Key Engineering Materials Vols. 70-73 (004) pp. 45-51 online at http://www.scientific.net 004 Trans Tech Publications, Switzerland Citation & Copyright (to be inserted by the publisher) Development and Characterization of a Flat-panel Detector-based Microtomography System Sang Chul Lee 1, Ho Kyung Kim, In Kon Chun 1, Myung Hye Cho 1, Min Hyoung Cho 1 and Soo Yeol Lee 1,* 1 Graduate School of East-West Medical Science, Kyung Hee University, Yongin 449-701, Korea School of Mechanical Engineering, Pusan National University, Pusan 609-735, Korea Keywords: CMOS, Flat-panel Detector, Microtomography, 3-D Imaging Abstract. In this paper, we describe the development and performance evaluation of a 3-dimentional (3-D) high-resolution x-ray microtomography (micro-ct) system. Unlike a conventional micro-ct, the developed system uses a flat-panel detector as a digital x-ray imager. The detector is a CsI:Tl (thallium-doped cesium iodide) scintillator coupled to an active-matrix photodiode array with a pixel pitch of 50 [µm]. Without geometric magnification, the spatial resolution of the detector is 7 [lp/mm] at 10 [%] of MTF (modulation-transfer function). The overall efficiency of the detector for the input x-ray signal-to-noise ratio (SNR) has been measured to be about 50 [%] with the x-ray source operating at 60 [kvp] and 1-mm-thick Al filtration. For fast 3-D cone-beam image reconstruction, the Feldkamp algorithm has been realized in a distributed parallel processing system composed of multiple personal computers. The signal and noise properties in tomograms have been measured with quantitative phantoms and the measurement results are found to conform well to the theoretical models. From the measurements, it has been also found that the spatial resolution in a tomogram is almost determined by the detector resolving power. Some high-resolution imaging results are shown to demonstrate the capability of the developed system in bio-medical and industrial applications. Introduction High-resolution imaging modalities have been greatly developed for small-animal imaging in bio-medical studies [1]. With the advent of desktop micro-focus x-ray sources and highly sensitive x-ray detectors, x-ray micro computed tomography (micro-ct) becomes one of relevant small-animal imaging modalities. In addition to bio-medical applications, the micro-ct can also be greatly used for industrial purposes such as non-destructive evaluation (NDE) of solder joints in the integrated printed circuit boards (PCBs) and ball-grid-array (BGA) packages []. In NDE of PCBs or BGA packages in a quality control procedure, 3-dimensional (3-D) imaging is very essential to detect the tiny flaws that cannot be visualized in conventional x-ray projection system. In a micro-ct, the x-ray detector is the most important component on which the micro-ct performance is largely relied. In previous studies [3,4], a lens-coupled CMOS detector was used in bio-medical and NDE applications. However, it had poor image quality due to very low lens-coupling efficiency (~0.05 [%]). In this study, we have applied a flat-panel detector to a micro-ct that has cone-beam imaging geometry. We briefly describe the micro-ct system design in terms of detector selection. Performance evaluation results of the complementary metal-oxide-semiconductor (CMOS) flat-panel detector in a micro-ct are presented in comparison with a lens-coupling method. The signal and noise characteristics of the developed micro-ct system are addressed with theoretical models and corresponding experimental results. Imaging performances of the developed micro-ct are also demonstrated with small sample imaging results. * Correspondence to sylee01@khu.ac.kr Licensed to KYUNGHEE U (sylee01@khu.ac.kr) - Yongin, Kyungki - Korea, Republic of All rights reserved. No part of the contents of this paper may be reproduced or transmitted in any form or by any means without the written permission of the publisher: Trans Tech Publications Ltd, Switzerland, www.ttp.net. (ID: 163.180.145.148-8/08/04,06:01:05)
46 Advances in Nondestructive Evaluation Fig. 1. A schematic diagram of the developed micro-ct system. The main components are a micro focus x-ray source, a rotational sample holder, and a CMOS flat-panel detector. The source-to-detector distance (SDD) and source-to-object distance (SOD) determine the system geometric parameters such as magnification ratio, cone-beam angles (φ 1 and φ ), and field-of-view (FOV) in imaging. Image reconstructions are performed with a parallel computing system consisting of four PCs, which is not shown in this diagram. System Design Main components of the system are a microfocus x-ray source, a rotational sample holder, and a -D x-ray imager. The x-ray source continuously irradiates a sample and the CMOS detector acquires -D projection data at a given frame (integration) time. The angular velocity of the sample holder and the detector integration time are computer-controlled. The microfocus x-ray source (L811-01, Hamamatsu, Japan) is a sealed tube with a fixed W target and 00-µm-thick Be exit window. The emitted x-ray beam angle is 43 [ ]. The source has a variable focal spot size from 5 [µm] to 50 [µm] depending on the applied tube power (kvp ma). An additional filtration with a 1-mm-thick Al plate was introduced. The digital imager (C794, Hamamatsu, Japan) is a matrix-addressed photodiode array fabricated by 0.6 [µm] CMOS process to read optical photons emitted from the overlying CsI:Tl (thallium-doped cesium iodide) scintillator. The fill factor, fractional area of the photo-sensitive region in a pixel, is as high as 79 [%] in spite of the small pixel size of 50 [µm]. The active area of the detector is 10 10 [mm ]. The CsI:Tl has a columnar structure with a diameter of about 10 [µm] and the thickness of 00 [µm]. The packing density is about 80 [%]. For fast volume image reconstruction with the cone-beam projection data, a parallel data processing system has been realized with four personal computers, each one equipped with dual CPUs (Athlon MP 00+, AMD, US). A host computer, one of the four personal computers, controls the whole system and evenly distributes processing tasks to all the computers including itself. For the image reconstructions, we adopted Feldkamp cone-beam algorithm [5] filtering the projection data with Ram-Lak filter, which is a band-limiting filter derived from the ramp function [6]. A schematic diagram of the developed system is shown in Fig. 1. The micro-ct system has been typically operated at the tube voltage of 60 [kvp]. System geometrical magnification ratio has been fixed to (source-to-detector distance, SDD = 49 [mm]). Radiation doses, addressed in this study, are based on the measurement of exposure in the air at the position of axis-of-rotation (AOR) with a calibrated ionization chamber (Victoreen 6000-58, Innovision, US). To take into account the dosimetry aspects, the measured exposures are expressed in the air kerma with an exposure-to-air kerma conversion factor of 8.767 [mgy/r] at an x-ray spectrum of 60 [kvp]. The frame integration time of the CMOS detector to acquire a single projection data is normally 50 [ms].
Key Engineering Materials Vols. 70-73 47 Physical Imaging Characteristics of the CMOS Flat-panel Detector We have evaluated the spatial resolution of the detector in terms of modulation-transfer function (MTF) that is the Fourier transform of the line-spread function (LSF). MTF describes the contrast resolution with respect to the spatial frequency. In order to measure the aliasing-free MTF of the detector, we obtained a projection image of a 10-µm-wide slit (I.I.E. GmbH, Aachen, Germany) with the slit tilted by. [ ] to the vertical axis. With the tilted slit image we can obtain a cut view of the slit with the sampling frequency enough to avoid the aliasing effect. Fig. shows the measured MTF, with an MTF of a lens-coupled CMOS detector for a comparison. The resolving powers of the flat-panel and lens-coupled detectors are up to ~7 [lp/mm] and up to ~ [lp/mm] at 10 [%] of MTF, respectively. It is noted that the lens-coupled CMOS detector has much poor MTF than the small pixel pitch of 10 [µm] could have. It is due to the low coupling efficiency of the lens system in a high spatial frequency range. Fig.. MTF curves of two detector configurations. The flat-panel detector and the lens-coupled detector have pixel pitches of 50 [µm] and 10 [µm], respectively. Fig. 3. QADs as a function of stage for the two detector configurations. Inefficiency of the lens-coupling method can be explained by observing a quantum accounting diagram (QAD), which is a plot of the effective number of quanta at a given spatial frequency as a function of physical stage number [7]. In an x-ray spectrum from W target operated at 60 [kvp], we calculated QAD considering many physical parameters, such as quantum absorption efficiency, scintillation gain, escape probability, coupling efficiency, charge conversion efficiency, and fill factor. Parameters regarding the scintillator were based on the Monte Carlo simulation and other parameters were referred to the published data [8,9]. As shown in Fig. 3, there are two quantum sinks in the lens-coupled detector configuration: one at the x-ray absorption stage in the scintillator and the other at the light collecting stage as expected. Because information-carrying quanta are lost irrecoverably in each quantum sink, the quantum sink limits the signal-to-noise ratio (SNR) of the whole system. On the contrary, the flat-panel detector configuration shows only one quantum sink at the first stage, which is desirable in imaging systems. Since the detective quantum efficiency (DQE) is a quantum utilization efficiency of an imaging system, i.e., effective fraction of incident x-ray photons contributing to image SNR, the DQE is now believed to be a most objective measure of imaging system performance. In order to calculate the DQE, measurement of noise-power spectrum (NPS) is necessary. For the NPS evaluation, the image obtained with uniform irradiation on the detector was divided into 100 non-overlapping sections, each of which was 18 18 in matrix size. After correcting low-frequency drifts in each image section, we computed the -D Fourier transform of the detected image signal in each section. The process was applied to every section and the -D Fourier transform results were then averaged to get NPS. Based on the MTF and NPS measurements, we calculated the DQE as shown in Fig. 4. For a comparison, the
48 Advances in Nondestructive Evaluation DQE calculation based on the Monte Carlo simulation with the cascaded linear-systems theory [7] is also shown, which conforms to the measurements quite reasonably. Through the Monte Carlo analysis of the CsI:Tl scintillator, it was found that the quantum absorption efficiency, A Q = 0.47 and the Swank factor, A S = 0.84 when the energy absorption fluctuation was only taken into account. Since the DQE at zero-spatial frequency is calculated by A Q A S = 0.40, we can note that the overall SNR performance of the CMOS flat-panel detector is governed by the CsI:Tl scintillator. The low value of A Q in this study even with the x-ray spectrum of 60 [kvp] is due to the 1-mm-thick Al protection enclosure of the detector. If we replace the Al enclosure with a radiation-transparent material, e.g., polycarbonate, we could obtain a further enhancement in the DQE as depicted in Fig. 4. Fig. 4. Measured and calculated DQE curves for an x-ray spectrum of 60 [kvp] and an irradiation of 17 [mr]. Open circles show the meaured DQE result. Solid line is the calculated DQE based on the cascaded linear-system theory. Dotted line is the expected DQE in the new detector design. Fig. 5. SNR as a function of the absorbed dose in water. The theoretical SNR model agrees well with the measurements obtained from the water phantom having a diameter of 8 [mm]. Tomographic Imaging Performances Image quality in a quantum-limited imaging system such as the flat-panel detector in this study is mainly determined by the interacting x-ray photon statistics. Therefore, analysis of statistical variations in CT image reconstruction is important in determining the imaging performance of the developed system. With a water-filled bath having a diameter, d = 8 [mm], the SNR has been measured with respect to the dose. In order to analyze the measured data, we have calculated the SNR by using the theory [10], which was originally developed based on the parallel-beam acquisition geometry and later was verified to work also well in fan-beam and cone-beam geometries [11]. In CT reconstruction using the Ram-Lak filter, the statistical variance model of linear attenuation coefficient, σ µ, can be expressed in terms of the isotropic voxel size, a, and total number of photons during a scan, N: π =, (1) 1 a N σ µ where N is related to the absorbed dose, D as follows: µ d / p h AQ D e N =. () E ( µ / ρ) en In Eq., E is the photon energy, µ the linear attenuation coefficient, µ en /ρ the mass energy-absorption coefficient, p the pixel aperture of the detector, and h the slice thickness. In this study, the photon
energy was assumed to be monoenergetic at 30 [kev]. Fig. 5 shows both the measured and calculated SNR or µ/σ µ as a function of dose. The doses in the measurements were based on the air kerma and were corrected with the µ en /ρ of water. The theory agrees well with the measured data. The slight difference is presumably due to the use of the ideal pixel pitch and slice thickness in the calculation. Contrast-to-noise ratio (CNR) was analyzed with a quantitative phantom, which consists of six low-contrast inserts with the diameter of 5 [mm] immersed in a water bath. The inserts are made of commercial electronic density phantoms (Model 76-430, Nuclear Associates, NY, US) such as plastic water (1.03 [g/cm 3 ]), nylon (1.15 [g/cm 3 ]), polyethylene (0.95 [g/cm 3 ]), acryl (1.18 [g/cm 3 ]), polystyrene (1.11 [g/cm 3 ]), and polycarbonate (1.18 [g/cm 3 ]). A transaxial image of the contrast phantom obtained at 86 [mgy] of air kerma is shown in Fig. 6. From the acquired images we have calculated the CNR as a function of dose. From the measurement results, we can notice that the developed micro-ct can differentiate less than 36 CT numbers at the dose of 95 [mgy]. The spatial resolving power of the developed micro-ct was evaluated with the phantom made of an 18-µm-thick Al foil attached on an acrylic plate. The evaluation method is the same as the work by Boone [1]. Fig. 7 shows the measured total system MTF of the system at the magnification ratio of. The total system MTF is affected by various physical parameters such as the focal spot size, the magnification ratio, the reconstruction algorithm, and the detector resolving power. As derived by Holdsworth et al [11], each MTF component was calculated and the total MTF was estimated as follows: MTF system ( f ) = MTF = sinc{ f MTF focal ( f ) MTF b ( M 1) / M} MTF detector Key Engineering Materials Vols. 70-73 49 ( f ) detector ( f ) MTF detector recon ( f ) ( f ) sinc ( f a), (3) where f is the spatial frequency, b the focal spot size, and M the magnification ratio. The estimation agrees well with the measurement. From the result and Eq. (3), we can notice that the system MTF is almost determined by the detector resolving power. Fig. 6. A transaxial image of the low-contrast phantom obtained with air kerma of 86 [mgy]. Fig. 7. Meaured MTF curve of the micro-ct system (circle). The predicted MTF curve (solid line) is also shown to account for the physical parameters affecting the system MTF. The most significantly affecting parameter is the detector MTF (dotted line). Applications The developed micro-ct has been applied to both bio-medical and NDE stuides. In Fig. 8(a) we have shown a 3-D rendered image of a rat femur. Due to the small voxel size of 0 0 50 [µm 3 ], we can clearly observe the fine bony structures inside the femur. High-resolution bone imaging is very important in osteoporosis studies. In Fig. 8(b), we have shown a 3-D rendered skeletal image of a rat. Owing to the large field of view of the flat panel detector, we can image the whole volume of the rat in
50 Advances in Nondestructive Evaluation a scan. The voxel size in the image is 60 60 00 [µm 3 ]. In addition to the bony structure imaging, the mirco-ct can be applied to many other areas, such as vascular imaging and soft tissue imaging with aids of contrast agents and small-animal PET/CT for attenuation correction. Fig. 8(c) shows a 3-D rendered image of a gullwing connector on a printed circuit board (PCB). The voxel size is 15 15 5 [µm 3 ]. We can clearly see the 3-D shape of lead solderings on the connector. In Fig. 8(d), we show a cross-sectional image of the gullwing connector. The image was taken at the interface between the connector and the PCB to which lead soldering had been applied. The voxel size is 15 15 15 [µm 3 ]. Summary A CsI:Tl coupled CMOS flat-panel detector has been successfully applied to the micro-ct system. Through the experimental measurements and theoretical calculations, we have found that the overall performances of the detector such as the spatial resolution and DQE are mainly determined by the first x-ray detection stage, the CsI:Tl converting layer. The spatial resolution of the system is governed by the detector resolving power in the present geometry and configuration. The minimum resolvable contrast of the micro-ct is less than 3.6 [%] at about 1.0 [%] of the LD 50/30 (whole-body radiation dose that would kill 50 [%] of an exposed population within 30 days of exposure) level if we assume that LD 50/30 = 10 [Gy] for a mouse. In order to reduce the voxel noise and improve the CNR without increasing radiation dose, enhancement of the detector quantum efficiency is very essential. In future studies, we plan to adopt flat-panel detectors optimized for specific imaging tasks to improve the performances. Fig. 8. (a) A 3-D rendered image of a rat femur. (b) A 3-D rendered skeletal image of a rat. (c) A 3-D rendered image of a gullwing connector on a PCB. (d) A cross-sectional image of the gullwing connector whose cross-sectional plane is shown as the dotted line in (c). Acknowledgement This work supported by a grant of KISTEP, Republic of Korea (M-0305-04-0005).
References Key Engineering Materials Vols. 70-73 51 [1] R. Weissleder and U. Mahmood: Radiology Vol. 19 (001), p. 316. [] G.C. Neubauer: IEEE Trans. Comp. Packag. Manufact. Technol. Vol. 0 (1997), p. 111. [3] S.W. Lee, H.K. Kim, G. Cho, Y.H. Shin and Y.Y. Won: IEEE Nucl. Sci. Vol. 48 (001), p. 1503. [4] H.K. Kim, S.C. Jeon and G. Cho: Conference Record of 001 IEEE Nucl. Sci. Symp. San Diego, Nov. (001). [5] L.A. Feldkamp, L.C. Davis and J.W. Kress: J. Opt. Soc. Am. A. Vol. 1 (1984), p. 61. [6] G.N. Ramachandran and A.V. Lakshminarayanan: Proc. Natl. Acad. Sci. Vol. 68 (1971), p. 36. [7] I.A. Cunningham, M.S. Westmore and A. Fenster: Med. Phys. Vol. 1 (1994), p. 417. [8] H.K. Kim, G. Cho, S.W. Lee, Y.H. Shin and H.S. Cho: IEEE Nucl. Sci. Vol. 48 (001), p. 66. [9] S. Hejazi and D.P. Trauernicht: Med. Phys. Vol. 4 (1997), p. 87. [10] H.H. Barrett, S.K. Gordon and R.S. Hershel: Comput. Biol. Med. Vol. 6 (1976), p. 307. [11] D.W. Holdsworth, M. Drangova and A. Fenster: Med. Phys. Vol. 0 (1993), p. 449. [1] J.M. Boone: Med. Phys. Vol. 8 (001), p. 356.