See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/45629831

Measurement of the linear attenuation coefficients of breast tissues by

synchrotron radiation computed tomography

Article  in  Physics in Medicine and Biology · September 2010

DOI: 10.1088/0031-9155/55/17/008 · Source: PubMed

CITATIONS READS

71 1,495

11 authors, including:

Rongchang Chen Renata Longo

Shanghai Institute of Applied Physics University of Trieste
61 PUBLICATIONS   757 CITATIONS    227 PUBLICATIONS   4,785 CITATIONS   

SEE PROFILE SEE PROFILE

Luigi Rigon Diego Dreossi

University of Trieste Sincrotrone Trieste S.C.p.A.
169 PUBLICATIONS   3,427 CITATIONS    141 PUBLICATIONS   2,924 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Quantification of the quality improvement of mammography images in phase contrast modality using synchrotron radiation beams as compared with conventional
and digital detectors mammography using X-rays beams. View project

x-ray imaging View project

All content following this page was uploaded by Luigi Rigon on 13 August 2014.

The user has requested enhancement of the downloaded file.

Home Search Collections Journals About Contact us My IOPscience

Measurement of the linear attenuation coefficients of breast tissues by synchrotron radiation

computed tomography

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

2010 Phys. Med. Biol. 55 4993

(http://iopscience.iop.org/0031-9155/55/17/008)

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

Download details:
IP Address: 159.226.222.25
The article was downloaded on 12/08/2010 at 02:16

Please note that terms and conditions apply.

IOP PUBLISHING PHYSICS IN MEDICINE AND BIOLOGY

Phys. Med. Biol. 55 (2010) 4993–5005 doi:10.1088/0031-9155/55/17/008

Measurement of the linear attenuation coefficients of

breast tissues by synchrotron radiation computed
tomography

R C Chen1,2,3 , R Longo2,3 , L Rigon3 , F Zanconati4 , A De Pellegrin4 ,

F Arfelli2,3 , D Dreossi3,5 , R-H Menk3,5 , E Vallazza3 , T Q Xiao1 and
E Castelli2,3
1 Shanghai Institute of Applied Physics, Chinese Academy of Sciences, Shanghai,

People’s Republic of China

2 Department of Physics, University of Trieste, Trieste, Italy
3 INFN, Sezione di Trieste, Trieste, Italy
4 Department of Pathologic Anatomy, University of Trieste, Trieste, Italy
5 Sincrotrone Trieste SCpA, Trieste, Italy

E-mail: rongchang.chen@gmail.com

Received 11 April 2010, in final form 24 June 2010

Published 11 August 2010
Online at stacks.iop.org/PMB/55/4993

Abstract
The measurement of the linear attenuation coefficients of breast tissues is
of fundamental importance in the field of breast x-ray diagnostic imaging.
Different groups have evaluated the linear attenuation coefficients of breast
tissues by carrying out direct attenuation measurements in which the
specimens were thin and selected as homogeneous as possible. Here,
we use monochromatic and high-intensity synchrotron radiation computed
tomography (SR CT) to evaluate the linear attenuation coefficients of surgical
breast tissues in the energy range from 15 to 26.5 keV. X-ray detection is
performed by a custom digital silicon micro-strip device, developed in the
framework of the PICASSO INFN experiment. Twenty-three human surgical
breast samples were selected for SR CT and histological study. Six of them
underwent CT, both as fresh tissue and after formalin fixation, while the
remaining 17 were imaged only as formalin-fixed tissues. Our results for fat
and fibrous tissues are in good agreement with the published values. However,
in contrast to the published data, our measurements show no significant
differences between fibrous and tumor tissues. Moreover, our results for fresh
and formalin-fixed tissues demonstrate a reduction of the linear attenuation
coefficient for fibrous and tumor tissues after fixation.

0031-9155/10/174993+13$30.00 © 2010 Institute of Physics and Engineering in Medicine Printed in the UK 4993
4994 R C Chen et al

1. Introduction

Computed tomography (CT) using monochromatic x-rays allows reconstruction of linear

attenuation coefficient maps of the investigated samples (Nuzzo et al 2002). In 1997,
Dilmanian explored the potential for clinical research of CT with synchrotron radiation (SR)
monochromatic x-rays using the preclinical multiple energy CT (MECT) system at the National
Synchrotron Light Source (NSLS, Brookhaven, NY, USA) (Dilmanian et al 1997). Their
results show that MECT immediately above the iodine K-edge has a twofold to threefold
advantage over conventional CT. Similar to other SR laboratories, such as the European
Synchrotron Radiation Facility (ESRF, Grenoble, France) and the Shanghai Synchrotron
Radiation Facility (SSRF, Shanghai, China), a CT setup has been developed at the SYRMEP
(SYnchrotron Radiation for MEdical Physics) beamline at the ELETTRA SR facility in Trieste
(Italy), in order to exploit the advantages of SR CT. Firstly, the monochromatic beam yields
reconstructed CT images free of beam hardening artifacts (Tafforeau et al 2006), and allows
image calibration in order to evaluate linear attenuation coefficient distribution within the
sample. Secondly, due to the small source dimensions and limited divergence of SR, spatial
resolution in SR CT is limited mainly by the detector (Baruchel et al 2006). Thirdly, the high
flux available at SR sources allows rapid CT data acquisition with high spatial resolution,
resulting in precise mapping of sample internal structures (Trtik et al 2007). Fourthly, the
high degree of (lateral) coherence of the SR sources allows implementation of phase-sensitive
imaging techniques (Cloetens et al 1996).
At the SYRMEP beamline, mammography with SR, which utilizes the free propagation
phase-contrast imaging technique, is under clinical evaluation (Dreossi et al 2008), and the
knowledge of the linear attenuation coefficients of glandular tissue, tumor tissue and fat
tissue in the mammographic energy range is useful in both model development and energy
optimization. Hammerstein and co-workers evaluated the linear attenuation coefficients of
fat and fibrous breast tissues for the first time in the energy range of 10 to 50 keV, by
determining the chemical composition of normal breast tissues (Hammerstein et al 1979). In
their comprehensive work, the measurement of linear attenuation coefficients is an intermediate
step of their mammography absorbed-dose calculation. Later, Johns and Yaffe investigated the
linear attenuation coefficients of breast tissues using a high-purity germanium spectroscopy
system and a beam of 120 kV constant potential x-rays in the energy range between 18 and
110 keV. They measured the energy spectra of two bremsstrahlung beams: one transmitted
through the tissue and a second, or reference, beam transmitted through plastics identical to
an empty specimen holder. They observed that tumor tissues have a slightly higher linear
attenuation coefficient than fibrous tissues, which is a normal mammary gland, for energies
lower than 30 keV, and the differences tend to increase with decreasing energy (Johns and
Yaffe 1987). Carroll and co-authors measured the linear attenuation coefficients of fat, fibrous
and various malignant breast tumor tissues utilizing SR monochromatic x-ray in the energy
range of 14 to 18 keV. Their results showed a wide overlap of the linear attenuation coefficient
between fibrous and cancer tissues, even if the mean value of cancer tissue is larger than that of
fibrous tissue (Carroll et al 1994). Both Johns and Yaffe and Carroll and colleagues conducted
their studies by direct attenuation measurements of thin samples selected as homogeneous as
possible.
Different groups have already investigated breast tumors by means of SR (Arfelli et al
1998, Chapman et al 1998, Takeda et al 1998, Pisano 2000, Fiedler et al 2004, Kao et al 2009).
In particular, the mammography program at the SYRMEP beamline at ELETTRA (Arfelli et al
2000, Dreossi et al 2008) conducted a feasibility study for breast SR CT using high-efficiency
silicon pixel detectors, read out by a single-photon counting electronic chain (Pani et al 2004).
Measurement of the linear attenuation coefficients of breast tissues by SR CT 4995

A promising work in breast SR CT was published by Bravin and co-authors using the x-ray
diffraction-enhanced imaging (DEI)-CT technique at ESRF (Bravin et al 2007, Keyrilainen
et al 2008). The DEI-CT images of tumor-bearing breast tissue were compared with the
images of histological examination, clinical screen-film mammograms and clinical CT scans,
showing that large contrast enhancement is achieved by using the DEI-CT method. These
studies are based on formalin-fixed tissues, as are a large number of other papers (Takeda
et al 2004); however, to our knowledge no data are available on the effects of fixation on the
attenuation properties of breast tissues.
In this work, we acquired SR CT data of both fresh and formalin-fixed surgical breast
samples in the energy range between 15 and 26.5 keV using monochromatic and high-flux
SR in combination with a silicon edge-on micro-strip detector developed in the framework of
the PICASSO (Phase Imaging for Clinical Application with Silicon detector and Synchrotron
radiation) experiment funded by Istituto Nazionale di Fisica Nucleare (INFN). Reconstructed
SR CT images have been calibrated in order to obtain quantitative linear attenuation coefficient
maps; the linear attenuation coefficients of surgical breast fat, fibrous and tumor tissues have
been investigated. Moreover, for the first time to our knowledge, the linear attenuation
coefficients of fresh and formalin-fixed breast tissues have been compared.

2. Materials and methods

The work described here has been carried out following the Directive 2004/23/EC of the
European Parliament and of the Council of 31 March 2004 on setting standards of quality and
safety for the donation, procurement, testing, processing, preservation, storage and distribution
of human tissues. According to the standard procedures of the University Hospital of Trieste
(Italy), written informed consent had been obtained from all patients/donors.

2.1. The SYRMEP beamline

The experiments were performed at the SYRMEP beamline at the ELETTRA SR facility in
Trieste, Italy. The SYRMEP beamline aims at evaluating the effectiveness of synchrotron-
based techniques in medical radiology with particular interest in mammography and, more in
general, in the imaging of biological and biomedical samples, either in planar or CT modalities.
Since 1996, studies have been carried out on mammographic phantoms and in vitro breast
tissue samples. In the frame of a worldwide unique clinical phase-contrast SR mammography
program (Castelli et al 2007), in vivo exams have been carried out since 2006.
Figure 1 shows a schematic layout of the SYRMEP beamline. The radiation source
is one of the bending magnets of an ELETTRA storage ring. A double Si [1 1 1] crystal
monochromator provides photon energy ranging from 8.5 to 35 keV with a 0.2% energy
bandwidth. When ELETTRA is operated at 2.0 GeV (as it was during the experiments
described herein) the critical energy is 3.2 keV, which implies a negligible contribution of the
higher harmonics for working energies of 15 keV and above. Moreover, this contribution can
be further diminished by suitably de-tuning the second crystal of the monochromator, yielding
a practically monochromatic beam. At the experimental area, located approximately 23 m
from the source, the beam cross-section is 210 mm (H) × 4 mm (V), and the flux of mono-
energetic photons measured at the sample position at 17 keV is around 108 photons mm−2 s−1.
A precision rotational stage is installed to perform CT data acquisitions in discrete angular
steps while the detector is kept stationary in front of the beam (Arfelli et al 1996, Abrami
et al 2004, Pani et al 2004).
4996 R C Chen et al

Figure 1. Schematic layout of the SYRMEP beamline at ELETTRA.

2.2. The PICASSO detector

In this experiment, a prototype of the PICASSO detector was used for CT data acquisition. The
PICASSO project is developing the fourth generation of silicon micro-strip detectors for the
SYRMEP beamline that can acquire low-dose mammographic images in a few seconds with
high spatial and contrast resolution. The PICASSO detector is a 50 μm pitched linear array
silicon micro-strip detector, operated in the ‘edge-on’ configuration. Based on Hamamatsu
silicon sensors (Hamamatsu Photonics K.K., Hamamatsu City, Japan), it matches the laminar
geometry of the beam providing a pixel array, where the aperture of each pixel is determined
by the strip pitch (50 μm) and the detector thickness (300 μm) (Rigon et al 2009).
Moreover, the detector is operated in single-photon counting mode, and it is read out by
high-rate electronics based on the Mythen-II application-specific integrated circuit (ASIC),
developed by the Paul Scherrer Institut (PSI) detector group for powder diffraction experiments
at the Swiss Light Source (Mozzanica et al 2009, Bergamaschi et al 2009). Each pixel is
wire-bonded to one channel of the Mythen-II ASICs, and its signal is processed individually
throughout the read-out electronics. On one hand, the single-photon counting approach allows
us to maximize the contrast resolution (preserving the quantum nature of the information
carried by the photon beam) and to overcome the limitations in the signal dynamic range,
which are typical of charge-integration detectors such as charge-coupled devices (CCDs) and
flat panels (Vallazza et al 2008, Rigon et al 2009). On the other hand, one possible drawback
of the single-photon counting approach is the pile-up of single photon pulses at high fluxes.
However, as demonstrated elsewhere (Rigon et al 2009), the PICASSO detector can handle
fluxes as high as 1.2 × 106 photons/pixel/s with only about 10% losses due to pile-up.
Anyway, since saturation could be highly detrimental in this study, CT scans were performed
with fluxes about 8 × 104 photons/pixel/s in the background and 4 × 104 photons/pixel/s
behind the samples, so that losses due to pile-up were about 0.8% and 0.4%, respectively, and
thus negligible.

2.3. Breast tissue samples

Twenty-three human surgical breast samples were selected for SR CT and histological study.
Six of them underwent CT both as fresh tissue and after formalin fixation, while the remaining
17 were imaged only as formalin-fixed tissues. The samples were prepared from specimens of
quadrantectomy or total mastectomy and were derived from surgical material sent to Pathology
Unit of University Hospital of Trieste (Italy) according to local guidelines for histological
examination. The studied tumors were all ductal carcinoma but two, one apocrine carcinoma
Measurement of the linear attenuation coefficients of breast tissues by SR CT 4997

and one angiosarcoma, were probably induced by radiation therapy. The tumor grading was G2
or G3; micro-calcifications were present in three samples. A large necrosis area was present
in two tumors, a moderate necrosis was reported in two cases while necrosis was absent in the
remaining samples.
During SR CT, the samples, which were 1 cm thick and possessed diameters between 1 and
3 cm, were sealed in cylindrical (Ø 3 cm) polyethylene containers. Having acquired SR CT data
of fresh breast samples, they were immediately fixed in 4% neutral buffered formalin for more
than 24 h, and subsequently a second SR CT was performed. Prior to all, SR CT’s planar scout
radiographs had been acquired in order to determine the appropriate vertical position within
the breast samples. Histological sections were obtained at the same vertical position at which
the SR CT images had been acquired before. The histological material was prepared according
to routine histological protocols (dehydration in alcohol, xylene clarification, imbibition and
inclusion in paraffin); specimens obtained were cut with microtome into 4–6 μm thick slices
and stained with hematoxylin–eosin (H&E). The acquisition of photographic histological
sections was performed with an Olympus BX-41 microscope, equipped with a 2.5× objective.
All data presented herein have been collected during three different experimental runs
carried out in intervals of 3 to 4 months.

2.4. Data acquisition and reconstruction

In order to measure the value of the linear attenuation coefficient directly from the reconstructed
SR CT images, a thorough calibration is needed. Therefore, three reference materials, pure
liquid water, ethanol and glycerol, had been used to calibrate the voxel values of the SR CT
images. These liquids and thus homogenous materials had been selected because their linear
attenuation coefficients are well known and can be obtained from many databases. Moreover,
the values of their linear attenuation coefficients cover the typical range observed within breast
tissues.
We acquired SR CT data for each sample, together with the three calibration materials,
at 15, 17, 19, 23, 26 and 26.5 keV with sufficient angular sampling within 180◦ (typically
1200–3000 projections, depending on the sample size). The use of 26.5 keV besides 26 keV is
merely due to a strong artifact detected in the beam at 26 keV during the second experimental
run. The exposure time per projection was 4 s for 15 keV and 1 s for higher energies. Our
study aimed to obtain the best image quality, regardless of dose constraints. In fact, the
air entrance dose was in the range of 0.2 to 6 mGy per projection, substantially higher than
that delivered in other breast CT experiments (Pani et al 2004, Keyrilainen et al 2008). In
order to avoid free propagation phase-contrast effects, the sample-to-detector distance was
set to 20 cm, which is small enough to neglect phase contrast in the PICASSO detector with
50 μm resolution (Olivo and Speller 2006). After the dark-field and flat-field corrections, the
SR CT images were reconstructed using a standard filtered back-projection algorithm using a
Gen-Hamming filter.

2.5. Calibration method and linear attenuation coefficients measurement

The XMuDat (version 1.0.1) x-ray database was used to calibrate the voxel values of the SR
CT images (XMuDat 1998). Using the XMuDat database the theoretical linear attenuation
coefficients of the three calibration materials have been calculated for different energies and
the voxel values measured from the SR CT images have been calibrated against the calculated
theoretical linear attenuation coefficients. Within the SR CT images, a region of interest (ROI)
4998 R C Chen et al

Figure 2. Experimental voxel values plotted versus theoretical linear attenuation coefficients for the
three calibration materials. A linear fit provides the calibration equation. The standard deviations
of the measured mean values are 5.0 × 10−7, 8.5 × 10−7 and 8.8 × 10−7 (a.u.), respectively, for
ethanol, water and glycerol. Thus the error bars are smaller than the markers.

has been assigned to each of the calibration materials comprising a total of 6 × 103 voxels and
encompassing the specific mean values and standard deviations of the materials.
Depicted in figure 2 is the functional dependence of the measured voxel values derived
from the SR CT images versus the theoretical linear attenuation coefficients for the three
calibration materials at 23 keV. The straight line corresponds to a linear fit, yielding the
calibration equation with an R-squared value of 1, which means an excellent alignment of the
three points.
The histological images, which were first investigated by a medical doctor who has more
than 15 years of experience in breast pathology, act as references to determine different breast
tissues in the SR CT images. Thus, the linear attenuation coefficients of breast tissues were
measured from the calibrated SR CT images. For each type of tissue and breast sample, five
different ROIs (in the order of 104 pixels) were analyzed yielding the mean value and standard
deviation that were then used in data analysis.

3. Results and discussion

3.1. Comparison between histological sections and SR CT images

Out of the twenty-three formalin-fixed human breast samples, three exemplary cases are
discussed in the following sections.
Measurement of the linear attenuation coefficients of breast tissues by SR CT 4999

(a) (b)

Figure 3. Breast sample 15955. (a) Histological image; (b) SR CT image at 19 keV.

(a) (b)

Figure 4. Breast sample 16037. (a) Histological image; (b) SR CT image at 26 keV. The blue pen
marker in the histological section identifies a knife cut of the tissue visible in the SR CT which
was used as a fiducial marker.

(a) (b)

Figure 5. Breast sample 16085. (a) Histological image; (b) SR CT image at 23 keV.
5000 R C Chen et al

3.1.1. Sample 15955. This sample was obtained from an 80 years old female who underwent
mastectomy. The histological diagnosis was a poorly differentiated ductal invasive carcinoma
staged as pT2N1a according to the Tumor Nodes and Metastases (TNM) classification system
(Greene et al 2002).
Shown in figure 3(a) is the image of an H&E-stained 5 μm thick histological section. The
field of view is about 5 cm × 3 cm. A large tumor embedded in a background of fatty tissue is
prominent. Moreover, normal breast gland (fibrous tissue and ducts), skin and another small
malignant nodule (bottom-right) are evident.
In the SR CT image (figure 3(b)), acquired at a photon energy of 19 keV, the spiculated
structure of the tumor is more pronounced than in the histological image. This is due to the
fact that the CT reconstruction is averaged over a 300 μm thick slice, while the histology cut
is only 5 μm. In the SR CT image, fat tissue features lower voxel values, and thus lower linear
absorption coefficients, than fibrous and tumor tissues, which are not distinguishable. Ring
artifacts are visible, stemming from beam instability during data acquisition; however areas
containing such artifacts were considered in the data analysis.

3.1.2. Sample 16037. This sample was obtained from a quadrantectomy performed in a
57 years old female. The histological exam demonstrated a poorly differentiated ductal
invasive carcinoma staged as pT2N0 according to the TNM classification system.
The histological section (figure 4(a)) shows a tumor in fatty tissue: the central part of the
lesion is extensively necrotic probably due to rapid growth and subsequent lack of nutrition.
Only at the periphery of the tumor, vital cells and normal breast dense tissue are recognizable.
Again the SR CT image (figure 4(b)), acquired at a photon energy of 26 keV, reveals that
fat tissue encompasses lower voxel values, and thus lower linear absorption coefficients, than
fibrous and tumor tissues, which are not distinguishable. As in all complementary imaging
modalities, the morphological differences are due to the different sample treatments and
environments.

3.1.3. Sample 16085. This sample was prepared from mastectomy performed in a 37 years
old female. The case was concluded as a poorly differentiated ductal invasive carcinoma staged
as pT2N3a according to the TNM classification system. The histological stained section and
the tomographic images acquired at a photon energy of 23 keV are shown in figure 5.
The histological section (figure 5(a)) contains adipose subcutaneous tissue, fibrous
mammary gland, fat and a solid malignant tumor. The SR CT image (figure 5(b)) reproduces
the same morphology of the histological section. Also at this photon energy, fibrous and
tumor tissues are not distinguishable; however, due to their high x-ray absorption the scattered
micro-calcifications are well prominent. These were also detectable in the histological section,
but only at a higher magnification in the optical microscope.

3.2. Linear attenuation coefficients of breast tissues

At first the SR CT images of the breast sample were calibrated on a pixel-to-pixel basis
using the linear calibration equation shown in figure 2. In this fashion each voxel value
represents the linear absorption coefficient (in cm−1) at a specific location. According to the
histological sections, five ROIs of different tissue types were assigned in the calibrated SR
CT images for each of those mean values, and thus the mean linear absorption coefficient and
standard deviations were calculated. Special care was taken to exclude areas containing ring
artifacts which were stemming from beam instability during data acquisition rather than using
algorithms for the removal of ring artifacts (Munch et al 2009) that might alter the mean values
Measurement of the linear attenuation coefficients of breast tissues by SR CT 5001

Table 1. Linear attenuation coefficient of breast tissues.

Linear attenuation coefficient (cm−1)

Tissue type 15 keV No 17 keV No 19 keV No 23 keV No 26 keV No 26.5 keV No

Fat Minimum 0.781 0.579 0.451 0.340 0.300 0.294

Mean 0.794 3 0.594 13 0.488 18 0.357 17 0.303 8 0.299 9
Maximum 0.811 0.606 0.513 0.373 0.311 0.308
σ – 0.009 0.017 0.010 0.007 0.006
Fibrous Minimum 1.617 1.159 0.865 0.604 0.473 0.470
Mean 1.659 2 1.193 9 0.918 13 0.612 12 0.480 6 0.473 6
Maximum 1.700 1.226 0.946 0.622 0.486 0.479
σ – 0.019 0.021 0.006 0.005 0.004
Tumor Minimum 1.564 1.149 0.866 0.598 0.476 0.457
Mean 1.608 2 1.183 9 0.920 14 0.615 14 0.483 7 0.476 7
Maximum 1.653 1.223 1.005 0.660 0.492 0.507
σ – 0.020 0.034 0.017 0.006 0.016
No: number of samples investigated.

of voxel values. For the same reason micro-calcifications, if present, were never included in
the selected ROIs.
Table 1 reports the mean linear attenuation coefficients for the three types of breast tissue
at different energies using the method described above. The linear attenuation coefficients of
the apocrine carcinoma and the angiosarcoma are in the range of the values obtained for ductal
carcinoma samples, and therefore all the tumor values have been processed as part of the same
group. As is well known, the linear attenuation coefficients of fat tissues are obviously smaller
than those of fibrous and tumor tissues, and the difference decreases with increasing energy:
for instance, the value of the fat-to-fibrous ratio increases from 0.48 at 15 keV to 0.63 at
26.5 keV. However, there are no significant differences in mean values between fibrous and
tumor tissues. The necrotic regions of the two samples with a large central necrotic area
are well defined in the histological images, but are not differentiated from the surrounding
growing region of the tumors themselves in the CT images (see figure 4). Table 1 shows that
the mean value of tumor tissue is smaller than that of the fibrous tissue at 15 and 17 keV, while
the reverse is observed from 19 to 26.5 keV. However, it should be noticed that the distribution
of values of tumor and fibrous tissues are overlapping.
Only a small number of samples were acquired at 15 keV, because at this energy the sample
is much more absorbing than at higher energies. Thus, the exposure time necessary to acquire
adequate statistics increases sharply, requiring about 5 h for one SR CT data acquisition,
compared to less than 1 h at higher energies. In turn, the long acquisition time exposed the
measurement to the chance of being spoiled by possible beam instability.
Table 1 also presents the overall number of samples investigated for each type of breast
tissue and for each energy. A different number of samples of each tissue were investigated,
since a sample may not contain all different kinds of breast tissues. Furthermore, artifacts
appear in few SR CT images, hindering the selection of a reasonably large area where the
linear attenuation coefficients could be measured. In this case, these data were excluded from
the analysis. The standard deviations are not presented for the 15 keV measurements due to
the small number of data at this energy.
Figure 6 shows a box-plot of the distribution of the linear attenuation coefficients for fat,
fibrous and tumor tissues for all the samples at different x-ray energies. It is not surprising
5002 R C Chen et al

Figure 6. The box-plot of linear attenuation coefficients for fat, fibrous and tumor tissues, shown
separately at each energy, of all the samples at different energies. (Each marker shows a box
encased by two outer lines known as whiskers. The box represents the middle 50% of the data
sample; the single line inside the box represents the median of the entire sample. The remaining
50% of the sample is contained between the box and the whiskers; the upper and lower lines
denote, respectively, the highest and smallest values of the entire sample.)

to see that tumor tissue features a wider distribution than fibrous tissue, because in this study
we investigated tumors of different tissue compositions (cellularity, fibrosis, necrosis, etc). In
contrast the distribution for the fat is rationally stable for all different energies. Data obtained
at 15 keV were not included in the box-plot, because the number of samples measured (two for
fibrous and tumor tissues, three for fat tissue) was not sufficient to allow statistical treatment.

3.3. The effect of formalin on the linear attenuation coefficients

As mentioned above, our study also encompassed the evaluation of the effect of the fixation
procedure on the linear attenuation coefficients obtained from the SR CT images. For these
data, sets of six breast samples were acquired prior to and after fixation. The data analysis
procedure was similar to the method described above. In order to prevent damage to the tissue
structures of fresh breast tissues, we limited this aspect of the study to three x-ray energies.
Table 2 demonstrates the measured values at three different x-ray energies. It turns out
that within the error bars the linear attenuation coefficient for fat tissue is not affected by
the fixation procedure. For tumor and fibrous tissues, however, a significant decrease of
the linear absorption coefficient for the fixed samples is observed at least for the three x-ray
energies involved in this aspect of the study. This difference is related to the mechanisms of
formaldehyde binding to fibrous and tumor tissues that do not apply to fat tissue.
Measurement of the linear attenuation coefficients of breast tissues by SR CT 5003

Figure 7. Comparison of the measured linear attenuation coefficients with the published data.

Table 2. Linear attenuation coefficient comparison between fresh and formalin-fixed breast tissues.

Linear attenuation coefficient (cm−1)

17 keV 19 keV 23 keV
Tissue type Status Mean σ Mean σ Mean σ

Fat Fresh 0.590 0.012 0.471 0.011 0.342 0.005

Fixed 0.593 0.011 0.483 0.008 0.346 0.003
Fixed/fresh 1.005 0.028 1.025 0.029 1.012 0.017
Fibrous Fresh 1.249 0.016 0.936 0.015 0.618 0.007
Fixed 1.189 0.017 0.918 0.008 0.604 0.002
Fixed/fresh 0.952 0.018 0.981 0.018 0.977 0.012
Tumor Fresh 1.252 0.011 0.946 0.013 0.622 0.004
Fixed 1.184 0.016 0.915 0.006 0.600 0.003
Fixed/fresh 0.946 0.015 0.967 0.015 0.964 0.008

3.4. Comparison with the published data

We compared our findings to data published earlier by Hammerstein et al (1979) and Johns
and Yaffe (1987). The latter measured the linear attenuation coefficients of fat, fibrous and
infiltrating ductal carcinoma tissues from 18 to 110 keV. For the former, the values of the
linear absorption coefficient for adipose and fibrous tissues were calculated with the XMuDat
(version 1.0.1) x-ray database (XMuDat 1998), utilizing the chemical composition of adipose
tissue and mammary gland published in the cited paper, and the data-source option ‘JH
Hubbell, SM Seltzer, NISTIR 5632, 1995’.
5004 R C Chen et al

Depicted in figure 7 are our measured linear attenuation coefficients of fat, fibrous and
tumor tissues versus the x-ray energy in the range from 15 keV to 26.5 keV. For comparison,
data from the aforementioned sources are presented as well. It turns out that our measured
values of fibrous tissues are in good agreement with the published findings, while our results
for fat tissue are consistent with Johns and Yaffe’s but slightly smaller than Hammerstein’s.
Regarding the tumor tissue, the results of Johns and Yaffe show a little difference with respect
to fibrous tissue, a difference which is, however, not confirmed by our measurements. As
presented in table 2, a systematic error is associated with our data due to formalin fixation,
and which according to our measurements is smaller than 5%.

4. Conclusions

Our experiment demonstrates that SR CT allows the evaluation of the linear attenuation
coefficients in samples of mixed composition, as long as pixel size and slice thickness are
small enough to acquire homogeneous regions in the samples, thereby avoiding partial-volume
effects.
The measured linear attenuation coefficients of different breast tissues for energy between
15 and 26.5 keV are in good agreement with the previously published data. We find no evidence
of a difference in the linear attenuation coefficients between tumor and fibrous tissues. This
is in agreement with the experience of breast radiology: dense breast tumor is sometimes
invisible in x-ray imaging (Sardanelli et al 2007), even if tomosynthesis is applied (Karellas
and Vedantham 2008). Since a significant difference in contrast between fibrous and tumor
tissues could not be demonstrated, this study suggests that tumor identification should rather
focus on the visualization of tissue structure, for instance increasing spatial resolution and
possibly using phase-contrast techniques.
Moreover, we compared the linear attenuation coefficients of fresh and formalin-fixed
breast tissues, and we found that the value for both tumor and fibrous tissues of fresh samples
is larger than that for formalin-fixed samples. Depending on the x-ray energy, differences up
to 5% were found. Fat tissues remained unchanged. In preliminary studies of new imaging
techniques, formalin-fixed tissues were commonly used (Takeda et al 2004, Bravin et al
2007, Keyrilainen et al 2008); our results confirm the reliability of this practice in absorption
imaging, as long as small systematic errors are acceptable.

Acknowledgments

The authors wish to thank Franco Vittur for providing the calibration materials. They are
indebted to Dr Ivana Schiavon for her valuable contribution in setting the histological samples.
They would also like to thank Nicola Sodini for helpful discussion. RCC and TQX were
supported by the National Basic Research Program (973 Program) (no 2010CB834301) and
the Chinese Academy of Sciences Key Project of International Co-operation (no GJHZ09058).
RCC was also supported by ICTP TRIL programme.

References

Abrami A et al 2004 Medical applications of synchrotron radiation at the SYRMEP beamline of ELETTRA Nucl.
Instrum. Methods A 548 221–7
Arfelli F et al 1996 Digital mammography at the Trieste synchrotron light source IEEE Trans. Nucl. Sci. 43 2061–7
Arfelli F et al 1998 Low-dose phase contrast x-ray medical imaging Phys. Med. Biol. 43 2845–52
Arfelli F et al 2000 Mammography with synchrotron radiation: phase-detection techniques Radiology 215 286–93
 Measurement of the linear attenuation coefficients of breast tissues by SR CT 5005

Baruchel J, Buffiere J Y, Cloetens P, Di Michiel M, Ferrie E, Ludwig W, Maire E and Salvo L 2006 Advances in
synchrotron radiation microtomography Scr. Mater. 55 41–6
Bergamaschi A, Cervellino A, Dinapoli R, Gozzo F, Henrich B, Johnson I, Kraft P, Mozzanica A, Schmitt B and Shi
X 2009 Photon counting microstrip detector for time resolved powder diffraction experiments Nucl. Instrum.
Methods A 604 136–9
Bravin A et al 2007 High-resolution CT by diffraction-enhanced x-ray imaging: mapping of breast tissue samples
and comparison with their histo-pathology Phys. Med. Biol. 52 2197–211
Carroll F E et al 1994 Attenuation of monochromatic x-rays by normal and abnormal breast tissues Invest.
Radiol. 29 266–72
Castelli E et al 2007 Clinical mammography at the SYRMEP beam line Nucl. Instrum. Methods A 572 237–40
Chapman D, Pisano E, Thomlinson W, Zhong Z, Johnston R E, Washburn D, Sayers D and Malinowska K 1998
Medical applications of diffraction enhanced imaging Breast Dis. 10 197–207
Cloetens P, Barrett R, Baruchel J, Guigay J P and Schlenker M 1996 Phase objects in synchrotron radiation hard
x-ray imaging J. Phys. D: Appl. Phys. 29 133–46
Dilmanian F A et al 1997 Single- and dual-energy CT with monochromatic synchrotron x-rays Phys. Med.
Biol. 42 371–87
Dreossi D et al 2008 The mammography project at the SYRMEP beamline Eur. J. Radiol. 68 S58–62
Fiedler S, Bravin A, Keyrilainen J, Fernandez M, Suortti P, Thomlinson W, Tenhunen M, Virkkunen P and Karjalainen-
Lindsberg M L 2004 Imaging lobular breast carcinoma: comparison of synchrotron radiation DEI-CT technique
with clinical CT, mammography and histology Phys. Med. Biol. 49 175–88
Greene F, Balch C, Haller D and Morrow M 2002 AJCC Cancer Staging Manual 6th edn (New York: Springer)
Hammerstein G R, Miller D W, White D R, Masterson M E, Woodard H Q and Laughlin J S 1979 Absorbed radiation
dose in mammography Radiology 130 485–91
Johns P C and Yaffe M J 1987 X-ray characterization of normal and neoplastic breast tissues Phys. Med.
Biol. 32 675–95
Kao T, Connor D, Dilmanian F A, Faulconer L, Liu T, Parham C, Pisano E D and Zhong Z 2009 Characterization of
diffraction-enhanced imaging contrast in breast cancer Phys. Med. Biol. 54 3247–56
Karellas A and Vedantham S 2008 Breast cancer imaging: a perspective for the next decade Med. Phys. 35 4878–97
Keyrilainen J et al 2008 Toward high-contrast breast CT at low radiation dose Radiology 249 321–7
Mozzanica A, Bergamaschi A, Dinapoli R, Gozzo F, Henrich B, Kraft P, Patterson B and Schmitt B 2009 MythenII:
a 128 channel single photon counting readout chip Nucl. Instrum. Methods A 607 250–2
Munch B, Trtik P, Marone F and Stampanoni M 2009 Stripe and ring artifact removal with combined wavelet—Fourier
filtering Opt. Express 17 8567–91
Nuzzo S, Peyrin F, Cloetens P, Baruchel J and Boivin G 2002 Quantification of the degree of mineralization of bone
in three dimensions using synchrotron radiation microtomography Med. Phys. 29 2672–81
Olivo A and Speller R 2006 Experimental validation of a simple model capable of predicting the phase contrast
imaging capabilities of any x-ray imaging system Phys. Med. Biol. 51 3015–30
Pani S et al 2004 Breast tomography with synchrotron radiation: preliminary results Phys. Med. Biol. 49 1739–54
Pisano E D 2000 Current status of full-field digital mammography Radiology 214 26–8
Rigon L, Arfelli F, Astolfo A, Bergamaschi A, Dreossi D, Longo R, Menk R H, Schmitt B, Vallazza E and Castelli
E 2009 A single-photon counting ‘edge-on’ silicon detector for synchrotron radiation mammography Nucl.
Instrum. Methods A 608 S62–5
Sardanelli F et al 2007 Multicenter comparative multimodality surveillance of women at genetic-familial high risk
for breast cancer (HIBCRIT study): interim results Radiology 242 698–715
Tafforeau P et al 2006 Applications of x-ray synchrotron microtomography for non-destructive 3D studies of
paleontological specimens Appl. Phys. A 83 195–202
Takeda T, Momose A, Ueno E and Itai Y 1998 Phase-contrast x-ray CT image of breast tumor J. Synchrotron
Radiat. 5 1133–5
Takeda T, Wu J, Tsuchiya Y, Yoneyama A, Lwin T T, Aiyoshi Y, Zeniya T, Hyodo K and Ueno E 2004 Interferometric
x-ray imaging of breast cancer specimens at 51 keV x-ray energy Japan. J. Appl. Phys. 43 5652–6
Trtik P, Dual J, Keunecke D, Mannes D, Niemz P, Stahli P, Kaestner A, Groso A and Stampanoni M 2007 3D imaging
of microstructure of spruce wood J. Struct. Biol. 159 46–55
Vallazza E, Arfelli F, Bruni F, Castelli E, Longo R, Pontoni C, Rigon L and Rokvic T 2008 A double layer silicon
detector for single photon counting Proc. 10th ICATPPT Conf. (Singapore: World Scientific) pp 700–5
International Atomic Energy Agency 1998 XMuDat http://www-nds.iaea.org/reports/nds-195.htm

View publication stats