Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2023 Apr 1.
Published in final edited form as: IEEE Trans Radiat Plasma Med Sci. 2021 Jun 7;6(4):385–392. doi: 10.1109/trpms.2021.3087465

Performance Characteristics of a Dual-Sided Position-Sensitive Sparse-Sensor Detector for Gamma-ray Imaging

William C J Hunter 1, Donald Q DeWitt 1, Robert S Miyaoka 1
PMCID: PMC8974312  NIHMSID: NIHMS1724922  PMID: 35372738

Abstract

Purpose:

We characterize the performance of a dualsided position-sensitive sparse sensor (DS-PS3) array detector for positron emission tomography (PET). The DS-PS3 detector is designed as a high performance, cost effective PET detector for organ-specific imaging systems (e.g., brain, breast, etc.).

Methods:

Two sparse 4-by-4 arrays of silicon photomultipliers (18.5% SiPM fill-factor) coupled through segmented light guide are used to readout a 15-by-15 array of 2-mm-pitch, 20-mm-long LSYO crystals. Uniform flood data were used for crystal identification, depth determination, and position-dependent energy resolution. Intrinsic-spatial and depth-of-interaction (DOI) resolutions were determined by stepping a collimated gamma-ray source over the front and side, respectively.

Results:

We measured an average intrinsic spatial resolution of 2.14 ± 0.07 mm full width at half maximum (FWHM). DOI FWHM resolution varied from 2.2 mm for crystals over sensors to 5.3 mm for crystals between sensors. Average DOI resolution was 3.6 ± 0.8 mm FHWM. Average energy resolution for the detector module was 16.6% with a range of 11.3% to 25.8%.

Conclusions:

We have demonstrated use of a dual-sided sparse sensor arrays to enable low-cost high-performance decoding of three-dimensional positioning within a PET detector using an 18.5% sensor fill-factor.

I. Introduction

This work examines the Dual-Sided Position-Sensitive Spare-Sensor (DS-PS3) detector in an effort to develop lower-cost high-performance PET detectors for organ-specific imaging system (e.g., for out-patient setting).

While most commercial PET systems sold are either whole body clinical PET/CT systems or preclinical small animal systems, there has been a developing market for organ-specific PET systems, mainly for human brain or breast imaging. Recent growth in high resolution PET systems dedicated for human neuroimaging has been fueled by the US government’s BRAIN Initiative, by the development of 18F-based neuroPET imaging agents for imaging of amyloid and tau proteins [1] and by a new 18F-based PET blood flow tracer, 18F-4-Fluoroantipyrine [2]. The research potential for ambulatory neuroPET has also generated a lot of enthusiasm for the development of lightweight, highly portable neuroPET imaging systems. Currently there are at least two commercial vendors of dedicated neuroPET imaging systems along with a number of research neuroPET imaging systems under development [3-14].

Similar to dedicated neuroPET imaging systems, there have been a number of dedicated research breast PET imaging systems developed over the years [15-22] including some with combined X-ray CT capability. A few of these dedicated breast PET imaging systems were commercialized and a couple are still actively being sold. While the early focus on the development of dedicated breast PET imaging systems was for the detection and characterization of small breast lesions, there is growing interest in using breast PET as a quantitative biomarker to assess tumor response to breast cancer therapies [23-25]. The advantages of dedicated organ-specific PET imaging systems are lower cost, small footprint, higher intrinsic spatial resolution and higher detection efficiency versus clinical whole body PET imaging systems. Lastly, an excellent overview of the historical and current development of organ-specific PET instrumentation is available in [26].

The DS-PS3 detector design reduces the number of photosensors and digital acquisition channels per detector, offering a significant cost savings at a time when photosensors and front-end electronics of a fully populated device is about one third of the full system cost. A sparse array of silicon-photomultiplier (SiPM) sensors is placed on two sides of a pixelated scintillator array in order to resolve interaction position (including depth) and energy of gamma-ray interactions. We note prior work that used a single-sided sparse sensor layout design in which custom light guides were used to enhance detector decoding performance [27].

Our use of a sparse sensor array to read out crystals required light sharing to resolve crystal elements smaller than the sensor pitch. However, in doing so with a continuous (non-segmented) light guide, we found it difficult to resolve the edge crystal elements from its interior neighbor (results not shown). Therefore, in this work, we also used optical barriers in the light guide between edge and interior crystals as a means to better distinguish edge crystals in a flood map.

The design of the DS-PS3, including scintillator type, detector thickness, and dual-sided readout, is targeted at as a non-TOF (time of flight) PET detector. As such, we evaluate the DS-PS3 performance in terms of 3D position and energy resolution only. The initial performance goals for the DS-PS3 detector were to be able to support <2 mm image resolution throughout the useful imaging field of view for dedicated organ-specific PET imaging systems (e.g., human brain or breast imaging).

II. Materials and Methods

Detector Design

Two 4-by-4 arrays of sparsely spaced SiPMs (Hamamatsu S13360-3050VE) are used for dual-sided readout of an LYSO scintillation crystal array (Fig. 3). The LYSO crystal array is segmented in 15-by-15 pixels with 2.0-mm pitch. Each LYSO pixel is 1.917 by 1.917 by 20 mm3 with sides that are lapped with 200-grit abrasive and ends (top and bottom) polished with 1200-grit abrasive. Roughness for the side treatment was selected from a limited parametric study of surface roughness on depth sensitivity using a set of three 3-by-3 mm2 crystal samples with varied side treatment (200, 300, and 400 grit); coarser grit improved depth sensitivity while finer grit improved net light output and energy resolution. We chose 200 grit for good depth sensitivity but decided not to use an even-coarser grit to avoid further signal loss. A 50-μm-thick reflector (LuMirror™, E60 by Toray®) is glued between LYSO pixels with Dymax® OP-20 optical adhesive. The finished LYSO array with reflectors is 30.0-mm by 30.0-mm wide by 20.0-mm tall.

Fig. 3.

Fig. 3.

Open in a new tab

Top: Side-view of a multi-layer DS-PS3 detector with side reflector removed to reveal the LYSO array that is sandwiched from above and below by custom-light guide (CSLG) and sparse-sensor-array layers (PCB-S). Middle Left: DS-PS3 detector module assembled with cooling block and PCB-A front-end electronics. Middle Right: Driver-amplifier PCB. Bottom: Data acquisition electronics board.

Two-sided readout of the scintillation crystals enables identification of the crystal of interaction as well as depth of interaction (DOI) [28, 29]. We use a custom segmented light-guide (CSLG) made of borosilicate-crown glass (Schott® BK-7) between the scintillator array and the SiPM sensors to enhance position decoding, especially for the edge crystals. Segments of the CSLG were cut perpendicular to the entrance surface with a diamond-blade saw and glued back together with 50-μm-thick Lumirror (a polyester reflector by Toray®) between segments; an optical adhesive (Dymax® OP-20) was used for this purpose. Optimal thickness of the CSLG depends on sensor spacing; several commercially available thicknesses ranging from 1.6 mm to 3.0 mm were tested. We selected a 2.2-mm CSLG for best crystal-peak resolution in a 2D flood histogram. A laser-cut mirror film (3M® ESR) covers the area on the output face of the light guide between sensors to preserve light that would otherwise be lost. A photograph of the 15-by-15 LYSO crystal array, custom light guide, sparse sensor SiPM array and mirror film mask is shown in Fig. 1. Dimensions and spacing of the sensors, light guides, and LYSO crystals are shown in Fig. 2.

Fig. 1.

Fig. 1.

Open in a new tab

Photograph of 15-by-15 LYSO crystal array, custom light guide assembly, prototype sparse-sensor-array printed circuit board (PCB) and ESR mirror film mask. Photograph of completed DS-PS3 detector is shown in Fig. 3.

Fig. 2.

Fig. 2.

Open in a new tab

DS-PS3 design: A 15-by-15 array of 2-mm-wide LYSO (thin-lined square grid) is used. The segmented light guide between LYSO and sensors appear as thick-lined rectangles. Horizontal spacing of the 4-by-4 sparsely spaced 3-mm-wide SiPMs are shown at the bottom. Mirror-film reflector (gray) fills the space between sensors on the light guide exit surface. Horizontal and vertical dimensions are similar, and all dimensions are in mm with 0.01mm precision.

The final detector assembly shown in Fig.3 is a sandwich of LYSO crystal array, light guides (CSLG), and two 20-mil (0.5-mm) thick printed circuit boards (PCB-S) with sparse SiPM arrays. We read out 32 SiPM and 2 thermistor signals per detector on two printed circuit boards (PCB-A) on the sides of the module. Currents from the SiPMs pass through transimpedance amplifiers on the PCB-A. At a future point, we will multiplex signals on the PCB directly after signal buffering. However, at present we are reading out all channels and are multiplexing in software.

A liquid-cooled aluminum block is coupled to the rear PCB-S board and the two side PCB-A boards. A solid copper ground plane on the PCB-A boards is used to thermally couple front and rear PCB-S boards.

SiPM signals are passed to a second-stage amplifier-driver board (PCB-D), which serves two purposes. First, signals are summed for event triggering and timing. Second, each SiPM signal undergoes single-ended to differential conversion before they are passed to the data-acquisition board (UW DAQ v.2) [30]. The UW DAQ v.2 data-acquisition board digitizes 64 signal channels with a 60-MHz sampling rate. Signals are integrated for 32 samples (533 nsec). The UW DAQ v.2 also has a 65th channel sampled at 300 MHz that is used for triggering and timing.

A complete detector module and supporting electronics are shown in Fig. 3. The detector has been designed so that it can be tightly packed in either a panel-based arrangement [18, 31] as shown Fig. 4a, or in a detector ring geometry as shown in Fig. 4b.

Fig. 4a.

Fig. 4a.

Open in a new tab

Artist rendering of tightly packed detectors supporting panel-based detector geometry.

Fig. 4b.

Fig. 4b.

Open in a new tab

Photograph of a single ring of DS-PS3 modules (a work in progress for the NINDS Brain Initiative).

Event estimation

We first estimate depth of interaction (DOI) by partitioning the Depth-Signal-Ratio (DSR) distribution for this dual-sided-readout detector. DSR is computed as the sum of entrance-face sensor signals divided by the sum of sensor signals from both top and bottom. Next, we estimate the crystal of interaction (lateral position of event) using a depth-dependent segmentation of the signal centroid (a.k.a. crystal map). We then improve our depth estimation, by recomputing DSR using a subset of sensors for the given crystal of interaction; specifically, sensors with at least 50% of the maximum sensor signal (on average) and the complimentary set of sensors on the opposite side (top or bottom) are used to compute DSR. Finally, for a given crystal and depth, energy is estimated as the gain-corrected signal sum normalized by the calibrated 3D-position-dependent photopeak energy.

Calibration

We measure detector response for a large ensemble of events in order to later estimate position and energy of interactions when imaging. We do so by using a 30-cm-distant 50-μCi Ge-68 point source to flood-illuminate the DS-PS3 detector. Roughly 200,000 un-filtered events are collected for each crystal (fewer for crystal-array edges due to scattered secondaries that escape). Measurement is performed in a temperature-controlled dark box at 12.5°C.

Calibration is performed iteratively, starting with identification of peaks in the crystal map for all events without energy filtering (Fig. 5). Events are partitioned by crystal and a photopeak is determined from the gain corrected sum-signal. A course energy window of −25% to +50% of the photopeak is applied and the remaining data is used to refine peak positions in the crystal map. Events are re-partitioned by crystal. Next energy photopeaks for each crystal are refined.

Fig. 5.

Fig. 5.

Open in a new tab

Map of the 2D signal-centroid distribution, and partitions equidistant from crystal peaks used to identify crystal of interaction.

The first iteration of the depth signal ratio (DSR) is then computed using all sensors (not just the local sensor). Events for each crystal are equally partitioned by DSR. The mean depth for each DSR bin is computed assuming a monotonic relation between DSR and interaction depth, and assuming that interaction depth is exponentially distributed with an attenuation coefficient of 0.0834/mm (combined attenuation coefficient for photoelectric and Compton interactions in LYSO). Depth-dependence of crystal peak positions and photopeak energies are computed. Events for each depth in each crystal are then energy windowed. This process of crystal positioning, depth binning, and energy windowing is iterated until positions and energies have converged. We finally compute the mean signal for each sensor in each crystal at all depths and use this information to select the group of sensors for computing local DSR for each crystal.

We note that full calibration of the 3-dimensional positioning within the detector module is done via flood illumination of the detector module from its entrance surface. Using our methods, detector modules can be calibrated and adjusted while residing within the imaging system. Side illumination is not required to calibrate the DOI positioning estimator.

Characterization

Two scans of a collimated source are performed to characterize detector performance. One scan with the source in front of the entrance face (Fig. 6) is used to characterize lateral spatial resolution (i.e., detector array intrinsic spatial resolution). A second scan with the source to the side (Fig. 7) is used to characterized depth-positioning and energy-resolution performance.

Fig. 6.

Fig. 6.

Open in a new tab

Electronic-collimation setup used for lateral-positioning performance.

Fig. 7.

Fig. 7.

Open in a new tab

Electronic-collimation setup used for depth-positioning and energy-resolution performance. The position of the source in this figure is not accurate and was actually shifted to the right by 2 cm for the depth characterization.

A small, distant coincidence detector is used to electronically-collimate a 0.25-mm 50-μCi Na-22 point source. The concidence detector is a 0.8mm-by-8.0mm wide and 10 mm thick LYSO crystal that was read out by a compact metal-dynode PMT (Hamamatsu R9880U-110) to readout this single-element LYSO coincidence detector. Both acquisitions were performed in a temperature-controlled dark box at 12.5°C.

For characterization of lateral resolution, the source was placed 3 cm from the entrance face of the DS-PS3 detector and 6 cm from the coincidence detector; the 8-mm dimension of the coincidence crystal was parallel to the rows of the DS-PS3 crystal array. In this manner coincidence events for one beam position covered 20% of a single row (0.4mm width) and was aligned to cover two whole columns (4 mm width). The beam for this entrance-face scan was translated in 0.2mm steps along the row dimension.

For characterization of depth resolution, the source was placed 5 cm from the side surface of the DS-PS3 ad 4 cm from the coincidence detector; in this case, the 8-mm dimension of the coincidence crystal was parallel to the width of the crystals and the 0.8-mm dimension along the depth direction. The coincidence beam in this configuration was 1.0 mm in depth and 10.0 mm in width, fully covering 4 crystal elements and partially covering 2 additional crystals. The beam for this scan over the side was translated in 0.2mm steps along the depth dimension.

III. Results

We show in Fig. 8 a flood map from normal entrance-face illumination of the DS-PS3. Also shown are line profiles that are interpolated on a spline curve passing through crystal peaks for rows (blue) and columns (red). We find crystal elements are well resolved with an average peak-to-valley ratio of 5.9. We also note that the best crystal decoding occurs for crystals that are not directly viewed by individual SiPM dies.

Fig. 8.

Fig. 8.

Open in a new tab

Flood 2D and 1D profiles

A coincidence-collimated fan beam as shown in Fig. 6 was stepped over the entrance face of the DS-PS3. Photopeak events (±15%) were partitioned by crystal as shown in Fig. 13. The resulting counts versus position were mapped for each crystal and the full-width at half-max (FWHM) intrinsic resolution was evaluated. 1D count profiles for six crystals are shown in Fig. 9. The FWHM intrinsic spatial is measured to be 2.14 ± 0.07 mm.

Fig. 13.

Fig. 13.

Open in a new tab

Depth-averaged energy resolution using collimated source from side

Fig. 9.

Fig. 9.

Open in a new tab

Intrinsic spatial resolution using a collimated source scanned across the front of the crystal array.

The process of depth decoding by equipartitioning of DSR for each crystal is shown in Fig. 10a. Depth resolution is assessed by scanning in depth a coincidence-collimated fan beam through the side of the DS-PS3 (as shown in Fig. 7). On the left side of Fig. 10b, events are grouped by depth (DSR bin) and a depth-dependent flood map (after the positions and energies have converged) are shown; differences in these the flood maps versus DSR bin are subtle and best captured by the migration of the crystal peak positions (shown on the right of Fig. 10b). In Fig. 11 we show the resulting crystal map of events incident from the side at two fan-beam positions. The distribution of DSR at several depths separated by 4-mm are shown for 4 representative crystals in Fig. 12. The color of the boxes, as used in Figs. 10a, 11, 12, and 13, indicate which crystals each DSR distribution profile represents. Depth resolution for each crystal is computed as a function of depth by multiplying the DSR distribution FWHM by 4 mm and dividing by the DSR peak separation about each depth. Depth resolution for each crystal is then averaged over all depths. We find an average depth resolution of 3.6 mm across all crystals and depths. We also show an energy histogram for these four representative crystals at a central depth in Fig. 13. An average energy resolution of 16.6% with a range from 11.3% to 25.8% was measured across crystals at this depth.

Fig. 10a.

Fig. 10a.

Open in a new tab

Left: signal-centroid event distribution using flood illumination. Right: distribution of depth signal ratio (DSR) for one quadrant of 15-by-15 crystal array; the DSR range graphed is [0,1]. The width of the signal ratio is partitioned for depth of interaction decoding (i.e., wider is better). Circled on the 2D map and outlined on the DSR distribution are crystals with Row (R) and Column (C) indices of R5C1 (Red), R5C8 (Purple), R8C1 (Blue), and R8C8 (Green) for which we subsequently illustrate the depth and energy performance.

Fig. 10b.

Fig. 10b.

Open in a new tab

Left: Depth-dependent flood maps after the positions and energies converged. Right: color-coded peak position versus depth (blue is DSR bin #1 and red is DSR bin #10).

Fig. 11.

Fig. 11.

Open in a new tab

Left: collimated source at side of DS-PS3 in line with crystal row 5 (top) and row 8 (bottom); spacing between coincidence detector, source and DS-PS3 are not to scale. Right: resulting signal-centroid event distribution with same four crystals indicated in Fig. 10a.

Fig. 12.

Fig. 12.

Open in a new tab

Depth of interaction resolution of four reference crystals using collimated source from side (see Fig. 10)

IV. Discussion

Our use of a sparse sensor array to read out crystals required light sharing to resolve crystal elements smaller than the sensor pitch. However, in doing so with a continuous (non-segmented) light guide, we found it difficult to resolve the edge crystal elements from its interior neighbor (results not shown). Therefore, in this work, we used optical barriers in the light [27] guide between edge and interior crystals as a means to better distinguish edge crystals in a flood map. We believe that adding additional light barriers in the light guide could improve overall decoding performance, especially for crystals directly viewed by SiPM dies. On the other hand, current crystal decoding performance is excellent and for the current detector it was not deemed necessary.

We see a migration of crystal-peak positions versus depth in Fig. 10b right. Although there does not appear to be an obvious connection, one possible explanation is a spatial-dependent temperature effect on SiPM gains. A thermistor was located in the front center of each sensor board (PCB-S). Using pairs of calibrated thermistors in this fashion, we found temperatures on the PCB adjacent to the sensors during normal detector operation to be within 1°C on the front and back PCB-S boards. However, we do not know what the differences between the PCB-S temperature and the internal temperature of the SiPMs are; we might expect this difference for SiPMs close to the center on the front and back to be close to the temperature difference measured by the thermistors on the front and back PCB-S. Additionally, we also do not know if there are different temperature gradients across the the front and back PCB-S. This is a topic for further investigation.

The average energy resolution for all the crystals in the array was 16.6% with a range from 11.3% to 25.8%. The crystals directly viewed by SiPM dies had the best energy resolution. The single central crystal, with the poorest light collection efficiency, had 25.8% energy resolution. The range in the energy resolution for the individual crystals can easily be reduced by placing an additional SiPM element over the center crystal. This would only increase the total number of SiPM sensors from 32 to 34 but would make the overall performance of the detector module more uniform (i.e., both energy resolution and DOI positioning resolution). This is something that we are considering for future designs.

Using a coincidence-collimated beam incident upon the DS-PS3, we have shown the Depth Signal Ratio (DSR) to be monotonic with depth and that a broader spread of DSR signifies improved depth resolution, Fig. 10a right. For this study, we have assumed the depth distribution of events equipartitioned by DSR are exponentially distributed. However, energy filtration of Compton multiple-interaction Compton-escape events result in a non-exponential distribution [32]. This effect will bias DOI distribution, enlarging the first and last depth bins.

We have stated that the DS-PS3 design was intended for non-TOF PET for organ-specific applications. In an organ-specific application, the smaller field of views would require a coincidence-timing resolution of 300 psec or better to be useful for TOF. However, our preliminary optical-ray-trace simulations indicate the added variance of optical pathlengths in a sparse-readout detector design would not achieve this goal. A more detailed simulation or experimental verification would need to be performed to verify this conclusion. We did not do so in the current study due to our use of a non-TOF data-acquisition system.

V. Conclusion

We have reported the performance of a first-generation design for the Dual-Sided Position-Sensitive Sparse-Sensor (DS-PS3) detector. All crystals of a 2.0-mm pitch 15-by-15 array of LYSO crystals were well resolved, with an average peak-to-valley ratio of 5.9 for the crystal peaks in the 2D-histogram of the energy-weighted-centroid positions. The average intrinsic spatial resolution was 2.14 mm. The average depth resolution was 3.6 mm, and average energy resolution was 16.6%.

In this work, we have demonstrated use of a dual-sided sparse sensor arrays to enable low-cost high-performance decoding of three-dimensional positioning within a PET detector using an 18.5% sensor fill-factor. Therefore, even when using dual-sided readout, our DS-PS3 detector uses less than 40% of the sensors that are used in a tightly packed SiPM array in a single-sided readout device.

In future work, we will look to further improve performance verses cost by refining photosensor positions and fill factor. In particular, we think positioning the sparse photosensors to be centered on scintillator pixels (rather than between) will help to better separate the few crystal peaks that were previously symmetrically placed about the sensor. Adjusting sensor placement in this way should also enable us to decrease the crystal pixel size and still maintain crystal identification in all cases. In addition, hardware multiplexing methodologies are being explored to minimize the number of signal channels that need to be collected without compromising 3-dimensional positioning performance and studies are being designed to explore the coincidence timing characteristics of DS-PS3 detectors.

Acknowledgments

This work was supported by R21EB020420.

VI. Bibliography

  • [1].Okamura N, Harada R, Furukawa K, Furumoto S, Tago T, Yanai K, Arai H and Kudo Y, "Advances in the development of tau PET radiotracers and their clinical applications," Ageing Res. Rev, vol. 30, p. 107–113, Sept. 2016. [DOI] [PubMed] [Google Scholar]
  • [2].Sako K, Diksic M, Kato A, Yamamoto YL and Feindel W, "Evaluation of [18F]-4-Fluoroantipyrine as a New Blood Flow Tracer for Multiradionuclide Autoradiography," J. Cerebr. Blood F. Met, vol. 4, no. 2, pp. 259–263, 1984. [DOI] [PubMed] [Google Scholar]
  • [3].Herzog H, Langen K-J, Weirich C, Kops ER, Kaffanke J, Tellmann L, Scheins INJ, Stoffels G, Fischer K, Caldeira L, Coenen HH and Shah NJ, "High resolution Brain PET combined with simultaneous MRI," Nuklearmedizin, vol. 50, no. 2, p. 74–82, 2011. [DOI] [PubMed] [Google Scholar]
  • [4].Shah N, Oros-Peusquens A-M, Arrubla J, Zhan K, Warbrick T, Mauler J, Vahedipour K, Romanzetti S, Felder J, Celik A, Rota-Kops E, Iida H, Langen K-J, Herzog H and Neuner I, "Advances in multimodal neuroimaging: Hybrid MRPET and MR-PET-EEG at 3 T and 9.4 T," J. Magn. Reson, vol. 229, p. 101–115, Apr. 2013. [DOI] [PubMed] [Google Scholar]
  • [5].Caldeira L, Weirich C, Almeida P and Herzog H, "Reconstruction of PET data acquired with the Brain PET using STIR," in IEEE Nucl. Sci. Symp. Med. Imag. Conf. (NSS/MIC), Anaheim, CA, USA, 2012. [Google Scholar]
  • [6].Hong KJ, Choi Y, Jung JH, Kang J, Hu W., Lim HK, Huh Y, Kim S, Jung JW, Kim KB, Song MS and Park H-W, "A prototype MR insertable brain PET using tileable GAPD arrays," Med. Phys, vol. 40, no. 4 Art. no. 04250 , 2013. [DOI] [PubMed] [Google Scholar]
  • [7].Nishikido F, Obataa T, Shimizub K, Sugab M, Inadamaa N, Tachibanaa A, Yoshidaa E, Itoa H and Yamaya T, "Feasibility of a brain-dedicated PET-MRI system using four-layer DOI detectors integrated with an RF head coil," Nucl. Instrum. Methods Phys. Res. Section A, vol. 756, p. 6–13, 2014. [Google Scholar]
  • [8].Del Guerra A, Ahmad S, Avram M, Belcari N, Berneking A, Biagi L, Bisogni MG, Brandl F, Cabello J, Camarlinghi N, Cerello P, Choi C-H, Coli S, 0 SC, Fleury J, Gagliardi V and Gira G, "TRIMAGE: A dedicated trimodality (PET/MR/EEG) imaging tool for schizophrenia," Eur. Psychiatry, vol. 50, pp. 7–20, Apr, 2018. [DOI] [PubMed] [Google Scholar]
  • [9].Ahmad S, Fleury J, de la Taille C, Seguin-Moreau N, Dulucq F, Martin-Chassard G, Callier S, Thienpont D and Raux L, "Triroc: A multi-channel SiPM read-out ASIC for PET/PET-ToF application," IEEE Trans. Nucl. Sci, vol. 62, no. 3, p. 664–668, Jun. 2015. [Google Scholar]
  • [10].Benlloch JM, González AJ, Pani R, Preziosi E, Jackson C, Murphy J, Barberá J, Correcher C, Aussenhofer S, Gareis D, Visvikis D, Bert J, Langstrom B, Farde L, Toth M, Haggkvist J, Caixeta FV, Kullander K and Somlai-Sc, "The MINDVIEW project: First results," Eur. Psychiatry, vol. 50, p. 21–27, Apr. 2018. [DOI] [PubMed] [Google Scholar]
  • [11].González-Montoro A, Aguilar A, Canilleras G, Conde P, Hernández L, Vidal LF, Galasso M, Fabbri A, Sánchez F, Benlloch JM and González AJ, "Performance study of a large monolithic LYSO PET detector with accurate photon DOI using retroreflector layers," IEEE Trans. Radiat. Plasma Med. Sci, vol. 1, no. 3, p. 229–237, May 2017. [Google Scholar]
  • [12].Grogg KS, Toole T, Ouyang J, Zhu X, Normandin MD, Li Q, Johnson K, Alpert NM and El Fakhri G, "National electrical manufacturers association and clinical evaluation of a novel brain PET/CT scanner," J. Nucl. Med, vol. 57, no. 4, p. 646–652, 2016. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [13].Moliner L, Rodríguez-Alvarez M, Catret J, González A, Ilisie V and Benlloch M, "NEMA Performance Evaluation of CareMiBrain dedicated brain PET and Comparison with the whole-body and dedicated brain PET systems.," Sci. Rep, vol. 9, no. no. 15484, 2019. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [14].Won JY, Park H, Lee S, Son J-W, Chung Y, Ko GB, Kim KY, Song J, Seo S, Ryu Y, Chung J-Y and Lee JS, "Development and Initial Results of a Brain PET Insert for Simultaneous 7-Tesla PET/MRI Using an FPGA-Only Signal Digitization Method," Trans Med Imag, vol. 1, no. 1, p. 1, 2021. [DOI] [PubMed] [Google Scholar]
  • [15].Weinberg IN, Beylin D, Zavarzin V, Yarnall S, Stepanov PY, Anashkin E, Narayanan D, Dolinsky S, Lauckner K and Adler LP, "Positron emission mammography: High- resolution biochemical breast imaging,," Technol. Cancer Res. Treat, vol. 4, no. 1, p. 55–60, 2005. [DOI] [PubMed] [Google Scholar]
  • [16].Raylman RR, "Positron emission tomography-guided biopsy with a dedicated breast scanner: Initial evaluation," IEEE Trans. Nucl. Sci, vol. 56, no. 3, p. 620–624, 2009. [Google Scholar]
  • [17].Moliner L, Gonzalez AJ, Soriano A, Sanchez F, Correcher C, Orero A, Carles M, Vidal LF, Barbera J, Caballero L, Seimetz M, Vazquez C and Benlloch JM, "Design and evaluation of the MAMMI dedicated breast PET," Med. Phys, vol. 39, no. 9, p. 5393–5404, 2012. [DOI] [PubMed] [Google Scholar]
  • [18].MacDonald LR, Hunter WCJ, Kinaha PE and Miyaoka RS, "Effects of Detector Thickness on Geometric Sensitivity and Event Positioning Errors in the Rectangular PET/X Scanner," IEEE Trans. Nucl. Sci, vol. 60, no. 5, pp. 3242–3252, 2013. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [19].Miyake KK, Matsumoto K, Inoue M and e. al, "Performance evalu-ation of a new dedicated breast PET scanner using NEMA NU4-2008 standards," J Nucl Med., vol. 55, no. 7, p. 1198–1203, 2014. [DOI] [PubMed] [Google Scholar]
  • [20].Hsu DFC, Freese DL, Reynolds PD, Innes DR and Levin CS, "Design and performance of a 1 mm3 resolution clinical PET system comprising 3-D position sensitive scintillation detectors," IEEE Trans. Med. Imag, vol. 37, no. 4, p. 1058–1066, Apr. 2018. [DOI] [PubMed] [Google Scholar]
  • [21].MacDonald L, Edwards J, Lewellen T, Haseley D, Rogers J and Kinahan P, "Clinical Imaging Characteristics of the Positron Emission Mammography Camera: PEM Flex Solo II," J Nucl Med, vol. 50, p. 1666–1675, 2009. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [22].Wu Y, Bowen SL, Yang K and e. al., "PET characteristics of a dedi-cated breast PET/CT scanner prototype," Phys. Med. Biol, vol. 54, no. 13, p. 4273–4287, 2009. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [23].Humbert O, Lasserre M, Bertaut A, Fumoleau P, Coutant C, Brunott F and Cochet A, "Breast Cancer Blood Flow and Metabolism on Dual-Acquisition 18F-FDG PET: Correlation with Tumor Phenotype and Neoadjuvant Chemotherapy Response," J. Nucl Med, vol. 59, no. 7, pp. 1035–1041, Jul 2018. [DOI] [PubMed] [Google Scholar]
  • [24].Connolly RM, Leal JP, Solnes L, Huang C-Y, Carpenter A, Gaffney K, Abramson V, Carey LA, Liu MC, Rimawi M, Specht J, Storniolo AM, Valero V, Vaklavas C, Krop IE, Winer EP, Camp M, Miller RS and Wolff AC, "TBCRC026: Phase II Trial Correlating Standardized Uptake Value With Pathologic Complete Response to Pertuzumab and Trastuzumab in Breast Cancer," J. Clin. Onc, vol. 37, no. 9, pp. 714–722, 2019. [DOI] [PMC free article] [PubMed] [Google Scholar] [Retracted]
  • [25].Hathi DK, Li W, Seo Y, Flavell RR, Kornak J, Franc BL, Joe BN, Esserman LJ, Hylton NM and Jones EF, "Evaluation of primary breast cancers using dedicated breast PET and whole-body PET," Sci. Rep, vol. 10, no. 1, p. 21930, December 2020. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [26].González AJ, Sánchez F and Benlloch JM, "Organ-Dedicated Molecular Imaging Systems," IEEE Trans. Radiat. Plasma Med. Sci, vol. 2, no. 5, pp. 388–403, September 2018. [Google Scholar]
  • [27].Song T, Wu H, Komarov S, Siegel S and Tai C, "A sub-millimeter resolution PET detector module using a multi-pixel photon counter array," Phys. Med. Biol, vol. 55, p. 2573–2587, 2010. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [28].Moses WW, Derenzo SE, Melche CL and Manente RA, "A room temperature LSO/PIN photodiode PET detector module that measures depth of interaction.," IEEE Trans. Nucl. Sci, vol. 42, p. 1085–1089, 1995. [Google Scholar]
  • [29].Shao Y, Meadors K, Silverman R, Farrell R, Cirignano L, Grazioso R, Shah K and Cherry S, "Dual APD array readout of LSO crystals: optimization of crystal surface treatment," IEEE Trans. Nucl. Sci, vol. 49, no. 3, pp. 649–654, June 2002. [Google Scholar]
  • [30].Lewellen TK, DeWitt D, Miyaoka RS and Hauck S, "A Building Block for Nuclear Medicine Imaging Systems Data Acquisition," IEEE Trans. Nucl. Sci, vol. 61, no. 2, p. 79–87, Feb 2014. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [31].Qi J, Kuo C, Huesman RH, Klein GJ, Moses WW and Reutter a. B. W., "Comparison of rectangular and dual- planar emission mammography scanners," IEEE Trans. Nucl. Sci, vol. 49, no. 5, p. 2089–2096, October 2002. [Google Scholar]
  • [32].Hunter WC, Barret HH and Furenlid LR, "Calibration Method for ML Estimation of 3D Interaction Position in a Thick Gamma-Ray Detector," IEEE Trans. Nucl. Sci, vol. 56, no. 1, p. 189–196, Feb 2009. [DOI] [PMC free article] [PubMed] [Google Scholar]

RESOURCES