Placement error identification method, gamma ray detector system, and placement error identification program

Abstract

To identify an error in the arrangement of detector elements included in a gamma-ray detector ring.SOLUTION: An arrangement error identification method according to an embodiment includes the steps of: acquiring detector element event data associated with radiation events in a plurality of detector elements included in a gamma-ray detector ring; identifying assembly events between the detector elements based on the detector element event data; calculating a metric based on each of the assembly events between the detector elements; and identifying an error in the arrangement of the detector elements based on the metric.SELECTED DRAWING: Figure 8

Description

[Technical Field]

[0001] The embodiments disclosed in this specification and the drawings relate to a method for identifying a placement error, a gamma ray detector system, and a placement error identification program. [Background technology]

[0002] Gamma ray detector systems, such as Positron Emission Tomography (PET) scanners, typically include multiple detector elements, such as crystals, modules, and detector units. These detector elements can be positioned incorrectly during manufacturing or system service. For example, detector elements can be inserted with an incorrect orientation or cable connections can be swapped. To avoid imaging artifacts due to inaccurate layout, it is important to verify the actual layout of the detector elements in the system.

[0003] Often, the layout is verified by visually inspecting the detector element cabling and orientation, but such inspections can be unreliable. While image verification is also possible, this method does not provide immediate error feedback without lengthy calibration procedures or reconstructions. Often, if a problem is not detected by the time the system is fully calibrated, some or all of the lengthy calibration process must be repeated, which is time-consuming and expensive.

[0004] Therefore, in view of the above-mentioned problems, a more efficient approach for identifying placement errors between detector elements in a gamma ray detector system would be beneficial. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] US Patent Application Publication No. 2016 / 0299240 [Patent Document 2] Chinese Patent Application Publication No. 107456235 [Patent Document 3] US Patent Application Publication No. 2019 / 0070437 Summary of the Invention [Problem to be solved by the invention]

[0006] One of the problems to be solved by the embodiments disclosed in this specification and the drawings is to identify placement errors of detector elements included in a gamma ray detector ring. However, the problems to be solved by the embodiments disclosed in this specification and the drawings are not limited to the above problem. Problems corresponding to the effects of each configuration shown in the embodiments described below can also be positioned as other problems. [Means for solving the problem]

[0007] An embodiment of a method for identifying placement errors includes obtaining detector element event data associated with radiation events in a plurality of detector elements included in a gamma ray detector ring, identifying assembly events between the detector elements based on the detector element event data, calculating a metric based on each of the assembly events between the detector elements, and identifying placement errors of the detector elements based on the metric. [Brief explanation of the drawings]

[0008] The present application will be better understood by referring to the description given in a non-limiting manner in conjunction with the accompanying drawings, in which: [Figure 1] FIG. 1 is a perspective view showing an example of a PET scanner. [Figure 2] FIG. 2 is a schematic diagram of a PET scanner and associated hardware according to an exemplary embodiment of the present disclosure. [Figure 3] FIG. 3 shows the crystal, module, detector unit and detector ring in relation to each other. [Figure 4] FIG. 4 is a diagram showing an example of a coincidence event caused by inter-detector scattering. [Figure 5] FIG. 5 shows the series of gamma rays that accompany the beta decay of Lu-176. [Figure 6] FIG. 6 shows an example of a coincidence event caused by background radiation in a crystal. [Figure 7A] FIG. 7A shows an example of an expected layout between detector elements. [Figure 7B] FIG. 7B shows an example of a measured layout between detector elements. [Figure 8] FIG. 8 is a flow chart illustrating a step-by-step example embodiment for detecting placement errors. [Figure 9] FIG. 9 shows an example of an assembly event generated by using fixed sub-regions within a detector ring. [Figure 10] FIG. 10 shows an example of an assembly event generated by centering a geometric subregion on a collision with maximum energy. [Figure 11A] FIG. 11A shows an example of an unfolded two-dimensional detector ring in a crystal (or pixel) view, a module view, and a detector unit view. [Figure 11B] FIG. 11B is a diagram illustrating various alignment errors that can occur in a detector system. [Figure 11C] FIG. 11C shows an example of an average distance map at the crystal / pixel level. [Figure 11D] FIG. 11D shows an example of an average distance map at the module level. [Figure 11E] FIG. 11E shows an example of an average distance map at the detector unit level. [Figure 12] FIG. 12 is a diagram showing a first example of random assembly. [Figure 13] FIG. 13 is a diagram showing a second example of random assembly. [Figure 14A] FIG. 14A shows the average distance map when the assembly region is a fixed quadrant. [Figure 14B] FIG. 14B shows the average distance map when the assembly region has crystals of size rmax=62. [Figure 14C] FIG. 14C shows the average distance map when the assembly region has crystals of size rmax=16. [Figure 15] FIG. 15 shows the average distance and standard deviation values ​​for each row constituting a defect-free detector ring for different assembly area sizes. [Figure 16A] FIG. 16A shows the placement error when two modules are swapped. [Figure 16B] FIG. 16B shows an average distance map obtained using a coincidence time window of one time constant (Δt=t1). [Figure 16C] FIG. 16C shows an average distance map obtained using a coincidence time window of three time constants (Δt=3t1). [Figure 16D] FIG. 16D shows an average distance map obtained using a coincidence time window of 8 time constants (Δt=8t1). [Figure 16E] FIG. 16E shows an average distance map obtained using a coincidence time window of 23 time constants (Δt=23t1). [Figure 17A] FIG. 17A shows the average distance map of extracted true assembly events. [Figure 17B] FIG. 17B shows the average distance map of random assembly events. [Figure 18] FIG. 18 is a diagram showing an example of C(r) in an infinitely large detector. [Figure 19] FIG. 19 shows an example of C(r) for a detector that is not infinitely large. [Figure 20]FIG. 20 shows an example of a mean distance map of random assembly events pre-calculated for a fixed region, where the subregions are quadrants. [Figure 21] FIG. 21 shows the average distance map for a defect-free system with no correction applied, and the average distance map after Method A has been applied. [Figure 22] FIG. 22 shows the average distance map for a system with swapped modules with no correction applied, and after Method A has been applied. [Figure 23] FIG. 23 shows the average distance map for a defect-free system with no correction applied, and the average distance map after Method B has been applied. [Figure 24] FIG. 24 shows the average distance map for a system with swapped modules with no correction applied, and the average distance map after Method B has been applied. [Figure 25] FIG. 25 shows an example of a mean distance map of random assembly events pre-computed using a geometric cut of a centered region. [Figure 26] FIG. 26 is a flow chart illustrating the steps of another exemplary embodiment for detecting placement errors. [Figure 27] FIG. 27 shows an example of ten precisely positioned detector elements. [Figure 28A] FIG. 28A shows the (accurate) detector system layout, the corresponding relative correlation map, an example of a method for converting relative correlation to effective distance, and the resulting effective distance map. [Figure 28B] FIG. 28B shows an (inaccurate) detector system layout, the corresponding relative correlation map, an example of a method for converting relative correlation to effective distance, and the resulting effective distance map. DETAILED DESCRIPTION OF THE INVENTION

[0009] To address at least one problem identified in known techniques, the present disclosure discloses techniques for acquiring radiation events in a gamma ray detector system and processing the radiation event data to identify relative positioning or placement errors of detector elements in the detector system.

[0010] Inter-detector scattering and background radiation from the scintillator crystals can cause coincidence events. These coincidence events between multiple detector elements can be tracked and processed. Detector elements designed to be adjacent to each other are expected to share more of these coincidence events, while detector elements designed to be spaced apart from each other are expected to share very few of these coincidence events.

[0011] In one embodiment, the average distance between assembly events (coincidence events caused by inter-detector scattering and background radiation from the scintillator crystals) for each detector element is measured to determine if a placement error exists. In another embodiment, another distance metric, such as median distance, modal distance, etc., is used.

[0012] In another embodiment, the relative correlation between assembly events for each detector element is measured to determine placement error.

[0013] Alignment errors regarding different detector elements, such as errors between crystals, modules or detector units, can be identified.

[0014] Additionally, techniques are disclosed for removing noise and / or bias resulting from random assembly events (similar to, but slightly different from, random coincidence events; see detailed description for more details).

[0015] In another embodiment, machine learning systems and / or look-up tables (LUTs) are used to specifically identify and correct types of layout errors.

[0016] It should be noted that the summary above does not define every embodiment and / or every incrementally novel aspect of the invention(s) set forth in this disclosure or the claims. Instead, this summary merely provides a preliminary description of different embodiments and corresponding points of novelty. Additional details and / or possible aspects of the invention(s) and embodiments will be presented to the reader by the detailed description of the disclosure and corresponding drawings, as further described below.

[0017] The following disclosure presents many different embodiments or examples for implementing different features of the provided subject matter. To simplify the disclosure, specific examples of components and arrangements are described below. Of course, these are merely examples and are not intended to be limiting. Furthermore, although the disclosure may repeat reference numerals and / or letters in various examples, this repetition is for purposes of simplicity and clarity and does not in itself indicate a relationship between the various embodiments and / or configurations described.

[0018] The order of different steps described herein is presented for clarity. In general, these steps can be performed in any suitable order. Furthermore, although different features, techniques, configurations, etc. are described in different places herein, it is intended that the concepts can be implemented independently of one another or in combination with one another. Accordingly, the present invention may be embodied and disclosed in many different ways. While this disclosure describes a PET detector system including scintillator crystals, photosensors, and associated electronics to illustrate various embodiments, these concepts can be applied to similar ionizing radiation detector systems, such as gamma-ray and X-ray detectors. Examples of other systems to which the present disclosure may be applied include direct conversion semiconductor detectors using materials such as cadmium telluride, cadmium zinc telluride, silicon, and germanium.

[0019] The methods of the present disclosure may be implemented within a PET scanner, as shown in FIGS. 1 and 2. Accordingly, FIGS. 1 and 2 show a PET scanner 100 including multiple gamma-ray detectors (GRDs) 101 (e.g., GRD1, GRD2, through GRDN), each configured as a rectangular detector module. According to one embodiment, each PET detector ring that forms a circular bore 102 around the gantry 104 includes 40 GRDs. In another embodiment, there are 48 GRDs, with more GRDs used to create a larger bore size for the PET scanner 100. The GRDs include scintillator crystal arrays for converting gamma rays into scintillation photons (e.g., at optical, infrared, and ultraviolet wavelengths) that are detected by a photodetector. Each GRD can include a two-dimensional array of individual detector crystals that absorb gamma radiation and emit scintillation photons. The scintillation photons can be detected by a two-dimensional array of photomultiplier tubes (PMTs), also located in the GRD. A waveguide can be positioned between the array of detector crystals and the PMTs. Furthermore, each GRD can include multiple PMTs of various sizes, each positioned to receive scintillation photons from multiple detector crystals. Each PMT can generate an analog signal indicating when a scintillation event occurs and the energy of the gamma ray that generates the detection event. Furthermore, photons emitted from one detector crystal can be detected by multiple PMTs, and based on the analog signal generated by each PMT, the detector crystal corresponding to the detection event can be determined using, for example, Anger logic and crystal decoding. However, Anger calculations are not necessarily required when there is a one-to-one correspondence between the crystals and the photodetectors. For detectors with a one-to-one correspondence between the crystals and the photodetectors, silicon photomultiplier tubes (SiPMs) or avalanche photodiodes (APDs) are common choices for photodetectors.

[0020] FIG. 2 shows a schematic diagram of a PET scanner system having a GRD positioned to detect gamma rays emitted from an object (OBJ). The GRD can measure the timing, position, and energy corresponding to each detected gamma ray. In one embodiment, the gamma ray detectors are arranged in a PET detector ring as shown in FIGS. 1 and 2 and described herein. It is understood that the single PET detector ring of FIG. 2 can be extrapolated to include any number of PET detector rings along the axial length of the PET scanner. The detector crystals can be scintillator crystals, which have individual scintillator elements arranged in a two-dimensional array. The scintillator elements can be any known scintillation material (e.g., LYSO, BGO, LSO). The PMTs can be arranged so that light from each scintillator element can be detected by multiple PMTs for Anger calculation and crystal decoding of scintillation events.

[0021] 2 shows an example of the arrangement of the PET scanner 100, in which an object OBJ to be imaged is placed on a table 206, and GRD modules GRD1 through GRDN are arranged around the object OBJ and the table 206. The GRDs may form a PET detector ring or may be fixedly connected to a circular bore 202 that is fixedly connected to a gantry 204. The gantry 204 houses many of the components of the PET scanner. The PET scanner gantry 204 also includes an open aperture defined by the cylindrical bore 202 through which the object OBJ and the table 206 can pass. Gamma rays emitted in opposite directions from the object OBJ due to annihilation events can be detected by the GRDs, and timing and energy information can be used to determine coincidence counts for gamma ray pairs.

[0022] 2 further illustrates circuitry and hardware for acquiring, storing, processing, and distributing gamma ray detection data. The circuitry and hardware include a processor 207, a network controller 203, a memory 205, and a data acquisition system (DAS) 208. The PET imaging device further includes a data channel that transmits detection measurements from the GRD to the DAS 208, the processor 207, the memory 205, and the network controller 203. The data acquisition system 208 can control the acquisition, digitization, and routing of detection data from the detector. In one embodiment, the DAS 208 controls the movement of the table 206. The processor 207 performs functions including identification of placement errors, pre-reconstruction processing of detection data, image reconstruction, and post-reconstruction processing of image data.

[0023] According to one embodiment, the processor 207 of the PET scanner 100 of Figures 1 and 2 can be configured to perform the methods described herein. The processor 207 can comprise a CPU that can be implemented as separate logic gates, such as an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA), or another Complex Programmable Logic Device (CPLD). The FPGA or CPLD implementation may be coded in VHDL, Verilog, or any other hardware description language, and the code may be stored in electronic memory directly integrated into the FPGA or CPLD or in a separate electronic memory. Furthermore, the memory 205 can be a hard disk drive, a CD-ROM drive, a DVD drive, a flash drive, RAM, ROM, or any other electronic storage known in the art. The memory 205 may also be non-volatile, such as ROM, EPROM, EEPROM, or flash memory. The memory 205 may also be volatile, such as static or dynamic RAM, and a processor, such as a microcontroller or microprocessor, may be provided to manage the electronic memory as well as the interaction between the FPGA or CPLD and the memory.

[0024] Alternatively, a CPU within processor 207 may execute a computer program including a set of computer-readable instructions for performing the methods described herein, the program being stored in any of the non-transitory electronic memory and / or hard disk drives, CDs, DVDs, flash drives, or any other known storage media described above. Furthermore, the computer-readable instructions may be provided as a utility application, background daemon, or operating system component, or a combination thereof, executing in conjunction with a processor, such as an Intel Xeon processor or an AMD Optetron processor, and an operating system, such as Microsoft Vista, UNIX, Solaris, LINUX, Apple, MAC-OS, and other operating systems known to those skilled in the art. Furthermore, a CPU may be implemented as multiple processors that function cooperatively to execute instructions in parallel.

[0025] In one embodiment, the PET scanner may include a display for displaying the reconstructed images, etc. The display may be an LCD display, a CRT display, a plasma display, an OLED, an LED, or any other display known in the art.

[0026] A network controller 203, such as an Intel Ethernet PRO network interface card from Intel Corporation, USA, can interface between various components of the PET imaging device. Furthermore, the network controller 203 can also interface with an external network. As will be apparent, the external network can be a public network, such as the Internet, or a private network, such as a LAN or WAN network, or any combination thereof, and can include a PSTN or ISDN subnetwork. The external network can also be wired, such as an Ethernet network, or wireless, such as a cellular network, including EDGE, 3G, and 4G wireless cellular systems. The wireless network can also be WiFi, Bluetooth, or any other known wireless form of communication.

[0027] FIG. 3 highlights a subset of components within PET scanner 100 that are generally referenced in this disclosure. Module 301 is generally the smallest element capable of independently operating as a PET detector and includes scintillator crystals and photosensors (and optional electronics). For example, module 301 in FIG. 3 includes an array of crystals. Detector unit 302 may be composed of multiple modules 301. For example, detector unit 302 in FIG. 3 includes eight modules 301. Detector ring 303 may be composed of multiple detector units 302. For example, detector ring 303 in FIG. 3 includes ten detector units. The detector ring forms a circular bore 102 around gantry 104, as previously shown in FIG. 1. A detector element is defined herein as a crystal, module 301, or detector unit 302.

[0028] Electrically and / or optically independent detector elements can often form coincidence events due to inter-detector scattering or energy escape of multi-stage background radiation in the crystal. These multiple detector events between detector elements (i.e., between crystals, modules, detector units) can be used to verify the relative positioning of the detector elements and to detect layout errors.

[0029] FIG. 4 illustrates an example of inter-detector scattering. A first detector module 401 and a second detector module 402 are shown, each containing multiple crystals 403 and multiple photosensors 404. An external source of high-energy gamma rays 405 (such as Cs-137 (662 keV) or Ge-68 (a coincidence 511 keV pair)) enters one of the crystals 403 in the first module 401, causing a first energy deposition 406 (i.e., collision). Compton-scattered gamma rays 407 then cause a second collision 408 in the second detector module 402. The first collision 406 and the second collision 408 occur at approximately the same time on two different detector elements. This inter-detector scattering can result in coincidence events between two or more detector elements.

[0030] Additionally, some scintillator materials are radioactive (i.e., the scintillator crystals generate background / intrinsic radiation). For example, Lu-176 naturally occurs in LYSO or LSO (two common PET scintillators). As shown in FIG. 5, Lu-176 emits a series of gamma rays nearly simultaneously as it decays by beta radiation. Typically, the energy from the beta radiation is contained in the element where the decay occurred, but occasionally the gamma rays can be absorbed by other detector elements. This type of background radiation in the crystals can also cause coincidence events. For example, FIG. 6 shows the decay 601 of Lu-176 in the first detector module 602. The decay 601 results in an energy deposition 603 of 202 keV in the first module 602, an energy deposition 604 of 88 keV in the first module 602, and an energy deposition 605 of 307 keV in the second detector module 606. It can be seen that energy can be deposited at nearly the same time in two detector elements, thereby forming a coincidence event.

[0031] Because these types of coincidence events typically occur between detector elements that are closer together than between detector elements that are farther apart, these coincidence events can be tracked, measured, and compared to a threshold or designed layout to estimate placement errors.

[0032] As an example and introduction to the concept, Figures 7A and 7B illustrate an embodiment of the present disclosure where the layout error is a cabling error (misconnected modules). Figure 7A shows the designed / expected layout. In this example, module A 701 is connected to channel 1 702, module B 703 is connected to channel 2 704, module C 705 is connected to channel 3 706, and module D 707 is connected to channel 4 708. Each module is connected to its respective channel via cable 709, which is part of a data acquisition system 710. In this example, in module C 705, a high-energy incident gamma ray 711 may cause a first collision 712, and in module D 707, a Compton scattered gamma ray 713 may cause a second collision 714. The first collision 712 is recorded by channel 3 706, and the second collision 714 is recorded by channel 4 708. The resulting data may be processed to determine the distance between collisions as, for example, 1 (i.e., |4-3|=1). This result may typically be expected when module C 705 and module D 707 are adjacent. FIG. 7B shows the measured / realized layout. Again, each module is connected to a channel via a cable, which is part of the data acquisition system. However, in this example, two cables have been swapped: module A 701 is connected to channel 1 702, module B 703 is connected to channel 3 706, module C 705 is connected to channel 2 704, and module D 707 is connected to channel 4 708. Again, in this example, a high-energy incident gamma ray 711 may cause a first collision 712 at module C 705, and a Compton scattered gamma ray 713 may cause a second collision 714 at module D 707. However, in this example, the first collision 712 is recorded by channel 2 704 and the second collision 714 is recorded by channel 4 708 because the cables are connected incorrectly.The resulting data may be processed to incorrectly determine the distance between collisions as, for example, 2 (i.e., |4-2|=2). In one embodiment, a predetermined threshold may be established so that errors can be identified (e.g., by a processor) and appropriate adjustments / corrections can be made. Adjustments and corrections may include manually swapping incorrectly connected cables or adjusting one or more look-up tables (LUTs) used to associate each detector element with its respective physical location. In this exemplary embodiment, only a single event is considered when identifying defects. In another exemplary embodiment, defects are identified after processing many events to minimize the effects of random statistical fluctuations.

[0033] The above-described concepts can be applied in many different embodiments. The flowchart in Figure 8 outlines an exemplary embodiment of the present disclosure. First, in step 805, radiation event data (i.e., impacts) are acquired. The acquired radiation event data may include position information (e.g., the location of the detector element where the impact occurred), the time of each impact, and the amount of energy deposited by each impact.

[0034] In step 810, the acquired radiation event data is used to identify coincidence events between detector elements. Coincidence events can be identified in real time via a coincidence circuit or offline via a comparison of time measurements in software. In the context of conventional PET scanning, coincidence events typically refer to events resulting from positron annihilation along the line of response (LOR) between two detectors. In the discussion of this disclosure, coincidence events primarily focus on coincidence events caused by inter-detector scattering or energy escape due to multi-stage background radiation in the scintillator crystal. Thus, the identified coincidence events are events caused by inter-detector scattering and / or background radiation. Under normal operation, coincidence events caused by inter-detector scattering or energy escape due to multi-stage background radiation in the scintillator crystal are more likely to occur in nearby detector elements (rather than detector elements at opposite ends of the LOR), and the frequency of these coincidence events typically decreases as the distance between the detector elements increases. To distinguish these specific coincidence events from conventional coincidence events that typically occur between detectors along the LOR, geometric constraints (i.e., geometric cuts that define subregions) may be applied to the detector ring. In this manner, collisions that occur between multiple detector elements within the coincidence time window and their individual geometric constraints (i.e., subregions) may be identified or referred to as assembly events. The process of collecting assembly events is referred to as assembling events.

[0035] According to one embodiment, a geometric cut may divide a detector ring into fixed sub-regions. For example, as shown in FIG. 9 , a detector ring 901 may be divided into a first quadrant 902, a second quadrant 903, a third quadrant 904, and a fourth quadrant 905. Although collisions A 906, B 907, and C 908 may all occur within one coincidence window (Δt), collisions A 906 and B 907 are assembled into a single assembly event when they occur only within the same quadrant 903. As a result, collision C 908 is not assembled into a single assembly event with collisions A 906 and B 907 because it occurs outside the sub-region defined by the geometric cut. Only coincidence events occurring within each sub-region are assembled into a single assembly event. The fixed sub-regions defined by the geometric cuts may all be of equal size, or may be various sub-regions of different sizes.

[0036] FIG. 10 illustrates an example of a geometric constraint applicable in an exemplary embodiment of the present disclosure. For example, for each identified coincidence event, a geometric region may be positioned to center the single event with the highest energy among the group of collisions occurring within the coincidence window. For example, as shown in FIG. 10, collision A 1001, collision B 1002, collision C 1003, and collision D 1004 may be detected within a given coincidence time window. Of the four collisions, collision A 1001 has the highest energy. If collision A 1001 occurs within detector ring 1006 at an angle θ from zero-degree reference point 1005, all collisions occurring within the range of θ+90° to θ−90° are assembled into an assembly event that includes collision A 1001. In this example, the assembly event includes collision B 1002 and collision C 1003. Collision D 1004 is not included in the assembled event because it did not occur within the +90° θ to -90° θ range of collision A 1001. Of course, in other embodiments, other tolerances besides 90 degrees can be used as appropriate.

[0037] 8, in step 815, the average distance between all assembled events that occurred for each detector element (in this case, a scintillator crystal) is calculated. In other embodiments, a different metric may be used, such as median distance or modal distance, and different detector elements (e.g., modules, detector units) may be used to calculate the distance between detector elements (e.g., between modules or detector units).

[0038] To elaborate on the latter point, consider collisions detected at different levels of detector elements, here (1) the crystal / pixel level, (2) the module level, or (3) the detector unit level. The crystal / pixel level is the lowest level, followed by the module level and then the detector unit level. Other levels may be defined as appropriate. Each of these levels refers to the detector element used in the position calculation. For example, if the average distance between detector elements is calculated at the crystal level, then the position of each crystal is used in the position calculation. Alternatively, if the average distance is calculated at the module level or the detector unit level, then only the position of the module or detector unit (respectively) is used in the position calculation. This concept of levels can also be applied when calculating correlations between detector elements. Below, such examples are described in detail with reference to subsequent figures.

[0039] In one embodiment, calculations at any level can detect positioning or cabling errors at that same level. For example, if calculations are performed at the module level, it is possible to detect swapped modules. Lower levels can also detect positioning or cabling errors at higher levels. For example, swapped modules within the same detector unit can be detected at the crystal level, but not at the detector unit level.

[0040] In order to be able to detect rotation of the detector elements, calculations must be performed at a level lower than the rotated detector elements. For example, in order to be able to detect rotation errors of the modules, calculations must be performed at a lower level, in this example at the crystal level. Similarly, in order to detect rotation at the detector unit level, calculations must be performed at the module level or at the crystal level.

[0041] 8, in step 820, the average distance between assembly events (in this example, between crystals) for each detector element is displayed on an average distance map. The collected distance metric data (in this case, the average distance between assembly events) is displayed on the map to aid in visualization of the collected results. Note that, as noted above, step 820 is optional and may be omitted in other embodiments without affecting the functionality of the method.

[0042] Thus, FIG. 11A shows a two-dimensional detector ring unfolded in three different levels of view: a crystal (or pixel) level view, a module level view, and a detector unit view. As shown in FIG. 11A, the detector ring may include 48 detector units, each detector unit including 5×1 modules, and each module including 12×12 crystals. FIG. 11B shows various intentionally created alignment errors for the detector ring, and FIGS. 11C-11E show the results obtained at three different levels. For example, FIG. 11C shows an average distance map at the crystal level, where all defects are visible by the lighter colored detector elements. FIG. 11D shows an average distance map at the module level, where module swaps and detector unit level defects are visible (by the lighter colored detector elements), but module rotations are not. Furthermore, FIG. 11E shows an average distance map at the detector unit level. In this map, only the detector unit swaps are visible (by the lighter colored detector units).

[0043] In step 825, defects are identified by values ​​exceeding a threshold level. For example, layout errors can be identified by visually inspecting the collected data (e.g., inspecting an average distance map). Any errors can be identified by comparing the data to a predetermined threshold level. In one embodiment, the predetermined threshold level may be set low enough to identify all defects but high enough to provide a low false positive rate. For example, noise and bias may be used as determining factors when setting the predetermined threshold level for defect detection. Instead of explicitly setting a threshold, in step 825, a trained operator may inspect the image and identify defects based on their experience observing many examples of defective and non-defective systems.

[0044] In addition to the above, random assembly events are an important source of noise and bias. However, before defining random assembly, it is necessary to understand initiation events. An initiation event is an event that traditionally results in a single collision (i.e., all energy is deposited in a single crystal; although the energy may be deposited in separate, separate processes, such as Compton scattering with photoelectric absorption, they occur nearly simultaneously and are not determined as separate detection events by the detector if they occur in the same pixel). A single positron annihilation typically results in two initiation events because the two 511 keV gamma rays emitted from the annihilation traditionally result in two collisions, each of which strikes a different crystal, resulting in a separate initiation event. However, Lu-176 decay within a scintillator crystal in a detector typically results in a single initiation event. Even though Lu-176 decay can produce a cascade of beta particles and gamma rays, this decay often produces only a single collision as the beta particle, and the gamma cascade is typically absorbed within the scintillator crystal.

[0045] In addition to the above, while an initiation event often results in only one collision, it is possible for a single initiation event to result in multiple collisions. For example, as illustrated in FIG. 6, a single Lu-176 decay can result in some energy being deposited in the crystal where the decay occurred, while another portion of the energy is deposited in a neighboring crystal by a gamma ray escaping the first crystal. Another example of an initiation event that results in multiple collisions is a 511 keV gamma Compton scattering in one crystal (and depositing some energy in the process), and a scattered gamma ray depositing energy in another crystal (as illustrated in FIG. 4). Thus, the goal of the assembly process is to identify collisions (i.e., assemble events) that result from the same initiation event.

[0046] As described above, random assembly occurs during the assembly process. Random assembly occurs when an event is assembled from collisions resulting from two or more initiating events. FIG. 12 shows an example of random assembly. In the example of FIG. 12, a first initiating event 1201 (e.g., a 511 keV incident gamma ray) causes a first collision 1202 in a first detector module 1203, and a second initiating event 1204 (e.g., a 511 keV incident gamma ray) causes a second collision 1205 in a second detector module 1206, with both collisions occurring within the same coincidence time window. As a result, the assembly process generates a random assembly event composed of the first collision 1202 and the second collision 1205. FIG. 13 shows another example of random assembly. 13, a first initiation event 1301 (e.g., Lu-176 decay) causes a first collision 1302 in a first detector module 1303, and a second initiation event 1304 (e.g., Lu-176 decay) causes a second collision 1305 in a second detector module 1306, with both collisions occurring within the same coincidence time window. Again in this example, the assembly process generates a random assembly event composed of the first collision 1302 and the second collision 1305, even though the first collision 1302 and the second collision 1305 are caused by different initiation events.

[0047] One way to avoid significant data contamination due to random assembly is to use a source with relatively low radioactivity. The number of random assemblies increases with the square of the radioactivity. Lu background is a good choice for scintillators in which it exists (such as LYSO, LSO, and LGSO). The activity of the Lu background is low enough to prevent significant problems from random assembly, but provides a high enough event rate so that sufficient data can be acquired within acquisition times of a few seconds to a few minutes. Furthermore, the Lu background is advantageous because it is always present (i.e., no external source is required), and the half-life of Lu-176 is very long, making the Lu background fraction essentially constant over the lifetime of the PET scanner.

[0048] In one embodiment, the region and / or subregion created through the geometric cut in which assembly events are identified is referred to as the assembly region. The assembly region needs to be large enough (geometrically) to cover all possible defect types. If there are no constraints on the possible defect types, events may be assembled throughout the entire ring. If the source is a positron emitter that results in consecutive 511 keV gamma rays of coincidence, a geometric cut must be made to the assembly region to eliminate the majority of coincidences caused by separate 511 keV gamma rays, so that collisions from different 511 keV gamma rays do not assemble into a single event. By limiting the assembly event region, the occurrence of random assembly is less likely, thereby reducing the contribution of random assembly to noise and / or bias. Therefore, one way to reduce the contribution of random assembly to bias and noise is to limit the size of the assembly region used in the assembly process (e.g., the size of the assembly region is set to the minimum size sufficient to cover all desired defect types). In other words, the assembly area only needs to be slightly larger than the maximum deviation from the intended detector element placement error that needs to be detected. For example, when detecting rotated modules, the assembly area only needs to be slightly larger than the diagonal length of one module. As another example, when detecting transposed modules, the assembly area only needs to be slightly larger than the longest dimension of the module. In an exemplary embodiment, slightly larger may be the longest dimension or diagonal length of the detector element under test for placement errors, rounded to the nearest whole pixel (e.g., if the maximum diagonal length of the module is 16.97 pixels, the assembly area would have a radius of 17 pixels). Of course, slightly larger than the maximum deviation could have other meanings, such as increasing the longest dimension or diagonal length by a factor of 1-25%, or increasing the longest dimension or diagonal length by a fixed number of pixels.

[0049] The selection of the geometric assembly range can often be selected based on knowledge of the manufacturing or service process. Detector layout errors often result from defects occurring over a known distance range. For example, two adjacent modules may be swapped by incorrectly connecting cables, or multiple modules within a detector unit may be swapped due to incorrect cabling. In these examples, the maximum defect range is set to a scale close to the longest dimension of the module or detector unit, respectively. In another example, two adjacent detector units may be swapped by incorrectly connecting cables. Detector unit cable swaps over longer distances are relatively easy to detect by visual inspection. In this example, the maximum defect range is set to a scale close to the longest dimension of the detector unit. In a third example, the module may be rotated 90 degrees or 180 degrees. In this example, the maximum defect range is set to a scale close to the longest diagonal dimension of the module. In a fourth example, the detector unit may be rotated 180 degrees. In this example, the maximum defect range is set to a scale close to the longest diagonal dimension of the detector unit.

[0050] For illustration, FIGS. 14A-14C show the results of reducing bias and noise by confining the assembly area. The same defects as in FIG. 11B are created. The expanded two-dimensional view of the detector unit shown in each of FIGS. 14A-14C corresponds to the detector unit of FIG. 11A. Here, the detector unit has a width of 12 pixels and a length of 60 pixels, and the modules have a width of 12 pixels and a length of 12 pixels. In one embodiment, FIG. 14A shows the result when the assembly area is divided into multiple quadrants, each covering 12 modules / detector units in the entire axial and transaxial directions. Placement errors are still visible, as indicated by the pronounced outlines around incorrectly placed modules and detector units, but correctly placed elements show some noise and bias, as indicated by different colors. In FIG. 14B, the assembly area is further constrained to have a radius of 62 pixels (the approximate diagonal length of the detector unit is √(12 2 +60 2 ) = ∼61.2 pixels, which can be rounded up to 62 pixels. The defect is now more visible as indicated by the increased contrast between incorrectly and correctly placed detector elements. In Figure 14C, the assembly area is further restricted to 18 pixels (the approximate diagonal length of the module is √(12 2 +12 2) = ∼16.7 pixels, which can be rounded up to 17. The defects are clearly visible, as indicated by the strong contrast between incorrectly and correctly placed detector elements. As shown throughout Figures 14A-14C, noise and bias visibly decrease as the assembly area decreases. This can be further observed in Figure 15, which contains a series of plots showing the effect on bias and noise of limiting the assembly area. For example, the series of plots shows the average distance and standard deviation along each row for a defect-free detector ring. The black curve represents the bias, and the gray dashed envelope represents the noise. As shown in Figure 15, bias and noise decrease as the assembly area decreases.

[0051] Bias from random assembly can also be estimated and removed, allowing the threshold level to be lowered. By considering that true and random assembly events contribute to the average distance map, the average distance map can be written as a weighted sum of two components, as shown in Equation (1) below.

number

[0052] Furthermore, it can be written as the following equation (2).

number

number

[0053] According to one embodiment, if the data is acquired as a single collision list (i.e., event assembly is done offline in a later processing step), the data from a single acquisition can be assembled at multiple values ​​of Δt, where Δt must be large enough to assemble all true events. The value chosen for Δt depends on the calibration state of the system. If timing channel offset and energy walk calibration have not yet been performed, a larger value of Δt must be used to ensure that all true assembly events are assembled. For uncalibrated systems, a Δt in the range of 3-10 ns is often reasonable, but the selection depends on the details of the system. For each pixel, [d i (Δt) vs.Δt] to obtain d_true i , d_rand i , and αT can be determined. The same approach can be used if the event assembly is done in real time, but it should be noted that multiple data sets need to be acquired.

[0054] According to one embodiment of the present disclosure, the above-described approach is as follows. First, a data set is acquired from a detector ring with a known defect, such as when the first and second modules 1601 and 1602 are swapped, as shown in FIG. 16A. FIGS. 16B-16E show data processed to generate an average distance map using different coincidence windows Δt from the detector described in FIG. 16A. In the examples of FIGS. 16B-16E, data was acquired using only Lu background, which has a significantly lower event initiation rate. Of course, the effect of random assembly can be greater when an external radiation source is used. In FIG. 16B, the acquired data has a coincidence window of one time constant (Δt=t1). In FIG. 16C, the acquired data has a coincidence window of three time constants (Δt=3t1). In FIG. 16D, the acquired data has a coincidence window of eight time constants (Δt=8t1). In Figure 16E, the acquired data has a coincidence window of 23 time constants (Δt = 23t1). In Figures 16B-16E, each average distance map reflects the crystal level. Similar techniques can be applied when using other metrics or other levels, such as the module level or the detector unit level. Also, note that the assembly regions in Figures 16B-16E are divided into fixed quadrants. The acquired data is then fitted pixel by pixel to produce the true assembly average distance map (d_true i ) is extracted. The swapped modules are clearly visible as the periphery of the adjacent squares is brighter. Figure 17B shows the average distance map (d_rand i ) is shown. Using four quadrants as assembly regions results in four repetitions of the pattern.

[0055] In one embodiment, the average distance map (d_rand iThe general pattern of Δt can also be predetermined as in equation (4) below to avoid dependence on experimental conditions and parameters (e.g., radioactivity and Δt).

number

number

[0056] Furthermore, if a fixed assembly region, such as a division into quadrants, is used, the following equation (6) can be used:

number

[0057] An example of results from a preliminary calculation of the average map of random assembly events is shown in Figure 20. Here, the fixed regions are the four quadrants of the detector ring. Note the similarities between Figure 20 and Figure 17B, even though they were extracted in different ways. Also, note some symmetries that can be exploited to reduce the computational load.

[0058] For a selected set of experimental conditions (e.g., isotopes, radioactivity, etc.), it can be written as follows:

number

[0059] One way to obtain β (hereafter referred to as Method A) is by selecting the value of β that minimizes the variation, which for a chosen set of experimental conditions (isotopes, radioactivity, etc.) can be written as follows:

number

[0060] FIG. 21 shows the mean distance map for a defect-free system without correction (top plot labeled "No Correction") and the true mean distance map estimated using the value of β obtained by Method A (bottom plot labeled "Minimum Change"). Additionally, FIG. 22 shows the mean distance map for a system with a swapped module (same defect as in FIG. 16A) without correction (top plot labeled "No Correction") and the true mean distance map estimated using the value of β obtained by Method A (bottom plot labeled "Minimum Change"). Therefore, using β obtained by Method A allows the threshold for detecting defects to be set to a more sensitive value.

[0061] A second method for obtaining β (hereinafter referred to as Method B) can be written as in the following equation (9).

number

number

[0062] P i is easily measured, so β i To estimate the random assembly rate R for each pixel, i It is necessary to estimate R i is ∫{1~r max}Since it is proportional to C(r)·dr, it can be written as the following equation (11).

number

number

number

number

number

[0063] According to one embodiment, FIG. 23 shows the mean distance map for a defect-free system without correction (top plot labeled "No correction") and the true mean distance map estimated using the values ​​of β obtained with Method B (bottom plot labeled "Correction with β from Method B") (the r used threshold= 30 pixels). Additionally, FIG. 24 shows the mean distance map for a system with swapped modules (same defect as in FIG. 16A) without correction (top plot labeled "No correction"), and the true mean distance map estimated using the value of β obtained with Method B (bottom plot labeled "Correction with β from Method B"). By using the selection of β obtained with Method B, it is possible to set the threshold for detecting defects to a more sensitive value.

[0064] Furthermore, if β is determined for one set of conditions (activity and Δt), new values ​​of β can be estimated for different choices of activity and Δt. Given that β is determined for an activity Activity0 and a coincidence window Δt0, a new value of β for each pixel i is calculated by substituting equations (16) and (17) below into equation (18) and recalculating the average value of β to obtain equation (19).

number

number

number

number

[0065] According to an exemplary embodiment of the present disclosure, the above-described method can also be applied to geometric cuts centered on maximum energy collisions. Referring back to FIG. 19, y=Q / 2 can be used to calculate the angle involving C(r). FIG. 25 shows the results of a map of random assembly pre-computed using the above-described method for a centered region. Also, note several symmetries that can be exploited to reduce the computational load.

[0066] In another exemplary embodiment, bias can be removed by obtaining an uncorrected distance map (i.e., no bias correction applied) from one or more systems free of known defects. If multiple systems free of known defects are available, their individual uncorrected distance maps may be averaged to generate an average uncorrected distance map. In either case, a distance map obtained by a new system (i.e., a system under test) not yet known to be free of defects can be corrected by normalizing the distance map with a distance map resulting from the absence of known defects. That is, the distance map can be divided by the known-defect-free map element by element to arrive at a corrected or normalized distance map. In another exemplary aspect, a distance map from a new system can be corrected by subtracting the known-defect-free distance map element by element. Both methods produce a corrected distance map with very little bias.

[0067] Another way to deal with bias and its effect on threshold levels is to have a threshold level map (i.e., threshold levels that vary with position). The way the map varies with position is determined by the bias. The threshold level map can be determined empirically or by calculating bias levels as described above. As an example of an empirical method, the threshold level map can be determined by averaging results from several defect-free scanners and then increasing the threshold levels by a multiple of the local standard deviation to prevent false positives.

[0068] Since defects are equivalent to changes in the LUT, they can be corrected by using different LUTs according to exemplary embodiments of the present disclosure. Furthermore, real data can be used to generate data with any conceivable defects. A dataset with many different defects can be generated, which can then be fed into a machine learning application, enabling the ability to identify defects and determine how to correct them. Training data can be generated by generating a LUT with defects, which is used to associate each detector element with its physical location.

[0069] According to one exemplary embodiment of the present disclosure, a correlation metric can be used to measure the relative correlation of two detector elements, with reference now to the flow diagram of Figure 26. This method is similar to the previously described embodiment, except that a correlation score is used instead of a distance metric.

[0070] First, in step 2605, radiation events are acquired, which may include the crystal's inherent multi-phase background radiation or coincidence events between detector elements using an external radiation source. The acquired radiation event data may include position information (e.g., the location of the detector element where the collision occurred), the time the collision occurred, and the amount of energy deposited for each collision in the PET detector.

[0071] In step 2610, the acquired radiation event data is used to identify coincidence events with an appropriately selected geometric cut. Similar logic as described above is applicable when selecting the geometric cut in this embodiment. Optionally, coincidence events may be processed with time or energy calibration, and even filter events with additional time or energy cuts.

[0072] Next, in step 2615, the relative correlation between the detector elements is determined. Such correlation is normalized to minimize effects from variable detector efficiency, background intensity, and exposure to external energy sources. One way to define correlation is to calculate the fraction of coincidence events relative to the total number of events detected by the two detectors, as in Equation (20) below:

number

[0073] Next, in step 2620, the correlation (or effective distance) data is converted into a correlation (or effective distance) score. As an example, FIG. 27 shows a layout of 10 precisely positioned detector modules (each module is numbered 1 to 10). The correlation between all neighboring modules can be measured and calculated. This correlation is the module-level correlation. As an example, for the layout shown in FIG. 27, the correlation score can be calculated using the following equation (21):

number

[0074] Returning to FIG. 26, in step 2625, the correlations (or effective distances) between detector elements are stored and displayed on a map. This allows the map to be compared to a predefined template to identify layout errors. FIG. 28A shows an example layout 2801 consisting of 10 correctly positioned modules (i.e., boards) labeled 1 through 10. Correlation data can be obtained and displayed on correlation map 2802. The correlation score for this map 2802 is 0.99. The correlation data can also be converted to effective distance map 2803 using conversion table 2804. The effective distance score for this map is 21. FIG. 28B shows an example layout 2805 in which boards 8 and 10 have been incorrectly swapped. Correlation data can be obtained for this layout and displayed on correlation map 2806. The correlation score for this map is 0.58. The correlation data can also be converted to effective distance map 2807 using table 2804. The effective distance score for this map is 30. These suspect correlation and effective distance scores serve to indicate placement errors.

[0075] If all detector elements are accurately positioned, the correlation score will be at its maximum value (high correlation between elements). If an effective distance score is used, the effective distance will be at its maximum value (low distance between assembled events). A predetermined threshold can be used on the overall score to determine whether the layout is correct. If no threshold is applicable, scores for all possible layout configurations can be calculated, and if the default configuration does not have an extreme score (highest value for correlation score, lowest value for effective distance) compared to the other configurations, a layout defect is likely to exist. Due to system symmetry, multiple configurations may all give extreme scores. In this case, the configuration most similar to the default configuration can be used as the correct one. Furthermore, as in the previous embodiment, correlation can be calculated at different levels (e.g., crystal level, module level, detector unit level).

[0076] Depending on the manufacturing uniformity of the detector, some calibration of the detector may be required. Detectors are generally divided into two categories. The first category of detectors, referred to as Category 1 detectors, are designed so that the signal from the scintillator crystal is shared across multiple photosensor / electronics channels (i.e., the scintillator pitch is different from the photosensor pitch). The second category of detectors, referred to as Category 2 detectors, minimizes the shared signal from the scintillator crystal so that the scintillator pitch is equal to the photosensor pitch. Generally, Category 1 detectors require position calibration before individual scintillator elements can be determined for impact. On the other hand, Category 2 detectors can identify individual crystals for impact without any calibration. The methods described in this disclosure are effective for properly calibrated Category 1 detectors as well as Category 2 detectors.

[0077] In view of the above, full position calibration can be slow and may sometimes require manual review and adjustment of segmentation results. However, for the purposes of the present invention, full position calibration of Category 1 detectors is not always required. It is only required that spatial information within the module be available to calculate distance maps or correlation metrics at a level lower than the module level. Therefore, partial position calibration of Category 1 detectors can also be performed, which allows the methods described in this disclosure to be implemented without requiring a full calibration process.

[0078] Below is an example of how partial position calibration can be performed to achieve sufficient accuracy and generate distance maps or correlation metrics at a level lower than the module level. In this example, the module is divided into nine regions in a 3x3 pattern. In general, anything with a division larger than 2x2 will work, and the number of divisions in the X and Y directions does not need to be the same.

[0079] The first step begins with a flood histogram (two dimensions). A flood histogram is typically created by flooding a detector with incident radiation, calculating the location (X, Y) for each detected event, for example, using Unger logic (which often includes significant spatial distortions), and creating a two-dimensional histogram of the locations (X, Y) of all impacts. Next, the histograms in the X direction are summed to reduce the flood histogram to one dimension. Next, a cumulative sum of the floods reduced to one dimension is calculated, and then the cumulative sum is normalized so that the final value of the cumulative sum is 1 (unit). The Y values ​​where the cumulative sum is closest to 1 / 3 and 2 / 3 are found, designated Y1 and Y2, respectively. These values ​​of Y1 and Y2 are used as the boundaries to define three horizontal regions: Region A, Region B, and Region C. For each of these regions, the flood histogram is summed in the Y direction to reduce to one dimension. For each region, calculate the cumulative sum of the floods dropped in one dimension, then normalize the cumulative sum so that the final value of the cumulative sum is 1 (unit). For each region, find the X values ​​for which the cumulative sum is closest to 1 / 3 and 2 / 3. Call these values ​​X1(R) and X2(R), respectively, where R represents the region label (A, B, C). Use these values ​​of X1(R) and X2(R) as boundaries to define three regions inside each of the original horizontal regions. These nine regions can be assigned (X', Y') values, similar to a crystal ID LUT, or can be used as submodule region LUTs to generate X' and Y' LUTs. One of these types of LUTs (submodules or combinations of X' and Y') can then be used to assign X' and Y' locations to each collision (X, Y) for the purpose of calculating submodule-level distance maps or correlation metrics in Category 1 detectors.

[0080] It should also be noted that PET systems typically require energy and timing offset calibration to achieve sufficient energy and timing resolution prior to clinical use. The techniques described in this disclosure enable a method for identifying placement errors before calibrating the system. In one embodiment, a distance metric is used to assist in identifying placement errors. For multiple element detection events, the distance metric and time metric can be related to each other based on the speed of light because the multiple element events require gamma rays (emitted or scattered by the first crystal) to travel the distance between the crystal elements associated with the multiple element events. However, it is important to consider the effect of timing offset calibration. Timing offset calibration is required due to manufacturing tolerances, such as differences in cable length or timing responses of different photosensors within the detector. Prior to calibration, timing offsets in the system are variable over a range of + / - 1 ns or more, and the standard deviation of timing offsets is often in the range of hundreds of picoseconds. Often, placement errors that need to be detected are at the module level (e.g., swapped or rotated modules). Considering that the typical lateral dimension of a module is approximately 2 cm, and assuming that gamma rays in multiple element events typically travel to neighboring modules, the typical distances involved in the calculations are on the order of 4 cm or less. The speed of light is approximately 2.998 × 10 10 The typical distance corresponds to approximately 130 ps or less of the gamma-ray flight time. This value is a fraction of the typical offset value measured before timing offset calibration, so the timing difference measured for module-level placement errors before timing offset calibration is difficult to correlate with distance due to the large timing offset error.

[0081] In the foregoing description, specific details have been set forth, such as a particular method and system for detecting PET detector placement errors. However, it should be understood that the technology herein may be practiced in other embodiments that depart from these specific details, and that such details are for purposes of explanation and not limitation. The embodiments disclosed herein are described with reference to the accompanying drawings. Similarly, for purposes of explanation, specific numbers, materials, and configurations are set forth to provide a thorough understanding. Nevertheless, embodiments may be practiced without such specific details. Components having substantially the same functional structure are indicated by similar reference characters, and therefore, any redundant description may be omitted.

[0082] Various techniques are described as multiple separate operations to facilitate understanding of various embodiments. The order of description should not be construed as implying that these operations are necessarily order dependent. In fact, these operations need not be performed in the order presented. The operations described may be performed in a different order than in the described embodiments. Various additional operations may be performed and / or described operations may be omitted in additional embodiments.

[0083] Those skilled in the art will also understand that many variations can be made to the operation of the techniques described above while still achieving the same objectives of the present invention. Such variations are intended to be encompassed by the scope of the present disclosure. Accordingly, the foregoing description of embodiments of the present invention is not intended to be limiting. Rather, any limitations to embodiments of the present invention are presented in the following claims.

[0084] According to at least one of the embodiments described above, it is possible to identify placement errors of detector elements included in a gamma ray detector ring.

[0085] Embodiments of the present disclosure may also be defined in the following parenthetical descriptions.

[0086] (1) A method for identifying detector element placement errors in a gamma ray detector ring, the method including: obtaining detector element event data associated with radiation events in detector elements of the gamma ray detector ring; identifying assembly events between detector elements of the gamma ray detector ring based on the detector element event data; calculating a metric based on each of the assembly events between detector elements of the gamma ray detector ring; and identifying detector element placement errors based on the calculation.

[0087] (2) The method of (1), wherein the detector element event data includes the accumulated energy of each of the radiation events, the time of energy accumulation of each of the radiation events, and the energy accumulation location of each of the radiation events.

[0088] (3) A method according to (1) or (2), wherein the assembly event between detector elements is a coincidence event occurring between two or more detector elements within a subregion of the geometric cut of the gamma ray detector ring.

[0089] (4) Any of the methods (1) to (3), wherein the detector element event data includes the accumulated energy of each of the radiation events, and the subregion of the geometric cut for each of the assembly events is centered on the radiation event among the radiation events having the greatest accumulated energy.

[0090] (5) Any of methods (1) to (4) in which the subarea of ​​the geometric cut of the gamma-ray detector ring is slightly larger than the diagonal length of the detector element.

[0091] (6) Any of methods (1) to (5) in which the sub-area of ​​the geometric cut of the gamma-ray detector ring is slightly larger than the longest dimension of the detector element.

[0092] (7) Any of the methods (1) to (6), wherein the calculation calculates a distance metric that defines the distance between assembly events for each detector element as the metric.

[0093] (8) The method according to any one of (1) to (7), wherein the calculation calculates a relative correlation score between detector elements as a metric.

[0094] (9) Any of the methods of (1) to (8), further comprising: calculating a relative correlation score for all detector element layout combinations; identifying a precise detector element layout based on the relative correlation scores for all detector element layout combinations; and adjusting one or more lookup tables used to associate each detector element with its physical location based on the precise detector element layout.

[0095] (10) Any of the methods (1) to (9), further comprising collecting radiation event data in multiple time coincidence windows, identifying contributions to the metric from random assembly events based on the collection, and removing the contributions from the random assembly events from the metric.

[0096] (11) Any of the methods (1) to (10), further comprising collecting radiation event data from a gamma ray detector ring that is known to be free of defects, and removing contributions from random assembly events from metrics of the gamma ray detector ring based on the collection.

[0097] (12) Any of the methods (1) to (11), further comprising generating training data by generating a faulty lookup table used to associate each of the detector elements with its physical location, and training a machine learning system using the training data to identify placement errors.

[0098] (13) Any of the methods (1) to (12), further comprising performing a partial position calibration prior to acquisition when the gamma ray detector ring is a Category 1 detector.

[0099] (14) Any of the methods (1) to (13), wherein the relative correlation score is a comparison between a number of coincidence events between the first detector element and the second detector element to the sum of the total number of radiation events in the first detector element and the second detector element.

[0100] (15) The method according to any one of (1) to (14), wherein the detector element is a crystal.

[0101] (16) Any of the methods (1) to (15), wherein the detector element is modular.

[0102] (17) Any of the methods (1) to (16), wherein the detector element is a detector unit.

[0103] (18) A gamma ray detector system for identifying detector element placement errors in a gamma ray detector ring, comprising: processing circuitry configured to acquire detector element event data associated with radiation events in detector elements of the gamma ray detector ring; identify assembly events between detector elements of the gamma ray detector ring based on the detector element event data; and calculate metrics based on the assembly events between detector elements of the gamma ray detector ring to identify detector element placement errors.

[0104] (19) The system of (18), wherein the processing circuitry is further configured to train a machine learning system based on the training data generated by generating the faulty lookup table to identify placement errors.

[0105] Thus, the foregoing description discloses and describes merely exemplary embodiments of the present invention. As will be understood by those skilled in the art, the present invention may be embodied in other specific forms without departing from its spirit or essential characteristics. Accordingly, the disclosure of the present invention is intended to be illustrative, but not limiting, of the scope of the present invention as well as the remaining claims. The present disclosure, including any readily discernible variations of the teachings herein, defines in part the scope of the following claims, such that the subject matter of the invention is not available to the public. [Explanation of symbols]

[0106] 100 PET scanners 101, GRD1, GRD2~GRDN gamma ray detectors 207 processors 301 Module 302 Detector Unit 303 Detector Ring 401, 602, 1203, 1303 First detector module 402, 606, 1206, 1306 Second detector module 403 Crystal 404 Photo Sensor

Claims

1. acquiring detector element event data associated with radiation events in a plurality of detector elements included in the gamma ray detector ring; identifying assembly events between the detector elements based on the detector element event data; calculating a metric based on each of the assembly events between the detector elements; identifying placement errors of the detector elements based on the metric; A method for identifying placement errors, comprising:

2. the detector element event data includes an accumulated energy of each of the radiation events, a time of energy accumulation of each of the radiation events, and a location of energy accumulation of each of the radiation events; The method of identifying placement errors according to claim 1 .

3. the inter-detector element assembly event is a coincidence event occurring between two or more detector elements within a sub-region of a geometric cut of the gamma ray detector ring; 3. A method for identifying placement errors according to claim 1 or 2.

4. the detector element event data includes an accumulated energy of each of the radiation events, and a sub-region of the geometric cut for each of the assembly events is positioned to center a radiation event having a maximum accumulated energy among the radiation events. The method of identifying placement errors according to claim 3.

5. a sub-area of ​​the geometric cut is larger than the longest dimension of the detector element; 5. A method for identifying placement errors according to claim 3 or 4.

6. the metric is a distance metric that defines the distance between assembly events for each detector element; A method for identifying placement errors according to any one of claims 1 to 5.

7. the metric is a relative correlation score between the detector elements; A method for identifying placement errors according to any one of claims 1 to 5.

8. calculating the relative correlation scores for all detector element layout combinations; identifying a detector element layout based on the relative correlation scores for all of the detector element layout combinations; adjusting one or more look-up tables used to associate each detector element with its physical location based on the detector element layout; The method of claim 7 further comprising:

9. collecting radiation event data in a plurality of time coincidence windows; identifying contributions to the metric from random assembly events based on the collection; removing the contribution from random assembly events from the metric; The method of any one of claims 1 to 8, further comprising:

10. collecting radiation event data from an error-free gamma ray detector ring; removing contributions from random assembly events from the metrics of the gamma ray detector ring based on the radiation event data; The method of any one of claims 1 to 9, further comprising:

11. generating training data by generating an erroneous lookup table used to associate each of the detector elements with a distinct physical location; using the training data to train a machine learning system to identify the placement errors; The method of any one of claims 1 to 10, further comprising:

12. and if the gamma ray detector ring is a Category 1 detector in which the crystal pitch and the photosensor pitch are different, performing a position calibration before acquiring the detector element event data. A method for identifying placement errors according to any one of claims 1 to 11.

13. the relative correlation score is a ratio of the number of coincidence events between a first detector element and a second detector element to the sum of the total number of radiation events at the first detector element and the second detector element; 9. A method for identifying placement errors according to claim 7 or 8.

14. the detector element is a crystal; A method for identifying placement errors according to any one of claims 1 to 13.

15. the detector element is a module; A method for identifying placement errors according to any one of claims 1 to 13.

16. the detector element is a detector unit; A method for identifying placement errors according to any one of claims 1 to 13.

17. acquiring detector element event data associated with radiation events in a plurality of detector elements included in the gamma ray detector ring; identifying assembly events between the detector elements based on the detector element event data; calculating a metric based on the assembly events between the detector elements; Identifying placement errors of the detector elements based on the metric. A gamma ray detector system comprising a processing circuit.

18. the processing circuitry trains a machine learning system that identifies the placement errors using training data generated by generating an erroneous lookup table used to associate each of the detector elements with a distinct physical location.

18. A gamma ray detector system according to claim 17.

19. acquiring detector element event data associated with radiation events in a plurality of detector elements included in the gamma ray detector ring; identifying assembly events between the detector elements based on the detector element event data; calculating a metric based on each of the assembly events between the detector elements; identifying placement errors of the detector elements based on the metric; A placement error identification program that causes a computer to execute the above.

Images