows detecting significant treatment errors (>5% of the prescribed dose). RESULTS - CONCLUSION: Applying this methodology we concluded that a tolerance level of 6.5% (k=2) can be associated with our method. With this value, it is demonstrated that in many cases differences of 5% (or less) on the prescribed dose can be detected. This study demonstrates clearly that in vivo transit dosimetry is not able to detect all the treatment errors but remains an ultimate and efficient tool in many situations.Cancer/Radiothérapie 10/2013; · 1.48 Impact Factor
- [Show abstract] [Hide abstract]
ABSTRACT: Aim To present practical examples of our new algorithm for reconstruction of 3D dose distribution, based on the actual MLC leaf movement. Background DynaLog and RTplan files were used by DDcon software to prepare a new RTplan file for dose distribution reconstruction. Materials and methods Four different clinically relevant scenarios were used to assess the feasibility of the proposed new approach: (1) Reconstruction of whole treatment sessions for prostate cancer; (2) Reconstruction of IMRT verification treatment plan; (3) Dose reconstruction in breast cancer; (4) Reconstruction of interrupted arc and complementary plan for an interrupted VMAT treatment session of prostate cancer. The applied reconstruction method was validated by comparing reconstructed and measured fluence maps. For all statistical analysis, the U Mann–Whitney test was used. Results In the first two and the fourth cases, there were no statistically significant differences between the planned and reconstructed dose distribution (p = 0.910, p = 0.975, p = 0.893, respectively). In the third case the differences were statistically significant (p = 0.015). Treatment plan had to be reconstructed. Conclusion Developed dose distribution reconstruction algorithm presents a very useful QA tool. It provides means for 3D dose distribution verification in patient volume and allows to evaluate the influence of actual MLC leaf motion on the dose distribution.Reports of Practical Oncology and Radiotherapy 01/2014;
3D IN VIVO DOSIMETRY USING MEGAVOLTAGE CONE-BEAM CT AND EPID
WOUTER VAN ELMPT, M.SC., SEBASTIAAN NIJSTEN, M.SC., STEVEN PETIT, M.SC., BEN MIJNHEER, PH.D.,
PHILIPPE LAMBIN, M.D., PH.D., AND ANDRE´DEKKER, M.SC., PH.D.
Department of Radiation Oncology (MAASTRO), GROW Research Institute, University Medical Centre Maastricht, Maastricht,
Purpose: To develop a method that reconstructs, independently of previous (planning) information, the dose
delivered to patients by combining in-room imaging with transit dose measurements during treatment.
Methods and Materials: A megavoltage cone-beam CT scan of the patient anatomy was acquired with the patient
in treatment position. During treatment, delivered fields were measured behind the patient with an electronic por-
tal imaging device. The dose information in these images was back-projected through the cone-beam CT scan and
used for Monte Carlo simulation of the dose distribution inside the cone-beam CT scan. Validation was performed
usingvariousphantoms forconformalandIMRT plans.Clinical applicabilityisshownforahead-and-neckcancer
patient treated with IMRT.
Results: For single IMRT beams and a seven-field IMRT step-and-shoot plan, the dose distribution was recon-
structed within 3%/3mm comparedwith the measured or planned dose. A three-dimensional conformal plan, ver-
ified using eight point-dose measurements, resulted in a difference of 1.3 ± 3.3% (1 SD) compared with the
reconstructed dose. For the patient case, planned and reconstructed dose distribution was within 3%/3mm for
about 95% of the points within the 20% isodose line. Reconstructed mean dosevalues, obtained fromdose–volume
histograms, were within 3% of prescribed values for target volumes and normal tissues.
Conclusions: We present a new method that verifies the dose delivered to a patient by combining in-room imaging
with the transit dose measured during treatment. This verification procedure opens possibilities for offline adap-
tive radiotherapy and dose-guided radiotherapy strategies taking into account the dose distribution delivered dur-
ing treatment sessions.
? 2009 Elsevier Inc.
Megavoltage cone-beam CT, dose verification, in vivo dosimetry, EPID dosimetry, Monte Carlo dose calculation.
Verification of the dose delivered during external beam ra-
diotherapy is of importance to guarantee accurate delivery
of treatment. There are two major sources of uncertainty
that may cause the planned dose to be different from the ac-
tual dose delivered to the patient. The first possible source of
discrepancy is related to the patient model (planning CT
scan) in which the dose distribution is planned and calcu-
lated. This planning CT scan is usually made on a conven-
tional CT scanner a few days or weeks before the start of
treatment. Between the time of obtaining the planning CT
scan and the treatment, as well as during the course of treat-
ment, changes in patient anatomy might occur (e.g., weight
loss or tumor shrinkage). A second source of uncertainty is
accelerator, limitations of the dose calculation algorithm in
the treatment planning system (TPS), or dose delivery errors
of the linear accelerator during treatment (e.g., differences
between planned and actual leaf positions of the multileaf
In the past, various research groups have investigated the
methods either verify the patient anatomy by using in-room
imaging—for example, using the recent developments in
kilovoltage (kV) and megavoltage (MV) cone-beam CT
imaging—or verify the dose delivered using another dose
calculation engine (e.g., Monte Carlo). Combination of both
in-room imaging and a dose calculation based on the actual
delivered treatment fields will yield the ultimate verification
of the dose delivery: a reconstruction of the dose delivered
in the actual patient anatomy for a particular fraction, i.e.,
3D in vivo dosimetry.
Reprint requests to: Wouter van Elmpt, M.Sc., Department of Ra-
diation Oncology (MAASTRO), University Medical Centre Maas-
tricht, Dr. Tanslaan 12, NL-6229 ET Maastricht, The Netherlands.
Tel: (+31) 884455666; Fax: (+31) 884455667; E-mail: wouter.
Conflict of interest: none.
Acknowledgments—We thank Stein Fekkes, M.Sc. for building the
Received Sept 4, 2008, and in revised form Nov 17, 2008.
Accepted for publication Nov 21, 2008.
Int. J. Radiation Oncology Biol. Phys., Vol. 73, No. 5, pp. 1580–1587, 2009
Copyright ? 2009 Elsevier Inc.
Printed in the USA. All rights reserved
0360-3016/09/$–see front matter
The aim of this study was to present and validate a model
used for the 3D in vivo dose verification based on information
gathered during treatment, i.e., the patient anatomy measured
the patient using an electronic portal imaging device (EPID)
calibrated for dosimetry purposes (8). The method was vali-
and inhomogeneous phantoms for both 3D conformal and in-
tensity-modulated fields. The clinical applicability of our
model is demonstrated for a head-and-neck cancer patient
treated with IMRT.
METHODS AND MATERIALS
Treatment delivery and portal image acquisition
A Siemens Oncor linear accelerator (Siemens Medical Solutions,
quired using an amorphous silicon-type EPID (Siemens OptiVue
1000ST) with the standard acquisition software implemented in the
Siemens Coherence workspace. Images had matrix sizes of 512 ?
512 or 1024 ? 1024 pixels with a 16-bit gray-level depth and were
converted into portal dose images using an in-house developed cali-
bration model. This model has been extensively described elsewhere
(9).FortheIMRTtreatmentfields, weappliedthisconversion model
ments are then summed to a portal dose image for the entire beam.
Electron density from MV cone-beam CT
MV cone-beam CT scans were made following the clinical
protocol using a total number of 8 or 15 monitor units (MUs).
The projection images of the MV cone-beam CT-scan were filtered
using an anisotropic diffusion filter, corrected for scattered radiation
and the photon energy dependence of the response of the EPID. The
projection images were then reconstructed to a 3D volume using
a voxel-driven filtered back-projection algorithm based on the Feld-
kamp-algorithm. Subsequently, the CT numbers were converted
into electron densities using a linear conversion based on a point
measurement of the CT number inside a water phantom relative to
the electron density of water. This calibration procedure has been
described in detail by Petit et al. (11). The cone-beam CT scans
were cropped to the patient contour to remove residual noise from
the images outside the patient.
Back-projection of energy fluence through an MV
cone-beam CT scan
The transit portal dose images of the treatment fields were mea-
An iterative scheme calculated the patient-scattered radiation in the
portal dose images using the density information of the MV cone-
beam CT scan. In a first step, the radiological (i.e., water equivalent)
thickness t(x,y) was projected from the target to every point (x,y) in
the detector plane. This thickness map t(x,y) was defined as the sum
of the electron density distribution reð r!Þ in the MV cone-beam CT
scan along a diverging rayline l
the detector plane, relative to the electron density of water rwater:
by correcting for the scattered radiation S(x,y) produced inside the
!from the target to the point (x,y) in
reð r!Þd l
patient and the attenuation of the photon beam by the patient. The
dose distribution I0(x,y) that would be measured in the absence of
the patient was derived from the measured portal dose distribution
I(x,y) behind the patient by correcting for scatter and attenuation
Iðx;yÞ ? Sðx;yÞ
with m(x,y,t) the effective attenuation coefficient that takes into ac-
count possible beam hardening or softening effects due the presence
of the patient inside the beam, and differences in off-axis beam
The scattered dose S(x,y) exiting the patient was calculated using
the equivalent homogeneous phantom (EHP) concept (13) and
pencil beam scatter kernels
in detail in a previous publication (12).
Because the calculation of the incident dose distribution I0(x,y)
and the scattered dose distribution S(x,y) depend on each other, an
iterative loop was introduced. The earlier steps (i.e., Eq. 2 and 3)
were repeated three times, which was necessary to ensure conver-
gence of the iterative scheme (11, 14). The scattered dose distribu-
tion was set to zero for the first loop.
The dose distribution I0(x,y) is then converted into an energy flu-
ence distribution I0(x,y)
I0ðx0;y0Þ$Kðx0? x0;y0? y0;tðx;yÞÞdx0dy0;
by correcting for the phantom scatter inside the (artificial) layer of
water (tw= 5.0 cm) using the energy deposition kernel KJ(x,y).
Values for this kernel were taken from van Elmpt et al. (15).
This step was necessary because our EPID calibration model con-
verts an EPID image into a dose distribution at a depth tw= 5 cm
inside a water phantom. The attenuation coefficient m(q) was set
to 4.68 10?2cm?1with the off-axis correction proposed by
Tailor et al. (16).
From the incident energy fluence distribution J0(x,y), a phase
space was sampled assuming that all photons originate from a point
source at the target of the linac. A 3D dose calculation that takes in-
homogeneities into account was started from this phase space using
the fast Monte Carlo code XVMC (17) using the electron density in-
formation of the MV cone-beam CT scan. Previously, the accuracy
ofthis dosereconstructionmethodforpretreatment verification(i.e.,
based on EPID images without an object and the planning CT-scan)
was assessed to be within 3% for homogeneous (15) and inhomoge-
neous phantoms (18), whereas its clinical applicability was also
A schematic overview of the various steps involved in the 3D
dose reconstruction procedure is shown in Fig. 1.
Phantom verification measurements
The method was tested using cubic and cylindrical shaped phan-
toms irradiated using a 6-MV photon beam with 3D conformal as
well as intensity-modulated fields. These experiments are described
in detail in the following sections.
Because the field-of-view (FOV) of the MV cone-beam CT scan
iscurrently limitedtoavolumeof approximately27?27 ?27cm3,
we choose phantoms with similar dimensions as applied for the ver-
ification of head-and-neck treatments. Other treatment sites require
3D in vivo dosimetry using MVCBCT and EPIDs d W. VAN ELMPT et al. 1581
either a larger FOV of the MV cone-beam CT or a combined regis-
tration of the planning CT scan with the MV cone-beam CT scan to
compensate for the missing anatomy outside the FOV, are the sub-
ject of further study, and are beyond the scope of this article (19).
Single IMRT field verification: A cubic phantom made of poly-
styrene with dimensions 29.0 ? 14.5 ? 12.8 cm3(Phantom 1) posi-
tioned symmetrically around the isocenter was irradiated with a
6-MV IMRT field consisting of five segments each delivered with
40 MUs having a field size of 20 ? 10 cm2reduced to 4 ? 10
cm2by closing the leaves in four steps of 4 cm. An MV cone-
beam CT scan having 8 MUs was used for imaging the phantom ge-
ometry. Dose verification was performed using film measurements
(EDR2, Eastman Kodak Company, Rochester, NY) calibrated for
relative dose measurements. Planned two-dimensional dose distri-
butions in a plane inside the phantom at 4.2-cm and 8.6-cm depth
butions. The film measurements were normalized to the recon-
structed dose value at a point in the center of the second segment.
Differences are quantified using the gamma evaluation (20) using
a 3% of the maximum dose and 3-mm distance-to-agreement crite-
rion within the 5% isodose line.
Three-dimensional conformal technique verification: To verify
the dose reconstruction also in inhomogeneous cases, a cylindrical
phantom of 11 cm diameter and 14 cm length was made of flexible
silicon material (r z 1.2 g/cm3), which included a low-density
foam region (r z 0.1 g/cm3) of 5 cm diameter and 3 cm length
(phantom 2), as well as a higher density PMMA disk (rz1.4
g/cm3) of 5 cm diameter and 1.5 cm length. A schematic picture
of the phantom is shown in Fig. 2. The phantom was irradiated
of 9.8 ? 13.0 cm2at 0?gantry angle and a virtual wedge of 45?; the
gantry angle, and the third field had a size of 7.6 ? 5.0 cm2with 100
MUs at 270?. To validate the 3D dose reconstruction method, cali-
brated Metal Oxide Semiconductor Field-Effect Transistor (MOS-
FET) detectors (TN-502RD, Thomson and Nielsen Electronics,
Ottawa, Canada) were positioned inside this phantom at eight points
located throughout the entire volume. The MOSFET measurements
were repeated three times to reduce the measurement error in MOS-
FET readings (21).
Comparison with the TPS: Verification of the 3D reconstructed
dose distribution was performed for a clinical IMRT plan. A 20-cm
diameter cylindrical phantom having a polystyrene wall of 5-mm
thickness and filled with water was irradiated using a clinical
were delivered consisting of in total 92 segments and 528 MUs. The
reconstructed dose distribution was compared with the 3D dose
distribution obtained from the clinically applied treatment planning
system (XiO 4.3.4, CMS, St. Louis, MO). A 2D gamma analysis
within the 20% isodose line was performed in the transversal, coro-
nal, and sagittal slice through the isocenter using a 3% of the maxi-
mum dose and a 3-mm distance-to-agreement criterion.
Clinical case study
As an example to illustrate how our 3D dose reconstruction
method is applied in clinical practice, we used the treatment of a na-
sopharyngeal cancer patient with an 8-field step-and-shoot IMRT
technique having 484 MUs delivered in 97 segments resulting in
56 Gy in 28 fractions. To verify the 3D dose distribution delivered
to this patient, an MV cone-beam CT scan using 8 MUs was made
during the seventh fraction and the treatment fields were captured
using the EPID. A 3D dose reconstruction inside the cone-beam
CT scan of this patient was performed. One of the beams included
a couch rotation. Because of a possible collision between the gantry
and the EPID, it was not possible to capture this treatment field,
hence a portal dose image without a patient in the beam was used
in the calculation of the energy fluence in Eq. 4. The reconstructed
dose distribution was compared with the planned dose distribution
inside the kV planning CT scan using 2D gamma analysis applying
a 3% of the maximum dose and 3-mm distance-to-agreement crite-
rion evaluated within the 20% isodose line. Because the dose recon-
struction was performed early during treatment, no large anatomic
differences are expected, and dose–volume histograms (DVHs)
were calculated using structures delineated in the planning CT scan.
Single IMRT field verification: Two planes extracted from
the 3D dose distribution inside the MV cone-beam CT scan
were compared with the measured film dose distributions.
A single plane extracted from the dose distribution at 8.6
cm depth is shown in Fig. 3, together with the measured
film dose distribution. A gamma evaluation for values larger
than 5% of the maximum dose was used to compare both
dose distributions. Good agreement was observed as shown
by median gamma values of 0.149 and 0.141 for the film po-
sitioned at 4.2 and 8.6 cm, respectively. The maximum
Fig. 1. Schematic overview of the various steps in the three-dimensional (3D) dose reconstruction method.
EPID = electronic portal imaging device.
1582 I. J. Radiation Oncology d Biology d PhysicsVolume 73, Number 5, 2009
gamma value of a single pixel for both planes was 0.89 and
Three-dimensional conformal technique verification: The
eight dose values obtained with the MOSFETs were com-
pared with the dose values reconstructed at the position of
these detectors for Phantom 2. The average difference be-
tween the eight MOSFET measurements and 3D recon-
structed values was small, on average 1.3 ? 3.3% (1 SD)
expressed relative to the maximum dose of 2 Gy. The largest
observeddeviation was6.4%forasuperficiallyplaced MOS-
FET adjacent to the high-density region. A slice of the MV
cone-beam CT scan through the isocenter with the dose dis-
tribution is shown in Fig. 2, together with the position of two
IMRT verification: To verify the accuracy of the 3D dose
reconstruction method for an IMRT treatment, the 3D recon-
structed dose distribution was compared with the TPS
planned dose distribution for a homogenous water-filled cyl-
inder irradiated with the seven-field IMRT technique. Dose
distributions were in good agreement with each other, indi-
cated by the low values of the 2D gamma analysis in the
various planes. The mean gamma value within the 20% iso-
dose line was 0.253, 0.293, and 0.359 for the transversal,
coronal, and sagittal plane, respectively. The percentage of
points with a gamma value smaller than unity for these
planes was 100%, 99.8%, and 100%, respectively. Also
scored were the gamma values of the upper percentile,
which were 0.74, 0.79, and 0.84, for the transversal, coronal,
and sagittal plane, respectively. Figure 4 shows the dose dis-
tribution of the TPS calculation overlaid with the 3D dose
reconstruction method for the three orientations through
Fig. 2. Schematic picture of the cylindrical phantom (Phantom 2) having a low-density (yellow) and high-density (red)
region (left) and a slice through the MV cone-beam CT scan with the reconstructed dose distribution (right). The white
crosses indicate the position of a Metal Oxide Semiconductor Field-Effect Transistor (MOSFET) detector.
Fig. 3. Dose distribution in a plane at 6.8 cm depth inside the cubic phantom (Phantom 1). Film measurement (left)
and reconstructed dose distribution (middle). The gamma evaluation (3% max. dose/3 mm) is shown at the right.
3D = three-dimensional.
3D in vivo dosimetry using MVCBCT and EPIDs d W. VAN ELMPT et al.1583
Figure 5 shows the planning CT scan with the planned
dose distribution and the MV cone-beam CT scan with the
3D reconstructed dose. From an imaging point of view, no
large anatomic differences canbeobserved between theplan-
ning CT scan and the online cone-beam CTscan. Toquantify
possible dosimetric differences, the dose distribution has
been compared using a gamma evaluation. Good agreement
is obtained between the planned and measured dose distribu-
tion indicated by low gamma values. For the transversal and
sagittal slice shown in Fig. 5, 93.5% and 95.2% of the points,
respectively, had a gamma value < 1 with a median gamma
value of 0.364 and 0.292, respectively. The gamma evalua-
tion reveals some small dose differences in the regions
around the air cavities.
fraction that was verified are scaled to the situation for 28
fractions to make a comparison with the planned DVH. For
the PTV, a small underdosage (0.9%) was observed in
Fig. 4. Transversal, coronal, and sagittal slice of the MV cone-beam CT scan of a 20-cm diameter water-filled cylinder
withthedose distributioncalculated bythe treatment planningsystem(solid lines) and thethree-dimensional reconstructed
dose distribution (dashed lines).
Fig. 5. Transversal (top row) and sagittal (bottom row) slices through the planning CT scan (left) with the dose calculated
by the treatment planning system TPS, the MV cone-beam CT-scan with the reconstructed dose distribution (middle) and
are visible at the edges of the air cavities.
1584I. J. Radiation Oncology d Biology d PhysicsVolume 73, Number 5, 2009
mean dose between planned (56.4 Gy) and reconstructed
dose (55.9 Gy) for the high-risk PTV. The mean dose in
the low-risk PTV was 1.7% lower, 51.8 Gy vs. 50.9 Gy for
theplanned and reconstructed dosedistribution,respectively.
For the normal structures, also no large differences were ob-
served. For instance, the maximum dose in the spinal cord
was 25.7 Gy vs. 25.9 Gy, the mean dose in the brain 9.0
Gy vs. 9.2 Gy, the mean dose in the left eye bulb 35.9 Gy
vs. 34.3 Gy, and for the right eye 23.2 Gy vs. 23.5 Gy for
the planned vs. reconstructed dose, respectively.
3D dose distribution delivered to the patient that is com-
pletely independent of the treatment planning process, both
with respect to patient anatomy and the dosimetric aspects.
For the actual patient anatomy in treatment position, an
MV cone-beam CT scan, calibrated for dose calculation, is
used. For the fields delivered during treatment, the transit
portal dose images behind the patient are acquired as a mea-
sure of the dose delivered during treatment. Differences be-
tween plannedand delivered
incorporated in this step, and in addition, possible errors or
inaccuracies in the dose calculation of the treatment planning
system are taken into account by using an independent dose
calculation engine based on Monte Carlo simulations.
The advantage and novelty of this 3D dose reconstruction
method compared with other dose recalculation methods
(7, 19, 22, 23) is that the actual delivered beams are used to
reconstruct the dose delivered to the patient. If the MV
cone-beam CT scan information is already used for adjusting
the patient setup, then no extra dose has to be delivered to the
patient. Strategies for taking the imaging dose of the MV
cone-beam CT scan into account during IMRT optimization
were described by Morin et al. (24). The method does not re-
quire additional measurements, patient treatment time, or
equipment and can be performed with a modern linear accel-
erator equipped with an EPID and (MV) cone-beam CT tech-
nology. Other imaging modalities, including in-room
conventional CT or kV cone-beam CT technology, can also
be used for our dose reconstruction method provided that
CT images are calibrated into electron density values (23).
The time needed to reconstruct the 3D dose for the head-
and-neck case was approximately 2 hours for an eight-field
IMRT step-and-shoot treatment. Conversion ofthe portalim-
ages to portal dose images takes less than 10 min, the correc-
tion of the cupping artifact in the MV cone-beam CT images
takes approximately 1 hour on a 2.4-GHz Intel Core 2 Duo
processor, and the dose calculation using the Monte Carlo
XVMC code takes less than 20 min on the same machine
for a statistical accuracy better than 2%. This procedure is
fast enough for offline verification of a fractionated treatment
in which the 3D reconstructed dose information has to be
available before the next fraction, typically the next day.
measured dose (film and MOSFET data) and the 3D recon-
structed dose distribution. The somewhat larger differences
between the MOSFET dose values and the 3D reconstructed
the MOSFET measurements, e.g., due to the nonoptimal re-
tion factor differences, and aging of the MOSFET (21).
The 3D dose reconstruction method consists of various
method calculates portal dose values with an uncertainty of
Fig. 6. Dose–volume histograms of the planned and reconstructed dose distribution of the low- and high-risk planning
target volume and various normal structures.
3D in vivo dosimetry using MVCBCT and EPIDs d W. VAN ELMPT et al.1585
less than 3% (1 SD) (9), the calibration of the MV cone-beam
CT scan into electron density values results in maximum dif-
cer patients showed differences smaller than 2% (1 SD) in the
dividual uncertainties would resultin anupper limiton theor-
der of 4% (1 SD) for the total uncertainty in our dose
reconstruction method. This indicates that the method is suit-
able for detecting dose differences of 5%–7% (1 SD).
A cone-beam CT scan is typically made before treatment
while the measured portal images detect intrafraction motion
of organs during treatment. Our method reconstructs the de-
livered energy fluence in this patient model acquired before
treatment. To take into account intrafraction motion in the
3D dose reconstruction model, volumetric imaging during ir-
radiation must be performed. However, such 3D imaging
techniques during treatment are currently not available.
For the clinical case, small differences can be observed be-
tween the 3D dose distribution calculated by the TPS using
tion using the MV cone-beam CT data. The reconstructed
dose values using the Monte Carlo simulation are in good
agreement with the dose values calculated using the superpo-
sition algorithm of the TPS, although overall, a slightly lower
(approximately 1%–2%) dose is observed. Considering that
many deviations may occur during treatment compared with
the planning (e.g., anatomic changes, variation in output of
these small differences indicate that a good overall quality is
achieved for the entire treatment delivery during this treat-
ment fraction. Small differences, predominantly around the
air cavities, occur in low-density regions where nonequilib-
rium of the scattered photons and electrons exists. Around
the air cavities, the position of the reconstructed isodose lines
differs somewhatfrom the position of theisodoselines calcu-
lated by theTPS,whichare more collapsedinside thepatient.
ergy deposition by the TPS for such configurations.
In clinical practice, it is sometimes difficult to compare
directly the reconstructed dose distribution at a slice of an
in-room cone-beam (or repeated conventional) CT scan
with the original dose distribution calculated in a planning
CT scan. Because changes in both anatomy and dose distri-
bution may occur, it may not always be sufficient to apply
simple rigid registration models for a comparison on
a voxel-to-voxel basis. A simple but cumbersome solution
would be to delineate the structures of interest in the in-
room CT scan and calculate the DVH parameters of interest.
interest maybe difficult becausetheimagequalityof in-room
imaging modalities is generally less than that of a CT scanner
used for planning and contrast enhancement, often in combi-
nation with PET and MR information. Another solution that
might be of interest is the use of nonrigid deformation
models. However, this approach has its limitations. Voxels
may disappear or appear because changes in tumor or normal
tissue volume may occur during the course of treatment. This
is an active area of research and needs further investigation
with respect to our specific application (25). With this final
step, a cumulative measurement of the actual dose delivered
to the patient in the target volume and normal tissue is possi-
ble, and this may benefit offline adaptive radiation therapy
(ART) and eventually true dose-guided radiotherapy.
In this article, we have described a new model that is able
to perform 3D in vivo dosimetry based on in-room MV cone-
beam CT imaging and a dose measurement during treatment
using an EPID. The accuracy was assessed using phantom
measurements, and a comparison was made with the dose
calculations performed by the treatment planning system.
These verification measurements indicate that the error was
smaller than 3% or 3 mm for most of the points in the 3D
reconstructed volume for both conformal and IMRT treat-
ments. The clinical applicability of our model for 3D
in vivo dosimetry was demonstrated for a head-and-neck can-
cer patient treatment with IMRT.
1. Lee L, Le QT, Xing L. Retrospective IMRT dose reconstruction
based on cone-beam CT and MLC log-file. Int J Radiat Oncol
Biol Phys 2008;70:634–644.
2. McDermott LN, Wendling M, Nijkamp J, et al. 3D in vivo
dose verification of entire hypo-fractionated IMRT treatments
using an EPID and cone-beam CT. Radiother Oncol 2008;86:
3. van Elmpt W, Nijsten S, Mijnheer B, et al. The next step in pa-
tient-specific QA: 3D dose verification of conformal and inten-
sity-modulated RT based on EPID dosimetry and Monte Carlo
dose calculations. Radiother Oncol 2008;86:86–92.
4. Chen J, Morin O, Aubin M, et al. Dose-guided radiation therapy
with megavoltage cone-beam CT. Br J Radiol 2006;79(spec no.
5. van Zijtveld M, Dirkx ML, de Boer HC, et al. 3D dose recon-
struction for clinical evaluation of IMRT pretreatment verifica-
tion with an EPID. Radiother Oncol 2007;82:201–207.
6. Louwe RJ, Damen EM, van Herk M, et al. Three-dimensional
dose reconstruction of breast cancer treatment using portal
imaging. Med Phys 2003;30:2376–2389.
7. van Zijtveld M, Dirkx M, Heijmen B. Correction of conebeam
day.’’ Radiother Oncol 2007;85:195–200.
8. van Elmpt W,McDermott L, Nijsten S,et al. Aliterature review
of electronic portal imaging for radiotherapy dosimetry. Radio-
ther Oncol 2008;88:289–309.
9. NijstenSM,vanElmptWJ,Jacobs M,etal.Aglobalcalibration
model for a-Si EPIDs used for transit dosimetry. Med Phys
10. McDermott LN, Nijsten SM, Sonke JJ, et al. Comparison of
ghosting effects for three commercial a-Si EPIDs. Med Phys
11. Petit SF, van Elmpt WJ, Nijsten SM, et al. Calibration of mega-
voltage cone-beam CT for radiotherapy dose calculations:
1586 I. J. Radiation Oncology d Biology d Physics Volume 73, Number 5, 2009
Correction of cupping artifacts and conversion of CT numbers
to electron density. Med Phys 2008;35:849–865.
12. van Elmpt WJ, Nijsten SM, Mijnheer BJ, et al. Experimental
verification of a portal dose prediction model. Med Phys
13. Pasma KL, Heijmen BJ, Kroonwijk M, et al. Portal dose image
(PDI) prediction for dosimetric treatment verification in radio-
therapy. I. An algorithm for open beams. Med Phys 1998;25:
14. Spies L, Partridge M, Groh BA, et al. An iterative algorithm for
reconstructing incident beam distributions from transmission
measurements using electronic portal imaging. Phys Med Biol
15. van Elmpt WJ, Nijsten SM, Schiffeleers RF, et al. A Monte
Carlo based three-dimensional dose reconstruction method de-
rived from portal dose images. Med Phys 2006;33:2426–2434.
16. Tailor RC, Tello VM, Schroy CB, et al. A generic off-axis en-
ergy correction for linac photon beam dosimetry. Med Phys
17. Fippel M. Fast Monte Carlo dose calculation for photon beams
based on the VMC electron algorithm. Med Phys 1999;26:
18. van Elmpt WJ, Nijsten SM, Dekker AL, et al. Treatment verifi-
cation in the presence of inhomogeneities using EPID-based
three-dimensional dose reconstruction. Med Phys 2007;34:
19. Aubry JF, Pouliot J, Beaulieu L. Correction of megavoltage
cone-beam CT images for dose calculation in the head and
neck region. Med Phys 2008;35:900–907.
20. Low DA, Harms WB, Mutic S, et al. A technique for the quan-
titative evaluation of dose distributions. Med Phys 1998;25:
21. Scalchi P, Francescon P. Calibration of a MOSFET detection
system for 6-MV in vivo dosimetry. Int J Radiat Oncol Biol
22. Morin O, Chen J, Aubin M, et al. Dose calculation using mega-
voltage cone-beam CT. Int J Radiat Oncol Biol Phys 2007;67:
23. Yoo S, Yin FF. Dosimetric feasibility of cone-beam CT-based
treatment planning compared to CT-based treatment planning.
Int J Radiat Oncol Biol Phys 2006;66:1553–1561.
24. Morin O, Gillis A, Descovich M, et al. Patient dose consider-
ations for routine megavoltage cone-beam CT imaging. Med
25. Orban de Xivry J, Janssens G, Bosmans G, et al. Tumour delin-
eation and cumulative dose computation in radiotherapy based
on deformable registration of respiratory correlated CT images
of lung cancer patients. Radiother Oncol 2007;85:232–238.
3D in vivo dosimetry using MVCBCT and EPIDs d W. VAN ELMPT et al.1587