Computed tomography with single-shot dual-energy sandwich detectors

Transcription

1 Computed tomography with single-shot dual-energy sandwich detectors Seung Ho Kim, a Hanbean Youn, b,c Daecheon Kim, a Dong Woon Kim, a Hosang Jeon, b,c Ho Kyung Kim a,c a School of Mechanical Engineering, Pusan National University, Busan 46241, South Korea b Department of Radiation Oncology, Pusan National University Yangsan Hospital, Yangsan, Gyeongsangnam-do 50612, South Korea c Center for Advanced Medical Engineering, Pusan National University, Busan 46241, South Korea ABSTRACT Single-shot dual-energy sandwich detector can produce sharp images because of subtraction of images from two sub-detector layers, which have different thick x-ray converters, of the sandwich detector. Inspired by this observation, the authors have developed a microtomography system with the sandwich detector in pursuit of highresolution bone-enhanced small-animal imaging. The preliminary results show that the bone-enhanced images reconstructed with the subtracted projection data are better in visibility of bone details than the conventionally reconstructed images. In addition, the bone-enhanced images obtained from the sandwich detector are relatively immune to the artifacts caused by photon starvation. The microtomography with the single-shot dual-energy sandwich detector will be useful for the high-resolution bone imaging. Keywords: Computed tomography, microtomography, micro-ct, sandwich detector, dual-energy imaging, single-shot dual-energy imaging, mouse imaging 1. INTRODUCTION We previously described the multilayer ( sandwich ) detector by stacking two flat-panel detectors (FPDs) and demonstrated its prospect for motion-artifact-free single-shot dual-energy imaging (DEI) by obtaining bone and soft-tissue images of a postmortem mouse. 1, 2 While the front FPD measures relatively low energy, as shown in Fig. 1, the rear one measures relatively high energy because of x-ray beam hardening through the front FPD. Onto the same CMOS photodiode platform, thus, we placed a thicker scintillator in the rear FPD than the front one to achieve high quantum efficiency with the relatively higher-energy x-ray spectrum. An intermediate copper (Cu) filter can be used to further increase spectral separation between the two FPD measurements, which may provides a better contrast-to-noise performance in the subtracted images. As shown in Fig. 2, the conventional dual-shot image obtained by switching the applied tube voltages (40/70 kvp) showed superior signal-to-noise ratio (SNR) performance but the single-shot image obtained from the sandwich detector was almost as good and had the advantage of being less susceptible to motion artifacts. We also observed that the single-shot method showed better SNR at higher spatial frequencies (e.g. edge regions and bone details) than the double-shot method. The reason can be explained by the inherent unsharp masking effect of the sandwich detector; 2 the rear FPD with a thicker scintillator provides a blurrier image than the front FPD, hence subtraction of the two images enhances edges in the resultant image. Since the physical characteristics of bone are affected by numerous factors, such as age, hormones, arthritis, and exercise, 3 a preclinical study with small animals with respect to those various factors would be helpful for theoretical and computational models of bone. 4 Moreover, longitudinal, histological studies without sacrificing animals are essential. 5, 6 In these regards, the x-ray microtomography (or micro-ct) has been popular for smallanimal bone studies Inspired by the aforementioned observation, we have applied the sandwich-detector concept to the microtomography (or micro-ct) for small-animal bone imaging. hokyung@pusan.ac.kr; phone ; fax

2 Figure 1. Conceptual illustration of the sandwich detector operation. Figure 2. Comparison of dual-energy postmortem mouse images obtained from (a) dual-shot (40/70 kvp switching) and (b) single-shot (using the sandwich detector at 70 kvp) methods. 2. MATERIALS AND METHODS As shown in Fig. 3, we have developed a bench-top micro-ct system with the sandwich detector. During continuous x-ray irradiation, the object rotates on its axis by an amount of prescribed step angle and then the rotation stays until the sandwich detector produces two projection images. These motion and image readout were computer-controlled and lasted till a single rotation completed. The distances from the x-ray focal spot to the detector (dsd ) and to the axis of rotation (dsa ) were computer-controlled variables. The traveling ranges of the object jig were 600, 300, and 75 mm along the x, y, and z directions, respectively. The x-ray source (Series 5000 XTF5011, Oxford Instruments, Inc., US) employed a tungsten anode and could operate up to the maximum power of 50 Watts. The tungsten x-ray spectra were further tailored by an additional 1-mm thick aluminum filter. According to the manufacturer, the nominal focal-spot size was mm. Each FPD layer constituting the sandwich detector employed a combination of a Gd2 O2 S:Tb phosphor screen for conversion of x-ray into optical quanta and a photodiode array for detection of them. The thicknesses of the front and rear phosphors were 34 and 67 mg cm 2, respectively. The same photodiode arrays (RadEye1TM, Teledyne Rad-icon Imaging Corp., Sunnyvale, US) were used for the front and rear FPDs. The pixel pitch of the photodiode array was mm, and the magnification of the pixel pitch of the rear detector was negligible. The active area of the sandwich detector was mm2. For a small-animal imaging, we prepared a postmortem mouse phantom ( 40 g) by replacing blood by paraformaldehyde as shown in Fig. 3. Bone-enhanced tomographic images may be obtained by reconstructing bone-enhanced projection data fbone (r) = FDK {WPF PR }, (1) where Pj denotes the projection data in a matrix form obtained from the jth FPD layer and W is a diagonal matrix consisting of weighting factors determined at each projection angle w(θ). The operator FDK { } implies the approximate filtered backprojection operation11 with the Hann filter. The w(θ) was determined by minimizing

3 Figure 3. Picture describing the micro-ct system with the sandwich detector. The enlarged views show the sandwich detector and a postmortem mouse phantom. contrast between the soft tissue to be subtracted and background. 2 If the weighting factor is independent upon the rotation angle, W becomes a scalar w. Head part of a mouse was scanned using two different designs of sandwich detector; one design used no intermediate filter and the other used a Cu filter with a thickness of 0.3 mm. Irradiation x-ray spectrum was from a tungsten target at 50 kvp/1 mm aluminum filter. 360 projection views were obtained for a single circular scan and they were used for reconstruction. All the reconstructed images were calibrated into the Hounsfield units (HU) using separate scans of water phantom. 3. RESULTS AND DISCUSSION Figure 4 compares projection images obtained from the each FPD layer of the two designs of sandwich detectors (i.e. one design included a 0.3 mm-thick Cu filter and the other did not) and their resultant DE images for the postmortem mouse phantom. The images were displayed with the level of their mean value (µ) and a window of two times their standard deviation (σ) (other images below were displayed with the same level and window). It was observed that the projections from the front FPD were sharper than those from the rear FPD as the font FPD employed a thinner phosphor than the rear FPD. Comparing Figs. 4(b) with (e), use of the intermediate filter resulted in higher noise in the rear FPD image, and the reason could be explained by the reduction in the number of x-ray photons reaching the rear FPD due to the attenuation through the filter layer. Weighted logarithmic subtraction successfully provided bone-enhanced images as shown in Figs. 4(c) and (f). Tomographic images reconstructed using each projection dataset, as exemplary shown in Fig. 4, are summarized in Fig. 5. The characteristics observed from the projection data were well reflected into the tomographic images. Comparing Figs. 5(c) with (f), use of the intermediate filter gave rise to a more reduction of soft tissues. It was also observed that the DE tomographic images showed less streak artifacts due to photon starvation compared to the images obtained using the front FPD. As shown in Fig. 6, the regions indicated as boxes in Fig. 5(d) have been investigated in detail. The rear FPD with a thick phosphor provided a blur image as shown in Fig. 6(b), and it became noisier, as shown in Fig. 6(e), when the Cu filter was additionally used. The DE tomographic image obtained form the sandwich detector without any filter layers showed the best visual performance for bone details. Figure 7 compares profiles extracted along the line A A (as possible as we can), as designated in Fig. 5(a), for each reconstruction image. Bone signal in the reconstructed image with projections from the front FPD was the largest, the rear FPD the second, and then the DE results. Although the DE tomographic image signals were noisy, they consisted mostly of bone signals. As observed from Figs. 4(e) and 5(e), the images obtained from the rear FPD of the sandwich detector with the Cu filter were noisy.

4 Figure 4. Projection images obtained from the each FPD layer of the sandwich detectors without and with a Cu filter layer and their resultant DE images for the postmortem mouse phantom. (L/W = µ/ ± 2σ).

5 Figure 5. Comparison of tomographic images reconstructed from the corresponding projection dataset. Figure 6. Enlarged images indicated by the boxes Fig. 5(d) for more detailed displays of tomographic images.

6 Figure 7. Profiles extracted along the line A A, as designated in Fig. 5(a), for each corresponding reconstruction image. In the present work, the bone-enhanced tomographic image was obtained by reconstructing weighted-subtraction sinogram. Alternatively, the bone-enhanced images may also be obtained by subtraction of two tomographic images reconstructed with projections obtained from the front and rear FPDs f bone (r) = wfdk {P F } FDK {P R }. (2) This approach is under progress and comparison with the present method will be made. Furthermore, the boneenhanced DE tomographic images may be combined with the images reconstructed with projections obtained from the front FPD as the conventional unsharp masking digital image processing. The authors anticipate the resultant images will be conventional tomographic images which include more pronounced bone details. These further studies will be a separate future study. 4. CONCLUSION Bone-enhanced tomographic images have been obtained using dual-energy sandwich detectors for a postmortem mouse phantom, and they outperformed the tomographic images obtained from the conventional detectors (i.e. the front and rear flat-panel detectors constituting the sandwich detectors) for bone details. Although use of an intermediate filter, which was placed between the front and rear flat-panel detectors, resulted in less residual soft tissues in the reconstructed bone-enhanced images, it degraded the visual image quality of bone details because of increased noise. Optimal filter design in terms of material and thickness is required for a more tissue separability and less noise performance in images. ACKNOWLEDGMENTS This work was supported by the National Research Foundation (NRF) grants funded by the Korea government (MSIP) (No. 2013M2A2A and No. 2014R1A2A2A ). REFERENCES 1. S. Yun, J. C. Han, D. W. Kim, H. Youn, H. K. Kim, J. Tanguay, and I. A. Cunningham, Feasibility of active sandwich detectors for single-shot dual-energy imaging, Proc. SPIE 9033, pp T 90335T 8, 2014.

7 2. J. C. Han, H. K. Kim, D. W. Kim, S. Yun, H. Youn, S. Kam, J. Tanguay, and I. A. Cunningham, Singleshot dual-energy x-ray imaging with a flat-panel sandwich detector for preclinical imaging, Cur. Appl. Phys. 14(12), pp , J.-T. Hsu, Y.-J. Chen, J.-T. Ho, H.-L. Huang, S.-P. Wang, F.-C. Cheng, J. Wu, and M.-T. Tsai, A comparison of micro-ct and dental CT in assessing cortical bone morphology and trabecular bone microarchitecture, PLoS ONE 9, p. e107545, N. A. Hanson and C. M. Bagi, Alternative approach to assessment of bone quality using micro-computed tomography, Bone 35(1), pp , E. L. Ritman, Molecular imaging in small animals - Roles for micro-ct, J. Cell. Biochem. 87(S39), pp , S. C. Lee, H. K. Kim, I. K. Chun, M. H. Cho, S. Y. Lee, and M. H. Cho, A flat-panel detector based micro- CT system: Performance evaluation for small-animal imaging, Phy. Med. Biol. 48(24), pp , M. J. Paulus, S. S. Gleason, S. J. Kennel, P. R. Hunsicker, and D. K. Johnson, High resolution x-ray computed tomography: An emerging tool for small animal cancer research, Neoplasia 2(1-2), pp , D. W. Holdsworth and M. M. Thornton, Micro-CT in small animal and specimen imaging, Trends Biotechnol. 20(8), pp. S34 S39, E. L. Ritman, Micro-computed tomography - Current status and developments, Annu. Rev. Biomed. Eng. 6(1), pp , D. Clark and C. Badea, Micro-CT of rodents: State-of-the-art and future perspectives, Phys. Medica 30(6), pp , L. A. Feldkamp, L. C. Davis, and J. W. Kress, Practical cone-beam algorithm, J. Opt. Soc. Am. A 1, pp , Jun 1984.