You are on page 1of 10

Journal of Instrumentation

A Monte Carlo simulation study on effects of energy weighted dual-


energy subtraction images with indirect photon-counting detector in
mammography
To cite this article: C.-H. Baek et al 2020 JINST 15 C03015

View the article online for updates and enhancements.

This content was downloaded from IP address 130.238.7.40 on 10/03/2020 at 17:39


Published by IOP Publishing for Sissa Medialab
Received: October 7, 2019
Revised: January 3, 2020
Accepted: February 3, 2020
Published: March 10, 2020
21st International Workshop on Radiation Imaging Detectors
7–12 July 2019
Crete, Greece

A Monte Carlo simulation study on effects of energy

2020 JINST 15 C03015


weighted dual-energy subtraction images with indirect
photon-counting detector in mammography

C.-H. Baek,a S.-J. Lee,b,c I.-H. Chod and D. Kimd,1


a Department of Radiological Science, Kangwon National University,
Samcheok, Republic of Korea
b Department of Radiological Science, Dongseo University,

Busan, Republic of Korea


c Center for Radiological Environment & Health Science, Dongseo University,

Busan, Republic of Korea


d Department of Radiological Science, Eulji University,

Seongnam, Republic of Korea


E-mail: goldcollar011@gmail.com

Abstract: The purpose of this study is to design an indirect photon-counting detector using a GAGG
scintillator with Monte Carlo simulation and to evaluate the efficacy of the result using an energy
weighted subtraction method. Factors to validate the efficacy of energy weighted subtraction were
the type of scintillator and the profile of the x-ray spectrum. Images were obtained using GAGG
and CsI scintillators with incident x-ray spectra split from continuous x-rays and were processed
using energy weighted subtraction and energy subtraction. The composition of the phantom for
image acquisition consisted of calcium embedded in the breast tissue, and the phantom images were
analyzed by comparing the contrast-to-noise ratio (CNR). Based on the simulation results, GAGG
demonstrates improved CNR compared to CsI, and the energy weighted subtraction method yields
similar or higher CNR values compared to the energy subtraction method except in the case of using
the CsI detector in a continuous x-ray spectrum. Moreover, the CNR values were higher for split
x-rays than for continuous x-rays. In conclusion, GAGG detectors, the energy weighted subtraction
method, and a split spectrum can be used to enhance the contrast of images while augmenting the
fine calcification of simulated breast images.
Keywords: X-ray mammography and scinto- and MRI-mammography; X-ray radiography and
digital radiography (DR)

1Corresponding author.

c 2020 IOP Publishing Ltd and Sissa Medialab


https://doi.org/10.1088/1748-0221/15/03/C03015
Contents

1 Introduction 1

2 Materials and methods 2


2.1 X-ray beam generation 2
2.2 Phantom and detector 3
2.3 Simulation design 3
2.4 Energy weighted dual-energy subtraction 3

2020 JINST 15 C03015


2.4.1 Energy weighting method 3
2.4.2 Energy subtraction 4
2.4.3 Energy weighted Subtraction 4
2.4.4 Image evaluation 5

3 Results 5

4 Discussion and conclusion 7

1 Introduction

Based on the statistics of the Korea Central Cancer Registry published in 2016, breast cancer
ranks first in women’s cancer incidence [1]. Breast cancer has a high early diagnosis rate and a
high rate of survival following treatment at early stages, which highlights the importance of early
cancer diagnosis in these cases. Mammography is the most widely used screening method for
detecting breast cancer; however, the poor contrast between adipose and glandular tissue in the
breast reduces the overall accuracy of breast cancer diagnosis [2]. To overcome this issue, dual-
energy mammography is used. Dual-energy mammography can enhance the visualization of the
signal against background tissues through dual x-ray irradiations. Also, a photon-counting system
based direct mode was used to reduce radiation dose to the breast and increases the accuracy of
material separation in simulation study [3].
To overcome such problems, the direct photon-counting detector based on the semiconductor
for dual-energy x-ray imaging was introduced to distinguish x-ray energies with a single exposure
and eliminate energy information interference using an energy window. Additionally, the x-ray
detection efficiency is high due to the high atomic numbers of the underlying materials such as
CZT and CdTe. Despite these advantages, there exist several issues including the problem of charge
sharing due to the reduction of pixel size for high spatial resolution images, pulse overlap when
used in high-dose environments such as diagnostic x-ray screening, reduction of uniformity due to
defects in the semiconductor itself, and the challenge of large-scale production [4].
In order to solve the problems of uniformity reduction and large area production which are the
major limitations of current semiconductor-based photon-counting detectors, it is necessary to design
an indirect photon-counting detector using the scintillator and the silicon photomultiplier device.

–1–
Conventional indirect detector for mammography is energy-integration mode. Recently, the
indirect photon-counting detector for x-ray imaging is introduced. They reported two-energy-
level photon-counting performance by design of a photon-counting image sensor [5]. For breast
imaging, an indirect photon-counting detector was developed with combinations of scintillator and
complementary metal-oxide-semiconductor (CMOS) for digital breast tomosynthesis (DBT) [6].
For breast imaging, CsI is the most widely used detector scintillator material in mammography
and is used as a charge accumulation pixel detector for detecting a large number of x-rays [7].
However, CsI scintillators have the disadvantages of low density, long decay time, and deliques-
cence. Recently, the adoption of GAGG scintillators as radiation detectors in medical imaging has
increased due to their higher density, shorter attenuation times, and non-deliquescence compared

2020 JINST 15 C03015


to conventional CsI-based detectors [8].
Based on these advantages of GAGG scintillators, the potential exists for the development
of photon-counting-based mammography devices with improved uniformity and larger-scale with
silicon photomultiplier devices. Therefore, the purpose of this study is to design a scintillator
based indirect photon-counting detector using Monte Carlo simulation and evaluate the efficacy of
the result. The advantages of GAGG scintillators was evaluated through energy-weighted image
analysis obtained using GAGG and CsI scintillators. Additionally, the result of energy-weighted
subtraction (EW DES) images was evaluated using the x-ray spectrum split by a Tin filter.

2 Materials and methods

2.1 X-ray beam generation


The SRS-78 code was used to generate an x-ray beam with a 30 kV tube voltage and a Rhodium target.
The generated x-ray beams were input to the TASMIP code [9] to generate two types of spectrums,
as shown in figure 1 (a) and (b). The spectrum in figure 1(a) is the shape of the spectrum produced
after the transmission through Rhodium filter with a thickness of 1HVL into the front part of the x-
ray beam produced by the SRS-78 code. Figure 1(b) illustrates a split x-ray spectrum produced after
transmission through a 6HVL-thick Tin filter. Each spectrum in figure 1 was entered into the GATE
Monte Carlo simulation code to simulate the x-ray beam irradiation. In figure 1(a), the range of
energy windows for photon-counting was 13–22 keV at low energies and 23–28 keV at high energies.
The energy ranges in figure 1(b) were 13–23 keV at low-energies and 24–28 keV at high-energies.

Figure 1. X-ray spectra generated using Rh target (a) and Sn filter (b). (a) is the continuous x-ray spectrum
and (b) is the split x-ray spectrum.

–2–
2.2 Phantom and detector
The detector consists of CsI or GAGG materials and has dimensions of 50.0 × 25.0 mm2 a thickness
of 2.0 mm, and a pixel size of 0.5 × 0.5 mm2 . The phantom is composed of the breast tissue in
a cylindrical form with a diameter of 46 mm and a thickness of 40 mm, and contained calcium, a
calcification material, each with a diameter of 4 mm, 7 mm and 10 mm and a thickness of 5 mm.

2.3 Simulation design


In this study, a photon-counting detector system was simulated using Geant4 Application for
Tomographic Emission (GATE) version 6.0 [10]. The focal spot to detector distance (FDD) was set

2020 JINST 15 C03015


to 650.0 mm. The breast phantom was placed between the source and the detector, and the x-ray
beam was irradiated using the spectrums shown in figures 1 (a) and (b). In order to apply energy
subtraction (DES) and energy weighting subtraction (EW DES), radiation was exposed with and
without the phantom, respectively.
The image was acquired by irradiating x-rays in the range of 13–22 keV for low energies
with the continuous x-ray beam of figure 1(a) in units of 1 keV. For high energy, the energy was
irradiated with x-rays in the range of 23–28 keV, and images of each energy were obtained. In
addition, figure 1(b) shows the split x-rays where low energy in the range of 13–23 keV and high
energy in the range of 24–28 keV images were acquired respectively.

2.4 Energy weighted dual-energy subtraction


2.4.1 Energy weighting method
The energy weighting method can be derived from the difference in the linear attenuation coefficients
of two different materials and the difference in the thicknesses of the two materials as discussed by
Shikhaliev et al. [11].

Figure 2. Schematic illustration for the energy weighting factor. The background material has a thickness of
L and an attenuation coefficient of µb , and includes a constant element with a thickness of d and attenuation
coefficient of µs .

As shown in figure 2, it is assumed that the thickness of the background material is L and
the linear attenuation coefficient of the background material is µb , and that the background ma-

–3–
terial includes the signal material of thickness d with the linear attenuation coefficient µs . The
transmittance of the background material and the signal material can be expressed as the following:

Tb = Ib (E)/I0 (E) = exp(−µb (E) · L) (2.1)


Ts (E) = Is (E)/I0 (E) = exp{−µs (E) · (L − d) − µs (E) · d} (2.2)

Here, Tb (E) is the transmission of photons transmitted through the background material, and Ts (E)
is the transmission of photons transmitted through the background material and the signal material.
I0 (E) is the number of photons of the x-ray beam incident on the detector in the air, and Ib (E)
is the number of photons incident on the detector after passing through the background material.

2020 JINST 15 C03015


Is (E) refers to the number of photons incident on the detector after passing through the background
material and the signal material. Based on Equations (2.1) and (2.2), the energy weighting factor
W(E) can be expressed as follows:
Tb − Ts 1 − exp{(µb (E) − µs (E)) · d}
W(E) = = (2.3)
Tb + Ts 1 + exp{(µb (E) − µs (E)) · d}
Equation (2.3) above is an energy-weighted factor, which can be obtained using the linear atten-
uation coefficient and thickness of the background and signal material, and by multiplying the
image for each energy pixel-by-pixel, the energy-weighted image can be obtained as shown in the
Equation (2.4) as follows. Õ
F= W(Ei ) · P(Ei ) (2.4)
i

F is an energy-weighted image with a weighting factor and Ei represents energy. P(Ei ) is an x-ray
image obtained from energy Ei .

2.4.2 Energy subtraction


The energy subtraction technique uses two base materials, as discussed in a previous study [12].
The basis materials in this study were breast tissue and calcium. The equation to compute the
calcium-enhanced image is as follows:

DESCa = ln(P0L /PL ) − wb · ln(P0H /PH ) (2.5)

P0L is the image that was obtained with the detector in air at a low energy region and PL is the
image that was transmitted through an object at the corresponding energy. P0H is the image that was
obtained with the detector in air at a high energy region and PH is the image that was transmitted
through an object at the corresponding energy. wb is a weighting factor that minimizes the signal
of the breast tissue as background material and to enhance the calcium as signal.

2.4.3 Energy weighted Subtraction


Energy weighted subtraction was performed using Equations (2.4) and (2.5). Using this method,
the energy subtraction method was applied after the energy weighting was performed in the low
and high energy images. The final equation for energy weighted subtraction of calcium-enhanced
images is as follows.
EWDESCa = ln(F0L /FL ) − wFb · ln(F0H /FH ) (2.6)

–4–
F0L and F0H are the energy-weighted images that were obtained with the detector in air at low and
high-energy regions, while FL and FH are the energy-weighted images that were obtained with
the detector transmitted through objects at low- and high-energy regions, respectively. wFb is a
weighting factor that minimizes the signal of the background material in the energy weighted image.

2.4.4 Image evaluation


Image evaluation was performed using the CNR, which can quantitatively analyze the relationship
between contrast and noise on images. The CNR of the energy subtracted image and the energy
weighted subtracted image can be obtained by the following equation:

2020 JINST 15 C03015


|Ss − Sb |
CNR = (2.7)
SDb
Ss is the average value of the calcium concentration in the energy subtracted image and energy
weighted subtracted image. Sb and SDb are the average value and standard deviation of breast
tissue as a background material, respectively.

3 Results

Shown in figure 3(a), the energy-dependent linear attenuation coefficient of breast tissue and calcium
as the background material and the signal material are obtained by extracting data from NIST [13].
Energy-weighting factors were calculated by substituting the linear attenuation coefficient of breast
tissue and calcium and the thickness of calcification into Equation (2.3). The energy weighting
factors of breast and calcification material calculated by Equation (2.3) are shown in figure 3(b).

Figure 3. The linear attenuation coefficients of CaCO3 and breast tissue (a). Energy weighting factors in
accordance with energy for CaCO3 and breast tissue (b).

Energy-weighted images were obtained by inserting the calculated weighting factor into Equa-
tion (2.4) and the corresponding energy weighting factor at each energy. The weight (wb ) of
breast tissue, the background material of Equation (2.5) used to obtain DES images, is the linear
attenuation coefficient of the background material at low energy divided by the linear attenuation
coefficient at high energy in continuous x-rays.
In split x-rays, breast tissue weights were obtained by dividing the linear attenuation coefficient
of the background material at low energy by the linear attenuation coefficient of the background

–5–
material at high energy. EW DES images were calculated using Equation (2.4) and the weight (wFb )
of breast tissue was calculated as in the DES method.
Figures 4 and 5 show images using DES and images using EW DES using Equations (2.5)
and (2.6) and in addition, the degree of enhancement of images according to the types of CsI and
GAGG scintillators. figure 4 shows the image obtained using a continuous x-ray spectrum, and
figure 5 shows the image obtained using a split x-ray spectrum.

2020 JINST 15 C03015


Figure 4. Conventional DES (a) and EW DES (b) images with CsI scintillator. Conventional DES (c) and
EW DES images (d) with GAGG scintillator. All images were exposured by continuous x-ray spectrum with
Rh filter.

Figure 5. Conventional (a) and EW DES (b) images with CsI scintillator. Conventional (c) and EW DES (d)
images with GAGG scintillator. All images were exposured by using split x-ray spectrum with Sn filter.

The results of the CNRs in breast tissue with calcium is compared between the conventional
DES and the EW DES images measured in figures 4 and 5. In the results of continuous x-rays,
the CNR of the EW DES images were 11.50 and 12.47 for CsI and GAGG, respectively, compared
to 11.86 and 12.52 in the DES images. When using GAGG to obtain EW DES images, the CNR
increased by 8.48% and in the DES image the CNR increased by 5.55%.

–6–
In results of the split x-rays, the CNR of the EW DES images were 14.95 and 15.28, and in DES
images the CNR was 14.59 and 15.15 for CsI and GAGG, respectively. Using GAGG, the CNR
increase was 2.17% in the EW DES images and 3.88% in DES images. The CNR enhancement of
split x-rays were 30.04 and 22.48% in EW DES images, and 22.98 and 21.02% in DES images for
CsI and GAGG, respectively.

4 Discussion and conclusion

In this study, we have designed a scintillator based indirect photon-counting detector system using

2020 JINST 15 C03015


Monte Carlo simulation and verified the advantages of GAGG scintillators by analyzing the EW
DES images obtained using GAGG and CsI scintillators. The results of EW DES images and
DES images were evaluated using x-ray spectrum split by the Tin filter, and the advantages in the
application of EW DES methods using the split x-ray spectrum was demonstrated.
GAGG showed better CNR than CsI when the image was acquired using a continuous x-ray
spectrum. The CNR increased 8.48% in EW DES image acquisition and 5.55% in energy-reduction
images. Using GAGG and a split x-ray spectrum, the CNRs were increased by 2.17% in EW DES
images and 3.88% in DES images. Such results are due the high density and high light emission
characteristic of GAGG. Therefore, GAGG showed improved contrast on the EW DES image and
the DES image in comparison to the CsI used in the conventional mammography devices.
With the application of EW DES methods, the CNR of images were improved by 2.50 and
0.82% in CsI and GAGG, respectively, using the split x-ray spectrum. However, when the continuous
x-ray spectrum was used, the CNR of the EW DES method was reduced by −3.06% at CsI and
similar CNR values were observed in GAGG. This is due to the small difference in the value of
the energy weighted coefficient in continuous x-rays as shown in figure 3(b). Since the split x-rays
have large energy difference between low and high energy, the CNR enhancement result was 30.04
and 22.48% in the EW DES images and 22.98 and 21.02% in the DES images of CsI and GAGG,
respectively.
In our simulation study, contrast enhancement was achieved using GAGG scintillator and split
x-ray spectrums. However, it is expected to increase exposure time (> 1 s) for achieving high
spatial resolution because the split x-ray spectrum is produced by 6 HVL Tin filter. Therefore, the
limitation of the split x-ray spectrum is considered in clinical use. Also, the calcification size is
larger than 1 mm as mentioned in section 2.2 since we evaluate the contrast of the image using image
metric as CNR. Although appropriate lesion and background material size is needed to obtain CNR
value, the calcification size would be considered in clinical interest.
In conclusion, contrast enhancement of images can be achieved using GAGG detectors, split
x-ray spectrums, and GAGG detectors and therefore increase the contrast between calcification
materials and normal breast tissue during dual-energy mammography.

Acknowledgments

This work was supported by the National Research Foundation of Korea (NRF) grant funded by the
Ministry of Science and ICT (Grant No. NRF-2017R1C1B5017626).

–7–
References

[1] https://ncc.re.kr/cancerStatsList.ncc.
[2] A.D. Laidevant, S. Malkov, C.I. Flowers, K. Kerlikowske and J.A. Shepherd, Compositional breast
imaging using a dual-energy mammography protocol, Med. Phys. 37 (2010) 164.
[3] Y. Lee, S. Lee, S. Kang and J. Eom, Dose optimization for dual-energy contrast-enhanced digital
mammography based on an energy-resolved photon-counting detector: A Monte Carlo simulation
study, Radiat. Phys. Chem. 132 (2017) 46.
[4] L. Ren, B. Zheng and H. Liu, Tutorial on X-ray photon counting detector characterization, J. Xray
Sci. Technol. 26 (2018) 1.

2020 JINST 15 C03015


[5] B. Dierickx, Q. Yao, N. Witvrouwen, D. Uwaerts, S. Vandewiele and P. Gao, X-ray photon counting
and Two-Color X-ray imaging using indirect detection, Sensors 16 (2016) 764.
[6] H. Jiang, J. Kaercher and R. Durst, Indirect-detection single-photon-counting x-ray detector for breast
tomosynthesis, Proc. SPIE 9783 (2016) 97833P.
[7] A. Mackenzie, L.M. Warren, M.G. Wallis, J. Cooke, R.M. Given-Wilson, D.R. Dance et al., Breast
cancer detection rates using four different types of mammography detectors, European Radiology 26
(2015) 874.
[8] J. Kataoka et al., Handy Compton camera using 3D position-sensitive scintillators coupled with
large-area monolithic MPPC arrays, Nucl. Instrum. Meth. A 732 (2013) 403.
[9] J. Punnoose, J. Xu, A. Sisniega, W. Zbijewski and J.H. Siewerdsen, Technical note: spektr 3.0 — A
computational tool for x-ray spectrum modeling and analysis,
Med. Phys. 43 (2016) 4711.
[10] S. Jan, D. Benoit, E. Becheva, T. Carlier, F. Cassol, P. Descourt et al., GATE V6: a major
enhancement of the GATE simulation platform enabling modelling of CT and radiotherapy, Phys.
Med. Biol. 56 (2011) 881.
[11] P.M. Shikhaliev, Energy-resolved computed tomography: first experimental results, Phys. Med. Biol.
53 (2008) 5595.
[12] M. Koutalonis, H. Delis, G. Spyrou, L. Costaridou, G. Tzanakos and G. Panayiotakis,
Contrast-to-noise ratio in magnification mammography: A Monte Carlo study, Phys. Med. Biol. 52
(2007) 3185.
[13] https://www.nist.gov/pml/productsservices/physical-reference-data.

–8–

You might also like