A computer-implemented method for identifying and locating radiation events, and a pixelated radiation detector for performing the method.

Abstract

The present invention relates to a computer-implemented method (200) for radiative event localization of a pixelated radiative detector (10), the pixelated radiative detector (10) comprising at least one scintillator array (24) having scintillator array elements (26) arranged in an (m)×(n) array in a light-sharing mode and an optical sensor array (28) having an optical sensor (30) arranged in a (q)×(z) array and coupled to the scintillator array (24) for determining the spatial intensity distribution of scintillating photons, wherein scintillating light is emitted by the scintillator array (24) in response to an incident radiative event at a light-conversion location. The computer-implemented method (200) includes the following steps: - sampling (72) the spatial intensity distribution of scintillation photons emitted by the scintillator array (24) in response to a plurality of incident radiation events; - performing at least one cluster analysis (76) based on the sampled spatial intensity distribution of the scintillation photons (80) to obtain clusters (84) of radiation events attributed to the scintillator array elements (26), wherein the dimension of the sampled spatial intensity distribution of the scintillation photons corresponds to the (q)×(z) dimension of the optical sensor array (28); and - determining the location of the radiation events based on at least one cluster analysis (76).

Description

Technical Field

[0001] This invention relates to a computer-implemented method for identifying and locating radiation events, and to pixelated radiation detectors, such as gamma-ray detectors for performing the method. The invention also relates to medical imaging apparatuses including pixelated radiation detectors. Background Technology

[0002] Pixelated radiation detectors, also known as imaging detectors, are used in nuclear medicine imaging systems, such as single-photon emission computed tomography (SPECT) and positron emission tomography (PET) imaging systems, as well as in other applications. These imaging devices typically use gamma-ray and X-ray imaging detectors to acquire imaging data, such as gamma-ray or photon imaging data. Such imaging detectors typically acquire a projection of the distribution of radionuclides emitted from the object being imaged (e.g., a patient). Scintillation detectors are commonly used to represent imaging detectors for gamma-ray or X-ray imaging.

[0003] An array of scintillator crystals or scintillator crystals coupled to an array of photosensitive elements forms a scintillator detector. When an object is introduced into the imaging region of the imaging detector and emits photons or charged particles such as electrons, alpha particles, ions, or high-energy photons, the scintillator scintillates in response to the incident particles emitted by the object, i.e., emits flashes. Therefore, the scintillator is configured to emit photons, which are captured by the photosensitive element, i.e., the photodetector, and then read out by dedicated readout electronics.

[0004] Image detectors can rotate around a patient to acquire multiple projections, creating a multidimensional image of the structure of interest or photons transmitted through the object. Similarly, multiple non-rotatable detectors can also be used to acquire multiple projections. These systems are configured to provide information about where and when specific particles are emitted, which medical or other imaging devices can utilize by introducing a substance that emits particles or by otherwise inducing particle emission at a specific region. For example, if a patient is administered a radioactive tracer that emits a particular type of particle (possibly in response to a metabolic response), an image can be generated as a representation of the location where these particles are emitted. Alternatively, gamma-ray detectors can also detect gamma rays emitted by a gamma-ray source and interacting with an object (e.g., a patient) en route to the detector. Image data representing the location and time of capture of scintillating photons can be generated based on the intensity distribution of charge or spatial intensity distribution (also referred to as charge distribution) on the photodetector. In other words, the temporal and spatial location of incident particles in the scintillator can be determined.

[0005] In modern PET scanners, spatial resolution is crucial, and it depends on a variety of other factors, such as the design of the photodetector or scintillator, data processing, the algorithms used, the calibration of different components, material properties (size, mass, etc.), external conditions, or other influences. For example, the spatial resolution achievable with a given detector design depends heavily on the size of the detector elements (e.g., the size of an optical sensor array or scintillator array when using an array).

[0006] The scintillator included in the detector can, for example, comprise a single (monolithic) block, which produces a continuous distribution (light distribution) of scintillating photons emitted in response to an incident particle to be sampled and analyzed. Alternatively, the scintillator can comprise an array of small crystal elements, which results in an inherent spatial resolution of the imaging detector. If the scintillator comprises a higher number of crystal elements in a given region, the position of the incident particle can be determined with greater accuracy because this increases the resolution. However, this necessitates the correct identification of the higher number of crystal elements.

[0007] One method for identifying crystals interacting with incident particles relies on a dedicated photodetector element configured to read out individual scintillator crystal elements in an array of crystal elements. Another method uses light sharing, where several optical sensors read out several crystal elements. To identify individual crystal elements struck by incident particles, the distribution of scintillator light across multiple photodetector elements can then be evaluated. To improve the detection and accurate identification of crystal elements, a light guide, i.e., an optically transparent material, can be used to disperse the scintillator light across several pixels of the optical sensor array. To identify scintillator array elements that have been struck, the distribution of scintillator photons, i.e., scintillator light or scintillator flashes, on the photosensitive elements of the optical sensor array can be analyzed.

[0008] Furthermore, the energy of the incident particles can be determined. However, if light sharing is used instead of reading out each scintillator array element individually, it is generally more difficult to extract the correct parameters (timing, energy, and location of the collision). On the other hand, the required number of photosensitive elements in the photodetector (photodetector pixel) and the complexity of the data acquisition system can be significantly reduced, which may result in lower device costs.

[0009] For example, if each scintillator crystal array element were read out individually, the same number of photosensitive elements in the photodetectors (photodetector pixels) and electronic channels would be required. Using a light-sharing method, the number of required photodetector pixels and electronic channels can be reduced by an order of magnitude.

[0010] However, using light sharing can increase the computational workload and time in gamma-ray detectors using light-sharing crystals because the energy of the incident gamma rays must be extracted from the signal set of all affected photosensitive elements from the photodetector. Additionally, this presents a significant calibration hurdle because the gamma-ray incidence must first be reconstructed from the data before calibration (e.g., separate scintillator-based spectral calibration and timing calibration) can be performed.

[0011] For localization, the most widely used method is anger localization, which involves determining the centroid or center of mass of the distribution. For example, Wei et al. disclosed the influence factor of a two-dimensional localization map on a photomultiplier tube detector block designed using quadrant sharing technology in 2011. Alternatively, Wei et al. disclosed crystal identification in a two-layer offset DOI-PET detector using hierarchical peak tracking based on SVD and mean drift algorithms in 2016. However, this method has drawbacks such as missing signals caused by, for example, the dead time of one or more photosensitive elements. It may also suffer from defects such as air bubbles in the glue or positional shifts that may occur during the fabrication of detector elements (e.g., scintillator crystal elements may be glued together to form an array, and additional light-sharing devices may be glued to the crystal array).

[0012] Alternative methods involve semi-physical modeling and optimization using model parameters for any given incident scintillation event. For example, in Lerche et al.'s 2011 paper, *Maximum Likelihood Based Positioning and Energy Correction for Pixelated Solid-State PET Detectors*, Proceedings of the Nuclear Science Symposium and Medical Imaging Conference, pp. 3027–3029, the authors proposed an alternative method for determining the position of incident gamma rays and extracting the corresponding parameters. This method is based on the maximum likelihood approach. The most probable light transition position in the scintillator array coupled to the photodetector array under light-sharing mode is determined by comparing the resulting light distribution with a predetermined distribution of different light transition positions in the scintillator. The most probable position, corresponding to the most similar light distribution, is used as an estimate of the light transition position in the scintillator of the incident gamma ray. The authors demonstrate that the resolution of medical images can be improved by using the maximum likelihood position estimation method.

[0013] Another solution, as proposed in EP3033636, is a calibration method in which the centroid location and accumulated energy are determined for a set of simultaneously emitted scintillating photons. Cluster analysis is then performed to obtain clusters attributable to gamma-ray events of the scintillator array elements. The accumulated spatial intensity distribution of scintillating photons in the scintillator array in response to emitted gamma rays is determined by accumulating the spatial intensity distribution for the clusters.

[0014] The centroid localization method described in EP3033636 relies on parameterization of a low-dimensional data space, particularly a 2D data space or a 2D data space plus one-dimensional energy parameterization. This has the inconvenience of reducing cluster separation performance, thus negatively impacting the localization accuracy of radiation events.

[0015] Therefore, the object of the present invention is to eliminate or at least mitigate some of the disadvantages mentioned above.

[0016] More specifically, the object of the present invention is to provide a computer-implemented method for pixelated radiation detectors with improved clustering separation performance for more accurate localization of radiation events. Summary of the Invention

[0017] This objective is achieved by means of a computer-implemented method for radiation event localization of a pixelated radiation detector, the pixelated radiation detector comprising at least one scintillator array having scintillator array elements arranged in an (m)×(n) array in a light-sharing mode and an optical sensor array having optical sensors arranged in a (q)×(z) array and coupled to the scintillator array for determining the spatial intensity distribution of scintillating photons. In response to an incident radiation event at a light-conversion location, scintillating photons are emitted by the scintillator array. The computer-implemented method includes the following steps:

[0018] - In response to multiple incident radiation events, the spatial intensity distribution of scintillating photons emitted by the scintillator array is sampled;

[0019] - Perform at least one cluster analysis based on the spatial intensity distribution of the sampled scintillation photons to obtain clusters attributable to radiation events of the scintillator array elements, wherein the dimension of the spatial intensity distribution of the sampled scintillation photons corresponds to the (q)×(z) dimension of the optical sensor array, and

[0020] - The location of radiation events is determined based on at least one cluster analysis.

[0021] In this implementation, cluster analysis is repeated based on clusters obtained through previously performed cluster analysis.

[0022] In the implementation, the first cluster analysis and the second cluster analysis use the same clustering algorithm or different clustering algorithms.

[0023] In implementation, cluster analysis includes the use of standard clustering algorithms.

[0024] In this implementation, the clustering analysis, or each clustering analysis, is based on an unsupervised machine learning clustering algorithm.

[0025] In this implementation, the clustering analysis, or each clustering analysis, is based on a density-based spatial clustering algorithm.

[0026] In this implementation, cluster analysis includes the following steps:

[0027] -Limiting the cluster domain boundary,

[0028] - Parameterize the cluster domain edges.

[0029] - Save the obtained parameters in the calibration data array.

[0030] - Apply the parameters to the spatial intensity distribution of sampled scintillating photons sensed by the optical sensor array (q)×(z); and

[0031] -Based on the previous calibration, data is obtained divided into (m)×(n) domains.

[0032] In one embodiment, the optical sensor is arranged to read out scintillation data from each scintillator element of at least one scintillator array.

[0033] In one implementation, at least one clustering analysis is performed based on light intensity samples to obtain clusters attributable to radiative events of the scintillator array elements. The spatial intensity distribution of scintillator photons in the matrix of each scintillator array element is based on these clusters.

[0034] Another aspect of the invention relates to a pixelated radiation detector for performing the computer-implemented method as described above. The pixelated radiation detector includes a region in which a radiation event may occur, means arranged to detect the radiation event, and a computer operatively connected to the means. The means includes one or more detector module arrays, each array including a plurality of detector modules and a detector module array readout means connected to read the output of each detector module array. The detector module array readout means includes a processing unit for storing and / or processing acquired data.

[0035] In this implementation, each of the detector modules includes a plurality of scintillator units. Each scintillator unit includes a scintillator array having dimensions of (m) × (n), an optical sensor device for detecting light from the scintillator array, and a scintillator unit output interface connected to the detector module readout device.

[0036] In one embodiment, the scintillator array includes scintillator elements. At least one optical sensor of the optical sensor array is associated with two or more scintillator elements. The optical sensor array defines a (q)×(z) array that is related to the size of the scintillator array in a manner where (q)<(m) or (z)<(n), or (q)<(m) and (z)<(n).

[0037] In this implementation, the acquired data includes information about the (q)×(z)-dimensional intensity distribution of photons, the identifier of the optical sensor, and at least one timestamp. The temporally and spatially sampled intensity distribution of scintillating photons can be provided using either the (q)×(z) timestamp or multiple timestamps for each (q)×(z) optical sensor.

[0038] Another aspect of the present invention relates to a medical imaging apparatus comprising a pixelated radiation detector as described above.

[0039] Another aspect of the present invention relates to a non-transitory computer-readable storage medium for storing instructions that, when executed by a processor, perform the method described above. Attached Figure Description

[0040] The invention will be better understood through the description of embodiments given by way of example and illustrated by the accompanying drawings, in which:

[0041] Figure 1 A flowchart illustrating a computer-implemented method for locating radiation events is shown;

[0042] Figure 2 It shows a feedback loop Figure 1 Flowchart of the method;

[0043] Figure 3a A 2D centroid diagram of the clusters obtained through initial cluster analysis is shown;

[0044] Figure 3b It shows that by using the method based on Figure 3a A 2D centroid diagram of clusters obtained from the initial cluster analysis and subsequent cluster analysis.

[0045] Figure 4a A perspective view of a scintillator unit, including an optical sensor array and a scintillator array, is shown.

[0046] Figure 4b yes Figure 4a Side view;

[0047] Figure 4c It shows Figure 4a Top view;

[0048] Figure 4d A top view of the optical sensor array is shown;

[0049] Figure 4e It shows Figure 4d Bottom view of the optical sensor array;

[0050] Figure 4f A perspective view of the scintillator element is shown;

[0051] Figure 5a A side view of the detector module is shown;

[0052] Figure 5b It shows Figure 5a Bottom view of the detector module readout device;

[0053] Figure 5c A top view of the detector module readout device is shown;

[0054] Figure 6a A perspective view of the detector module array is shown;

[0055] Figure 6b A bottom view of a detector module array without a detector module array readout device is shown;

[0056] Figure 6c A top view of the detector module array readout device is shown; and

[0057] Figure 7 A perspective view of a pixelated radiation detector designed for medical use is shown. Detailed Implementation

[0058] Figure 7 An embodiment of a medical scanning setup system 1 is illustrated. System 1 includes a positron emission tomography (PET) scanner, a pixelated radiation detector 10, one or more detector rings 16 of an external imaging region 12, and an imaging device 14. The imaging region 12 is configured to receive the head of a subject resting on a subject support 4. The subject, who has received a radiopharmaceutical injection, begins to emit gamma rays. These gamma photons are detected by the imaging device 14, which has multiple detector module arrays 18 circumferentially mounted on the imaging device 14. Each detector module array 18 includes multiple detector modules 20 mounted on a detector module array readout device 50, such as... Figure 5a and Figures 6a to 6c As shown. The detector module array readout device 50 includes at least one detector module array processing unit 52. In one embodiment, the detector module readout device 50 is a motherboard having at least one integrated processor 52.

[0059] Figure 6aAn example of a detector module array 18 is shown. Regarding... Figure 6b The image shows a rear view without the detector module array readout device. The detector module 20 includes a connector 48 on the rear side, which connects to the detector module array readout device 50, i.e. Figure 6c The motherboard shown has a connector 54. Connector 54 may also provide current and cooling. One or more detector module processing units 46 are represented. In one embodiment, the detector module processing unit 46 is an ASIC chip configured to store and digitize the output data stream. Each detector module 20 includes, as shown... Figure 5a The plurality of scintillator units 22 shown are mounted on detector module 20 and communicatively coupled to detector module readout device 44 via connector 42. Figure 6b Only three detector modules 20 of the array are shown, making the rear side of the multiple scintillator units 22 visible, and connector 42 is shown. Output data from the optical sensor is transmitted via connector 42 to detector module processing unit 46 for further processing.

[0060] This modular design allows for greater versatility in end-design during manufacturing. For example, assembling different types of detectors in different sizes based on the same sub-components provides faster and ultimately more efficient production of medical devices with integrated detector imaging systems.

[0061] Figure 4a An example of a scintillator unit 22 including a scintillator array 24 comprising (m) × (n) scintillator elements 26 is shown. Generally, a scintillator is a material capable of absorbing ionizing radiation such as X-rays or gamma rays, converting a portion of the absorbed energy into visible or ultraviolet photons. The conversion process typically generates short photon pulses corresponding to each radiation event interacting with the scintillator material. The light pulses deposited in the scintillator elements are generated by, for example... Figure 4d The optical sensor 30 within the optical sensor array 28 shown senses and converts the signals into electrical signals. For the detection of X-rays and gamma rays, such as the 511 keV gamma rays used in PET, an inorganic single-crystal scintillator 26 is used because of their typically higher density and atomic number, which results in better detection efficiency. The scintillator element 26 can be, for example, a LYSO crystal scintillator.

[0062] In other embodiments, inorganic scintillator elements such as LSO or BGO crystals can be used for the scintillator unit.

[0063] Scintillators can typically be liquids or solids, organic or inorganic, crystalline or amorphous. Organic liquid and plastic scintillators are commonly used to detect beta particles and fast neutrons. In other embodiments, inorganic materials can be used for the scintillator.

[0064] In one embodiment, the scintillator elements 26 are arranged in a 6×6 array to form a shape such that... Figure 4c The scintillator array 24 is shown. The scintillator array 24 has a oriented imaging region 12 ( Figure 7 The surface of the scintillator array 24 is referred to herein as the top surface. The top surface is not limited to spatial positioning. The surface opposite the top surface of the scintillator array 24 is referred to as the rear surface of the scintillator array. This rear surface faces the optical sensor array 28 having the optical sensor 30, such as, for example... Figure 4b As shown.

[0065] The optical sensors 30 of the optical sensor array 28 are arranged in a light-sharing mode relative to the scintillator array 24. The light-sharing method is widely used to overcome the limitation of one-to-one coupling between optical sensors and scintillator elements.

[0066] In one embodiment, an optical sensor 30 is coupled to four scintillator elements 26. In this embodiment with 6×6, or 36 scintillator elements 26, in a 1:4 light-sharing mode, the optical sensor array 28 is a 3×3 array, i.e., nine optical sensors 30 for 36 scintillator elements.

[0067] In other embodiments, scintillator elements with different shapes, such as triangular shapes, can be used, wherein the optical sensor can cover six scintillator elements in a light-sharing mode. In other embodiments, smaller scintillator elements can be used, making other coupling ratios feasible and within the scope of the invention. For the sake of simplicity and completeness in describing the invention, a 1:4 coupling ratio between the optical sensor 30 and the scintillator elements is described.

[0068] like Figure 4b and Figure 4e As shown, one side of the optical sensor array 28 faces the scintillator element 26, while the other side includes a scintillator unit output interface 40, which includes a connector 42 for data transmission of readout data from the optical sensor. The connector 42 may also provide current and cooling means.

[0069] The optical sensor array 28 is any suitable photodetector, such as a photomultiplier tube (PM), a microchannel plate photomultiplier tube (MCPT), or a silicon-based photomultiplier tube (SiPM) in this exemplary embodiment.

[0070] Optionally, the processing unit may be additionally mounted on the scintillator unit output interface 40. This processing unit at this stage will allow for preprocessing of the acquired sensor data. For example, in embodiments with a high coupling ratio between the scintillator element and an optical sensor, preprocessing of the acquired sensing data via the optical sensor may be necessary.

[0071] An optical sensor array 28, which is in a light-sharing mode with the scintillator array 24, is fixed to the scintillator crystal array 28 directly or via a light guide 32 arranged between the optical sensor array 28 and the scintillator array 24. Figure 4b As shown. The light guide 32 can be, for example, air, glass, acrylic glass, sapphire, or a suitable adhesive with light-guiding properties.

[0072] In one embodiment, one or more scintillator units 22 are releasably mounted on the detector module readout device 44, such as Figure 5a As shown. The detector module readout device 44 has an output interface 40 facing the scintillator unit with connector 42. Figure 4e On one side of the detector module readout device 44, such as... Figure 5c As shown, connector 48 is coupled to a corresponding connector 42 of scintillator unit 22. Connector 48 provides data transmission of readout data from scintillator unit 22. Connector 48 may also provide current and cooling means.

[0073] In one embodiment, the connector 42 of the output interface 40 of each of the four scintillator units 22 is inserted into the corresponding connector 48 of the detector module readout device 44.

[0074] In an exemplary embodiment, six detector modules 20 are mounted on the motherboard 50 to form a detector module array 18, such as... Figures 6a to 6c As shown.

[0075] In one embodiment, one or more of these detector module arrays 18 are mounted on the detector ring 16 of the imaging device 14 of the pixelated radiation detector 10 of the medical scanning setup system 1, such as Figure 7 As shown. The detector module array readout device 50 includes connectors and mounting means for connecting the module array 18 to an imaging device and an external processing unit such as a computer 60.

[0076] The modular structure of the scintillator unit 20 can be regarded as a single building block, requiring only adjustments to the detector module array readout device 50 and the scintillator unit output interface 40, i.e., if the detector module size or detector module array size needs to be adapted to different detector architectures.

[0077] During a radiation event, light pulses deposited in the scintillator element 26 are sensed by the optical sensor 30 within the optical sensor array 28 and converted into electrical signals. Multiple radiation events converted into electrical signals form readout data processed by the method of the present invention.

[0078] Figure 1An exemplary flowchart of a computer-implemented method 200 is shown, which processes readout data to locate radiation events in a particular scintillator element based on electrical signal samples obtained from optical sensors 30 within an optical sensor array 28.

[0079] Data captured by the optical sensor 30 within the optical sensor array, corresponding to radiation events in a single scintillator element 26 within the corresponding scintillator array 24, is transmitted via connector 42 to the corresponding detector module processing unit 46.

[0080] The detector module processing unit 46 samples 72 the spatial intensity distribution of scintillating photons emitted by the scintillator array from the acquired data 80. In one embodiment, the sampling corresponds to recording a portion of the photon intensity distribution at a selected location, which is designed to maximize optical detection efficiency, maximize optical sharing separation power (i.e., the distance between data clusters), maximize the system's time-of-flight (ToF) capability, or follow certain symmetries to simplify post-processing, or other methods for selecting a good location. Sampling and recording data on each detector module processing unit 46 allows data to avoid accumulation on a centralized data processing unit. The digitization process of the sensed analog data can advantageously be performed at the detector module level.

[0081] In one implementation, the calibration step includes performing cluster analysis 76 on the sampled data to obtain a first set of clusters. Examples of these obtained clusters are shown in... Figure 3a As shown in the image.

[0082] The dimension of the sampled data is (q) × (z) – the dimension of the optical sensor array. The analysis includes cluster identification through supervised, semi-supervised, or unsupervised machine learning. This first clustering analysis provides location information for radiation events to distinguish the optical sensors sensing these events. These clusters also represent the dimension of the optical array 28. In an exemplary embodiment, a 3 × 3 optical array is used, i.e., 9 clusters.

[0083] Compared to "marginalizing" 2D / (2+1)D or other parameterized data spaces, the advantage of using a high (q)×(z) dimensional data space lies in preserving "more volume" in higher dimensions. This is a mathematical property of high-dimensional spaces. The greater volume between data points facilitates their separation.

[0084] Furthermore, depending on the supervised, semi-supervised, or unsupervised machine learning algorithm employed, the "cluster domain edges" are defined using centroids, decision trees, or other methods. The format of the domain edge representation largely depends on the method used to determine the clusters. The domain edges are parameterized, and these parameters are stored in the data array, specifically the calibration data array 84.

[0085] In one implementation, a second clustering analysis 78 is performed on the previously clustered data obtained through clustering analysis 76 in the first calibration step to obtain further differentiation of radiation events. Clustering analysis 78 includes identifying other clusters among the clusters obtained in the analysis of step 76. Figure 3b The clusters in the clustering analysis are shown, providing differentiation of radiative events at the level of each scintillator crystal 26. Then, in the second step, the calibration data array 84, containing the parameterized domain edges, is directly applied to the fast (q) × (z) dimensional data 86. Based on the previous calibration, this divides the data 94 into (m) × (n) domains. Therefore, the resulting cluster data 86 has a total dimension of (m) × (n), or 36 in the example embodiment, corresponding to the dimension of the scintillator array 28.

[0086] In one implementation, the clustering analysis of step 76 is performed on one or more external data processing units, such as external computers, servers, or the cloud. The resulting calibration data array 84 can be stored on the detector module array processing unit 52. Therefore, the clustering analysis step 78 can also be performed at the detector module level within the detector module processing unit 52.

[0087] Figure 2 Alternative implementations are shown in which n cluster analyses are performed on the obtained data. These additional steps can improve the accuracy of cluster separation and thus achieve higher accuracy in locating data associated with radiation events. The same or different types of clustering algorithms can be used.

[0088] The clustering analysis in step 76 and / or in step 78 can be based on machine learning algorithms. While many algorithms may be suitable for computer-implemented methods of radiological event localization, density-based and / or hierarchical clustering methods have proven to be more robust while providing better results.

[0089] In one implementation, cluster analysis 76 and optional second cluster analysis 78 are based on a density-based spatial clustering (DBSCAN) algorithm with noisy applications.

[0090] List of reference numerals

[0091] Medical Scan Setup System 1

[0092] Subject support 4

[0093] Pixelated radiation detector 10

[0094] Imaging area 12

[0095] Imaging device 14

[0096] Detector ring 16

[0097] Detector module array 18

[0098] Detector Module 20

[0099] Scintillator unit 22

[0100] Scintillator array 24

[0101] Scintillator element 26

[0102] crystal

[0103] Optical sensor array 28

[0104] Optical sensor 30

[0105] 32-channel optical fiber

[0106] Glass

[0107] air gap

[0108] glue

[0109] Scintillator unit output interface 40

[0110] Connector 42

[0111] Detector module readout device 44

[0112] Detector module processing unit 46

[0113] Connector 48

[0114] Detector module array readout device 50

[0115] Detector module array processing unit 52

[0116] Connector 54

[0117] Computer 60

Claims

1. A computer-implemented method (200) for radiation event localization for a pixelated radiation detector (10) comprising at least one scintillator array (24) with scintillator array elements (26) arranged in an (m) x (n) array in a light sharing mode and an optical sensor array (28) with optical sensors (30) arranged in a (q) x (z) array and coupled to the scintillator array (24) for determining a spatial intensity distribution of scintillating photons, wherein, In response to an incident radiation event at a light conversion location, the scintillation photons are emitted by the scintillator array (24), wherein the computer-implemented method (200) comprises the step of sampling (72) a spatial intensity distribution of the scintillation photons emitted by the scintillator array (24) in response to a plurality of incident radiation events, characterized in that the computer-implemented method (200) further comprises the steps of: - performing at least one cluster analysis (76) on the sampled spatial intensity distribution of scintillation photons (80) to obtain clusters (84) of radiation events attributed to scintillator array elements (26), wherein the dimensionality of the sampled spatial intensity distribution of scintillation photons corresponds to the (q) x (z) dimensionality of the optical sensor array (28), and - determining a localization of the radiation events based on the at least one cluster analysis (76).

2. The computer-implemented method (200) of claim 1, wherein, The first cluster analysis and the second cluster analysis (78) are performed, and wherein the second cluster analysis is based on the clusters (84) obtained by the previously performed first cluster analysis.

3. The computer-implemented method (200) of claim 2, wherein, The first cluster analysis and the second cluster analysis (78) use the same cluster algorithm or different cluster algorithms.

4. The computer-implemented method (200) according to any one of claims 1 to 3, wherein, Each cluster analysis of the at least one cluster analysis comprises using a standard cluster algorithm.

5. The computer-implemented method (200) according to any one of claims 1 to 3, wherein, Each cluster analysis of the at least one cluster analysis is based on a supervised, semi-supervised or unsupervised machine learning cluster algorithm.

6. The computer-implemented method (200) according to any one of claims 1 to 3, wherein, Each cluster analysis of the at least one cluster analysis is based on a density-based spatial cluster algorithm.

7. The computer-implemented method (200) according to any one of claims 1 to 3, wherein, Each cluster analysis of the at least one cluster analysis comprises the steps of: - defining cluster domain edges, - parameterizing the cluster domain edges, - saving the obtained parameters in a calibration data array, - applying the parameters to the sampled spatial intensity distribution of scintillation photons sensed by the optical sensor array (q) x (z), and - obtaining data divided into (m) x (n) domains from previous calibrations.

8. The computer-implemented method (200) according to any one of claims 1 to 3, wherein, The optical sensor (30) is arranged to read out scintillation data from each scintillator array element (26) of the at least one scintillator array (24).

9. The computer-implemented method of claim 8, wherein, The at least one cluster analysis (76) is based on light intensity samples to obtain clusters of radiation events attributed to scintillator array elements (26), and wherein a matrix (94) of the spatial intensity distribution of scintillation photons for each scintillator array element (26) is based on the clusters.

10. A pixelated radiation detector (10) for performing the computer-implemented method (200) according to any one of claims 1 to 9, the pixelated radiation detector (10) comprising an imaging area (12) in which radiation events can occur, an imaging device (14) arranged to detect radiation events, and a computer (60) operably connected to the imaging device (14), wherein, The imaging device (14) comprises one or more detector module arrays (18), each comprising a number of detector modules (20) and a detector module array readout device (50) connected to read the output of each detector module array (18), wherein the detector module array readout device (50) comprises a processing unit (52) for storing and / or processing acquisition data.

11. The pixelated radiation detector (10) according to claim 10, wherein Each of the detector modules (20) comprises a number of scintillator units (22), each scintillator unit (22) comprising a scintillator array (24) having dimensions (m) x (n), an array of optical sensors (28) for detecting light from the scintillator array (24), and a scintillator unit output interface (40) connected to a detector module readout (44).

12. The pixelated radiation detector (10) according to claim 11, wherein The scintillator array (24) comprises scintillator array elements (26), wherein at least one optical sensor (30) of the array of optical sensors (28) is associated with two or more scintillator array elements (26), wherein the array of optical sensors (28) defines a (q) x (z) array that is related to the size of the scintillator array (24) in a way that (q) < (m) or (z) < (n), or (q) < (m) and (z) < (n).

13. The pixelated radiation detector (10) according to claim 12, wherein The acquisition data comprises information on a (q) x (z) dimensional intensity distribution of the photons, an identifier of the optical sensor (30), and at least one time stamp, providing an intensity distribution of the time and spatially sampled scintillating photons.

14. A medical imaging apparatus (8) comprising a pixelated radiation detector (10) according to any one of claims 10 to 13.

15. A non-transitory computer readable storage medium (46, 52, 62) for storing instructions which, when executed by a processor, perform the method (200) according to any one of claims 1 to 9.

Images