Download DETERMINATION OF CT-TO-DENSITY CONVERSION

Medical Dosimetry, Vol. 30, No. 3, pp. 145-148, 2005
Copyright © 2005 American Association of Medical Dosimetrists
Printed in the USA. All rights reserved
0958-3947/05/$–see front matter
doi:10.1016/j.meddos.2005.05.001
DETERMINATION OF CT-TO-DENSITY CONVERSION RELATIONSHIP
FOR IMAGE-BASED TREATMENT PLANNING SYSTEMS
CHENG B. SAW, PH.D., ALPHONSE LOPER, M.S., KRISHNA KOMANDURI, PH.D.
TONY COMBINE, M.S., SAIFUL HUQ, PH.D., and CAROL SCICUTELLA, D.O.
Department of Radiation Oncology, University of Pittsburgh Medical Center Cancer Centers, Pittsburgh, PA
(Received 18 May 2004; accepted 23 May 2005)
Abstract—The implementation of tissue inhomogeneity correction in image-based treatment planning will
improve the accuracy of radiation dose calculations for patients undergoing external-beam radiotherapy. Before
the tissue inhomogeneity correction can be applied, the relationship between the computed tomography (CT)
value and density must be established. This tissue characterization relationship allows the conversion of CT value
in each voxel of the CT images into density for use in the dose calculations. This paper describes the proper
procedure of establishing the CT value to density conversion relationship. A tissue characterization phantom
with 17 inserts made of different materials was scanned using a GE Lightspeed Plus CT scanner (120 kVp). These
images were then downloaded into the Eclipse and Pinnacle treatment planning systems. At the treatment
planning workstation, the axial images were retrieved to determine the CT value of the inserts. A region of
interest was drawn on the central portion of the insert and the mean CT value and its standard deviation were
determined. The mean CT value was plotted against the density of the tissue inserts and fitted with bilinear
equations. A new set of CT values vs. densities was generated from the bilinear equations and then entered into
the treatment planning systems. The need to obtain CT values through the treatment planning system is very
clear. The 2 treatment planning systems use different CT value ranges, one from ⴚ1024 to 3071 and the other
from 0 to 4096. If the range is correct, it would result in inappropriate use of the conversion curve. In addition
to the difference in the range of CT values, one treatment planning system uses physical density, while the other
uses relative electron density. © 2005 American Association of Medical Dosimetrists.
Key Words: CT scanning, Tissue characterization, Electron density, Treatment planning system, Tissue inhomogeneity.
to have accurate radiation dose calculations, in particular, to the lung. Studies have shown that uncorrected
treatment plans can produce radiation dose errors exceeding 30% the prescribed dose.8,9 The implementation
of tissue inhomogeneity correction is facilitated by the
availability of image-based treatment planning systems.
Here, the tissue inhomogeneity is derived by converting
the CT value in each voxel into radiological parameters
such as relative electron density or physical density. The
CT value and radiological parameter relationship is typically established empirically by scanning a tissue characterization phantom. This paper reports the proper procedure of relating CT values to radiological parameters
and implementing them in the treatment planning
systems.
INTRODUCTION
Image-based treatment planning has become the standard
of practice in radiotherapy.1–7 Patient data are now acquired within a short time using the new generation fast
helical computed tomography (CT) scanners. The CT
images are then downloaded into a treatment planning
system. At the treatment planning workstation, the CT
image set is assigned as the primary set of patient data for
treatment planning. Next, the tabletop is removed from
the CT images and the external patient contours are
extracted using edge-detection technique. In addition, the
contours of the internal organs can be identified. Additional tools such as image fusion and co-registration are
available to allow visualization of targets seen on other
tomographic modalities superimposed onto the CT image set for delineation. Based on this information, a
virtual patient can be constructed for 3-dimensional
treatment planning. Inverse treatment planning for intensity-modulated radiation therapy can be implemented if
the target volumes are drawn, and dose goal and dose
constraint objectives are defined for these targets and
organs at risk.
Tissue inhomogeneity correction should be applied
METHODS AND MATERIALS
The physical features of the tissue characterization
phantom, CIRS Model 062 (CIRS Tissue Simulation
Technology, Norfolk, VA), are shown in Fig. 1. It is
designed with a height of 270 mm and a width of 330
mm for an abdominal scan. However, it can be separated
into a smaller size head phantom with a radius of 90 mm.
The holes to accommodate the inserts are arranged in 2
concentric rings with radii of 60 mm and 115.3 mm.
Each ring will have 8 holes equally spaced around the
Reprint requests to: Cheng B. Saw, Ph.D., Medical Physics
Division (Room 541), UPMC Cancer Pavilion, 5150 Centre Avenue,
Pittsburgh, PA 15232. E-mail: [email protected]
145
146
Medical Dosimetry
Fig. 1. Physical features of the tissue characterization phantom,
CIRS Model 062, capable of adapting for head and body scans.
ring. There is also a hole at the center of the phantom;
hence, it can accommodate a total of 17 inserts.
The commercially available tissue characterization
phantom comes with 17 inserts and a titanium insert as
an option as listed in Table 1. The relative electron
density ranges from 0.190 to 1.512 or correspondingly
from 0.195 g/ml to 1.609 g/ml for physical density.
Inserts that can accommodate thermoluminescence dosimeters (TLDs) or films are also available upon request.
The arrangement of the insert materials is shown in Fig. 2.
The GE Lightspeed Plus scanner is a third-generation (the tube and detector array are attached to the same
rotating frame and move simultaneously) multislice
scanner with a 70-cm bore size aperture and 496.9-mm
scanning field of view (SFOV). It has a 16-row detector
system and an 8-row data acquisition system (DAS) to
allow for simultaneous acquisition at different longitudinal locations as opposed to single-row detector array of
conventional single-slice CT scanners.10 –12 With this
system, the scanner provides 8 data rows of output,
allowing for simultaneous acquisition of as many as 8
axial images per gantry rotation. The 16-row detector is
segmented into 16 physically separated cells in the lon-
Table 1. Tissue descriptions and densities
Materials
Syringe water
Lung (inhale)
Lung (exhale)
Breast (50/50)
Dense bone 800 mg/ml HA
Trabecular bone
Liver
Muscle
Adipose
Titanium
Location in
Figure 2
1
2
3
4
5
6
7
8
9
Physical
Density
(g/ml)
Relative
Electron
Density
1.000
0.195
0.495
0.991
1.609
1.161
1.071
1.062
0.967
4.507
1.000
0.190
0.489
0.976
1.512
1.117
1.052
1.043
0.952
3.735
Volume 30, Number 3, 2005
Fig. 2. Arrangement of tissue inserts within the tissue characterization phantom.
gitudinal direction to provide post-patient collimation of
the x-ray beam. The actual detector cell size is 2.19 mm
in the longitudinal direction. The effective cell size is
1.25 mm in the longitudinal direction at the isocenter.
These cells in the longitudinal direction can be summed
into a “macro-cell” giving the thickness of 1.25 mm (1
cell), 2.50 mm (2 cells), 3.75 mm (3 cells), and 5.0 mm
(4 cells). The number of macro-cell in the longitudinal
direction forms the macro-row. The maximum scanned
width for a 16-row detector would be 20 mm (4 ⫻ 5.0
mm). The scanner has a shorter focal spot to imaging
isocenter distance of 541 mm compared to 630 mm in a
single-slice scanner. The focal spot to detector distance is
949 mm. The system is equipped with large and small
focal spots with dimensions of 1.2 ⫻ 1.2 mm and 0.9 mm
⫻ 0.7 mm, respectively.
The scanner has the ability to generate images of
various thickness, made possible with the helical scanning mode. However, the image thickness is dependent
on the pitch, which is defined as the ratio of the table
travel per gantry revolution to the beam collimation and
number of detectors in a macro-row. For example, if a
scanning of 4 slices is obtained using 4 detector macrorow ⫻ 2 pitches, then reconstruction of images of thickness from 1⫻ to 4⫻ the detector macro-cell size is
possible.
Before performing the CT scan, the tissue characterization phantom was set up such that the laser passed
through the center of the phantom. Next, the scanning
table was arranged such that the internal CT laser passed
through the center of the phantom; the CT scanner was
then set to zero. A typical scanning parameter of 120
kVp used in standard scanning protocols with 5-mm slice
thickness was used. After the scan, CT images of the
tissue characterization phantom were downloaded into
the network and forwarded to the Varian Eclipse and the
Philip Pinnacle treatment planning systems.
At each treatment planning system, these CT images were converted into treatment planning images. The
CT-to-density conversion relationship ● C. B. SAW et al.
147
CT images were then retrieved and a region of interest
(approximately 300 mm2 for large inserts and 30 mm2
for small inserts) was drawn on the central portion of
each insert to determine the mean CT value and its
standard deviation. The CT values were then plotted
against the relative electron density or physical density
of the tissue. The plot was fitted with 2 linear equations
for CT values from the ⫺1000 to 0 –50 range and the
above 0 –50 range.13
RESULTS
Figure 3 shows the axial CT image of the tissue
characterization phantom. Because the axial CT image
was generated through the filtered backprojection reconstruction technique, streaking artifact may be present,
due to its neighboring high attenuation insert. However,
as seen in the figure, there are clearly no artifacts, indicating that the CT value of each insert is minimally
influenced by its neighboring value. Regions of interest
were drawn on the central portion of the inserts to
determine the mean CT values and their standard
deviations.
Figure 4 shows the CT-to-density conversion or
tissue characterization curve for the Varian Eclipse treatment planning system. The relationship is plotted as the
mean CT values vs. relative electron densities as required
by the dose calculation algorithm. In addition, the range
of CT values were found to be from ⫺1000 to 3071. The
plot shows scattered data points but its behavior can be
predicted to consist of 2 linear relationships for CT value
ranges from ⫺1000 to 0 and above 0. In addition, it
should be recognized that the slope of these straight lines
and the point of inflection change somewhat from CT
scanner to CT scanner.8 The fitted equations has correlation coefficients, R2, of greater than 0.99.
Fig. 3. Axial image of the tissue characterization phantom
showing little to no streaking artifacts.
Fig. 4. Tissue characterization curve for the Varian Eclipse
treatment planning system using relative electron densities.
On the other hand, the tissue characterization curve
for the Philips Pinnacle treatment planning system (Philips Medical, Andover, MA) is totally different compared
to the Varian Eclipse treatment planning system (Varian,
Andover, MA). As shown in Fig. 5, the mean CT values
are plotted against the physical densities of the tissues.
The use of physical densities instead of relative electron
densities should be noted for the Philips Pinnacle treatment planning system. In addition, the range of CT
values was found to be from 0 to 4096. This CT value
range is different from the Varian Eclipse treatment
planning system. The CT value of the image set for the
Philips Pinnacle treatment planning system is derived by
adding a numerical value of 1000 to the CT value of the
image set from the CT scanner. As a standard definition
for CT scanner, the CT value is defined to be ⫺1000 for
air and 0 for water. Again, Fig. 5 shows scattered data
points but can be fitted with bilinear relationships, one
relationship from CT values of 0 to 1000 and the other
from over 1000.
Fig. 5. Tissue characterization curve for the Philips Pinnacle
treatment planning system using physical densities.
148
Medical Dosimetry
DISCUSSION
The human body is composed of various tissues and
cavities that have radiological properties that are different from water. In some instances, foreign materials such
as metallic and non-metallic prostheses may be implanted into patients to improve their quality of life.14 –18
The radiation dose and its distribution are affected by
these inhomogeneities, and because treatments are becoming more conformal, there are opportunities for geometrical misses of the isodose coverage of the target. To
maximize the benefit of radiation therapy, the dose delivered to all irradiated tissues in the presence of these
inhomogeneities should be predicted accurately. AAPM
Report No. 85 has alluded to the need for the implementation of tissue inhomogeneity corrections.14
The implementation of tissue inhomogeneity corrections in image-based treatment planning system requires prior knowledge of the radiological properties of
these tissues, cavities, and prosthetic materials. With the
advent of CT, it is possible to derive the density information in vivo and incorporate it into the dose calculation
process. A number of published articles have shown a
correlation between the radiological properties and CT
values.8,19,20 As such, radiological properties of tissues,
cavities, and prosthetic materials for treatment planning
systems are derived in vivo and exclusively through the
CT values.
It is necessary to determine the CT-to-density conversion curve empirically because the CT values are
dependent on the individual scanner parameters such as
kVp/filtration and reconstruction algorithm. The tissue
characterization phantom is solely designed to establish
the CT-to-density conversion curve. Appropriate scanning parameters must be chosen to minimize the effect of
kVp and slice thickness on the CT-to-density conversion
curve. The kVp chosen should be the typical values used
in scanning protocols for radiotherapy patients. After the
scan, the CT images must be downloaded into the individual treatment planning system before determining the
mean CT value and its standard deviation. This is important because the CT values used may be different for
different treatment planning systems. Furthermore, the
type of radiological parameters required can be different,
such as relative electron density vs. physical density of
tissues, again depending on the dose calculation algorithms. After the mean CT values are determined, the
data must be fitted with the bilinear equations. After the
fit, these equations will be used to generate the CT value
Volume 30, Number 3, 2005
vs. the radiological parameter data for input into the
conversion file in the treatment planning system. This
last step is important otherwise the dose algorithm will
interpolate between measured scattered data points.
REFERENCES
1. Purdy, J.A. Intensity-modulated radiation therapy (Editorial). Int.
J. Radiat. Oncol. Biol. Phys. 35:845– 6; 1996.
2. Saw, C.B.; Ayyangar, K.M.; Enke, C.A. MIMiC-based IMRT-Part
I (Editorial). Med. Dosim. 26:1; 2001.
3. Saw, C.B.; Ayyangar, K.M.; Enke, C.A. MLC-based IMRT-Part II
(Editorial). Med. Dosim. 26:111–12; 2001.
4. Intensity Modulated Radiation Therapy Collaborative Working
Group. Intensity modulated radiotherapy: Current status and issues
of interest. Int. J. Radiat. Oncol. Biol. Phys. 51:880 –914; 2001.
5. Saw, C.B.; Zhen, W.; Ayyangar, K.M.; et al. Clinical Aspects of
IMRT - Part III (Editorial). Med. Dosim. 27:75–7; 2002.
6. Ezzell, G.A.; Galvin, J.M.; Low, D.; et al. Guidance document on
delivery, treatment planning, and clinical implementation of
IMRT: Report of the IMRT subcommittee of the AAPM radiation
therapy committee. Med. Phys. 30:2089 –115; 2003.
7. Orton, C.G.; Chungbin, S.; Klein, E.E.; et al. Study of lung density
corrections in a clinical trail (RTOG 88-08). Int. J. Radiat. Oncol.
Biol. Phys. 41:787–94; 1998.
8. Constantinou, C.; Harrington, J.C. An electron density calibration
phantom for CT-based treatment planning computers. Med. Phys.
19:325–7; 1992.
9. Mackie, T.R.; El-Khatib, E.; Battista, J.; et al. Lung dose corrections for 6 and 15 MV X-rays. Med. Phys. 12:327; 1985.
10. Hu, H. Multi-slice CT: Scan and reconstruction. Med. Phys. 16:
5–18; 1999.
11. McCollough, C.H.; Zink, F.E. Performance evaluation of a multislice CT system. Med. Phys. 16:2223–30; 1999.
12. McCann, C.; Alasti, H. Comparative evaluation of image quality
from three CT simulations scanners. J. Appl. Clin. Med. Phys.
5:55–70; 2004.
13. McCullough, E.C.; Holmes T.W. Acceptance testing computerized
radiation therapy treatment planning system: Direct utilization of
CT scan data. Med. Phys. 12:237– 42; 1985.
14. AAPM Report No. 85. Tissue Inhomogeneity Corrections for
Megavoltage Photon Beams. Madison: Medical Physics Publishing, 2004.
15. Reft, C.; Alecu, R.; Das, I.J.; et al. Dosimetric considerations for
patients with hip prostheses undergoing pelvic irradiation: Report
of the AAPM Radiation Therapy Committee Task Group 63. Med.
Phys. 30:1162– 82; 2003.
16. Hazuka, M.B.; Stroud, D.N.; Adams, J.; et al. Prostatic thermoluminescent dosimeter analysis in a patient treated with 18 MV
X-rays through a prosthetic hip. Int. J. Radiat. Oncol. Biol. Phys.
25:339 – 43; 1993.
17. Niroomand-Rad, A.; Razavi, R.; Thobejane, S.; et al. Radiation
dose perturbation at tissue-titanium dental interfaces in head and
neck cancer patient. Int. J. Radiat. Oncol. Biol. Phys. 34:475– 80;
1996.
18. Klein, E.E.; Kuske, R.R. Changes in photon dose distributions due
to breast prostheses. Int. J. Radiat. Oncol. Biol. Phys. 25:541–9;
1993.
19. Schneider, U.; Pedroni, E.; Lomax, A. The calibration of CT
Hounsfield units for radiotherapy treatment planning. Phys. Med.
Biol. 41:111–24; 1996.
20. El-Khatib, E. Computed tomography for densitometry of lung:
Application in radiotherapy. Med. Dosim. 12:31– 8; 1987.