Home Search Collections Journals About Contact us My IOPscience

Scatter correction method for cone-beam CT based on interlacing-slit scan

This content has been downloaded from IOPscience. Please scroll down to see the full text.

2014 Chinese Phys. B 23 098106

(http://iopscience.iop.org/1674-1056/23/9/098106)

View the table of contents for this issue, or go to the journal homepage for more

Download details:

IP Address: 169.230.243.252
This content was downloaded on 21/02/2015 at 10:35

Please note that terms and conditions apply.

 Chin. Phys. B Vol. 23, No. 9 (2014) 098106

Scatter correction method for cone-beam CT

based on interlacing-slit scan∗
Huang Kui-Dong(黄魁东)a)† , Zhang Hua(张 华)b) , Shi Yi-Kai(史仪凯)b) ,
Zhang Liang(张 亮)a) , and Xu Zhe(徐 哲)a)
a) Key Laboratory of Contemporary Design and Integrated Manufacturing Technology of the Ministry of Education,
Northwestern Polytechnical University, Xi’an 710072, China
b) School of Mechanical Engineering, Northwestern Polytechnical University, Xi’an 710072, China

(Received 4 November 2013; revised manuscript received 3 March 2014; published online 16 July 2014)

Cone-beam computed tomography (CBCT) has the notable features of high efficiency and high precision, and is
widely used in areas such as medical imaging and industrial non-destructive testing. However, the presence of the ray
scatter reduces the quality of CT images. By referencing the slit collimation approach, a scatter correction method for
CBCT based on the interlacing-slit scan is proposed. Firstly, according to the characteristics of CBCT imaging, a scatter
suppression plate with interlacing slits is designed and fabricated. Then the imaging of the scatter suppression plate is
analyzed, and a scatter correction calculation method for CBCT based on the image fusion is proposed, which can splice
out a complete set of scatter suppression projection images according to the interlacing-slit projection images of the left
and the right imaging regions in the scatter suppression plate, and simultaneously complete the scatter correction within the
flat panel detector (FPD). Finally, the overall process of scatter suppression and correction is provided. The experimental
results show that this method can significantly improve the clarity of the slice images and achieve a good scatter correction.

Keywords: cone-beam computed tomography, scatter correction, interlacing slits, image fusion
PACS: 81.70.Tx, 87.57.Q–, 87.64.Bx, 87.57.C– DOI: 10.1088/1674-1056/23/9/098106

1. Introduction geometry conditions of the X-ray imaging, can reduce the

amount of scatter reaching the detector by means of adding
In a cone-beam computed tomography (CBCT) system,
hardware devices. The hardware methods mainly include
the projection images of a tested object are collected from a flat
collimator, air gap, [9] bucky, [10] radiopaque strips, honey-
panel detector (FPD) when the cone beam X-ray source is on,
comb grid, [11] and so on. The software methods, which han-
then serial slice images are reconstructed with the correspond-
dle the projection images based on digital image processing
ing reconstruction algorithm from these images. The CBCT
methods, obtain the scatter distribution through the analysis
has many advantages, such as high scan speed, the same space
of the images themselves and the estimation of the tested
distinguishing ratio in the slice inner and between slices, and
object properties, and adopt the post-processing methods to
high precision. The CBCT has been extensively applied in
correct the scatter. The software methods mainly include
medical imaging and industrial non-destructive testing. [1,2] convolution, [12,13] deconvolution, [14] the Monte Carlo (MC)
The scatter is an important factor which can influence the simulation method, [15–17] and so on. The MC method, which
image quality in the CT imaging process. Due to the exis- uses the computer simulation to obtain the image scatter distri-
tence of the scatter, the image contrast is reduced, [3] and the bution, is an effective and accurate scatter correction method,
image details become blurred. It is very negative for the de- but is mostly limited to theoretical studies because of its time-
tection of tiny structures or defects. The scatter intensity and consuming calculation.
distribution are not only related with the ray energy, but also As a result of using plane array detectors in the CBCT,
related with the material, structure, thickness, area, and even the scatter field shows a two-dimensional distribution. Then
the whole imaging environment. As there are so many uncer- satisfactory results cannot be obtained through a single hard-
tain factors, the scatter correction is still a focused problem to ware or software method. So, there have been some hybrid
be solved. [4–7] correction methods combining hardwares and softwares, for
The scatter suppression and correction techniques can be example, the beam-stop array, [18] beam-hole array, [19] beam
generally divided into two types: hardware methods and soft- attenuation array, [20] stationary beam blocker, [21,22] moving
ware methods. [8] The hardware methods, which consider the blocker strips, [23,24] attenuation plate imaging, [25] and the pri-
∗ Projectsupported by the National Science and Technology Major Project of the Ministry of Industry and Information Technology of China (Grant
No. 2012ZX04007021), the Young Scientists Fund of the National Natural Science Foundation of China (Grant No. 51105315), the Natural Science Ba-
sic Research Program of Shaanxi Province, China (Grant No. 2013JM7003), and the Northwestern Polytechnical University Foundation for Fundamental
Research, China (Grant Nos. 3102014KYJD022 and JC20120226).
† Corresponding author. E-mail: [email protected]

© 2014 Chinese Physical Society and IOP Publishing Ltd 　http://cpb.iphy.ac.cn

http://iopscience.iop.org/cpb　

098106-1
 Chin. Phys. B Vol. 23, No. 9 (2014) 098106
mary ray modulation correction method. [26,27] In addition, in the plate needs to be larger than the detector’s imaging win-
someone has simulated the scatter from the perspective of sec- dow.
ondary radiation. [28] These methods, which are usually under The X-ray source of the experimental CBCT system is
some approximation assumptions, have achieved good scatter Y.TU 450-D02 of YXLON, whose focus is 2.5/5.5 mm and
correction results in the medical CT imaging. However, the maximum voltage is 450 kV. The FPD is PaxScan 2520 of
applicability of the methods used in the middle- and the high- Varian. Based on the above analysis, taking the ray shielding,
energy X-ray scans of the industrial CT imaging needs to be economy, and workability of the scatter suppression plate with
researched and tested further. interlacing slits into account, we designed the plate shown in
A new method of CBCT scatter suppression and correc- Fig. 1, with slit width W = 8 mm, slit spacing S = 15 mm, plate
tion based on the interlacing-slit scan is proposed in this arti- thickness T1 = 40 mm; the material used is lead-antimony al-
cle. This method can complete the scatter suppression outside loy (5% antimony).
the FPD and the scatter estimation within the FPD at the same
time, so the scatter correction result, which is more in line with L
the actual scatter condition in CBCT imaging systems, can be L1 L2 L1
obtained. The feasibility and practicality of this method are
verified through the experiments.

2. Scheme design of scatter suppression scan

H
2.1. Design of the scatter suppression plate with interlac-
ing slits
The basic idea of scatter suppression based on the B
A
interlacing-slit scan is to make use of the good inhibition of (a)
the slits on the scatter ray, image the two slit regions respec-
tively (the slits of the two regions are interlaced), and then

S
splice out a complete set of scatter suppression projection im-

W
ages. According to the characteristics of the CBCT imaging X-ray

S
W
system based on an FPD, we analyze and design the scatter
suppression plate with interlacing slits from the following as-
pects. T1 T2
(i) The suppression plate must be sufficient to fully shield D

the X-ray transmission in the range of voltage used. On this

(b) FPD
basis, the thicker the suppression plate is, the better the scatter
Fig. 1. (color online) Design diagram of the scatter suppression plate
suppression result can be. with interlacing slits: (a) front structures, (b) structural diagram of the
(ii) The suppression plate is divided into two imaging re- slits in the imaging region A.
gions which have some parallel slits, and the slits of the two
areas remain interlacing. There are two placement options of the scatter suppres-
(iii) Theoretically, the narrower the slits are, the better the sion plate with interlacing slits. One is to put it between the
scatter suppression result can be. However in order to make tested object and the FPD (position 1), the other is to put it
full use of the initial ray intensity of the ray source, the width between the tested object and the ray source (position 2). For
of the slits cannot be smaller than the size of the ray source position 1, the X-rays pass through the tested object first and
focus. then pass through the plate. In this case, only a small num-
(iv) To obtain a complete image of the tested object and to ber of scatter photons can pass through the slits to the FPD.
satisfy the need of the splice computing of the projection im- For position 2, the X-rays are collimated by the slits and then
ages, the width of the blocker strips between the slits should interact with the tested object, and both scatter photons and
be smaller than the width of the slits. penetration photons can reach the FPD. We choose position 1
(v) Taking the imaging cone angle of the CBCT scan to place the plate and make it as close as possible to the FPD.
into account, the slit angles need to concentrate on the ray The scatter suppression plate needs to be moved in the
source focus, which can make the penumbral regions of the process of the CBCT scan. In order to achieve the automa-
slits smallest under the condition of cone beam imaging. tion and continuity of the scan, we fix the plate on a transla-
(vi) To ensure no reduction of the detector’s effective tional worktable, whose movement control is integrated with
imaging area, the slit coverage of the left or right imaging area the CBCT scan system. In addition, in order to suppress the
098106-2
 Chin. Phys. B Vol. 23, No. 9 (2014) 098106
scatter from peripheral objects, the beam-limited block is fixed 3. Imaging analysis and scatter correction calcu-
on the exit of the ray source, and the non-image area of the lation
FPD is shielded. The design schematic diagram and the ac-
The CBCT scan experiment with the scatter suppression
tual system picture of the scatter suppression plate imaging
plate is carried out for an aluminum part. The scan voltage is
are shown in Fig. 2.
180 kV, the exposure is 0.2178 mA·s, the number of collected
ray shielding plate projection images of step (ii) and step (v) is both 9, and those
(a)
scatter suppression plate of step (iii) and step (iv) are both 360. The image resolution
is 1024×1024. One of each group of the projection images
beamlimited block
scanned with the interlacing slits is shown in Fig. 3, and the
tested object
grayscale comparison of the interlacing-slit scan is shown in
Xray Fig. 4.
rotary worktable
translational worktable
FPD BA C
(b)

(a) (b)

Fig. 2. (color online) Installation diagram of the scatter suppression

plate with interlacing slits: (a) schematic diagram, (b) actual system
picture.

2.2. Steps of the scatter suppression scan with interlacing

slits
(c) (d)
The CBCT scatter suppression based on the interlacing-
slit scan needs four groups of projection images, and the spe- Fig. 3. (color online) Projection images of interlacing-slit scan
(grayscale window: [98, 2827]): (a) projection image of slit A, (b) pro-
cific steps are as follows. jection image of slit A and the part, (c) projection image of slit B, (d)
(i) Determine the scan parameters of the CBCT system, projection image of slit B and the part.

and keep them unchanged during the entire scan.

(ii) Collect some projection images of slit A, and obtain It can be seen from Fig. 4(a) that compared with slit imag-
one projection image after averaging them, which is called im- ing and open imaging, the grayscale in the slit region decreases
age 1. 150–230 approximately. Since all scan parameters are consis-
(iii) Place the tested object on the rotary worktable, and tent, this part of decreasing grayscale can be considered as the
then collect a group of projection images of slit A plus the ob- suppression of the scatter photons by the slits.
ject through a circle scan, which are called image 2. From Fig. 4(b), it can be seen that there are apparent out-
(iv) Keep the tested object still (the starting position is puts in the blocker strip regions of the scatter suppression
the same as that in step 3), move the scatter suppression plate, plate. However, according to the design parameters of the
and then collect a group of projection images of slit B plus the plate, the regions are not likely to be irradiated by penetration
object through a circle scan, which are called image 3. rays. So the outputs must be the contribution of the scatter. As
(v) Remove the tested object, collect some projection im- the range of the visible light scatter of the scintillator is small
ages of slit B, and obtain one projection image after averaging and the width of the blocker strip is bigger, the main source
them, which is called image 4. of the outputs must be the ray scatter, which is caused by the
Image 1 and image 4 are used to obtain the fusion coef- related structures within the FPD. It is called the inner scatter
ficient image, generally only a little time is needed to collect of the FPD in this article.
the images. If the same or similar scan parameters are used, It can be seen from Fig. 4(b) that there are obvious dif-
the fusion coefficient image obtained before can be reused in ferences between the blocker strip region’s covered and not
the scatter correction, in this case, step (ii) and step (v) can be covered parts. The grayscale of the blocker strip region’s cov-
omitted. ered part is significantly lower, and the thicker the part is, the
098106-3
 Chin. Phys. B Vol. 23, No. 9 (2014) 098106
lower the grayscale will be. This means that the scatter dis- Based on the four groups of images obtained through the
tribution within the FPD is not uniform. It is related with the scatter suppression scan with interlacing slits for CBCT and
distribution of the received radiation in adjacent slit regions. the above-described imaging analysis, we propose a scatter
correction calculation method for the CBCT based on image
3000
slit imaging fusion, and the specific steps are as follows.
open imaging (i) According to image 1 and image 4, calculate the inner
2000
scatter field Sw without the tested object as follows.
Grayscale

1) Search for the maximum grayscale pixel in image 1

and calculate its neighborhood average, which is denoted as q.
1000 2) Determine the segmentation threshold of the blocker
strip regions, g = q/η, where η is general 7–9.
(a) 3) Generate two template images F1 and F2 which are the
0
0 200 400 600 same size as image 1, and set the initial values of them to 0.
Pixel sequence 4) Traversal image 1, and if the neighborhood grayscale
300 average of the pixel is less than g, set the corresponding pixel
in F1 to 1.
5) Traversal image 4, and if the neighborhood grayscale
200 average of the pixel is less than g, set the corresponding pixel
Grayscale

in F2 to 1.
6) Obtain S1 by multiplying image 1 and F1 correspond-
100
ingly, and obtain S2 by multiplying image 4 and F2 corre-
position B
position C spondingly.
(b) 0 by adding S and S correspondingly.
7) Obtain Sw 1 2
0
0 200 400 600 8) Make a copy of Sw 0 , called S , calculate the neighbor-
w
Pixel sequence 0 , and
hood summation t of the pixel whose grayscale is 0 in Sw
Fig. 4. (color online) Grayscale comparison on the lines in Fig. 3(b): (a) slit the number of non-zero pixels in the neighborhood is n, then
imaging and open imaging on position A, (b) position B and position C.
the grayscale of the pixel is s = t/n, store s to the correspond-
Due to the existence of the inner scatter of the FPD, the ing location in Sw .
grayscale of the slit regions must include the contribution of (ii) Calculate fusion coefficient image R according to im-
the inner scatter. According to the experimental analysis, the age 1 and image 4 as follows.
distribution of the inner scatter is relatively smooth, so it is 1) Plus image 1 and image 4 correspondingly and then
possible to make use of the inner scatter of the blocker strip minus 2Sw to obtain image J.
regions in slits A and B to acquire the integral scatter within 2) Set up J as the average value of J, then Rx,y = J/Jx,y .
the FPD in some manner. It is noteworthy that the projection (iii) According to every corresponding angle projection
images of both slits A and B (regardless of including or ex- image of image 2 and image 3, which are called I1 and I2 ,
cluding the part) contain an inner scatter field. So it needs to calculate the inner scatter field Sy with the tested object as fol-
be removed in the calculation of the scatter correction. lows.

image 1
slits A FPD scatter fusion
template
field image coefficients
image without object image

image 4
slits B

slice
image 2 FPD
images
slits A+ scatter FPD scatter reconstru
object field scatter
images corrected cted with
corrected
images scatter
with images
corrected
image 3 object
images
slits B+
object

Fig. 5. Overall flowchart of the scatter correction based on an interlacing-slit scan.

098106-4
 Chin. Phys. B Vol. 23, No. 9 (2014) 098106
1) Obtain S1 by multiplying I1 and F1 correspondingly, sion and correction projection images (FDK reconstruction al-
and obtain S2 by multiplying I2 and F2 correspondingly. gorithm commonly), the scatter corrected slice images of the
2) Obtain Sy0 by adding S1 and S2 correspondingly. tested object can be obtained. The overall flowchart of the
3) Make a copy of Sy0 , called Sy , calculate the neighbor- scatter correction based on the interlacing-slit scan is shown
hood summation t of the pixel whose grayscale is 0 in Sy0 , and in Fig. 5.
the number of non-zero pixels in the neighborhood is n, then
the grayscale of the pixel is s = t/n, store s to the correspond- 4. Experiment and analysis
ing location in Sy . Figure 6 is the images from the scatter suppression and
(iv) Plus I1 and I2 correspondingly and then minus 2Sy to correction experiment for the CBCT based on the interlacing-
obtain image C0 . slit scan. It can be seen that the grayscale of the scatter cor-
(v) Obtain the final scatter suppression and correction im- rected image decreases significantly, and the absolute value
age by multiplying C0 and R correspondingly. decreased in the blank area is a little larger than that in the
By the CBCT reconstruction with a set of scatter suppres- area with the tested object.

(a) (b) (e) (f)

3000
original
corrected
Grayscale

2000

1000

(g)
(c) (d)
0
0 200 400 600
Pixel sequence
Fig. 6. (color online) Experimental results of CBCT scatter suppression and correction based on interlacing-slit scan: (a) FPD scatter field image without
object, (b) FPD scatter field image with object, (c) fusion coefficient image, (d) FPD scatter corrected image, (e) scatter corrected image, (f) original projection
image, and (g) grayscale comparison on the same line A. Grayscale window: (a) [264, 325], (b) [128, 251], (c) [0.52, 1.3], (d) , (e), and (f) [98, 2827].

Figures 7 and 8 show the comparison of the 500th and rected slice images is amplified correspondingly.
the 572th slice images before and after the scatter correction, Table 1 is the numerical comparison of slice image qual-
where rectangle A is the comparison position of the signal to
ity of the 500th and the 572th slice images before and after the
noise ratio (SNR), [29] rectangle B is the comparison position
scatter correction in Figs. 7 and 8. It can be clearly seen from
of the contrast to noise ratio (CNR) [29] and average gradient
(AG), [30] and line C is the comparison position of grayscale. It the table that the SNR of the slice images processed by the
can be seen from the pictures that the contour sharpness of the scatter correction decreases significantly, and that of the 500th
scatter corrected slice images is improved significantly, and slice image, which comes from a thicker part, decreases more
this improvement is more outstanding in the parts of inner con- than that of the 572th slice image, which comes from a thin-
cave contours. On the other hand, we observe that the noise ner part. This is consistent with the previous analysis. While
in the scatter corrected slice images has a certain degree of
the other two evaluation items, CNR and AG, both increase
magnification, this is because the interlacing-slit scan method
obviously, especially in the parts of inner concave contours,
decreases the amount of X-ray photons received by the FPD
and the noise of the penetration rays is relatively stable, the AG has a great raise. This means that the scatter correction
proportion of the noise in the projection images is increased is directly effective to improve the sharpness of the internal
after the scatter correction, and the noise in the scatter cor- contour structures.
098106-5
 Chin. Phys. B Vol. 23, No. 9 (2014) 098106

C
A B
A
B C

(a) (b)
(a) (b)

0.06
original
0.06 original
corrected
corrected
0.04
Grayscale

0.04

Grayscale
0.02 0.02
(c)

0 (c) 0

0 10 20 30 40 0 10 20 30 40
Pixel sequence Pixel sequence
Fig. 7. (color online) Comparison of the 500th slice image before and af- Fig. 8. (color online) Comparison of the 572th slice image before and af-
ter scatter correction (grayscale window: [–0.012, 0.068]): (a) original ter scatter correction (grayscale window: [–0.012, 0.068]): (a) original
slice image, (b) scatter corrected slice image, (c) grayscale comparison slice image, (b) scatter corrected slice image, (c) grayscale comparison
on the same location. on the same location.

Table 1. Numerical comparison of slice image quality before and after the scatter correction.
SNR CNR AG
Slice number
value increased/% value increased/% value increased/%
original 13.08 / 6.095 / 0.001560 /
500th
corrected 9.597 –26.63 6.810 11.73 0.003546 127.31
original 21.89 / 6.337 / 0.001157 /
572th
corrected 18.66 –14.76 7.479 18.02 0.002168 87.38

the experiment environment and the correction method un-

changed. This can make the grayscale of the blank area in slit
imaging consistent with that in open imaging. The purpose is
to make the ray strength received by the FPD the same before
and after the scatter suppression. Figure 9 is the scatter correc-
tion comparison of the 500th and the 572th slice images before
(a) (b) and after the increasing exposure. No apparent difference can
be seen from the images, but we know that increasing the ex-
posure can reduce the noise. The numerical comparison of
slice image quality is shown in Table 2 (the evaluation regions
are the same as those in Figs. 7 and 8). It can be seen from
the table that SNR and CNR are both improved, and SNR is
more obviously improved. AG cannot completely distinguish
(c) (d) structures and noises, so there is a slight lowering of AG. By a
further analysis, we can see that although increasing exposure
Fig. 9. (color online) Scatter correction comparison of slice images with
different exposure levels (grayscale window: [–0.012, 0.068]): (a) the can compensate for the loss of the ray strength caused by the
500th slice with 0.2178 mA·s, (b) the 500th slice with 0.2343 mA·s, (c) scatter suppression, the method not only suppresses the outer
the 572th slice with 0.2178 mA·s, (d) the 572th slice with 0.2343 mA·s.
scatter but also corrects the inner scatter of the FPD, and de-
In order to test the cause analyzed above for the SNR re- creases the image grayscale in the form. The exposure should
duction in the scatter corrected slice image, we increase the be further increased if we want to achieve the SNR level as
scan exposure from 0.2178 mA·s to 0.2343 mA·s, keeping open imaging.

098106-6
 Chin. Phys. B Vol. 23, No. 9 (2014) 098106
Table 2. Numerical comparison of slice image quality with different exposure levels.
SNR CNR AG
Slice number
value increased/% value increased/% value increased/%
0.2178 mA·s 9.597 / 6.810 / 0.003546 /
500th
0.2343 mA·s 10.19 6.18 6.915 1.54 0.003517 –0.82
0.2178 mA·s 18.66 / 7.479 / 0.002168 /
572th
0.2343 mA·s 19.87 6.48 7.675 2.62 0.002138 –1.38

5. Conclusion [4] Ding G X, Duggan D M and Coffey C W 2007 Phys. Med. Biol. 52
1595
Based on the analysis of the existing methods, a scatter [5] Morin O, Gillis A, Chen J, Aubin M, Bucci M K, Roach M 3rd and
correction method for CBCT based on the interlacing-slit scan Pouliot J 2006 Med. Dosim. 31 51
[6] Kwan A L, Boone J M and Shah N 2005 Med. Phys. 32 2967
is proposed in this article. Firstly, according to the charac-
[7] Siewerdsen J H and Jaffrey D A 2001 Med. Phys. 28 220
teristics of CBCT imaging, a scatter suppression plate with [8] Boone J M 1986 Med. Phys. 13 319
interlacing slits is designed and fabricated. Then a scatter cor- [9] Neitzel U 1992 Med. Phys. 19 475
rection calculation method for CBCT based on image fusion is [10] Graham S A, Moseley D J, Siewerdsen J H and Jaffray D A 2007 Med.
Phys. 34 2691
proposed, which can splice out a complete set of scatter sup- [11] Coates A M, Dixon A M and Huckle J 2012 Radiography 18 250
pression projection images and simultaneously complete the [12] Li Q L, Yan B, Li L, Sun H S and Zhang F 2012 High Power Laser and
scatter correction within the FPD. Finally, the overall process Particle Beams 24 2235
[13] Naimuddin S, Hasegawa B H and Mistretta C A 1987 Med. Phys. 14
of scatter suppression and correction is provided. The experi- 330
mental results show that this method can significantly improve [14] Deng Z S, Ni D P, Jiang D Z, Zhao Z H, Xiao F and Xu L J 1997 J.
the clarity of the slice images and achieve a good scatter cor- Biomed. Eng. 14 137 (in Chinese)
[15] Jia P X, Zhang F, Yan B and Bao S L 2010 Chin. Phys. B 19 087802
rection. If the interlacing slits are dense enough, the scatter
[16] Mettivier G, Lanconelli N, Meo S L and Russo P 2012 IEEE Trans.
suppression result of CBCT can be close to that of the fan- Nucl. Sci. 59 2008
beam CT in theory. The method is also suitable for small-focus [17] Baer M and Kachelrie M 2012 Phys. Med. Biol. 57 6849
and micro-focus CT systems. [18] Ning R, Tang X Y and Conover D 2004 Med. Phys. 31 1195
[19] Schörner K, Goldammer M and Stephan J 2011 Nucl. Instr. Meth. Phys.
This method may cause SNR loss to a certain extent, Res. B 269 292
which can be compensated by increasing the exposure. In ad- [20] Zhang D H, Hu D C, Huang K D and Kong Y M 2009 Chin. Mech.
Eng. 20 639 (in Chinese)
dition, due to the projection cone angle of the slits in the scatter
[21] Niu T Y and Zhu L 2011 Med. Phys. 38 6027
suppression plate, this method can only be used when the dis- [22] Dong X, Jia X, Niu T Y and Zhu L 2012 Proc. SPIE Medical Imag-
tance between the ray source and the detector is constant or ing 2012: Physics of Medical Imaging, February 4–9, 2012 San Diego,
USA, p. 831319
has only small changes. For the problem that two circle scans
[23] Wang J, Mao W and Solberg T 2010 Med. Phys. 37 5792
of interlacing slits may lower the detection efficiency, we may [24] Ouyang L, Song K and Wang J 2013 Med. Phys. 40 071903
increase the scan speed and the exposure, which can further [25] Xie S P and Luo L M 2011 Acta Electron. Sin. 39 1708 (in Chinese)
improve this method. [26] Zhu L, Bennett N R and Fahrig R 2006 IEEE Trans. Med. Imaging 25
1573
[27] Gao H W, Fahrig R, Bennett N R, Sun M S, Star-Lack J and Zhu L
References 2010 Med. Phys. 37 934
[1] Luo Z Y, Yang X Q, Meng Y Z and Deng Y 2010 Acta Phys. Sin. 59 [28] Wils P, Létang J M and Bruandet J P 2012 J. Electron. Imaging 21
8237 (in Chinese) 021113
[2] Wang X C, Yan B, Liu H K, Li L, Wei X and Hu G E 2013 Acta Phys. [29] Zhang H, Huang K D, Shi Y K and Li M J 2012 CT Theory and Appli-
Sin. 62 098702 (in Chinese) cations 21 247 (in Chinese)
[3] Guo X L, Tang Z W, Zhang H and Hu G S 2010 J. Tsinghua Univ. (Sci. [30] Huang K D, Zhang D H, Li M J and Zhang H 2013 Acta Phys. Sin. 62
& Tech.) 50 450 210702 (in Chinese)

098106-7