Optimization of a Silicon Drift Detector Module for a Gamma Camera

Transcription

1 Journal of the Korean Physical Society, Vol. 54, No. 2, February 2009, pp Optimization of a Silicon Drift Detector Module for a Gamma Camera Byung Jun Min, Yong Choi, Tae Yong Song, y Seung Han Shin, Jin Ho Jung, Key Jo Hong, Joon Young Choi, Yearn Seong Choe, Kyung-Han Lee and Byung-Tae Kim Department of Nuclear Medicine, Samsung Medical Center, Sungkyunkwan University School of Medicine, Seoul Young Bok Ahn Department of Electronic Engineering, Konkuk University, Seoul (Received 24 December 2008) The aim of this study was to optimize the design parameters of a silicon drift detector (SDD) module for a gamma camera. The examined design parameters were the reection coecients, the crystal size, the energy threshold eect, the light guide thickness and the sensor packing ll factor. The silicon drift detector module proposed in this study was composed of hexagonally packed 72 SDDs and a single-slab NaI(Tl) crystal with a light guide that was optically coupled to the SDD sensors. Monte Carlo simulations using DETECT2000 were performed to generate necessary data for the optimization. The Poisson noise was calculated analytically and was applied to the simulation. The eects of the reection coecients (RC) of the crystal's side surfaces and of the crystal size were examined. A Cramer-Rao (CR) lower bound analysis was conducted to determine the light guide thickness and the energy thresholds providing best spatial resolutions were estimated. A white reector (RC = 0.9) on the side surface improved the light collection eciency (LCE) versus black paint (RC = 0.1) while the spatial resolution was less sensitive to the reection coecient. We found that crystal size should be smaller (about 4 mm) than the SDD size to prevent event overlapping. Though a thinner light guide provided a better CR lower bound, a 2-mm light guide was suggested for the mechanical support and for NaI(Tl) crystal encapsulation. The intrinsic spatial resolution was best at a 2 % energy threshold regardless of the noise level. An inter-sdd gap of 2-mm was chosen to obtain a ll factor of 95 % without any technical problems in the SDD manufacturing. The results of this study will be useful in the development of a compact high-resolution gamma camera based on an array of SDDs coupled to a single-slab scintillator. PACS numbers: Mc, Wk, Pw, Pm Keywords: Scintillator, Silicon drift detector, Gamma camera, Monte Carlo simulation I. INTRODUCTION Silicon drift detectors (SDDs) have the primary advantages of high quantum eciency, compact physical size, low output capacitance and good energy resolution, which make them attractive for gamma camera applications [1{10]; hence, they could potentially replace photomultiplier tubes (PMTs) in high-resolution gamma cameras [11{16]. Our group developed a 7-channel SDD detector module with a high spatial resolution of 0.74-mm full width at half maximum (FWHM) [17]. The detector module consists of an array of hexagonally-packed SDDs (hexagon shaped with about 1 cm in diameter) which is optically coupled to a single-slab scintillator. ychoi@skku.edu; Tel: ; Fax: ; y Current address: Mallinckrodt Institute of Radiology, Washington University in St. Louis, Missouri, U.S.A Since a compact and high-resolution gamma camera detector module employing 72 SDDs is being developed, an investigation of the design parameters that sustain the advantage of the high spatial resolution of the SDD module was required. The design parameters of the scintillation crystals utilized in conventional PMT-based gamma cameras are well known [18{20]. However, these parameters may be unsuitable for SDD gamma cameras because of the drawbacks of SDD sensors, which require optics optimization. The drawbacks of the SDDs are a low signal-to-noise ratio (SNR) and gain uctuations due to changes in the bias voltage and the temperature. In addition, their lower intrinsic SNR means that a fast low-noise preampli- er is required at each SDD output [21]. High-intrinsicspatial-resolution SDD sensors could promote the introduction of an improved scintillation crystal design unsuitable for conventional PMT. Additionally, since gamma cameras based on PMT detector modules have

2 -598- Journal of the Korean Physical Society, Vol. 54, No. 2, February 2009 Fig. 2. Simulation geometry used to examine the eect of the reection coecient on the spatial resolution and the energy resolution. Light sources were generated at ve positions; position 1 was located at the edge of the crystal and the pitch of each light position was 4-mm. Fig. 1. The gamma camera detector module consists of 72 hexagonal SDDs. The inter-sdd gap has a 3-mm dead space. Spatial resolution and light collection studies were performed using a Monte Carlo simulation tool, DETECT. very high SNRs, all PMT signals from a detector module are used to calculate the gamma-ray event localization and the energy resolution. For SDD gamma camera detector modules, however, each SDD sensor has a relatively high electronic noise, which may be altered by bias voltage, shaping time and temperature. Furthermore, a suitable restriction on the use of SDD signals is necessary to minimize the serial (voltage) and the parallel (current) noise. For the above reasons, a new design is needed to obtain the best performance when SDD array modules are used. The aim of this study was to optimize the design parameters for a gamma camera detector module consisting of 72 channel SDD detectors coupled to a NaI(Tl) crystal via a light guide. The reectivity of the scintillation crystal's side surfaces, crystal size, eect of the energy threshold on the spatial resolution, the light guide thickness and the sensor packing ll factor were investigated using Monte Carlo simulation studies. II. MATERIALS AND METHODS 1. Simulation Geometry reectivity and crystal size on the spatial and the energy resolution. The thickness of the optical grease between the NaI(Tl) crystal and the light guide was xed at 0.1- mm and its refractive index was Light sources were generated at a 2.9-mm depth from the top surface of the crystal, which was the mean interaction depth that provided a 50 % detection possibility. 2. Silicon Drift Detector Electrical Noise The leakage current of each SDD, as simulated in this study, was about 3 na/cm 2 with a total capacitance of 2 pf at room temperature. The Poisson distribution noise was calculated as the equivalent noise charge (ENC) by using the leakage current and the capacitance information as shown in Eq. (1) [24{26] and this was applied to simulation results: ENC 2 = A1 2KT gm C2 1 T +A2 2 f C 2 T + b f + A3qI L ; (1) 2 where A is the coecient of the CR-shaper, K is Boltzmann's constant, T is the temperature (Kelvin), gm is the transconductance, is the excess noise factor, C T is the total capacitance, a f and b f are frequency noise factors, q is the unit charge, I L is the leakage current and is the shaping time. The eect of temperature on the electronic noise was estimated analytically. The silicon drift detector (SDD) module proposed in this study consisted of a 72 hexagonal SDD array (Figure 1) coupled to a 10-mm-thick NaI(Tl) crystal. The sensitive area of the SDD array was 114 mm mm. The crystal light guide thicknesses were changed from 1 to 6 mm. A Monte Carlo simulation using DETECT2000 [22, 23] was conducted to investigate the eect of side surface 3. Reectivity of the Scintillation Crystal's Side Surfaces Five locations, shown in Figure 2, were selected to study the eect of the crystal's side surface reectivity on the spatial resolution and on the light collection eciency (LCE). A total of 2,000 gamma-ray interactions

3 Optimization of a Silicon Drift Detector Module for a Gamma Camera { Byung Jun Min et al Table 1. Light collection eciency at gamma detection positions 1 to 5. RC Position 1 Position 2 Position 3 Position 4 Position % 46 % 54 % 58 % 60 % % 61 % 66 % 69 % 70 % % 81 % 83 % 84 % 84 % 2 lb(x) 0 B@ nx 1 i(x) S i CA 1 ; (3) (x) Fig. 3. Monte Carlo simulation geometry to determine the crystal size. Case 1 illustrates an overhanging type and the scintillation crystal size was decreased to Case 4. were generated at ve gamma detection positions 4-mm apart. The LCE is dened by Collected P hotons LCE = 100 % (2) Generated P hotons Black absorber, half reector and white reector on the side surfaces were simulated using reection coecients (RC) of 0.1, 0.5 and 0.9, respectively. The eects of side surface reectivity on the LCE and the gamma-ray event overlapping were investigated. where S i (x) is the mean output of the i-th SDD as a function of the scintillation position x, which is known as a light response function. The S i (x) function was smoothed to avoid excessive noise in its derivative. The light response function, an input of the CR lower bound calculation, was obtained when light guide thicknesses were changed from 1 to 6 mm. 6. Energy Thresholds The total sum of all SDD signals in one detector module was dened as an energy signal. The cut-o value of an energy signal was dened as the energy threshold that excluded SDD signals lower than the reference value. The energy threshold values providing the best spatial resolution were determined using a line prole analysis. Since the SDD sensor had a relatively high electronic noise when the bias voltage and the temperature were changed, two dierent noise levels of 50 and 100 electrons were added to each threshold image. 4. Determination of the Crystal Size The eect of crystal size was investigated by changing the size from an overhanging type (Case 1) to a ush type (Case 4), as demonstrated in Figure 3. The side surface simulated as a white reector (RC = 0.9) provided the best results (Section III. 1). The overlap of gamma ray events in the crystal edges was examined and the spatial resolutions of Case 1 to Case 4 were compared 7. Fill Factor The SDD array has a dead space due to inter-sdd gaps (Figure 1) and this dead space aects the light collection eciency. The LCE was estimated when the inter-sdd gap was changed from 0 mm to 3 mm. The acceptable ll factor was selected to be 95 % in this study. III. RESULTS AND DISCUSSION 5. Light Guide Thickness A Cramer-Rao (CR) lower bound analysis [27] using Eq. (3) was applied to determine optimum light guide thickness: 1. Reectivity of the Scintillation Crystal's Side Surfaces A white reector (RC = 0.9) on the side surface markedly improves the light collection eciency (LCE),

4 -600- Journal of the Korean Physical Society, Vol. 54, No. 2, February 2009 Fig. 4. The eect of the reection coecient (RC) on the side surface of the scintillation crystal. The RC was changed from 0.1 (top), to 0.5 (middle), to 0.9 (bottom). Planar images of 5 point sources (left) and their line proles (right). Fig. 5. The eect of the crystal size on the spatial resolution and the event overlapping. The side surface was simulated as a white reector (RC = 0.9). Cases 1 to 4 shows the planar images (left) and their line proles (right). which aects the energy resolution (Table 1). For example, the LCE improved from 60 % to 84 % by changing the RC from 0.1 to 0.9 at position 5 (Figure 4). The side surface reectivity of the NaI(Tl) scintillation crystal did not aect event overlapping at the crystal edges, as demonstrated in Figure 4. The white re- ector treatment on the side surface of the scintillation crystal provided slightly better spatial resolution than treatment with a black absorber. The intrinsic spatial resolution for a white reector was slightly better than that for a black absorber with values of 0.72 and 0.75 mm full width at half maximum (FWHM), respectively, at position 5. Hence, the side surfaces of the scintillation crystals should be coated with a white reector to improve the spatial resolution and the LCE. 2. Determination of the Crystal Size The eect of the crystal size on the spatial resolution is demonstrated in Figure 5. The gure shows that test points 1 to 3 (Figure 3) overlap each other and cannot be resolved. Half of the SDD size (8 mm) from the detector edge could not be used for event localization due to limitations of the Anger-logic positioning algorithm. The line proles illustrate that oversized scintillation crystals cause event overlapping and an edge blurring eect. Furthermore, light positions 1 and 2 intruded upon light position 3 in Cases 1 and 2. However, light position 3 in Case 3 was not overlapped because there was no outer light contamination. This result indicates that the crystal size should not be larger than the SDD area and, in fact, that it needs to be little smaller. Smaller sized (about 4 mm from the SDD edge) scintillation crystals were favorable in preventing unwanted event overlapping and edge blurring. Furthermore, smaller crystal sizes reduce manufacturing costs. 3. Light Guide Thickness Figure 6 shows the Cramer-Rao (CR) lower bound as a function of the light guide thickness between a crystal and the SDD array. The result suggests that the light guide, located between the crystal and the SDD, needs to be as thin as possible. Although the scintillation light must be spread for the Anger-logic positioning algorithm, excessive spread

5 Optimization of a Silicon Drift Detector Module for a Gamma Camera { Byung Jun Min et al Fig. 6. Cramer-Rao (CR) lower bound as a function of the light guide thickness. Fig. 8. The eect of inter-sdd gap on the light collection eciency. encapsulation. 4. Energy Threshold An accurate positioning estimate is an important consideration for a high-resolution gamma camera. If all the SDD signal outputs contribute to the position determination, the spatial resolution could be deteriorated due to the low signal-to-noise ratio (SNR) of the SDDs. Therefore, an energy threshold was employed to eliminate some of the noisy SDD signal from the positioning computation. The accuracy of event localization was improved by applying an energy threshold of up to 2 %, as shown in Figure 7. At higher energy thresholds, however, the positioning algorithm suered from event localization error due to a lack of light-sharing information. An energy threshold value of 2 % provided the best intrinsic spatial resolution with electrical noise when a 100- mm 2 SDD and a 10-mm thick scintillation crystal were used. The intrinsic spatial resolution at the center eldof-view (FOV) at a 2 % energy threshold was 0.79-mm in FWHM. Fig. 7. The eect of the energy threshold. The energy threshold value was changed from 0 % to 5 %. The 2 % energy threshold of total light collection provided the most accurate positional information. could deteriorate the positioning accuracy. Since the sensitive area of a SDD sensor is small enough (100 mm 2 ), the results indicate that light broadening is not necessary, which contrasts with the situation for a conventional large PMT ( mm 2 ). A minimum 2-mm light guide for the SDD detector module is required for mechanical support and for NaI(Tl) crystal 5. Fill Factor Figure 8 shows the eect of the inter-sdd gap on the relative light collection eciency. The ll factor was 99 % with a 1-mm inter-sdd gap and 94 % with a 2-mm gap. The inter-sdd gap was investigated to provide a reference value for a SDD manufacturer. Although a smaller inter-sdd gap results in a higher ll factor, it is technically dicult to minimize the gap. An inter-sdd gap of 2 mm is feasible without SDD performance degradation and allows a ll factor of about 95 % to be obtained.

6 -602- Journal of the Korean Physical Society, Vol. 54, No. 2, February 2009 IV. CONCLUSION The purpose of this study was to optimize the design parameters of a silicon drift detector (SDD) module by using a Monte Carlo simulation. The SDD module could be utilized to develop a high-performance gamma camera and single photon emission computed tomography (SPECT), which are widely used to obtain functional and molecular images in experimental small animals and humans. Scintillation crystal reection coecients, sizes, light guide thicknesses, energy threshold eects and sensor packing ll factors were investigated. The results of this study indicate that the devised scintillation crystal design was optimum with a white reector on the side surface, a 2-mm light guide thickness and 4-mm smaller size from the edge of the SDD. The energy threshold should be 2 % for best event localization and the dead space between SDD sensors should be less than 2-mm in order to obtain a ll factor of over 95 %. The results of this study will be useful in the development of a gamma camera using a high-resolution photosensor, for example, a SDD detector coupled to a single-slab scintillator. ACKNOWLEDGMENTS This study was supported by a grant from the Ministry of Knowledge Economy and by a grant from the Radiation Technology Development Program through the Korea Science and Engineering Foundation funded by the Ministry of Education, Science and Technology ( ), Republic of Korea. REFERENCES [1] C. Fiorini, A. Longoni, F. Perotti, C. Labanti, P. Lechner and L. Struder, IEEE Trans. Nucl. Sci. 44, 2553 (1997). [2] C. Fiorini, F. Perotti, C. Labanti, A. Longoni and E. Rossi, IEEE Trans. Nucl. Sci. 46, 858 (1999). [3] P. Lechner, C. Fiorini, R. Hartmann, J. Kemmer, N. Krause, P. Leutenegger, A. Longoni, H. Soltau, D. Stotter, R. Stotter, L. Struder and U. Weber, Nucl. Instr. Meth. A 458, 281 (2001). [4] W. Metzger, J. Engdahl, W. Rossner, O. Boslau and J. Kemmer, IEEE Trans. Nucl. Sci. 51, 1631 (2004). [5] C. Fiorini, M. Bellini, A. Gola, A. Longoni, F. Perotti, P. Lechner, H. Soltau and L. Struder, IEEE Trans. Nucl. Sci. 52, 1165 (2005). [6] M. Marisaldi, C. Labanti, H. Soltau, C. Fiorini, A. Longoni and F. Perotti, IEEE Trans. Nucl. Sci. 52, 1842 (2005). [7] A. Castoldi and A. Galimberti, IEEE Trans. Nucl. Sci. 53, 526 (2006). [8] C. Fiorini, A. Gola, M. Zanchi, A. Longoni, P. Lechner, H. Soltau and L. Struder, IEEE Trans. Nucl. Sci. 53, 2392 (2006). [9] C. Fiorini, A. Longoni, M. Porro, F. Perotti, P. Lechner and L. Struder, Nucl. Instr. Meth. A 560, 148 (2006). [10] N. Y. Kim and C. S. Lee, J. Korean Phys. Soc. 53, 1201 (2008). [11] C. Fiorini, F. Perotti, E. Rossi, A. Longoni and C. Labanti, IEEE Trans. Nucl. Sci. 51, 1625 (2004). [12] C. Fiorini and F. Perotti, Rev. Sci. Instrum. 76, (2005). [13] C. Fiorini, A. Gola, A. Longoni, M. Zanchi, A. Restelli, F. Perotti, P. Lechner, H. Soltau and L. Struder, Nucl. Instr. Meth. A 568, 96 (2006). [14] A. Castoldi, C. Guazzoni, R. Hartmann and L. Struder, Nucl. Instr. Meth. A 582, 849 (2007). [15] C. Fiorini, A. Gola, M. Zanchi, A. Longoni, H. Soltau and L. Struder, Nucl. Instr. Meth. A 571, 126 (2007). [16] A. Gola, C. Fiorini, M. Porro and M. Zanchi, Nucl. Instr. Meth. A 571, 339 (2007). [17] J. Joung, K. Lee, D. Henseler, W. Metzger, Y. Choi, Y. B. Ahn and Y. K. Kim, J. Nucl. Sci. Technol. 45, 1347 (2008). [18] J. Joung, R. S. Miyaoka, S. Kohlmyer and T. K. Lewellen, IEEE Trans. Nucl. Sci. 47, 1104 (2000). [19] J.-H. Kim, Y. Choi, K.-S. Joo, B.-S. Sihn, J.-W. Chong, S. E. Kim, K. H. Lee, Y. S. Choe and B.-T. Kim, Phys. Med. Biol. 45, 3481 (2000). [20] A. P. Dhanasopon, C. S. Levin, A. M. K. Foudray, P. D. Olcott and F. Habte, IEEE Trans. Nucl. Sci. 52, 1439 (2005). [21] C. Fiorini and P. Lechner, IEEE Trans. Nucl. Sci. 49, 1147 (2003). [22] G. F. Knoll, T. F. Knoll and T. M. Henderson, IEEE Trans. Nucl. Sci. 35, 872 (1988). [23] F. Cayouette, D. Laurendeau and C. Moisan, Proc. SPIE 4833, 69 (2002) [24] G. Bertuccio, L. Fasoli, C. Fiorini and M. Sampietro, IEEE Trans. Nucl. Sci. 42, 1399 (1995). [25] L. Kurchaninov, Nucl. Instr. Meth. A 374, 91 (1996). [26] K. Hansen and C. Reckleben, IEEE Trans. Nucl. Sci. 51, 3845 (2004). [27] A. O. Hero, Technical Report 890, Lincoln Laboratory MIT, 1992.