The frequency split method for helical cone-beam reconstruction

G. Shechter
Philips Medical Systems and Technologies, Haifa, Israel
Th. Köhler
Philips Research Laboratories, Hamburg, Germany
A. Altman
Philips Medical Systems and Technologies, Haifa, Israel
R. Proksa
Philips Research Laboratories, Hamburg, Germany
共Received 6 January 2004; revised 14 April 2004; accepted for publication 27 May 2004;
published 14 July 2004兲
A new approximate method for the utilization of redundant data in helical cone-beam CT is pre-
sented. It is based on the observation that the original WEDGE method provides excellent image
quality if only little more than 180° data are used for back-projection, and that significant low-
frequency artifacts appear if a larger amount of redundant data are used. This degradation is
compensated by the frequency split method: The low-frequency part of the image is reconstructed
using little more than 180° of data, while the high frequency part is reconstructed using all data.
The resulting algorithm shows no cone-beam artifacts in a simulation of a 64-row scanner. It is
further shown that the frequency split method hardly degrades the signal-to-noise ratio of the
reconstructed images and that it behaves robustly in the presence of motion. © 2004 American
Association of Physicists in Medicine. 关DOI: 10.1118/1.1773622兴

Key words: cone-beam CT, cone-beam reconstruction, helical CT, PI, wedge

I. INTRODUCTION et al. and Shechter et al. independently.9,13 The focus of

The introduction of cone-beam computer tomography 共CT兲
systems offers a couple of benefits. Compared with a single
slice CT system, the time for data acquisition can be reduced,
the output of the x-ray tube is used more efficiently leading
to a simpler heat management of the tube, and there is no
longer need for compromising z resolution in order to
achieve an acceptable scan time. However, a major problem
using cone-beam CT systems is to find a proper reconstruc-
tion algorithm. A large number of publications within the last
few years has addressed the problem of reconstruction of
cone-beam projections acquired during a helical path of the
source-detector system. Exact or quasiexact algorithms exist
only for a PI and a 3-PI acquisition so far.1–5 However, from
a practical point of view, approximate methods still have
advantages for medium size cone angles, in particular if they
can use an arbitrary amount of redundant data, because using
redundant data is known to reduce sampling artifacts in the z
direction significantly6 and the sensitivity to patient
motion.7,8 Furthermore, most approximate methods can use
the full detector area at arbitrary pitch values, thus using all
doses applied to the patient, and support the possibility to use
the ray offset technique for resolution enhancement.9 An-
other advantage may arise from a reduced computational ef-
fort. The frequency split method may be applied to any of
these approximate methods. We apply it here exemplarily to
the WEDGE method.10,11
The basic idea of the frequency split method has been
applied first by Tanaka12 for three-dimensional 共3D兲 position
emission tomography 共PET兲 imaging and was applied to
cone-beam CT imaging by Kudo
Kudo et al. was the dose utilization, while Shechter et al.
introduced the concept in the context of high-resolution im-
aging using the ray-offset technique. Here, we apply the
technique to the WEDGE method and investigate the behav-
ior with respect to cone-beam artifacts, noise, and patient
motion.
This paper is organized as follows: In Sec. II, the
WEDGE method is reviewed. Furthermore, the frequency
split technique is introduced for the handling of redundant
data. Section III shows evaluation results for the method
in terms of cone-beam artifacts, noise properties, and
the sensitivity to patient motion. Section IV summarizes
the results.
II. THEORY
The x-ray source moves on a helical path
冉
S 共 ␭ 兲 ⫽ R cos ␭,R sin ␭, P
HR ␭
L 2␲ 冊 共1兲
around the object. The relative pitch P is the table travel per
rotation in units of the detector height H projected onto the
center of rotation and R the radius of the helix. A focus-
centered, cylindrical detector at distance L from the x-ray
source is used throughout this paper. The fan-angle ␤ 共with
⫺ ␤ 0 /2⭐ ␤ ⭐ ␤ 0 /2) and the height h 共with ⫺H/2⭐h⭐H/2)
are used to parametrize the measured data on the detector
area. The position of the detector is given by

2230 Med. Phys. 31 „8…, August 2004 0094-2405Õ2004Õ31„8…Õ2230Õ7Õ$22.00 © 2004 Am. Assoc. Phys. Med. 2230
2231 Shechter et al.: The frequency split method for helical cone-beam reconstruction
FIG. 1. Geometry of the data acquisition system 共left兲 and the beam shape
after WEDGE-beam transformation.
D 共 ␭, ␤ ,h 兲 ⫽S 共 ␭ 兲 ⫹ 共 ⫺L cos共 ␭⫹ ␤ 兲 ,⫺L sin共 ␭⫹ ␤ 兲 ,h 兲 .
共2兲
Cone-beam projection data are acquired according to
p 1 共 ␭, ␤ ,h 兲 ⫽ 冕0
⬁
f 共 S 共 ␭ 兲 ⫹lE 共 ␭, ␤ ,h 兲兲 dl,
where E(␭, ␤ ,h) is the unit vector pointing from the source
position S(␭) to the detector element D(␭, ␤ ,h).
A. The original WEDGE method
We shortly review the main tasks to be done for the
WEDGE method.
The first step in the WEDGE method is a parallel rebin-
ning, which is also called WEDGE-beam-transform.10 Each
row of the detector is handled separately. The new data are
parametrized by the view angle ␪, which is the angle of the
parallel lines projected onto the xy plane, the distance of the
lines from the rotation axis u and finally the unchanged co-
ordinate h, see also Fig. 1:
␪ ⫽␭⫹ ␤ ,
u⫽R sin ␤ ,
p 2 共 ␪ ,u,h 兲 ⫽p 1 共 ␪ ⫺arcsin共 u/R 兲 ,arcsin共 u/R 兲 ,h 兲 .
Geometrically, the cone beam is transformed into a
wedge-shaped beam, which gave rise to the name of the
transform.
Due to the focus centered, cylindrical geometry of
the physical detector, it appears that the line integrals
within each row of the rebinned detector correspond to
lines which are truly parallel to each other. The next step
in reconstruction is the usual weighting with the cosine
of the cone angle,
L For a low pitch helical scan, the phenomenon of inter-
p 3 共 ␪ ,u,h 兲 ⫽ p 2 共 ␪ ,u,h 兲 .
冑h 2
⫹L 2
These data are filtered using a modified band-limited ramp
filter. The ramp-filter H ramp( ␯ ) itself is calculated as de-
scribed in Ref. 14. This filter is multiplied in the Fourier
domain by another function M ( ␯ ) which is used to taper the
resolution of the final image. The filter step yields projection
data
Medical Physics, Vol. 31, No. 8, August 2004
2231
共3兲
FIG. 2. Exemplary overscan weighting functions for different illumination
windows according to Eq. 共13兲. Illumination windows 关 ␪ f : ␪ l 兴 from top to
bottom are 关⫺␲/2:␲/2兴, 关⫺2␲/3:2␲/3兴, 关⫺␲:␲兴, and 关⫺1.3␲:1.3␲兴.
p 4 共 ␪ ,u,h 兲 ⫽ p 3 共 ␪ ,u,h 兲 * 共 h ramp共 u 兲 * m 共 u 兲兲 , 共8兲
where h ramp and m correspond to the Fourier transform of
H ramp and M, respectively. Finally, the image is reconstructed
using 3D back-projection according to
冕
1
f 共 x 兲⫽ w 共 ␪ ,x 兲 p 4 共 ␪ ,u 共 ␪ ,x 兲 ,h 共 ␪ ,x 兲兲 d ␪ , 共9兲
2␲ I共 x 兲
where the functions u( ␪ ,x) and h( ␪ ,x) provide the
coordinates of the object point x projected onto the WEDGE
detector, and the integration is performed over the set
I(x) of all projection angles, where the object point
共4兲
x is within the cone. The weighting function w( ␪ ,x) has
共5兲 to fulfill the following normalization condition for all
␪ 0 苸 关 0: ␲ 兴 :
共6兲
冕 兺
I 共 x 兲 兵 i苸N 其
␦ 共 ␪ 0 ⫺i ␲ 兲 w 共 ␪ ,x 兲 d ␪ ⫽1. 共10兲
Various methods are possible for choosing the weighting
function. In Ref. 15, we set w( ␪ ,x)⫽1, if ␪ is within the PI
window and 0 elsewhere. This is called WEDGE-PI. Heus-
cher and Brown16 prefer to choose w( ␪ ,x) according to the
aperture of the ray passing x, which is called aperture
weighted WEDGE. Here, we calculate the weight from the
illumination window of the voxel. We call this procedure
angular weighted WEDGE.
共7兲
rupted illumination occurs.17 This means that some object
points enter and leave the cone several times and it leads to a
noncontinuous illumination interval. For the sake of simplic-
ity, we restrict ourselves to the so-called central illumination
window. This is the largest interval 关 ␪ f (x): ␪ l (x) 兴 which in-
cludes the well-known PI-illumination window I PI(x)
⫽ 关 ␪ PI
f (x): ␪ l (x) 兴 .
PI 18
In other words, ␪ f (x) is the smallest
angle for which 关 ␪ f (x): ␪ PI l (x) 兴 contains only projection
2232 Shechter et al.: The frequency split method for helical cone-beam reconstruction 2232

angles ␪ for which the point x is still in the wedge. ␪ l (x) is

defined, respectively. Let

␣共 x 兲⫽ 再 ␲, ␪ l 共 x 兲 ⫺ ␪ f 共 x 兲 ⬎2 ␲ ,
共 ␪ l 共 x 兲 ⫺ ␪ f 共 x 兲兲 ⫺ ␲ , else,
共11兲

be the overscan angle and

再 ␲
T 共 x 兲 ⫽ ␪ l共 x 兲 ⫺ ␪ f共 x 兲 ⫺ ␲
1, else
, ␪ l 共 x 兲 ⫺ ␪ f 共 x 兲 ⬎2 ␲
共12兲

the plateau value. The weighting function w( ␪ ,x) is then

chosen as
w 共 ␪ ,x 兲 FIG. 3. Exemplary MTF functions for the low-frequency image 共top兲, the
high-frequency image 共middle兲, and the overall MTF 共bottom兲. F 1 ⫽0.38,
␪⬍␪ f 共 x 兲,

冦
0, F 2 ⫽0.95, F 3 ⫽0.05, F 4 ⫽0.12.

T 共 x 兲共 ␪ ⫺ ␪ f 共 x 兲兲 / ␣ 共 x 兲 , ␪ f 共 x 兲⭐␪⬍␪ f 共 x 兲⫹␣共 x 兲,
⫽ T共 x 兲, ␪ f 共 x 兲 ⫹ ␣ 共 x 兲 ⭐ ␪ ⬍ ␪ l共 x 兲 ⫺ ␣ 共 x 兲 , WEDGE reconstruction, the product of the ramp with M ( ␯ )
would be used in the filter step. We use two parameters F 1
T 共 x 兲共 ␪ l 共 x 兲 ⫺ ␪ 兲 / ␣ 共 x 兲 , ␪ l共 x 兲 ⫺ ␣ 共 x 兲 ⭐ ␪ ⬍ ␪ l共 x 兲 ,
and F 2 to parametrize the shape of M. Here, F 1 is the start
0, ␪ l共 x 兲 ⭐ ␪ . and F 2 the end of a cosine shaped fall-off according to
共13兲 for ␯ ⭐F 1 ␯ Ny ,

The shape of this weighting function is illustrated in Fig. 2
for different illumination window widths.
B. The frequency split method
If the original WEDGE method is used without redundant
data, we observe only very little cone-beam artifacts, even
for large cone angles.15 However, if it is applied to scans
with a small pitch and a large cone angle, we observe sig-
nificant artifacts for both, aperture weighted WEDGE and
angular weighted WEDGE. This is shown in Sec. III for
angular weighted WEDGE. The artifacts have typically a
very low spatial frequency in planes perpendicular to the
rotation axis. This has been observed also for other 3D re-
construction algorithms.12,13 This motivated the following
approach to handle redundant data: Two separate images are
reconstructed. The first one contains the low spatial fre-
quency components in planes perpendicular to the rotation
axis. This is reconstructed using no or only little redundant
data. The second one contains the high spatial frequencies
and this is reconstructed using all redundant data. It may be
noted that one advantage of using redundant data is the re-
duction of aliasing artifacts in the xy plane. Since these ar-
tifacts show up in high spatial frequencies, the frequency
split method still maintains this feature. 共Sampling artifacts
in z direction,6 also known as windmill artifacts,19 are of
course not reduced by the quarter detector offset and are also
not expected to be suppressed by the frequency split method
since they show up in the medium frequency range.兲
A modification of the ramp filtering 关Eq. 共8兲兴 is tradition-
ally used to control the resolution in the xy plane by multi-
plication with a modulation transfer function 共MTF兲. The
frequency split is realized by a modification of this MTF
function. Suppose we want to achieve an MTF given in the
frequency domain as a function M ( ␯ ). For standard
冦
1,
M F 1 ,F 2 共 ␯ 兲 ⫽
1
2 冉
1⫹cos 冉
␲ 共 ␯ ⫺F 1 ␯ Ny兲
共 F 2 ⫺F 1 兲 ␯ Ny
冊冊 ,
共14兲
for F 1 ␯ Ny⬍ ␯ ⭐F 2 ␯ Ny ,
0, else,
with ␯ Ny being the Nyquist frequency. In the frequency split
method, we use two functions M Flf ,F ( ␯ ) and
3 4
M Fhf ,F ,F ,F ( ␯ ) to reconstruct the low-frequency and the
1 2 3 4
high frequency image, respectively. The two parameters F 3
and F 4 are used as F 1 and F 2 in Eq. 共14兲 to control the
resolution of the low-frequency image. The modulation
transfer function for the high-frequency image is simply
M Fhf
1 ,F 2 ,F 3 ,F 4
共 ␯ 兲 ⫽M F 1 ,F 2 共 ␯ 兲 ⫺M Flf 3 ,F 4 共 ␯ 兲 . 共15兲
The modulation transfer functions are illustrated in Fig. 3.
As Kudo et al.,13 we reconstruct the low frequency image
not from strictly nonredundant data, but we use a small
amount of redundant data ␻ 共which is called ⌬ by Kudo
et al.兲. The amount is another parameter for the reconstruc-
tion. The low frequency image is reconstructed using an an-
gular range of ␲⫹␻ centered around the PI illumination win-
dow I PI(x). The reason to use some redundant data is to
reduce the sensitivity of the algorithm to patient motion. This
will be demonstrated in Sec. III. The high frequency image is
always reconstructed using all redundant data.
III. EVALUATION
Head imaging is probably the most demanding applica-
tion for cone-beam algorithms because of the very complex
bone structures inside the head, which can easily cause cone-
beam artifacts. Thus, we picked a head phantom for the
evaluation of cone-beam artifacts. For the analysis of the
sensitivity to motion, we took the cardiac phantom described

Medical Physics, Vol. 31, No. 8, August 2004

2233 Shechter et al.: The frequency split method for helical cone-beam reconstruction 2233

TABLE I. Definition of the scanners used in the simulations. L is the distance TABLE II. RMS noise level in HU calculated in a circular region of interest
from the source to the detector, R the distance of the source to the rotation containing 5025 voxels on the left-hand side of the phantom. Images were
axis, and H the detector height, and P relative pitch. reconstructed using standard WEDGE. They are shown in Fig. 4.

No. rows 64 No redundant data All redundant data

No. columns 672
No. vertices/turn 1160 High pitch 6.17 5.83
L/mm 1040 Low pitch 9.62 5.76
R/mm 570
H/mm 73
Active pixel height/mm 1.055
Active pixel width/mm 1.117 In the frequency split method, some data are discarded.
Focal spot size/mm2 0.6⫻1.0 Thus, the signal to noise ratio may be reduced compared to
Full fan angle ␤ 0 26.1° the standard WEDGE, which would be a reduced dose utili-
Detector width/mm 473.8
Full cone angle 4.02°
zation of the method. As a measure for the dose utilization,
P for low pitch scan 0.6 we calculated the root-mean-square 共rms兲 deviation from the
P for high pitch scan 1.5 average reconstructed absorption value in a homogeneous,
circular region of 5025 voxels on the left-hand side of the
phantom. We do not aim at making an absolute statement
about the dose utilization. The aim is to evaluate the loss in
in Ref. 20. A scanner with 64 rows was used for the simula- signal-to-noise ratio of the frequency split method compared
tion. The detailed scanner geometry is given in Table I. Note with the standard WEDGE method.
that the fan angle of the system is unusually small. This is
due to the fact that the phantoms used in the study are both 1. Standard WEDGE
rather small and fit completely into the fan angle of 26°. Figure 4 shows reconstruction results for the standard
Furthermore, it should be noted that the height and width of WEDGE method at high and low pitch and different amount
the detector pixel is given as active height and width. These of redundant data usage. The rms noise values are summa-
quantities define the area, where x-ray photons are detected. rized in Table II.
In order to suppress aliasing in the xy plane, we mimicked a We observe that for the high-pitch case shown in the up-
dual-focus acquisition by using a larger active width than the per row of Fig. 4, the utilization of redundant data introduces
pixel spacing. Thus, the simulation can be seen as a direct or enhances low-frequency artifacts, i.e., dark or bright
simulation of the data, which are obtained in a real system shades, close to the inner ear on the right-hand side of the
after interleaving data from the two focus positions. The Ny- phantom. The same holds true for the low-pitch case shown
quist frequency ␯ Ny in our simulation is 12.94 cm⫺1.
The frequency split method is designed for low pitch he-
lical scans. Thus, we tested the algorithm for a low pitch
helical scan. However, in order to complete the study and to
demonstrate that the problems we observe for low pitch are
mainly related to the use of redundant data, we included
simulations at high pitch and reconstruction results for the
standard WEDGE method.
In Sec. III A, we will focus on pure cone-beam artifacts
using a static head phantom. At the same time, we measure
image noise in order to evaluate the dose utilization. In Sec.
III B, the behavior with respect to motion is analyzed.
A. Analysis of cone-beam artifacts and dose utilization
The forbild head phantom 共as given in www.imp.uni-
erlangen.de/forbild兲 was used for the cone-beam artifact
evaluation. The objects designed for resolution measurement
were omitted. We showed already in Ref. 9 that the resolu-
tion of the standard WEDGE method is maintained by the
frequency split method.
In order to test the dose utilization of the frequency split
method, we added Poisson distributed noise to the projec-
tions. The simulated nonattenuated number of photons was
0.6⫻106 and 1.5⫻106 per single detector pixel for the low- FIG. 4. Results for the standard WEDGE method at different pitches and for
different amount of redundant data usage at z⫽⫺6 mm. Top, results from
pitch and high-pitch dataset, respectively. For the pitch val-
the high pitch scan. Bottom, results for the low pitch scan. Left, no redun-
ues used, this implies that the dose applied to the phantom is dant data are used; right, full usage of redundant data. Level 35 HU, window
the same for both data sets. 100 HU.

Medical Physics, Vol. 31, No. 8, August 2004

2234 Shechter et al.: The frequency split method for helical cone-beam reconstruction
in the lower row. Here, the effect is even stronger because
approximately three times more data were used by the recon-
struction for the right image than for the left one. This causes
also the significant reduction in the noise level, see Table II.
If all redundant data are used, the resulting noise level is
almost the same for both pitches, as expected since the simu-
lated dose for the two cases is the same. The argument about
the same dose is strictly true only for object points on the
rotation axis, where the ratio illumination window is the
same as the ratio of the relative pitches. Since the region,
where we computed the noise level is slightly off-center, a
small difference in the noise level is expected.
2. Frequency split WEDGE
Figure 5 shows exemplarily the low frequency, the high
frequency, and the final image for the frequency split method
applied to the low-pitch scan. Results for the frequency split
method applied to the low-pitch scan are shown in Fig. 6 for
different values of the overscan angle ␻ and cutoff frequency
F 4 . The rms noise values are summarized in Table III.
We observe, that the low-frequency artifacts are com-
pletely suppressed by the frequency split method for both
cutoff frequencies and ␻⭐␲/3. For a larger overscan of
␻⫽2␲/3, the dark and bright shades near the inner ear, which
are present in the standard WEDGE reconstruction in the
lower right image of Fig. 4 become slightly visible.
The noise measurement confirms that for the low
cutoff frequency F 4 ⫽0.15, the dose utilization of the
frequency split method is virtually the same as for standard
Medical Physics, Vol. 31, No. 8, August 2004
2234
FIG. 5. Reconstruction results for the frequency split
method with F 3 ⫽0, F 4 ⫽0.3 and low-pitch scan of the
head phantom at z⫽⫺6 mm. Left, low-resolution im-
age using ␻⫽␲/5; middle, high-frequency image using
all redundant data; right, final image for WEDGE with
the frequency split method. Level 35 HU, window 100
HU for the low-resolution and the final image, level
⫺1000 HU, window 100 HU for the high-frequency
image.
WEDGE. However, for the higher cutoff frequency F 4
⫽0.3, we observe a small increase of the noise level by 3%.
B. Sensitivity to motion artifact
In order to analyze the sensitivity to patient motion, we
used the cardiac phantom as described in Ref. 20. The phan-
tom had 60 beats per minute and a rest period of 20%. The
simulated time for one rotation of the scanner was 420 ms.
The other scan parameters are the same as for the low pitch
scan of the head phantom. Results for the frequency split
method for different values of the overscan ␻ and the cutoff
frequency F 4 are shown in Fig. 7. For comparison, Fig. 8
shows the result for the standard WEDGE method. The goal
of the evaluation here is not to recover the small spheres
well—we do not evaluate a cardiac CT algorithm—but to get
as little artifacts as possible in the surrounding area. It might
be seen as a simulation of a thorax scan, where we do not
want to get artifacts inside the nonmoving tissue from a
moving stent placed in a coronary artery.
Comparing the upper and the lower row of Fig. 7, we
observe that the sensitivity to patient motion increases with
an increased cutoff frequency F 4 . This can be explained by
the fact that a low cutoff frequency implies that more fre-
quency components are reconstructed using the full overscan
range. The sensitivity is reduced if a larger overscan angle ␻
is used. However, the biggest decrease in motion artifacts is
achieved by going from ␻⫽0 共left column of Fig. 7兲 to
␻⫽␲/3 共middle column兲. Increasing the overscan angle fur-
ther to ␻⫽2␲/3 共right column兲, has only a little effect on the
FIG. 6. Results for the frequency split method and the
low-pitch scan of the head phantom at z⫽⫺6 mm. Top
row, F 3 ⫽0, F 4 ⫽0.15; bottom row, F 3 ⫽0, F 4 ⫽0.3.
Left, middle, right column, ␻⫽0, ␻⫽␲/3, ␻⫽2␲/3.
Level 35 HU, window 100 HU.
2235 Shechter et al.: The frequency split method for helical cone-beam reconstruction
TABLE III. RMS noise level in HU calculated in a circular region of interest
containing 5025 voxels on the left-hand side of the phantom. Images were
reconstructed using the frequency split method. They are shown in Fig. 6.
For all cases, we used F 3 ⫽0.
␻⫽0 ␻⫽␲/3 ␻⫽2␲/3
F 4 ⫽0.15 5.76 5.76 5.77
F 4 ⫽0.30 5.92 5.90 5.90
motion artifacts. This conclusion obtained from a visual in-
spection is also supported by a quantitative analysis of the
motion artifacts. We calculated the root-mean-square devia-
tion from the true background value in a rectangular region
indicated in Fig. 8. The results are summarized in Table IV.
The frequency split method applied to WEDGE becomes
the standard WEDGE method for the both limits, F 4 →0 and
␻→max. Thus, the result of the standard WEDGE method as
shown in Fig. 8 represent the best possible result that we can
obtain with the frequency split method applied to WEDGE.
It appears that an overscan angle of ␻⫽␲/3 is sufficient to
reach almost the same level of artifact suppression as the
standard WEDGE method.
IV. DISCUSSION
Like in many other approximate methods, the way redun-
dant data are used in the WEDGE method is motivated and
justified by two-dimensional 共2D兲 arguments: Rays passing a
certain object point are considered to be equivalent if they
are parallel in the projection onto the plane perpendicular to
the rotation axis. Of course, redundancy in cone-beam CT is
not related to line integrals but to Radon planes. As a matter
of fact, not all Radon planes are measured multiple times for
a low-pitch helical CT scan.17,5 In particular, planes, which
have only one intersection point with the helix, can be mea-
sured only once. Geometrically, these planes are almost per-
pendicular to the rotation axis. According to the 3D Fourier
slice theorem, these Radon planes contain predominantly in-
formation about spatial frequencies along the rotation axis.
Medical Physics, Vol. 31, No. 8, August 2004
2235
TABLE IV. Artifact levels in the motion phantom. The RMS deviation from
the true background value was calculated in the square indicated in Fig. 8.
The first three columns are related to the images shown in Fig. 7, full
overscan is equivalent to standard WEDGE shown in Fig. 8. For all cases,
we used F 3 ⫽0.
␻⫽0 ␻⫽␲/3 ␻⫽2␲/3 Full overscan
F 4 ⫽0.15 2.6 1.3 1.1 1.0
F 4 ⫽0.30 4.8 1.4 1.2 1.0
Alternatively, it is valid to say that these Radon planes con-
tain only low-frequency information in planes perpendicular
to the rotation axis.
The observation that the WEDGE method exhibits low-
frequency artifacts may be explained by the facts mentioned
above. In the WEDGE method, redundancies are treated like
in 2D CT. The fact that some Radon planes are not measured
multiple times is neglected. The structure of the resulting
artifacts can be related by the Fourier slice theorem to an
improper weighting of these Radon planes.
The frequency split method is a heuristic approach to
mitigate this effect. Cone-beam artifacts are reduced signifi-
cantly, see Figs. 5 and 6. This improvement is achieved over
a wide range for the parameter values F 4 and ␻.
The choice for the overscan parameter ␻ and the cutoff
frequency is obviously a trade-off between the performance
of the algorithm with respect to cone-beam artifacts, dose
utility, and robustness with respect to patient motion. If we
consider cone-beam artifacts, it appears that the overscan
must not be chosen large, see right-most images in Fig. 6.
The aspect of dose utilization pushes the cutoff frequency
down and the overscan to large values. Similarly, motion
artifacts are better suppressed for a low cutoff frequency F 4
and a large overscan angle, see Fig. 7. However, it appears
that these parameters are not very critical for the scanner that
we investigated here. With a proper choice of the cutoff fre-
quency and the overscan angle, namely F 4 ⫽0.15 and
␻⫽␲/3, we can maintain the good dose utilization and sup-
pression of motion artifacts of the WEDGE method, while
reducing cone-beam artifacts significantly.
FIG. 7. Results for the frequency split method and the
low-pitch scan of the cardiac phantom at z⫽3 mm. Top
row, F 3 ⫽0, F 4 ⫽0.15; bottom row, F 3 ⫽0, F 4 ⫽0.3.
Left, middle, right column, ␻⫽0, ␻⫽␲/3, ␻⫽2␲/3.
Level 0 HU, window 100 HU.
2236 Shechter et al.: The frequency split method for helical cone-beam reconstruction
FIG. 8. Result for the standard WEDGE method and the low-pitch scan of
the cardiac phantom at z⫽3 mm. The rectangle indicates the region where
the motion artifact level is calculated, see Table IV. Level 0 HU, window
100 HU.
Since Kudo et al.13 observed similar results for the aspect
of cone-beam artifacts and dose utility by applying the fre-
quency split method to a different cone-beam reconstruction
method,13 it seems to be a very generic approach to improve
image quality of approximative methods for low-pitch cone-
beam CT.
As already pointed out by Kudo et al.,13 the additional
computational cost is relatively low, because the reconstruc-
tion of the two images can be in fact combined into a single
back-projection loop, all the geometrical calculations during
back projection have to be carried out only once. The pre-
processing time is increased by a factor of approximately
1.5, because the most time consuming part here is the Fou-
rier transformation and we only need an additional inverse
Fourier transformation of the projection data.
V. SUMMARY
The angular weighted WEDGE method shows artifacts
for a simulated 64-row scanner if redundant data are used.
These artifacts are more serious if the pitch is chosen small.
The artifacts show up as bright or dark shades with very low
frequency components within the xy plane.
We introduced the frequency split method as a generic
method to suppress these artifacts. We applied it to the an-
gular weighted WEDGE reconstruction algorithm and ana-
lyzed it with respect to artifacts, noise, and sensitivity to
patient motion. The method introduces some additional pa-
rameters, which need to be selected carefully. In the case of
a 64-row scanner, we found a set of parameters, for which
the method suppresses artifacts very efficiently without in-
creasing the noise level. The algorithm appears to be quite
robust with respect to patient motion. For other scanner or
other base methods than WEDGE, other parameter values
may lead to the best trade-off between the mentioned quality
measures.
Medical Physics, Vol. 31, No. 8, August 2004
2236
1
K. C. Tam, S. Samarasekera, and F. Sauer, ‘‘Exact cone beam CT with a
spiral scan,’’ Phys. Med. Biol. 43, 1015–1024 共1998兲.
2
M. Defrise, F. Noo, and H. Kudo, ‘‘A solution to the long object problem
in helical cone-beam tomography,’’ Phys. Med. Biol. 45, 623– 643
共2000兲.
3
A. Katsevich, ‘‘Analysis of an exact inversion algorithm for spiral cone-
beam CT,’’ Phys. Med. Biol. 47, 2583–2597 共2002兲.
4
A. Katsevich, ‘‘Theoretically exact FBP-type inversion algorithm for spi-
ral CT,’’ SIAM 共Soc. Ind. Appl. Math.兲 J. Appl. Math. 62, 2012–2026
共2002兲.
5
C. Bontus, T. Köhler, and R. Proksa, ‘‘A quasiexact reconstruction algo-
rithm for helical CT using a 3-Pi acquisition,’’ Med. Phys. 30, 2493–2502
共2003兲.
6
T. Köhler, R. Proksa, C. Bontus, M. Grass, and J. Timmer, ‘‘Artifact
analysis of approximate cone-beam CT algorithms,’’ Med. Phys. 29,
51– 64 共2002兲.
7
C. Bontus, R. Proksa, J. Timmer, T. Köhler, and M. Grass, ‘‘Movement
artifacts in helical CT cone-beam reconstruction,’’ Proceedings of the
Sixth International Meeting on fully three-dimensional image reconstruc-
tion in radiology and nuclear medicine, 2001, pp. 199–202.
8
K. Taguchi, ‘‘Temporal resolution and the evaluation of candidate
algorithms for four-dimensional CT,’’ Med. Phys. 30, 640– 650
共2003兲.
9
G. Shechter, T. Köhler, A. Altman, and R. Proksa, ‘‘High-resolution im-
ages of cone beam collimated CT scans,’’ Nuclear Science Symposium
Conference Record, IEEE, 2003.
10
H. K. Tuy, ‘‘3D image reconstruction for helical partial cone beam scan-
ners,’’ in Proceedings of the 3D’99 Conference 共Egmond aan Zee, The
Netherlands, 1999兲, pp. 7–10.
11
G. Shechter, G. Naveh, A. Altman, R. Proksa, and M. Grass, ‘‘Cardiac
image reconstruction on a 16-slice CT scanner using retrospective ECG-
gated, multicycle 3d back-projection algorithm,’’ Proc. SPIE 5032, 1820–
1830 共2003兲.
12
E. Tanaka, S. Mori, K. Shimizu, E. Yosikawa, and H. Murayama, ‘‘Mov-
ing slice septa and pseudo 3-D reconstruction from multi-ring PET,’’
Phys. Med. Biol. 37, 661– 672 共1992兲.
13
H. Kudo, F. Noo, M. Defrise, and T. Rodet, ‘‘New approximate filtered
backprojection algorithm for cone-beam helical CT with redundant data,’’
Nuclear Science Symposium Conference Record, IEEE, 2003.
14
M. Magnusson, ‘‘Linogram and other direct Fourier methods for tomog-
raphic reconstruction,’’ Tech. Rep. Ph.D. thesis No. 320, Linköping Uni-
versity, 1993.
15
T. Köhler, C. Bontus, K. Brown, D. Heuscher, M. Grass, G. Shechter, and
R. Proksa, ‘‘Evaluation of helical cone-beam CT reconstruction algo-
rithms,’’ Nuclear Science Symposium Conference Record, IEEE, 2002,
pp. 1217–1220.
16
D. Heuscher and K. Brown, ‘‘Redundant data and exact helical cone-
beam CT,’’ Proceedings of the 3D’03 Conference, Saint Malo, France,
2003, pp. Tu AM2– 4.
17
R. Proksa, T. Köhler, M. Grass, and J. Timmer, ‘‘The n-PI-method
for helical cone-beam CT,’’ IEEE Trans. Med. Imaging 19, 848 – 863
共2000兲.
18
P. E. Danielsson, P. Edholm, J. Eriksson, M. Magnusson-Seger, and H.
Turbell, ‘‘The original PI-method for helical cone-beam CT,’’ Proceed-
ings of the 3D’99 Conference 共Egmond aan Zee, The Netherlands, 1999兲,
pp. 3– 6.
19
M. D. Silver, K. Taguchi, I. A. Hein, K. S. Han, M. Kazama, and I. Mori,
‘‘Windmill artifact in multislice helical CT,’’ Proc. SPIE 5032, 1918 –
1927 共2002兲.
20
M. Kachelriess, S. Ulzheimer, and W. A. Kalender, ‘‘ECG-correlated im-
age reconstruction from subsecond multi-slice spiral computed tomogra-
phy scans of the heart,’’ Med. Phys. 27, 1881–1902 共2000兲.