Tumor Tracking in the Absence of Radiopaque Markers
Ross I. Berbeco1, Hassan Mostafavi2, Gregory C. Sharp1, Steve B. Jiang1
1
Department of Radiation Oncology, Massachusetts General Hospital and Harvard Medical School, Boston, MA, USA
Ginzton Technology Center, Varian Medical Systems, Inc., Palo Alto, CA USA
2
Abstract
Due to a risk of pneumothorax, physicians are reluctant to implant radiopaque markers within patients’ lungs for the purpose of
radiographic or fluoroscopic tumor localization. It is, therefore, necessary to develop methods to find the positional information
directly, without surrogate markers. We have explored several algorithms for separating tumor motion from the stationary
anatomy. The motion, itself, is exploited to reveal the tumor’s positional information. We show that the tumor motion may be
tracked in the absence of radiopaque markers, for specific cases. More robust methods, coupled with better data, may be needed
for a universal algorithm.
Keywords
Radiotherapy, tumor tracking, image guidance, IRIS
Introduction
A large concern in radiation oncology is the localization of
the tumor during radiation treatment [1-6]. This becomes
especially important when treating tumors located in a
patient’s abdomen or thorax, because of respiratory motion.
Several image guidance techniques have been proposed to
deal with this issue [7-14]. At our institution, we are building
an Integrated Radiotherapy Imaging System (IRIS)
consisting of orthogonal kV x-ray sources and flat panel
detectors, all mounted on a Varian Clinac 21EX gantry
(VMS) [7]. Our midterm goal is to use this system to
perform gated treatment and our long-term goal is real-time
adaptive radiotherapy.
difficult to delineate the edges in some anatomical locations.
To overcome this, a common practice is the surgical
placement of one or more radiopaque markers in or near the
tumor prior to treatment. These markers, which are visible in
kV x-ray images, act as a reliable surrogate for tumor
location [15-17]. However, there may be cases for which
physicians are reluctant to implant markers. For example,
there is a concern that the current technique for marker
implantation in the lung has an unacceptable risk of causing
pneumothorax in the patient [18-22]. For these cases, it
would be in the patient’s best interest to image the tumor
directly, without markers.
Tumor density is very close to tissue density, making it
The interior of a patient’s lungs, being mostly air, acts a
Inside
Outside
Figure 1a: A lateral view of a patient with a liver tumor.
Radiopaque markers have been surgically implanted near the
tumor. The white box represents a region of interest within the
patient. The black box represents a region of interest outside the
patient.
Figure 1b: An anterior-posterior view of a patient with a lung tumor.
The tumor is visible within the white-dotted region of interest.
good contrast to tumors. In many cases, lung tumors can be
located without the aid of radiopaque markers. To illustrate
this point, figure 1a shows a lateral kV x-ray image of a
patient with a liver tumor. The tumor shape (within the white
dotted box) is indiscernible from the surrounding healthy
tissue. Only the markers (in the upper left corner of the box)
show where the tumor is. Figure 1b shows an anteriorposterior image of a different patient with a visible lung
tumor. No markers are needed to visually locate the tumor
mass in this image.
We have investigated several different approaches for
detecting motion in fluoroscopic images and locating lung
tumors without the prior implantation of radiopaque
markers. Our work shows the feasibility of building an
algorithm for automatically tracking lung tumors with kV xrays. This work will be further improved by the use of IRIS
as well as the integration of other intelligence from an RPM
system and/or a prior CT scan.
Materials and Methods
Motion induced intensity fluctuations
The patient in figure 1a underwent a fluoroscopic simulation
session in our clinic prior to radiation treatment. A C series
Ximatron was used to acquire images at 30 frames/sec. The
patient has a liver tumor that cannot be seen in the image
shown. However, as the patient breathes, the relative image
intensity in the region of the tumor may change due to organ
motion and deformation. The white box shows the region of
interest inside the patient within which the tumor is moving.
The image shown was chosen at inhale to show that the
diaphragm does not enter the region of interest at any point
during the breathing cycle. In our clinic, we use the
automatic brightness control (ABC) feature to partially
account for global intensity changes. To show that the
intensity change inside the region of interest is not due to the
ABC, we also tracked the intensity change in a region
outside the patient. Figure 2 shows the relative average
intensity values for the pixels within the regions of interest
inside and outside of the patient.
The oscillatory pattern inside the patient matches the
patient’s breathing pattern. This shows that there is
information about the internal motion that can be extracted
from fluoroscopic images even if the tissue densities are very
similar. This could be used as a basis for creating a signal
that serves as a surrogate for sensing the tumor motion.
Although this algorithm could not be used for detecting
tumor location directly, it demonstrates the principle that
motion information can be obtained from fluoroscopic data
in which a tumor is not visible without the aid of radiopaque
markers. More advanced methods (like the ones presented in
this paper) are needed to determine tumor locality.
Motion enhancement and detection for tumor gating,
tracking and position monitoring
The feasibility of using motion enhancement techniques to
highlight moving organs and find the associated tumor
position has been studied. Moving and stationary features in
fluoroscopic images can be contrasted using several
breathing cycles. Real-time fluoroscopic images can than be
produced in which features are enhanced or suppressed
depending on their motion characteristics.
Respiratory synchronized fluoroscopy is acquired using
Varian’s RPM Respiratory Gating System. This allows
phase-based registration of the fluoroscopic sequence with a
4D treatment plan[23]. Frame averaging is used to form a
sequence in which each frame is the long-term average of the
frames up to that point. By subtracting this averaged image
from the current input frame, the moving structures are
enhanced in the current image while the contrast of the
stationary features is suppressed. Using this method, a
reference set of motion-enhanced image templates is
prepared for many phases of the breathing cycle. The target
contour for each phase of the breathing cycle, as defined by
the 4D treatment plan, is projected onto the corresponding
Relative Intensity
Intensity fluctuation as a function of time
1.2
Inside the patient
1.1
1.0
0.9
Outside the patient
0.8
0
100
200
300
400
500
f rame (30 f rames = 1sec)
Figure 2. The change of average image intensity of a region
of interest inside and outside the patient, as a function of time.
(a)
(c)
(b)
Figure 3. The original (a) and motion-enhanced
(b) fluoroscopic frames of a lung patient, and the
template (c) with the best match to the area
shown by the overlay box in (b).
Lat Position- Pixels
Results and Discussion
Max Correlation
450
1
400
0.9
350
0.7
300
Pixels
0.6
250
0.5
200
0.4
150
0.3
100
Max Correlation Value
0.8
0.2
50
0.1
0
0
1
8
15
22
29
36
43
50
57
64
71
78
85
92
99
Frame No.
Figure 4. An example of tracking results using a sequence of
templates. In addition to the pixel coordinates, the maximum
correlation value for template matching is also shown.
frame of the simulation fluoroscopy. This is then used to
select the target region of interest (ROI) that comprises the
reference template for that frame. During the treatment
delivery, the real-time fluoroscopic images are enhanced,
separating the moving tissues from stationary structures, and
then temporally and spatially registered with similarly
processed candidate reference sequences. The reference
template resulting in the best spatial match, based on cross
correlation template matching, defines the position of the
target ROI. Figure 3 shows the input and motion-enhanced
fluoroscopic frames of a lung patient. Figure 3c shows the
template with the best match to the area shown by the
overlay box in Figure 3b. Figure 4 shows the tracking results
using images taken at 10 frames/sec. This motion
enhancement and template-matching algorithm can be used
as a position monitoring system for both gating and tracking.
Template matching for feature-based tumor tracking
In the images of the patient shown in figure 1b, the tumor
mass can be seen even in the static image. As the patient
breathes, the tumor can be seen to move in a mostly cranialcaudal direction. Given the relatively distinct shape of this
tumor and the adjacent anatomy, template matching may be
performed on the fluoroscopic images. A sample image is
used to define the region of interest within which the tumor
will be moving and to select the template. The region of
interest in this and all subsequent images is normalized to
zero mean in order to have consistency among the data set.
The position of the template is described by its center-ofmass. For each frame, a simple convolution algorithm, using
the template as the convolution kernel, is applied to find the
best template match. The coordinates of the best match are
output and recorded. The recorded trace of the motion is
shown in figure 5. This breathing pattern fits the visually
available pattern in the fluoroscopic images.
We have shown that although tumors may not be
immediately visible in static lung radiographs, valuable
information may be gleaned from a fluoroscopic sequence of
image frames. The motion, for which we would like to
compensate, is the key to marker-less tumor tracking. Our
results show that it is feasible to isolate a moving tumor from
the rest of the anatomy by using a motion detection
algorithm. The motion enhancement method presented uses
the past history of the data set to suppress static anatomy
while enhancing objects that are in motion. Phase-based
template matching is then performed to find the location of
the desired object. The results, shown in figure 4, show a
periodic motion of the tumor consistent with respiration.
Unfortunately, the maximum correlation value tends to fall
off just after periods of lowest target motion. Methods of
prediction [24, 25] may help to improve this.
The feature-based template-matching algorithm presented
here is successful in tracking a lung tumor with distinctive
features. The case presented here shows the feasibility of
direct tumor tracking. It should be pointed out that to have
such an obvious mass is not always the case. For many,
perhaps even for most, patients, this technique will be unable
to effectively track the tumor. Further advancements and the
inclusion of information from other technologies may bolster
this technique and allow it to have more general relevance.
Conclusions
Valuable information about lung tumor motion can be
extracted from fluoroscopic images in the absence of
implanted radiopaque markers. Although the methods
presented here are not yet robust enough for clinical
implementation, we believe that subsequent improvements
may allow us to reliably track lung tumors without the aid of
implanted radiopaque markers. To perform this task in three
dimensions, which would be required for use during
treatment, a stereoscopic imaging system like the IRIS must
be used [7]. By combining all of the information available
from 4DCT[23], RPM, fluoroscopic simulation, and
orthogonal on-board imaging, we may be able to discern the
S-I tumor motion as a function of time
12.5
Distance (cm)
SI position - Pixels
12.0
11.5
11.0
10.5
10.0
0.0
2.0
4.0
6.0
8.0
10.0
Time (seconds)
Figure 5. The superior-inferior tumor motion, relative to an
arbitrary fixed point. The position at each time was determined by
the template-matching method.
tumor location throughout the imaging session. Once
tracking has been established, the treatment can be adapted
to its location by either gating the treatment beam or taking
the motion into account in the MLC leaf sequencing [13].
This presents a clear path towards marker-less 4D
radiotherapy for lung tumors.
14.
15.
References
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
Langen, K.M. and D.T. Jones, Organ motion and its
management. Int J Radiat Oncol Biol Phys, 2001.
50(1): p. 265-78.
Weiss, P.H., J.M. Baker, and E.J. Potchen,
Assessment of hepatic respiratory excursion. J Nucl
Med, 1972. 13(10): p. 758-9.
Suramo, I., M. Paivansalo, and V. Myllyla, Craniocaudal movements of the liver, pancreas and kidneys
in respiration. Acta Radiol Diagn (Stockh), 1984.
25(2): p. 129-31.
Davies, S.C., et al., Ultrasound quantitation of
respiratory organ motion in the upper abdomen. Br
J Radiol, 1994. 67(803): p. 1096-102.
Balter, J.M., et al., Uncertainties in CT-based
radiation therapy treatment planning associated
with patient breathing. Int J Radiat Oncol Biol Phys,
1996. 36(1): p. 167-74.
Shimizu, S., et al., Three-dimensional movement of a
liver tumor detected by high-speed magnetic
resonance imaging. Radiother Oncol, 1999. 50(3): p.
367-70.
Berbeco, R., et al., Integrated radiotherapy imaging
system (IRIS): design considerations of tumour
tracking with linac gantry-mounted diagnostic x-ray
systems and flat-panel detectors. Phys Med Biol,
2004. 49(2): p. 243-255.
Mackie, T.R., et al., Image Guidance for precise
conformal radiotherapy. Int J Radiat Oncol Biol
Phys, 2003. 56(1): p. 89-105.
Murphy, M.J., et al., The effectiveness of breathholding to stabilize lung and pancreas tumors
during radiosurgery. Int J Radiat Oncol Biol Phys,
2002. 53(2): p. 475-82.
Jaffray DA, et al., Flat-panel cone-beam computed
tomography for image-guided radiation therapy.
International Journal of Radiation Oncology,
Biology, and Physics, 2002. 53(5): p. 1337-49.
Shirato, H., et al., Physical aspects of a real-time
tumor-tracking system for gated radiotherapy. Int J
Radiat Oncol Biol Phys, 2000. 48(4): p. 1187-95.
Kubo, H.D. and B.C. Hill, Respiration gated
radiotherapy treatment: a technical study. Phys Med
Biol, 1996. 41(1): p. 83-91.
Neicu, T., et al., Synchronized moving aperture
radiation therapy (SMART): average tumour
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
trajectory for lung patients. Phys Med Biol, 2003.
48(5): p. 587-98.
Ford, E.C., et al., Evaluation of respiratory
movement during gated radiotherapy using film and
electronic portal imaging. International Journal of
Radiation Oncology*Biology*Physics, 2002. 52(2):
p. 522-531.
Gierga, D., et al., Correlation between external and
internal markers for abdominal tumors: implications
for respiratory gating. Int J Radiat Oncol Biol Phys,
2003. 57(2 Suppl): p. S186-7.
Kitamura, K., et al., Registration accuracy and
possible migration of internal fiducial gold marker
implanted in prostate and liver treated with realtime tumor-tracking radiation therapy (RTRT).
Radiother Oncol, 2002. 62(3): p. 275-81.
Shirato, H., et al., Feasibility of
insertion/implantation of 2.0-mm-diameter gold
internal fiducial markers for precise setup and realtime tumor tracking in radiotherapy. Int J Radiat
Oncol Biol Phys, 2003. 56(1): p. 240-7.
Topal, U. and B. Ediz, Transthoracic needle biopsy:
factors effecting risk of pneumothorax. Eur J Radiol,
2003. 48(3): p. 263-7.
Geraghty, P.R., et al., CT-guided transthoracic
needle aspiration biopsy of pulmonary nodules:
needle size and pneumothorax rate. Radiology,
2003. 229(2): p. 475-81.
Arslan, S., et al., CT- guided transthoracic fine
needle aspiration of pulmonary lesions: accuracy
and complications in 294 patients. Med Sci Monit,
2002. 8(7): p. CR493-7.
Laurent, F., et al., Percutaneous CT-guided biopsy
of the lung: comparison between aspiration and
automated cutting needles using a coaxial technique.
Cardiovasc Intervent Radiol, 2000. 23(4): p. 266-72.
Laurent, F., et al., CT-guided transthoracic needle
biopsy of pulmonary nodules smaller than 20 mm:
results with an automated 20-gauge coaxial cutting
needle. Clin Radiol, 2000. 55(4): p. 281-7.
Rietzel, E., et al., 4D computed tomography for
treatment planning. Int J Radiat Oncol Biol Phys,
2003. 57(2 Suppl): p. S232-3.
Sharp, G., et al., Prediction of respiratory tumour
motion for real-time image guided radiotherapy.
Phys Med Biol.
Murphy, M.J., J. Jalden, and M. Isaksson, Adaptive
filtering to predict lung tumor breathing motion
during image-guided radiation therapy. Proc. 16th
Int. Conf. on Computer Assisted Radiology (CARS
2002) (June 2002) ed H U Lemke, K Inamura, M W
Vannier, A G Farman and K Doi, 2002.