Int. J. Radiation Oncology Biol. Phys., Vol. 67, No. 2, pp.

630 – 638, 2007

Copyright © 2007 Elsevier Inc.
Printed in the USA. All rights reserved
0360-3016/07/$–see front matter

doi:10.1016/j.ijrobp.2006.10.007

PHYSICS CONTRIBUTION

CORRELATION BETWEEN INTERNAL FIDUCIAL TUMOR MOTION AND

EXTERNAL MARKER MOTION FOR LIVER TUMORS IMAGED WITH 4D-CT

A. SAM BEDDAR, PH.D.,* KRISTOFER KAINZ, PH.D.,* TINA MARIE BRIERE, PH.D.,*
YOSHIKAZU TSUNASHIMA, M.S.,* TINSU PAN, PH.D.,† KARL PRADO, PH.D.,* RADHE MOHAN, PH.D.,*
MICHAEL GILLIN, PH.D.,* AND SUNIL KRISHNAN, M.D.‡
Departments of *Radiation Physics, †Diagnostic Imaging, and ‡Radiation Oncology, The University of Texas M. D. Anderson
Cancer Center, Houston, TX

Purpose: We investigated the correlation between the motions of an external marker and internal fiducials
implanted in the liver for 8 patients undergoing respiratory-based computed tomography (four-dimensional CT
[4D-CT]) procedures.
Methods and Materials: The internal fiducials were gold seeds, 3 mm in length and 1.2 mm in diameter. Four
patients each had one implanted fiducial, and the other four had three implanted fiducials. The external marker
was a plastic box, which is part of the Real-Time Position Management System (RPM) used to track the patient’s
respiration. Each patient received a standard helical CT scan followed by a time-correlated CT-image acquisition
(4D-CT). The 4D-CT images were reconstructed in 10 separate phases covering the entire respiratory cycle.
Results: The internal fiducial motion is predominant in the superior–inferior direction, with a range of 7.5–17.5
mm. The correlation between external respiration and internal fiducial motion is best during expiration. For 2
patients with their three fiducials separated by a maximum of 3.2 cm, the motions of the fiducials were well
correlated, whereas for 2 patients with more widely spaced fiducials, there was less correlation.
Conclusions: In general, there is a good correlation between internal fiducial motion imaged by 4D-CT and
external marker motion. We have demonstrated that gating may be best performed at the end of the respiratory
cycle. Special attention should be paid to gating for patients whose fiducials do not move in synchrony, because
targeting on the correct respiratory amplitude alone would not guarantee that the entire tumor volume is within
the treatment field. © 2007 Elsevier Inc.

4D-CT imaging, Tumor motion, Respiratory gating, Internal fiducials, Hepatobiliary Cancer.

INTRODUCTION motion: (1) direct or indirect tumor monitoring with internal

Accounting for tumor motion in treatment planning and
delivery is one of the most recent and significant challenges
facing radiotherapy for the abdomen and thorax. Respirato-
ry-gated radiotherapy can be used to achieve a high dose
conformity around the target while reducing the overall
irradiated volume so as to reduce normal tissue complica-
tions (1– 4). Initial efforts were directed toward lung tumors
because of the tumor’s obvious respiratory motion, the poor
prognosis for lung cancer, and the deleterious effects of
radiotherapy on normal lung tissue including pneumonitis
and fibrosis (3). Recently, however, there is burgeoning
interest in respiratory gating for liver tumors as well. Gated
radiotherapy for the liver is still in its infancy in the United
States, and to our knowledge this technique has been used in
only a few institutions worldwide (1, 2, 5–10).
Two basic techniques have been used to track tumor
fiducials and fluoroscopic imaging (2, 5) and (2) indirect
tumor monitoring with external markers and/or respiratory
sensors (6, 11). The first technique enables gated treatments
because the tumor motion can be tracked nearly in real time.
The second method can be used for time-correlated com-
puted tomography (CT)-image acquisition (four-dimen-
sional CT [4D-CT]) imaging (12, 13) as well as gated
therapy but relies heavily on the ability of the respiratory
sensors to predict the tumor location reproducibly.
Although several studies have clarified the relationship
between the motion of an external marker and an internal
fiducial for specific breathing sensors (4, 10), these relation-
ships may differ for each type of sensor. In addition, most
studies have been performed with fluoroscopic units and not
the CT images used for treatment planning. The primary
purpose of this study was to investigate the correlation
between the external respiratory signal obtained using the

Reprint requests to: A. S. Beddar, Ph.D., Department of Radi- mdanderson.org

ation Physics, Unit 94, University of Texas M. D. Anderson Conflict of interest: none.
Cancer Center, 1515 Holcombe Boulevard, Houston, TX 77030. Received Aug 9, 2006, and in revised form Sept 29, 2006.
Tel: (713) 563-2609; Fax: (713) 563-2620; E-mail: abeddar@ Accepted for publication Oct 2, 2006.

630
 Correlation between internal fiducial tumor motion and external marker motion in 4D-CT for liver tumors ● A. S. BEDDAR et al. 631

Real-Time Position Management System (RPM, Varian
Medical Systems, Palo Alto, CA) and tumor motion for
liver cancer. Our second goal was to determine how the
motions of multiple fiducials correlate with each other when
implanted in the liver. Our third aim was to establish quali-
tative criteria to identify those patients who could benefit from
respiratory-gated radiotherapy. Finally, we hoped to demon-
strate the value of 4D-CT imaging for the management of
tumor motion for gated as well as nongated treatment delivery.
METHODS AND MATERIALS
Terminology: markers and fiducials
Throughout the discussion that follows, the following terminol-
ogy is used to distinguish the reference points that are external to
the patient from those that are internal. Traceable reference points
exterior to the patient are referred to as external markers or simply
markers and are described subsequently, whereas reference points
located inside the patient are referred to as fiducials.
Implanted fiducials consisted of gold cylinders used with the
ACCULOC image-guided radiation therapy (IGRT) system (Med-
Tec, Orange City, Iowa); each cylinder had a diameter of 1.2 mm
and a length of 3 mm. The high density of gold enables the
fiducials to be readily discernible within the CT images (albeit
with some beam-hardening artifacts), as well as within radiographs
obtained using 6-MV photon beams for gated treatment.
For some of our liver patients who received 4D-CT scans, postop-
erative surgical clips or stents were present. It was possible to localize
accurately the surgical clips or the tips of the stents within the CT data
as well, and we therefore also designated them as fiducials.
Brief description of the Varian RPM as the “external
marker” system
We presume that the motion of an external marker placed on the
patient is correlated primarily with the respiratory component of
the patient’s internal motion. The marker is an infrared reflective
circle mounted on one face of a plastic box, which is taped to the
patient’s abdomen. This box is a component of the Varian RPM
system and has been described before (14 –16). Additional com-
ponents include an infrared-sensitive video camera, which records
the location and motion of the marker, and a program that tracks
and later displays the location and motion of the marker. Data for
the marker’s motion with time (referred to hereafter as the trace)
can be monitored in real time during imaging or treatment proce-
dures or can be saved in ASCII format for offline analysis. The
system is shown in Fig. 1.
Description of the 4D-CT data acquisition
A thorough description of the techniques used to obtain and
reconstruct 4D-CT data are given by Pan et al. (12). The process
used in this study is briefly reviewed here. Before the 4D-CT
acquisition, the RPM box is placed on the patient’s abdomen at a
location that exhibits the maximum motion in the AP direction. A
series of axial CT scans in cine mode are obtained at consecutive
couch positions spaced 2 cm apart. The scanning duration at each
couch position must be long enough to encompass at least one full
cycle respiratory cycle. The images are reconstructed to produce at
least 10 CT images that can be binned into 10 phases covering the
entire breathing cycle. The cine CT images are subsequently sent,
along with the RPM trace recorded during the acquisition, to a GE
Advantage Workstation (AW) for processing. The AW software

Fig. 1. The Real-Time Position Management System (RPM): (a) infrared camera, (b) reflective box placed on the
abdomen of a patient. Sample RPM respiratory traces: (c) the default spacing between phases, with the 0% and 50%
phases shown in green, (d) the same trace with phase correction so that the 50% phase corresponds to the end of
expiration, and (e) amplitude gating at the end of expiration.
632 I. J. Radiation Oncology ● Biology ● Physics
Fig. 2. Digitally reconstructed radiograph of Patient 8, showing an
example of the placement of the fiducials within the liver. The
three fiducials are red, the liver is contoured in yellow, and the
tumor is contoured in blue.
associates each reconstructed image with a point on the RPM trace.
In this way, the cine CT images are associated with specific phases
of the marker’s motion. The standard RPM software identifies the
peaks of the respiratory traces and then equally subdivides each
respiratory cycle into 10 phases. The peaks are denoted as the “0%
phase” and correspond to the end of inspiration; however, the
crests do not necessarily correspond to the “50% phase” because
each cycle is evenly divided in time (Fig. 1c). To improve the
correspondence between the phase identification with the actual
peak and crest of the patient’s respiratory cycle, for some patients
we have used an internal software package (17) to identify the
crest as being the 50% phase and then separately subdivide each
Fig. 3. Axial, coronal, and sagittal views of the anatomy of Patient 1 exhibiting the location of the fiducial marker in
the free-breathing computed tomography series (in blue) and in the 50% phase (in red) within the liver.
Volume 67, Number 2, 2007
half of the respiratory cycle (Fig. 1d). Thus, we correct for the
mismatch between the end of expiration and the 50% phase.
Among the reconstructed 4D-CT cine images, those that corre-
sponded best to phases of the marker motion (or to phases of the
respiratory cycle) were identified. The phase-bins were taken in
10-percent increments of the full marker-motion period.
We should note that the 4D images are used only for internal
target volume (ITV) tumor contouring. Contouring of normal tissues
and organs for treatment planning is performed on a normal CT
image set that is obtained from a free-breathing (FB) helical scan
just before the 4D image acquisition.
Technique to locate the centers of each fiducial in the
4D-CT data
First, the 10 4D-CT image sets were exported to the Pinnacle
treatment planning system (Philips Medical Systems, Bothell,
WA). For each fiducial in each CT image set, a point of interest
(POI) was defined. For the first 4 patients in our study, we used a
single fiducial, and for the last 4 patients, we used three fiducials.
A typical example is shown in Fig. 2. The axial CT slice that
corresponded to the center of the fiducial was identified; within
this slice, the POI was positioned at the center of the fiducial, along
the lateral and anterior–posterior (AP) directions (i.e., anatomic
transverse plane). The sagittal view of the CT image set was then
used to refine the centering of the POI along the superior–inferior
(SI) direction. An example is shown in Fig. 3. The position
uncertainty of each POI is greatest along the SI direction, due to
the 2.5-mm axial slice thickness. In the lateral and AP directions,
beam-hardening artifacts complicate the POI placement; it is ex-
pected that the uncertainty of selecting the location along these
directions is approximately 0.5 mm.
RPM trace analyses and evaluation method
To analyze the correlation of the external marker’s position to
that of the fiducial implanted within the liver, a means is necessary
to normalize the RPM trace data to compare it directly with the
fiducial location.
 Correlation between internal fiducial tumor motion and external marker motion in 4D-CT for liver tumors
The RPM trace data were processed in the following manner
with the use of a MATLAB routine. Of the entire RPM trace data
recorded during the 4D-CT scan, the only section of the RPM trace
considered for analysis consisted of the full free-breathing cycles
during which the 4D-CT scan acquisition took place. A “full
free-breathing cycle” comprises the region of the RPM trace
between consecutive phase values of zero; over the course of a
free-breathing cycle, the phase value ranges from zero to 2␲.
Within the ASCII data for the RPM trace, a TTL “on” signal sent
from the CT scanner console to the RPM computer indicated when
the CT scanner’s X-ray tube was on, and the RPM-trace ASCII
data were flagged accordingly (1 vs. 0). For each free-breathing
cycle within the RPM trace region considered for analysis, the 0%,
10%, 20%, . . . “time-in-cycle” points (corresponding to phase
values of 0, ␲/5, 2␲/5, . . .) were located. The amplitude (y axis
value) of the RPM trace was recorded at each of these points, and
for each phase value ␾ the average RPM amplitude 具RPMraw典␾
was calculated.
Meanwhile, the SI position of the fiducial was measured, in the
manner described earlier, within the 4D-CT phase image data sets
(0%, 10%, 20%, . . .) corresponding to the RPM trace data. For the
patients with three fiducials, the average SI position of the three
fiducials relative to each midpoint of motion was used. The un-
certainty in the phase value is a measure of how “close” a given
reconstructed 4D-CT phase image is to a particular “canonical”
phase value (0%, 10%, 20%, . . .). For example, the GE Advantage
Workstation (AW) may associate, with a canonical phase of 50%,
a reconstructed cine image whose time stamp is most consistent
with the 50% phase, but which may not correspond exactly to the
50% phase; the cine image could be anywhere between the 45%
and 55% phases. In this case the matching tolerance is 5%. In
principle, the uncertainty in the phase can be determined at each
phase value by reviewing the AW results. In practice, we recon-
struct and export enough images to AW to keep the matching
tolerance to within 3%.
Once the SI coordinate of the fiducial 4DCT␾ and the average
RPM amplitude 具RPMraw典␾ were determined for each phase value
␾, the RPM trace data were renormalized in the following manner.
First, a plot of 4DCT␾ vs. 具RPMraw典␾ was generated, an example
of which is shown in Fig. 4. We assumed that not only would there
be a one-to-one correspondence between 4DCT␾ and 具RPMraw典␾,
but also that the correlation would be linear, as suggested by
Gierga et al. (18). A linear fit was applied to the data, and the slope
m and intercept b were recorded for the fit:
4DCT⌽ ⫽ m ⫻ 具RPMraw典⌽ ⫹ b. (1)
Fig. 4. Example of the fiducial’s superior–inferior (SI) coordinate
as a function of the amplitude of the raw Real-Time Position
Management System (RPM) trace. A linear fit to this line is used
to generate the renormalized RPM trace.
● A. S. BEDDAR et al. 633
The parameters m and b were used as the scale and offset
terms to convert the “raw” RPM trace amplitudes at each time t,
RPMraw(t), to amplitudes RPMnorm(t) that, in principle, should
correspond directly to the SI coordinate of the fiducial within the
patient:
RPMnorm(t) ⫽ m ⫻ RPMraw(t) ⫹ b. (2)
Each full free-breathing cycle of RPMnorm(t) was plotted, and
the SI coordinates of the fiducial position and their uncertainties
were overlaid on this graph. Such a plot would be useful in
addressing the following issues:
1. The correspondence between the marker’s AP motion and the
fiducial’s SI motion. Better agreement at certain phase values
may suggest that the RPM amplitude would more reliably
predict the position of the fiducial (and thus the target) at those
phase values.
2. The legitimacy of assuming a linear correspondence between
4DCT␾ and 具RPMraw典␾.
3. The consistency of the patient’s AP motion, overall free-
breathing cycles recorded during the 4D-CT scan. Consider-
able spread in the amplitude of RPMnorm(t) might suggest
considerable uncertainty in the SI position of the fiducial.
RESULTS
A single fiducial was first identified within the liver for
each of the first 4 patients considered in this study. Within
each of the 4D-CT image sets for all 10 individual phases as
well as for the free-breathing helical CT scan, the coordi-
nates of the fiducial (in the SI, AP, and lateral directions)
were determined as described earlier and plotted in Fig. 5.
(Note that the scale on the ordinate of the SI displacement
is twice as large as the ordinates of the AP and lateral
displacements. Therefore, the AP and lateral displacements
are not as large as the SI direction as they may appear in the
figure.) The locations of the fiducials relative to the dome of
the patient’s liver and midcoronal and midsagittal planes are
shown in Table 1. Also shown in Table 1 is the range of
motion of the fiducials. The lateral motion of the fiducial
tends to be minimal, less than 3 mm over the course of a full
respiration. This motion also does not show a strong depen-
dence with the phase of the respiratory motion. In the AP
direction, the displacement ranged from 1.2 mm to 3.5 mm
for all 4 patients. For Patients 2 and 4, the motion does
appear to correlate roughly with the respiratory phase. The
SI motion of the fiducial is significant for all 4 patients,
ranging between 7.5 mm and 10.5 mm. Furthermore, the
extrema of the fiducial motion occurs near the 90 – 0% and
the 40 – 60% phase values.
Also plotted in Fig. 5 are the coordinates of the fiducial as
measured within the FB image data set (dashed line). The
fiducial coordinates in the FB image set are mostly within
the range of the 4D data points with a maximum deviation
of 2 mm. Some differences in the positions of the marked
isocenters determined from three external BBs placed on the
skin of the patients are also observed (Table 2). The largest
difference seen between the FB and 4D-CT isocenters is
634 I. J. Radiation Oncology ● Biology ● Physics Volume 67, Number 2, 2007

Fig. 5. Fiducial coordinates in the superior–inferior (SI), anterior–posterior (AP), and lateral (Lat) directions for the first
4 patients in this study. The symbols represent the fiducial’s position in the 10 phases of the time-correlated computed
tomography (CT) series, and the dotted line represents the fiducial’s position in the free-breathing CT series. The
coordinates were normalized to the midpoint of the extrema in each direction. Note the scale for the coordinates in the SI
direction is twice as large as in the AP and lateral directions. Note also that the positive fiducial motion in the SI direction
is inferior to the diaphragm.

Table 1. Patient data, including the locations of the fiducials with respect to the dome of the liver in the superior–inferior direction and
midsagittal and midcoronal planes

Fiducial location (cm) Range of motion (cm)

Patients Dome Midsagittal plane Midcoronal plane SI Lat AP

1 5.8 ⫺8.8 ⫺0.8 0.80 0.16 0.12

2 9.7 ⫺1.4 6.5 1.00 0.23 0.35
3 6.0 ⫺3.5 4.7 0.75 0.25 0.29
4 10.3 ⫺6.9 1.3 1.05 0.20 0.25
5 2.3, 2.8, 3.8 ⫺4.6, ⫺4.7, ⫺2.5 6.3, 3.1, 4.1 1.08 0.11 0.60
6 9.0, 9.0, 9.7 ⫺2.2, ⫺2.1, ⫺0.3 5.4, 6.4, 6.7 0.80 0.16 0.25
7 7.0, 8.9, 13.6 ⫺4.2, ⫺3.7, ⫺8.4 1.0, ⫺3.7, ⫺2.3 1.75 0.50 0.87
8 4.2, 9.6, 10.0 1.1, 2.5, ⫺4.8 6.8, 9.9, 8.4 0.93 0.35 0.84

Abbreviations: SI ⫽ superior-inferior; AP ⫽ anterior-posterior; Lat ⫽ lateral.

Patients 1– 4 each had one fiducial, whereas Patients 5– 8 had three fiducials. The distances were calculated at the 50% phase (end-expiration).
A positive distance from the midsagittal plane indicates placement in the left side of the body, and a negative distance indicates placement in the
right side. A positive distance from the midcoronal plane indicates anterior placement of the fiducial, and a negative distance indicates posterior
placement. All fiducials were inferior to the dome of the liver. The range of fiducial motion in the SI, lateral (Lat), and anterior–posterior (AP)
directions is also given. For Patients 5– 8, this represents the maximum value for the three fiducials in each direction.
 Correlation between internal fiducial tumor motion and external marker motion in 4D-CT for liver tumors ● A. S. BEDDAR et al. 635

Table 2. Differences between isocenters in the free-breathing Table 3. The calculated slope (m) and intercept (b) used for the
and respiratory-based CT scans calculation of the renormalized Real-Time Position Management
System trace
Patients ⌬SI (cm) ⌬AP (cm) ⌬Lat (cm)
Patients m b
1 0.00 0.05 0.05
2 0.25 ⫺0.25 ⫺0.32 1 2.06 ⫺3.55
3 0.03 0.04 ⫺0.09 2 0.79 ⫺1.53
4 0.20 0.06 ⫺0.04 3 1.34 ⫺1.05
5 0.00 0.04 ⫺0.07 4 1.51 0.69
6 0.25 ⫺0.08 ⫺0.38 5 1.39 ⫺10.4
7 ⫺0.25 0.15 ⫺0.25 6 1.38 1.95
8 0.00 0.02 0.05 7 1.39 ⫺1.33
8 0.62 ⫺2.85
Abbreviations: SI ⫽ superior–inferior; AP ⫽ anterior–posterior;
Lat ⫽ lateral.

trace amplitudes at near the 100% phase agree well with

about 3 mm for Patient 2 and coincides with a deviation in
the FB fiducial coordinates from the 4D coordinates, par-
ticularly in the AP and lateral directions. This indicates that
the patient probably shifted slightly during simulation.
Figure 6 shows the renormalized RPM trace data along
with the SI coordinates of the fiducial for the first 4 patients.
The slopes and intercepts used in the renormalization equa-
tion for all patients are shown in Table 3. For these patients,
the renormalized RPM trace amplitude tends to correlate
best with the fiducial position, within the uncertainty of the
fiducial position, from the 0% phase through the 60% phase
(from end-inspiration to end-expiration). The differences
between the RPM trace and the fiducial position are 0.03–
0.06 mm for the 0 – 60% phases and (–1.4)–(– 0.06) mm for
the 70 –90% phases. This would suggest that for these
patients, the RPM trace is a reliable predictor for the fiducial
position. In particular, gating treatments during phases the
40 – 60% phases, using the RPM trace amplitude as a guide,
should yield suitable localization of the fiducial and thus
the target volume during treatment. For Patients 1 and 3, the
renormalized RPM amplitude underpredicts the fiducial po-
sition during inspiration. Given that the renormalized RPM
their amplitudes at the 0% phase, we may suspect that the
linear correlation between fiducial position and marker po-
sition breaks down in some fashion between end-expiration
and end-inspiration. A possible explanation for this general
trend may be that, during inhalation, the AP motion of the
abdomen lags behind the SI motion of the diaphragm and
the liver. Perhaps, during inhalation, the chest tends to
expand before the abdomen expands. Also, there may be
components of abdominal motion unrelated to respiration,
such as muscle tightening or loosening, that may be present
during either inhalation or exhalation. It should be noted
that if only the phase values corresponding to inhalation
(60 –90%) were used in the linear fit described by Eq. 1,
then the RPM trace would likely overpredict the fiducial
motion during exhalation.
The results for the 4 patients having three fiducials are
similar to those having only one fiducial. From Fig. 7, we
can see that, again, the amplitude of lateral motion is small
and shows little dependence on respiration. There is a stron-
ger dependence for AP motion, particularly for Patients
6 – 8. Finally, the greatest amplitude occurs for SI motion,
with a range of 8.0 –17.5 mm. This motion is well correlated

Fig. 6. Superior–inferior (SI) coordinates of the fiducials (diamonds) and renormalized Real-Time Position Management
System (RPM) traces (x) for the first 4 patients in this study. The RPM traces were renormalized using the linear fit
discussed in Methods and Materials.
636 I. J. Radiation Oncology ● Biology ● Physics Volume 67, Number 2, 2007

Fig. 7. Fiducial coordinates in the superior–inferior (SI), anterior–posterior (AP), and lateral (Lat) directions for the
second 4 patients in this study. Fiducial 1 (diamond) is the most superior, and Fiducial 3 (circle) is the most inferior.
Note the scale in the SI direction is twice as large as in the AP and lateral directions. The error bars are not shown. Note
also that the positive fiducial motion in the SI direction is inferior to the diaphragm.

with the respiratory phase for all 4 patients. More important,
for Patients 5 and 6, there is a strong correlation and
synchrony in the motion of all three fiducials. However, for
Patients 7 and 8, where the fiducials are much farther apart,
we found less correlation. Although beyond the scope of
this study, this may be related to deformation of the liver. A
study using deformable image registration, for example,
found that the motion of the diaphragm may not be a good
indicator of liver tumor motion (19). For Patient 8, the AP
motion of fiducial 2 has the same amplitude as the SI motion
and thus is an exception to the results for other patients and
fiducials. Figure 8 shows the results for the renormalized
RPM trace data along with the SI coordinates of the fiducial
for these 4 patients. For all patients, there appears to be a
good correlation between the renormalized RPM trace data
and the location of the fiducials for all phases.
DISCUSSION
Analysis of the data for the first 4 patients shows the
importance of comparing the FB and 4D isocenters before
treatment planning (Fig. 5 and Table 2). Of the 8 patients in
this study, 2 showed shifts in the isocenter of 3 mm or more.
At our institution, the tumor is contoured on the 4D image
set, but actual treatment planning is performed on the FB
scan. The main goal of performing a 4D-CT scan is to
reduce the uncertainty in the location of the tumor resulting
from respiration, thereby allowing a reduction on the mar-
gin of the CTV. One should therefore be careful not to
introduce additional uncertainty caused by patient shifts
between acquisition of the FB and 4D scans. This is of less
concern for patients undergoing gated radiotherapy because
the treatment isocenter is determined solely from the
positions of the fiducials during each treatment session.
Although treatment planning would ideally be performed on
the 4D image set (or an intensity projection of the data), it
is not currently common practice to do so. This is due, in
part, to the large amount of data required to generate an
image set, limiting the scan to an area covering the tumor
region within the liver but not the entire abdominal region
required for treatment planning calculations. Furthermore,
the 4D images are acquired with high-Z contrast agents
(Optiray 320, Mallinkrodt, St. Louis, Missouri) to enhance
 Correlation between internal fiducial tumor motion and external marker motion in 4D-CT for liver tumors
Fig. 8. Average superior–inferior (SI) coordinates of the fiducials (diamonds) and renormalized Real-Time Position
Management System (RPM) traces (x) for the second 4 patients in this study. The RPM traces were renormalized using
the linear fit discussed in Methods and Materials.
the borders of the tumor, whereas the FB scan is acquired
without any contrast agent.
As discussed earlier, the motion of the fiducials is pre-
dominantly in the SI direction. The range of motion is consis-
tent with earlier studies obtained with fluoroscopy (5, 7, 18)
and a most recent 4D-CT study of liver, spleen, and kidney
motion (20). Comparison of the relative locations of the
three fiducials implanted in the liver for Patients 5 and 6
(Fig. 7) shows a good correlation between the motion of the
fiducials throughout the respiratory cycle. Considering the
distances between the fiducials for these 2 patients (3.2 cm
max), our data suggest that one should expect the areas
delimited by closely spaced fiducials within the liver to
move in synchrony. However, for Patients 7 and 8, where
the fiducials are more widely spaced (8.6 cm max), there is
less correlation. This is an important result for gating. For
Patients 5 and 6, a “miss” in the phase of the gate, while still
at the correct RPM amplitude, would likely cover the tumor
volume because it would move as a unit and the relative
orientation of the fiducials will be phase independent.
However, for Patients 7 and 8, irradiating at the incorrect
phase, even if the external marker is at the correct ampli-
tude, would probably lead to some of the tumor lying
outside the treatment field because the fiducials would only
be in the correct orientation at the targeted phase. Further-
more, verification of the correct targeting during treatment
delivery (with acquisition of cine electronic portal images,
for instance) may be more accurate if more than one fiducial
is visible in the treatment field.
Analysis of the renormalized respiration trace and the
locations of the fiducials in the SI direction for all 8 patients
shows strongest correlation between external respiration
and internal motion during expiration, the 40 – 60% phases.
This would suggest that for these patients, the RPM trace is
a reliable predictor for the fiducial position. In particular,
gating treatments during the 40 – 60% phases, using the RPM
trace amplitude as a guide, should yield suitable localization of
● A. S. BEDDAR et al. 637
the fiducial and thus the target volume during treatment. Be-
cause the renormalized RPM amplitude tends to underpredict
the fiducial position during inspiration, end-inspiratory gating
would not be as ideal a choice as end-expiratory gating.
For Patients 2 and 5, there is a significant variation in the
amplitude of the renormalized RPM traces, with a greater
spread during both inspiration and expiration when com-
pared with the end of expiration. This tends to lead to some
choppiness in the reconstructed 4D image sets, which is
typically of only minor concern because the FB scan is used
for treatment planning. This is an issue, however, when the
patient’s breathing becomes too shallow during acquisition
at the tumors inferior or superior borders, because this could
lead to an underestimation in tumor volume. Thus it is
important to examine the respiration trace during image
reconstruction. When more than one image of the appropri-
ate phase is available, as is often the case, the image
coinciding with the more representative respiratory ampli-
tude should be selected.
Identifying the patients who may benefit from respirato-
ry-gated radiotherapy is a more difficult task. Patients 1, 3,
4, and 6 appear to be ideal candidates for respiratory gating
because of their consistent breathing cycles. Treatment de-
livery would be most efficient for such a patient. However,
patients with greater variation in their respiratory cycle
could benefit from respiratory-gated radiotherapy if the gate
is carefully controlled during treatment delivery. For a pa-
tient with inconsistent breathing, the tumor could be better
targeted if the beam is turned on only at the end of expira-
tion and if the amplitude is within a certain range (Fig. 1e).
CONCLUSIONS
Our study of 4D-CT images of implanted fiducials in the
liver shows that the motion of the fiducials is predominant
in the superior–inferior direction. This motion is well cor-
related with the respiratory motion measured by the RPM
638 I. J. Radiation Oncology ● Biology ● Physics
system with an external marker placed on the patient’s
abdomen, with the best agreement occurring at expiration.
Respiratory gating may therefore be most precise if per-
formed at the end of expiration (i.e., the 40 – 60% phases).
For patients with multiple implanted fiducials, the best
REFERENCES
1. Ohara K, Okumura T, Akisada M, et al. Irradiation synchro-
nized with respiration gate. Int J Radiat Oncol Biol Phys
1989;17:853– 857.
2. Shirato H, Shimizu S, Kitamura K, 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– 442.
3. Tada T, Minakuchi K, Fujioka T, et al. Lung cancer: Inter-
mittent irradiation synchronized with respiratory motion—
results of a pilot study. Radiol 1998;207:779 –783.
4. Berbeco RI, Nishioka S, Shirato H, et al. Residual motion of
lung tumours in gated radiotherapy with external respiratory
surrogates. Phys Med Biol 2005;50:3655–3667.
5. Dawson LA, Brock K, Kazanjian S, et al. The reproducibility
of organ position using active breathing control (ABC) during
liver radiotherapy. Int J Radiat Oncol Biol Phys 2001;51:
1410 –1421.
6. Minohara S, Kanai T, Endo M, et al. Respiratory gated irra-
diation system for heavy-ion radiotherapy. Int J Radiat Oncol
Biol Phys 2000;47:1097–1103.
7. Shirato H, Seppenwoolde Y, Kitamura K, et al. Intrafractional
tumor motion: Lung and liver. Semin Radiat Oncol 2004;14:
10 –18.
8. Wagman R, Yorke E, Ford E, et al. Respiratory gating for
liver tumors: Use in dose escalation. Int J Radiat Oncol Biol
Phys 2003;55:659 – 668.
9. Berbeco RI, Neicu T, Rietzel E, et al. A technique for respi-
ratory-gated radiotherapy treatment verifcation with an EPID
in cine mode. Phys Med Biol 2005;50:3669 –3679.
10. Tsunashima Y, Sakae T, Shioyama Y, et al. Correlation be-
tween the respiratory waveform measured using a respiratory
sensor and 3D tumor motion in gated radiotherapy. Int J
Radiat Oncol Biol Phys 2004;60:951–958.
Volume 67, Number 2, 2007
correlation occurs when the fiducials are closely spaced.
When performing respiratory gating, care should be taken to
irradiate not only at the correct amplitude but also at the
correct phase. Finally, one should verify that the isocenters
remain aligned from the FB to the 4D-CT scans.
11. Whyte RI, Crownover R, Murphy MJ, et al. Stereotactic
radiosurgery for lung tumors: Preliminary report of a Phase I
trial. Ann Thorac Surg 2003;75:1097–1101.
12. Pan T, Lee T-Y, Rietzel E, et al. 4D-CT imaging of a volume
influenced by respiratory motion on multi-slice CT. Med Phys
2004;31:333–340.
13. Chi P, Balter P, Luo D, et al. Relation of external surface to
internal tumor studied with cine CT. Med Phys 2006;33:
3116 –3123.
14. Keall PJ, Todor AD, Vedam SS, et al. On the use of EPID-
based implanted marker tracking for 4D radiotherapy. Med
Phys 2004;31:3492–3499.
15. Vedam SS, Kini VR, Keall PJ, et al. Quantifying the predict-
ability of diaphragm motion during respiration with a nonin-
vasive external marker. Med Phys 2003;30:505–513.
16. Mageras GS, Pevsner A, Yorke ED, et al. Measurement of
lung tumor motion using respiration-correlated CT. Int J Ra-
diati Oncol Biol Phys 2004;60:933–941.
17. Pan T. Comparison of helical and cine acquisitions for
4D-CT imaging with multi-slice CT. Med Phys 2005;32:
627– 634.
18. Gierga DP, Brewer J, Sharp GC, et al. The correlation be-
tween internal and external markers for abdominal tumors:
Implications for respiratory gating. Int J Radiat Oncol Biol
Phys 2005;61:1551–1558.
19. Brock K, Dawson LA, Sharpe MB, et al. Feasibility of a novel
deformable image registration technique to facilitate classifi-
cation, targeting, and monitoring of tumor and normal tissue.
Int J Radiat Oncol Biol Phys 2006;64:1245–1254.
20. Brandner ED, Wu A, Chen H, et al. Abdominal organ motion
measured using 4D CT. Int J Radiat Oncol Biol Phys 2006;
65:554 –560.