Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2026 Jan 17.
Published in final edited form as: Phys Med Biol. 2025 Jan 17;70(2):10.1088/1361-6560/ada084. doi: 10.1088/1361-6560/ada084

High Spatial Resolution PET Detectors Based on 10 mm × 10 mm Linearly-Graded SiPMs and 0.5 mm Pitch LYSO Arrays

Jiahao Xie 1,*, Haibo Wang 1, Stefano Merzi 2, Giovanni Paternoster 2, Alberto Gola 2, Jinyi Qi 1, Simon R Cherry 1, Junwei Du 1,*
PMCID: PMC12084082  NIHMSID: NIHMS2075531  PMID: 39689427

Abstract

Objective:

Position-sensitive silicon photomultipliers (PS-SiPMs) are promising photodetectors for ultra-high spatial resolution small-animal positron emission tomography (PET) scanners. This paper evaluated the performance of the latest generation of linearly-graded SiPMs (LG-SiPMs), a type of PS-SiPM, for ultra-high spatial resolution PET applications using LYSO arrays from two vendors.

Approach:

Two dual-ended readout detectors were developed by coupling LG-SiPMs to both ends of the two LYSO arrays. Each LG-SiPM has an active area of 9.8 × 9.8 mm2. Both LYSO arrays consist of 20 × 20 arrays of 0.44 × 0.44 × 20 mm3 polished LYSOs with a pitch of 0.5 mm. The performance of the two detectors was compared in terms of flood histogram, energy resolution, timing resolution, and depth-of-interaction (DOI) resolutions.

Main results:

Flood histograms showed clear identification of all LYSO elements except for some edge crystals due to the larger size of the LYSO arrays compared to the active area of the LG-SiPMs and the misalignment between LG-SiPMs and LYSO arrays in the assembled detectors. At a bias voltage of 37.0 V, the detectors utilizing the Tianle LYSO array and EBO LYSO array provided energy resolutions of 17.5 ± 2.2 % and 18.6 ± 2.0 %, timing resolutions of 0.75 ± 0.03 ns and 0.78 ± 0.03 ns, and DOI resolutions of 2.16 ± 0.15 mm and 2.31 ± 0.12 mm, respectively.

Significance:

The results presented in this paper demonstrate that the new generation LG-SiPMs are promising photodetectors for ultra-high spatial resolution small-animal PET scanner applications.

1. Introduction

Small-animal positron emission tomography (PET) scanners are pivotal functional molecular imaging tools in pre-clinical studies, where spatial resolution and sensitivity are key characteristics that strongly affect the quantitative accuracy and precision of PET imaging. Over the last two decades, numerous developments in small-animal PET scanners have been aimed at enhancing spatial resolution and sensitivity (Dupont and Warwick 2009, Miyaoka and Lehnert 2020). However, despite these efforts, none of these scanners have simultaneously reached the physical limitations of spatial resolution (~0.5 mm) while achieving high sensitivity (>10% at the center of the scanner) (Du and Jones 2023).

Spatial resolution significantly limits the ability to visualize and quantify radiotracer distributions in small subjects or organs of interest, particularly in scenarios such as imaging the murine brain and utilizing image-derived input functions (IDIF) for kinetic modeling studies (Karakatsanis et al 2013, Yang et al 2016). Previous studies have demonstrated that small-animal PET scanners with ~0.5 mm spatial resolution can be developed using lutetium–yttrium oxyorthosilicate (LYSO) arrays with ~0.5 mm pitches (Yamamoto et al 2016, Yang et al 2016). However, these scanners have a low sensitivity (1% at the center) due to the limitations of the detectors and geometry, making them impractical for routine use. To achieve ~0.5 mm spatial resolution across the field-of-view (FOV), the detectors should have good intrinsic resolution along with good depth-of-interaction (DOI) information to alleviate the parallax effect; to achieve high sensitivity, the detectors must utilize thick crystals such as ~20 mm LYSO to have high detection efficiency (Du and Jones 2023). Furthermore, these detectors must have a high packing fraction to minimize dead space between the detectors in the scanner to achieve high sensitivity.

Pixelated silicon photomultiplier (SiPM) arrays are the most commonly used photodetectors in state-of-the-art PET scanners and have demonstrated their ability to resolve crystal arrays with a pitch down to ~0.3 mm (Kuang et al 2019, Niu et al 2022, Liu et al 2023). However, resolving the small crystal elements requires light guides, which unfortunately leads to 1) a loss of scintillation photons reaching the photodetectors, thereby reducing the crystal identification ability of the detectors, which is more pronounced with thicker light guides, and 2) wider crystal spots in the flood histogram, making it more challenging to resolve smaller crystals (Du et al 2015a, Xie et al 2024). Additionally, the crystal arrays are significantly smaller than the size of SiPM arrays (Song et al 2008, Yamamoto et al 2013, Liu et al 2023), resulting in reduced packing fraction and low sensitivity when these detectors are used to construct a scanner.

Our previous studies have shown that dual-ended readout detectors based on position-sensitive photodetectors, such as position-sensitive avalanche photodiodes (PS-APDs) and position-sensitive SiPMs (PS-SiPMs), combined with thick and finely segmented LYSO arrays, are promising candidates for developing small-animal PET scanners with high and uniform spatial resolution across the FOV and high sensitivity (Yang et al 2016, Du et al 2018, Lai et al 2021). Specifically, our group has developed a dedicated scanner for mouse brain studies, and a better than 0.5 mm spatial resolution has been achieved at the center of the scanner (Yang et al 2016). However, due to the low gain, high bias voltage (~1800 V), and instability of the used PS-APD, developing a PET scanner with long axial FOV is challenging. Hence, this mouse brain PET scanner has a 7 mm axial FOV, resulting in a low sensitivity of ~0.6% at the center of the scanner. Compared to APDs, SiPMs offer much higher gain, lower bias voltage (<100 V), and greater robustness, making them much easier to use. In collaboration with Radiation Monitoring Devices (RMD, Watertown, USA) and Fondazione Bruno Kessler (FBK, Trento, Italy), we developed different versions of the PS-SiPMs (Schmall et al 2012, Du et al 2013, 2018). These early studies highlighted the potential of the PS-SiPM as promising photodetectors for ultra-high spatial resolution small-animal PET scanners or other dedicated PET scanners. Specifically, working with FBK, we developed a type of PS-SiPM named linearly-graded SiPMs (LG-SiPMs) by using double quenching resistors for each microcell and in-chip position-encoding circuit (Du et al 2015b). Two generations of LG-SiPMs were developed and evaluated for high-resolution PET applications using FBK’s red-green-blue high-density (RGB-HD) SiPM technology (Du et al 2015b, 2018). Using the second generation of LG-SiPM, we developed a dual-ended readout detector using 2 × 2 LG-SiPM arrays coupled to both ends of an LYSO array with a 0.5 mm pitch and a 20 mm thickness (Du et al 2018). All elements of the LYSO array were clearly resolved, and a 3.8 mm DOI resolution was achieved. However, the active area of 7.5 × 7.5 mm² of this second-generation LG-SiPM is too small for developing scanners, necessitating extensive readout electronics. Hence, a third generation of LG-SiPM with an active area of 9.8 × 9.8 mm² was developed recently, and FBK’s near-ultraviolet-sensitive and high-density (NUV-HD) SiPM technology was utilized to improve the performance of the LG-SiPM, too (Gola et al 2019, Acerbi et al 2024).

This study evaluated the performance of this third-generation LG-SiPM. To further improve the detector’s performance, LYSO arrays, fabricated using the same method (Andreaco et al 2007, James et al 2009) but from two different vendors, Tianle Photonics Co., Ltd (Sichuan, China) and EBO Optoelectronics Technology (Ningbo, China), were compared in terms of flood histogram quality, energy resolution, timing resolution, and DOI resolution.

2. Materials and Methods

2.1. Dual-Ended Readout Detector

The LG-SiPMs used in this study, shown in Figure 1 (left), had an active area of 9.8 × 9.8 mm² and were fabricated using FBK’s NUV-HD SiPM technology with 25 μm microcells. Each LG-SiPM has four cathode outputs that enable calculating the 2D position of the triggered microcells using the center-of-gravity (COG) algorithm and one anode dedicated to biasing the LG-SiPM. A dedicated timing signal can also be extracted from the anode using the alternating coupling (AC) method. The breakdown voltage of the LG-SiPM is 31.5 V at 20°C, and its peak photodetection efficiency (PDE) occurs at ~400 nm wavelength, closely matching the emission wavelength of LYSO crystals (Du et al 2009, Mao et al 2008), thereby enhancing performance of the PET detector.

Figure 1.

Figure 1.

Open in a new tab

Photographs of (left) an LG-SiPM, (middle) LYSO arrays with a 0.5 mm pitch size from two vendors, and (right) an assembled dual-ended detector based on two LG-SiPMs and one LYSO array. The four outer lateral surfaces of the EBO LYSO array were wrapped with Toray E60 first and then a layer of aluminum film to protect the array.

The two 20 × 20 arrays of 0.44 × 0.44 × 20 mm³ LYSOs from Tianle Photonics Co., Ltd (Sichuan, China) and EBO Optoelectronics Technology (Ningbo, China) both have a 0.5 mm pitch, and all six surfaces of the crystal elements were polished, as displayed in Figure 1 (middle). Inter-crystal reflectors, 50 μm thick Toray E60 sheets (Toray Industries, Inc., Japan), were glued to the LYSO elements using 5 μm thick optical glue. For convenience, these two LYSO arrays will be referred to as the Tianle LYSO array and the EBO LYSO array throughout this paper.

Compared to the LYSO array used in our previous studies, which were fabricated by assembling individually prepared LYSO elements (Du et al 2018), these two new LYSO arrays were fabricated using a method dedicated to producing high-resolution crystal arrays (Andreaco et al 2007, James et al 2009), enhancing the performance of the detectors. Briefly, the LYSO arrays were fabricated using the following process:

  1. A block of LYSO was cut to a thickness slightly thicker than 20 mm.

  2. The block was cut into LYSO slices with a thickness of 0.44 mm, and the slices were polished.

  3. Reflector sheets (Toray E60 sheets) were inserted between the slices, and the slices were glued together using optical glue with a thickness of 5 μm, resulting in a 1 × 20 array of LYSO slices with a 0.5 mm pitch.

  4. The 1 × 20 array of LYSO slices was cut orthogonally to the original slices, resulting in each slice being a 1 × 20 array of 0.44 × 0.44 × 20 mm³ LYSOs, and each slice was polished.

  5. Reflectors (Toray E60 sheets) were inserted between the slices, and the slices were glued together using optical glue with a thickness of 5 μm, resulting in a 20 × 20 array of 0.44 × 0.44 × 20 mm³ LYSO elements.

  6. Reflectors (Toray E60 sheets) were added to the LYSO array’s four outermost lateral surfaces.

  7. The two ends of the 20 × 20 LYSO array were polished to ensure the two ends were flat.

The LYSO arrays were coupled to the center of the LG-SiPM arrays using optical grease BC-631 (Saint-Gobain, France). For consistent alignment of the LYSO array relative to the LG-SiPMs across different experiments, 3D-printed holders were utilized, as shown in Figure 1(right).

2.2. Experimental Methods

2.2.1. Flood Histogram, Energy Resolution, and Timing Resolution

The diagram of the experimental setup for flood histogram, energy resolution, and timing resolution measurements is illustrated in Figure 2. The eight cathode outputs of the two LG-SiPMs in one dual-ended readout detector were first amplified using trans-impedance amplifiers based on AD8056s (Analog Device Inc., USA), and then shaped by a spectroscopy amplifier N568B (CAEN S.p.A., Italy). The outputs of the N568B were digitized using a digitizer DT5740D (CAEN S.p.A., Italy) at a speed of 62.5 mega samples per second (MSPS). The two anode outputs of the two LG-SiPMs were first amplified using trans-impedance amplifiers based on AD8045s (Analog Device Inc., USA), and then summed together as the timing signal, which was fed into a constant fraction discriminator (CFD) ORTEC 584 (AMETEK ORTEC Inc., USA) for timing information pick-off, used as the start signal for a time-to-amplitude converter (TAC model 566, AMETEK ORTEC Inc., USA).

Figure 2.

Figure 2.

Open in a new tab

Diagram of the experimental setup for flood histogram, energy resolution, and timing resolution measurements. The distance from the 22Na source to the dual-ended readout detector was made longer than the distance to the reference detector to ensure a uniform gamma photon distribution across the LYSO array.

A reference detector, consisting of a Hamamatsu PMT R13449 and a LYSO cylinder with a diameter of 20 mm and a thickness of 5 mm, was used as a coincidence detector to collect coincidence events. The PMT output was amplified first and then split into two channels: one for energy and the other for timing measurement. The energy channel was also shaped using the N568B and then digitized using the digitizer DT5740D. The timing channel was fed into a CFD ORTEC 584 for timing information pick-off, used as the stop signal for the TAC ORTEC 566. The output of the TAC was also digitized by the digitizer DT5740D.

The gamma photon’s interaction position in 2D (x and y), deposited energy (E), and DOI information (DOI) were calculated using the following formulas:

x=12R1−L1R1+L1+R2−L2R2+L2y=12T1−B1T1+B1+T2−B2T2+B2 (1)
E1=B1+T1+R1+L1,E2=B2+T2+R2+L2 (2)
E=E1+E2 (3)
DOI=aE1−E2E1+E2+b (4)

where Li, Ri, Ti and Bii=1,2 were the four digitized cathode signals from LG-SiPM1 and LG-SiPM2 (Figure 1, right). E1 and E2 were the sum of four digitized cathode signals from LG-SiPM1 and LG-SiPM2, respectively. Parameters a and b were used to model the relationship between the DOI ratio based on two energy signals from two LG-SiPMs and the DOI position (Du et al 2020).

The dual-ended readout detector was placed in a light-tight box, and the temperature of the air inside the box was maintained at 13 ± 0.3°C by blowing cool-dry air into the box using an FTS XR AirJet Sample Cooler (model XR 401, SP Industries Inc., USA). A thermocouple thermometer (model DIGI-SENSE WD-60010-10, Cole-Parmer Instrument Company, USA) continuously monitored the temperature inside the box throughout the experiment.

A 29.3 μCi 22Na source with a 0.25 mm active diameter (model MMS06-022, Eckert & Ziegler Isotope Products, USA) was used as the radiation source. To find the optimal bias voltage for the flood histogram, measurements were performed at bias voltages ranging from 35.5 to 38.5 V with 0.5 V intervals, and a 200–1000 keV energy window was applied to each crystal to select events. A wide energy window was selected as the developed detector is intended for small-animal PET applications. All inter-crystal scatter (ICS) events were included in our data analysis. Given the small size of the crystal elements, the ICS event ratio was expected to be very high. Discarding ICS events would lead to a significant loss in sensitivity when using these detectors to construct a scanner.

To quantitatively compare the flood histograms, flood histogram quality, k, was calculated as the ratio of the centroid distance to the full width at half maximum (FWHM) of the crystal spots in the flood histogram following the method described in Du et al 2016 using the formula:

k=NΠiNki,kstd=1N−1∑i=1Nki−k2 (5)

where ki was the flood histogram quality of each crystal (Du et al 2016), and N was the number of crystals. A larger k and a smaller kstd indicate better flood histogram quality, facilitating larger separation between crystals and better crystal identification. The optimal bias voltage was the one that provided the highest flood histogram quality value.

Using look-up tables generated from the flood histograms, events were assigned to individual crystals. The energy resolution for each crystal was calculated by determining the ratio of the FWHM to the peak position of the 511 keV photopeak, obtained from a Gaussian fit to the energy spectrum. The energy resolution of the detector was then determined by averaging the energy resolutions across all crystals in the detector. Saturation correction was not applied, as it is not feasible for dual-ended readout detectors (Du et al 2019). However, given the limited number of scintillation photons reaching the photodetectors and due to the small size of the crystal elements and the small microcell size (25 µm) of the LG-SiPMs, we believe the impact of saturation can be neglected.

Similarly, the timing resolution for each crystal was determined by measuring the FWHM of the timing spectra, also derived from a Gaussian fit to the timing spectrum. The timing resolution of the detector was calculated as the average of the timing resolutions across all crystals in the detector.

2.2.2. DOI resolution

The DOI resolution was measured using the optimal bias voltage identified from the flood histogram experiment. The DOI experimental setup, shown in Figure 3, was similar to the setup used for flood histogram, energy resolution, and timing resolution measurements (Figure 2), except the reference detector was replaced by a Hamamatsu PMT R13349 coupled to a 0.5 × 20 × 20 mm³ polished LYSO slab, and the dual-ended detectors were irradiated from one lateral side. The beam width was ~ 0.5 mm, estimated based on the geometry of the setup. DOI resolution was assessed at nine steps ranging from 2 mm to 18 mm with a 2 mm increment. To ensure sufficient event detection by the crystals on the right side of the LYSO array, 2,000,000 events were acquired at each depth. A 200-1000 keV energy window was also applied to each crystal to select events.

Figure 3.

Figure 3.

Open in a new tab

Diagram of the experimental setup for DOI resolution measurement.

The DOI resolution was calculated for each crystal and each depth by the FWHM of the Gaussian distribution of the estimated gamma interaction positions derived from a Gaussian fit. The DOI resolution of the detector was then calculated as the average DOI resolution across all crystals and depths.

3. Results

3.1. Flood Histogram

Figure 4 shows the flood histograms obtained from both detectors at the optimal bias voltage of 37.0 V. All crystals can be clearly resolved except for some edge crystals located at the two outermost left columns and two bottom rows due to the LYSO arrays having a cross-section (10 × 10 mm²) slightly larger than the active area of the LG-SiPMs (9.8 × 9.8 mm²) and the slight misalignment between the LYSO array and the LG-SiPMs. Therefore, the central 16 × 16 crystals were selected from the 20 × 20 crystals to compare the performance of the two detectors based on the two different LYSO arrays, as shown in Figure 5.

Figure 4.

Figure 4.

Open in a new tab

Flood histograms of the two detectors based on (left) the Tianle LYSO array and (right) the EBO LYSO array. The bias voltage was 37.0 V. The crystals in the two outmost left columns and two bottom rows were not well resolved.

Figure 5.

Figure 5.

Open in a new tab

Flood histograms of the two detectors based on (left) the selected 16 × 16 Tianle LYSO array and (right) the selected 16 × 16 EBO LYSO array.

Figure 6 shows the flood histogram quality obtained at various bias voltages for the two detectors based on the two LYSO arrays. The flood histogram quality values of the two detectors both increased initially and then decreased with the increasing bias voltage. This phenomenon was attributed to the change in noise and signal with the rising bias voltage, where the signal increased more than noise at lower bias voltages, but noise increased more at higher bias voltages (Du et al 2015b, Acerbi et al 2024). The best flood quality values were obtained at a bias voltage of 37.0 V for both detectors, and the detector based on the EBO LYSO array provided slightly better flood histogram quality. The flood histogram qualities obtained at a 37.0 V bias voltage for detectors based on the Tianle LYSO array and EBO LYSO array were 2.82 ± 0.28 and 2.90 ± 0.34, respectively.

Figure 6.

Figure 6.

Open in a new tab

Flood histogram quality versus bias voltage. The error bars represent the standard deviation across all crystals.

3.2. 511 keV Photopeak Position and Energy Resolution

Figures 7 and 8 show the 511 keV photopeak position and energy resolution for the central 16 × 16 LYSOs obtained at an optimal bias voltage of 37.0 V. The detector based on the Tianle LYSO array had a higher 511 keV photopeak position and slightly better energy resolution. Figure 9 presents the average 511 keV photopeak position and energy resolution across the central 16 × 16 LYSOs obtained at different bias voltages. As the bias voltage increased, the 511 keV photopeak position increased significantly from both detectors. Meanwhile, the best energy resolutions from both detectors were obtained at the bias voltage of 37.0 V, and the energy resolutions from both detectors were similar across various voltages. The average 511 keV photopeak position and energy resolution obtained at a bias voltage of 37.0 V across the selected 256 crystals were 3353 ± 392 and 2674 ± 208 in Analog-to-Digital Converter (ADC) channel for detectors based on the Tianle and the EBO LYSO arrays, respectively, while the corresponding average energy resolutions across the selected crystals were 17.5 ± 2.2% and 18.6 ± 2.0%, respectively.

Figure 7.

Figure 7.

Open in a new tab

511 keV photopeak position for the central 16 × 16 crystals of detectors based on (left) the Tianle LYSO array and (right) the EBO LYSO array. The bias voltage was 37.0 V.

Figure 8.

Figure 8.

Open in a new tab

Energy resolution for the central 16 × 16 crystals of the detectors based on (left) the Tianle LYSO array and (right) the EBO LYSO array. The bias voltage was 37.0 V.

Figure 9.

Figure 9.

Open in a new tab

The (top left) average 511 keV photopeak position, (top right) average FWHM of the 511 keV photopeak, and (bottom) average energy resolution across the selected 256 crystals versus bias voltage. The error bars represent the standard deviation across all crystals.

Although the detector based on the Tianle LYSO arrays exhibited a much higher 511 keV photopeak position, the energy resolutions of the two detectors were similar (Figure 9), which was attributed to the Tianle array having wider 511 keV photopeak (Figure 9(right)), due to differences in the LYSO material. To further investigate this, the average 511 keV photopeak position, the average FWHM of the 511 keV photopeak, and the average energy resolution across all crystals at different depths - calculated using data obtained during DOI measurements - are shown in Figure 10. Similarly, while the Tianle LYSO arrays showed higher 511 keV photopeak positions, they also exhibited higher FWHM values, resulting in an energy resolution that was not significantly better than the EBO array.

Figure 10.

Figure 10.

Open in a new tab

The (top left) average 511 keV photopeak position, (top right) average FWHM of the 511keV photopeak, and (bottom) average energy resolution across the selected 256 crystals versus depth, which were obtained using the data from DOI measurements. The error bars represent the standard deviation across all crystals.

3.3. Timing Resolution

Figure 11 shows the timing resolution for the central 16 × 16 crystals obtained at the optimal bias voltage of 37.0 V. The timing resolutions were crystal location dependent: the central crystals outperformed the edge crystals. Additionally, the timing resolution for each crystal showed a pattern that corresponded to the 511 keV photopeak position (Figure 7). For each detector, crystals had better timing resolution when they had a higher signal amplitude. Even though the detector based on the Tianle LYSO array had a much higher signal, the timing resolution improvements were minimal. This may be because the Tianle LYSO had a higher light yield but a slower decay time than the EBO LYSO array. The average timing resolution obtained at a bias voltage of 37.0 V was 0.75 ± 0.03 ns and 0.78 ± 0.03 ns for the detectors based on the Tianle LYSO array and EBO LYSO array, respectively.

Figure 11.

Figure 11.

Open in a new tab

Timing resolution for the central 16 × 16 crystals of the detector based on (left) the Tianle LYSO array and (right) the EBO LYSO array. The bias voltage was 37.0 V.

Figure 12 shows the average timing resolutions across the 256 selected crystals obtained at different bias voltages. The timing resolution of the two detectors based on the Tianle LYSO array and EBO LYSO array remained similar across the tested voltages.

Figure 12.

Figure 12.

Open in a new tab

Average timing resolution obtained at different bias voltages for the detector based on (left) the Tianle LYSO array and (right) the EBO LYSO array. The error bars represent the standard deviation across all crystals.

3.4. DOI resolution

Figure 13 shows the DOI resolution across the nine depths for the central 16 × 16 crystals at the optimal bias voltage of 37.0 V. The DOI resolutions of the crystals located on the left side of the LYSO array were better because the LYSO array was irradiated from the left side (Figure 3). The average DOI resolutions across the 256 crystals and 9 depths were 2.16 ± 0.29 mm and 2.31 ± 0.39 mm for detectors with the Tianle LYSO array and the EBO LYSO array, respectively.

Figure 13.

Figure 13.

Open in a new tab

DOI resolutions for the central 16 × 16 crystals of the detectors based on (left) the Tianle LYSO array and (right) the EBO LYSO array at 37.0 V bias voltage. The 22Na was located at the left side of the detectors. Hence, the crystals close to the source showed a better DOI resolution.

Figure 14 shows the average DOI resolution across the 256 crystals versus the depth. Better DOI resolution was obtained at depths closer to the two ends of the LYSO arrays, corresponding to previous studies (Du et al 2019). Additionally, the detector based on the Tianle LYSO arrays showed slightly more uniform DOI resolution across the 9 depths.

Figure 14.

Figure 14.

Open in a new tab

The average DOI resolution across the selected 256 crystals versus depth. The error bars represent the standard deviation values. The error bars represent the standard deviation across all crystals.

4. Discussion

Two dual-ended PET detectors based on the two LYSO arrays from different vendors were developed and compared. Both detectors exhibited similar performance in terms of flood histogram, energy resolution, timing resolution, and DOI resolution, with the detector based on the Tianle LYSO array providing slightly better overall performance. The discrepancy between these two detectors was caused by a combination of factors, including slight differences in the LYSO material and different LYSO fabrication methods, including the polishing method and optical glue used by the two different LYSO providers (Ren et al 2014, Kang et al 2020).

The LG-SiPM, a type of PS-SiPM, was used in our studies. Compared to traditional pixelated SiPM arrays, the LG-SiPM offers significantly higher intrinsic spatial resolution, making it more suitable for developing ultra-high-spatial resolution PET detectors and PET scanners (Acerbi et al 2024). Compared to our previous studies using LG-SiPMs based on FBK’s RGB technology (Du et al 2015b, 2018), which has peak photodetection efficiency (PDE) at ~500 nm wavelength (Serra et al 2013), the new generation LG-SiPMs used in this paper utilized NUV technology with a peak PDE at ~400 nm wavelength (Gola et al 2019), aligning more closely with the emission wavelength of LYSO crystals, which enhanced the overall performance of the detectors (Du et al 2009, Mao et al 2008), especially for crystal identification ability (Table I). Additionally, the active area of this latest version of LG-SiPM was also doubled compared to the previous generation, reducing the required complexity of the readout electronics and making it suitable for developing high-sensitivity PET scanners.

Compared to the LYSO array used in Du et al. 2018, a different LYSO array fabrication method was used to improve the performance of the detector (Table I). In our previous study, to fabricate the LYSO array economically, individual LYSO elements were cut and polished first, and then all LYSO elements were assembled to fabricate the LYSO array. In this way, if the performance of one LYSO element is not satisfactory, it can be replaced before assembling into the LYSO array. However, it is burdensome to glue and align the crystal elements, and it is challenging to cut the Toray reflector precisely to match the size of the LYSO elements. To improve the quality of LYSO arrays, the LYSO arrays used in this study were fabricated following the method described in Andreaco et al 2007 and James et al 2009. The advantage of this method is that it is easier to align the crystal arrays, and the reflector can match the size of the LYSO elements properly. On the other hand, the disadvantage is that if one LYSO element slice is broken during the polishing process, this LYSO slab needs to be discarded, potentially increasing the manufacturing cost of the LYSO array. One way to increase the yield of the array is to use unpolished lateral surfaces or partially polished lateral surfaces. However, the effect of surface treatments needs to be investigated to determine the effect of any changes.

The flood histograms showed that all the crystals in the two detectors were clearly resolved (Figure 4), except for some edge crystals due to the LYSO arrays having a cross-section (10 × 10 mm²) that was slightly larger than the active area of the LG-SiPMs (9.8 × 9.8 mm²) and the minor misalignment between the LYSO arrays and the LG-SiPMs in the assembled detectors. The two detectors had similar flood histograms, and the best flood histograms were both obtained at a bias voltage of 37.0 V. At the optimal bias voltage, the flood histogram qualities of detectors based on the Tianle and EBO LYSO arrays were 2.82 ± 0.28 and 2.90 ± 0.34, respectively, which were also better than our previous studies (Du et al 2018) and much better than other high spatial resolution detectors based on pixelated SiPM arrays (Kuang et al 2019, Liu et al 2023). The flood histogram also showed that LYSO arrays with even smaller pitches can be resolved using this latest LG-SiPM. In the future, we will develop detectors using LYSO arrays with ~0.25 mm pitch for ultra-high spatial resolution PET scanners dedicated to mouse or rat brain studies.

The energy resolutions and timing resolutions of the two detectors were similar, which were ~18% and ~0.7 ns, respectively. Although the detector based on the Tianle LYSO array provided a higher 511 keV photon peak position, its FWHM was also higher. As a result, the energy resolutions were not significantly better than those of the detector based on the EBO LYSO array. This outcome may be attributed to differences in the properties of the LYSO materials. The timing resolution and energy resolution were worse than those used for whole-body clinical PET due to the high aspect ratio of the crystals (Akamatsu et al 2020, Cates and Levin 2018). The aspect ratio for the crystals in this study was ~45 compared with ~6 for whole-body PET scanners. However, as the detectors are designed for small-animal PET scanners, the effect of timing resolution and energy resolution on the image quality is less critical than for clinical scanners.

The detector based on the Tianle LYSO array showed slightly better DOI resolution (2.16 ± 0.29 mm and 2.31 ± 0.38 mm for detectors based on the Tianle and EBO LYSO arrays, respectively), especially at the center of the LYSO array. Additionally, because the radiation source was positioned at the left-hand side of the detector, crystals from the left columns exhibited better performance than those from the right columns (Figure 13), consistent with our previous results (Du et al 2018, Du 2021).

Compared to high-resolution DOI-encoding detectors that use 0.5 mm pitch and 20 mm thick crystal arrays with pixelated SiPM arrays (Liu et al 2023), the detector developed in this study provided similar energy and DOI resolutions, but significantly better crystal identification ability, and superior timing resolution. Additionally, the readout electronics used in this study are much simpler, as the position encoding circuit is integrated within the LG-SiPMs.

The LG-SiPMs used in this study had large dead space around the edge of the package due to wire bonding being used to lead the signals of the LG-SiPM to the printed circuit board in order to reduce the cost. To reduce the dead space, through-silicon via (TSV) technology can be used (Parellada-Monreal et al 2023) and will be applied in future LG-SiPM development.

5. Conclusion

Two high spatial resolution dual-ended readout PET detectors, based on our latest generation LG-SiPMs coupled to both ends of LYSO arrays with a 0.5 mm pitch and 20 mm thickness, were developed and compared. All LYSO elements in the two detectors were clearly resolved, except for some edge crystals due to the LYSO arrays having cross-sections (10 × 10 mm²) slightly larger than the active area of the LG-SiPMs (9.8 × 9.8 mm²) and the slight misalignment between the LG-SiPMs and LYSO arrays. The results showed that the two detectors, based on LYSO arrays from two different vendors, Tianle Photonics Co., Ltd (Sichuan, China) and EBO Optoelectronics Technology Co., Ltd (Ningbo, China), had similar performance, and both are promising candidates for ultra-high spatial resolution small-animal PET scanners. The results also indicate that even more finely segmented LYSO arrays could likely be resolved with these latest generation LG-SiPMs.

TABLE I.

Performance Comparison of Detectors Based on Different Generations of LG-SiPMs

LG-SiPM Generation 1st 2nd 3rd
Active area (mm2) 4 × 4 7.6 × 7.6 9.8 × 9.8
Microcell size (µm) 45 20 25
Technology RGB RGB-HD NUV-HD
LYSO array Pitch (mm) 0.5
Crystal size (mm3) 0.45 × 0.45 × 6 0.445 × 0.445 × 20 0.44 × 0.44 × 20
Source Agile Engineering Crystal Photonics Tianle EBO
Detector performance Flood histogram quality 2.49 ± 0.72 1.91 ± 0.26 2.82 ± 0.28 2.90 ± 0.34
Energy resolution (%) 20.6 ± 3.3 21.8 ± 4.9 17.5 ± 2.2 18.6 ± 2.0
Timing resolution (ns) 0.66 ± 0.03 0.96 ± 0.08 0.75 ± 0.03 0.78 ± 0.03
DOI resolution (mm) NONE 4.04 ± 1.34 2.16 ± 0.15 2.31 ± 0.15
Temperature (°C) 10 10 13
Reference (Du et al 2015b) (Du et al 2018) This study
Open in a new tab

Acknowledgments

This work was funded by NIH grant R01 EB031961.

Footnotes

Conflicts of Interest

The authors disclose no potential conflicts of interest.

Reference

  1. Acerbi F, Merzi S and Gola A 2024. Position-Sensitive Silicon Photomultiplier Arrays with Large-Area and Sub-Millimeter Resolution Sensors 24 4507. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Akamatsu G, Takyu S, Yoshida E, Iwao Y, Tashima H, Nishikido F and Yamaya T 2020. Evaluation of a Hamamatsu TOF-PET Detector Module with 3.2 mm Pitch LFS Scintillators and a 256-channel SiPM Array IEEE Transactions on Radiation and Plasma Medical Sciences 1–1 [Google Scholar]
  3. Andreaco MS, Williams CW, Moyers JC and Vaigneur K 2007. Method for producing a high resolution detector array Online: https://patents.google.com/patent/US7157014/en [Google Scholar]
  4. Cates JW and Levin CS 2018. Evaluation of a clinical TOF-PET detector design that achieves ⩽100 ps coincidence time resolution Phys. Med. Biol 63 115011. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Du J 2021. Performance of Dual-Ended Readout PET Detectors Based on BGO Arrays and BaSO4 Reflector IEEE Transactions on Radiation and Plasma Medical Sciences 1–1 [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Du J, Ariño-Estrada G, Bai X and Cherry SR 2020. Performance comparison of dual-ended readout depth-encoding PET detectors based on BGO and LYSO crystals Phys. Med. Biol 65 235030. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Du J, Bai X, Gola A, Acerbi F, Ferri A, Piemonte C, Yang Y and Cherry SR 2018. Performance of a high-resolution depth-encoding PET detector module using linearly-graded SiPM arrays Phys. Med. Biol 63 035035. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Du J, Bai X, Liu C-C, Qi J and Cherry SR 2019 Design and evaluation of gapless curved scintillator arrays for simultaneous high-resolution and high-sensitivity brain PET Phys. Med. Biol 64 235004. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Du J and Jones T 2023. Technical opportunities and challenges in developing total-body PET scanners for mice and rats EJNMMI Physics 10 2. [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Du J, Schmall JP, Di K, Yang Y, Judenhofer M, Bec J, Buckley S, Jackson C and Cherry SR 2015a. Design and optimization of a high-resolution PET detector module for small-animal PET based on a 12 × 12 silicon photomultiplier array Biomed. Phys. Eng. Express 1 045003 [Google Scholar]
  11. Du J, Schmall JP, Yang Y, Di K, Dokhale PA, Shah KS and Cherry SR 2013. A Simple Capacitive Charge-Division Readout for Position-Sensitive Solid-State Photomultiplier Arrays IEEE Transactions on Nuclear Science 60 3188–97 [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Du J, Wang Y, Zhang L, Zhou Z, Xu Z and Wang X 2009. Physical Properties of LYSO Scintillator for NN-PET Detectors 2009 2nd International Conference on Biomedical Engineering and Informatics 2009 2nd International Conference on Biomedical Engineering and Informatics pp 1–5 [Google Scholar]
  13. Du J, Yang Y, Bai X, Judenhofer MS, Berg E, Di K, Buckley S, Jackson C and Cherry SR 2016. Characterization of Large-Area SiPM Array for PET Applications IEEE Trans Nucl Sci 63 8–16 [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Du J, Yang Y, Berg E, Bai X, Gola A, Ferri A, Zorzi N, Piemonte C and Cherry SR 2015b. Evaluation of linearly-graded SiPMs for high resolution small-animal PET Biomed. Phys. Eng. Express 1 045008 [Google Scholar]
  15. Dupont P and Warwick J 2009. Kinetic modelling in small animal imaging with PET Methods 48 98–103 [DOI] [PubMed] [Google Scholar]
  16. Gola A, Acerbi F, Capasso M, Marcante M, Mazzi A, Paternoster G, Piemonte C, Regazzoni V and Zorzi N 2019. NUV-Sensitive Silicon Photomultiplier Technologies Developed at Fondazione Bruno Kessler Sensors (Basel) 19 308. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. James SS, Yang Y, Wu Y, Farrell R, Dokhale P, Shah KS and Cherry SR 2009. Experimental characterization and system simulations of depth of interaction PET detectors using 0.5 mm and 0.7 mm LSO arrays Phys. Med. Biol 54 4605. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Kang HG, Yamaya T, Han YB, Song SH, Ko GB, Lee JS and Hong SJ 2020. Crystal surface and reflector optimization for the SiPM-based dual-ended readout TOF-DOI PET detector Biomed. Phys. Eng. Express 6 065028. [DOI] [PubMed] [Google Scholar]
  19. Karakatsanis NA, Lodge MA, Tahari AK, Zhou Y, Wahl RL and Rahmim A 2013. Dynamic whole body PET parametric imaging: I. Concept, acquisition protocol optimization and clinical application Phys Med Biol 58 7391–418 [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Kuang Z, Wang X, Fu X, Ren N, Yang Q, Zhao B, Zhang C, Wu S, Sang Z, Hu Z, Du J, Liang D, Liu X, Zheng H and Yang Y 2019. Dual-ended readout small animal PET detector by using 0.5 mm pixelated LYSO crystal arrays and SiPMs Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment 917 1–8 [Google Scholar]
  21. Lai Y, Wang Q, Zhou S, Xie Z, Qi J, Cherry SR, Jin M, Chi Y and Du J 2021. H2RSPET: a 0.5 mm resolution high-sensitivity small-animal PET scanner, a simulation study Phys. Med. Biol 66 065016. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Liu Z, Mungai S, Niu M, Kuang Z, Ren N, Wang X, Sang Z and Yang Y 2023. Edge effect reduction of high-resolution PET detectors using LYSO and GAGG phoswich crystals Phys. Med. Biol 68 065010. [DOI] [PubMed] [Google Scholar]
  23. Mao R, Zhang L and Zhu R-Y 2008. Emission Spectra of LSO and LYSO Crystals Excited by UV Light, X-Ray and γ-ray IEEE Transactions on Nuclear Science 55 1759–66 [Google Scholar]
  24. Miyaoka RS and Lehnert AL 2020. Small animal PET: a review of what we have done and where we are going Phys. Med. Biol 65 24TR04. [DOI] [PubMed] [Google Scholar]
  25. Niu M, Liu Z, Kuang Z, Wang X, Ren N, Sang Z, Wu S, Cong L, Sun T, Hu Z and Yang Y 2022. Ultra-high-resolution depth-encoding small animal PET detectors: Using GAGG and LYSO crystal arrays Medical Physics 49 3006–20 [DOI] [PubMed] [Google Scholar]
  26. Parellada-Monreal L, Acerbi F, Ficorella A, Franzoi A, Gola A, Merzi S, Nawaz A, Ruzzarin M and Paternoster G 2023. 3D integration technologies for custom SiPM: From BSI to TSV interconnections Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment 1049 168042 [Google Scholar]
  27. Ren S, Yang Y and Cherry SR 2014. Effects of reflector and crystal surface on the performance of a depth-encoding PET detector with dual-ended readout Medical Physics 41 072503. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Schmall JP, Du J, Yang Y, Dokhale PA, McClish M, Christian J, Shah KS and Cherry SR 2012. Comparison of large-area position-sensitive solid-state photomultipliers for small animal PET Phys. Med. Biol 57 8119–34 [DOI] [PubMed] [Google Scholar]
  29. Serra N, Ferri A, Gola A, Pro T, Tarolli A, Zorzi N and Piemonte C 2013. Characterization of new FBK SiPM technology for visible light detection J. Inst 8 P03019 [Google Scholar]
  30. Song TY, Wu H, Komarov S, Siegel SB and Tai Y-C 2008. Sub-millimeter resolution PET detector module using Multi-Pixel Photon Counter array 2008 IEEE Nuclear Science Symposium Conference Record 2008 IEEE Nuclear Science Symposium Conference Record pp 4933–7 Online: https://ieeexplore.ieee.org/document/4774347 [Google Scholar]
  31. Xie J, Wang H, Cherry SR and Du J 2024. PET Detectors Based on Multi-Resolution SiPM Arrays IEEE Transactions on Radiation and Plasma Medical Sciences 8 493–500 [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Yamamoto S, Watabe H, Kanai Y, Watabe T, Kato K and Hatazawa J 2013. Development of an ultrahigh resolution Si-PM based PET system for small animals Phys. Med. Biol 58 7875. [DOI] [PubMed] [Google Scholar]
  33. Yamamoto S, Watabe H, Watabe T, Ikeda H, Kanai Y, Ogata Y, Kato K and Hatazawa J 2016. Development of ultrahigh resolution Si-PM-based PET system using 0.32mm pixel scintillators Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment 836 7–12 [Google Scholar]
  34. Yang Y, Bec J, Zhou J, Zhang M, Judenhofer MS, Bai X, Di K, Wu Y, Rodriguez M, Dokhale P, Shah KS, Farrell R, Qi J and Cherry SR 2016. A Prototype High-Resolution Small-Animal PET Scanner Dedicated to Mouse Brain Imaging Journal of Nuclear Medicine 57 1130–5 [DOI] [PMC free article] [PubMed] [Google Scholar]

RESOURCES