Dynamic imaging of the lungs using x-ray phase contrast.

Abstract

High quality real-time imaging of lungs in vivo presents considerable challenges. We demonstrate here that phase contrast x-ray imaging is capable of dynamically imaging the lungs. It retains many of the advantages of simple x-ray imaging, whilst also being able to map weakly absorbing soft tissues based on refractive index differences. Preliminary results reported herein show that this novel imaging technique can identify and locate airway liquid and allows lung aeration in newborn rabbit pups to be dynamically visualized.

Full text

Page 1

INSTITUTE OF PHYSICS PUBLISHING
PHYSICS IN MEDICINE AND BIOLOGY
Phys. Med. Biol. 50 (2005) 5031–5040
doi:10.1088/0031-9155/50/21/006
Dynamic imaging of the lungs using x-ray phase
contrast
R A Lewis1,3, N Yagi2, M J Kitchen3, M J Morgan3, D Paganin3,
K K W Siu3, K Pavlov3, I Williams3, K Uesugi2, M J Wallace4, C J Hall5,
J Whitley6and S B Hooper4
1Monash Centre for Synchrotron Science, Monash University, Victoria 3800, Australia
2SPring-8/JASRI, Mikazuki, Hyogo 679-5198, Japan
3School of Physics, Monash University, Victoria 3800, Australia
4Department of Physiology, Monash University, Victoria 3800, Australia
5Daresbury Laboratory, Warrington WA4 4AD, UK
6Department of Primary Industries, 475 Mickleham Road, Attwood, Victoria 3049, Australia
Received 31 May 2005, in final form 8 August 2005
Published 13 October 2005
Online at stacks.iop.org/PMB/50/5031
Abstract
High quality real-time imaging of lungs in vivo presents considerable
challenges. We demonstrate here that phase contrast x-ray imaging is capable
of dynamically imaging the lungs. It retains many of the advantages of simple
x-ray imaging, whilst also being able to map weakly absorbing soft tissues
based on refractive index differences. Preliminary results reported herein show
that this novel imaging technique can identify and locate airway liquid and
allows lung aeration in newborn rabbit pups to be dynamically visualized.
1. Introduction
Dynamic imaging of the small airways in the lung remains a largely unfulfilled goal of
biomedical imaging. In x-radiographs, the air filled airways are only poorly visible; however,
x-ray imaging has some advantages and is usually the first imaging method employed in
the diagnosis of lung pathologies. It is simple to perform, low cost, yields good spatial
resolution and the resulting images yield useful diagnostic information.
imaging modalities, such as MRI and CT, produce 3D images of soft tissue none can currently
produce high quality images of small airways in the lung.
Phase contrast x-ray imaging utilizes refractive index variations (phase information) in
addition to conventional absorption information to produce x-ray images of tissues with
greatly enhanced contrast, compared with conventional x-ray absorption techniques. It is able
to achieve this because the parameter describing the phase shift of the x-rays is generally more
thanthreeordersofmagnitudegreaterthantheabsorptiontermoverthediagnosticx-rayenergy
range(20keV–90keV).Asaresult,thephaseshiftofanx-raybeampropagatingthroughtissue
can be much larger than the absorption change and if phase effects are rendered visible they
can enhance image contrast. An important consequence of this is that phase contrast images
Although other
0031-9155/05/215031+10$30.00© 2005 IOP Publishing LtdPrinted in the UK5031

Page 2

5032R A Lewis et al
can be recorded with significantly lower dose than conventional images (Lewis 2004) which
is particularly important for both longitudinal studies and dynamic studies where repeated
imaging is required.
Several techniques have been employed to realize phase contrast x-ray imaging, but in the
experiments reported here, we utilized the simplest method which is known as propagation-
based phase contrast imaging (Wilkins et al 1996, Snigirev et al 1995). This modality is a
formofin-lineholography, whichhasalternativelybeennamed‘refractionenhanced imaging’
(Suzuki et al 2002, Yagi et al 1999).The method is straightforward and differs from
conventional x-radiography only in that the object and detector are separated by a sufficient
distance to allow the refracted rays to diverge from the undeviated ones. The amount of phase
contrast depends upon the object–detector separation. If this is small, as in the conventional
x-ray imaging, the deviated and the undeviated x-rays are not resolved by the detector and
no phase contrast is observed. The technique works with both coherent synchrotron radiation
(Snigirev et al 1995) and conventional micro-focus x-ray sources (Wilkins et al 1996).
However, the unique properties of synchrotron radiation, in particular its brightness, make
it ideal for these experiments. The small source size and inherent collimation yields excellent
phase contrast, whilst the large beam intensity allows monochromatic radiation and short
exposure times. With this configuration and a high-speed detector, it is possible to acquire
dynamic images with relative ease.
It has previously been shown that phase contrast x-ray imaging is capable of producing
remarkable images of the lung (Kitchen et al 2004, 2005b, 2005a, Yagi et al 1999). The effect
of propagation distance and the resulting phase contrast is illustrated in figure 1. It shows
x-ray images of a rabbit pup euthanized 48 h after birth. Figures 1(a) and (b) differ only in the
object–detector distance, which was 7 cm for figure 1(a) and 3 m for figure 1(b). The distance
of7cmwastheminimumpracticalforourexperimentalconfiguration imposedbymechanical
constraints and at that distance the image contains predominantly absorption contrast. The
3 m propagation distance allows significant phase effects to be rendered visible as intensity
modulations at the detector so that the resulting image is a combination of absorption and
phase contrast.
The enhanced contrast provided by the 3 m propagation distance over the 7 cm distance
is particularly evident in figures 1(c) and (d). The phase contrast is predominantly visible at
the edges of objects and this is particularly clear at the edge of the bubble visible at the bottom
right of figure 1(d). The ‘S’ curve corresponding to a black–white fringe pair can be seen
indicated by an arrow in the 3 m intensity profile (figure 1(e)). Note that there is no effect in
the 7 cm profile showing that the bubble is predominantly a phase object. The same is true of
the small peripheral airways which are clearly visible in figure 1(d) but not in figure 1(c).
In the main body of the lungs, the phase contrast image is rather more complex and takes
the form of a speckle pattern as can be seen in figures 1(b) and (d). The primary cause of this
pattern is the result of multiple refraction through many layers of aerated alveoli (Kitchen et al
2004) and is not visible in absorption images; see figures 1(a) and (c). The magnitude of the
phase contrast enhancement for the lungs is illustrated in figure 1(e) at pixel values below 70.
Here it can be seen that the phase contrast takes the form of a large oscillatory signal about the
average absorption level. The interpretation of this phase contrast signal yields information
on the projected thickness of material and hence the degree of local lung aeration (Kitchen
et al 2005b). In other words, using suitable phase retrieval algorithms, it is possible to exploit
the phase contrast signal to improve the projected thickness measurement normally provided
by absorption contrast.
Inthispaper, wereportanextension ofthistechnique toinclude dynamic studiesallowing
the lungs of rabbit pups to be imaged during the first minutes after birth. In particular, we

Page 3

Dynamic imaging of the lungs using x-ray phase contrast5033

(a)

(b)


(c)

(e)
(d)
500
600
700
800
900
1000
1100
1200
020 406080100
Pixel
120 140160 180 200
Intensity (Arb.Units)
Profile along line in (c) 7cm
Profile along line in (d) 3m
Edge of ribEdge of rib
Bubble edge
Figure 1. The effect of sample-to-detector propagation distance on images of a newborn rabbit
pup. Images are recorded using beamline 20B2 at the SPring-8 Japanese synchrotron facility using
an energy of 25 keV, an exposure time of 500 ms and a CCD x-ray camera having an effective pixel
size of 12 µm. (a) Distance = 7 cm, (b) distance = 3 m, (c) enlarged area marked by rectangle
in (a), and (d) enlarged area marked by rectangle in (b). (e) Intensity profiles as indicated by the
white lines in (c) and (d).
demonstrate that visualization of the lung using this technique is critically dependent upon the
presence of air within the airways as the liquid-filled fetal lung is not visible.
2. Lung fluid clearance
Lung diseases are major contributors to mortality and morbidity in people of all ages and
nationalities, yet we are unable to diagnose many of these disease states before they become

Page 4

5034R A Lewis et al
clinically evident and well advanced. Improved lung imaging techniques would enhance the
ability to detect many lung disease states at a much earlier stage, thereby greatly increasing
the therapeutic opportunities for treatment. In performing the experiments reported here, we
have focused on the need to better understand the transition to air-breathing at birth. The
survival of newborn infants is critically dependent upon the ability of the lungs to immediately
(within minutes) take over the role of respiratory gas exchange after birth. Before birth, gas
exchange occurs across the placenta and the lungs are filled with a liquid that is produced by
the fetal lungs and which plays a vital role in their development (Hooper and Harding 1995).
At birth, the lungs stop secreting liquid and the airways are cleared to allow the entry of air
into the lungs with the onset of ventilation. This dynamic process is not well understood and
resultsin acomplex interaction of physical and biochemical processes thatprecipitate changes
in pulmonary physiology that allow it to successfully adopt the role of a gas exchange organ
(Hooper and Harding 2005). However, it is apparent that the resuscitation, ventilation and
transition to extra-uterine life of infants that are born preterm are complicated by the failure
of the lungs to clear liquid from the airways at birth; similar problems can be experienced by
term infants that are delivered by caesarean section (Mason and Shannon 1997).
The ability to study lung aeration at birth and the factors that regulate, or enhance it,
is restricted by the difficulty of accurately measuring the rate and pattern of lung aeration.
Instead, most previous studies have focused on measuring rates of lung liquid clearance using
inert indicators added to the airway liquid. However, these studies have provided little or no
information on the rate or pattern of lung aeration within different regions of the lung or the
factors that influence it.
3. Experimental technique
In this study, we have imaged the lungs of rabbit pups delivered by caesarean section at the
SPring-8 synchrotron radiation research facility, Japan, on beamline 20B2 of the Biomedical
ImagingCentre(Gotoetal2001). Experimentswereperformedinaccordancewithregulations
set by the SPring-8 Animal Care and Use Committee and approved by Monash University
Animal Ethics Committee. A Si(311) double-bounce monochromator aligned to the (111)
atomic planes was used to select a narrow bandwidth of the bending magnet radiation.
Following previous studies (Suzuki et al 2002), 25 keV x-rays were chosen to provide both
good phase and absorption contrast from the rabbit pup lungs. All objects were positioned in
hutch 3 of beamline 20B2, approximately 210 m downstream of the synchrotron source. The
detector was placed a further 3.0 m downstream to record the phase contrast images. Images
were recorded with a Hamamatsu phosphor charge-coupled device (CCD) detector (C4742-
95HR) with 5.9 µm pixels. It had a measured spatial resolution of approximately 25 µm and
an active area of 24.0 × 15.7 mm2. With these optics at this distance the monochromatic beam
covered an area of approximately 19.8 × 10.1 mm2.
Following image acquisition, custom software was used to correct for dark current and
non-uniform beam intensity effects. This was achieved by the usual method of recording a
flat field with identical illumination to the experimental image but with no object and a dark
image with the x-ray shutter closed. Identical exposure times were used for the experimental
and the correction images. Both the experimental image and the flat field image were first
corrected by subtracting the dark image and then the dark corrected experimental image was
divided by the dark corrected flat field image.
Anaesthesia of pregnant rabbits at 30–31 days of gestation was induced by a 12 mg kg−1
bolus and maintained by a 100 mg kg–1h–1infusion (i.v.) of Rapinovet (10 mg ml−1propofol,
Schering). Rabbit pups were then delivered one at a time by caesarean section. Two imaging

Page 5

Dynamic imaging of the lungs using x-ray phase contrast5035
protocols were employed: Firstly, static images of pups were recorded post mortem, with
the pups being either euthanized before the onset of air-breathing, following a fixed number
of breaths, or at a set time interval after the onset of ventilation. Each pup was euthanized
via an intraperitoneal injection of sodium pentobarbitone (0.5 ml; 150 mg) and following
delivery of the last pup, the doe was also euthanized (1 g i.v., sodium pentabarbitone). The
technical parameters for imaging were optimized during static experiments to develop the
second functional imaging protocol that was used to image the lung aeration dynamically in
individual live pups. The pups were continuously imaged at 4 s intervals for the first hour
after birth.
For the static imaging, the rabbit pups were mounted vertically and imaging the entire
thorax required the tiling together of several images. Exposure times per frame for all imaging
were 588 ms corresponding to a surface entrance dose of approximately 1 mGy per frame.
In the dynamic studies, the pups were imaged lying prone and a smaller field of view was
accepted since tiling was clearly not possible.
As a means of quantifying the volumetric change of lung fluid in the dynamic study, we
have utilized a phase retrieval algorithm for evaluating the projected object thickness from a
single-phase contrast image (Paganin et al 2002). This reconstruction approach assumes that
the object consists of a single material and that the phase (φ) and intensity (I) directly after
the object are related to its thickness (T) in a trivial way. That is, φ = 2πδT/λ where δ is
the difference between the refractive index of the material and unity (δ = 1 − n) and λ is the
x-ray wavelength; I = I0exp(−µT) where I0is the intensity incident on the sample and µ is
the attenuation coefficient. The implementation of this algorithm for biological samples has
been previously discussed elsewhere (Kitchen et al 2005b).
Clearly the thorax of a rabbit pup is not composed of a single material. However, in a
region of interest that is fixed relative to the anatomy throughout the period of interest, it is
reasonabletoassumethatthechanges inprojectedthicknessaredominatedbytheclearanceof
lung liquid. Such an assumption means that whilst the projected thickness cannot be explicitly
retrieved, the changes observed will yield an indicative measure of the relative changes in lung
liquid. Other errors arise from animal motion, image noise and our additional assumption that
the lung liquid is pure water with consequent uncertainties in the magnitudes of δ and µ. For
these studies, µ for water at 25 keV was taken as 50.7 m−1as calculated from the National
Institutes of Standards and Technology (NIST) database (Chantler et al 2003). At this energy,
the uncertainty can be as much as 5% from the NIST theoretical value (Midgley2005). δ was
computed as 3.81 × 10−7using atomic scattering factors given by NIST using the method
outlined by Kitchen et al (2004). For each frame of the dynamic sequence, the ‘projected
thickness’ from the region of interest (ROI) was calculated and summed over the ROI to yield
an effective volume. The ‘volume’ calculated from the first image was then subtracted from
successive volume calculations to yield a relative change. Given the assumption that the major
change in the object’s composition is the replacement of lung liquid by air, the subtraction
minimizes the errors caused by the presence of multiple materials.
4. Results
The results from imaging a sequence of pups euthanized at birth and five different time points
after birth are displayed in figure 2. The first image is of a pup euthanized before the onset of
air-breathing and, as the lungs and trachea are liquid-filled, no structures of the respiratory tree
arevisible. Thisisbecausebothphaseandabsorptioncontrastbetweenfluidintheairwaysand
tissue are negligible compared with that between air and tissue. However, within the first 12
breaths after birth, the trachea and large bronchi become clearly visible as they fill with air.

Page 6

5036R A Lewis et al


(a)
(b) (c)

(d)(e)(f)
Figure 2. Phase contrast images of newborn rabbit pups. The times indicate when the pups
were euthanized after birth: (a) Fetal (before 1st breath), (b) 12 breaths, (c) 5 min, (d) 15 min,
(e) 30 min, (f) 1 h.
As time progresses, the gas exchanging regions of the lungs gradually aerate revealing the
terminalairspacesasaspeckledintensitypattern. Thepatternoflungaerationwasnotuniform
across the entire lung, with different regions aerating at different rates. This is particularly
evident in figure 2(c) and was attributed to body position during postnatal ventilation; the
highest vertical regions of the lung aerated first with aeration of the lower regions being
significantly delayed.
Figure 3 shows selected frames from an x-ray movie from the dynamic study recorded
during the first hour following birth of a single live rabbit pup. As with the static images
of the euthanized rabbit pups in figure 2, the initial images clearly reveal the major airways
and the branching structures down to the tertiary bronchi. The speckle pattern only appears
once the smaller airways and alveolar structures become aerated. The animal from which
these images were recorded displayed shallow breathing for the first few minutes after birth
and then the inspiratory effort became visibly more pronounced, leading to accelerated
rates of aeration. It can be seen from figure 3 that as time and the number of inhalations
increase, both the visibility and the brightness of the speckle pattern of the lungs increase
dramatically.
We have performed time-dependent volumetric analyses of the two regions, indicated by
the white boxes in figure 3(a) in order to quantify the amount of air within the regions of
interest covering portions of the upper and lower lungs. We have also employed the phase
retrievalprocessaspreviouslydescribed(Kitchenetal2005b). Sincetheanimalwasbreathing
normally with no synchronization between inspiratory efforts and imaging, it was necessary to

Page 7

Dynamic imaging of the lungs using x-ray phase contrast5037

0s 120s

240s 360s
480s 600s
Figure 3. Time series of a breathing rabbit pup showing selected frames at various times after the
onset of imaging. Breathing commenced 30 s earlier. Exposure time: 588 ms. The white boxes
define the ROI for volumetric analysis and measure 0.72 × 3.30 mm2.
ensurethattheregionsofinterestremainedinthesameplacerelativetoanatomicallandmarks,
such as ribs. This was accomplished by moving the ROIs in accordance with a simple cross-
correlation between image frames to determine the local chest wall movement. The results of
the time-dependent volumetric analysis are presented in figure 4. Initially, the data show the
lungstobeaeratingslowlybutamajorincreaseinairvolumeoccursafterapproximately400s.
This time point correlated with a substantial increase in the pup’s breathing activity. With
increased activity, the volume measurements fluctuate dramatically due to the chest expansion
and image blurring. At these points the uncertainty in volume is as much as 60%, yet the time

Page 8

5038R A Lewis et al
Volume Change (mm^3)
0
-5
-10
-15
-20
Time (s)
6004002000
H2O Volume Change (Chest area: 0.72x3.30 mm^2 )
Lower Lung
Upper Lung
Figure 4. Volume change of liquid in lung tissue calculated using the phase retrieval technique
described in the text. The selected ROIs corresponding to this volume are labelled in figure 3 with
white rectangles.
averaged value gives us an indication of the real volume change within the ROI. The gaps in
the data are due to the time required by the current system to allow data to be written to disk
once the computer memory has filled.
5. Discussion
Previous studies have demonstrated that phase contrast x-ray imaging is able to elucidate
the fine structure of the lung much more clearly than conventional absorption-based imaging
(Kitchen et al 2004, Yagi et al 1999). In simple terms, the technique can be thought of as
providing better contrast of an aerated lung by exploiting refraction effects at the air–tissue
interfaces. Inthesmallairwaysandalveoli,multiplerefractioneffectsgiverisetoaspeckle-like
pattern (Kitchen et al 2004). In the larger structures of the trachea and major bronchi, single
refraction gives rise to a dark fringe on the inner side of the airway surface, which serves
to enhance the edges of these structures. Consequently, propagation-based phase contrast
x-ray imaging provides a straightforward way of visualizing lung aeration and hence lung
liquid removal from the immediate neonatal period until a few hours after birth.
We have demonstrated that the time course of air entry into the lungs can be studied to
some extent by euthanizing pups at various times after birth (figure 2). Whilst this protocol
has some advantages over imaging a live animal in that the lack of movement allows high
signal to noise to be achieved, there are clearly limitations to this approach. Each animal only
provides a single time point and it seems highly likely that the pattern of aeration will differ
for different pups. It is therefore impossible to follow the progress of aeration precisely by
this method and so a time resolved, continuous observation on a live pup is required to provide
accurate physiological information over the whole process.
Dynamicphasecontrastimagingoflungaerationdemandsthatsufficientfluxbeavailable
to see the details of the lung structure at the required time resolution. The time resolution is
driven not only by the exposure time necessary (a function of the x-ray flux and energy for a
given object), but also the readout time of the detector. The latter period is not insignificant for

Page 9

Dynamic imaging of the lungs using x-ray phase contrast5039
high-resolution CCD devices, being comparable to or even considerably longer than the actual
exposure time. In the present study, the time resolution possible was primarily restricted by
the readout speed of the CCD detector to no faster than 1.7 frames per second.
At this frame rate, the effect of breathing movements on the images is readily apparent,
as can be seen in figure 3 in the 600 s image. Nonetheless, this study demonstrates that
phase contrast imaging can provide important physiological information in real time. It is
motion artefacts that give rise to the apparently high noise level in the volumetric analyses
beyond 400 s (figure 4). During breathing the images blur and the expansion and contraction
of the chest changes the lung shape, making analysis more complex. From the relative
positions of the spinal cord and sternum, we have made an assessment of the chest rotation
and attempted to account for this in the positioning of the ROI. Despite these difficulties, the
prototype volumetric analysis not only provides quantitative information on the rate of lung
liquid clearance and lung aeration, but also suggests that body position and gravity influence
the process; this is shown in figure 4 which demonstrates that the decrease in liquid volume is
significantly delayed within the lower regions of the lung.
A further limitation to dynamic imaging is posed by the intrinsic collimation of the
synchrotron x-ray beam. At the Biomedical Imaging Centre of beamline 20B2 at SPring-8,
the beam has a maximum width of approximately 300 mm and 20 mm in height at 25 keV
and becomes somewhat smaller at higher x-ray energies. The beam size is a consequence of
the natural beam divergence and the length of the beamline and is sufficient to image small
animals. The height of the beam at the specimen could be increased by diffraction from an
asymmetrically cut crystal; however, this would have the effect of reducing the flux density
and hence compromising the time resolution. The frame rate could be increased by reducing
the number of pixels in the image. Higher time resolution can therefore be achieved by either
viewing a small area or reducing the spatial resolution by pixel binning.
Forphysiologicalstudies,oneshouldideallybeabletofollowtheairflowduringbreathing
at sufficiently high time and spatial resolutions. This requires a framing rate higher than 100
frames per second. Inorder toachieve this, anincrease inthebeam intensitywillbe necessary.
This could be achieved by the use of a multipole wiggler instead of a bending magnet or by
broadening the energy bandpass of the incident beam, since the phase contrast effect is only
weakly dependent upon the degree of monochromaticity (Wilkins et al 1996). Improvements
in the x-ray detector efficiency (around 20% in the current study) and the CCD camera (for
rapid readout whilst maintaining high spatial resolution) will also be necessary to achieve this
goal whilst maintaining a low dose to the animals.
Furtherworkisalsorequiredtodeterminethemostadvantageousconditionsforobserving
lung aeration, including optimizing the propagation distances and x-ray energies used.
6. Conclusions
We have demonstrated that propagation-based phase contrast imaging has potential for real-
time, dynamic imaging of lung function. The technique is simple to implement, provided
one has access to a synchrotron, and provides much improved contrast in the soft tissues over
conventional x-ray imaging methods. Static images of euthanized rabbit pups revealed that
the technique is able to provide detail down to the terminal airways. Dynamic imaging yielded
semi-quantitative information on the rate of liquid clearance from, and aeration of, the lung
and demonstrated the importance of body position on these processes.
Current technical limitations are largely detector based.
sensitivity of the detector limit the dynamic imaging frame rate to a few frames per second,
whilst the active area limits the field of view to less than 2 × 2 cm. Larger area detectors with
The readout time and the

Page 10

5040R A Lewis et al
sufficient spatial resolution (∼50 µm) to reveal phase contrast are available at SPring-8 and
future work will utilize such devices.
Acknowledgments
We gratefully acknowledge support from the Access to Major Facilities Program, which is
fundedbytheCommonwealthofAustraliaandSPring-8beamtimeawardnumber2004A0556-
NL3-np.Marcus Kitchen acknowledges an Australian Postgraduate Award, Karen Siu,
an Australian Synchrotron Research Program Fellowship and Konstantin Pavlov a Monash
Synchrotron Fellowship.
References
Chantler C T, Olsen K, Dragoset R A, Kishore A R, Kotochigova S A and Zucker D S 2003 X-Ray Form Factor,
Attenuation and Scattering Tables (version 2.0) [Online] (NIST)
Goto S et al 2001 Construction and commissioning of a 215 m-long beamline at SPring-8 Nucl. Instrum. Methods
Phys. Res. A 467–8 682–5
Hooper S B and Harding R 1995 Fetal lung liquid: a major determinant of the growth and functional development of
the fetal lung Clin. Exp. Pharmacol. Physiol. 22 235–47
Hooper S B and Harding R 2005 Role of aeration in the physiological adaptation of the lung to air-breathing at birth
Curr. Resp. Med. Rev. 1 185–95
Kitchen M J et al 2005a Phase contrast x-ray imaging of mice and rabbit lungs: a comparative study Br. J. Radiol.
at press
Kitchen M J, Paganin D, Lewis R A, Yagi N and Uesugi K 2005b Analysis of speckle patterns in phase contrast
images of lung tissue Nucl. Instrum. Methods Phys. Res. A 548 240–6
Kitchen M J, Paganin D, Lewis R A, Yagi N, Uesugi K and Mudie S T 2004 On the origin of speckle in x-ray phase
contrast images of lung tissue Phys. Med. Biol. 49 4335–48
Lewis R A 2004 Medical phase contrast x-ray imaging: current status and future prospects Phys. Med. Biol. 49
3573–83
Mason R and Shannon J 1997 Alveolar type II cells The Lung: Scientific Foundations ed R Crystal, J West, E Weibel
and P Barnes (Philadelphia: Lippincott-Raven)
Midgley S M 2005 Measurements of the x-ray linear attenuation coefficient for low atomic number materials at
energies 32–66 and 140 keV Radiat. Phys. Chem. 72 525–35
Paganin D, Mayo S C, Gureyev T E, Miller P R and Wilkins S W 2002 Simultaneous phase and amplitude extraction
from a single defocused image of a homogeneous object J. Microsc. 206 33–40
Snigirev A, Snigireva I, Kohn V, Kuznetsov S and Schelokov I 1995 On the possibilities of x-ray phase contrast
microimaging by coherent high-energy synchrotron radiation Rev. Sci. Instrum. 66 5486–92
Suzuki Y, Yagi N and Uesugi K 2002 X-ray refraction-enhanced imaging and a method for phase retrieval for a
simple object J. Synchrotron Radiat. 9 160–5
Wilkins S W, Gureyev T E, Gao D, Pogany A and Stevenson A W 1996 Phase-contrast imaging using polychromatic
hard x-rays Nature 384 335–8
Yagi N, Suzuki Y, Umetani K, Kohmura Y and Yamasaki K 1999 Refraction-enhanced x-ray imaging of mouse lung
using synchrotron radiation source Med. Phys. 26 2190–3