PUBLISHED BY IOP PUBLISHING FOR SISSA MEDIALAB. Imaging performance of silicon photomultipliers coupled to BGO and CsI:Na arrays

Transcription

1 PUBLISHED BY IOP PUBLISHING FOR SISSA MEDIALAB RECEIVED: August 2, 2013 REVISED: October 16, 2013 ACCEPTED: November 27, 2013 PUBLISHED: December 12, 2013 Imaging performance of silicon photomultipliers coupled to BGO and CsI:Na arrays S. David, a,1 M. Georgiou, a,b E. Fysikopoulos, c N. Belcari d,e and G. Loudos a a Department of Biomedical Engineering, Technological Educational Institution of Athens, Athens, Greece b Department of Nuclear Medicine, Medical School, University of Thessaly, Larissa, Greece c School of Electrical and Computer Engineering, National Technical University of Athens, Athens, Greece d Department of Physics E. Fermi, University of Pisa, Pisa, Italy e Istituto Nazionale di Fisica Nucleare (INFN), Sezione di Pisa, Pisa, Italy sdavid@teiath.gr ABSTRACT: The aim of this study is to investigate the imaging performance of a silicon photomultiplier array (ArraySL-4) photodetector for possible PET and potentially SPECT applications using BGO and CsI(Na) pixellated scintillators. Our main objectives are: i) the comparison of the ArraySL-4 to the older version SensL s SPMArray4 photo detector in terms of energy resolution and peak to valley ratio of a row profile in the flood image and ii) the study of the effect of different coupling schemes using ultra transmitting glass windows of various thicknesses. We acquired raw images from two pixellated scintillators (BGO with mm 3 and CsI:Na with mm 3 pixel sizes) irradiated with 511 kev and kev g-rays from a 22 Na source. The SiPM array detector allowed the clear visualization of the discrete 2 2 mm 2 pixellated BGO and 1 1 mm 2 CsI:Na scintillator elements at room temperature (no cooling). The energy resolution of the new SensL ArraySL-4 detector for the mm 3 BGO pixellated scintillator array is improved for rather 6 percentage points (energy resolution improvement equal to 22%) and the peak to valley ratio is measured higher for both scintillator arrays (for BGO 68% (1.7 ) and for CsI:Na 154% (2.5 )) compared with SPMArray4. The clear identification of the 1 1 mm 2 CsI:Na scintillator elements provides evidence that the combination of those SiPMs with even smaller arrays can be used as an efficient imaging detector module. Optical coupling significantly improves image uniformity, while the use of BK7 ultra transmitting glass window with 1.35 mm thickness provided the best measure energy resolution equal to 21.5%. KEYWORDS: Gamma detectors (scintillators, CZT, HPG, HgI etc); Gamma camera, SPECT, PET PET/CT, coronary CT angiography (CTA); Photon detectors for UV, visible and IR photons (solidstate) (PIN diodes, APDs, Si-PMTs, G-APDs, CCDs, EBCCDs, EMCCDs etc) 1 Corresponding author. c 2013 IOP Publishing Ltd and Sissa Medialab srl doi: / /8/12/p12008

2 Contents 1 Introduction 1 2 Materials and methods 2 3 Results and discussion ArraySL-4 system performance and comparison with the previous SPMArray4 photo detector Optical coupling 7 4 Conclusions 9 1 Introduction SiPM arrays are becoming detectors of choice for compact, as well as whole body PET and SPECT systems due to their small size and attractive characteristics [1]. They have high gain (similar to that of PMTs) and are operated at low bias voltages (< 80 V) [2]. Moreover, they are insensitive to magnetic fields and thus good candidates for hybrid PET/MR and SPECT/MR imaging systems. In addition, they show excellent timing resolution (below 200 ps) [3, 4], which makes them suitable for Time-of-Flight (ToF) PET applications [5, 6]. Astonishing progress has been made over the past 5 years, with design and performance improvements in arrays produced by SiPMs manufacturers. Nowadays the SiPM s can have more than 4000 microcells per pixel and can provide almost linear response at most of the nuclear imaging applications [7 9]. However, the readout of individual SiPM arrays pixels requires a large number of processing channels, which can lead to complexity in scaling up the detector design or require the design and fabrication of specific ASICs. While the development of readout systems digitizing signals from each SiPM pixel (typically 2 3 mm in size) is pursued by the majority of the developers, the readout schemes that allow the substantial reduction in the number of the output channels, and therefore the required number of digitizing analog-to-digital converters (ADC) channels, have several benefits including hardware simplicity and cost reduction [10]. To this direction, an attractive approach is the use of resistor dividers, which multiplex the number of channels of each detector module. In this way it is possible to have only four analog output signals [11], which contain the position and energy information of each detected event. The idea of using division resistive networks was introduced in position sensitive photomultiplier tubes and has been successfully transferred to other discrete detectors such SiPMs and one of its benefits is that it can be extended to more than one arrays [12 14]. This study has three objectives: firstly, the evaluation of the performance of the new SensL s silicon photomultiplier array (ArraySL-4) using a division resistive network for applications in nuclear imaging. To this aim, we acquired raw images from a pixellated BGO scintillator with 1

3 2 2 5 mm 3 pixel size irradiated with 511 kev and kev g-rays from a 22 Na source. In addition we tested a CsI:Na with mm 3 pixel size, since the small size of the crystal pixel allows us to test the limits of crystals identification. We have applied a symmetric resistive charge division circuit (Charge-SCD) reducing the 16 pixel outputs to only 4 position signals. Secondly, the comparison of the newer ArraySL-4 to the older version of the SensL s SPMArray4 photo detector, in terms of energy resolution and peak to valley ratio of a row profile in the flood image. The major difference between the two detector versions is that the new SensL s SiPM array (ArraySL-4) has larger number of microcells per pixel (4774) compared to those (3640) of the old one (SPMArray4), which results to a larger dynamic range. According to the manufacturer, the second-generation of SL family silicon photomultiplier detectors has four times higher signal-tonoise ratio compared to the previous generation and better uniformity. The output uniformity of ArraySL-4 is ±10% compared to the ±40% of the SPMArray4 detector, whereas standard photomultiplier tubes have ±33% output uniformity [16]. The behavior of the older SensL s detector array (SPMArray4) has been extensively evaluated in a previous study by our group [17]. Thirdly, to study the effect of different coupling schemes using ultra transmitting glass windows of various thicknesses and silicone grease. At the moment few studies report how different coupling schemes affect detectors performance [18, 19]. Moreover, most groups report the imaging performance of SiPMs arrays when coupled to LSO or LYSO arrays for PET applications but there are no published studies demonstrating the imaging performance of BGO, which can be interesting for specific applications, such as low cost dedicated PET systems using BGO. Detailed information about the energy and timing resolution properties of BGO scintillator coupled to various SiPMs can be found at [20]. Taking into account that ArraySL-4 has a peak wavelength at 500 nm, it better matches BGO, which has a peak at 480 nm, compared to LYSO, whose peak is at 420 nm. 2 Materials and methods SensL s scalable silicon photomultiplier array (ArraySL-4) is a commercially available, solid-state array detector based on silicon photomultiplier technology [16]. It consists of 16 pixel elements covering an active area of mm 2. Each pixel has 4774 square microcells connected in parallel, with individual cell side equal to 35 µm. The reported dimension of 35um cell side of the SensL s device describes (according to the manufacturer) the real effective area, which is similar to 50um cell side of Hamamatsu MPPC. The pixel array is over-molded with epoxy to completely encapsulate the pixels, bondwires and substrate bondpads. A 20-pin grid array (PGA) is employed for electrical I/O s to the user s printed circuit board. This optical detector array offers 4 side tileable packaging to allow to be tiled for larger area detection systems. The ArraySL-4 is sensitive to visible light in the range of 400 nm to 850 nm with peak photon detection efficiency (PDE) 14% at 500 nm [16]. The PDE value referred includes crosstalks and afterpulses. The particular ArraySL-4 detector works properly at a low bias voltage of V as it is recommended by the manufacturer. The 16 output signals of the SiPM array are further reduced to 4 position signals through a two-stage charge division resistive network (Charge-SCD resistive readout). First, the incoming charges from the 4 4 SiPM array are equally split into X and Y directions using a symmetric 2

4 2D decoupling resistive matrix [12], which reduces the 16 readout channels to 8 channels (4 rows and 4 columns). Those 8 readout channels are individually amplified and shaped [17] and finally are further reduced to 4 outputs (Xa, Xb, Yc, Yd) using a resistive division network of amplitude weighting resistors. The centroid position (X, Y) of the incident light pulse distribution is calculated using Anger s equations (2.1). A summed signal from the four position signals was used to provide the signal which is proportional to photons energy (2.2). X b X = X a, Y = Y c Y d (2.1) X a + X b Y c +Y d E = X a + X b +Y C +Y d (2.2) More details about the resistive network that was used can be found in a previous study of our group in [17]. The 4 position signals (X a,x b,y c and Y d ) were amplified and then digitized using a free running ADC [21, 22]. The amplifiers shape the input signals taking into account the ADC sampling rate. A FPGA (Spartan 6 LX16) was used for triggering and signal processing of the pulses [23]. Data were transferred to a standard PC via ethernet link. We acquired raw images and energy histograms of a BGO and a CsI:Na pixellated scintillator under 22 Na excitation. The BGO array consists of 6 6 scintillator elements with mm 3 pixels and 2.3 mm pitch. The CsI:Na array has scintillator crystal elements with mm 3 independent crystals and 1.2 mm pitch and it is encapsulated in an aluminum box with a 1 mm glass window. Both scintillator arrays were manufactured by Hilger Crystals [24]. More information about scintillators characteristics can be found in [25]. Most SiPM arrays have a limited active area, since their pixels cover only part of the detector module s surface. Taking into account that those arrays are not usually aligned one by one with the scintillators array elements, their optimal coupling remains challenging and has an effect on detector s sensitivity, image quality as well as other performance parameters. The use of Anger logic for position calculation is in principle improved when the signals from more SiPM pixels contribute. When using a light guide, the scintillator light from an individual scintillator pixel is distributed to more SiPM pixels increasing the uniformity of light distribution and allowing its detection by more SiPM pixels, despite the dead regions of the array. This effect does not exist when we have one by one scintillator to optical detector elements coupling, but in this case the size of scintillator pixels has to match the size of SiPM modules, which minimizes the possible scintillators that can be used. The different coupling schemes of the BGO scintillator array to the ArraySL-4 entrance window that were studied are: a) wet optical coupling (two components silicone grease Sylgard ) and b) coupling with ultra transmitting glass windows of 0.7, 1.0, and 1.35 mm thickness [26], with optical grease to all optical surfaces namely thereafter as BK7 0.7 mm, BK7 1.0 mm and BK mm. The refractive index of the BK ultra transmitting glass windows was n d = at 480 nm. A1µ Ci 22 Na source was utilized to all measurements positioned 11 mm above the surface of the scintillator. The scintillator blocks, the SiPM array and the division resistor network were enclosed into a light tight black box. All measurements underwent room temperature. The SiPM were supplied by a stabilized double channel AC/DC power supply (MPS-6005L-2). 3

5 Figure 1. Raw images (up) and normalized horizontal profiles (down) of the (left) ArraySL-4 SiPM array and (right) SPMArray4 photo detector coupled with mm 3 BGO scintillator crystal with optical grease. 3 Results and discussion 3.1 ArraySL-4 system performance and comparison with the previous SPMArray4 photo detector Figure 1 shows two raw images and the normalized horizontal profiles (produced under excitation of g-rays from a 22 Na source) of the ArraySL-4 and SPMArray4 photo detectors coupled with mm 3 BGO scintillator crystal array with optical grease. The horizontal profiles of both raw images indicate a clear visualization of the 6 discrete scintillator elements. In addition, the ArraySL-4 provides lower background and clearer pixels identification, as it can be observed in both images and profiles. The decrease in sensitivity at the edges of field of view, which is observed in raw images, is produced by a non uniform irradiation from the 22 Na source (positioned in all cases only 11 mm far away above from the center of the scintillator array). In fact, the radioactivity of the controlled 22 Na source that was available was very low (1µCi) and data acquisition at larger distances was rather time consuming. Simulations with GATE (Monte Carlo simulation platform based on Geant4) have confirmed this hypothesis and they demonstrated that when the source is placed at a larger distance the uniformity is restored. Figure 2 presents a typical energy spectrum of one central crystal element, obtained with the mm 3 BGO scintillator coupled by the ArraySL-4 SiPM array with optical grease. 4

6 Figure 2. Energy spectrum of one central crystal element of the mm 3 BGO scintillator coupled with ArraySL-4 SiPM array photo detector. The energy resolution was measured equal to 22% at 511 kev photopeak. Both full energy peaks related to 22 Na gamma ray energies are clearly illustrated despite the small thickness of the BGO scintillator (5 mm). A large number of low energy pulses corresponding to photons not fully absorbed in the detector (i.e Compton scattered photons) are not detected due to the lower energy threshold trigger logic that we used. The kev/511 kev peak ratio for the ArraySL-4 detector is equal to 2.36 demonstrating linear operation with small saturation effect at high energies. CsI scintillators doped either with Tl or Na activators are currently used as state of the art crystals in SPECT and not in PET imaging. The evaluation of CsI:Na scintillator under 22 Na excitation was carried out in order to check the performance of two SiPM arrays with smaller 1 1 mm 2 pixels too. The imaging performance of the new SiPM detector is clearly improved compared to the old version as it is presented in figure 3. Although the CsI:Na array consist of pixel elements (scintillator elements inside the active field of view of the entrance window of the SiPM array), in the case of ArraySL-4 detector the horizontal profiles of the raw images show only 9 discrete elements; the first two elements appear as one. Moreover, the last two scintillator elements are poorly separated, as it can be also observed in the horizontal profile. This can be mainly explained by the placement of the CsI:Na crystal array, which was slightly larger in dimensions than the SiPM array entrance window. This assumption can be also supported by the fact that a slight rotation on the crystal s position is observed; as a consequence the upper left edge pixels are better placed in the FOV and are more visible compared to the bottom left edge pixels. On the other hand, the upper right pixels are less visible compared to the bottom right ones. The raw image produced by the SPMArray4 photo detector coupled with 1 1 mm 2 cross section CsI:Na pixellated crystals illustrates a not clear visualization of the discrete crystal elements. From the pixels very few scintillator elements are clearly distinguished in figure 3 (right). The improved performance of the new ArraySL-4 can be clearly demonstrated from the data presented below in table 1. Table 1 summarizes the results obtained with both scintillator arrays coupled with the ArraySL-4 and the SPMArray4 photo detectors, in terms of energy resolution and peak to valley ratios. The energy resolution values referred in table 1 were the best values achieved between the 4 central crystal elements of the sipm arrays. The peak to valley ratio of a row profile in the flood 5

7 Figure 3. Raw images (up) and normalized horizontal profiles (down) of the (left) ArraySL-4 SiPM array and (right) SPMArray4 photo detector coupled with mm 3 CsI:Na scintillator array with optical grease. Table 1. Peak to valley ratios and energy resolution values of the ArraySL-4 & SPMArray4 optical detectors coupled with CsI:Na (1 1 5 mm 3 ) and BGO (2 2 5 mm 3 ) scintillator arrays. SiPM & Scintillator detector Values Average peak to valley ratio: ArraySL-4 & BGO 15.4 Average peak to valley ratio: SPMArray4 - BGO 9.14 Average peak to valley ratio: ArraySL-4 - CsI:Na 9.4 Average peak to valley ratio: SPMArray4 - CsI:Na 3.7 Energy kev: ArraySL-4 - BGO 21.2% Energy kev: SPMArray4 - BGO 27.2% image was calculated as the average of all the peaks to valleys ratio of a horizontal row profile in the flood image, to assess the accuracy of crystal pixels mapping. The ArraySL-4 detector presents higher peak to valleys ratio for both scintillator arrays (for BGO 68% (1.7 ) and for CsI:Na 154% (2.5 ) improvement). The best energy resolution achieved with the new SiPM detector is of 21.2%, which is improved for rather 6 percentage points (energy resolution improvement equal to 22%) when compared to the SPMArray4. 6

8 Table 2. Energy resolution 511 kev of different coupling schemes. ArraySL-4 & BGO Scintillator Coupling Energy Resolution Mean Values Standard Deviation Optical grease 23% ±0.97 BK7 0.7 mm 22.3% ±0.82 BK7 1 mm 22.8% ±0.55 BK mm 21.5% ± Optical coupling Figure 4 presents the raw images acquired with the discrete mm 3 pixellated BGO scintillator array. The image in the first row was obtained by coupling the scintillator to the SiPM detector entrance window, using only optical grease (two components Sylgard ), the second by using the BK7 0.7 mm ultra transmitting glass window and the third the BK7 1.0 mm and the last the BK mm. The same optical grease was used among all coupling surfaces. The horizontal profiles (second column of figure 4) of the images indicate a clear visualization of the discrete scintillator elements in all cases. The histograms on the right column show the energy spectrum as calculated from one central pixel. Energy resolution of about 22% and 17% was measured under 511 kev and kev excitation respectively. The ultra transmitting glass windows used for optical coupling kept almost the same peak to valley ratio in all cases of flood histograms. We also observed that as the glass thickness increases, the curve of the two central crystal elements is getting wider. Especially, an increase of 0.26 mm in the FWHM is measured while using the BK mm glass (0.78 mm) compared to coupling with only grease (0.52 mm). The comparison of the four images in figure 4 demonstrates that the raw image becomes more uniform when glass becomes thicker. When no glass is used the structure of the SiPM array becomes visible and the relative position of the crystal and SiPM pixels is evident in the raw image. It can be observed that by adding a glass of even 0.7 mm both the distance between the peaks that correspond to crystal pixels and the width of crystal profiles become uniform, resulting to a more uniform image, where the SiPM structure is not observable. This observation is rather important, especially when several SiPM arrays are coupled together. Then, as the dead regions in the field-of-view (FOV) increase, the thickness of the optical guide should increase. However, the effect on systems spatial resolution needs to be considered, thus when uniformity at raw images is ensured, further increase of the thickness of the coupling material should be avoided. The thickness of the glass does not seem to have a negative effect on energy resolution. On the contrary, the 1.35 mm BK7 glass improves slightly the system energy resolution (from 23% to 21.5%) and the uniformity of the energy response for the 511 kev peak. Table 2 summarizes the mean energy resolution values measured at 511 kev photopeaks of different discrete scintillator pixels of the raw images. The standard deviation mentioned in the table decreases as the thickness of the BK7 ultra transmitting glass window increases showing that the glass coupling minimizes the deviations in energy resolution response across the area of the SiPM detector. 7

9 Figure 4. Raw images, horizontal profiles and energy spectra acquired with the discrete mm 3 pixellated BGO scintillator array under 22 Na excitation. The image in the first row was produced by coupling the scintillator plate to the SiPM entrance window by using only optical grease, the second by using 0.7 mm glass (BK7 0.7 mm), the third by using 1.0 mm glass (BK7 1.0 mm) while the last with 1.35 mm (BK mm) ultra transmitting glass window. A quartz glass of similar thickness (1.3 mm) was used by another group to redistribute the light above the 2 mm dead areas in an arrangement of 2 2 SPMArray4 optical detectors [18]. In other study using MPPC arrays dry glass coupling even of 2 3 mm in thickness were used in order to map the scintillator elements above the gaps, as the active area of each pixel of the MPPC arrays are 3 3 mm 2 and the pixel pitch 4.05 mm [19]. The same pixel characteristics of the ArraySL-4 8

10 SiPM are mm 2 and 3.36 mm respectively. Another practical consideration is the fact that optical coupling with transmitting glass windows is very stable in SiPM arrays and can replace the light guide with thick optical grease slices that can create inhomogeneities due to melting after a long period. Although most groups have studied Lu based scintillators, BGO has some strong advantages including its high stopping power and availability, which lead to the construction of even commercially available lower cost detectors e.g. PETbox by Sofie Bioscience, Inc. [27]. Although the controversy between BGO and LYSO is still open, simulation and experimental results by this group show the advantages of using BGO arrays (with thickness of 5 mm to 7 mm), such as i) improved spatial resolution, due to decrease in parallax errors, ii) improvement of minimum detectable activity, due to the low intrinsic radioactivity of BGO, iii) reduced intercrystal scattering and iv) feasibility to construct compact small ring detectors. Imaging results with CsI:Na 1 1 mm 2 pixels show that the suggested readout approach can be used for even smaller pixels, at least for PET and possibly for SPECT [25]. Taking into account the recent availability of blue-sensitive SiPMs with peak wavelength at 420 nm, the coupling of CsI:Na to those arrays is expected to provide very good results for SPECT. Moreover, the successful application of the presented readout circuit, which has been used in both single Position Sensitive Photomultiplier Tubes (PSPMTs) and PSPMTs arrays, can be extended to larger blocks of SiPM arrays. This can lead to the design of efficient detectors, with significantly lower cost, by combining small SiPMs and BGO arrays with relatively little material. Our next steps include the optimization of the optical coupling between the scintillation crystal and the SiPM array entrance window and the assessment on larger blocks constructed by two or more ArraySL-4 detectors in an array. In this way we will investigate the limits of using only 4 position signals from a larger number of arrays. Moreover, we intend to apply this readout scheme to SPECT scintillators, since SiPMs can be used as a low cost alternative to PSPMTs in SPECT as well, provided that scintillators light output is sufficient to overcome SiPMs intrinsic noise. 4 Conclusions A clear visualization of both mm 3 pixellated BGO and mm 3 pixellated CsI:Na array elements was achieved under Na 22 excitation, with the applied resistive 4-channel readout circuit at room temperature. The charge-scd resistive network readout shows very good results, which are in accordance with results reported by other groups working with PSPMTs and which stated that the Charge-SCD is the optimum multiplexed electrical readout circuit among others [11 14]. The energy resolution of the new SensL ArraySL-4 detector for the mm 3 BGO pixellated scintillator array is improved for rather 6 percentage points (energy resolution improvement equal to 22%) and the peak to valley ratio is measured higher for both scintillator arrays (for BGO 68% (1.7 ) and for CsI:Na 154% (2.5 ) improvement) compared with SPMArray4. In the case of mm 3 pixellated CsI:Na the new generation detector presents 2.5 times better averaged peak to valley ratio. For clear visualization of crystal pixels optical coupling is necessary. The different coupling schemes tested, show similar performance, as the energy resolution and the peak to valley ratios remain almost constant. On the other hand, in cases that more than one SiPM arrays 9

11 are joined to form a larger field of view, the thickness of the optical guide is critical, to redistribute light above dead areas and detector edges and needs to be studied. For the mm 3 BGO scintillator, a thickness of BK mm light guide showed optimal characteristics. Acknowledgments This research has been co-funded by the European Union (European Social Fund) and Greek national resources under the framework of the Archimedes III: Funding of Research Groups in TEI of Athens project of the Education & Lifelong Learning Operational Programme. The authors would also like to thank COST Action TD1007 Bimodal PET-MRI molecular imaging technologies and applications for in vivo monitoring of disease and biological processes for facilitating equipment exchange between authors. References [1] W. Christoph et al., Calibration and stability of a SiPM-based simultaneous PET/MR insert, Nucl. Instrum. Meth. A 702 (2013) 50. [2] P. Buzhan et al., Silicon photomultiplier and its possible applications, Nucl. Instrum. Meth. A 504 (2003) 48. [3] M. Conti, Improving time resolution in time-of-flight PET, Nucl. Instrum. Meth. A 648 (2010) 194. [4] G. Llosa et al., Energy and Timing Resolution Studies With Silicon Photomultipliers (SiPMs) and 4-Pixel SiPM Matrices for PET, IEEE T. Nucl. Sci. 56 (2009) 543. [5] R. Vinke et al., Optimizing the timing resolution of SiPM sensors for use in TOF-PET detectors, Nucl. Instrum. Meth. A 610 (2009) 188. [6] H.S. Yoon et al., Initial results of simultaneous PET/MRI experiments with an MRI-compatible silicon photomultiplier PET scanner, J. Nucl. Med. 53 (2012) 608. [7] Hamamatsu webpage, [8] Photonique SA webpage, [9] SensL webpage, [10] AiT Instruments webpage, [11] V. Popov, Matrix output device readout system, U.S. Patent 6,747,263 (2004). [12] V. Popov, S. Majewski and B. Welch, A novel readout concept for multianode photomultiplier tubes with pad matrix anode layout, Nucl. Instrum. Meth. A 567 (2006) 319. [13] V. Popov, S. Majewski and A.G. Weisenberger, Readout electronics for multianode photomultiplier tubes with pad matrix anode layout, IEEE Nucl. Sci. Conf. R. (2003) [14] V. Popov, Advanced data readout technique for Multianode Position Sensitive Photomultiplier Tube applicable in radiation imaging detectors, 2011 JINST 6 C [15] S. Siegel, R.W. Silverman, Y. Shao and S.R. Cherry, Simple charge division readouts for imaging scintillator arrays using a multi-channel PMT, IEEE T. Nucl. Sci 43 (1996) [16] SensL, ArraySL-4 Scalable Silicon Photomultiplier Array, datasheet availabel online at ArraySL-4 v1p1.pdf. 10

12 [17] S. David, M. Georgiou, E. Fysikopoulos and G. Loudos, Comparison of three Resistor Network Division Circuits for the readout of 4x4 Pixel SiPM Arrays, Nucl. Instrum. Meth. A 702 (2013) 121. [18] Y. Wang et al., Design and performance evaluation of a compact, large-area PET detector module based on silicon photomultipliers, Nucl. Instrum. Meth. A 670 (2012) 49. [19] S. Majewski, J. Proffitt, A. Stolin and R. Raylman, Development of a resistive readout for SiPM arrays, IEEE Nucl. Sci. Conf. R. (2011) [20] T. Szczesniak et al., MPPC arrays in PET detectors with LSO and BGO scintillators, IEEE T. Nucl. Sci. 60 (2013) [21] Texas Instruments, 12-Bit Octal-Channel ADC Family Up to 65MSPS (2008) datasheet available online at [22] M. Streun, G. Brandenburg, H. Larue, E. Zimmermann, K. Ziemons and H. Halling, Pulse Recording by Free-Running Sampling, IEEE T. Nucl. Sci. 48 (2001) 524. [23] Xilinx Inc., SP601 Hardware User Guide, UG518(v1.4) (2010) datasheet available online at [24] Hilger Crystals webpage, [25] S. David et al., Initial results on SiPM array based on a symmetric resistive voltage division readout, IEEE Nucl. Sci. Conf. R. (2012) [26] K. Straub et al., Study of UT glasses for pixel identification performance in multi-anode PMT-based detectors for PET, Nucl. Instrum. Meth. A 718 (2013) 154. [27] Hui Zhang et al., Performance Evaluation of PETbox: a Low Cost Bench Top Preclinical PET Scanner, Mol. Imaging Biol. 13 (2011)