PET imaging method based on any detector gap structure
By expanding crystal bars in the PET system to fill gaps and using Monte Carlo simulation to generate a complete system matrix, the data loss problem caused by axial sparse structure is solved, the reconstruction accuracy and adaptability are improved, and consistency with the actual detection process is achieved. .
Patent Information
- Application Number
- CN202410914817.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-09
- Publication Date
- 2024-11-01
AI Technical Summary
In PET systems, gaps caused by axial sparse structures lead to data loss. The prior art makes up for by interpolation, but the accuracy is not high, and limited to sinogram data, making it difficult to realize real-time reconstruction based on list-mode data.
By expanding crystal bars to fill the gap, a complete system matrix is pre-generated, and combined with the judgment of the physical position of the gap, the emission situation is simulated, and the complete emission situation of axial single-layer image pixels is generated. The Monte Carlo simulation tool GATE is used for simulation to generate a complete system matrix , determine whether the LOR is located on the gap during reconstruction, and if so, it will not participate in the calculation.
It improves the accuracy of reconstructed images, reduces the computing resource requirements, adapts to the reconstruction needs of multiple mainstream PET data formats, and maintains consistency with the actual detection process.
Smart Images

Figure CN118873159A_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the technical field of PET imaging, and relates to a PET imaging method based on an arbitrary detector gap structure. Background Art
[0002] In the PET structural design, in order to achieve the expansion flexibility of the system and improve the packaging efficiency, a modular detector design is adopted, and there are gaps between modules, such as Figure 1 Furthermore, in the design of PET detectors, any detector gap structure design can be performed in the radial space and axial space of the detector. Therefore, the present invention focuses on a PET imaging system with any detector gap structure.
[0003] Positron emission tomography (PET) is the most sensitive imaging method for recording the molecular interactions of biomarkers, biochemical substances, and drugs in living organisms. The current development goal of PET is to achieve ultra-high resolution and ultra-high sensitivity, so as to meet the needs of precise imaging and dynamic imaging in clinical medicine. High-sensitivity long-axis PET systems have been a research hotspot in the PET field in recent years. This type of system improves sensitivity by increasing the axial field of view, which improves imaging quality while reducing injection dose and scanning time; and long-axis PET has a large axial coverage area and can perform simultaneous imaging of multiple organs throughout the body, which is of great significance for clinical applications.
[0004] In order to facilitate system expansion and reduce the manufacturing cost of long-axis systems, the detector is usually designed in a modular manner and the interval between modules is increased during assembly to increase the overall axial length, forming an axially sparse system structure. The PET system detects a pair of back-to-back gamma photons produced by annihilation based on the line of response (LOR). Photons that fall in the gap between modules cannot be detected, such as Figure 2 As shown in the figure, data is missing. Directly using the missing data for image reconstruction will also cause corresponding gaps in the axial direction of the image, affecting the imaging results. Therefore, in the axially sparse PET system, the deformation of the reconstructed image caused by data missing must be solved.
[0005] Most of the existing solutions for dealing with missing axial sparse PET data are based on interpolation of sinogram-type data. Sinogram data is obtained by accumulating LOR according to the angle, distance and number of layers to the center of the system, and is one of the basic data formats of PET. For missing detection data caused by axial gaps, the actual detector module is usually divided into layers according to the physical size of the gap. The sinogram data of the layer corresponding to the gap is interpolated from the data of the surrounding layers, thereby compensating for the missing detection data to a certain extent, and reconstructing it using the interpolated sinogram data.
[0006] Although the existing technology compensates for the missing data to a certain extent through interpolation, this method has two disadvantages. First, the compensated data is obtained by interpolation of the surrounding layer data, which is not accurate and cannot accurately reflect the real response of the system during the detection process, so the gaps in the image are still visible; second, this method is based on sinogram data, so reconstruction can only be based on sinogram data, which is not conducive to real-time reconstruction based on list-mode data, and its application has certain limitations. Summary of the invention
[0007] In view of the problems existing in the prior art, the purpose of the present invention is to provide a PET imaging method based on an arbitrary detector gap structure. The present invention will pre-generate a complete system matrix after expanding the gap and filling it with crystal strips to make up for the data loss; then judge the median of the complete system matrix in combination with the physical position of the gap, and set the value falling on the gap to zero, so as to keep consistent with the actual detection process and improve accuracy. In addition, the system matrix can be generated according to the reconstruction data format, so that the reconstruction process is not restricted by the data type.
[0008] The present invention fills the axial gap with an extended crystal strip to obtain a new axial structure; based on the new axial structure, the Monte Carlo simulation tool GATE is used to simulate the complete emission conditions corresponding to the pixels of the axial single-layer image, thereby generating a complete system matrix; during reconstruction, the system matrix corresponding to the pixels outside the simulation area is translated into the simulation area by translating the LOR of the response line to be tested according to the pixel axial index difference, and then the corresponding simulated response line LOR result is searched, and the response probability value in the complete system matrix corresponding to the simulated response line LOR is assigned to the LOR to be tested; during reconstruction, it is determined whether each LOR is located on the gap in the actual geometric structure, and if it is located on the gap, it will not be calculated, thereby maintaining consistency with the actual detection process.
[0009] The present invention simulates the emission of axial single-layer image pixels based on the Monte Carlo simulation tool GATE. During the simulation, the image pixel size corresponds to half of the axial size of the crystal bar, and the pixel is located on the left or right half of the crystal bar. The simulation data is statistically analyzed and organized into a system matrix according to the pixel drive or ray drive method.
[0010] For the detected LOR of the pixel in the non-simulated area, its LOR is translated according to the axial pixel index difference between the pixel and the simulated pixel, and the system matrix value of the LOR corresponding to it in the simulation result is found, which is the system matrix value of the pixel in the non-simulated area and its corresponding LOR.
[0011] When iterative reconstruction is performed based on the virtual new structure, whether the LOR falls on the seam in the real physical space is judged according to formula (9). If the condition is met, the LOR does not participate in the calculation during reconstruction.
[0012] The technical solution of the present invention is:
[0013] A PET imaging method based on an arbitrary detector gap structure, the steps comprising:
[0014] 1) Modeling the selected PET system to obtain a model corresponding to the PET system; in the model, the gaps in the axial and cross-sectional directions of each detector module in the PET system are filled with extended crystal strips as the complete detector structure of the model; performing simulation detection based on the complete detector structure of the model, and generating a complete system matrix of the PET system according to the simulation results;
[0015] 2) An object to be detected is placed in the PET system for detection, and then the detection data of the object to be detected is processed based on the complete system matrix of the PET system to reconstruct the PET image of the object to be detected.
[0016] Furthermore, the method for generating the complete system matrix of the PET system is:
[0017] 11) The PET system includes NumDet arranged along the axial direction Axial detector module rings, each of which consists of NumDet trans detector modules; each of the detector modules includes a NumCry*NumCry crystal bar array, wherein the axial size of the crystal bar is SizeAxial, and the PET system includes (NumDet Axial -1) axial gap, the length of the axial gap along the axial direction is SizeGap; the number of crystal strips along the axial direction in the PET system = the number of crystal strip rings NumRing = NumDet Axial*NumCry; in the model, each of the gaps is filled with an extended crystal bar according to the detector crystal material and crystal bar size, and the number of axial crystal bar rings is NumRing ′ The axial structure; the number of extended crystals contained in each of the axial gaps The number of crystal rings along the axial direction in the model is NumRing ′ =NumRing+(NumDet Axial -1)*NumCry gap ;
[0018] 12) The number of crystal rings in the axial direction is NumRing ′ A simulated detection is performed under the axial structure of the PET system, and then a complete system matrix of the PET system is generated according to the simulated detection results.
[0019] Further, in step 12), first, the number of axial crystal rings is NumRing ′ The emission conditions of the point source corresponding to each pixel in the axial single-layer image are simulated under the axial structure to obtain the simulated detection data corresponding to each pixel in the single-layer image, and then the complete system matrix of the PET system is generated according to the simulated detection data of each pixel.
[0020] Furthermore, a complete system matrix of the PET system is generated by pixel driving or ray driving according to the simulated detection data.
[0021] Furthermore, the method for reconstructing the PET image of the object to be tested is:
[0022] 21) Based on the axial translation invariance and symmetry of the PET system, for each pixel b located at the non-simulated crystal strip position in the detection data of the object to be tested, first determine whether the axial position of pixel b is located on the left half or right half of the crystal strip as the virtual pixel. If not on the same side, first mirror the LOR emitted by pixel b. Suppose the crystal strip ring number at one end of any LOR is zIndex0, then the new crystal strip ring number after mirroring is NumRing ′ -zIndex0; then translate the LOR according to the crystal ring number, and find the corresponding complete system matrix value in the simulation result; then calculate the difference Δindex between the pixel b and the simulated pixel axial index, and the sign of Δindex represents the direction of translation; then translate the LOR corresponding to the pixel b according to the difference Δindex, and the translated crystal ring number is zIndex ′ =NumRing ′-zIndex0+Δindex; if on the same side, directly translate pixel b according to the difference Δindex, and the translated crystal ring number is zIndex ′ =zIndex0+Δindex;
[0023] 22) Find the LOR that is consistent with the number of the crystal strip ring after the above translation in the simulation detection results of the simulation pixel as the system matrix value of the interaction of the LOR in the pixel b;
[0024] 23) Reconstructing the PET image of the object to be tested according to the system matrix value of each pixel obtained in step 22).
[0025] Further, the PET image of the object to be tested is reconstructed using an iterative reconstruction algorithm according to the system matrix value of each pixel obtained in step 22), and the specific method is as follows:
[0026] 231) In each iteration process, forward projection and back projection operations are performed; the formula for forward projection is Where I is the total number of pixels of the reconstructed image, v i is the estimated pixel value, q j is the estimated value of the j-th LOR count, a ij It represents the probability that the ray generated by the i-th radiation source recorded in the complete system matrix is detected by the j-th response line. The formula for back projection is: Where, J is the sum of the number of LORs;
[0027] 232) Calculate the number of any crystal strip RingIndex in the axial t-th detector module in the axial structure of the model ′ =RingIndex+(t-1)*NumCry;t=1,2,3,...,ND A ; RingIndex is the original ring number of the crystal strip in the PET system, RingIndex ′ Satisfy(RingIndex ′ +1)%(NumCry+NumCry gap )=0, it is considered to fall in the number RingIndex ′ If the LOR of a crystal strip hits the axial gap in real space, the contribution of the LOR of this strip to the current pixel is not considered.
[0028] Furthermore, based on the axial structure of the model, the Monte Carlo simulation tool GATE is used to perform simulated detection to obtain the complete system matrix corresponding to the axial single-layer image pixels, and then the complete system matrix of the PET system is generated according to the complete system matrix corresponding to the axial image pixels of each layer.
[0029] The advantages of the present invention are as follows:
[0030] The present invention expands and fills the axial gap in the PET system, and based on this structure, uses Monte Carlo simulation to pre-generate a complete system matrix, which includes all emission conditions of the radiation source and makes up for the data loss; the present invention only simulates and generates the system matrix of one layer of image pixels, and the system matrices of the remaining pixels are obtained by translation and mirroring, which greatly reduces the demand for computer resources for the pre-generated system matrix; in the present invention, the simulated data can be organized into a system matrix according to ray drive or pixel drive according to demand, so as to adapt to the reconstruction requirements of various PET mainstream data formats; when calling the system matrix in reconstruction, the present invention combines the physical position of the gap to judge the LOR in the complete system matrix, and the LOR falling on the gap does not participate in the calculation, so as to keep consistent with the detector response of the actual detection process and improve the reconstruction accuracy. BRIEF DESCRIPTION OF THE DRAWINGS
[0031] Figure 1 Schematic diagram of the axial gap of PET.
[0032] Figure 2 Schematic diagram of a photon falling on the slit.
[0033] Figure 3 The flowchart of the PET detector structure imaging method based on arbitrary axial gap.
[0034] Figure 4 Schematic diagram of axial simulation.
[0035] Figure 5 Obtain a flow chart for any corresponding system moment.
[0036] Figure 6 The results are directly reconstructed for the rod source without considering the axial gap.
[0037] Figure 7 This is the rod source reconstruction result using the present invention. DETAILED DESCRIPTION
[0038] The present invention is further described in detail below in conjunction with the accompanying drawings. The examples given are only used to explain the present invention but not to limit the scope of the present invention.
[0039] The present invention proposes a PET detector structure imaging method based on arbitrary axial gaps. This method virtually regards the gaps as normal crystal strips to form a new system structure. Based on the new structure, a complete system matrix corresponding to one axial layer of image pixels is generated. The corresponding system matrices at other pixel positions are obtained from simulation results using translational symmetry. During reconstruction calls, it is determined whether the current LOR falls on the gap in the real physical space. If so, it is set to 0 to achieve consistency with the detector response. The overall flow chart of this method is shown in the figure below. Figure 3shown.
[0040] 1. Virtual compensation of axial gap
[0041] Assume that the PET system consists of NumDet in the loop trans detector modules, the axial direction is determined by NumDet Axial detector modules, each of which is composed of a NumCry*NumCry array of crystal bars (the array of crystal bars of the detector module is two-dimensionally distributed, with NumCry crystal bars in the axial and radial directions), the axial size of the crystal bar is SizeAxial, the axial gap size of each detector module is SizeGap, and the axial size is (NumDet Axial -1) gaps. The number of axial crystal bars in the system, that is, the number of rings, is NumRing:
[0042] NumRing=NumDet Axial *NumCry (1)
[0043] like Figure 4 As shown, the axial gaps between the detector modules are virtually converted into extended crystal strips according to the axial size of the crystal strips. The number of extended crystal strips corresponding to each gap is:
[0044]
[0045] In the new system structure formed after the conversion, the number of system rings increases to:
[0046] NumRing ′ =NumRing+(NumDet Axial -1)*NumCry gap (3)
[0047] 2. Generation of a complete system matrix based on a fully populated detector structure
[0048] After virtually filling the axial gap of the detector module with an extended crystal strip, a ring number of NumRing is formed. ′ In the new structure, the crystal strips cover the entire axial field of view, and all gamma photons emitted by the radiation source can be detected, thus making up for the lack of data and generating a complete system matrix. Here, the Monte Carlo simulation tool GATE is used to simulate the detection under the new structure, and a complete system matrix is generated based on the simulation results.
[0049] Element a in the system matrix ijIt represents the probability that the ray generated by the i-th radiation source is detected by the j-th line of response. Each pixel in the reconstructed image represents a radiation source, and the purpose of image reconstruction is to obtain the intensity of the radiation source at each pixel. Since the system matrix in this scheme is pre-generated, in order to reduce the demand for computer storage resources, only the emission of the point source corresponding to one layer of image pixels in the axial direction is simulated, such as Figure 4 shown.
[0050] During reconstruction, the pixel size of the axial image is usually set to half the size of the crystal strip. Select the crystal strip at the center of the axial field of view and set the axial position of the simulated pixel to half of the crystal strip. Figure 4 The analog pixel is placed on the right half of the crystal strip. The center of the point source corresponding to the analog pixel is the pixel center, and the diameter of the point source is the side length of the pixel.
[0051] There are two main ways to convert simulation data into system matrix: pixel drive and ray drive, which correspond to sinogram data reconstruction and listmode data reconstruction respectively. When organizing the system matrix by pixel drive, statistics are performed pixel by pixel. Assuming the activity of the simulated point source is A, the total count that can be generated in the simulation time t is A*t, and the count c on the jth LOR corresponding to the i-th pixel is counted LOR , then the corresponding system matrix value is c LOR / At; when the organization is driven by rays, the statistics are performed one by one in LOR. Let the total count of all point source rays detected by the jth LOR be c Total , where the number of rays emitted by the i-th point source detected by the LOR is c pixel , then the system matrix value is c pixel / c Total .
[0052] 3. Obtaining the system matrix at non-analog pixels
[0053] An object to be tested is placed in the PET system for detection, and then the detection data of the object to be tested is processed based on the complete system matrix of the PET system to reconstruct a PET image of the object to be tested. Since the simulation results correspond to a single layer of pixels, all point sources must be involved in reconstructing the PET image. Based on the axial translation invariance and symmetry of the system, the emission of pixels at the non-simulated crystal bar position is obtained from the simulation results. For a certain pixel b at the non-simulated crystal bar position, first determine whether the axial position of pixel b is located on the left half or the right half of the crystal bar. Since the simulated pixel is located on the right half, when pixel b is located on the left half of the crystal bar, the LOR emitted by pixel b must first be mirrored. Suppose the crystal ring number at one end of any LOR is zIndex0, then the new ring number after mirroring is NumRing ′-zIndex0; the crystal ring number at the other end is obtained by the same operation.
[0054] Next, the LOR is translated according to the ring number, and the corresponding system matrix value is found in the simulation results. The difference Δindex between the axial index of pixel b and the simulated pixel is calculated, and the sign of Δindex represents the direction of translation. The LOR to be measured corresponding to pixel b is translated according to the difference. If pixel b is located on the left half of the crystal bar, the ring number after translation is:
[0055] zIndex ′ =NumRing ′ -zIndex0+Δindex (4)
[0056] If pixel b is located on the right half of the crystal strip, the ring number after translation is:
[0057] zIndex ′ =zIndex0+Δindex (5)
[0058] In the simulation detection results of the simulation pixel, the LOR with the same ring number after the above translation is found, and the system matrix value of the search result is the system matrix value of the interaction of the LOR in the b pixel. Figure 5 shown.
[0059] 4. Determination of the LOR of the real gap during PET image reconstruction
[0060] The above step 2 obtains the complete system matrix after the gap is filled with the extended crystal strip, and the complete system matrix based on the new system is reconstructed. However, there are gaps in the actual system. In order to make the reconstruction process consistent with the actual detection process, the LOR falling on the gap needs to be judged during reconstruction.
[0061] PET systems often use iterative reconstruction algorithms, and forward projection and back projection operations are required during the iteration process. The formula corresponding to the forward projection is as follows:
[0062]
[0063] Where I is the total number of pixels of the reconstructed image, v i is the estimated pixel value, q j is the estimated value of the j-th LOR count. The back-projection process is the opposite:
[0064]
[0065] Where J is the sum of the number of LORs. During reconstruction, the contribution of the LORs falling on the gaps should be discarded to be consistent with the actual detection process. In the new structure where the gaps are filled with extended crystal strips, the extended crystal strips are also numbered in the axial crystal array. Then, the number of any crystal strip in the t-th detector module in the axial direction in the new structure becomes:
[0066] RingIndex ′ =RingIndex+(t-1)*NumCry t=(1,2,3,...,ND A ) (8)
[0067] RingIndex is the original ring number of the crystal bar. When it is converted to the virtual new structure, RingIndex ′ When the following conditions are met, it can be considered to fall into the number RingIndex ′ The LOR of the crystal strip hitting the gap in real space:
[0068] (RingIndex ′ +1)%(NumCry+NumCry gap ) = 0 (9)
[0069] If the above conditions are met, the contribution of the LOR to the current pixel is not considered, so that the reconstruction process is consistent with the detection process when there is actually a gap.
[0070] The above method can significantly improve the axial distribution characteristics of the reconstructed image. Figure 6 , Figure 7 These are the results of column source sagittal plane reconstruction before and after the application of this method. It can be seen that the axial inhomogeneity in the image has been significantly improved.
[0071] In this scheme, GATE is used for Monte Carlo simulation, and GEANT4 or other simulation tools can also be used for simulation. In this scheme, the method for obtaining the system matrix can also be obtained by using the method of the complete system matrix, such as the experimental method, or the "analysis + simulation" method. The translational symmetry and rotational symmetry in this scheme can be selected with different degrees of symmetry. The axial gap and radial gap processing principles in this scheme are similar, and non-zero data are used instead of the detected data of 0 to compensate for the real gap detection process.
[0072] Although the specific embodiments of the present invention are disclosed for the purpose of illustration, the purpose is to help understand the content of the present invention and implement it accordingly, those skilled in the art will understand that various substitutions, changes and modifications are possible without departing from the spirit and scope of the present invention and the appended claims. Therefore, the present invention should not be limited to the content disclosed in the best embodiment, and the scope of the present invention is subject to the scope defined in the claims.
Claims
1. A PET imaging method based on an arbitrary detector gap structure, comprising the steps of: 1) Modeling the selected PET system to obtain a model corresponding to the PET system; in the model, the gaps in the axial and cross-sectional directions of each detector module in the PET system are filled with extended crystal strips as the complete detector structure of the model; performing simulation detection based on the complete detector structure of the model, and generating a complete system matrix of the PET system according to the simulation results; 2) An object to be detected is placed in the PET system for detection, and then the detection data of the object to be detected is processed based on the complete system matrix of the PET system to reconstruct the PET image of the object to be detected.
2. The method according to claim 1, characterized in that The method for generating the complete system matrix of the PET system is: 11) The PET system includes NumDet arranged along the axial direction Axial detector module rings, each of which consists of NumDet trans detector modules; each of the detector modules includes a NumCry*NumCry crystal bar array, wherein the axial size of the crystal bar is SizeAxial, and the PET system includes (NumDet Axial -1) axial gap, the length of the axial gap along the axial direction is SizeGap; the number of crystal strips along the axial direction in the PET system = the number of crystal strip rings NumRing = NumDet Axial *NumCry; in the model, each of the gaps is filled with an extended crystal bar according to the detector crystal material and crystal bar size, and the number of axial crystal bar rings is NumRing ′ The axial structure; the number of extended crystals contained in each of the axial gaps The number of crystal rings along the axial direction in the model is NumRing ′ =NumRing+(NumDet Axial -1)*NumCry gap ; 12) The number of crystal rings in the axial direction is NumRing ′ A simulated detection is performed under the axial structure of the PET system, and then a complete system matrix of the PET system is generated according to the simulated detection results.
3. The method according to claim 2, characterized in that In step 12), first, the number of axial crystal rings is NumRing ′ The emission conditions of the point source corresponding to each pixel in the axial single-layer image are simulated under the axial structure to obtain the simulated detection data corresponding to each pixel in the single-layer image, and then the complete system matrix of the PET system is generated according to the simulated detection data of each pixel.
4. The method according to claim 1, 2 or 3, characterized in that: A complete system matrix of the PET system is generated by pixel driving or ray driving according to the simulated detection data.
5. The method according to claim 1, 2 or 3, characterized in that: The method for reconstructing the PET image of the object to be tested is: 21) Based on the axial translation invariance and symmetry of the PET system, for each pixel b located at the non-simulated crystal strip position in the detection data of the object to be tested, first determine whether the axial position of pixel b is located on the left half or right half of the crystal strip as the virtual pixel. If not on the same side, first mirror the LOR emitted by pixel b. Suppose the crystal strip ring number at one end of any LOR is zIndex0, then the new crystal strip ring number after mirroring is NumRing ′ -zIndex0; then translate the LOR according to the crystal ring number, and find the corresponding complete system matrix value in the simulation result; then calculate the difference Δindex between the pixel b and the simulated pixel axial index, and the sign of Δindex represents the direction of translation; then translate the LOR corresponding to the pixel b according to the difference Δindex, and the translated crystal ring number is zIndex ′ =NumRing ′ -zIndex0+Δindex; if on the same side, directly translate pixel b according to the difference Δindex, and the translated crystal ring number is zIndex ′ =zIndex0+Δindex; 22) Find the LOR that is consistent with the number of the crystal strip ring after the above translation in the simulation detection results of the simulation pixel as the system matrix value of the interaction of the LOR in the pixel b; 23) Reconstructing the PET image of the object to be tested according to the system matrix value of each pixel obtained in step 22).
6. The method according to claim 5, characterized in that The PET image of the object to be tested is reconstructed using an iterative reconstruction algorithm according to the system matrix value of each pixel obtained in step 22), and the specific method is as follows: 231) In each iteration process, forward projection and back projection operations are performed; the formula for forward projection is Where I is the total number of pixels of the reconstructed image, v i is the estimated pixel value, q j is the estimated value of the j-th LOR count, a ij It represents the probability that the ray generated by the i-th radiation source recorded in the complete system matrix is detected by the j-th response line. The formula for back projection is: Where, J is the sum of the number of LORs; 232) Calculate the number of any crystal strip RingIndex in the axial t-th detector module in the axial structure of the model ′ =RingIndex+(t-1)*NumCry;t=1,2,3,...,ND A ; RingIndex is the original ring number of the crystal strip in the PET system, RingIndex ′ Satisfy(RingIndex ′ +1)%(NumCry+NumCry gap )=0, it is considered to fall in the number RingIndex ′ If the LOR of a crystal strip hits the axial gap in real space, the contribution of the LOR of this strip to the current pixel is not considered.
7. The method according to claim 1, characterized in that Based on the axial structure of the model, the Monte Carlo simulation tool GATE is used to perform simulated detection to obtain the complete system matrix corresponding to the axial single-layer image pixels, and then the complete system matrix of the PET system is generated according to the complete system matrix corresponding to the axial image pixels of each layer.
