ARTICLE IN PRESS

Nuclear Instruments and Methods in Physics Research A 583 (2007) 313–323

www.elsevier.com/locate/nima

Monte Carlo simulations of a high-resolution X-ray CT

system for industrial applications
A. Micelia,b,, R. Thierrya, A. Flischa, U. Sennhausera, F. Casalib, M. Simonc
a
Empa - Swiss Federal Laboratories for Materials Testing and Research, Electronics/Metrology Laboratory, Ueberlandstrasse 129,
Duebendorf 8600, Switzerland
b
Department of Physics, Università di Bologna, C. Berti Pichat 6/2, Bologna 40127, Italy
c
Hans Wälischmiller GmbH, SchieX stattweg16, 88677 Markdorf, Germany

Received 9 May 2007; received in revised form 28 August 2007; accepted 4 September 2007
Available online 12 September 2007

Abstract

An X-ray computed tomography (CT) model based on the GEANT4 Monte Carlo code was developed for simulation of a cone-beam
CT system for industrial applications. The full simulation of the X-ray tube, object, and area detector was considered. The model was
validated through comparison with experimental measurements of different test objects. There is good agreement between the simulated
and measured projections. To validate the model we reduced the beam aperture of the X-ray tube, using a source–collimator, to decrease
the scattered radiation from the CT system structure and from the walls of the X-ray shielding room. The degradation of the image
contrast using larger beam apertures is also shown. Thereafter, the CT model was used to calculate the spatial distribution and the
magnitude of the scattered radiation from different objects. It has been assessed that the scatter-to-primary ratio (SPR) is below 5% for
small aluminum objects (approx. 5 cm path length), and in the case of large aluminum objects (approx. 20 cm path length) it can reach up
to a factor of 3 in the region corresponding to the maximum path length. Therefore, the scatter from the object significantly affects
quantitative accuracy. The model was also used to evaluate the degradation of the image contrast due to the detector box.
r 2007 Elsevier B.V. All rights reserved.

PACS: 02.70.Uu; 81.70.Tx; 87.59.Fm; 81.70.q

Keywords: Monte Carlo simulation; GEANT4; X-ray beam; Cone beam computed tomography; Scatter; Area detector

1. Introduction industries is growing as equipment becomes increasingly

available.
X-ray computed tomography (CT) provides quantita- Nowadays there are different kinds of CT systems
tive, readily interpretable data and enables the inspection available on the market: fan-beam CT scanners using line
of structures that are not amenable to any other detectors and cone-beam CT scanners using area detectors.
nondestructive evaluation technique. As a result, CT has Fan-beam scanners generate stacks of slices which have to
become well established as an inspection, evaluation, be assembled to 3D volumetric data. The image quality
and analysis tool in industry [1]. Many of the applications obtained with such systems is high but the slice-by-slice
have been in the automotive industry, where mostly scanning process requires long acquisition time and there-
aluminum cast parts have to be inspected, but use in other fore is expensive [2]. Cone-beam CT scanners using area
detectors allow simultaneous 3D data acquisition with
Corresponding author. Empa - Swiss Federal Laboratories for
shorter scanning time. This advantage of increased volume
coverage is hampered by the growing contribution of
Materials Testing and Research, Electronics/Metrology Laboratory,
Ueberlandstrasse 129, Duebendorf 8600, Switzerland.
Compton scatter radiation to the measured signal that
Tel.: +41 44 823 4336; fax: +41 44 823 4579. leads to significant degradation of image quality. In fact,
E-mail address: [email protected] (A. Miceli). incident X-ray photons are scattered inside the object and,

0168-9002/$ - see front matter r 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.nima.2007.09.012
 ARTICLE IN PRESS
314 A. Miceli et al. / Nuclear Instruments and Methods in Physics Research A 583 (2007) 313–323

if reaching the detector, cause an increase in the registered In this work, a MC model for an X-ray CT system with
total intensity and lead to underestimation of the attenua- an area detector configuration for industrial applications
tion in each projection. The effect on CT images is a based on the GEANT4 code was developed. The simula-
decreasing of the low-contrast detectability and the tion was designed to describe the image formation in CT
presence of cupping and streak artifacts between image starting from the generation of the X-ray photons up to
regions of high attenuation [3–6]. their absorption in the sensitive detector. Detailed simula-
Knowledge of scatter distribution is therefore essential tion of the X-ray tube, filters, test objects, and detector
to optimize the design and acquisition parameters of were considered. The model was validated by comparison
cone-beam CT systems for industrial applications. More- with experimental measurements of test objects on a cone-
over, the information about the scatter distribution can be beam CT system. The CT model was used to calculate the
used for corrective image reconstruction algorithms [7,8]. magnitude and spatial distribution of the scattered radia-
Monte Carlo (MC) based scatter correction for image tion from different objects. Since in this work our aim was
reconstruction has already been performed in medical the study of the scattered radiation from the object, we
imaging [9]. adopted all the necessary measures to reduce the environ-
The amount of scattered radiation depends strongly on ment scatter. The first was the use of a source–collimator
the type of imaging system and detector geometry used as that reduces the aperture of the X-ray beam to the region
well as the object under study. Several authors have of interest, the second was the use of a lead shielding box
estimated the amount of scatter for medical scanners around the X-ray tube to reduce the leakage radiation of
[4,5,10] while studies for industrial CT systems are rather the tube. In this paper we show the profile of a test object
limited [11–13]. Studies on medical imaging X-ray systems acquired using several kinds of source–collimator to show
show that the scatter-to-primary ratio (SPR) increases with how the environment scatter, if not reduced, contributes to
the object size and the field of view [4,10,11]. the degradation of the contrast on the image. Moreover, we
The scatter contribution in medical CT has been evaluated, thanks to the MC simulation, the environment
simulated by several authors using analytical models scatter produced by the only detector box.
[14–16]. These models can be used with some success, but
result often in approximate estimates of the true scatter
distribution and therefore they are not suitable to simulate 2. Experimental setup and simulation procedure
complex CT geometries. A more general and accurate way
of estimating scatter contribution is to use MC simulation 2.1. CT system
[3,10,11,17,18]. Although the investigation of scatter
contribution using the MC method is time consuming The validity of the Geant4 based MC model was verified
compared to relatively simple mathematical modeling, the by comparing the simulated and measured distributions
widespread availability of high performance parallel from various test objects on a high resolution cone-beam
computing and Grid technology in addition to the X-ray CT scanner for industrial applications.
popularity of variance reduction techniques spurred the The CT system (Fig. 1) consists of an X-ray source, an
use of MC calculations especially when modeling complex X-ray source–collimator, a detector composed of a
geometries [19]. scintillator, a mirror and a CCD camera, and a four axis
The radiation scattered from the CT system structure manipulator. A filter of silver 1 mm thick (post-filter) is
and from the walls of the X-ray shielding room on the
detector (environment scatter) is an additional source of
noise and therefore artifact that affects the quality of the
CT image. The amount of this radiation depends on the
characteristics of the X-ray shielding room (dimensions
and materials of the walls), the geometry, materials and
working energy of the CT system, the aperture of the X-ray
beam, and the materials and geometry of the investigated
object. The environment scatter adds a uniform noise to
the image that decreases the image contrast and therefore
limits the low-contrast detectability [3]. The environment
scatter is less critical than the scatter from the object
because it does not change significantly with the projec-
tions. It can be assimilated to a constant signal and
subtracted to the data. Moreover, it is possible to reduce
the environment scatter placing the CT system in a bigger
Fig. 1. Picture of the CT system. The X-ray tube and its housing, the
shielding room, choosing absorbing materials for the walls object manipulator, the filter and the detector box are clearly visible.
of the X-ray room and optimizing the design of the The scintillator screen is placed behind the filter; the CCD camera is inside
structure housing the detector. the detector box.
 ARTICLE IN PRESS
A. Miceli et al. / Nuclear Instruments and Methods in Physics Research A 583 (2007) 313–323 315

placed in front of the scintillator at 1461 mm from the activated in the physics list for electrons were: ionization,
source. The scintillator, the mirror, and the CCD camera bremsstrahlung, and multiple scattering; for photons:
are placed inside a metallic box (detector box) of overall Rayleigh scattering, Compton scattering, and photoelectric
dimensions of 490  510  1000 mm3 (width  height  effect [23]. The cut value in range was set to 0.1 mm for
depth) with walls made of a 3 mm thick steel layer (inner photons and electrons.
layer), a 3 mm thick lead layer and a 5 mm thick steel layer The simulation is performed in two steps: the generation
(external layer). The source–detector distance is 1500 mm. of the X-ray spectrum of the tube taking into account the
A lead housing is placed around the X-ray tube to reduce anode angle, inherent and external filtration of the tube;
the leakage of the X-ray tube housing. The dimensions of and the generation of the projection of the object according
the X-ray shielding room are 3.6  2.9  3.7 m3 (width  to the acquisition setup (i.e. image of the energy deposited
height  depth). Two walls of the room are made of within the detector).
concrete and the other two of a sandwich of steel (3 mm), The simulations were run on a Pentium-IV-based
lead (25 mm), and steel (3 mm). personal computer with a 2.80 GHz microprocessor. The
computing time to reach a good statistic strongly depends
2.1.1. X-ray source on the X-ray beam aperture; it goes from 1 day in case of
The X-ray unit employed is a 450 kV X-ray generator small objects (approx. 5 cm path length, beam aperture of
manufactured by Comet AG (Model MXR 451) with 61) to several days in case of large objects (approx. 20 cm
2.3 mm iron and 1.0 mm copper inherent filtration. The path length, beam aperture of 181).
angle of the target, made of tungsten, is 301. The size
of the focal spot is 2.5 mm. The emission cone of the X-ray 2.2.1. Generation of the X-ray spectrum
source is 401. A 1.0 mm tungsten (alloy: HPM1750) The generation of the spectrum involves the simulation
attenuator is used to reduce the primary flux of low-energy of a monochromatic pencil electron beam hitting the tung-
photons. sten target at an angle of 301 with respect to the normal of
the anode surface and the passage of the produced X-ray
2.1.2. X-ray source–collimator spectrum through inherent filtration (2.3 mm Fe and
Three source–collimators were manufactured to investi- 1.0 mm Cu) and external filtration (1.0 mm W, alloy
gate the influence of beam aperture on image contrast: (i) a HPM1750). The radiation is retrieved within an angle of
100 mm thick rectangular lead source–collimator with 201 with respect to the central axis of the beam. Fig. 2
angular aperture 8.871  6.091 (horizontal  vertical), (ii) a shows the spectrum at 450 kV for the MXR-451 Comet
180 mm thick brass source–collimator with angular X-ray tube simulated using the X-ray tube characteristics
aperture 5.611  5.611 (big aperture), and (iii) a brass mentioned above. The spectrum was simulated with 2  109
source–collimator with angular aperture 3.151  3.151 primary electron histories. The validation of the simulated
(small aperture). The brass source–collimators can be spectra has been assessed through comparison with
inserted inside the lead source–collimator. Moreover the experimental data [24].
configuration without any source–collimator, which corre-
sponds to a beam aperture of half angle 201, is considered. 2.2.2. Image of the deposited energy distribution
In each studied configuration the test object was comple- The X-ray photons are emitted from the focal spot of
tely irradiated by the primary beam. diameter 2.5 mm, with energy sampled randomly from the
simulated spectrum, towards the object. Their direction is
2.1.3. Detector selected randomly from an isotropic distribution of angles
The X-ray converter is a Thallium-doped Cesium Iodine,
CsI(Tl), scintillator with a thickness of 2 mm manufa- 4
ctured by Hamamatsu. The effective scintillator area is
3.5
480  280 mm2. The 1 mm thick back plate is made of
aluminum. The converted photons are projected on a 451 3
Photons per keV

mirror and reflected on a CCD camera (Apogee Alta U32) 2.5

of 2184  1472 pixels. The pixel size is 6.8  6.8 mm2. 2
A NIKON 28 mm lens is mounted on the CCD camera. 1.5
The field of view is 524  353 mm2.
1

2.2. Description of the GEANT4 based Monte Carlo CT 0.5

simulation 0
0 50 100 150 200 250 300 350 400 450
The low-energy extension of the electromagnetic pro- Energy (keV)
cesses version 2.3 of the simulation toolkit GEANT4 Fig. 2. Simulated spectrum of the X-ray tube MRX-451. The electron
[20–22] was used to model the interactions of photons and energy was 450 keV, the inherent filtration was 2.3 mm Fe+1.0 mm Cu
electrons with matter down to 250 eV. The processes and the external filtration was 1 mm W (alloy HPM1750).
 ARTICLE IN PRESS
316 A. Miceli et al. / Nuclear Instruments and Methods in Physics Research A 583 (2007) 313–323

Fig. 3. Picture of the test objects for the validation of the model: (a) Al box, (b) small Al cylinder, and (c) Al step wedge.

in a cone of selected aperture. When the X-ray photons where u and v are the detector coordinates, Inorm(u,v) is the
reach the object they can undergo photoelectric effect, normalized image, Iair(u,v) is the image in absence of the
single Compton scattering, multiple Compton scattering, object and Idark(u,v) is the image without X-rays. A 4  4
and Rayleigh scattering. The X-ray photons that leave the binning was performed on the acquired images.
object in direction of the detector are filtered by the post- Influence of the X-ray beam aperture. To evaluate the
filter and the back plate of the scintillator. The photons influence of the X-ray beam aperture on the image contrast
interact with the scintillator by releasing energy to the we manufactured an aluminum box of size 50  65 
material with production of electrons. When simulating 50 mm3 with two holes of size 10  10  50 and 8  10 
the detector box, the photons that are not absorbed by the 50 mm3 along the axial direction with equal distance from
scintillator can interact with the detector box and back- the object center (Fig. 3a). Radiographies with both the
scatter into the scintillator. The energy and the position of brass source–collimators and the rectangular lead source–
production of the electrons within the scintillator together collimator were acquired. In addition, an acquisition
with the number of Compton and Rayleigh interactions of without source–collimator was performed.
the parent photon within the object were retrieved. The CsI Validation of the model. Four test objects were used to
was simulated as a bulk material with density 4.52 g/cm3. validate the model: (i) the aluminum box, (ii) the aluminum
The pixel size was 0.96  0.96 mm2. box containing two copper rods, (iii) a cylinder made of
The scatter images were de-noised using the Richard- aluminum of external diameter 66 mm and inner diameter
son–Lucy fit. The procedure, which has been used by Colijn 25 mm (small cylinder) (Fig. 3b), and (iv) a step wedge
[12], utilizes a maximum likelihood algorithm to retrieve the made of aluminum of overall dimensions 100  100 
original noise-free signal blurred by a Gaussian kernel. 20 mm3 having five steps of thickness [20–100] mm
Smooth estimates of scatter projections can be obtained (Fig. 3c). To quantitative evaluate the MC simulation we
from simulation with a low number of photons. This allows calculated the root mean square (RMS) of the difference
reducing the time needed for MC simulations. In the present between the experimental and simulated projections given
study, MC simulations were performed with 109 primary by the formula
photon histories in case of small objects and 2  109 in case qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi
 2
of large objects. The projections were de-noised using 10 RMSA ¼ h I Exp ðu; vÞ  I Sim ðu; vÞ iðu;vÞ2A (2)
iterations of the Richardson–Lucy fit. The standard devia-
tion of the Gaussian kernel was set to 30 detector pixels. and we compared the result to the uncertainty (s)
defined by
2.3. Evaluation and validation qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 2 ffi
sA ¼ h I Exp ðu; vÞ  hI Exp ðu; vÞiðu;vÞ2A iðu;vÞ2A þ
Measurements. The X-ray high voltage was set to 450 kV qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 2
and the current to 2 mA. The distance from the rotation h I Sim ðu; vÞ  hI Sim ðu; vÞiðu;vÞ2A iðu;vÞ2A ð3Þ
axis to the detector plane was 226 mm. All the acquired
images, Iacq(u,v), were normalized to remove some geo- where u and v are the detector coordinates, IExp(u,v) is the
metric effects and scanner non-uniformity, such as the experimental normalized image, ISim is the simulated
spatial irregularity of the source radiation and the non- image, and A is the region of interest (ROI).
uniformity of the detector response, using the formula Scatter from large objects. To evaluate the scatter from
large objects we simulated a 200 mm high aluminum
I acq ðu; vÞ  I dark ðu; vÞ cylinder of external diameter 200 mm and inner diameter
I norm ðu; vÞ ¼ (1)
I air ðu; vÞ  I dark ðu; vÞ 50 mm. In this case the source–object distance was
 ARTICLE IN PRESS
A. Miceli et al. / Nuclear Instruments and Methods in Physics Research A 583 (2007) 313–323 317

Fig. 4. Simulation setup. The previously simulated X-ray spectrum is the input to the simulation of the CT system. The object is the Al step cylinder.
The rectangular lead source–collimator and the detector box were simulated.

1000 mm and the pixel size was 1.92  1.92 mm2. The post- is the value of primary radiation corresponding to the
filter was not simulated. detector coordinates (u,v). Moreover, the spatial distribu-
Degradation of the image contrast due to the detector box. tion of the scatter was studied.
To evaluate the influence of the detector box on the image
contrast, we carried out a simulation in the geometric 3. Results and discussion
configuration of Fig. 4. The detector box was modeled using
the description given above. Also the rectangular source– In this section, the degradation of the image contrast due
collimator of lead was taken into account. The object was an to the environment scatter and the validation of the Monte
aluminum step cylinder with inner diameter 20 mm, 8 external Carlo model are presented. Eventually the validated model
diameters (40, 60, 80, 100, 120, 160, 200, and 220 mm) and is used to evaluate the scatter from the test objects and the
height of 160 mm. The full scintillator screen was irradiated degradation of the image contrast due to the detector box.
by the primary beam. The source–object distance was set to
1000 mm. Thereafter, we performed a simulation with the 3.1. Degradation of the image contrast due to the
same configuration but different composition of the walls of environment scatter
the detector box (3 mm thick lead layer and 5 mm thick steel
layer) to evaluate the influence of the inner layer of the As already mention in the introduction, the scattered
detector box on the image contrast. radiation reaching the detector is composed of the
Evaluation. The degradation of the image quality was scattered radiation from the object and the scattered
evaluated by calculating the image contrast given by radiation from the environment. The environment scatter
depends on the size and wall materials of the X-ray
hIðu; vÞiðu;vÞ2flat  hIðu; vÞiðu;vÞ2obj
C¼ (4) shielding room, on the materials of the structure that
hIðu; vÞiðu;vÞ2obj houses the detector, on the X-ray spectrum of the source
where I(u,v) is the value of the gray level corresponding to and, less strongly, on the object investigated. Since we were
the detector coordinates (u,v), flat is a ROI in the image interested in studying the scatter produced by the object we
where the X-ray flux is not attenuated by the object and obj adopted measures to reduce the environment scatter. We
is a ROI in the image where the X-ray flux has been used a source collimator to reduce the aperture of the
attenuated by the object. X-ray beam and a shielding box around the X-ray tube to
The corruption of projection data by scattered photons reduce the leakage of the tube. The profile along a
was investigated by calculating the SPR for each order of horizontal line of the test object, scanned using several
scatter, SPRn(u,v), defined by source–collimators (Fig. 5a), shows how the environment
scatter contributes to the degradation of the contrast on
Sn ðu; vÞ the image. The source–collimators used were: (i) the brass
SPRn ðu; vÞ ¼ (5)
Pðu; vÞ source–collimator of small aperture, (ii) the brass source–
and the total scatter-to-primary ratio, SPR(u,v): collimator of big aperture, and (iii) the rectangular lead
P source–collimator. Moreover an acquisition without any
S n ðu; vÞ X source–collimator was performed. The Fig. 5(b) shows
n
SPRðu; vÞ ¼ ¼ SPRn ðu; vÞ (6) that the value of the contrast decreases considerably with
Pðu; vÞ n
the aperture of the X-ray beam: it goes from 2.60 when
where Sn(u,v) is the value of the n-order scatter radiation the brass source–collimator of small aperture is used
corresponding to the detector coordinates (u,v) and P(u,v) (3.151  3.151) to 1.60 when no source–collimator is used
 ARTICLE IN PRESS
318 A. Miceli et al. / Nuclear Instruments and Methods in Physics Research A 583 (2007) 313–323

1 2.9

2.7 2.60

2.5

Contrast
2.3
Inorm

2.13
2.1

Without 1.9 1.82

Rectangular Pb
Brass Big aperture 1.7 1.64
Brass Small Aperture
0.1 1.5
-60 -50 -40 -30 -20 -10 0 10 20 30 40 50 60 Brass Small Aperture Brass Big Aperture Rectangular Pb Without

Position x (mm) Pre-collimator

Fig. 5. Influence of the environment scatter. (a) Horizontal profile of the Al box in correspondence to the hole of front size 8  10 mm2. (b) Values of the
contrast calculated from the profiles.

-60 1 -60 1 100

0.9 0.9
-40 -40
0.8 10-0.2
’ 0.8
Position y (mm)

A A
Position y (mm)

-20 -20

I norm
0.7 0.7
0 0
0.6 0.6 10-0.4
20 0.5 20
0.5
40 0.4 40 0.4 10-0.6 Simulated
0.3 Measured
60 60 0.3 a c b
-50 0 50 -50 0 50 -60 -40 -20 0 20 40 60
Position x (mm) Position x (mm) Position x (mm)

Fig. 6. Comparison between simulated and experimental radiographies for the Al box. (a) Simulated image. (b) Measured image. (c) Profile of the Al box,
corresponding to the ROI AA’.

0
1 10
-60 1 -60
0.9 0.9
-40 -40
0.8 0.8
Position y (mm)

Position y (mm)

A A’ -20 0.7
-20 0.7
I norm

0.6 0.6
0 0
0.5 0.5
20 0.4 20 0.4
40 0.3 40 0.3 -1
10
Simulated
0.2 0.2
60 Measured
60 d f e
0.1 0.1
-50 0 50 -50 0 50 -60 -40 -20 0 20 40 60
Position x (mm) Position x (mm) Position x (mm)

Fig. 7. Comparison between simulated and experimental radiographies for the Al box containing copper rods. (a) Simulated image. (b) Measured image.
(c) Profile of the Al box containing copper rods, corresponding to the ROI AA’.

(half angle: 201). The small dimensions of the X-ray installed in a shielding room the environment scatter can be
shielding room, the material of the walls (lined with steel evaluate acquiring a radiography of an object easy to
3 mm thick), the high energy of the X-ray beam, and the simulate, similar to the one we want to analyze and
CT system structure are the cause of this considerable performing the simulation with the same parameters of the
amount of environment scatter. The environment scatter experiment. The difference between the results represents
can be reduced placing the CT system in a bigger shielding the environment scatter.
room, optimizing the composition of the walls to reduce In paragraph 3.4 we simulated the environment scatter
the scattering and optimizing the structure that houses the due only to the detector box to study the influence of the
detector. When the system is not modifiable and already detector box on the degradation of the contrast.
 ARTICLE IN PRESS
A. Miceli et al. / Nuclear Instruments and Methods in Physics Research A 583 (2007) 313–323 319

3.2. Validation of the simulation Cu rods (Table 1, ROIs d–f) shows that the simulated and
measured data are in agreement. Fig. 8 shows the simu-
Figs. 6–9 show the comparison of simulated and lated (a) and measured (b) image of the Al cylinder, and
measured profiles, after normalization, for various test the comparison of simulated and measured profiles of the
objects. The profiles were calculated from the ROI AA’. central ROI. The values of the RMS selected from the
There is generally good agreement between the simulations profile, corresponding to the maximum path length and to
and the experimental results. The values of the RMS of the the center of the object, are smaller than the uncertainties
difference between the simulated and experimental data
calculated for the ROIs defined in Figs. 6–9 were compared
to the combined uncertainties (Table 1). Fig. 6 shows the Table 1
simulated (a) and measured (b) image of the aluminum box Root mean square of difference between experimental and measured
projections (RMS) and uncertainty for the selected ROIs
together with the comparison of measured and simulated
profiles presented in log–linear scale to magnify the ROI x1 (mm) x2 (mm) RMS (102) Uncertainty (102)
differences between simulated and measured results (c).
a 25 10 0.74 1.0
Although the profile calculated from the simulated image is
b 10 25 0.59 0.6
noisy it appears clearly that the simulated and measured c 3 3 1.71 3
profiles are in agreement. The values of the RMS d 25 10 0.71 1.0
calculated from the profile of the aluminum box in e 9 24 0.64 1.0
correspondence to the maximum path length of the object f 3 2 0.29 2.0
g 3 3 0.65 0.8
and the void hole (Table 1, ROIs a–c) are smaller than the
h 22 16 0.48 2.0
combined uncertainties. Fig. 7 shows the normalized i 16 22 0.68 2.0
simulated (a) and measured (b) image of the Al box j 52 37 0.80 1.0
containing the Cu rods, and the comparison of simulated k 30 15 0.74 0.9
and measured profiles. The comparison between the RMS l 7 8 0.47 0.6
m 17 34 0.46 0.6
and the uncertainties corresponding to the region of
n 39 48 0.57 0.6
maximum path length of the object and to the region with

0
-60 1 1 10
-60
0.9 0.9
-40 -40
0.8 0.8 10-0.2
Position y (mm)
Position y (mm)

-20 -20
A 0.7 0.7
A’
I norm

0 0.6
g
0 0.6
-0.4
10
20 0.5 20 0.5
h i
0.4 0.4
40 40
0.3 0.3 10-0.6 Simulated
60 60 0.2 Measured
0.2
-50 0 50 -50 0 50 -60 -40 -20 0 20 40 60
Position x (mm) Position x (mm) Position x (mm)

Fig. 8. Comparison between simulated and experimental radiographies for the Al cylinder. (a) Simulated image. (b) Measured image. (c) Profile of the Al
cylinder, corresponding to the ROI AA’.

-60 1 1
-60
0.9 0.9 10-0.1 j
-40 -40
0.8 0.8
10-0.3 k
Position y (mm)

Position y (mm)

-20 A A’ 0.7 -20 0.7

I norm

0.6 0.6
0 0 l
0.5 10-0.5
0.5
20 20
0.4 0.4
10-0.7 m n
40 0.3 40 0.3
Simulated
0.2 0.2 Measured
60 60 10-0.9
0.1 0.1
-50 0 50 -50 0 50 -60 -40 -20 0 20 40 60
Position x (mm) Position x (mm) Position x (mm)

Fig. 9. Comparison between simulated and experimental radiographies for the Al step wedge. (a) Simulated image. (b) Measured image. (c) Profile of the
Al step wedge corresponding to the ROI AA’.
 ARTICLE IN PRESS
320 A. Miceli et al. / Nuclear Instruments and Methods in Physics Research A 583 (2007) 313–323

(Table 1, ROIs g–i). Fig. 9 shows the normalized simulated orders scattering is shown in Fig. 10a for the aluminum
(a) and measured (b) image of the Al step wedge together box. The profile of the first-order scattering follows the
with the comparison of simulated and measured profiles of shape of the test object. The profiles of the second
the central ROI. Although the simulated profile appears to and higher-order scattering are approximately uniform.
be slightly higher than the measured profile in correspon- Fig. 10b–10e show the SPR1, SPR2, and SPR42 for the
dence to the step 100 mm thick, the Table 1, ROI n, shows aluminum box, respectively, without and with copper rods,
that the intensities are in agreement within the uncertainty the small cylinder, and the step wedge. As expected, the
limits. This is due to the fact that the flux is highly major contribution to the scattering is given by the first-
attenuated and the signal-to-noise ratio (SNR) is low. For order scattering. The SPR1, SPR2 and SPR42 are peaked
an accurate convergence of the result, it is necessary to in the region corresponding to the maximum path length
simulate a larger number of photon histories. because of the smaller probability of transmitted primary
photons. In the area corresponding to the maximum path
3.3. Magnitude and spatial distribution of scatter radiation length the value of the SPR1 for aluminum box is below
from the object 4%, SPR2 and SPR42 are both below 1% (Fig. 10b).
The SPR1 for the aluminum box containing copper rods is
Evaluation of scatter from small objects. The profile of 12% in the area corresponding to the copper rod and
first-order scattering, second-order scattering, and higher- below 4% in the area corresponding to the aluminum, the

10-2 0.04
order 1
0.035 order 2
order > 2
0.03
0.025
Inorm

SPR

10-3 0.02
0.015
order 1
0.01
order 2 0.005
order > 2
10-4 0
-60 -40 -20 0 20 40 60 -60 -40 -20 0 20 40 60
Position x (mm) Position x (mm)

0.12 0.07
order 1 order 1
order 2 0.06 order 2
0.1 order > 2 order > 2
0.05
0.08
0.04
SPR
SPR

0.06
0.03
0.04
0.02
0.02 0.01
0 0
-60 -40 -20 0 20 40 60 -60 -40 -20 0 20 40 60
Position x (mm) Position x (mm)

0.04
order 1
0.035 order 2
order > 2
0.03
0.025
SPR

0.02
0.015
0.01
0.005
0
-60 -40 -20 0 20 40 60
Position x (mm)

Fig. 10. (a) Simulated profiles of the first, second, and higher than second order scattered photons of the Al box. Scatter-to-primary ratio (SPR) for the
projection of (b) the Al box, (c) the Al box containing copper rods, (d) the Al cylinder, and (e) the Al step wedge.
 ARTICLE IN PRESS
A. Miceli et al. / Nuclear Instruments and Methods in Physics Research A 583 (2007) 313–323 321

100 1.5
order 1
order 2
order >2

10-1 1
Inorm

SPR

Total
Primary
10-2 Order > 2
0.5
Order 1
Order 2
10-3 0
-150 -100 -50 0 50 100 150 -150 -100 -50 0 50 100 150
Position x (mm) Position x (mm)

Fig. 11. Al cylinder of external diameter 200 mm and inner diameter 50 mm. Simulated images of the energy deposited by (a) total radiation, (b) primary
radiation, (c) first-order scattering, (d) second-order scattering (e), and higher-order scattering. The images of the scatter were de-noised using the
Richardson–Lucy fit. (f) Profiles of the Al cylinder, corresponding to the ROI AA’. (g) Scatter-to-primary ratio for each order of scatter calculated from
the profiles.

SPR2 is 3% in correspondence to the copper rod and below Table 2

1% to the aluminum, while SPR42 is below 1% in Contrast calculated from the profiles shown in Fig. 11f
correspondence to the copper rod and to the aluminum Orders of scattering included Ccenter CMax. path length
(Fig. 10c). The SPR1 for the small cylinder is 6.5% in the
area where the path lengths are maximum and 4.5% in 0 9378 250730
the central area, the SPR2 is, respectively, 1.2% and 1%, 1st 7275 140710
1st + 2nd 6374 11378
and SPR42 is below 1% (Fig. 10d). The SPR1 for the
All 4872 7375
step wedge is 3.5% in the area corresponding to the
maximum path length, the SPR2 and SPR42 are below 1%
(Fig. 10e).
Evaluation of scatter from large objects. CT of large interact within the object before being detected), first-order
aluminum objects, such as the aluminum cylinder of scattered radiation, second-order scattered radiation, and
maximum path length 150 mm, is a challenge because of higher-order scattered radiation and the corresponding
the low SNR. The probability for a photon to be profiles are displayed in Fig. 11. Due to the large object
transmitted without interaction through the cylinder is size, the multiple scattering plays a major role and in the
1.1% (the equivalent energy of the X-ray spectrum is region corresponding to the maximum path length its
250 keV). Moreover the scattered radiation, having low contribution to the image is higher than the signal. SPR1 is
energy, more likely interacts with the scintillator than the 30% in the center and 80% in the region corresponding to
primary photons. The images of the energy deposited by the maximum path length, SPR2 is 20% in the center and
total radiation, primary radiation (i.e. photons that do not 55% in the region corresponding to the maximum path
 ARTICLE IN PRESS
322 A. Miceli et al. / Nuclear Instruments and Methods in Physics Research A 583 (2007) 313–323

-150 A
B B’ 1
-100

Position y (mm)
0.8
-50
0 0.6

50 0.4
100 0.2
150 A’
-200 -100 0 100 200
Position x (mm)

1 Without detector box

1
Detector box Without detector box
Detector box
Detector box without inner
Detector box without inner layer

Inorm
Inorm

0.1 0.1

0.01 0.01
-140 -100 -60 -20 20 60 100 140 -180 -140 -100 -60 -20 20 60 100 140 180
Position y (mm) Position x (mm)

Fig. 12. (a) Image of the simulated Al cylinder step with the detector box. (b) Profile along the line AA’. (c) Profile along the line BB’.

length and SPR42 is 50% in the center and 140% in the Table 3
region corresponding to the maximum path length. The Contrast calculated from the profiles in Fig. 12b
degradation of the contrast due to the n-order of scattering Diameter CDetector CDetector CWithout
box box detector
in the center of the object and in the region corresponding ring (mm) without inner layer box
to the maximum path length calculated from the profile in
Fig. 11f is shown in Table 2. 40 0.8970.04 0.9070.05 0.9270.04
60 2.4670.09 2.570.1 2.5470.09
80 5.070.2 5.370.2 5.370.2
3.4. Degradation of the image contrast due to the detector 100 9.370.4 9.970.4 9.970.4
box 120 15.570.9 17.270.9 17.670.9
160 3472 4173 4272
The image of the step cylinder obtained when the 220 5373 8576 7875
detector box is simulated is displayed in Fig. 12(a). The
Figs. 12(b) and (c) show the comparison of the profile
extracted from the image in Fig. 12(a) with the profiles 4. Conclusions
obtained when the detector box is not simulated. The
profile obtained in the same condition of Fig. 4 but A Monte Carlo model of a cone beam X-ray CT system
considering the walls of the box composed of two layers for industrial applications has been developed and
(3 mm lead, 5 mm steel) is also shown. The value of the experimentally validated using CT scans of several test
attenuated intensity in the region corresponding to the objects. The simulated projections were in good agreement
maximum path length is 0.02% with the detector box, with the measured data. A source–collimator was used to
0.01% without the detector box and 0.01% with the reduce the environment scatter. The degradation of the
detector box without the inner layer. We conclude that contrast due to the environment scatter was found to be
the inner layer of steel is the major responsible of the 61.5% when no source–collimator is used. The model was
production of the scatter. Table 3 shows the contrast used to calculate the images of the first-order scattering,
calculated from the central vertical profile (Fig. 12b) for second-order scattering, and higher-order scattering for
each step. The degradation of the contrast due to the different objects. The total contribution of the scattered
detector box with walls composed of three layers is 68% in radiation to the signal in case of small aluminum objects is
correspondence to the ring of diameter 220 mm. Due to the lower than 5%. In case of objects of large dimensions, such
high attenuation, it was not possible to distinguish between as the aluminum cylinder of diameter 200 mm, the total
the ring of diameter 200 and 220 mm. contribution of the scattered radiation to the signal from
 ARTICLE IN PRESS
A. Miceli et al. / Nuclear Instruments and Methods in Physics Research A 583 (2007) 313–323 323

the object itself is 280% and leads to degradation of the [3] M. Endo, T. Tsunoo, N. Nakamori, K. Yoshida, Med. Phys. 28
image contrast of 30%. Therefore, the evaluation of the (2001) 469.
[4] P.C. Johns, M. Yaffe, Med. Phys. 9 (1982) 231.
multiple scattering for large objects is of primary im-
[5] P.M. Joseph, R.D. Spital, Med. Phys. 9 (1982) 464.
portance for the optimization of the CT system and for the [6] G.H. Glover, Med. Phys. 9 (1982) 860.
scattering correction. Moreover, the developed CT model [7] H. Kanamori, N. Nakamori, K. Inoue, E. Takenaka, Phys. Med.
enabled the evaluation of the degradation of the contrast Biol. 30 (1985) 239.
due to the photons backscattered from the detector box [8] N. Freud, P. Duvauchelle, S.A. Pistrui-Maximean, J.M. Letang,
onto the detector. The inner layer of 3 mm steel was found D. Babot, Nucl. Instr. and Meth. B 222 (2004) 285.
[9] N. Freud, J.M. Letang, D. Babot, Nucl. Instr. and Meth. B 227
to be the major source of scatter from the detector box. (2005) 551.
The developed CT model is therefore a versatile tool to [10] F. Inanc, J. Nondestruct. Eval. 18 (1999) 73.
evaluate the contribution of scattered radiation to the [11] W. Kalender, Phys. Med. Biol. 26 (1981) 835.
image and to optimize the CT system design. [12] A.P. Colijn, F.J. Beekman, IEEE Trans. Med. Imaging 23 (2004)
584.
[13] B. Ohnesorge, T. Flohr, K. Klingenbeck-Regn, Eur. Radiol. 9 (1999)
Acknowledgments 563.
[14] Y. Kyriakou, T. Riedel, W.A. Kalender, Phys. Med. Biol. 51 (2006)
This research was supported by the Swiss State 4567.
Secretariat for Education and Research and the European [15] J.M. Boone, J.A. Seibert, Med. Phys. 15 (1988) 721.
[16] M. Honda, K. Kikuchi, K.I. Komatsu, Med. Phys. 18 (1991)
Commission within the sixth framework project DETECT 219.
(New product design and engineering technologies based [17] H.P. Chan, K. Doi, Phys. Med. Biol. 28 (1983) 109.
on next generation computed tomography). The authors [18] J.M. Boone, V.N. Cooper, W.R. Nemzek, J.P. McGahan,
would like to thank J. Hofmann for fruitful discussions. J.A. Seibert, Med. Phys. 27 (2000) 2393.
[19] M.R. Ay, H. Zaidi, Phys. Med. Biol. 50 (2005) 4863.
[20] J. Apostolakis, S. Giani, M. Maire, P. Nieminen, M.G. Pia,
References L. Urban, INFN/AE 99/18 (1999).
[21] S. Agostinelli, et al., Nucl. Instr. and Meth. A 506 (2003) 250.
[1] D.C. Copley, J.W. Eberhard, A.G. Mohr, JOM 46 (1994) 14. [22] J. Allison, et al., IEEE Trans. Nucl. Sci. NS-53 (2006) 270.
[2] A. Flisch, A. Obrist, J. Hofmann, in: DGZfP-Proceedings BB 84-CD, [23] Physics Reference Manual, Geant4 Users’ Documents (2006).
Paper 17 (2003). [24] A. Miceli, et al., Nucl. Instr. and Meth. A (2007).