Content uploaded by Marian Willner
Author content
All content in this area was uploaded by Marian Willner on Aug 31, 2015
Content may be subject to copyright.
Non-invasive Differentiation of Kidney
Stone Types using X-ray Dark-Field
Radiography
Kai Scherer
1
, Eva Braig
1
, Konstantin Willer
1
, Marian Willner
1
, Alexander A. Fingerle
2
, Michael Chabior
1
,
Julia Herzen
1
, Matthias Eiber
2
, Bernhard Haller
3
, Michael Straub
4
, Heike Schneider
5
, Ernst J. Rummeny
2
,
Peter B. Noe
¨
l
2
& Franz Pfeiffer
1
1
Lehrstuhl fu
¨
r Biomedizinische Physik, Physik-Department & Institut fu
¨
r Medizintechnik, Technische Universität Mu
¨
nchen, Garching,
Germany,
2
Department of Radiology, Technische Universität Mu
¨
nchen, Munich, Germany,
3
Department of Medical Statistics and
Epidemiology, Technische Universität Mu
¨
nchen, Munich, Germany,
4
Department of Urology, Technische Universität Mu
¨
nchen,
Munich, Germany,
5
Department of Clinical Chemistry and Pathobiochemistry, Technische Universität Mu
¨
nchen, Munich, Germany.
Treatment of renal calculi is highly dependent on the chemical composition of the stone in question, which is
difficult to determine using standard imaging techniques. The objective of this study is to evaluate the
potential of scatter-sensitive X-ray dark-field radiography to differentiate between the most common types
of kidney stones in clinical practice. Here, we examine the absorption-to-scattering ratio of 118 extracted
kidney stones with a laboratory Talbot-Lau Interferometer. Depending on their chemical composition,
microscopic growth structure and morphology the various types of kidney stones show strongly varying,
partially opposite contrasts in absorption and dark-field imaging. By assessing the microscopic calculi
morphology with high resolution micro-computed tomography measurements, we illustrate the
dependence of dark-field signal strength on the respective stone type. Finally, we utilize X-ray dark-field
radiography as a non-invasive, highly sensitive (100%) and specific (97%) tool for the differentiation of
calcium oxalate, uric acid and mixed types of stones, while additionally improving the detectability of
radio-lucent calculi. We prove clinical feasibility of the here proposed method by accurately classifying renal
stones, embedded within a fresh pig kidney, using dose-compatible measurements and a quick and simple
visual inspection.
T
he correct identification of the different renal calculi commonly found in the human body is of essential
importance for the correct diagnosis, prognosis and therapy of many common diseases of the genitourinary
system. For example, while urinary acid stones can occur in any healthy subject, struvite stones indicate an
infection within the patient. In therapy, lithotripsy can be successfully administered for the uric acid type of
kidney stones, while other types of calculi are more resistant to this type of therapy
1,2
. Nevertheless, while standard
imaging methods like computed tomography and sonography are helpful in localizing calculi in the body, they
only yield modest results in the correct identification of the stone type
3,4
. An examination of the patient’s urine, or
a removal of exemplary stones followed by a histological work-up is required in most cases. Currently, dual energy
CT is evaluated with some success
5
.
Recent developments in phase-sensitive X-ray imaging
6–9
have broadened the horizon of X-ray image contrast
generation and are currently being evaluated for clinical application in a variety of diagnostic fields
10
. Among
these, X-ray dark-field imaging
11
attracted particular interest, being sensitive towards structural changes in the
micro-morphology of tissue, as for instance associated with pathological processes of breast and lung tissue
12–14
.
In contrast to absorption-based imaging, which solely relies on the reduction of beam intensity when introducing
a specimen, dark-field contrast is generated by diffuse angular deflections of the X-ray wave-front when being
scattered at inherent sub-structures. By resolving the scatter associated reduction of a phase-grating induced
interference pattern, the dark-field signal strength can be quantified, as illustrated in Fig. 1 (for a detailed
description of the technique, see Ref. 8). The dark-field signal has been shown to be highly dependent not only
on the chemical composition of the imaged sample, but decisively also on the sample’s morphological structure
on the micrometer scale
15,16
, well below the resolution limit of commonly used imaging detectors.
The idea underlying our present work is to try to discriminate uric acid, calcium oxalate and mixed types of
stones from each other within a radiographic imaging mode, on the ground of the complementarity of their
OPEN
SUBJECT AREAS:
UROGENITAL DISEASES
IMAGING
Received
2 November 2014
Accepted
11 March 2015
Published
Correspondence and
requests for materials
should be addressed to
K.S. (kai.scherer@tum.
de)
SCIENTIFIC REPORTS | 5 : 9527 | DOI: 10.1038/srep09527 1
1 April 20155
absorption and dark-field contrasts, which is based on differences
between their morphological and chemical compositions. While the
absorption and dark-field images will be obtained from a lab-based
radiography setup, the micro-morphological information (which are
used to illustrate the generation of dark-field signal strength) will be
assessed using highly resolving micro-CT.
Results
Analytical description. Formally, the measured projection value in
absorption contrast
m can be written as
m~{ln T~
ð
L
0
m(z) dz, ð1Þ
where the transmission T 5 I/I
0
is obtained from the measured
intensity I, relative to a reference intensity I
0
measured without the
kidney stone, m being the absorption per unit length and L the stone
thickness. As shown by Bech et al.
17
, under the simplified assumption
of ideally random scattering, the dark-field signal in projection mode
can similarly be written as
E~{ln D~i
ð
L
0
E(z) dz, ð2Þ
where the dark-field signal D 5 V/V
0
can be obtained from the
interferometric visibilities V and V
0
with and without stone,
respectively, E being the linear diffusion coefficient, quantifying the
scattering per unit length, and i being a setup-specific constant.
To account for the problem of overlaying structures in projection
mode, we can formally describe the kidney stone as consisting of a
perfectly homogeneous material along each projection path, and
assign to this hypothetical material an ‘‘effective’’ absorption and
scattering coefficient m
eff
and E
eff
, respectively,
m~m
eff
:
L and
E~i
:
E
eff
:
L: ð3Þ
The effective coefficients m
eff
and E
eff
are thus defined as a
weighted average of the contribution of the absorption and scattering
coefficients along each projection path
m
eff
~
1
L
ð
L
0
m(z) dz and E
eff
~
1
L
ð
L
0
E(z) dz: ð4Þ
Interpreted in this way, the ratio of the projection values
m
E
~
m
eff
i
:
E
eff
~c; ð5Þ
can be seen to be independent of the total kidney stone thickness L.
Thus, in this approximation, we assume that there is a linear rela-
tionship between the measured
E (E
eff
) and
m (m
eff
) values, and that
the slope c relating the two parameters is constant and characteristic
for each kidney stone type. The simultaneous measurement of
absorption and scattering thus allows the cancellation of the thick-
ness dependence in projection mode, as well as the identification and
classification of different kidney stones by using the obtained slope c
as a binary classifier.
Absorption characteristics of renal calculi. The effective absorption
coefficient m
eff
is proportional to Z
3{4
eff
, whereas the effective atomic
number Z
eff
of the composite is mostly determined by the heaviest
element in the kidney stone
18
. Thus, with respect to absorption, two
classes of kidney stones can immediately be differentiated: the uric
acid type of stones on the one hand, and the calcified stones (the
oxalate, brushite and apatite/dahllite) on the other hand. While the
heavy calcium ion in the calcified stones leads to a strong absorption
signal ( Z
eff
< 14–16, large m
eff
), the uric acid stones contain only low
Z elements like carbon, nitrogen and oxygen which implies a small
absorption signal (Z
eff
< 7, small m
eff
), respectively. Struvite and
cystine have intermediate atomic numbers (Z
eff
< 10–12).
Nevertheless, those chemicals mostly occur only in combination
with other crystallite phases and thus usually fall into the mixed
stone category, with effective absorption coefficients m
eff
ranging
between the uric acid and calcium oxalate class.
Scattering characteristics of renal calculi. The classification of
kidney stones by their scattering properties is more difficult, since
the morphological structure and stone formation heavily depend on
the mineralogical composition, the time varying chemical com-
position of the urine, the location and time of formation, the
presence of growth inhibitors and catalyses, the inclusion of
organic matrix, among others
19,20
. Thus, for the sake of simplicity,
micro-CT investigations are restricted to pure uric acid and calcium
oxalate stones in the following, signifying two micro-morphological
extremes:
Uric acid type of kidney stones are known to grow in a layer-wise
manner as concentric rings around a crystallite core
19
. This multi-
shell structure is well reflected by the micro-CT measurement of an
uricite stone as shown in Fig. 2A. The inner structure of the uricite
stone displays a high textural irregularity comprising various grain
sizes rutted with ring-like structures of higher optical density.
Further, the exterior exhibits a distinct surface roughness containing
large cavities and sharp edges. Since scatter predominately originates
at boundaries with locally changing density and structure, uric acid
stones are expected to have a large effective scatter coefficient E
eff
.
Figure 1
|
Contrast generation in grating-based X-ray dark-field imaging. (A) Typical setup with a conventional X-ray source, an source grating
G
0
, a phase grating G
1
, an analyzer grating G
2
and a flat panel detector. (B) Diffuse X-ray scatter originating from sub-structures of the kidney stone
manifest themselves in a local reduction of a phase-grating induced interference pattern. By analysing the dark-field signal, information on the kidney
stone micro-morphology, well below the detector resolution limit can be retrieved.
www.nature.com/scientificreports
SCIENTIFIC REPORTS | 5 : 9527 | DOI: 10.1038/srep09527 2
In contrast, in the case of calcium oxalate stones crystal forming is
typically driven by a slow and regular crystalline growth (the exact
stone formation is complex and dependent on many factors, among
others the presence of crystallization cores and the ratio of mono-
hydrate to di-hydrate)
20
. As a consequence, especially in the case of
calcium oxalate mono-hydrate, stones exhibit a relatively homogen-
eous micro-structure with wedges rounding off and forming a
smooth exterior
21
. This corresponds well with the micro-CT mea-
surements of a 90% mono-hydrate and 10% dihydrate stone featur-
ing a strongly uniform micro-morphology with only minor
structural disturbances in the form of some air cavities as shown in
Fig. 2B. The high degree of regularity in grain size, a steady optical
density throughout the stone and smooth stone surface manifest
themselves in a very small effective scatter coefficient E
eff
.
Mixed types of stones comprise a more variable crystalline growth
pattern, being more irregular in shape and structure than calcium
oxalate stones, hence are expected to yield an intermediate effective
scatter coefficient E
eff
.
X-ray dark-field radiography of renal calculi. Grating-based
transmission and dark-field radiographies of an excised calcium
oxalate and uric acid stone can be seen in Fig. 3A & B. While the
calcium oxalate stone (top left inlay) exhibits a relatively high
absorption (low transmission T, large m
eff
) and weak scattering
signal (high dark-field D, small E
eff
), directly inverse observations
are made in case of the uric acid stone (top right inlay), which is in
accordance with previously discussed chemical and morphological
stone properties. Based on the complementarity of both image
signals, the two calculi can be clearly and easily differentiated in
radiographic mode, as shown in Fig. 3C in false color, by deriving
the thickness-independent dark-field-to-transmission ratio D/T.
Afterwards, we measured a fresh pig kidney with both stones
embedded in the inside with an imaging dose of 5.2 mGy (bottom
inlays), to demonstrate the potential of kidney stone assessment via
X-ray dark-field radiography as a future in-vivo application.
Consequently, a superior visual differentiation and discrimination
between the two calculi and surrounding kidney tissue could be
achieved also within native tissue. Notice that images were normal-
ized with respect to surrounding tissue/material in order to com-
pensate for signal not directly arising from the kidney stones
themselves.
Statistical analysis of renal calculi classification. To evaluate this
trend statistically, i.e. review the potential to differentiate uric acid
and calcium oxalate from each other and also discriminate the latter
from mixed stone types, a cohort of 118 stones was analysed. Each
stone was segmented, normalized with respect to the background
and separately analysed by generating a scatterplot of
m versus
E
using every pixel of each stone as data-points, as exemplary shown
for three stones in Fig. 4A. Here, data points close to the origin are
associated with the margins of the respective stone, while the
maximal values of
m and
E correspond to the region where the
stone is vastest. Each point cloud belonging to a specific stone was
then analysed separately using linear regression through the origin,
evaluating the slope c (used as binary classifier) and the coefficient of
determination R
2
of each stone. For each class, the obtained slopes
were then arithmetically averaged to obtain a mean slope value
c
which is characteristic for each class. For uric acid stones, we
obtain
c
U
~0:13+0:01, for the calcium oxalate
c
O
~0:97+0:44,
and for the mixed stones
c
M
~0:31+0:13. The error is obtained as
the standard deviation of the slopes ensemble of each class. The
coefficient of determination was used as an indicator of the
goodness-of-fit and calculated to R
2
5 0.80 6 0.16 over the full
sample collective, thus justifying the linear proportionality
assumption expressed by Eq. 5.
We investigated the distributions of slope values c for each of the
118 stones in more detail by using a box-whisker diagram as shown
in Fig. 4B. In addition to significantly differing median values (black
dash), no overlap in interquartile data (50% of the data set as indi-
cated by boxes) was observed for either of the three stone classes.
Further, except two outliers (circles) all uric acid stones exhibited
exclusively flatter slopes c than in comparison with the calcium oxal-
ate stones.
Diagnostic performance of renal calculi classification. To verify
whether the decisive change in effective absorption
m to scatter power
E with respect to stone class yields sufficient diagnostic value, receiver
operating characteristic (ROC) analyses were carried out. The ROC
curve provides combinations of specificity vs. sensitivity when using
the slope c as a binary classifier with varying threshold as shown in
Fig. 4C. The diagnostic performance of renal calculi assessment via
dark-field imaging was estimated by the area under the ROC curve
(AUC). In the case of uric acid and calcium oxalate stones, a nearly
Figure 2
|
Comparison of the uric acid and calcium oxalate stone micro-morphology. (A) The volumetric micro-CT rendering of an uric acid-type
kidney stone reveals a highly concentric growth structure, accompanied by a particular rough stone surface. The zoom-in of the tomographic slice shows
high textural and optical irregularity induced by the multi-shell structure. (B) The volumetric micro-CT rendering of the calcium oxalate (90% mono-
hydrate and 10% di-hydrate) stone, showing a strongly homogeneous micro-structure with a smooth stone exterior. The zoom-in of the tomographic
slice reveals fine, regularly distributed crystallite cores on the micrometer scale. The non-uniformity in optical density and structure in the case of the
uricite stone compared with the calcium oxalate stone manifests itself in a significantly increased effective scatter coefficient E
eff
. For an animated
volumetric micro-CT rendering of both stones find Movie S1.
www.nature.com/scientificreports
SCIENTIFIC REPORTS | 5 : 9527 | DOI: 10.1038/srep09527 3
unambiguous discernability between both classes was found using X-
ray dark-field radiography, quantified by a AUC value of 0.99 (95%
bootstrap confidence interval of 0.98 to 1). An optimal threshold
value (Youden-Index), maximizing the sum of diagnostic sensitivity
(100%) and specificity (97%), was found for c 5 0.26. Besides, also
mixed type of renal stones were found to be distinguishable with a
high accuracy from both uric acid (AUC of 0.94, 95% bootstrap
confidence interval of 0.84 to 1) and calcium oxalate stones (AUC
of 0.93, 95% bootstrap confidence interval of 0.88 to 0.97).
Discussion
Here, we have shown that the comparison of absorption and dark-
field signal strength can determine the composition of different cal-
culi classes of the genitourinary system. Our study was able to estab-
lish a clear trend in the absorption-to-scattering ratio, which we
could directly assign to chemical and morphological differences of
calcium oxalate and uric acid stones. We further deepened this cor-
relation by means of statistical analyses and scatter plots. A simple
visual inspection of the dark-field-to-transmission signal strength
was presented to allow a quick and convenient determination of
the stone type, compatible with clinical routine. Finally, receiver
operator characteristics including 118 stones from 18 patients
revealed an outstanding diagnostic performance of dark-field radio-
graphy for the accurate differentiation of pure uric acid and calcium
oxalate calculi as well as discrimination of mixed types of stones. To
secure our statistical findings and further clarify the origins of dark-
field contrast with respect to stone micro-morphology, especially
aiming at complicated, rare and mixed stone types, more work
including an increased sample collective and patient cohort is to be
performed in the near future. Follow-up studies will focus on the
deduction of composite-specific classifier values, enabling a more
detailed differentiation of mixed types of stones into their chemical
sub-groups.
Besides, we consider X-ray dark-field radiography to provide a
superior detection sensitivity towards certain renal stones in com-
parison to conventional radiography, due to the demonstrated com-
plementarity of absorption- and scatter-based imaging. While uric
acid stones are usually entirely radio-lucent, which involves a high
risk of being overlooked in conventional radiography and CT (see
Fig. 3A), they are clearly revealed and delineated by the dark-field
signal (see Fig. 3B)
22
. Also in the case of mixed types of stones, the
dark-field signal strength exceeds the respective absorption entity by
far (c=1, Fig. 4B), which is of major clinical interest taking into
account that only 60% of all renal stones are radiopaque
23
.
As this initial study was aimed at determining the potential of
dark-field imaging in the differentiation of kidney stones in the sense
of a proof-of-principle study, mostly excised stones were measured
within an ex-vivo framework. In a first step, we could successfully
verify our classification scheme by fully scanning a fresh pig kidney
with two manually embedded stones, while keeping the dose applied
considerably low at 5.2 mGy. Although this value may not directly
apply to a full abdomen scan, it is in the same order of magnitude as
clinical dose values (0.7 mGy and 8.0 mGy in case of an abdomen
radiogram and CT, respectively)
24
. Thus we are convinced that dose-
compatible abdomen dark-field radiography could be achieved, con-
sidering that an optimization of interferometer efficiency by tuning
several setup entities (grating height and quality, duty cycle, beam
energy and filtration) would imply a significant decrease in dose,
while maintaining equivalent image quality (for a more detailed dose
discussion, see Refs. 25, 26).
For the purpose of further pursuing clinical transferability of the
proposed method in the near-term, we modelled a first medical
meaningful scenario mimicking an abdomen phantom: A second
pig kidney, with both stones embedded was placed within a 11 cm
water-bath and measured at a high-energy laboratory setup running
at 100 kVp, which is the energy range of tube voltages used in com-
mercial systems. The preliminary obtained measurements shown in
Figure 3
|
Visual classification scheme for the discrimination of uric acid and calcium oxalate renal stones using X-ray dark-field radiography.
(A) Transmission images T of an calcium oxalate (top left), uric acid stone (top right) and a pig kidney with both stones embedded (bottom), taken at
40 kVp tube voltage. (B) Corresponding dark-field images D. (C) Since both stones show opposite absorption and scatter characteristics, the dark-
field-to-transmission ratio D/T allows a simple visual differentiation of stone class and discrimination from surrounding kidney tissue in false color.
Notice that the uric acid stone appears radio-lucent, while yielding high contrast in the dark-field image.
www.nature.com/scientificreports
SCIENTIFIC REPORTS | 5 : 9527 | DOI: 10.1038/srep09527 4
Fig. 5 proves that an inversion of dark-field-to-transmission signal
when comparing uric acid and calcium oxalate stones is still existent,
even in the case of very high X-ray energies (E) and significant beam-
hardening. Future studies have to investigate, whether this does hold
true for the differentiation from mixed types of stone, considering
that m
eff
and E
eff
are proportional to E
23
and E
22
, respectively
15
.An
increase in mean energy namely implies the drawback of converging
and shrinking scatter plots offering a correspondingly reduced dia-
gnostic performance.
Further an animal model is envisioned to address concerns regard-
ing the impact of structures overlying the kidney in radiographic
imaging mode and discriminability of small and initially growing
calculi in-vivo
27
. Especially in the case of inhomogeneous structure
compositions or strongly absorbing tissue underlying the kidney
stone data normalization could be ambiguous, resulting in a limited
accuracy in the determination of the stone type classifier c. Finally, X-
ray dark-field radiography is susceptible to superposing renal stones;
hence an additional lateral radiogram of the abdomen or the imple-
mentation of advanced techniques such as tomosynthesis would be
necessary to support diagnostic reliability. Besides, current technical
limitations need to be challenged, such as the fabrication of bended,
large-field-of-view gratings with high aspect ratios, in order to sig-
nificantly reduce scan-time and secure a successful implementation
of non-invasive kidney stone assessment via X-ray dark-field radio-
graphy into clinical routine.
Methods
Study design. A broad range of different kidney stones comprising various stone sizes
was acquired in order to have a representative sample collective. Nevertheless, the
classification of kidney stones is in general complicated by the fact that most kidney
stones in practice are rarely composed of a single pure chemical material, but are
instead a mixture of various components with widely differing composition
28
. Thus in
this proof-of-principle study we focus on the differentiation of three classes of kidney
stones only: the pure uric acid stones, the pure calcium oxalate stones, and the mixed
stone class including composites of brushite, carbon apatite and struvite. The
different kidney stone types that occur in practice are summarized in table 1 by
compound name, chemical formula and mineralogical name.
The samples in our measurements were acquired by the Klini kum rechts der Isar,
Department of Radiology. Each patient had their renal stone(s) removed following
the common clinical practice with respect to their individual diagnosis and indica-
tion. Written and informed consent was obtained from all patients. Nine patients
were found with a mixture of Whewellite and Weddelite, four patients with uric acid
stones, and five patients with mixed stones types. From these patients, we obtained 68
oxalate stones, 10 uric acid stones, and 40 mixed stones, thus a total of 118 renal
calculi was accessed and imaged.
Figure 4
|
Statistical analyses adjudge superior diagnostic performance of kidney stones assessment via X-ray dark-field radiography. (A) Scatter plot
showing the ratios of
m versus
E for every image pixel of an exemplary uric acid (blue), calcium oxalate (red) and mixed types of kidney stones (yellow).
Each data-cloud was fitted by a linear regression as indicated by dashed lines, and its slope c used as a binary stone type classifier. (B) Box-whisker
plot showing the distributions of slope values c, determined for the uric acid, calcium oxalate and mixed stone collective. In addition to strongly differing
mean values (black line) no overlap in interquartile data was found (box). (C) Receiver operator characteristic analyses on the data presented in (B), show
a highly sensitive and specific differentiation of all three stone classes using X-ray dark-field radiography (Area under curve . 0.9).
www.nature.com/scientificreports
SCIENTIFIC REPORTS | 5 : 9527 | DOI: 10.1038/srep09527 5
The composition of each kidney stone was determined by Fourier-transform
infrared spectroscopy (FTIR)
29
, using a Spectrum 100 system by Perkin Elmer,
Beaconsfield, UK. The exact chemical composition of the calculus was determined by
comparing the recorded spectrum with tabulated spectra. Components of stones were
given in percentages, in which a concentration of more than 90% of one component
was regarded as pure. More details on the study design, patient selection routine,
patient examination and exact composition of the kidney stone collective can be
found in Ref. 5.
Statistical analysis. Statistical analysis was carried out using the statistical software R
and its library pROC
30,31
. Outliers were considered in the ROC analysis. Two thousand
stratified bootstrap samples were drawn for each ROC analysis to estimate 95%
confidence intervals for the area under the ROC curve.
Micro-CT setup. We performed micro-CT measurements at a commercial GE
VtomeX system, using a reflection tube with a voltage setting of 100 kVp at a current
of 10 mA and a voxel-size of 20 3 20 3 20 mm
3
. For the tomographic scan we took
1200 projections over 360 degrees, with an exposure time of 1 s each. Datasets were
reconstructed using a standard filtered backprojection, rendered in Volume Graphics
VGStudio MAX and analyzed visually. An example volume rendering can be seen in
Fig. 2.
X-ray dark-field interferometer. X-ray dark-field radiography was conducted with a
compact laboratory setup using a three-grating Talbot-Lau interferometer (effective
pixel-size of 85 3 85 mm
2
)
9
. The source is a Nonius FR 591 rotating anode tube with
molybdenum target, operated at 40 kVp and 70 mA. The beam-splitter grating G
1
is a
p/2-shifting binary phase-grating with a design energy of 27 keV. The interferometer
is built in an asymmetric geometry with periods of 10 mm, 4.8 mm and 3.24 mm for
G
0
, G
1
and G
2
, respectively. The setup length is 1570 mm, with inter-grating distances
of l 5 1060 mm between G
0
and G
1
and d 5 510 mm between G
1
and G
2
,
corresponding to the third Talbot order at the design energy. The samples are
positioned 80 mm downstream of G
1
. The contributions of absorption and scattering
in the projections are separated using a phase-stepping technique
8
, with 14 phase-
steps and 4 seconds exposure time each. Transmission and dark-field signals are then
obtained from the raw phase-stepping projection data using a Fourier analysis
approach. Exemplary transmission and dark-field projections can be seen in Fig. 3.
For the statistical analysis, we normalized the projections with respect to the sample
holder, in order to remove all contributions not originating from the kidney stone. All
stones were then segmented from the background by intensity thresholding and their
E and
m-values were tabulated for each single pixel for further analysis. For the
purpose of validating in-vivo feasibility of renal calculi assessment via dark-field
radiography, one calcium oxalate and one uric acid stone were additionally embedded
and subsequently measured within a fresh pig kidney (Fig. S1). We derived
comprehensive images of the pig kidney with a field-of-view of 12.8 3 12.8 cm
2
by
stitching 16 single projections taken with 5 phase-steps and 1 second exposure time
each (for a more detailed description, see Reference 25). Each projection was scanned
within 20 seconds (overall scan time of 320 seconds).
Imaging dose. The total air kerma of the pig kidney measurement was determined to
be 10.5 mGy (incident air kerma rate 2.1 mGy/s) with a Dosimax plus/RQX-detector
system. To give a rough estimate of the effective dose deposited in the kidney, we
calculated the mean glandular dose of a 100% glandular breast tissue equivalent
(underlying consideration are based on the fact that glandular breast and kidney
tissue yield similar mass density coefficients)
32
. We obtained a mean dose of 5.2 mGy,
by multiplying the total incident air kerma with a Monte-Carlo based conversion
factor of 0.56 and a correction factor accounting for 100% glandular breast tissue of
0.9. Values arise from a half-value layer (Al) of 0.8 mm and a kidney thickness of
3cm
33,34
.
Potential clinical implementation. In a first step, the radiologist marks the renal
stone in the digital radiogram. Since stone edges yield excessive scatter corresponding
regions are excluded from the data analysis. Afterwards the region surrounding the
kidney stone is automatically selected in the radiographs and the respective mean
signal values in both the absorption and dark-field channel calculated, by which the
renal stone is subsequently normalized. This is feasible since the kidney stone is
distinctively thinner than the patient. In a next step, the scatter plot is generated from
Figure 5
|
In-Vivo transferability study at 100 kVp tube voltage using a preliminary abdomen phantom. (A) High-energy transmission image T of a pig
kidney, with manually embedded uric acid and calcium oxalate stones, placed within a 11 cm water-bath. (B) Corresponding dark-field image D,
revealing concentric growth rings of the uric acid stone. (C) The dark-field-to-transmission D/T signal enables a clear differentiation of uric acid (blue)
and calcium oxalate (red), consistent with the 40 kVp measurements shown in Fig. 3.
Table 1
|
Overview of the different types of kidney stones relevant for this study, with chemical name, formula and mineralogical name.
Stoichiometry adapted from Ref. 28
compound name chemical formula mineralogical name
Calciumoxalate-Monohydrate CaC
2
O
4
? H
2
O Whewellite
Calciumoxalate-Dihydrate CaC
2
O
4
? 2H
2
O Weddelite
Uric acid C
5
H
4
N
4
O
3
Uricite
Uric acid dihydrate C
5
H
4
N
4
O
3
? 2H
2
O none
Ammonium urate (NH
4
)C
5
H
3
N
4
O
3
none
Sodium urate monohydrate NaC
5
H
3
N
4
O
3
? H
2
O none
Calcium hydrogenate phosphate CaH(PO
4
) ? 2H
2
O Brushite
Carbonate hydroxylapatite Ca
5
(PO
4
, CO
3
)
3
(OH) Apatite/Dahllite
Magnesium ammonium phosphate Mg(NH
4
)(PO
4
) ? 6H
2
O Struvite
Cystine C
6
H
12
N
2
O
4
S
2
none
www.nature.com/scientificreports
SCIENTIFIC REPORTS | 5 : 9527 | DOI: 10.1038/srep09527 6
E and
m and the binary stone classifier c extracted as the mean slope of the data points,
by which the respective stone type can be determined from a look-up table. Finally, by
reading the maximum ordinate of the scatter plot (
m) the thickness of kidney stone in
beam direction can be approximated.
1. Brown, J. Diagnostic and treatment patterns for renal colic in US Emergency
Departments. Int. Urol. Nephrol. 38,87–92 (2006).
2. Coe, F. L., Evan, A. & Worcester, E. Kidney stone disease. J. Clin. Invest. 115,
2598–2608 (2005).
3. Dalla Palma, L., Pozzi-Mucelli, R. & Stacul, F. Present-day imaging of patients
with renal colic. Eur. Radiol. 11,4–17 (2001).
4. Bellin, M. F. et al. Helical CT evaluation of the chemical composition of urinary
tract calculi with a discriminant analysis of CT-attenuat ion values and density.
Eur. Radiol. 14, 2134–2140 (2004).
5. Eiber, M. et al. Targeted dual-energy single-source CT for characterisation of
urinary calculi: experimental and clinical experience. Eur. Radiol. 22, 251–258
(2012).
6. Davis, T. J., Gao, D., Gureyev, T. E., Stevenson, A. W. & Wilkins, S. W. Phase-
contrast imaging of weakly absorbing materials using hard X-rays. Nature 373,
595–598 (1995).
7. Wilkins, S. W., Gureyev, T. E., Gao, D., Pogany, A. & Stevenson, A. W. Phase-
contrast imaging using polychromatic hard X-rays. Nature 384, 335–338 (1996).
8. Weitkamp, T. et al. X-ray phase imaging with a grating interferometer. Opt.
Express 13, 6296–6304 (2005).
9. Pfeiffer, F. et al. Phase retrieval and differential phase-contrast imaging with low-
brilliance X-ray sources. Nat. Phys. 2, 258–261 (2006).
10. Bravin, A., Coan, P. & Suortti, P. X-ray phase-contrast imaging: from pre-clinical
applications towards clinics. Phys. Med. Biol. 58,1–35 (2013).
11. Pfeiffer, F. et al. Hard-X-ray dark-field imaging using a grating interferometer.
Nat. Mater. 7, 134–137 (2008).
12. Hauser, N. et al. A Study on Mastectomy Samples to Evaluate Breast Imaging
Quality and Potential Clinical Relevance of Differential Phase Contrast
Mammography. Invest. Radiol. 49, 131–137 (2014).
13. Schleede, S. et al. Emphysema diagnosis using X-ray dark-field imaging at a laser-
driven compact synchrotron light source. Proc. Nat. Acad. Sci. 109, 17880–17885
(2012).
14. Michel, T. et al. On a dark-field signal generated by micrometer-sized
calcifications in phase-contrast mammography. Phys. Med. Biol. 58, 2713–2732
(2013).
15. Yashiro, W., Terui, Y., Kawabata, K. & Momose, A. On the origin of visibility
contrast in X-ray Talbot interferometry. Opt. Express 18, 16890–16900 (2010).
16. Lynch, S. K. et al. Interpretation of dark-field contrast and particle-size selectivity
in grating interferometers. Appl. Optics 50, 4310–4319 (2011).
17. Bech, M. et al. Quantitative X-ray dark-field computed tomography. Phys. Med.
Biol. 55, 18, 5529–5539 (2010).
18. Tomasz, K., Podgrski, P., Guziski, M., Czarnecka, A. & Tupikowski, K. Novel
Clinical Applications of Dual Energy Computed Tomography. Adv. Clin. Exp.
Med. 21, 831–841 (2012).
19. Grases, F., Costa-Bauz, A. & Garcia-Ferragut, L. Biopathological crystallization: a
general view about the mechanisms of renal stone formation. Adv. Colloid
Interface Sci. 74, 169–194 (1998).
20. Grases, F., Costa-Bauz, A., Ramis, M., Montesinos, V. & Conte, A. Simple
classification of renal calculi closely related to their micromorphology and
etiology. Clin. Chim. Acta 322,29–36 (2002).
21. Stoller, M. & Meng, M. Urinary Stone Disease: The Practical Guide to Medical and
Surgical Management. Springer (2007).
22. Smith, R., Levine, J. & Rosenfeld, A. Helical CT of urinary tract stones.
Epidemiology, origin, pathophysiology, diagnosis, and management. Radiol. Clin.
N. Am. 35, 911–52 (1999).
23. Smith, R. & Varanelli, M. Diagnosis and Management of Acute Ureterolithiasis:
CT Is Truth. Radiol. Clin. N. Am. 175,3–6 (2000).
24. Mettler, A., Huda, W., Yoshizumi, T. & Mahesh, M. Effective Doses in Radiology
and Diagnostic Nuclear Medicine: A Catalog. Radiology 248
, 254–263 (2008).
25. Scherer, K. et al. Bi-Directional X-ray Phase-Contrast Mammography. Plos One 9,
93502 (2014).
26. Koehler, T., Engel, K. & Roessl, E. Noise Properties of Grating-Based X-ray Phase
Contrast Computed Tomography. Med. Phys. 38, 106–116 (2011).
27. Tapfer, A. et al. Experimental results from a preclinical X-ray phase-contrast CT
scanner. Proc. Nat. Acad. Sci. 109 15691–15696 (2012).
28. Stoller, M. & Meng, M. Urinary Stone Disease: The Practical Guide to Medical and
Surgical Management Chapter 5: Structure and Compositional Analysis of Kidney
Stones, Ian Mandel and Neil Mandel. Humana Press (2007).
29. Krafft, C., Steiner, G., Beleites, C. & Salzer, R. Disease recognition by infrared and
Raman spectroscopy. Bio. Med. Phys. Biomed. 2,13– 28 (2009).
30. R Core Team. R: A language and environment for statistical computing. R
Foundation for Statistical Computing (2014).
31. Robin, X., Turck, N., Hainard, A., Tiberti, N. & Lisacek, F. pROC: an open-source
package for R and S1 to analyze and compare ROC curves. BMC Bioinformatics
12, 77 (2011).
32. Woodard, H. & White, D. The Composition of Body Tissues. Br. J. Radiol. 59,
12091218 (1986).
33. Dance , D. R. Monte carlo calculation of conversion factors for the estimation of
mean glandular breast dose. Phys. Med. Biol. 35 , 1211–1219 (1990).
34. Dance, D. R. et al. Additional factors for the estimation of mean glandular breast
dose using the UK mammography dosimetry protocol. Phys. Med. Biol. 45,3225
(2000).
Acknowledgments
We thank Richard Clare, Irene Zanette, Susanne Grandl, Karin Hellerhoff and Klaus
Achterhold for fruitful discussions and revision of the manuscript as well as Friedrich Prade
and Florian Schaff for helping with the high-energy measurements.
Author contributions
K.S., E.B., K.W. and M.W. performed the measurements. K.S., P.N., E.R. and F.P. conceived
the study. M.E., A.F., M.S. and H.S. provided and analyzed the sample collective. B.H.
performed the statistical analysis. K.S., M.C., J.H. and A.F. wrote the manuscript.
Additional information
Funding: We acknowledge financial support through the DFG Cluster of Excellence
Munich-Centre for Advanced Photonics (MAP), the DFG Gottfried Wilhelm Leibniz
program and the European Research Council (ERC, FP7, StG 240142). Part of this work was
carried out with the support of the Karlsruhe Nano Micro Facility (KNMF, www.kit.edu/
knmf), a Helmholtz Research Infrastructure at Karlsruhe Institute of Technology (KIT).
Supplementary information accompanies this paper at http://www.nature.com/
scientificreports
Competing financial interests: The authors declare no competing financial interests.
How to cite this article: Scherer, K. et al. Non-invasive Differentiation of Kidney Stone
Types using X-ray Dark-Field Radiography. Sci. Rep. 5, 9527; DOI:10.1038/srep09527
(2015).
This work is licensed under a Creative Commons Attribution 4.0 International
License. The images or other third party material in this article are included in the
article’s Creative Commons license, unless indicated otherwise in the credit line; if
the material is not included under the Creative Commons license, users will need
to obtain permission from the license holder in order to reproduce the material. To
view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/
www.nature.com/scientificreports
SCIENTIFIC REPORTS | 5 : 9527 | DOI: 10.1038/srep09527 7
