Content uploaded by Zdenko Zápražný
Author content
All content in this area was uploaded by Zdenko Zápražný
Content may be subject to copyright.
2012 JINST 7 C03005
PUBLISHED BY IOP PUBLISHING FOR SI SS A MEDIALAB
RECEIVED:September 13, 2011
REV ISED:December 21, 2011
ACCEPTED:January 31, 2012
PUBLISHED:March 2, 2012
13th INTERNATIONAL WORKSHOP ON RADIATION IMA GI NG DETECTORS,
3–7 JU LY 2011,
ETH ZURICH, SWITZERLAND
Phase contrast imaging of lightweight objects using
microfocus X-ray source and high resolution CCD
camera
Z. Zaprazny,a,1D. Korytar,aV. Ac,bP. Konopkaa,dand J. Bieleckic
aInstitute of Electrical Engineering, Slovak Academy of Sciences
Dubravska cesta 9, 841 04 Bratislava, Slovak Republic
bAlexander Dubcek University of Trencin
Studentska 2, 911 50 Trencin, Slovak Republic
cInstitute of Nuclear Physics, Polish Academy of Sciences
Radzikowskiego 152, 31-342 Krakow, Poland
dInstitute of Materials Science, Slovak University of Technology,
J. Bottu 25, 917 24 Trnava, Slovak Republic
E-mail: zdenko.zaprazny@savba.com
ABS TR ACT: Modern laboratory X-ray imaging systems with microfocus source and CCD camera
give us the possibility to move some of modern imaging techniques from synchrotrons to labora-
tories. Spatially coherent X-rays emitted from microfocus source traverse a sample with a phase
shift depending on real part of refractive index. Beam deflection induced by the local change of
refractive index may be expressed as dark-bright contrast on the edges in final projection. These
phenomena lead to increase of spatial resolution of X-ray projections but may also lead to unpleas-
ant artifacts in Computerized Tomography (CT) unless reconstruction program can separate phase
and absorption contributions. In this paper, several results of phase contrast imaging are presented.
KEYWORDS: Computerized Tomography (CT) and Computed Radiography (CR); X-ray radiogra-
phy and digital radiography (DR)
1Corresponding author.
c
2012 IOP Publishing Ltd and Sissa Medialab srl doi:10.1088/1748-0221/7/03/C03005
2012 JINST 7 C03005
Contents
1 Introduction 1
1.1 Motivation 1
2 Experimental setup 3
3 Results 3
3.1 Absorption mode and artifacts 3
3.2 Phase contrast mode 5
4 Discussion and conclusion 6
1 Introduction
In non-destructive studies of materials there is an effort to image the internal structure of an object
by using not only conventional 2D X-ray radiography but also using high resolution 3D tomog-
raphy which is based on reconstruction of multiple 2D projections-radiographs at various angular
positions of the object. This set of radiographs, called projection data, is processed with a recon-
struction algorithm to generate the tomogram of the specimen. Tomogram is a three dimensional
representation of the specimen structure. The smallest three dimensional unit in the tomogram is
called voxel. The projection data are collected within cone beam geometry through a full 360o
rotation of the sample. The optimal number of projections is given by NØ= (π/2)NW, where NW
is the number of pixels in the horizontal line of the detector. This number ensures that there is
adequate angular sampling to accurately reconstruct the specimen [1].
1.1 Motivation
New approaches that can detect X-ray phase shifts have been taken into account by developing
a variable X-ray optical system to study modern phase-sensitive X-ray imaging techniques. In
some cases X-ray imaging methods based on X-ray absorption fail because absorption contrast is
undetectable. Particularly if object of interest is composed of materials with low atomic number
and linear absorption coefficients are very similar. The behaviour of X-rays penetrating through
a sample can be described using a complex index of refraction [2]. In the X-ray region, the index of
refraction ndeviates only slightly below unity. It can be written as n=1−δ−iβ, where δis linked
with phase shift and βrelated to linear absorption coefficient. The phase-shift term δ(10−6–10−8)
is approximately 1000 times greater than β(10−9–10−11) for X-ray energies of 10–100 keV (e.g.
for mammography X-ray energy range of 15–25 keV) [3] for low atomic number samples.
Based on this knowledge, it is possible to recognize several imaging modes. If we have spa-
tially coherent X-rays, in the “in-line” imaging space behind the sample we can recognize several
–1–
2012 JINST 7 C03005
Figure 1. Representation of different regions depending on the distance Zof the imaging plane (X-ray
detector) from the object. wis the largest dimension of the observed structure of the object perpendicular to
the direction of propagation. NFis the dimensionless Fresnel number and λis the wavelength.
a) b) c)
Figure 2. a) Projection of a fly with predominance of absorption contrast. Z=8.8 mm, SDD=508.8 mm,
SOD=500 mm, MF=1.02. b) Phase contrast is starting to be visible with increasing the distance Z. Z=208.8
mm, SDD=370.8 mm, SOD=162 mm, MF=2.29. c) Phase contrast is observable, brightness on the edges
(minimum of intensity profile) due to the deviation of the beam in place with the change of the refractive
index. Z=258.8 mm, SDD=290.8 mm, SOD=32 mm, MF=9.09.
regions characterized by the dimensionless Fresnel number NF=w2/Zλ, where wis the dimen-
sion of the observed object perpendicular to the direction of propagation, Zis the distance between
sample and imaging plane and λis the wavelength [4]. By positioning the imaging plane behind
the object and changing distance Z we go through individual areas with different Fresnel numbers
as shown in figure 1. Figure 2shows radiographic projections obtained in three different regions
in imaging space [5]. With the increase of the distance Z the visibility of fine structures increases.
For example, a 40 micrometers sized structure of the fly in figure 2b) at the distance Z=208.8 mm
is for the X-ray energy 17 keV in the Fresnel region (NF=1.05) and for 5 keV in the Fraunhofer
region (NF=0.31). All these projections were taken using accelerating voltage of the X-ray tube
80 kV, current 100 µA and exposure time 20 s, without filtering of X-ray spectrum. We used cone
beam projections, where increasing of distance Z causes the geometrical magnification of an object
with the magnification factor MF=SDD/SOD, where SDD is source-detector distance and SOD
source-object distance.
–2–
2012 JINST 7 C03005
Motivation for this work comes out from the above examples illustrating the increase of visi-
bility of finer structures of lightweight samples due to phase contrast effects. Special emphasis is
given to phase contrast computerized tomography because phase effects, including edge enhance-
ment are not common in conventional CT systems.
2 Experimental setup
Key requirements for higher resolution X-ray computerized tomography are: high resolution X-
ray detector, high resolution sample holder and a small spot size of the X-ray source. A Newport
high resolution rotation stage with minimal incremental motion of 0.0002ofor the sample holder
will allow us to take precise sequential CT projections. We used a high resolution CCD X-ray
camera from Photonic Science as a detector. Basic parameters of the camera are: 1392 ×1040
[pixels], input pixel size: 6.4 ×6.4 [µm], active area: 10x8 [mm], scintillator: Gd2O2S:Tb3(5
mg/cm2) with optimal energy response of 5 keV to 17 keV. The focus size of the X-ray source with
a transmission tungsten anode (80kV, 100 µA) is declared as of 8 µm and the source emits 39o
conical beam. All system components are installed in an enclosed radiation leak protected 2.6 m
long optical bench with shock absorbers between legs and floor ground.
An important prerequisite for high resolution CT is high spatial resolution of the 2D projec-
tions which is possible by using an optical magnification of the object. With the present set-up it
is possible to achieve the spatial resolution of the whole X-ray imaging system down to 3 µm [5].
There is a possibility to express the so-called CT voxel size as the ratio of CCD pixel size and mag-
nification factor of the imaging setup. Our imaging system with microfocus X-ray source allows
us to obtain the magnification factor of 140. For example, imaging distance required to obtain a
magnification of 40 times is 480 mm because of shorter focal length (12 mm) compared to con-
ventional type of X-ray source. Moving CCD camera on linear bearings to obtain magnification
between 1.02 to over 140 allows us to regulate the voxel size between 5.8 µm and 0.05 µm. This
value does not match the spatial resolution because X-rays emitted from the finite source show a
blur aberration, which destroys volume resolution significantly above the minimal voxel size of
tomograms. At magnification factor 140 the lateral coherence lengths of the X-ray tungsten energy
spectrum are in the range of 0.3 µm to 13 µm. This property of X-rays allows phase contrast.
3 Results
Several samples of various density and composition were used to evaluate imaging properties of
presented experimental set-up for 2D radiography and 3D tomography. Results of CT analysis of a
sample containing materials with higher atomic number, Al-Cu alloy, are presented in section 3.1.
Biological samples are presented in section 3.2.
3.1 Absorption mode and artifacts
Lower energy X-rays passing through a sample are more absorbed than higher energy X-rays and
the polychromatic X-ray beam becomes harder. The effect is called beam hardening. This phe-
nomenon causes the so-called cupping effect in reconstructed slices. Figure 4a) shows a slice of
Al-Cu alloy sample and figure 4b) corresponding cupping profile. Beam hardening phenomenon
–3–
2012 JINST 7 C03005
Figure 3. Experimental setup: (1) Hamamatsu microfocus X-ray source, (2) Newport goniometer, (3) X-ray
MiniFDI camera, Photonic Science.
a) b)
Figure 4. a) Reconstructed tomography slice of an Al-Cu alloy without filtering of X-ray spectrum. 80
kV, 100 µA; exposure time 10 s, 2250 projections with angular step of 0.16owith geometry: SDD=272.8
mm, SOD=21 mm, MF=13. Darker regions represent the areas with lower density of alloy. A reconstructed
density value in the centre of the sample is seemingly lower than at the sample edges — beam hardening
artifact. b) Intensity profile of yellow horizontal region indicated in a) showing cupping shape with the sharp
peaks on the edges of the profile.
causes non-linear detector element response and it is one of several causes of the appearance of
ring artifacts [6]. The presence of such artifacts complicates quantitative analysis and post process-
ing such as noise reduction or voxel segmentation. To minimize this effect of beam hardening, it is
possible to pre-filter the X-ray spectrum of the beam. Typically, the filter is chosen to have a similar
composition to the specimen and with a thickness about 37% (e−1)the diameter of the specimen.
Using of filtering usually leads to a significant increase in the duration of the experiment, because
each projection needs more time to exposure [1].
In our case aluminium filter with a thickness of 0.25 mm was used. The exposure time was
necessary to be increased to 30 s. Unexpectedly, there was practically no suppression of cupping
effect in the intensity profile observed. This observation can be explained by the fact that the
–4–
2012 JINST 7 C03005
a) b)
Figure 5. a) Single projection of the light-weight biological sample (fly’s head, 1 mm height). Phase
contrast (edge effect) is observable, b) surface visualization using Avizo software after CT reconstruction
based on X-ray absorption tomography using Octopus software.
suppression of the beam hardening effect could not have sufficient response because of low working
energy range (5keV to 17keV) of our X-ray camera.
3.2 Phase contrast mode
The light-weight biological samples usually do not show beam hardening effect, so they do not
require filtration of X-ray spectrum. Figure 5shows one of 720 projections of light-weight sample
(fly’s head) with magnification factor of 2.5. SOD=307 mm and SDD=755.8 mm. They were
taken using full power of X-ray tube with voltage of 80 kV and current 100 µA, with exposure
time of 20 s, without filtration. For flat-field and dark-field corrections only auto-corrections in
camera software were used. Reconstruction of the object was done using Octopus software from
inCT [7] and object’s surface was visualized using AVIZO software from VSG - Visualization
Sciences Group [8] (see figure 5b). The effect of phase contrast, occurring at the object’s edges
in the 2D projections (see figure 5a) did not have observable influence on the image quality in the
final 3D reconstruction of sample volume (surface visualization in figure 5b). For example, bright
stripes which can appear close to the 3D reconstructed structure, known as ray aliasing [9] are not
observable in figure 5b).
In CT slices of the sample (see figure 6a) it is possible to recognize structural areas with wall
thicknesses of about 14 micrometers. For objects of this dimension the Fresnel number at distance
Z=449 mm is less than one at energies between 5 and 17 keV (energy response of CCD camera).
It can be easily ascertained that the 14 micrometers sized structure is in the Fresnel and Fraunhofer
regions where the phase effect is more significant. Strong contrast at the edges of the internal
structures is due to this effect. We would like to emphasize that this strong contrast is preserved
after the CT reconstruction. Additional weak contrast dark/white stripes in figure 6a) are the only
and minor artifacts observed after reconstruction.
One of the most suitable candidates to study the internal structure of light-weight objects
is a wooden sample. It has a very regular internal structure as shown in figure 7. Results of
measurements of the spatial resolution presented in [5] are confirmed by this CT experiment. We
can recognize a wall thickness of the wood cells about 3 µm (see figure 7a). SDD=386.8 mm and
–5–
2012 JINST 7 C03005
a) b)
Figure 6. a) CT reconstructed slice obtained with Octopus CT reconstruction software, b) cross-section
through the sample obtained with Avizo visualizing software.
a) b)
Figure 7. a) A CT slice of a piece of spruce wood obtained with Octopus reconstruction software and with
VG-studio visualization software [10]. b) 3D visualization and cross-section through the sample using also
VG-studio software.
SOD=95 mm. Magnification factor was 4. Number of projections was 2000 at full power of X-ray
tube 80 kV and 100 µA.
4 Discussion and conclusion
In the case of non-biological object (figure 4) made of Al-Cu alloy reconstruction was done by
using standard reconstruction algorithms. It showed artifacts, which complicate voxel segmenta-
tion in post processing by 3D visualization. Although 2D visualization of individual slices was
sufficient for technologists, it was difficult to obtain quantitative information on spatial distribution
of the denser regions at voxel segmentation (see the darker areas in figure 4a)).
Increased visibility of some features of the biological object in figures 5–7is caused by edge
enhancement due to the phase effect. It is very important to note that this edge effect is preserved
after the CT reconstruction (see figure 6). Standard reconstruction algorithms are based on X-ray
absorption tomography, where projections are formed by measuring the attenuation of radiation
that passes through a physical specimen at different angles. It could be assumed that phase effect
due to the tilt of the beam in place with the change of the refractive index may generate artifacts in
–6–
2012 JINST 7 C03005
reconstructed volume. However, figures 5–7show only weak ray aliasing (weak dark/white stripes
in figure 6a). Fine structures of the objects are preserved and well resolved in the final CT results.
Presented results show that our system is particularly suitable for light-weight and non-metallic
objects such as biological objects, plastics, wood, paper, etc., where phase contrast helps to increase
the visibility of the finest structures of the object. The obtained results show that the light-weight
materials are not so prone to artifacts in final tomograms. Phase contrast X-ray Computerized
Tomography is of our special interest because it is an emerging imaging technique that can be
implemented not only in synchrotron laboratories but also in home laboratories using microfocus
sources.
Acknowledgments
This work is the result of the ESF project implementation, CENTE - 2nd stage, ITMS code
26240120019, supported by the R&D Operational Program funded by the ERDF, (0.6). This work
was also supported by grant of Science and Technology Assistance Agency no. APVV SK-PL-
0059-09, APVV-0459-06, and Scientific Grant Agency of the Ministry of Education of Slovak
Republic and the Slovak Academy of Sciences No. VEGA- 2/0153/10 and 2/0192/10 and COST
Action MP0601.
References
[1] A. Sakellariou et al., X-ray tomography for mesoscale physics applications,Physica A 339 (2004)
152.
[2] R. Fitzgerald, Phase-sensitive X-ray imaging,Phys. Today. 53 (2000) 23.
[3] X. Wu and H. Liu, Phase-space formulation for phase-contrast X-ray imaging,Appl. Optics. 44
(2005) 5847.
[4] A. Bronnikov, Phase-contrast CT: Fundamental theorem ans fast image reconstruction algorithms,
Proc. SPIE 6318 (2006) 27.
[5] Z. Z´
apraˇ
zn´
y et al., Experience with imaging by using of microfocus X-ray source,J. Electrical Eng.
61 (2010) 287.
[6] J. Sijbers and A. Postnov, Reduction of ring artefacts in high resolution micro-CT reconstructions,
Phys. Med. Biol. 49 (2004) N247.
[7] http://www.inct.be/en/home.
[8] http://www.vsg3d.com/about-vsg-visualization-sciences-group.
[9] J.F. Barret and N. Keat, Artifacts in CT: Recognition and Avoidance,Radiographics 24 (2004) 1679.
[10] http://www.volumegraphics.com/en/products/vgstudio-max.html.
–7–