Radiotherapy and Oncology 89 (2008) 156–163

www.thegreenjournal.com

Cervix cancer branchytherapy

Consequences of random and systematic reconstruction

uncertainties in 3D image based brachytherapy in cervical cancer
Kari Tanderupa, Taran Paulsen Hellebustb, Stefan Langc, Jørgen Granfeldtd,
Richard Pötterc, Jacob Christian Lindegaarda, Christian Kirisitsc,*
a
Department of Oncology, Aarhus University Hospital, Denmark, bDepartment of Medical Physics, Rikshospital-Radiumhopital
Health Trust, Oslo, Norway, cDepartment of Radiotherapy, Medical University of Vienna, Austria, dDepartment of Mathematical Sciences,
Aarhus University, Denmark

Abstract
Background and purpose: The purpose of this study was to evaluate the impact of random and systematic applicator
reconstruction uncertainties on DVH parameters in brachytherapy for cervical cancer.
Material and methods: Dose plans were analysed for 20 cervical cancer patients with MRI based brachytherapy.
Uncertainty of applicator reconstruction was modelled by translating and rotating the applicator. Changes in DVH
parameters per mm of applicator displacement were evaluated for GTV, CTV, bladder, rectum, and sigmoid. These data
were used to derive patient population based estimates of delivered dose relative to expected dose.
Results: Deviations of DVH parameters depend on direction of reconstruction uncertainty. The most sensitive organs
are rectum and bladder where mean DVH parameter shifts are 5–6% per mm applicator displacement in ant–post
direction. For other directions and other DVH parameters, mean shifts are below 4% per mm. By avoiding systematic
reconstruction errors, uncertainties on DVH parameters can be kept below 10% in 90% of a patient population.
Systematic errors of a few millimetres can lead to significant deviations.
Conclusion: Comprehensive quality control of afterloader, applicators and imaging procedures should be applied to
prevent systematic errors in applicator reconstruction. Random errors should be minimised by using small slice
thickness. With careful reconstruction procedures, reliable DVH parameters for target and OAR’s can be obtained.
c 2008 Elsevier Ireland Ltd. All rights reserved. Radiotherapy and Oncology 89 (2008) 156–163.

Keywords: Brachytherapy; 3D image guidance; Applicator reconstruction; DVH; Uncertainties

In recent years brachytherapy for locally advanced cervi-
cal cancer has shown significant development in the field of
3D image based treatment planning [1–3], and the first clin-
ical evidence strongly indicates that the new approach has
potential to increase local control without increasing mor-
bidity [4]. Recommendations for 3D image based brachy-
therapy in cervical cancer have been worked out by the
GEC-ESTRO working group with emphasis on target defini-
tion [5] and on use of dose volume histogram (DVH) param-
eters [6]. MRI is advocated as the preferred image modality
[7,8].
One of the main steps in the 3D image based procedure is
applicator reconstruction, which denotes the procedure of
defining the applicator source channels in the images. This
step is crucial since the dose calculation is based on the
geometry of source positions. When using CT for imaging,
the source channels are readily visualised either by using
X-ray markers or by exploiting the contrast between the
air in the source channels and the applicator material.
However, with MR images the visualisation of the applicator
is more challenging. The source channel cannot be discrim-
inated from the applicator material since both structures
appear dark in the images. Insertion of dummy catheters
containing fluid contrast material (for instance water, oil,
gadolinium or copper sulphate) into the source channels is
complicated by the fact that the diameter at the entrance
of the source channel is only around 2 mm in diameter for
most kinds of commercial intracavitary applicators for cer-
vical cancer [9]. The low signal intensity emitted from this
small volume complicates a precise definition of the entire
source path. Furthermore, MR compatible titanium applica-
tors may induce geometric distortions in the region around
the titanium material [10].
Applicator reconstruction for MRI based brachytherapy
can either be performed directly in the image series used
for contouring and dose planning (at the moment mainly
T2 weighted MRI), or it can be based on reconstruction in
images with better visualisation of the source channels


0167-8140/$ - see front matter c 2008 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.radonc.2008.06.010
 K. Tanderup et al. / Radiotherapy and Oncology 89 (2008) 156–163 157

(e.g. radiographs, CT, 3D MRI or T1 weighted MRI) followed
by fusion to the contouring/dose planning series (T2
weighted MR images).
Reconstruction uncertainties can be divided into two
categories: systematic and random uncertainties. System-
atic uncertainties are the ones that have the same influ-
ence in all brachytherapy fractions. Examples are MR
distortions and incorrect reconstruction procedures. Ran-
dom uncertainties are the ones that differ between brach-
ytherapy sessions. They are related to the visibility of the
applicator in the images and to the accuracy of image fu-
sion if this is applied during applicator reconstruction. MR
distortions can also add to random uncertainties when re-
lated to patient specific interaction with the magnetic
field. Reconstruction uncertainties are generally most pro-
nounced in the direction perpendicular to slice orientation
(longitudinal direction) due to the large pixel size in this
direction. The limited resolution reduces the visibility of
the applicator, and uncertainties of image fusion will also
be most pronounced in this direction. This direction is
along the uterine axis for para-transversal imaging (images
oriented according to the orientation of the applicator
[7]), and along the longitudinal axis of the patient in case
of transversal imaging.
Reconstruction uncertainties result in uncertainties on
DVH parameters. DVH parameters for tumour and organs
at risk (OAR) are essential throughout the 3D image based
procedure as they are used for target dose prescription,
during 3D dose plan optimisation, and to evaluate and re-
port dose distributions. The purpose of this study was to
evaluate the impact of random and systematic reconstruc-
tion uncertainties on DVH parameters.
and for the bladder, rectum, sigmoid, the D0.1cc, D2cc,
according to the guidelines of the GEC-ESTRO working group
[6]. Dose plans were analysed in BrachyVision (ver. 7.1.35,
Varian).
For the 10 cases with intracavitary applicators with no
interstitial component, translational reconstruction uncer-
tainties were simulated by maintaining the dwell times,
and translating the whole applicator (and dose distribution)
by ±3 mm in 3 spatial perpendicular directions: longitudinal
(along tandem axis), transversal, and anterior–posterior
(ant–post) (Fig. 1). Additionally, applicator translation of
±5 mm was performed in the longitudinal direction. Finally,
±15° rotation of the ring was applied in the transversal plane
– corresponding to a shift of ±4 mm for dwell positions in-
side a ring with a source channel diameter of 30 mm. DVH
parameters were read out for each translation/rotation.
For the 10 combined interstitial–intracavitary cases, only
shifts in longitudinal direction were performed.
Linear relation between applicator shift and DVH param-
eter shift was evaluated graphically for all displacements
and all patients.

Materials and methods

Impact of reconstruction uncertainties on DVH
parameters
Twenty patients (treated at General Hospital of Vienna)
with locally advanced cervical cancer were analysed in or-
der to evaluate the impact of reconstruction uncertainties
on DVH parameters. Radiotherapy was administered as a
combination of external beam radiotherapy (EBRT) (45 Gy
at 1.8 Gy per fraction) and 4 sessions of HDR brachytherapy
(7 Gy each session, prescribed to HR-CTV). Two different
applicators were used: (1) standard MRI compatible tandem
ring applicator (Nucletron, Veenendaal) made of carbon fi-
ber (10 patients), (2) combined interstitial and intracavitary
tandem ring applicator consisting of a modified ring applica-
tor and needles inserted directly into the tumour through
peripheral holes drilled in the ring [1,7] (10 patients). All
patients were scanned using MR (T2 weighted) with the
brachytherapy applicator in situ. Gross Tumour Volume
(GTV), High-Risk Clinical Target Volume (HR-CTV), bladder,
rectum and sigmoid were contoured in transversal slices
according to the guidelines of the GEC-ESTRO working group
[5]. Manual 3D dose planning was performed to optimise
dose coverage of the target and to spare OAR’s. The follow-
Fig. 1. (A) Translations of ring applicator in longitudinal (±3 mm,
ing DVH parameters were used in the quantification: for the
±5 mm), lateral (±3 mm), anterior/posterior (±3 mm) directions. (B)
targets, the D100 and D90 for both the HR-CTV and the GTV; Rotation in ring plane (±15°).
158 Random and systematic uncertainties in brachytherapy
Simulation of DVH parameters in a patient
population
A model for the variation of a specific DVH parameter for
a selected organ over a sequence of fractions is described
now.
The change of the DVH parameter in the brachytherapy
fraction k when the applicator is displaced by Dxi (i = 1, 2,
3, 4 for longitudinal, lateral, ant–post, rotation) is denoted
by DDqi, k where Dq denotes D100, D90, D0.1cc or D2cc.
When the DVH parameter shift is proportional to applica-
tor displacement, the relative change of Dqi, k can be ex-
pressed as (the coefficient b depends on the organ):
DDqi;k
¼ bi Dx i;k
Dqk
For some organs and directions of applicator displace-
ment the linear relationship of Eq. (1) cannot describe the
DVH parameter shift. For instance a symmetrical HR-CTV
(with regard to tandem) will show decrease in D100 and
D90 for both left/right and ant/post applicator shifts. In
such cases the DVH parameter shift will be proportional to
the norm of the displacement:
DDqi;k
¼ bi jDx i;k j
Dqi;k
The proportion of patients with linear (Eq. (1)) and non-
linear (Eq. (2)) relationship, respectively, was determined
for all DVH parameters and all directions.
Dqk is the planned value of the DVH parameter for brachy-
therapy fraction k and is assumed to be Dq/n, where Dq is the
planned value for the whole treatment, and n is the number of
fractions. Thus the relative change of the DVH parameter for
the displacement in the ith direction in the kth brachytherapy
fraction to the total planned value of the DVH parameter is:
DDqi;k 1 DDqi;k 1
¼ bi Dx i;k or ¼ bi jDx i;k j
Dq n Dq n lb, i and rb, i were estimated from the distribution of b’s.
Summed over all fractions and directions the total rela-
tive change of DVH parameters is (in case of linear relation-
ship for all directions):
DDq X n X 4
1 used at Aarhus University Hospital. Reconstruction was per-
¼ bi Dxi;k
Dq k¼1 i¼1
n formed in 10 patients by two independent observers in T2
In case there are non-linear components this is incorpo-
rated in Eq. (4) according to the proportion of patients.
Reconstruction uncertainties were assumed to be distrib-
uted according to a normal distribution with the mean equal
to the systematic reconstruction error and the variance cor-
responding to the random reconstruction uncertainty:
Dx i  Nðlx;i ; r2x;i Þ
Across the patient population, the coefficients corre-
sponding to each organ were assumed to be distributed
according to a normal distribution:
bi  Nðlb;i ; r2b;i Þ
The distribution of DDq/Dq is a sum of products of two
normal distributions according to Eq. (4). Since this distribu-
tion cannot be described in a simple analytic way, numerical
simulations were undertaken to assess the distribution of
DDq/Dq. In house developed software written in R (ver
2.5.0, http://www.r-project.org/) was used for numerical
simulations.
The distribution of the DVH parameters depends on four
parameters in each direction i: lx,i, rx,i, lb,i, rb,i. The
choice of parameters will be discussed below under the
heading ‘‘Input parameters for the model’’. First, we will
explain how we summarize the result of a simulation. For
each DVH parameter and a choice of parameters 10.000 sim-
ulation were run – representing 10.000 patients in a patient
population. The result of the simulation is a whole distribu-
tion of relative DVH parameter shifts which quantify how
much the delivered dose deviates from the expected dose.
ð1Þ
In order to summarize this distribution into a single number
we have chosen to focus on the 90 % of the patient popula-
tion with least unfavourable deviations. This is similar to the
approach used by van Herk et al. [11]. For target structures
under-dosage should be avoided and we calculate the mini-
mum relative Dq for 90% of the patient population. In this
case, our summary of the target DVH distribution could be
for instance that ‘‘D100 (or D90) will be at least 94% of ex-
pected dose for 90% of the patients’’. For OAR over-dosage
must be avoided, and we calculate the maximum relative Dq
ð2Þ
for 90% of the patient population – for instance ‘‘D0.1cc (or
D2cc) will not exceed 107% of expected dose for 90% of the
patients’’. Brachytherapy schedules of 1 and 4 fractions
were simulated.
Input parameters for the model
The linear coefficients, b, were estimated from the re-
sults by linear regression for those parameters depending
linearly on applicator displacement (Eq. (1)). For the pa-
tients showing decreasing DVH parameter for applicator dis-
placements in both positive and negative directions (Eq.
(2)), the b was estimated as the mean decrease per mm.
ð3Þ
Random uncertainties depend on the reconstruction and
imaging procedure. Random uncertainties corresponding
to ring rotation and lateral/ant–post displacements have
been evaluated for the applicator reconstruction method
ð4Þ
images with 5 mm slice thickness (aided by fusion with T1
images (3 mm slice thickness)). The difference between
the applicator position for the two observers was assessed
for the 10 patients resulting in a standard deviation of
0.5 mm in both lateral and ant–post direction and 7° (corre-
sponding to 1.8 mm) for the rotation. Simulations were per-
formed using these values for random uncertainties. In
ð5Þ order to investigate whether DVH parameter shifts would in-
crease drastically with increased random uncertainties in
lateral and ant–post directions, simulations were repeated
with 0.8 mm SD in these directions. With 0.8 mm SD more
than 1/3 of the reconstructions would be wrong by more
ð6Þ than 1 mm in lateral or ant–post directions. Systematic
uncertainties of both 0 mm and 2 mm were simulated for
lateral and ant–post directions.
Reconstruction uncertainties in the longitudinal direction
depend on slice thickness. The impact of reconstruction
 K. Tanderup et al. / Radiotherapy and Oncology 89 (2008) 156–163
uncertainties in the longitudinal direction was investigated
in more detail by performing a series of simulations where
systematic and random uncertainties were varied in the
ranges: [  5 mm, 5 mm] (l) and [0.5 mm, 3 mm] (r) in
steps of 1 and 0.5 mm, respectively.
In this study, we considered 10% to be an acceptable limit
for deviation of DVH parameters from expected value. This
means that the delivered dose should be at least 90% of ex-
pected dose for target structures, and should not exceed
110% of expected dose in OAR. Reconstruction uncertainties
giving rise to at least 90% of the patients satisfying the limit
of 10% were considered as acceptable reconstruction
uncertainties.
Results
Impact of reconstruction uncertainties on DVH
parameters
There was no significant statistical difference between
the effect of longitudinal applicator displacement for intra-
cavitary and intracavitary–interstitial applicators (t-test:
p > 0.05) except for D0.1cc for the sigmoid (t-test: p =
0.012). In this case, the intracavitary–interstitial implants
showed a larger impact of longitudinal applicator displace-
ment with 2.9% pr mm as compared to 1.9% per mm. This
is explained by the fact that the tips of interstitial needles
are located closer to the sigmoid than the intracavitary
applicator. All further data presented belongs to the 10
intracavitary cases and these were also used for the model-
ling of DVH parameters in a patient population.
Fig. 2 shows examples of the percentage change of DVH
parameters as a function of intracavitary applicator dis-
placement for D90 of HR-CTV, and D2cc of rectum. DVH
Fig. 2. Relative DVH shift (%) as a function of applicator displacement for D90 in HR-CTV and for D2cc in rectum. Positive direction of x-axis
corresponds to: applicator displacement in cranial, posterior and dexterior directions, respectively.
159
shifts depended linearly on applicator translation and rota-
tion in the majority of the patients. Good linear relation was
seen in more than 90% of the cases for both GTV and OAR’s.
For HR-CTV there was a non-linear relationship for displace-
ments in lateral and in ant–post directions for about half of
the patients, as can be seen in Fig. 2. In these cases the DVH
parameters decreased whichever direction the applicator is
moved into, and the dependence on applicator displace-
ment could be well described according to Eq. (2).
Absolute mean DVH shifts (%) per mm applicator dis-
placement are shown in Fig. 3 for intracavitary applicators.
The largest shifts of DVH parameters appeared for bladder
and rectum when the applicator was displaced in ant–post
direction, with a mean change of 5% and 6% per mm for
D2cc and D0.1cc, respectively. For all other parameters and
directions the mean shift was below 4% pr. mm. It was a
general trend that the influence of applicator shifts was lar-
ger for D100 and D0.1cc parameters as compared to D90 and
D2cc, respectively. Rotation in transversal plane had limited
impact for all target and OAR DVH parameters.
Modelling of DVH parameters in a patient
population
Actual DVH parameters relative to expected values are
shown in Table 1 for the following random uncertainties
(SD): 0.5 mm (lateral), 0.5 mm (ant–post), 1.25 mm (longi-
tudinal), and 7° (rotational). The minimum dose for 90% of
the patients is shown for target DVH parameters, and the
maximum dose for 90% of the patients is shown for OAR
DVH parameters. When systematic reconstruction uncer-
tainties are absent, 90% of the patients will receive at least
92–98% of expected dose (D100 and D90) to the GTV and
HR-CTV, and the dose to OAR (D0.1cc and D2cc) will not ex-
ceed 102–109% of expected dose for 90% of the patients.
160 Random and systematic uncertainties in brachytherapy

errors (<10% deviation for at least 90% of the patients).

Examples are shown for HR-CTV (D90) and rectum (D2cc)
for both 1 and 4 fractions of brachytherapy. Tolerances on
reconstruction uncertainties are tighter for rectum than
for GTV, HR-CTV, bladder and sigmoid, and also tighter
for 1 fraction as compared to multiple fraction schedules.

Fig. 3. DVH shift (%) per mm of applicator displacement (absolute
mean and standard deviation) for GTV, HR-CTV: D90, and for
rectum, bladder, sigmoid: D2cc.
Lateral shifts and partly longitudinal shifts of 2 mm can be
tolerable whereas systematic shifts in ant–post direction
lead to significant changes. D100 is more vulnerable to
reconstruction uncertainties than D90 for both GTV and
HR-CTV, and D0.1cc is more vulnerable than D2cc for all
OAR’s. Increasing random uncertainties from 0.5 to
0.8 mm (SD) in lateral and ant–post directions (data not
shown) caused an increase/decrease of less than 1 percent-
age point for all parameters, except for bladder and rectum
D0.1cc, where there was an increase of almost 2 percentage
points. Impact of reconstruction uncertainties on DVH
parameters is larger for 1 fraction than for 4 fraction
schedules.
In Fig. 4, the delivered dose (relative to expected values)
for 90% of the patients is shown as a function of longitudinal
systematic and longitudinal random reconstruction uncer-
tainties. Reconstruction uncertainties in the red areas will
lead to significant deviations (>10%) of DVH parameters for
a significant fraction of the patients (>10%), whereas the
green areas indicate a ‘‘safe’’ region for reconstruction
Discussion
In this study it was shown that target and OAR DVH
parameters for 90% of the patients deviate with less than
10% when systematic reconstruction uncertainties are
avoided. However, if present, systematic errors have signif-
icant impact on DVH parameters, with errors in ant–post
direction being particularly critical. Errors in this direction
may occur if wrong applicator dimensions are used when
defining the source channels and dwell positions related to
the outer applicator surface. Another source of systematic
errors in this direction could be geometric distortions in
MR with the use of titanium applicators. Such distortions
have been described in detail for prostate seeds [10], and
for intracavitary applicators [9].
Systematic errors can – to a large extend – be avoided
by quality control. Hellebust et al. [12] pointed out that ac-
tual source positions may systematically deviate by up to
2.5 mm from the indications of commercial X-ray marker
wires in the ring. Source positioning in the applicator should
be assessed by autoradiography [12,13], and this informa-
tion should be implemented into the reconstruction proce-
dures. The geometry of the applicator should be assessed
by 3D phantom scans to verify the relation between the out-
er surface of applicator and the source channels. Image dis-
tortions with titanium applicators in MR should be
investigated by fusion of CT and MR phantom scans. Finally,
the procedures in the dose planning systems should be well

Table 1
Minimum dose for 90% of the patients is shown for target DVH parameters, and maximum dose for 90% of the patients is shown for OAR DVH
parameters, relative to expected values
1 Fraction 4 Fractions
Systematic uncertainty (mm) Lateral 0 2 0 0 0 2 0 0
Ant–post 0 0 2 0 0 0 2 0
Long. 0 0 0 2 0 0 0 2
GTV D100 92% 87% 88% 90% 96% 90% 90% 92%
D90 96% 92% 90% 95% 98% 93% 91% 96%
HR-CTV D100 94% 88% 86% 88% 97% 90% 88% 91%
D90 95% 89% 91% 89% 97% 91% 93% 92%
Rectum D0.1cc 109% 111% 123% 118% 105% 107% 119% 114%
D2cc 107% 108% 118% 115% 104% 105% 114% 111%
Bladder D0.1cc 106% 109% 120% 110% 103% 106% 117% 108%
D2cc 105% 106% 116% 109% 102% 104% 115% 107%
Sigmoid D0.1cc 106% 108% 113% 110% 103% 106% 112% 108%
D2cc 105% 106% 111% 109% 102% 105% 110% 107%
Parameters are listed for 1 and 4 fraction schedules, and for systematic uncertainties of 0 and 2 mm in different directions. The following
random uncertainties (SD) were used: 0.5 mm (lateral); 0.5 mm (ant–post); 1.25 mm (longitudinal); 7° (rotational). For instance 90% of the
patients are ensured at least 89% of planned D90 to HR-CTV with a systematic uncertainty of 2 mm in longitudinal direction for a 1 fraction
schedule.
 K. Tanderup et al. / Radiotherapy and Oncology 89 (2008) 156–163 161

Fig. 4. The curves show the percentage dose (relative to expected values) that 90% of patients will receive as a function of random and
systematic longitudinal reconstruction uncertainties. (A and B), 1 fraction schedule; (C and D), 4 fraction schedule. Reconstruction errors in
the red area should be avoided (>10% dose deviation for at least 10% of the patients). The green area indicates a ‘‘safe’’ region of longitudinal
reconstruction errors (<10% dose deviation for at least 90% of patients).

tested. Correct geometry of library applicators is especially
important.
Random errors can be reduced by optimising the imaging
procedure, as improved visibility of the applicator evidently
reduces random reconstruction errors. Slice thickness should
be as small as possible without compromising the MR acquisi-
tion time since this would increase the risk of patient move-
ment during the image acquisition. In case image fusion is
used during reconstruction, it should be taken into account
that fusion uncertainties also adds to the random errors.
Fractionated brachytherapy is less vulnerable to recon-
struction uncertainties since random uncertainties tend to le-
vel out over several fractions. This means that tolerances on
reconstruction uncertainties are tighter for brachytherapy
schedules with few fractions. However, even for one fraction
schedules DVH parameters can be considered as stable.
The impact of longitudinal reconstruction uncertainties
on DVH parameters for the combined interstitial–intracavi-
tary applicator was very similar to the standard tandem-ring
applicator. Only slightly larger impact was seen for D0.1cc for
the sigmoid, because dose gradients are steeper in the
vicinity of needles. Lateral, ant–post displacements and
rotations were not simulated for this applicator. However,
reconstruction uncertainties in these directions are ex-
pected to be smaller with this applicator, because the signal
(or rather: lack of signal) from the needles helps to guide
the positioning of the applicator.
A limitation of this study is that DVH parameters were
simulated with the use of random reconstruction uncertain-
ties specific for our department. However, a moderate in-
crease of random uncertainties in lateral and ant–post
directions had relatively limited impact on the accuracy of
most DVH parameters (less than 1 percentage point). On
the other hand, it should be emphasised that for reconstruc-
tion procedures associated with much larger random uncer-
tainties, the accuracy of DVH parameters evaluated in this
study will no longer be valid.
Applicability of results with CT imaging and with
other applicators
This study addresses the ring applicator used with MR
imaging, but the results are also relevant for CT imaging
and partly also for other applicators. The impact of recon-
struction uncertainties on DVH parameters does not depend
on image modality, and would be the same for CT. However,
reconstruction uncertainties are smaller for CT which gen-
erally leads to more stable DVH parameters for CT based
dose planning. Hellebust et al. [12] evaluated dose in points
close to target (point A), bladder and rectum. They found
little variability of point doses for different reconstruction
methods and applicator orientations. This confirms that
reconstruction uncertainties are small with CT. However,
it should be kept in mind that CT is inferior to MR with re-
gard to target delineation [8].
162 Random and systematic uncertainties in brachytherapy
The impact of applicator displacement on DVH parame-
ters is determined by the dose gradients in the periphery
of the target and adjacent to OAR’s. The pear shaped iso-
dose curves and the dose gradients are in general quite
similar for different kinds of applicators. This would mean
that the impact of reconstruction uncertainties derived in
this paper is expected to apply to other kinds of applica-
tors such as mould, and tandem-ovoid applicators. How-
ever, if the geometry is generally very different from
that in our patient population (packing, bladder filling
etc.), then the linear coefficients, b, may be different.
This can only be tested by repeating the same kind of
investigation in another patient population. Furthermore,
the configuration of source positions differs between appli-
cators, and the pattern of reconstruction uncertainties can
differ. The results of this study are only applicable if
reconstruction uncertainties are similar to those simulated
in this study.
Comparison with EBRT and relevance of margins in
brachytherapy
Applicator reconstruction defines the relation between
applicator and target/OAR’s. Reconstruction uncertainties
in brachytherapy are of the same type as the uncertainties
in EBRT which are related to positioning the target/OAR’s
correctly with respect to the isocenter of the accelerator.
Uncertainties in EBRT can also be divided into random and
systematic uncertainties as discussed by Van Herk et al.
[11]. In conventional EBRT, these uncertainties are compen-
sated for by adding a margin to the CTV. The size of the pre-
scription dose plateau is thereby increased, and the dose
delivered to the CTV will be unaffected by geometric errors
which are smaller than the size of the margins. In brachy-
therapy it is not possible to create a dose plateau in lateral
and ant–post directions, since the dose gradients cannot be
manipulated to become less steep. Application of margins in
lateral and ant–post directions will be equivalent to a dose
escalation, and the uncertainty of DVH parameters will re-
main the same since relative dose gradients are not changed
with dose escalation. The use of margins in brachytherapy
can not be based on a simple translation of concepts from
EBRT, and the concept of creating a PTV around the CTV
by applying a margin is therefore not appropriate in
brachytherapy.
In specific situations, where the applicator geometry al-
lows the use of additional dwell positions above the target,
a homogenous dose plateau can partly be obtained in longi-
tudinal direction. This creates a robustness of the dose plan
in the longitudinal direction where the reconstruction
uncertainties are largest. At the moment it might be appro-
priate to use standard loadings in intracavitary brachyther-
apy to start dose shaping [1]. These standard loadings with
dwell positions in the vaginal applicators and up to the tip of
the tandem lead to the typical pear shaped isodose. This
dose distribution might over-treat parts of the uterus cra-
nial to the CTV, but it is more stable related to longitudinal
movements. It seems reasonable to keep the dose distribu-
tion from the standard loading and change it only at regions
where dose constraints have to be fulfilled. Conformity
might be less important than a good coverage.
In EBRT, van Herk’s margin recipe [11] can be applied to
ensure that the minimum target dose is at least 95% for at
least 90% of the patients. However, in clinical practice,
margins are often smaller in the direction of OAR’s – for in-
stance towards the rectum in prostate EBRT – and this will
lead to larger uncertainties in target dose [14]. This study
showed that for target dose (D100 and D90) in brachyther-
apy at least 92–98% of expected dose can be obtained for
90% of the patients when systematic reconstruction errors
are controlled. However, contouring uncertainties were
not included in this analysis, and the impact of these may
be larger in brachytherapy than in EBRT due to the steep
dose gradients of brachytherapy.
Whereas EBRT positioning uncertainties and target DVH
parameters has been the subject of many investigations the
same aspect has not been comprehensively investigated for
OAR’s. The uncertainty of DVH parameters for OAR’s in EBRT
is complicated since it depends on both the topography and on
the dose distribution [15,16]. In future work, it would be
interesting to evaluate the impact of EBRT planning and
set-up uncertainties on the dose distribution in pelvic OAR
and to compare this with the accuracy of brachytherapy.
Conclusion
Comprehensive quality control of applicators and imag-
ing procedures should be used to prevent systematic
uncertainties in applicator reconstruction. Systematic
reconstruction uncertainties can lead to significant uncer-
tainties on DVH parameters. Furthermore, random errors
should be minimised by using small slice thickness
(65 mm) and by having clear and reproducible guidelines
for reconstruction. With careful applicator reconstruction
procedures, the related uncertainty on DVH parameters
for both target structures and OAR’s can be kept below
5–10% for schedules of 1 brachytherapy fraction and 2–
5% for 4 fraction schedules.
Acknowledgements
Asena Kuzucan, Vienna Medical University, is acknowledged for
her comprehensive work at the dose planning system. Varian has
kindly supported Vienna Medical University with free software.
Funding was received from Danish Cancer Society, Agnes Nie-
buhr Andersson Cancer Research Foundation, Irma and Kaj Morten-
sens Memorial Foundation, DAFKO (Danish Graduate School in
Clinical Oncology).
* Corresponding author. Christian Kirisits, Department of Radio-
therapy, Medical University of Vienna, Währinger Gürtel 18-20, A –
1090 Vienna, Austria. E-mail address: Christian.Kirisits@meduniwie-
n.ac.at
Received 25 October 2007; received in revised form 16 June 2008;
accepted 19 June 2008; Available online 7 August 2008
References
[1] Kirisits C, Pötter R, Lang S, Dimopoulos J, Wachter-Gerstner N,
Georg D. Dose and volume parameters for MRI-based treat-
ment planning in intracavitary brachytherapy for cervical
cancer. Int J Radiat Oncol Biol Phys 2005;62:901–11.
 K. Tanderup et al. / Radiotherapy and Oncology 89 (2008) 156–163
[2] Lindegaard JC, Tanderup K, Nielsen SK, Haack S, Gelineck J.
MRI-guided 3D optimization significantly improves DVH param-
eters of pulsed-dose-rate brachytherapy in locally advanced
cervical cancer. Int J Radiat Oncol Biol Phys 2008;71:756–64.
[3] De Brabandere M, Mousa AG, Nulens A, Swinnen A, Van
Limbergen E. Potential of dose optimisation in MRI-based PDR
brachytherapy of cervix carcinoma. Radiother Oncol 2007.
[4] Pötter R, Dimopoulos J, Georg P, et al. Clinical impact of MRI
assisted dose volume adaptation and dose escalation in
brachytherapy of locally advanced cervix cancer. Radiother
Oncol 2007;83:148–55.
[5] Haie-Meder C, Pötter R, Van Limbergen E, et al. Recommen-
dations from gynaecological (GYN) GEC-ESTRO working group
(I): concepts and terms in 3D image based 3D treatment
planning in cervix cancer brachytherapy with emphasis on MRI
assessment of GTV and CTV. Radiother Oncol 2005;74:235–45.
[6] Pötter R, Haie-Meder C, Van Limbergen E, et al. Recommen-
dations from gynaecological (GYN) GEC ESTRO working group
(II): concepts and terms in 3D image-based treatment planning
in cervix cancer brachytherapy-3D dose volume parameters
and aspects of 3D image-based anatomy, radiation physics,
radiobiology. Radiother Oncol 2006;78:67–77.
[7] Dimopoulos JC, Schard G, Berger D, et al. Systematic evalu-
ation of MRI findings in different stages of treatment of
cervical cancer: potential of MRI on delineation of target,
pathoanatomic structures, and organs at risk. Int J Radiat
Oncol Biol Phys 2006;64:1380–8.
[8] Viswanathan AN, Dimopoulos J, Kirisits C, Berger D, Pötter R.
Computed tomography versus magnetic resonance imaging-
based contouring in cervical cancer brachytherapy: results of a
163
prospective trial and preliminary guidelines for standardized
contours. Int J Radiat Oncol Biol Phys 2007;68:491–8.
[9] Haack S, Lindegaard JC, Nielsen SK, Gelineck J, Tanderup K.
Applicator reconstruction in MRI 3D image-based dose planning
of brachytherapy for cervical cancer. Brachytherapy
2008;7:167.
[10] Wachowicz K, Thomas SD, Fallone BG. Characterization of the
susceptibility artifact around a prostate brachytherapy seed in
MRI. Med Phys 2006;33:4459–67.
[11] van Herk M, Remeijer P, Rasch C, Lebesque JV. The probability
of correct target dosage: dose-population histograms for
deriving treatment margins in radiotherapy. Int J Radiat Oncol
Biol Phys 2000;47:1121–35.
[12] Hellebust TP, Tanderup K, Bergstrand ES, Knutsen BH, Røislien
J, Olsen DR. Reconstruction of a ring applicator using CT
imaging: impact of reconstruction method and applicator
orientation. Phys Med Biol 2007;52:4893–904.
[13] Venselaar J, Pérez-Calatayud J. A practical guide to
quality control of brachytherapy equipment, ESTRO,
Brussels: 2004.
[14] de Crevoisier R, Tucker SL, Dong L, et al. Increased risk of
biochemical and local failure in patients with distended
rectum on the planning CT for prostate cancer radiotherapy.
Int J Radiat Oncol Biol Phys 2005;62:965–73.
[15] McKenzie A, van Herk M, Mijnheer B. Margins for geometric
uncertainty around organs at risk in radiotherapy. Radiother
Oncol 2002;62:299–307.
[16] Stroom JC, Heijmen BJ. Limitations of the planning organ at
risk volume (PRV) concept. Int J Radiat Oncol Biol Phys
2006;66:279–86.