A gamma photon detection device and a positron annihilation angle correlation measurement device

Abstract

The present invention discloses a gamma photon detection device and a positron annihilation angle correlation measurement device, which relate to the field of nuclear detection technology and solve the problems of high performance requirements for instruments and low measurement efficiency in existing measurement methods. The gamma photon detection device includes: a Na22 radiation source for generating positrons; two fluorescence detectors, respectively covered with scintillation crystal plates coupled, respectively arranged parallel and symmetrically on both sides of the sample to be tested; the scintillation crystal plates and the fluorescence detectors cooperate with each other to detect the first two-dimensional array and the second two-dimensional array corresponding to the first gamma photon and the second gamma photon propagating in the reverse direction generated by the annihilation of electrons and positrons in the sample to be tested; the gamma photon detector is used to detect the third gamma photon generated when the Na22 radiation source generates positrons, so as to serve as an exposure trigger for the two fluorescence detectors; a memory is used to store the two two-dimensional arrays corresponding to the gamma photon pairs detected by the gamma photon detection device.

Description

Technical Field

[0001] The present invention relates to the field of nuclear detection technology, and in particular to a gamma photon detection device and a positron annihilation angle correlation measurement device. Background Art

[0002] The positron annihilation angular correlation technique has unique sensitivity in measuring electron state density and electron momentum. After the positron is thermalized in the material, it annihilates with the electron. According to the law of conservation of momentum, the two characteristic gamma rays in opposite directions produced by the annihilation will deflect a certain angle θ on the 180° line. The range of the deflection angle θ is generally within 1°. The θ distribution curve is the angular correlation spectrum, which has very high requirements on the angular resolution of the spectrometer, generally requiring sub-mrad (milliradian) level, which can reflect the distribution of electron momentum in the material.

[0003] In the prior art, high-precision instruments are usually used to collimate and locate the light beam to determine the effective event, and then statistically obtain the angle-related one-dimensional curve, or use the gamma photon positioning algorithm to obtain the coordinates of the photon incident point on the detector, and then reverse the photon propagation direction to calculate the deflection angle.

[0004] The prior art has at least the following defects: first, high angular resolution is achieved through high-precision instruments and equipment, but its counting efficiency is low and the disturbance signal is strong, which increases the complexity of measurement and poor operability; second, the use of positioning algorithms can effectively solve the problem of resolution being limited by equipment accuracy, but commonly used positioning algorithms have low accuracy and low efficiency in processing large amounts of data. For example, the center of gravity method has low positioning accuracy, and the maximum likelihood method has a high time overhead when processing data. Summary of the invention

[0005] In view of the above analysis, the present invention aims to provide a gamma photon detection device and a positron annihilation angle correlation measurement device to solve the problems of high performance requirements on instruments and low measurement efficiency in existing measurement methods.

[0006] In one aspect, the present invention discloses a gamma photon detection device, comprising:

[0007] Na22 radioactive source, used to produce positrons;

[0008] Two fluorescence detectors, respectively covered and coupled with scintillation crystal plates, are respectively arranged in parallel and symmetrically on both sides of the sample to be tested; the scintillation crystal plates cooperate with the fluorescence detectors to detect a first two-dimensional array and a second two-dimensional array corresponding to the first gamma photons and the second gamma photons propagating in the opposite direction generated by the annihilation of electrons and positrons in the sample to be tested;

[0009] A gamma photon detector, used to detect the third gamma photon generated when the Na22 radiation source generates positrons, so as to serve as an exposure trigger for the two fluorescence detectors;

[0010] The memory is used to store two two-dimensional arrays corresponding to the gamma photon pairs detected by the gamma photon detection device.

[0011] On the basis of the above scheme, the present invention also makes the following improvements:

[0012] Furthermore, when the gamma photon detector detects the third gamma photon, the two fluorescence detectors are triggered to be exposed, and the exposure time of the two fluorescence detectors is set to be in the order of hundreds of nanoseconds, thereby detecting effective gamma photon pairs to obtain corresponding two-dimensional arrays.

[0013] Furthermore, the gamma photon detector is arranged in the middle and upper part of the two fluorescence detectors and is perpendicular to the two fluorescence detectors;

[0014] The position setting of the gamma photon detector cannot block the detection of the first gamma photon or the second gamma photon by the fluorescence detector.

[0015] Furthermore, the detection solid angle of the gamma photon detection device, the size of the scintillation crystal plate, and the distance between the two scintillation crystal plates satisfy the following relationship:

[0016]

[0017] Wherein, α represents the arc of the detection solid angle, a represents the side length of the scintillation crystal plate, and L represents half of the distance between two scintillation crystal plates.

[0018] Furthermore, the fluorescence detector is a microchannel multiplier tube or a silicon photomultiplier tube array module equipped with a photocathode.

[0019] Furthermore, two samples to be tested with exactly the same shape and material are placed on both sides of the Na22 radiation source to form a sandwich structure.

[0020] On the other hand, the present invention further discloses a positron annihilation angle correlation measurement device, comprising: a measurement system, and a gamma photon detection device as described in any one of the above items; wherein the measurement system comprises:

[0021] A sample acquisition module, used for acquiring a sample training set, each sample in the sample training set includes two two-dimensional arrays corresponding to the detection of gamma photon pairs and the incident position coordinates of the gamma photon pairs;

[0022] A model training module optimizes the deep neural network model based on the sample training set;

[0023] A positioning module, used to input the first two-dimensional array and the second two-dimensional array into the optimized deep neural network model to obtain the incident position coordinates of the first gamma photon and the second gamma photon respectively;

[0024] The processor is used to obtain the corresponding three-dimensional annihilation angle according to the incident position coordinates and the position coordinates of the sample to be tested, and obtain the two-dimensional positron annihilation angle correlation spectrum corresponding to the sample to be tested according to the three-dimensional annihilation angle.

[0025] On the basis of the above scheme, the present invention also makes the following improvements:

[0026] Further, the step of obtaining a two-dimensional positron annihilation angle correlation spectrum corresponding to the sample to be tested according to the three-dimensional annihilation angle includes:

[0027] Obtaining a first projection angle θ1' and a second projection angle θ2' of the stereoscopic annihilation angle on the first projection plane and the second projection plane respectively;

[0028] Based on the correspondence of multiple groups of incident position coordinates, multiple groups of θ1' values ​​and θ2' values ​​are obtained, and the distribution of θ1' values ​​and θ2' values ​​is obtained by counting statistics, and then the positron annihilation angle correlation spectrum is obtained.

[0029] Furthermore, the first projection plane and the second projection plane are perpendicular to each other, and both the first projection plane and the second projection plane are perpendicular to the plane where any fluorescence detector is located.

[0030] Furthermore, the deep neural network model includes an input layer, a first convolutional layer, a first ReLU activation layer, a first pooling layer, a second convolutional layer, a second ReLU activation layer, a second pooling layer, a first fully connected layer, a second fully connected layer and an output layer connected in sequence;

[0031] The two-dimensional array corresponds to an image matrix representing the number of pixels in two dimensions of the image;

[0032] The first convolution layer includes 16 3×3 convolution kernels, which are used to perform convolution operations on the image matrix to obtain 16 feature matrices accordingly;

[0033] The first pooling layer is used to downsample the 16 feature matrices output by the first ReLU activation layer, and correspondingly obtain 16 feature matrices after dimensionality reduction;

[0034] The second convolution layer includes 32 3×3 convolution kernels, which are used to perform convolution operations on the 16 feature matrices after dimensionality reduction, and correspondingly obtain 32 feature matrices;

[0035] The second pooling layer is used to further downsample the 32 feature matrices output by the second ReLU activation layer, so as to obtain 32 feature matrices after dimensionality reduction;

[0036] The first fully connected layer includes 32 neurons, each of which is connected to each neuron of the second pooling layer, the second fully connected layer includes 16 neurons, each of which is connected to each neuron of the first fully connected layer, and the output layer includes two neurons, each of which is connected to each neuron of the second fully connected layer, and outputs the incident position coordinates corresponding to the two-dimensional array;

[0037] The first ReLU activation layer and the second ReLU activation layer both use rectified linear units as activation functions.

[0038] Compared with the prior art, the present invention can achieve at least one of the following beneficial effects:

[0039] 1. The gamma photon detection device proposed in the present invention utilizes the third gamma photon generated when the Na22 radiation source generates positrons, and utilizes the gamma photon detector to detect the third gamma photon as a trigger for triggering the exposure of the fluorescence detector, thereby reducing the requirements for the output signal efficiency of the fluorescence detector.

[0040] 2. The positron annihilation angular correlation measurement device proposed in the present invention combines angular correlation spectrum measurement with deep learning for the first time, and uses a trained deep neural network model to process the two-dimensional array corresponding to the two gamma photons obtained by detection to directly obtain the corresponding incident position coordinates, and then obtain the angular correlation spectrum. The measurement device is simplified, and the measurement accuracy is improved while ensuring it. The defects of low efficiency, high complexity and poor operability when using high-precision instruments for detection are avoided, and the defects of low positioning accuracy and low data processing efficiency when using positioning algorithms are avoided.

[0041] 3. The positron annihilation angular correlation measurement device proposed in the present invention simulates the gamma photon detection device and sets the emission position and incident angle of the gamma photon pairs to detect the gamma photon pairs to obtain two two-dimensional arrays, thereby obtaining a sample training set, and training the deep neural network model based on the sample training set. The sample acquisition method is simple and the sample data has high accuracy, which is beneficial to improving the optimization effect of the deep neural network model and improving the accuracy of angular correlation spectrum measurement.

[0042] In the present invention, the above-mentioned technical solutions can also be combined with each other to achieve more preferred combination solutions. Other features and advantages of the present invention will be described in the subsequent description, and some advantages can become obvious from the description, or can be understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained through the contents particularly pointed out in the description and the drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] The drawings are only for the purpose of illustrating particular embodiments and are not to be considered limiting of the present invention. Like reference symbols denote like components throughout the drawings.

[0044] Figure 1 This is a flow chart of a positron annihilation angle correlation measurement method based on deep learning according to an embodiment of the present invention;

[0045] Figure 2 is a schematic diagram of a gamma photon detection device according to an embodiment of the present invention;

[0046] Figure 3 It is a three-dimensional schematic diagram of a gamma photon pair propagating in the reverse direction according to an embodiment of the present invention;

[0047] Figure 4 A schematic diagram of the relationship between the incident position coordinate x set in an embodiment of the present invention and the incident position coordinate x' obtained by calculation;

[0048] Figure 5 A schematic diagram of the relationship between the incident position coordinate y set in the embodiment of the present invention and the incident position coordinate y' obtained by calculation;

[0049] Figure 6 Schematic diagram of a positron annihilation angle correlation measurement device according to an embodiment of the present invention.

[0050] Reference numerals:

[0051] 1-fluorescence detector; 2-scintillating crystal plate; 3-gamma photon detector; 4-sample to be tested; 110-sample acquisition module; 120-model training module; 130-positioning module; 140-processor. DETAILED DESCRIPTION

[0052] The preferred embodiments of the present invention are described in detail below in conjunction with the accompanying drawings, wherein the accompanying drawings constitute a part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not used to limit the scope of the present invention.

[0053] Method Embodiment

[0054] A specific embodiment of the present invention discloses a positron annihilation angle correlation measurement method based on deep learning. Figure 1 As shown, the method includes:

[0055] S110, obtaining a sample training set, wherein each sample in the sample training set includes two two-dimensional arrays corresponding to the detection of a gamma photon pair and the incident position coordinates of the gamma photon pair.

[0056] S120. Optimize the deep neural network model based on the sample training set.

[0057] S130, using a gamma photon detection device to detect first gamma photons and second gamma photons generated by annihilation of electrons and positrons in the sample to be tested, and correspondingly obtaining a first two-dimensional array and a second two-dimensional array, wherein the two-dimensional array contains incident position information of the gamma photons.

[0058] S140. Obtain incident position coordinates of the first gamma photon and the second gamma photon respectively based on the first two-dimensional array and the second two-dimensional array using the optimized deep neural network model.

[0059] S150, obtaining a corresponding three-dimensional annihilation angle according to the incident position coordinates and the position coordinates of the sample to be tested, and obtaining a two-dimensional positron annihilation angle correlation spectrum corresponding to the sample to be tested according to the three-dimensional annihilation angle.

[0060] Preferably, Figure 2 As shown, the gamma photon detection device includes:

[0061] The Na22 radioactive source is used to generate positrons. The Na22 radioactive source and the sample to be tested are located at the same position. Specifically, in practical applications, in order to allow more positrons generated by the Na22 radioactive source to enter the sample to be tested and annihilate with the electrons therein, two samples to be tested with exactly the same shape and material are placed on both sides of the Na22 radioactive source to form a sandwich structure. Since the volume of the Na22 radioactive source is very small and negligible, the Na22 radioactive source and the sample to be tested can be considered to be located at the same position.

[0062] Two fluorescence detectors 1 are respectively covered with scintillation crystal plates 2 coupled thereto, and are respectively arranged parallel and symmetrically on both sides of the sample to be tested 4. Specifically, the two scintillation crystal plates are used to detect the first gamma photon γ1 and the second gamma photon γ2 generated by the annihilation of electrons and positrons in the sample to be tested, which propagate in the opposite direction and generate corresponding fluorescence signals. The two fluorescence detectors 1 are used to detect the corresponding fluorescence signals to obtain the corresponding first two-dimensional array and second two-dimensional array; the size of the scintillation crystal plate is the same as the size of the fluorescence detector. Preferably, the material of the scintillation crystal plate is lutetium yttrium silicate scintillation crystal (LYSO), which can generate a large number of fluorescence signals after receiving gamma photons, which is beneficial to detection; in addition, the greater the thickness of the scintillation crystal plate, the more conducive it is to the energy deposition of gamma photons, but it will also cause the transmission attenuation of fluorescence to a certain extent, so it should be weighed according to the actual measurement situation. Specifically, the thickness range of the scintillation crystal plate is [5mm, 7mm], and preferably, it is set to 6mm.

[0063] Specifically, the fluorescence detector is a microchannel multiplier tube equipped with a photocathode (MCP-PMT) or a silicon photomultiplier tube array module (SiPM).

[0064] Considering that in order to obtain the two-dimensional array corresponding to the effective gamma photon pairs within the time coincidence in the prior art, the time for the detector detecting the gamma photons to continuously output the signal is in the order of hundreds of nanoseconds, which greatly increases the performance requirements for the corresponding detectors. The present invention solves this problem by adopting the following technical means:

[0065] When a Na22 radioactive source produces a positron, it can also produce a third gamma photon with an energy of 1.28 MeV.

[0066] Therefore, a gamma photon detector 3 is arranged in the middle and upper part of the two fluorescence detectors 1, and the gamma photon detector is perpendicular to the two fluorescence detectors to detect the third gamma photon γ3 generated when the Na22 radiation source generates positrons, so as to serve as an exposure trigger for the two fluorescence detectors. Specifically, when the gamma photon detector detects the third gamma photon, the two fluorescence detectors are triggered to be exposed, and the exposure time of the two fluorescence detectors is set to the order of hundreds of nanoseconds, so that the effective gamma photon pairs can be detected to obtain the corresponding two-dimensional array; in addition, by controlling the multiple switches of the Na22 radiation source, multiple groups of two-dimensional arrays corresponding to the gamma photon pairs are obtained accordingly.

[0067] Preferably, the smaller the distance between the gamma photon detector and the sample to be measured is, the more conducive it is to improving measurement accuracy. However, it should be noted that the position of the gamma photon detector cannot block the detection of the first gamma photon or the second gamma photon by the fluorescence detector.

[0068] Preferably, two two-dimensional arrays corresponding to the gamma photon pairs detected by the gamma photon detection device are stored in corresponding memories for subsequent positioning and calling.

[0069] Preferably, in step S110, the sample training set is obtained by:

[0070] The gamma photon detection device is simulated, and the emission position and incident angle of the gamma photon pair are set in the simulation program to detect the gamma photon pair and obtain two two-dimensional arrays. Preferably, the gamma photon detection device and the gamma photon transmission are simulated using Geant4 open source software; specifically, the input items set include: two gamma photon sources (located at the same position), the energy of the gamma photons generated by each gamma photon source is 0.511 MeV, the direction (i.e., angle) of the gamma photons emitted by each gamma photon source, and the position of the gamma photon source, the type, size, and placement of the fluorescence detector and the scintillation crystal plate are set according to the gamma photon detection device. The output items set include: the two-dimensional distribution of the reaction deposition energy (i.e., the deposition energy of the fluorescence signal) detected in the detector, i.e., a two-dimensional array.

[0071] According to the set emission position and incident angle of the gamma photon pair, the incident position coordinates of the two gamma photons on the two fluorescence detectors are obtained respectively.

[0072] The two 2D arrays corresponding to the gamma photon pairs and the incident position coordinates form a sample;

[0073] By changing the incident angle of the gamma photon pair, multiple samples are obtained, and then a sample training set is obtained.

[0074] Preferably, obtaining the incident position coordinates of the two gamma photons on the two fluorescence detectors respectively according to the emission position and the incident angle of the set gamma photon pair specifically includes:

[0075] like Figure 3 As shown, θ1, They represent the two angles corresponding to the incident path of the first gamma photon γ1 in the polar coordinate system; θ2, They respectively represent the two angles corresponding to the incident path of the second gamma photon γ2 in the polar coordinate system, where the emission position coordinate of the gamma photon pair is the coordinate zero point; the incident position coordinate is obtained by calculating the following formula:

[0076] The incident position coordinates of the first gamma photon are A(x1,y1,-L):

[0077] x1=L×tanθ1,

[0078]

[0079] The incident position coordinates of the second gamma photon are B(x2,y2,L):

[0080] x2=L×tanθ2,

[0081]

[0082] Wherein, L represents half of the distance between the two scintillation crystal plates, and its value can be set according to specific requirements.

[0083] Preferably, the deep neural network model includes an input layer, a first convolutional layer, a first ReLU activation layer, a first pooling layer, a second convolutional layer, a second ReLU activation layer, a second pooling layer, a first fully connected layer, a second fully connected layer and an output layer connected in sequence.

[0084] Specifically, the two-dimensional array corresponds to an image matrix representing the number of pixels in two dimensions of the image. The image matrix is ​​input to the input layer, and the image edge is padded with 0 to obtain a 24×35 image matrix, which is then input to the first convolutional layer.

[0085] The first convolutional layer includes 16 3×3 convolution kernels, which are used to perform convolution operations on the aforementioned image matrix, corresponding to obtaining 16 24×35 feature matrices, which are processed through the first ReLU activation layer so that the output of some neurons is 0, increasing the nonlinearity between the layers of the deep neural network to prevent overfitting.

[0086] The first pooling layer is used to downsample the 16 feature matrices output by the first ReLU activation layer, and obtain 16 feature matrices after dimensionality reduction. Specifically, a 2×2 sliding window is used to slide the value of each feature matrix, and the maximum value in the sliding window is used as the corresponding matrix element, thereby obtaining a 23×34 feature matrix.

[0087] The second convolutional layer includes 32 3×3 convolution kernels, which are used to perform convolution operations on the 16 feature matrices after dimensionality reduction, and correspondingly obtain 32 23×34 feature matrices.

[0088] The second pooling layer is used to further downsample the 32 feature matrices output by the second ReLU activation layer, and obtain 32 feature matrices after dimensionality reduction. Specifically, based on the same principle as the first pooling layer, a 2×2 sliding window is used for sliding window value selection, and the maximum value in the sliding window is used as the corresponding matrix element, thereby obtaining a 22×33 feature matrix. Specifically, the second ReLU activation layer has the same function as the first ReLU activation layer, which is used to prevent overfitting. Preferably, the first ReLU activation layer and the second ReLU activation layer both use a rectified linear unit as the activation function.

[0089] The first fully connected layer contains 32 neurons, each of which is connected to each neuron in the second pooling layer. The second fully connected layer contains 16 neurons, each of which is connected to each neuron in the first fully connected layer. The output layer includes two neurons, each of which is connected to each neuron in the second fully connected layer, and outputs the incident position coordinates corresponding to the two-dimensional array. Specifically, each neuron in the first fully connected layer, the second fully connected layer, and the output layer is connected to each neuron in the previous layer, so as to ensure that all features output by the previous layer are received.

[0090] Preferably, in S120, the step of optimizing the deep neural network model based on the sample training set includes:

[0091] The obtained sample training set is randomly divided into multiple sample training subsets.

[0092] Based on multiple sample training subsets, loss functions corresponding to multiple deep neural network models are obtained.

[0093] For each loss function, the gradient descent method is used to obtain the deep neural network model parameters that minimize the loss function value, thereby obtaining the corresponding deep neural network model.

[0094] Based on the multiple loss functions, multiple deep neural network models are obtained, and the deep neural network model corresponding to the minimum loss function value is used as the optimized deep neural network model;

[0095] Among them, the loss function is specifically:

[0096]

[0097] L=(l1,l2,......l N ) T

[0098] l i =(S i -T i ) 2 ,i∈[1,N],

[0099] In the above formula, L(ST) represents the loss function, S i represents the vector composed of the predicted incident position coordinates of the first gamma photon and the second gamma photon obtained based on the deep neural network model, T i represents a vector consisting of the actual incident position coordinates of the first gamma photon and the second gamma photon in the sample, and N represents the number of samples in each sample training subset.

[0100] In order to better demonstrate the accuracy of the positioning by using the deep neural network model proposed in the present invention according to the two-dimensional array, the set incident position coordinates are compared with the calculated incident position coordinates, as shown in FIG. Figure 4 As shown, the horizontal axis represents the value of x in the incident position coordinate obtained by setting the incident angle, and the vertical axis represents the value of x' in the incident position coordinate calculated according to the deep neural network model. Figure 4 Each point in the equation represents the value of x' in the incident position coordinates obtained by the deep neural network model when the incident position coordinates x takes a certain value. The corresponding straight line x'=1.0046x+6.617×10 is obtained by fitting multiple points. -4 , it can be seen from its expression that the value of x' is basically equal to the value of x; in addition, Figure 5 As shown, the horizontal axis represents the value of y in the incident position coordinate obtained by setting the incident angle, and the vertical axis represents the value of y' in the incident position coordinate calculated according to the deep neural network model. Figure 5Each point in the equation represents the value of y' in the incident position coordinate calculated by the deep neural network model when y in the incident position coordinate takes a certain value. Multiple points are fitted to obtain the corresponding straight line y'=1.0046y+6.617×10 -4 As can be seen from the expression, the y' value is substantially equal to the y value. Therefore, the calculation and positioning based on the two-dimensional array using the deep neural network model proposed in the present invention has higher accuracy.

[0101] Preferably, in order to achieve the measurement of the deflection angle within the range of 1°, the detection solid angle of the gamma photon detection device is set to be greater than or equal to 35 mrad.

[0102] Preferably, the detection solid angle, the size of the scintillation crystal plate, and the distance between the two scintillation crystal plates satisfy the following relationship:

[0103]

[0104] Wherein, α represents the arc of the detection solid angle, a represents the side length of the scintillation crystal plate, and L represents half of the distance between two scintillation crystal plates.

[0105] In the actual detection process, a gamma photon detection device is used to detect and obtain two two-dimensional arrays corresponding to gamma photons, and then processed based on the optimized deep neural network model to obtain the corresponding incident position coordinates. Multiple detections are performed by controlling the multiple switches of the Na22 radiation source to obtain multiple groups of two-dimensional arrays, and then the corresponding multiple groups of incident position coordinates are obtained.

[0106] Preferably, in S150, the corresponding stereo annihilation angle is obtained according to the incident position coordinates and the position coordinates of the sample to be tested (coordinate zero point), and the two-dimensional positron annihilation angle correlation spectrum corresponding to the sample to be tested is obtained according to the stereo annihilation angle, specifically including:

[0107] According to the incident position coordinates of the first gamma photon and the second gamma photon and the position coordinates of the sample O(0, 0, 0), the corresponding stereo annihilation angle is obtained by:

[0108]

[0109] The first projection angle θ1' and the second projection angle θ2' of the stereoscopic annihilation angle are obtained on the first projection plane and the second projection plane, respectively; wherein the first projection plane and the second projection plane are perpendicular to each other, and the first projection plane and the second projection plane are both perpendicular to the plane where any fluorescence detector is located. Exemplarily, in the xyz three-dimensional coordinate system, the two fluorescence detectors are parallel to the xy plane, the first projection plane is parallel to the yz plane, and the second projection plane is parallel to the xz plane. Accordingly, θ yz is the first projection angle θ1', θ xzThe specific formula for the second projection angle θ2' is as follows:

[0110]

[0111]

[0112] Based on the correspondence of multiple groups of incident position coordinates, multiple groups of θ1' values ​​and θ2' values ​​are obtained, and the distribution of θ1' values ​​and θ2' values ​​is obtained by counting statistics, and then the positron annihilation angle correlation spectrum is obtained.

[0113] Device Embodiment

[0114] Another embodiment of the present invention discloses a positron annihilation angle correlation measurement device. Since the working principle of the device is the same as that of the above method embodiment, the repetitive parts can be referred to the above method embodiment, and will not be repeated here.

[0115] Specifically, Figure 6 As shown, the positron annihilation angle correlation measurement device includes a measurement system and a gamma photon detection device.

[0116] Preferably, the gamma photon detection device is used to detect first gamma photons and second gamma photons generated by annihilation of electrons and positrons in the sample to be tested, and obtain a first two-dimensional array and a second two-dimensional array respectively.

[0117] The measurement system includes:

[0118] The sample acquisition module 110 is used to acquire a sample training set, where each sample in the sample training set includes two two-dimensional arrays corresponding to the detected gamma photon pairs and the incident position coordinates of the gamma photon pairs.

[0119] The model training module 120 optimizes the deep neural network model based on the sample training set.

[0120] The positioning module 130 is used to obtain the incident position coordinates of the first gamma photon and the second gamma photon respectively based on the first two-dimensional array and the second two-dimensional array by using the optimized deep neural network model.

[0121] The processor 140 is used to obtain the corresponding three-dimensional annihilation angle according to the incident position coordinates and the position coordinates of the sample to be tested, and obtain the two-dimensional positron annihilation angle correlation spectrum corresponding to the sample to be tested according to the three-dimensional annihilation angle.

[0122] Preferably, Figure 2 As shown, the gamma photon detection device includes:

[0123] Na22 radioactive source, used to produce positrons.

[0124] Two fluorescence detectors are respectively covered and coupled with scintillation crystal plates, and are respectively and symmetrically arranged on both sides of the sample to be tested; the two scintillation crystal plates are used to detect the first gamma photons and the second gamma photons propagating in the opposite direction generated by the annihilation of electrons and positrons in the sample to be tested to generate corresponding fluorescence signals; the two fluorescence detectors are used to detect the corresponding fluorescence signals to obtain the corresponding first two-dimensional array and the second two-dimensional array; the size of the scintillation crystal plate is the same as the size of the fluorescence detector.

[0125] The gamma photon detector is arranged in the middle and upper part of the two fluorescence detectors and is perpendicular to the two fluorescence detectors. It is used to detect the third gamma photon generated when the Na22 radiation source generates positrons, so as to serve as an exposure trigger for the two fluorescence detectors.

[0126] Compared with the prior art, the positron annihilation angular correlation measurement method based on deep learning and the positron annihilation angular correlation measurement device disclosed in the embodiments of the present invention combine angular correlation spectrum measurement with deep learning for the first time, and use the trained deep neural network model to process the two-dimensional array corresponding to the two gamma photons detected to directly obtain the corresponding incident position coordinates, and then obtain the angular correlation spectrum, simplifying the measurement device, while ensuring the measurement accuracy. Improve the measurement accuracy, avoid the defects of low efficiency, high complexity and poor operability when using high-precision instruments for detection, and also avoid the defects of low positioning accuracy and low data processing efficiency when using positioning algorithms. Secondly, a Na22 radioactive source is used to generate a third gamma photon generated by the positron, and a gamma photon detector is used to detect the third gamma photon as a trigger for triggering the exposure of the fluorescence detector, thereby reducing the requirements for the efficiency of the fluorescence detector output signal. In addition, by simulating the gamma photon detection device and setting the emission position and incident angle of the gamma photon pairs, two two-dimensional arrays are obtained to detect the gamma photon pairs, thereby obtaining a sample training set, and the deep neural network model is trained based on the sample training set. The sample acquisition method is simple and the sample data has high accuracy, which is conducive to improving the optimization effect of the deep neural network model and improving the measurement accuracy of the angular correlation spectrum.

[0127] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by any technician familiar with the technical field within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention.

Claims

1. A gamma photon detection device, characterized in that: include: Na22 radioactive source, used to produce positrons; Two fluorescence detectors, respectively covered and coupled with scintillation crystal plates, are respectively arranged in parallel and symmetrically on both sides of the sample to be tested; the scintillation crystal plates cooperate with the fluorescence detectors to detect a first two-dimensional array and a second two-dimensional array corresponding to the first gamma photons and the second gamma photons propagating in the opposite direction generated by the annihilation of electrons and positrons in the sample to be tested; A gamma photon detector, used to detect the third gamma photon generated when the Na22 radiation source generates positrons, so as to serve as an exposure trigger for the two fluorescence detectors; The memory is used to store two two-dimensional arrays corresponding to the gamma photon pairs detected by the gamma photon detection device.

2. The gamma photon detection device according to claim 1, characterized in that: When the gamma photon detector detects the third gamma photon, the two fluorescence detectors are triggered to be exposed, and the exposure time of the two fluorescence detectors is set to be in the order of hundreds of nanoseconds, so as to detect effective gamma photon pairs and obtain corresponding two-dimensional arrays.

3. The gamma photon detection device according to claim 2, characterized in that: The gamma photon detector is arranged in the middle and upper part of the two fluorescence detectors and is perpendicular to the two fluorescence detectors; The position setting of the gamma photon detector cannot block the detection of the first gamma photon or the second gamma photon by the fluorescence detector.

4. The gamma photon detection device according to claim 3, characterized in that: The detection solid angle of the gamma photon detection device, the size of the scintillation crystal plate, and the distance between the two scintillation crystal plates satisfy the following relationship: Wherein, α represents the arc of the detection solid angle, a represents the side length of the scintillation crystal plate, and L represents half of the distance between two scintillation crystal plates.

5. The gamma photon detection device according to claim 1, characterized in that: The fluorescence detector is a microchannel multiplier tube or a silicon photomultiplier tube array module equipped with a photocathode.

6. The gamma photon detection device according to claim 1, characterized in that: Two samples to be tested with exactly the same shape and material are placed on both sides of the Na22 radiation source to form a sandwich structure.

7. A positron annihilation angle correlation measurement device, characterized in that: include: A measurement system, and a gamma photon detection device according to any one of claims 1 to 6; wherein the measurement system comprises: A sample acquisition module, used for acquiring a sample training set, each sample in the sample training set includes two two-dimensional arrays corresponding to the detection of gamma photon pairs and the incident position coordinates of the gamma photon pairs; A model training module optimizes the deep neural network model based on the sample training set; A positioning module, used to input the first two-dimensional array and the second two-dimensional array into the optimized deep neural network model to obtain the incident position coordinates of the first gamma photon and the second gamma photon respectively; The processor is used to obtain the corresponding three-dimensional annihilation angle according to the incident position coordinates and the position coordinates of the sample to be tested, and obtain the two-dimensional positron annihilation angle correlation spectrum corresponding to the sample to be tested according to the three-dimensional annihilation angle.

8. The positron annihilation angle correlation measurement device according to claim 7, characterized in that: The step of obtaining a two-dimensional positron annihilation angle correlation spectrum corresponding to the sample to be tested according to the three-dimensional annihilation angle comprises: Obtaining a first projection angle θ′1 and a second projection angle θ′2 of the stereoscopic annihilation angle on the first projection plane and the second projection plane respectively; Based on the correspondence of multiple groups of incident position coordinates, multiple groups of θ′1 values ​​and θ′2 values ​​are obtained, and the distribution of θ′1 values ​​and θ′2 values ​​is obtained by counting statistics, and then the positron annihilation angle correlation spectrum is obtained.

9. The positron annihilation angle correlation measurement device according to claim 8, characterized in that: The first projection plane and the second projection plane are perpendicular to each other, and both the first projection plane and the second projection plane are perpendicular to the plane where any fluorescence detector is located.

10. The positron annihilation angle correlation measurement device according to any one of claims 7 to 9, characterized in that: The deep neural network model includes an input layer, a first convolutional layer, a first ReLU activation layer, a first pooling layer, a second convolutional layer, a second ReLU activation layer, a second pooling layer, a first fully connected layer, a second fully connected layer and an output layer connected in sequence; The two-dimensional array corresponds to an image matrix representing the number of pixels in two dimensions of the image; The first convolution layer includes 16 3×3 convolution kernels, which are used to perform convolution operations on the image matrix to obtain 16 feature matrices accordingly; The first pooling layer is used to downsample the 16 feature matrices output by the first ReLU activation layer, and correspondingly obtain 16 feature matrices after dimensionality reduction; The second convolution layer includes 32 3×3 convolution kernels, which are used to perform convolution operations on the 16 feature matrices after dimensionality reduction, and correspondingly obtain 32 feature matrices; The second pooling layer is used to further downsample the 32 feature matrices output by the second ReLU activation layer, so as to obtain 32 feature matrices after dimensionality reduction; The first fully connected layer includes 32 neurons, each of which is connected to each neuron of the second pooling layer, the second fully connected layer includes 16 neurons, each of which is connected to each neuron of the first fully connected layer, and the output layer includes two neurons, each of which is connected to each neuron of the second fully connected layer, and outputs the incident position coordinates corresponding to the two-dimensional array; The first ReLU activation layer and the second ReLU activation layer both use rectified linear units as activation functions.

Images