Computerized

Medical Imaging
and Graphics
PERGAMON Computerized Medical Imaging and Graphics 26 (2002) 153±158
www.elsevier.com/locate/compmedimag

Comparison of ¯at-panel detector and image-intensi®er detector

for cone-beam CT
Rika Baba a,*, Yasutaka Konno a, Ken Ueda b, Shigeyuki Ikeda b
a
Central Research Laboratory, Hitachi Ltd., 1-280 Higashi-koigakubo, Kokubunji-shi, Tokyo 185-8601, Japan
b
Research and Development Center, Hitachi Medical Corporation, 2-1, Shintoyofuta, Kashiwa, Chiba 277-0804, Japan
Received 9 August 2001; revised 11 January 2002; accepted 11 January 2002

Abstract
We evaluated a ¯at-panel detector (FPD) (scintillator screen and a-Si photo-sensor array) for use in a cone-beam computed tomography
(CT) detector and compared it with an image-intensi®er detector (IID). The FPD cone-beam CT system has a higher spatial resolution than
the IID system. At equal pixel sizes, the standard deviation of noise intensity of the FPD system is equal to that of the IID system. However,
the circuit noise of the FPD must be reduced, especially at low doses. Our evaluations show that the FPD system has a strong potential for use
as a cone-beam CT detector because of high-spatial resolution. q 2002 Elsevier Science Ltd. All rights reserved.
Keywords: Cone-beam computed tomography; Flat-panel detector; Image intensi®er; Angiography; Three-dimensional imaging

1. Introduction view. Accordingly, we evaluated a FFD and compared it

with an IID for use in a cone-beam CT detector.
Cone-beam computed tomography (CT) uses a two-
dimensional X-ray detector, and a cone- or pyramid-shaped
X-ray beam. In addition to ¯uoroscopy and radiography 2. Methods
functions, it can acquire a full set of projection images
2.1. Detectors
during a single rotation scan, and reconstruct an isotropic
high-spatial resolution three-dimensional image.
The FPD (provided by Varian Medical Systems, Inc.)
A cone-beam CT system using several X-ray tubes and
consists of a scintillator screen and a-Si photo-sensor
detectors (DSR) was developed as an ultrafast CT, espe-
array (Fig. 1). An X-ray beam is converted to an optical
cially for heart, lungs, and circulation [1]. On the other
signal by the scintillator, and detected by a-Si sensor
hand, the system with an X-ray tube and a detector has
array. The FPD does not generate veiling glare or image
been developed especially for angiography [2±5]. We
distortion and has a smaller imaging pixel than that of the
have developed a new cone-beam CT system using an X-
IID.
ray image intensi®er and charge coupled device (CCD) tele-
The IID consists of a 16-inch X-ray image intensi®er,
vision camera as the detector. We have evaluated the system
optics, and a CCD camera. An X-ray beam is converted to
for two years; it has been shown to be very effective for
an optical signal by the input phosphor screen, and then
planning operations and con®rming their results, for ortho-
converted to electrons by the photocathode screen. Electrons
pedic imaging [6±10]. However, it is necessary to enlarge
are accelerated through the electric ®eld inside the image
its ®eld of view and improve its spatial resolution by future
intensi®er and converted to an optical signal at the output
development.
phosphor screen. The optical signal is adjusted by the optical
A ¯at-panel detector (FPD) can form a higher spatial
iris and detected by the CCD. The read-out image contains
resolution image than an image-intensi®er detector (IID)
geometrical distortion and a blurring component of veiling
[11±15]. In addition, it can greatly enlarge the ®eld of
glare generated at the image intensi®er [16±18].

2.2. Experimental systems

* Corresponding author. Tel.: 181-42-323-1111; fax: 181-42-327-7681.
E-mail address: r-baba@crl.hitachi.co.jp (R. Baba). We have developed a stationary type cone-beam CT
0895-6111/02/$ - see front matter q 2002 Elsevier Science Ltd. All rights reserved.
PII: S 0895-611 1(02)00008-3
154 R. Baba et al. / Computerized Medical Imaging and Graphics 26 (2002) 153±158
Fig. 1. IID and FPD.
experimental system, in which a detector and an X-ray tube
are stationary and a subject is rotated on a turntable. On the
other hand, we developed a cone-beam CT prototype
system, in which an IID and an X-ray tube were placed on
arms sticking out from a CT gantry disk; they were rotated
during scanning and the subject was stationary, lying on a
bed (Fig. 2). The images obtained by the stationary type and
gantry type are equivalent. Our FPD system evaluated in
this paper was the stationary type, whereas our reference
system using the IID was the gantry type (Fig. 3).
The FPD acquires a 12-bit 960 £ 768 pixel projection
image at 15 frames/s (Table 1). The pixel size of the FPD
is 0.254 mm. The IID acquires a 12-bit 512 £ 512 pixel
projection image at 60 frames/s. After geometrical correc-
tion, the pixel size of the image from the IID is set to 0.8 or
0.307 mm. The pixel size of 0.8 mm is the maximum which
®ts the maximum ®eld of view (16-inch mode), and
0.307 mm is the minimum which ®ts the minimum ®eld
of view (6-inch mode).
A 40-lines/cm anti-scatter grid can be set in front of the
detector to reduce scattered X-rays. In these experiments,
the grid was used for only the large pixel size (0.8 mm)
because the S/N was decreased by the grid when the pixel
size was small.
Fig. 2. Cone-beam CT.
Fig. 3. Experimental set ups.
The FPD system forms two-dimensional images with a
pixel size of 0.254 mm. These images are reconstructed into
a three-dimensional image with a 0.169-mm voxel and 162-
mm diameter £ 130-mm height cylindrical ®eld of view. In
order to avoid resolution depression in image reconstruction
stage, the voxel size, the pitch of the reconstructed three-
dimensional image, was adjusted to the effective pixel size
at the rotation center. The value was calculated by the pixel
size of detector divided by the magni®cation factor of the
image at the rotation center. The IID system forms two-
dimensional images with a pixel size of 0.8 mm. Then,
these images are reconstructed into a three-dimensional
image with a 0.533-mm voxel and a 273-mm diameter
sphere ®eld of view. Those with a pixel size of 0.307 mm
are reconstructed into a three-dimensional image with a
0.205-mm voxel and a 105-mm diameter sphere ®eld of
view. The reconstruction process is performed by using
the Feldkamp algorithm with the standard Shepp & Logan
®lter [19].
3. Experiments and results
3.1. Spatial resolution
A 165-mm diameter cylindrical water phantom contain-
ing a high-contrast bar pattern was used to evaluate the
contrast resolution on the plane perpendicular to the
rotation-axis (X±Y plane) and parallel to it (Y±Z plane).
During a scan, 576 projection images were obtained and
several reconstructed images were added in order to evalu-
ate spatial resolution without in¯uence of the streak artifact
and noise.
A 0.35-mm bar pattern can be resolved on both the X±Y
and Y±Z planes of the reconstructed image taken by the FPD
system when the voxel size is 0.169 mm (Fig. 4(a) and (b)).
A 0.65-mm bar pattern can be resolved on both the X±Y and
Y±Z planes of the reconstructed image taken by the IID
system when voxel size is 0.533 mm (Fig. 5(a) and (b)). A
0.35-mm bar pattern can be resolved on the X±Y plane and a
 R. Baba et al. / Computerized Medical Imaging and Graphics 26 (2002) 153±158 155

The measured standard deviations of the CT value of

image taken by the 0.169-mm voxel FPD system are plotted
as B in Fig. 6. Based on these data, equation of the variance,
which was square of the standard deviation, was derived. It
consists of a ®rst term inversely proportional to the incident
X-ray dose to the detector, and a second term independent of
it. The standard deviation calculated using the equation is
plotted as a solid line. The derived equation is considered
reliable because the solid line ®ts the measured data (B)
well. The standard deviation of the CT value of image
taken by the 0.533-mm voxel FPD system was calculated
using the equation and is plotted as a broken line.
The measured standard deviations corresponding to the
0.205-mm voxel IID system are plotted as W, and those
corresponding to the 0.533-mm voxel system are plotted
Fig. 4. Reconstructed images of high-contrast bar-pattern phantoms by FPD as K. The optical iris of the IID system was set so that X-
system (120 kV, 160 mA, 5 ms, 576 projections, 1.7-mm thick). (a) X±Y ray intensity passing through the air was not saturated at the
plane (0.169-mm pitch), (b) Y±Z plane (0.169-mm pitch).
standard dose in a clinical trial of the system (230 mAs). As
a result, the iris was not controlled. Thus, performance of
0.40-mm one on the Y±Z plane when voxel size is 0.205 mm IID system could be compared to that of FPD system with-
(Fig. 5(c) and (d)). The spatial resolution of the FPD system out iris control.
is thus higher than that of the IID system. The ®gure shows that the standard deviations correspond-
ing to the FPD system are nearly equal to those correspond-
3.2. Contrast resolution ing to the IID system with dose more than 144 mAs, but
they are signi®cantly different at 46.1 mAs. The contrast
The contrast resolution in relation to the total clinical-use
resolution of the FPD system is equal at high and medium
dose of the cone-beam CT (120 kV, 14.4±288 mAs) was
doses but higher at low dose.
investigated. A 200-mm diameter cylindrical water phan-
tom was used as the subject. During a scan, 288 projection 3.3. Noise reduction method
images were obtained, and the standard deviation of the
pixel intensity in the central water region on the mid- The noise of the reconstructed image, which consists of
plane image perpendicular to the rotation-axis was the ®rst term of X-ray quantum noise and the second term of
measured. circuit noise, was derived as in Eq. (1).
v
uXN   X N  2
1u
t 1 s… 1 † 1
p 1
V{Y…x; y; z†} N iˆ1 noi exp{ 2 mLi …x; y; z†} iˆ1
c oi noi exp{ 2 mLi …x; y; z†}
ˆ " # …1†
Y…x; y; z† 1 X N X
1
0 0
m {L …k †w…k 2 k†}
N iˆ1 k 0 ˆ2 1 i

Table 1
Parameters of experimental systems

IID system FPD system

Distance between source and 1200 1245

detector (mm)
Distance between source and 800 830
rotation center (mm)
Measured image matrix (pixel) 512 £ 512 (12 bit) 960 £ 768 (12 bit)
Pixel size of detector (mm) 0.800 or 0.307 0.254
Frame rate (frame/s) 60 15
Number of projections (images) 288 or 576 288 or 576
Reconstructed image matrix (voxel) 512 £ 512 £ 512 960 £ 960 £ 768
Reconstructed ®eld of view 273 or 105-mm diameter sphere 162-mm-diameter
£ 130-mm cylinder
Reconstructed pitch (mm) 0.533 or 0.205 0.169
156 R. Baba et al. / Computerized Medical Imaging and Graphics 26 (2002) 153±158
Fig. 5. Reconstructed images of high-contrast bar-pattern phantoms by IID
system (120 kV, 160 mA, 5 ms, 576 projections, 3-mm thick). (a) X±Y
plane (0.533-mm pitch), (b) Y±Z plane (0.533-mm pitch), (c) X±Y plane
(0.205-mm pitch), (d) Y±Z plane (0.205-mm pitch).
Here, Y…x; y; z† is the reconstructed data at a point of interest,
and V{Y} is the variance of Y. N is the number of projection
image, noi the X-ray dose applied from an X-ray source at
ith angle, Li the X-ray beam path through the subject, m the
X-ray absorption coef®cient, s (1 ) the offset noise of a
detector, coi the optical gain, and w(k) is convolution ®lter.
If both X-ray dose (noi) and optical gain (coi) are controlled
at each projection angle, each term can be minimized
individually.
As optical gain (coi) increases, the second term becomes
smaller for the same X-ray dose. Our developed IID system
has a high-speed variable iris, and optical gain was
controlled at each projection angle so that there is no satura-
tion in the area of the subject. The standard deviations of
noise intensity corresponding to the 0.533-mm voxel IID
system were measured without and with optical-gain
control. Those without and with the control were 48.9 and
23.2 at 230 mAs, respectively. It is thus clear that this
control method can effectively reduce noise.
The FPD system cannot control the optical gain (coi). We
Fig. 6. Standard deviation in transaxial images of 200-mm diameter cylind-
rical-water phantom (120 kV).
Fig. 7. Projection images of hand phantom. (a) IID measured image, (b) IID
corrected image, (c) FPD measured image, (d) FPD corrected image.
will reduce the image noise by controlling the electrical
circuit gain instead of the optical gain. The development
is required to increase the electrical circuit gain and
decrease the electrical circuit noise of the detector in
order to reduce the image noise equal to that of the IID
system using the iris control. And furthermore, it is neces-
sary to increase the bit number of the analog-to-digital
converter. Generally, when the X-ray beam path through
the subject is shorter or longer at each projection angle,
it is possible to reduce more image noise by controlling
the X-ray dose (noi) with circuit gain because the ®rst term
can be minimized.
3.4. Phantom imaging
A right-hand phantom was imaged by the FPD system
(0.254-mm pixel) and the IID system (0.307-mm pixel) at
a low total X-ray dose of 46.1 mAs (Fig. 7). Pixel defect,
offset, and sensitivity of images taken by the FPD system
were corrected, and offset, veiling glare, sensitivity, and
distortion of the those taken by the IID system were also
corrected.
The same reconstruction process for both systems was
performed by using the Feldkamp algorithm with the stan-
dard Shepp & Logan ®lter. The reconstructed images with
0.169-mm voxel were obtained (Fig. 8). The slice image of
the FPD system has a higher spatial resolution but more
noise than that of the IID system. In spite of noise, the
image of the FPD system shows the small gaps between
the bones as clearly as that taken by the IID system. The
detailed structures are described more clearly in the
images by the FPD system than the IID system ( ! ).
Volume-rendering images taken by the FPD system are
obviously more accurate than those taken by the IID
system are ( ! ).
 R. Baba et al. / Computerized Medical Imaging and Graphics 26 (2002) 153±158
Fig. 8. Reconstructed images of hand phantom (120 kV, 50 mA, 3 ms, 288
projections 0.169-mm voxel). (a) IID coronal image, (b) IID volume-
rendering image, (c) FPD coronal image, (d) FPD volume-rendering image.
140-mm diameter acrylic-cylinder phantom composed of
a pig's coronary artery ®lled with contrast materials was
imaged by the FPD system. The corrected projection
image and maximum-intensity projection image clearly
show the vessels (Fig. 9). About 0.5-mm diameter vessels
could be distinguished ( ! ).
4. Conclusions
The FPD cone-beam CT system has a higher spatial
Fig. 9. FPD maximum-intensity-projection image of pig's coronary artery
®lled with contrast materials (120 kV, 100 mA, 5 ms, 288 projections,
0.169-mm voxel).
157
resolution than the IID system. At equal pixel sizes, the
standard deviation of noise intensity of the FPD system is
equal to that of the IID system. However, the circuit noise of
the FPD must be reduced so that the standard deviation of its
noise will become equal to that of the IID using the iris
control, especially at low doses. Moreover, the X-ray-dose
control method is effective to improve the S/N of three-
dimensional images.
Our evaluations show that the FPD system has a strong
potential for use as a cone-beam CT detector because of
high-spatial resolution. As the next stage of our research,
a FPD cone-beam CT system with a large ®eld of view and
high acquisition rate will be developed.
References
[1] Ritman EL, Kinsey H, Robb RA, Gilbert BK, Harris LD, Wood EH.
Three-dimensional imaging of heart, lungs, and circulation. Science
1980;210(17):273±80.
[2] Ning R, Kruger RA. Computer simulation of image intensi®er-based
computed tomography detector: vascular application. Med Phys
1988;15(2):188±92.
[3] Saint-Felix DS, Picard C, Ponchut C, Romeas R, Rougee A, Trousset
Y. Three dimensional X-ray angiography: ®rst in vivo results with a
new system. Proc SPIE 1993;1897:90±8.
[4] Sekihara K, Kawai H, Yamamoto K, Kumazaki T. Cone-beam CT
angiography. Proc JAMIT Front 1995:23±8.
[5] Ning R, Wang X, Conover DL, Tang X. An image intensi®er-based
volume tomographic angiography imaging system. Proc SPIE 1997;
3032:238±46.
[6] Baba R, Ueda K, Ueki H, Umetani K. Ellipsoid scan: chest cone-beam
CT with a large ellipsoidal view ®eld using a 16-inch X-ray image
intensi®er. Proc SPIE 1997;3032:349±57.
[7] Baba R, Ueda K, Kawai H, Kuba A, Takagi H. Development of three-
dimensional imaging method with cone-beam X-ray. MEDIX
1999;31:42±7.
[8] Baba R, Ueda K, Kawai H, Oka Y, Yamasaki K. High-S/N cone-beam
CT with exposure- and camera-input-level control. Radiology
1998;209(P): 284 see also p. 581.
[9] Baba R, Ueda K, Kuba A, Kohda E, Shiraga N, Sanmiya T.
Development of a subject-standing-type cone-beam CT for chest
and orthopedic imaging. Radiology 1999;213(P):194 see also p.
501.
[10] Baba R, Ueda K, Kuba A, Kohda E, Shiraga N, Sanmiya T. Devel-
opment of a subject-standing-type cone-beam computed tomography
for chest and orthopedic imaging, in preparation.
[11] Colbeth RE, Allen MJ, Day DJ, Gilblom DL, Harris R, Job ID,
Klausmeier-Brown ME, Pavkovich J, Seppi EJ, Shapiro EG, Wright
MD, Yu JM. Flat panel imaging system for ¯uoroscopy applications.
Proc SPIE 1998;3336:376±87.
[12] Yamada S, Umazaki H, Takahashi A, Honda M, Shiraishi K, Rudin S,
Bednarek DR, Yang CJ, Wang Z, Gopal A. Image quality evaluation
of a selenium-based ¯at-panel digital X-ray detector system based on
a animal studies. Proc SPIE 2000;3977:429±36.
[13] Granfors PR, Aufrichtig R. Performance of a 41 £ 41-cm 2 amorphous
silicon ¯at panel X-ray detector for radiographic imaging applica-
tions. Med Phys 2000;27(6):1324±31.
[14] Ning R, Chen B, Yu R, Conover D, Tang X, Ning Y. Flat panel
detector-based cone-beam CT volume CT angiography imaging:
system evaluation. IEEE Trans Med Imag 2000;19(9):949±63.
[15] Jaffray DA, Siewerdsen JH. Cone-beam computed tomography with a
158 R. Baba et al. / Computerized Medical Imaging and Graphics 26 (2002) 153±158
¯at-panel imager: initial performance characterization. Med Phys
2000;27(6):1311±23.
[16] Rudin S, Bednarek DR, Wong R. Accurate characterization of image
intensi®er distortion. Med Phys 1991;18(6):1145±51.
[17] Fahrig R, Moreau M, Holdsworth W. Three-dimensional computed
tomographic reconstruction using a C-arm mounted XRII: correc-
tion of image intensi®er distortion. Med Phys 1997;24(7):1097±
106.
Rika Baba received her Master's degree of Agriculture from Kyoto
University in 1991. Since 1991, she has been working for Central
Research Laboratory, Hitachi Ltd. Her current research interests
include three-dimensional medical imaging and processing.
Yasutaka Konno received his Master's degree of Science from Tokyo
University in 1998. Since 1998, he has been working for Central
Research Laboratory, Hitachi Ltd. His current research interests include
designing X-ray detector for medical imaging.
[18] Seibert JA, Nalcioglu O, Roeck W. Removal of image intensi®er
veiling glare by mathematical deconvolution techniques. Med Phys
1985;12(3):281±8.
[19] Feldkamp LA, Davis LC, Kress JW. Practical cone-beam algorithm.
J Opt Soc Am A 1984;1(6):612±9.
Ken Ueda received his Master's degree of Engineering from Tokyo
University in 1972. He worked for Central Research Laboratory, Hita-
chi Ltd from 1972 to 1999. Since 1999, he has been working for
Research and Development Center, Hitachi Medical Corporation. His
current research interests include three-dimensional medical imaging
and processing.
Shigeyuki Ikeda received his Bachelor's degree of Science from The
University of Electro-Communications in 1982. Since 1982, he has
been working for Research and Development Center, Hitachi Medical
Corporation. He has developed digital radiography system including
¯at panel detector system.