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

Download details:

IP Address: 125.214.163.2

This content was downloaded on 13/10/2013 at 04:21

Please note that terms and conditions apply.

Detective quantum efficiency of a silicon microstrip photon-counting detector having edge-on

geometry under mammography imaging condition

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

2011 JINST 6 C12006

(http://iopscience.iop.org/1748-0221/6/12/C12006)

Home Search Collections Journals About Contact us My IOPscience

2011 JINST 6 C12006

PUBLISHED BY IOP PUBLISHING FOR SISSARECEIVED: September 6, 2011

ACCEPTED: November 23, 2011PUBLISHED: December 5, 2011

13th INTERNATIONAL WORKSHOP ON RADIATION IMAGING DETECTORS,3–7 JULY 2011,ETH ZURICH, SWITZERLAND

Detective quantum efficiency of a silicon microstripphoton-counting detector having edge-on geometryunder mammography imaging condition

S. Yun,a H.K. Kim,a,1 H. Youn,a O. Joe,a S. Kim,b J. Park,b D.G. Kang,c Y.H. Sung,c

J. Marchal,d J. Tanguaye and I.A. Cunninghame

aSchool of Mechanical Engineering, Pusan National University,Busan 609-735, Republic of Korea

bSemiconductor Device Laboratory, Samsung Advanced Institute of Technology,Yongin, Gyeonggi-Do 446-712, Republic of Korea

cMedical Imaging Systems Group, Samsung Advanced Institute of Technology,Yongin, Gyeonggi-Do 446-712, Republic of Korea

dDiamond Light Source Ltd., Harwell Science and Innovation Campus,Didcot, Oxfordshire OX11 0DE, U.K.

eImaging Research Laboratories, Robarts Research Institute,100 Perth Drive, London, Ontario N6A 5K8,Canada

E-mail: hokyung@pusan.ac.kr

ABSTRACT: We have investigated the image quality of a silicon microstrip detector system oper-ated in single-photon counting mode under mammography imaging condition. The detector has anedge-on geometry with a tilting angle of 5 degrees to the normal direction of X-ray incidence. It iscomposed of four modules and each module employs 256 silicon microstrips. Using a slanted-edgeknife technique, the modulation-transfer function (MTF) without aliasing was determined. Noise-power spectrum (NPS) was determined using two-dimensional (2D) Fourier analysis on the line-scanned 2D images. Based on the measured MTF and NPS results, detective quantum efficiency(DQE) was calculated. These systematic procedures were repeated at various energy thresholds.Asymmetric MTF properties between two perpendicular directions were observed because of thescan motion. Spectral densities in NPS were white for spatial frequencies. The best DQE valuearound zero-spatial frequency was about 0.7. It was observed that the DQE was independent of thelevel of X-ray exposure, which is desirable for low-dose mammography.

KEYWORDS: Si microstrip and pad detectors; X-ray detectors; X-ray mammography and scinto-and MRI-mammography

1Corresponding author.

c© 2011 IOP Publishing Ltd and SISSA doi:10.1088/1748-0221/6/12/C12006

2011 JINST 6 C12006

Contents

1 Introduction 1

2 Materials and methods 12.1 Description of the Si microstrip detector system 12.2 Analysis of image quality 2

3 Results 3

4 Discussion and conclusion 4

1 Introduction

An early diagnosis of breast cancer from mammograms is difficult due to small-size lesion (e.g.micro-calcification) and the density similarity between the malignant mass and the surrounding nor-mal tissue. These difficulties may partly be remedied with an increased X-ray exposure since imagequality is proportional to radiation dose in some degree, but this approach should be avoided consid-ering the patient dose. Therefore, digital mammography detectors require a high spatial-resolvingpower and high dynamic range even with low X-ray exposure [1]. Solid-state detectors operatedin photon-counting mode are attractive in this regard because: (1) the direct-conversion detectorbased on semiconductor material provides superior spatial resolution to the indirect-conversion de-tector based on scintillator; (2) the dynamic range in the photon-counting approach is theoreticallyinfinite; and (3) electronic noise, which degrades the signal-to-noise performance in images, canbe suppressed by the thresholding technique [1, 2].

We have developed a silicon (Si) microstrip detector array for line-scanned imaging and in-vestigated its image quality for mammography. As objective image-quality metrics, modulation-transfer function (MTF), noise-power spectrum (NPS) and detective quantum efficiency (DQE)were considered. Although the overall results were comparable to the conventional digital mam-mography detectors, we gained the potential operation in a lower dose. Detailed information aboutthe detector system, measurement procedures and image quality results are given in this paper.

2 Materials and methods

2.1 Description of the Si microstrip detector system

The detector consists of four Si microstrip-array modules and each module comprises a 256-microstrip array. The dimension of a microstrip is 0.095 mm wide, 0.5 mm height and 10 mmlong. The detector system is designed so that the X-ray beam is incident along the longitudinaldirection of the detector. This is known as the “edge-on” geometry and it improves the X-ray

– 1 –

2011 JINST 6 C12006

Figure 1. Schematics showing the experimental setup and the geometry of the silicon microstrip photon-counting detector.

interaction efficiency of the detector [2, 3]. Since the dead zone, which is insensitive to X-ray in-teraction, is located at the front edges of the microstrips, the detector was tilted to the X-ray beamincidence by 5 degrees as illustrated in figure 1. Therefore, the effective depth for X-ray interactionis about 5.7 mm. The detector is operated in photon-counting mode and has four energy thresholds.The maximum counting capability per strip is 7 × 105 counts per second. The readout chips weredesigned to allow a bit-depth of 18 bits. The electronic noise level is about 200 e− rms which isequivalent to 0.72 keV. In this study, one of thresholds is used to trim the electronic noise floor, andothers are set to 13, 15 and 22 keV, respectively.

As schematically illustrated in figure 1, the X-ray beam was collimated to provide a symmetricpixel format. Since the slit designed for beam collimation was measured at 0.1 mm, however, thepixel size is slightly asymmetric (0.095× 0.1 mm). To obtain two-dimensional (2D) images, a tray,on which an object to be investigated is located, is scanned, while the X-ray tube and the detectorare fixed.

2.2 Analysis of image quality

Objective quality of images obtained from the Si microstrip detector system was assessed in termsof MTF, NPS and DQE which describe the spatial resolution, noise variance and signal-to-noiseperformance of an X-ray detector, respectively, as a function spatial frequency [4]. The MTF wasmeasured for two perpendicular directions; line and scan directions as depicted in figure 1. ThisMTF measurement was performed using the slanted-edge method to avoid aliasing. The NPS wasevaluated by 2D Fourier analysis of 80 images with a size of 256× 256 pixels and one-dimensional(1D) NPS was extracted along line and scan directions. All the images used for evaluating MTFand NPS were corrected for offset levels and gain variations in images.

– 2 –

2011 JINST 6 C12006

The DQE was calculated by

DQE(u,v) =q0G2MTF2(u,v)

NPS(u,v), (2.1)

where q0 is the average distribution of X-ray quanta incident on the detector (or incident photonfluence) and G is the system gain relating q0 to the average pixel value d. u and v are Fourierconjugates of spatial variables in the Cartesian coordinates. The photon fluence was estimatedbased on the measured half-value layer (HVL) and the in-house X-ray spectrum simulator [5]. TheDQE analysis was compliant with IEC 62220-1-2 using a W/Al spectrum at 28 kVp [6].

3 Results

The average pixel value in the unit of digital numbers (DN) of the Si microstrip detector as afunction of exposure measured at the detector entrance surface is displayed in figure 2(a) withrespect to various threshold energies. From the linear least-squares regression analyses, the detectorresponses vary linearly with exposure levels for all threshold energies. As expected, the slope of theresponse curve, which is related to the detector gain, is reduced as the threshold energy is increased.

The MTF curves measured for the two different threshold energies of 13 and 22 keV areplotted in figure 2(b) for two perpendicular directions. The MTF along the scan direction is muchlower than that along the line direction. Although the asymmetric pixel shape in two directionscontributes to this observation, it is presumed that mismatch between the scan speed and slit widthwould be the main cause. The other observation is that higher threshold energy enhances the MTFperformance along the line direction. The dependence of MTF on energy threshold in microstripdetectors has been studied in reference [7]. Setting a threshold higher than half the X-ray energyresults in a loss of counts at strip edges. This improves the MTF by reducing the effective stripwidth. However the MTF measured in our case is far away from the one expected with a 95microns strip subject to charge-sharing effects only [c.f. figure 3(a)]. The MTF might be stronglydegraded by geometrical factors such as misalignment of the detector strips with the incident beamor focal spot size.

The NPS curves measured for the two different threshold energies of 13 and 22 keV are shownin figure 2(c) and (d), respectively, with respect to various exposure levels. As the exposure levelis increased, the noise spectral densities also increase due to the increased quantum noise. Spectraldensities are almost independent on spatial frequencies.

The DQE curves are displayed in figure 2(e) and (f). As shown in figure 2(e), the depen-dencies of the detector gain, MTF and NPS on threshold energies and evaluation directions arewell reflected into the DQE results. The dependence of DQE on the exposure level is shown infigure 2(f), and it is negligible for the range of exposure levels investigated in this study (1.1–4.3mR). Removing electronic noise from the imaging process by using pulse discrimination logic inthe photon-counting detector results in a quantum-noise-limited operation, hence independence ofDQE on the exposure levels.

It is noted that the measurement results of MTF, NPS and DQE from the threshold energy of15 keV are similar to those from 13 keV as observed from the measurement of the detector response[see figure 2(a)]. To avoid complexity, the results of MTF, NPS and DQE from the threshold energyof 15 keV are not shown in figure 2.

– 3 –

2011 JINST 6 C12006

Figure 2. Summary of image quality measurements. (a) The detector response as a function of exposurelevel for various threshold energies. (b) MTF measured in two perpendicular directions (line and scandirections) for various threshold energies. (c) NPS measured in two perpendicular directions for the thresholdenergy of 13 keV at various exposure levels. (d) NPS measured in two perpendicular directions for thethreshold energy of 22 keV at various exposure levels. (e) DQE measured in two perpendicular directionsfor various threshold energies. (f) DQE measured in the line direction at various exposure levels. It is notedthat solid and dotted lines designate the line and scan directions, respectively, in (b), (c), (d), and (e).

4 Discussion and conclusion

There are several factors affecting the spatial resolution of the detector system such as the pixelgeometry (i.e. aperture response), oblique X-ray incidence, Compton scattering, charge diffusion,and so on. To quantitatively investigate the degradation in the system MTF performance from theideal pixel geometry, the MTF data measured in the line direction for the threshold energy of 13keV are re-plotted in figure 3(a). The MTF due to pixel geometry can be described by the sinecardinal function or sinc function. Assuming that other signal spreading factors are independent of

– 4 –

2011 JINST 6 C12006

Figure 3. (a) MTF analysis using the pixel geometry and the Gaussian signal spreading model. (b) Compar-isons between the measured DQE(0) and theoretical quantum efficiency.

each other and each factor is described by the Gaussian model, we may describe the total systemMTF as

MTF(u) =∣∣∣∣sin(πau)

(πau)

∣∣∣∣× exp(−2(πσu)2

), (4.1)

where a is the aperture size and σ is the total standard deviation of the Gaussian signal spreadingmodel. From the regression analysis with eq. (4.1), we obtained that σ ' 0.058 mm. The MonteCarlo simulation showed negligible signal spreading due to X-ray interactions such as the Comptonscattering in the detector material (Si in this study). According to the simple charge diffusion modelconsidering the depth of interaction, the signal spreading due to charge diffusion was ∼ 0.015mm [8]. This confirms that the MTF degradation observed is not due to charge-sharing effectsbetween strips but more likely to geometrical factors such as misalignment of the strips or focalspot size effects. By analyzing the measured MTFs with respect to various angles of rotation in thedetector along each axis as well as magnification factors, the misalignment and focal size effectsare under investigation.

Figure 3(b) shows comparisons between the theoretical quantum efficiency α obtained fromthe Monte Carlo simulation and the measured DQE around zero-spatial frequency obtained fromeq. (2.1) or DQE(0) = d2/NPS(0)q0. Since the Swank noise factor can be given by

I =DQE(0)

α, (4.2)

we may extract the Swank factors of 0.80, 0.76 and 0.71 for threshold energies of 13, 15 and 22keV, respectively, from figure 3(b). The Swank noise in a detector can be described by the statisticalfluctuation in the number of collected charge carriers. If the number of charge carriers producedper X-ray photon absorption is large and the most of them are collected, the fluctuation is mainlygoverned by the stochastic nature of X-ray interactions. The charge-sharing effect can furtherdistort the number of collected charge carriers in a pixel. Therefore, the energy-dependent Swankfactors mainly appreciate energy-dependent X-ray interactions and charge-sharing effect [9].

We have developed the Si microstrip detector for the line-scanned digital mammography andinvestigated image quality in terms of MTF, NPS and DQE. The measured image quality is com-

– 5 –

2011 JINST 6 C12006

parable to those of commercial mammography detectors [10]. The potential for low-dose imagingcapability attracts the photon-counting detector as a digital mammography detector further.

Acknowledgments

This research was supported by Basic Science Research Program through the National ResearchFoundation of Korea (NRF) funded by the Ministry of Education, Science and Technology (2011-0009769).

References

[1] M.J. Yaffe, Detectors for digital mammography, in Digital mammography, E.D. Pisno et al. eds.,Springer, U.S.A. (2009).

[2] M. Aslund et al., Physical characterization of a scanning photon counting digital mammographysystem based on Si-strip detectors, Med. Phys. 34 (2007) 1918.

[3] L. Rigon et al., A single-photon counting “edge-on” silicon detector for synchrotron radiationmammography, Nucl. Instrum. Meth. A 608 (2009) S62.

[4] J.T. Dobbins III, Handbook of medical imaging. Volume 1: physics and psychophysics, SPIE Press,U.S.A. (2009), see chapter 3.

[5] C.-S. Shon et al., Computational toolset for X-ray spectral analysis, Key Eng. Mater. 321–323 (2006)1060.

[6] Medical electrical equipment — Characteristics of digital x-ray imaging devices — Part 1–2:Determination of the detective quantum efficiency — Detectors used in mammography, InternationalElectrotechnical Commission, IEC 62220-1-2 report, Switzerland (2008).

[7] J. Marchal, Theoretical analysis of the effect of charge-sharing on the detective quantum efficiency ofsingle-photon counting segmented silicon detectors, 2010 JINST 5 P01004.

[8] H.E. Nilsson et al., Monte Carlo simulation of charge sharing effects in silicon and GaAs photoncounting x-ray imaging detectors, IEEE Trans. Nucl. Sci. 51 (2004) 1636.

[9] J. Tanguay et al., The role of X-ray Swank factor in energy-resolving photon-counting imaging, Med.Phys. 37 (2010) 6205.

[10] P. Monnin et al., A comparison of the performance of digital mammography systems, Med. Phys. 34(2007) 906.

– 6 –