High resolution depth coding PET detector with quasi-cylinder light guide array

Abstract

The invention discloses a particle detection device and a manufacturing method thereof. The particle detection device includes: a scintillator array including a plurality of scintillator crystals; a plurality of detectors disposed at a first end of the scintillator array; and the plurality of quasi cylinders are arranged at the second end of the scintillator array. A pseudo-cylinder of the plurality of pseudo-cylinders is disposed as a particle that is redirected between the second ends of the crystals of the scintillator array. A first end of a first set of crystals of the scintillator array is disposed to direct particles to a first detector of the plurality of detectors, and a first end of a second set of crystals of the scintillator array is disposed to direct particles to a second detector substantially adjacent to the first detector.

Description

[0001] Divisional application

[0002] This invention is a divisional application of the invention patent application with application number 202080001116.0. The application date of the original application is February 14, 2020, the earliest priority date is February 15, 2019, the earliest priority number is 62 / 806,035, and the other priority date is October 16, 2019, the priority number is 62 / 915,676, and the invention name is "High-resolution depth-encoded PET detector with pseudo-pillar light guide array".

[0003] CROSS-REFERENCE TO RELATED APPLICATIONS

[0004] This application claims the benefit of U.S. Provisional Patent Application Nos. 62 / 806,035 and 62 / 915,676, filed on February 15, 2019 and October 16, 2019, respectively, in the U.S. Patent and Trademark Office, which are hereby incorporated by reference in their entireties.

[0005] Government Rights

[0006] This invention was made with government support under Grant No. EB024849 awarded by the National Institutes of Health. The U.S. Government has certain rights in this invention. Technical Field

[0007] The present invention relates generally to the field of radiation imaging and more particularly to positron emission tomography (PET). Background Art

[0008] PET molecular imaging is a powerful technique used primarily for diagnosis, treatment selection, therapy monitoring, and research in cancer[1] and neuropsychiatric disorders[2]. Despite its high molecular specificity, quantification, and clinical utility, PET, as an essential molecular imaging method, has not yet reached its full potential due to its relatively poor spatial resolution, currently on the order of 3–6 mm[3,4]. With this degree of spatial resolution, it is not possible to measure target density in small nodules and many brain regions of humans and rodents that are relevant to disease etiology and pathophysiology.

[0009] Depth-encoded PET detector modules have been developed to mitigate the parallax error (misalignment of the lines of response) in long scintillator crystals [5]. When using crystals with a small cross-section, it is possible to achieve such small-diameter PET rings that can reduce the component cost of each detector ring, provide a larger solid angle coverage to improve sensitivity, and reduce the contribution of annihilation gamma-ray collinearity to the spatial resolution [4,6]. Additionally, the interaction depth (DOI) information can be used to deconvolve the photon transport in long crystals, thereby improving the time resolution [7,8]. A depth-encoded detector based on dual-ended readout has achieved an optimal continuous DOI resolution of less than 2 mm [9,10]. High-resolution PET systems such as dedicated breast Clear-PEM

[11] have been developed using dual-ended DOI readout detectors, but compared with standard single-ended readout PET scanners, these systems require a huge number of readout electronics, so their commercialization cost is too high. Due to the use of glass light guides at the crystal readout interface, a recently developed high-resolution variant of these detectors shows relatively poor energy resolution and time resolution, which are necessary for accurate crystal identification

[12] . Alternative single-ended readout detector modules have been proposed to obtain DOI information, such as multi-layer phosphors [13,14], retroreflectors for modules with integral scintillators

[15] , and other custom reflector designs [16,17]. However, in all these designs, a trade-off is required between depth encoding, cost, scintillator-to-readout coupling ratio, crystal identification accuracy, energy resolution, and time resolution. To reduce these trade-offs, an ideal depth-encoded detector module is a module with single-ended readout, in which the crystal array is directly coupled to silicon photomultiplier (SiPM) pixels without any intermediate glass light guide, to minimize the sharing of the entire scintillation photons propagating downward among multiple pixels and maintain good time resolution. In addition, the upward-propagating photons, which do not contribute to the time information, should be redirected by bending their path 180° towards the nearest adjacent SiPM to maintain good energy resolution and DOI resolution and mimic the behavior of dual-ended depth-encoded readout detectors.

[0010] Therefore, to develop a practical and cost-effective high-resolution time-of-flight (TOF) PET scanner and achieve continuous DOI localization using single-ended readout, detector modules composed of a frosted polycrystalline scintillator array have been recently studied, with one side coupled to SiPM pixels at a ratio of 4 to 1 and the other side provided with a uniform glass light guide [8,18,19]. See U.S. Patent No. 10,203,419 to Frazao et al., the content of which is incorporated herein by reference. In these detector modules, the energy-weighted average method is used for crystal identification, which uses 1.53 x 1.53 x 15 mm 3crystals and 3×3 mm 2 SiPM pixels to achieve an energy resolution of 9% and a DOI resolution of 3 mm full width at half maximum (FWHM), respectively [8]. However, due to the lack of optically shared neighbors, the crystal identification of these arrays is poor along their edges and corners

[19] , which is a problem that must be solved because the pixels at the edges and corners include 75% of the 4×4 SiPM readout chips and 44% of the 8×8 SiPM readout chips, respectively. At the same time, when using a uniform glass light guide, since many upward-propagating photons are reflected back to the first column, the remaining photons are isotropically shared between adjacent columns in a Gaussian intensity distribution manner, resulting in low inter-crystal light sharing efficiency. The problem caused by isotropic light sharing is that the integrity of the distribution of low-intensity signals in many SiPMs will be severely affected by dark counts, which will lead to a decrease in energy resolution and DOI resolution. Summary of the Invention

[0011] To overcome the disadvantages of conventional systems, the present invention provides a particle detector and an operation method thereof based on a Prism-PET detector module.

[0012] Accordingly, various aspects of the present invention solve the above problems and disadvantages and provide the following advantages. One aspect of the present invention provides a particle detection device, which includes: a scintillator array including a plurality of scintillator crystals; a plurality of detectors disposed at the bottom end of the scintillator array; and a plurality of prisms disposed at the top end of the scintillator array. Each of the plurality of prisms is configured to redirect particles between the tops of the crystals of the scintillator array. The bottom ends of the first group of crystals of the scintillator array are configured to guide particles to a first detector among the plurality of detectors, and the bottom ends of the second group of crystals of the scintillator array are configured to guide particles to a second detector, and the second detector is generally adjacent to the first detector.

[0013] Another aspect of the present invention provides a particle detector, which includes: a scintillator array including a plurality of scintillator crystals, a plurality of detectors disposed on the bottom end of the scintillator array, a plurality of prisms disposed on the top end of the scintillator array, and at least one processor operably communicating with the plurality of detectors. The at least one processor includes a plurality of supervised machine learning algorithms, and the plurality of supervised machine learning algorithms are configured to perform 3D gamma-ray localization on at least one interaction position within at least one of the plurality of scintillator crystals. Brief Description of the Drawings

[0014] Through the following detailed description in conjunction with the drawings, the above and other aspects of the present invention, as well as the features and advantages of certain embodiments, will become more apparent, where:

[0015] Figure 1Shows a uniform glass light guide module in a known detection system and the light distribution using this light guide module,

[0016] Figure 2 Shows a detector module according to an embodiment of the present invention, which has a special pattern of a segmented quasi-cylindrical light guide,

[0017] Figure 3 Shows the use of Figure 1 The light distribution simulated in TracePro software using the module shown,

[0018] Figure 4 Shows the use of Figure 2 The light distribution simulated in TracePro software using the module shown according to an embodiment of the present invention,

[0019] Figure 5 Shows the arrangement of a 4-to-1 coupled Prism-PET module according to an embodiment of the present invention,

[0020] Figure 6 Is a perspective view showing in detail a quasi-cylindrical array according to an embodiment of the present invention,

[0021] Figure 7 Shows the arrangement of a 9-to-1 coupled Prism-PET module according to an embodiment of the present invention,

[0022] Figure 8 Provides a perspective view of a light guide array according to an embodiment of the present invention,

[0023] Figure 9 Shows the 4-to-1 coupled detector readout with uniform glass of a known detector module,

[0024] Figure 10 Shows the detector readings of a 4-to-1 coupled Prism-PET module according to an embodiment of the present invention,

[0025] Figure 11 Shows the detector readings of a 9-to-1 coupled Prism-PET module according to an embodiment of the present disclosure,

[0026] Figure 13A Provides a crystal identification histogram according to an embodiment of the present invention, which is based on the centroid and the energy histograms measured with and without DOI filtering, and compares the results of the 4-to-1 coupled detector readout with uniform glass of a known detector module and the results using the detector module,

[0027] Figure 13B Provides the DOI resolution of a uniform glass light guide module of a known detector module, Figure 13CThe DOI resolution of a 4:1 coupled Prism-PET module according to an embodiment of the present invention is provided, and Figure 13C a comparison of the DOI resolutions of the two is provided,

[0028] Figure 14A - 14D a measured DOI resolution map of a 4:1 coupled Prism-PET module according to an embodiment of the present invention is provided,

[0029] Figure 15A - 15D the sensitivity maps and dimensions of several different PET scanners according to embodiments of the present invention are provided,

[0030] Figure 16A - 16F a theoretical light distribution of Compton interactions side by side in a 4:1 coupled Prism-PET module according to an embodiment of the present invention is shown, and

[0031] Figure 17 measurements of photoelectric and Compton interactions in a 4:1 coupled Prism-PET module according to an embodiment of the present invention are shown. Detailed Description of the Invention

[0032] Certain embodiments of the present invention will be described in detail below with reference to the accompanying drawings. When describing the present invention, in order to clearly understand the concept of the present invention, the explanations of related functions or structures known in the art are omitted so as not to obscure the present invention with unnecessary details.

[0033] The present invention discloses a single-ended readout depth-encoded detector module that utilizes a special pattern of a segmented quasi-cylindrical light guide. Among the features of the Prism-PET detector module disclosed in various embodiments, at least three different quasi-cylindrical designs are utilized, namely a central quasi-cylinder, an edge quasi-cylinder, and a corner quasi-cylinder, each of which has a different predefined design to reduce edge and corner artifacts, thereby achieving uniform crystal identification performance.

[0034] The light sharing between crystals is limited to the crystals belonging to the nearest SiPM neighbors to create a deterministic and anisotropic inter-crystal light sharing pattern and maximize the signal-to-noise ratio of those SiPMs, thereby improving both the energy resolution and the DOI resolution simultaneously.

[0035] The segmented mode improves crystal identification by decoupling adjacent crystals, which would otherwise have similar readout patterns. The shape of each pseudocolumn is interchangeable, and embodiments of the pseudocolumn are generally shaped as at least one of a prism, an anti-prism, a frustum, a triangle, a domed column, a parallelepiped, a wedge, a pyramid, a frustum of a pyramid, a spherical section, a cuboid, and a pyramid with a rectangular base. For ease of reference, a right triangular prism is discussed in the present invention, which improves the light sharing rate between crystals, thereby improving crystal identification and DOI resolution.

[0036] When photons enter the hypotenuse of the right triangular prism, they are deflected by 180°, thus being effectively guided to adjacent crystals, which are coupled to different readout pixels according to different offset crystal-prism coupling schemes related to crystal-pixel coupling, as Figure 2 、 4 shown in 8, 10 to 12, and 13B to 16.

[0037] Figure 1 Shows a uniform glass light guide module of a known detector module, Figure 3 shows the light distribution using this light guide module simulated in TracePro software. Figure 2 Shows a detector module according to an embodiment of the present invention, which has a special pattern of a segmented pseudocolumn light guide.

[0038] Figure 4 , which shows the light distribution using the Figure 2 detector module in simulated in TracePro software, showing that the light sharing is limited to a 16×16 crystal array 120 coupled to the same pseudocolumn light guide array, thereby improving the light sharing rate between crystals for Prism-PET in the embodiments of the present invention.

[0039] Figure 5 Shows the arrangement of a Prism-PET module coupled in a 4-to-1 ratio according to an embodiment of the present invention. Figure 5 The lower left corner of shows a plan view of the relevant arrangement of 2×2 crystals of a scintillator array, which includes a plurality of scintillator crystals and SiPM pixels 140 of a plurality of detectors provided at the bottom end of the scintillator array. Figure 5 The upper right corner of shows three different pseudocolumn designs used in the Figure 5 embodiment. The central pseudocolumn 162, the edge pseudocolumn 168, and the corner pseudocolumn 166 have different designs for reducing edge and corner artifacts, thereby achieving uniform crystal identification performance. As Figure 6As shown, the three different quasi-cylinders are arranged in a predetermined manner on the top 122 of the scintillator array and are arranged to redirect particles between the tops of the crystals of the scintillator array.

[0040] Different Prism-PET detector modules were fabricated for comparison. The first Prism-PET detector 142 consists of a 16×16 array of lutetium yttrium silicate (LYSO) crystals of 1.4×1.4×20 mm 3 . The crystals are coupled to an 8×8 SiPM readout array 140 at a ratio of 4 to 1 on one side and to a uniform glass light guide on the opposite side (the radiation-receiving side), similar to the modules previously studied in the literature [8, 20]. The second Prism-PET detector 144 consists of the same crystal and readout geometry, replacing the conventional single uniform glass light guide with a quasi-cylinder light guide array having unique prism designs and layouts at the corners, edges, and center of the detector module to optimize the light sharing pattern ( Figures 5 - 8 ). The third Prism-PET detector has a quasi-cylinder light guide array and uses the same SiPM array as the other two detectors, but it uses an approximately 24×24 array of LYSO crystals of 0.96×0.96×20 mm 3 to achieve a 9 to 1 coupling ( Figure 7 and Figure 8 ). In the two Prism-PET detector modules, the scintillator crystals are coupled to the readout pixels and right-angled prisms at the same ratio.

[0041] The coupling scheme of the prisms is offset from the coupling scheme of the readout pixels such that each crystal is only coupled to other crystals belonging to different readout pixels ( Figure 5 ). When photons enter the quasi-cylinder after gamma-ray interactions in the crystals, due to the geometry of the right-angled triangular prism, the photons (i.e., particles 300) are effectively redirected to adjacent crystals, thereby increasing the light sharing rate between pixels ( Figure 2 ). The geometry of each quasi-cylinder depends on its position and is predefined to decouple adjacent crystals along the edges and corners to optimize the separation between crystals, otherwise these adjacent crystals would have similar readout patterns. In some embodiments of the present invention, the first group includes four crystals, the second group includes four crystals, and the first group and the second group share two adjacent crystals out of the four crystals. In some embodiments, only the shared crystals are arranged to direct particles to the first detector and the second detector.

[0042] As Figure 7 and Figure 8As shown, the first quasi - cylinder 162 among multiple quasi - cylinders is set as the particle between the tops 122 of a set of nine crystals of the redirecting scintillator array 120. In an embodiment of the present invention, the central crystal 139 in the set of nine crystals is set to direct the particle to four adjacent detectors 142, 144, 146, 148. The second quasi - cylinder 164 among the multiple quasi - cylinders is set as the particle between the tops of another set of nine crystals of the redirecting scintillator array. The first quasi - cylinder 162 and the second quasi - cylinder 164 are generally adjacent, and a set of nine crystals 132 and another set of nine crystals 134 are generally adjacent. In an embodiment of the present invention, the corner quasi - cylinders of the multiple quasi - cylinders are set to redirect the particles between the tops of a set of five crystals of the scintillator array. In an embodiment of the present invention, the edge quasi - cylinder 168 of the multiple quasi - cylinders is set to redirect the particles between the tops of a set of five crystals of the scintillator array.

[0043] Since the coupling scheme limits the light sharing between crystals to adjacent SiPMs that can enhance crystal identification, it can further enhance light sharing by matching the refractive index n between the scintillator columns, prisms, and coupling adhesives, thereby improving the DOI resolution and crystal identification. All prisms are made of SF10 glass with n = 1.767 (instead of BK7 with n = 1.53, which is the material for homogeneous glass light guides) and are coupled to the scintillator array using NOA170 adhesive with n = 1.7. Barium sulfate (BaSO4) is used as the reflector material between the crystals and the prisms because of its high spatial performance and no reduction in energy resolution or time resolution

[21] . The SiPM saturation effect, which is known to cause a positive shift in energy resolution and a negative impact on DOI resolution, has not been considered yet

[22] .

[0044] Figure 6 A perspective view is provided that provides details of the quasi - cylinder array according to an embodiment of the present invention.

[0045] Figure 6 A perspective view of the quasi - cylinder array, cross - sections of the quasi - cylinders and corresponding crystals, and individual views of the corner, edge, and central quasi - cylinders of the 4 - to - 1 - coupled Prism - PET module are provided. Figure 6 The bottom ends of the first set of crystals of the scintillator array shown are set to direct the particles to the first detector among the multiple detectors, and the bottom ends of the second set of crystals of the scintillator array are set to direct the particles to the second detector that is generally adjacent to the first detector.

[0046] Figure 7 The layout of a 9 - to - 1 - coupled Prism - PET module according to an embodiment of the present invention is shown. Figure 7The embedded diagram shows the predetermined readout pattern for each crystal of a single truncated pyramidal light guide belonging to a 9-to-1 coupling module.

[0047] Figure 8 A perspective view of the light guide array and truncated pyramid crystal array of a 9-to-1 coupled Prism-PET module according to an embodiment of the present invention is provided, as well as a cross-section of the truncated pyramid.

[0048] Advantages are demonstrated through experimental measurements in aspects such as crystal identification, energy resolution, and DOI resolution, including how Prism-PET achieves up to 9-to-1 crystal-readout coupling, which can be used to significantly improve spatial resolution without increasing the number of readout channels ( Figure 7 and 8 ).

[0049] Figure 9 Shows the readout of a 4-to-1 coupled detector using uniform glass (known configuration). Figure 10 Shows the detector readout of a 4-to-1 coupled Prism-PET module according to an embodiment of the present invention. Figure 11 Shows the detector readout of a 9-to-1 coupled Prism-PET module according to an embodiment of the present invention

[0050] The detector module consists of a LYSO crystal array (4-to-1 or 9-to-1) manufactured by X-Lum Co., Ltd. (Shanghai, China) coupled to an 8×8 SiPM array (Hamamatsu S13361-3050AE-08). Data acquisition is performed using application-specific integrated circuits (ASICs) of model TOFPET2 from PETsys Electronics and a readout board of model FEB / D v2. A 3 MBq Na-22 sodium point source (effective diameter of 5 mm) is used to uniformly expose the 4-to-1 coupled and 9-to-1 coupled detector modules with truncated pyramid light guide arrays to obtain their flood data. 10,000,000 events of the 4-to-1 module and 22,500,000 events of the 9-to-1 module (so that each crystal obtains the same number of events) are used for flood histograms.

[0051] Figure 12 Provides Figure 9 (Known configuration) of 4-to-1 coupled uniform glass, Figure 10 in the 4-to-1 coupled Prism-PET module and Figure 11 in the 9-to-1 coupled Prism-PET module of Gaussian histograms and filtered energy spectra.

[0052] Figure 12 The upper half is a one-dimensional Gaussian histogram, which shows Figures 9 to 11The module in [it] reads out pixels at the corners, edges, and centers in the x-direction, with crystal separation on the readout pixels. Figure 12 The lower half is from Figures 9 to 11 The filtered energy spectra of the central crystal in [it], with (13%, 9%, and 10%) DOI correction and without (20%, 14%, 16%) DOI correction.

[0053] Using a method similar to that described in reference

[18] , the DOI performance was experimentally measured on a per-crystal basis. Using a lead collimator, the detector module was exposed to a Na-22 light source at five different crystal depths (2, 6, 10, 14, and 18 mm). The light source was placed in a lead cylinder with a 1-mm pinhole. One side of the pinhole was aligned with the DOI-calibrated module, and the other side was aligned with a single 1.4×1.4×20 mm 3 crystal on the reference module. Coincidence events between these two modules were used to reject scattered events and only accept events along the expected response line. The DOI estimation parameter

[18] , w, was calculated and plotted as a histogram for all crystals. Then, a linear regression was used to transform the w histogram into DOI space to determine the slope between w and the ground-truth DOI, which should be the center of each Gaussian peak. The width of the Gaussian peak transformed into DOI space was used to calculate the DOI resolution of the crystal ( Figure 14A - 14D ). The DOI resolution is depth-dependent and equal to the FWHM of the Gaussian histogram. The overall DOI resolution specific to a crystal was calculated as the average of the DOI resolutions over the entire measurement depth

[18] . The typical central crystal on each module was used to calculate the DOI resolution of each module.

[0054] Compared with a standard uniform glass light guide module using the flood histogram of the fabricated module ( Figures 9 - 11 ), the spatial performance of the Prism-PET module of the present invention has characteristics. The glass light guide module is affected by edge and corner effects, resulting in poor crystal separation related to position. Prism-PET is able to achieve excellent crystal separation throughout the detector array without edge and corner artifacts, which is unprecedented in a 4-to-1 coupled detector module with single-ended TOF-DOI readout [8, 19, 20]. The 9-to-1 coupled Prism-PET module showed similar results ( Figure 11 ), demonstrating in a detector with 3.2×3.2 mm 2Uniform sub-millimeter crystal separation in the TOF-DOIPET detector module of SiPM pixels. The one-dimensional event localization histograms (along the x-direction) plotted confirm that the Prism-PET of the present invention has uniform crystal separation performance at the center, edges, and corners. Prism-PET also achieved energy resolutions of 14% and 16% respectively through DOI correction in 4:1 and 9:1 coupled modules, while the uniform glass light guide only achieved an energy resolution of 20% (see Figure 12 the figure at the bottom).

[0055] Figure 13 provides the DOI for the Figure 10 and Figure 11 modules in accordance with an embodiment of the present invention. Figures 13A - 13C The histograms in Figure 13A provide the DOI resolution calculated at the interaction depth, Figure 13B provides the DOI resolution in the central crystal of a 4:1 coupled detector module with uniform glass (known construction), Figure 13C provides the DOI resolution of the pseudocylindrical light guide. Figure 13C provides a comparison of the DOI resolution based on the light guides used, showing that the Prism-PET detector module of the present invention achieves a two-fold improvement in DOI resolution compared to the uniform glass light guide, a result obtained through experimental measurements on a single central crystal in each module. The measured DOI resolution of the glass light guide is 5 mm FWHM, which is in good agreement with previously reported results

[19] . The Prism-PET module achieved a 2.5 mm FWHM DOI localization, which is the best resolution ever reported using single-ended readout. The increased dependence of the w parameter on depth can be attributed to 1) a controlled and deterministic light sharing pattern within the pseudocylinder; 2) increased light transmission from the scintillator to the light guide due to refractive index matching; 3) enhanced deviation of the upward-propagating photon path by 180° due to the geometry of the right-angled triangular prism, all of which enhance light sharing between crystals coupled to the same pseudocylinder. DOI information can be used to simultaneously improve the time resolution and energy resolution, the former by deconvolving the depth-specific photon transport within the scintillator, and the latter by constructing depth-specific light peaks [8, 18]. Embodiments of the present invention achieved energy resolutions of 9% and 10% respectively in 4:1 and 9:1 coupled Prism-PET modules, and 13% energy resolution under uniform light guide after applying DOI-based correction. Note that after performing SiPM saturation correction, the DOI resolution and energy resolution values improve slightly and deteriorate respectively

[22] ; thus, the reported values more indicate the relative performance of the Prism-PET of the present invention compared to the uniform light guide module rather than the absolute performance in practice.

[0056] Figure 14A - 14D There is provided a DOI resolution curve graph according to an embodiment of the present invention, having a conversion from a DOI-specific w histogram to a DOI histogram showing the DOI resolution of a single crystal at each depth. Figure 14A There are provided histograms of the DOI estimation parameter w obtained at 2, 6, 10, 14, and 18 mm. Figure 14B A fit between w and DOI is provided by linear regression. Figure 14C There is provided by multiplying Figure 14A the w histogram in Figure 14B by the slope of the linear fit in Figure 14D to generate a DOI histogram. Figure 14C There is provided the DOI resolution at each acquired depth based on the width of the Gaussian distribution curve in

[0057] The most important parameter that may be considered when establishing a PET system is the gamma-ray detection sensitivity, which is directly related to the signal-to-noise ratio (SNR), thus determining the patient throughput, the delivered dose, and the image quality. Monte Carlo simulations performed using highly advanced software, such as GATE, are the most reliable method for modeling and calculating the system-level sensitivity. However, the relative improvement in sensitivity and the comparison between systems can be done analytically by calculating (a) the geometric sensitivity and the coincidence time resolution (CTR) based on time-of-flight readout (TOF) to calculate (b) the sensitivity gain, which is equivalent to the square of the SNR gain in formula (1)

[24] :

[0058]

[0059] where D is the diameter of the object to be imaged, and Δx is the length of the reconstructed segment along the line of response, which is proportional to the CTR (Δt) in formula (2):

[0060]

[0061] An example of a dedicated brain PET scanner that can be built with Prism-PET detector modules is a cylindrical ring with an axial length of 50 cm and a diameter of 25 cm.

[0062] Figure 15A - 15D There is provided a sensitivity map according to an embodiment of the present invention. Figure 15A There are provided the dimensions and geometric coverage of examples of the Siemens Biograph Vision scanner, the Explorer whole-body PET scanner, and the Prism-PET brain scanner. Figure 15B There is provided the geometric sensitivity of a point source located at the center of each scanner shown in Figure 15A Figure 15CA relative sensitivity gain is provided as a function of the temporal resolution. Figure 15D Provides an effective sensitivity gain, which is calculated as the geometric efficiency (e.g. Figure 15B ) and TOF sensitivity gain (as shown Figure 15C The product between .

[0063] Figure 15A The dimensions of a Prism-PET brain scanner according to an embodiment of the present invention are shown compared to the dimensions of a whole body scanner (Siemens Biograph Visions) and a whole body (Explorer) PET scanner example. A smaller ring diameter and a larger axial field of view greatly improve the geometric efficiency ( Figure 15B ), but this comes at the expense of greatly increased parallax errors and partial volume effects, which can be mitigated by performing a depth of interaction (DOI) readout

[26] . Therefore, scanners dedicated to small-diameter organs can only be built using detector modules with DOI localization capabilities, such as our Prism-PET module.

[0064] The DOI readings can also be used to recover the CTR of the TOF readings by deconvolving the DOI dependence of the fit (i.e., the difference in the optical path length of the photons) [8]. Assuming that the CTR reported here is the same as previously reported (about 150 ps), which is a safe lower bound estimate because our module has a better DOI resolution (2.5 mm compared to 3 mm), based on equation (1), Prism-PET can achieve a nearly 10-fold gain in TOF sensitivity when imaging an object with a diameter of about 20 cm, such as the human brain ( Figure 15C ). The TOF sensitivity gain for human brain imaging is slightly lower for Siemens Biograph Vision, which can achieve a CTR of about 220 ps

[25] , while for Explorer( Figure 15C ) is much lower, with CTR>400ps

[23] .

[0065] Figure 15D The overall effective sensitivity gain in human brain imaging when both geometric efficiency and TOF sensitivity gain are considered is shown. Based on the above calculations, the Prism-PET scanner in the embodiment of the present invention increases sensitivity by three times and four times compared to the Siemens BiographVision scanner and Explorer scanner, respectively.

[0066] Figure 16A - 16F Compton interaction according to an embodiment of the present invention is shown.

[0067] Regarding Compton interactions, due to its deterministic light sharing pattern, the Prism-PET of the present invention enables Compton scatter energy decomposition (and thus localization). Suppose we have a 16×16 lutetium LYSO crystal array, and its Prism-PET light guides are coupled to an 8×8 silicon photomultiplier (SiPM) pixel array in a 4-to-1 manner. Based on the approximation that each 511 keV gamma ray will generate signals on 4 different pixels due to light sharing, we can directly measure the light sharing rate between all the crystals belonging to the same pseudo-column using the photoelectric events from the flood data. Using this information, the energy of the primary interaction position (i.e., the recoil electron) and the energy of the secondary interaction position (i.e., the scattered gamma ray) are decomposed. Once the decomposed energies are obtained, two independent absorption events in the scintillator block can be located, and the scattering angle and DOI can be determined. For the Prism-PET module of the present invention, due to the change from the random light sharing pattern of photoelectric events to the deterministic pattern, side-by-side Compton scatter events can be identified (Figs. 16 and 17).

[0068] Figure 16A - 16C An example of Compton energy decomposition in a multi-crystal scintillator array provided with the Prism-PET of the present invention is given. Figure 16D An example is provided of Figure 16A - 16C the ratio of the light sharing fraction between pixel 1 and adjacent pixels as identified in. In one case, two pixels (2 and 4) are both adjacent to pixel 1, resulting in equal light sharing fractions, while in another case, pixel 3 is diagonal to pixel 1, resulting in a smaller light sharing fraction. (E) and (F) are the energy and DOI errors of the Compton interaction decomposition of the Prism-PET.

[0069] Classical Compton energy decomposition can be carried out in the following manner. The total energies EA and EB absorbed by the constituent elements A and B (scattered electron and recoil electron) are taken as the sum of the energies of all 4 SiPMs in Equation (3):

[0070]

[0071] where E A1 and E B1 are the maximum deposited energies in the SiPMs coupled to the crystal pixels of the interaction, while E A2,3,4 and E B2,3,4 are the deposited energies in the adjacent columns due to light leakage at the bottom (from the SiPM side) and light leakage through the prism-mirror light guide at the top. The experimental results in Figure 16A show four known parameters E 1-4 corresponding to the energies detected by each of the four pixels after side-by-side Compton scatter events, where the total gamma particle energy deposited is provided by Equation (4):

[0072] E γ = E A + E B (4)

[0073] Note the energies of the constituents in the Compton scattering event, i.e., E A1-4 and E B1-4 , which are unknown. Writing equations based on the measured energies, we obtain equation (5):

[0074] E1 = EA1 + EB4

[0075] E2 = EA2 + EB1

[0076] E3 = EA3 + EB2

[0077] E4 = EA4 + EB3 (5)

[0078] Four equations and eight unknowns are provided. However, the deposited energies in adjacent columns are correlated. Considering that Figure 16D the maximum deposited energy in the embedded figure in

[0079]

[0080] appears in the SiPM in the upper left corner. Assuming that the light sharing fraction with three neighbors depends on their proximity to the interaction crystal and using the Pythagorean theorem with the centers of the three neighbors as the vertices to form a right triangle, we obtain equation (6): 12 where, for example, d

[0081] E1 = EA1 + EB2

[0082] E2 = EA2 + EB1

[0083] E3 = 0.7EA2 + EB2

[0084] E4 = E A2 + 0.7E B2 (7)

[0085] Now we have four equations and four unknowns. It should be noted that in actual operation, due to some small and inevitable misalignments between the prism-mirror light guide and the scintillator columns, the light sharing fraction will have a spatial difference from the ideal situation shown in equation 6. However, as Figure 16D shown, by analyzing the light sharing fraction of individual photoelectric events obtained from the flood histogram experiment, the light sharing fraction across the entire array can be obtained empirically.Figure 16B and 16C depicts two decomposed elements of the measured side-by-side Compton scattering events based on the above analysis.

[0086] Given that our module has a DOI positioning function, we can express the DOI variable as formula (8):

[0087] wA = EA1 / EA

[0088] w B = E B1 / E B (8)

[0089] As Figure 16E and 16F shown, based on 200,000 experimental gamma events, our estimation errors for {E A1 , E B1} and {w A , w B} are approximately 10%. Using a convolutional neural network as an estimator can further reduce the error, especially since we can collect millions of gamma events as a training dataset using experimental data of flood histograms.

[0090] Figure 16A - 16D shows an example of how to decompose a Compton event into its constituent elements, where adjacent scintillators in two different SiPMs in the Compton event fully absorb the recoil electron and the scattered γ-ray. Calculating the DOI variable w using the classical Compton decomposition method results in a 11% full width at half maximum (FWHM) error ( Figure 16E ). Additionally, the FWHM energy error caused by Compton decomposition is 15% ( Figure 16F ).

[0091] Figure 17 shows the photoelectric interaction and Compton interaction according to an embodiment of the present invention, where above the diagram of the deterministic light sharing pattern of Prism-PET in the embodiment of the present invention, a diagram of the random light sharing pattern of the glass light guide is provided. Figure 17 shows the experimental results of several examples of Compton events absorbed in adjacent crystals in the Prism-PET module of the present invention and in the module with a flat glass light guide.

[0092] The light sharing pattern in the glass light guide module is random, so it is difficult (impossible in most cases) to decompose the detected energy into the constituent energies of the scattered photon and the recoil electron. Due to the geometric structure of the right-angled triangular prism, in the Prism-PET module, the light sharing pattern is deterministic, thus making it practical to decompose an event into its constituent energies according to the known inter-crystal light sharing ratio.

[0093] Accordingly, a particle detector is provided that includes: a scintillator array including a plurality of scintillator crystals; a plurality of detectors disposed at a bottom end of the scintillator array; a plurality of pseudocolumns disposed at a top end of the scintillator array; and at least one processor operably communicable with the plurality of detectors. The at least one processor includes a plurality of supervised machine learning algorithms, including a convolutional network and a regression network, and is configured to perform 3D gamma-ray localization of at least one interaction position within at least one of the plurality of scintillator crystals. The at least one processor is configured to recover at least one Compton event scattered between the plurality of scintillator crystals and to localize the at least one Compton event at the scintillator level based on the 3D gamma-ray localization. The at least one processor is further configured to determine a scattering angle based on the at least one Compton event and DOI information. The at least one processor is further configured to localize at least one Compton event based on decomposed energies of at least two interactions absorbed in the plurality of scintillator crystals, the decomposed energies being based on at least one light sharing pattern, the at least one light sharing pattern being based on positions of the plurality of scintillator crystals relative to the plurality of detectors and the plurality of pseudocolumns.

[0094] According to an embodiment of the present invention, the at least one light sharing pattern is mapped based on a light sharing rate between scintillator crystals of the same pseudocolumn. The light sharing rate is based on a predetermined geometry of at least one of the plurality of pseudocolumns; the mapping is based on measured photoelectric events, decomposed energies of at least one primary interaction and at least one secondary interaction, the at least one primary interaction being based on electron recoil, the at least one secondary interaction being based on gamma-ray scattering, in which the light sharing pattern is deterministic.

[0095] Accordingly, an economical and practical method for achieving high spatial resolution and high DOI resolution in a multi-crystal single-ended readout detector module is provided without introducing artifacts at edges and corners. Embodiments of the present invention can be used to implement depth encoding in clinical body PET scanners and whole-body PET scanners

[23] without increasing cost (the pseudocolumn light guide array accounts for less than 10% of the total cost of each Prism-PET module) and power consumption, while it can improve spatial resolution (by using 2×2×20 mm 3 crystals and 6×6 mm 2For the readout pixels, 9-to-1 coupling is performed, the sensitivity is increased (through the recovery of Compton scattering between crystals), and the timing resolution is improved (through DOI correction by timing jitter). For brain imaging with a smaller ring diameter, the 9-to-1 coupling ratio can achieve sub-millimeter spatial resolution, while expanding the axial field of view to nearly twice that of a body PET scanner can achieve the same geometric sensitivity gain as the Explorer whole-body PET scanner (Figure 15) [8, 23 - 26]. In addition, with a DOI resolution of 2.5 mm, the parallax error can be greatly reduced, and it is possible to achieve a coincidence time resolution of approximately 100 ps through DOI correction [8], which will achieve higher sensitivity and spatial resolution [24 - 26]. These advantages result in a practical, economical, and powerful method to achieve high spatial resolution and high sensitivity at relatively low doses for functional quantification and molecular imaging of many organs in the human body, including some important structures in the brain that existing PET scanners cannot handle, such as the raphe nuclei, cholinergic basal forebrain nuclei, locus coeruleus, and hypothalamic nuclei, all of which are considered to play crucial roles in the basic physiology and pathophysiology of common neurodegenerative and mental diseases [26 - 30]. The ability to visualize and quantify these and similar targets has the potential to revolutionize molecular imaging in the clinical and research fields, providing unprecedented tools for the early diagnosis and basic research of tumors and brain diseases.

[0096] Another advantage of the embodiments of the present invention is the ability to more accurately identify the initial interaction position of Compton scattering events, thereby further improving the spatial resolution and sensitivity (Figures 16 - 17). Traditionally, Compton detection is performed using multiple detector layers, but recent papers have outlined the criteria for localizing and decomposing Compton interactions using single-ended readout detectors and have pointed out that high-resolution DOI readings are a key feature for the recovery of Compton scattering

[31] . A uniform light guide is not optimal for this task because the SiPM pattern of a single event is random, while our Prism-PET module creates a deterministic light sharing pattern independent of the interaction position inside the main scintillator column ( Figures 1 - 4 and Figure 14). It is worth noting that Prism-PET can decompose side-by-side scattered photon time and recoil electron events into their constituent energies, spatial positions, and DOIs, which are the most likely and also the most difficult scattered events to analyze. The recovery of Compton scattering is particularly crucial for maintaining high sensitivity in detector modules with small scintillator crystals because as the crystal size decreases, scattered photons are more likely to be absorbed in crystals at positions different from the main interaction position

[32] .

[0097] Embodiments of the present invention provide a Prism-PET detector module that uses an effective 180° light-bending reflector for enhanced light sharing, achieving a true single-ended analogy for dual-ended depth-encoded readout. A DOI resolution of 2.5 mm FWHM is achieved, along with a coupling of up to 9:1 of scintillator to SiPM for high spatial resolution, while directly coupling the crystal array to SiPM pixels to minimize light leakage and maintain high photon detection efficiency, which is necessary for good time resolution. The top reflector consists of an optimized pattern of segmented quasi-columnar light guides for effectively redirecting the path of scintillation photons from the main crystal to a selected nearest SiPM, thus closely mimicking the operation of a dual-ended readout detector. This creates an anisotropic and deterministic pattern of signals that can be used to resolve side-by-side Compton scattering events into their constituent energy and DOI information for scatter recovery. Thus, high and uniform spatial resolution (9:1 coupling of crystals of about 1 mm; enhanced light sharing eliminates edge and corner artifacts; reduced spatial blurring caused by Compton-scattered photons through scatter recovery) is achieved in a compact system (DOI encoding eliminates parallax errors and allows a smaller ring diameter), high sensitivity (20 mm thick detector and Compton scatter recovery between crystals) is achieved, and good energy resolution and time resolution (especially after applying DOI correction) are achieved. Through this unique combination of features, an economical and compact TOF-DOI-Compton PET scanner can be developed based on Prism-PET modules for organ-specific functional and molecular imaging of small animals and humans.

[0098] Although the present invention has been shown and described with reference to certain aspects thereof, those skilled in the art will understand that various changes in form and detail may be made therein without departing from the spirit and scope of the invention as defined by the appended claims and their equivalents. In the absence of expressly using "means for..." or "step for...", the recitations in any of the following claims should not be construed as means-plus-function elements.

[0099] References

[0100] [1] Mankoff DA, Farwell MD, Clark AS, Pryma DA. Making molecular imaging a clinical tool for precision oncology: a review. JAMA Oncol. 2017;3:695–701.

[0101] [2]Slifstein M,Abi-Dargham A.Recent developments in molecular brainimaging of neuropsychiatric disorders.Semin Nucl Med.2017;47:54–63.

[0102] [3]Van Sluis J,De Jong J,Schaar J,et al.Performance characteristicsof the digital biograph vision PET / CT system.J Nucl Med.2019;60:1031–1036.

[0103] [4]Moses WW.Fundamental limits of spatial resolution in PET.NuclInstr and Meth A.2011;648:S236–S240.

[0104] [5]Miyaoka RS,Lewellen TK,Yu H,McDaniel DL.Design of a depth ofinteraction(DOI)PET detector module.IEEE Trans Nucl Sci.1998;45:1069-1073.

[0105] [6]Stickel JR,Cherry SR.High-resolution PET detector design:modellingcomponents of intrinsic spatial resolution.Phys Med Biol.2005;50:179–195.

[0106] [7]Seifert S,Schaart DR.Improving the time resolution of TOF-PETdetectors by double-sided readout.IEEE Trans Nucl Sci.2015;62:3–11.

[0107] [8]Pizzichemi M, Polesel A, Stringhini G, et al. On light sharing TOF-PET modules with 3mm depth of interaction and 157ps FWHM coincidence time resolution. Phys Med Biol. 2019;64:155008.

[0108] [9]Yang Y, Wu Y, Qi J, et al. A prototype PET scanner with DOI-encoding detectors. J Nucl Med. 2008;49:1132–40.

[0109]

[10] Kuang Z, Wang X, Fu X, et al. Dual-ended readout small animal PET detector by using 0.5mm pixelated LYSO crystal arrays and SiPMs. Nucl Instrand Meth A. 2019;917:1–8.

[0110]

[11] Abreu MC, Aguiar JD, Almeida FG, et al. Design and evaluation of the clear-PEM scanner for positron emission mammography. IEEE Trans Nucl Sci. 2006;53:71–77.

[0111]

[12] Du J, Yang Y, Bai X, et al. Characterization of large-area SiPM array for PET applications. IEEE Trans Nucl Sci. 2016;63:8–16.

[0112]

[13] Schmand M, Eriksson L, Casey ME, et al. Performance results of a new DOI detector block for a high resolution PET-LSO research tomograph HRRT. IEEE Trans Nucl Sci. 1998;45:3000–3006.

[0113]

[14] Seidel J, Vaquero JJ, Siegel S, et al. Depth identification accuracy of a three layer phoswich PET detector module. IEEE Trans Nucl Sci. 1999;46:485–490.

[0114]

[15] Gonzalez-Montoro A, Aguilar A, Canizares G, et al. Performance study of a large monolithic LYSO PET detector with accurate photon DOI using retroreflector layers. IEEE Trans Rad Plasma Med Sc. 2017;1:229–237.

[0115]

[16] Ito M, Lee MS, Lee JS. Continuous depth-of-interaction measurement in a single-layer pixelated crystal array using a single-ended readout. Phys Med Biol. 2013;58:1269–1282.

[0116]

[17] Kuang Z, Yang Q, Wang X, et al. A depth-encoding PET detector that uses light sharing and single-ended readout with silicon photomultipliers. Phys Med Biol. 2018;63:045009.

[0117]

[18] Pizzichemi M, Stringhini G, Niknejad T, et al. A new method for depth of interaction determination in PET detectors. Phys Med Biol. 2016;61:4679–4698.

[0118]

[19] Niknejad T, Pizzichemi M, Stringhini G, et al. Development of high-resolution detector module with depth of interaction identification for positron emission tomography. Nucl Instr and Meth A. 2017;845:684–688.

[0119]

[20] Stringhini G, Pizzichemi M, Ghezzi A, et al. Development of a high-resolution module for PET scanners. J Instrum. 2017;12:C02073–C02073.

[0120]

[21] Kuang Z, Wang X, Li C, et al. Performance of a high-resolution depth encoding PET detector using barium sulfate reflector. Phys Med Biol. 2017;62:5945–5958.

[0121]

[22] Otte AN, Barral J, Dolgoshein B, et al. A test of silicon photomultipliers as readout for PET. Nucl Instr and Meth A. 2005;545:705-715.

[0122]

[23] Badawi RD, Shi H, Hu P, et al. First human imaging studies with the explorer total-body PET scanner. J Nucl Med. 2019;60:299–303.

[0123]

[24] Surti S. Update on time-of-flight PET imaging. J Nucl Med. 2015;56:98-105.

[0124]

[25] Reddin JS, Scheuermann JS, Bharkhada D, et al. Performance evaluation of the SiPM-based Siemens Biograph Vision PET / CT system. In: Conference Record of the 2018 IEEE Nuclear Science Symposium and Medical Imaging Conference. Sydney, AU: IEEE; 2018.

[0125]

[26] Gong K, Majewski S, Kinahan PE, et al. Designing a compact high-performance brain PET scanner—simulation study. Phys Med Biol. 2016;61:3681-3697.

[0126]

[27] Hornung JP. The human raphe nuclei and the serotonergic system. J Chem. Neuroanat. 2003;26:331–343.

[0127]

[28] Mufson EJ, Ginsberg SD, Ikonomovic MD, DeKosky ST. Human cholinergic basal forebrain: Chemoanatomy and neurologic dysfunction. J Chem Neuroanat. 2003;26:233–242.

[0128]

[29] Betts MJ, Kirilina E, Otaduy MCG, et al. Locus coeruleus imaging as a biomarker for noradrenergic dysfunction in neurodegenerative diseases. Brain. 2019;142:2558–2571.

[0129]

[30] Barbosa DA, de Oliveira-Souza R, Santo FM, et al. The hypothalamus at the crossroads of psychopathology and neurosurgery. Neurosurg Focus. 2017;43:1–11.

[0130]

[31] Makek M, Bosnar D, L. Scintillator pixel detectors for measurement of Compton scattering. Condens Matter. 2019;4:24.

[0131]

[32] Hsu DF, Freese DL, Innes DR, Levin CS. Intercrystal scatter studies for a 1 mm 3 resolution clinical PET system prototype. Phys Med Biol. 2019;64:095024.

Claims

1. A particle detection device, comprising: A scintillator array, the scintillator array comprising a plurality of scintillator crystals; A plurality of detectors disposed at a first end of the scintillator array, wherein each detector corresponds to a plurality of scintillator crystals; And A plurality of pseudo-cylinders disposed at a second end of the scintillator array, wherein the pseudo-cylinders have different shapes according to the position of the pseudo-cylinders relative to the plurality of detectors, and each pseudo-cylinder is in contact with scintillator crystals that contact at least two different detectors, wherein the shape determines the number of at least two different detectors, the at least two different detectors being adjacent detectors, and wherein each pseudo-cylinder is arranged to redirect particles between the scintillator crystals in contact with the corresponding pseudo-cylinder; The shape is selected from the group comprising a central pseudo-cylinder, an edge pseudo-cylinder, and a corner pseudo-cylinder.

2. The particle detection device according to claim 1, wherein The shape is a central pseudo-cylinder, and the central pseudo-cylinder is arranged to be in contact with scintillator crystals that contact four adjacent detectors.

3. The particle detection device according to claim 1, wherein The shape is an edge pseudo-cylinder, and the edge pseudo-cylinder is arranged to be in contact with scintillator crystals that contact two adjacent detectors.

4. The particle detection device according to claim 1, characterized in that, The shape is a corner pseudo-cylinder, and the corner pseudo-cylinder is arranged to be in contact with scintillator crystals that contact three detectors.

5. The particle detection device according to claim 1, wherein The plurality of scintillator crystals are four scintillator crystals.

6. The particle detection device according to claim 1, wherein The plurality of scintillator crystals are nine scintillator crystals.

7. The particle detection device according to claim 1, wherein, The shape is a three-dimensional (3D) shape, and the 3D shape varies according to position.

8. The particle detection device according to claim 7, characterized in that, There are three different 3D shapes, and the pseudo-cylinders with different shapes are positioned to achieve substantially uniform scintillator crystal identification performance.

9. A particle detection device, comprising: A scintillator array, the scintillator array comprising a plurality of scintillator crystals; A plurality of detectors disposed at a first end of the scintillator array, wherein each detector corresponds to a plurality of scintillator crystals; And A plurality of pseudo-cylinders disposed at a second end of the scintillator array, wherein the pseudo-cylinders have different shapes according to the position of the pseudo-cylinders relative to the plurality of detectors, and each pseudo-cylinder is in contact with scintillator crystals that contact at least two different detectors, wherein the shape determines the number of at least two different detectors, the at least two different detectors being adjacent detectors, and wherein each pseudo-cylinder is arranged to redirect particles between the scintillator crystals in contact with the corresponding pseudo-cylinder; And At least one processor operably communicating with the plurality of detectors, Characterized in that the shape is selected from the group comprising a central pseudo-cylinder, an edge pseudo-cylinder, and a corner pseudo-cylinder; The at least one processor comprises a plurality of supervised machine learning algorithms, the algorithms being arranged to perform three-dimensional (3D) gamma-ray localization of at least one interaction position within at least one of the plurality of scintillator crystals.

10. The particle detection device according to claim 9, characterized in that, The at least one processor is further arranged to recover at least one Compton event scattered between the plurality of scintillator crystals and to localize the at least one Compton event at the scintillator level based on the 3D gamma-ray localization.

11. The particle detection device according to claim 9, characterized in that, The at least one processor is further arranged to determine a scattering angle based on at least one Compton event and depth of interaction (DOI) information.

12. The particle detection device according to claim 9, characterized in that, The at least one processor is further configured to locate at least one Compton event based on the resolved energies of at least two interactions absorbed in the plurality of scintillator crystals.

13. The particle detection device according to claim 12, wherein, The resolved energies are based on at least one light sharing pattern.

14. The particle detection device according to claim 13, wherein The at least one light sharing pattern is based on the positions of the plurality of scintillator crystals relative to the plurality of detectors and the plurality of pseudocolumns.

15. The particle detection device according to claim 13, characterized in that, The at least one light sharing pattern is plotted based on the light sharing rate between the scintillator crystals of the same pseudocolumn.

16. The particle detection device according to claim 15, wherein The light sharing rate is based on the predetermined geometry of at least one of the plurality of pseudocolumns.

17. The particle detection device according to claim 15, characterized in that, The plotting is based on the measured photoelectric events, the resolved energies of at least one primary interaction and at least one secondary interaction, and wherein the at least one primary interaction is based on electron recoil and the at least one secondary interaction is based on gamma ray scattering.

Images