CN109884682A

CN109884682A - A kind of generation method, device, equipment and the medium of crystal positions look-up table

Info

Publication number: CN109884682A
Application number: CN201910105044.3A
Authority: CN
Inventors: 李运达; 孙智鹏; 李明
Original assignee: Neusoft Medical Systems Co Ltd
Current assignee: Shenyang Zhihe Medical Technology Co ltd
Priority date: 2019-02-01
Filing date: 2019-02-01
Publication date: 2019-06-14
Anticipated expiration: 2039-02-01
Also published as: CN109884682B

Abstract

The embodiment of the present application discloses a kind of generation method of crystal positions look-up table, it provides a kind of mode of new generation crystal positions look-up table, utilize crystal positions distributed model, generate the corresponding crystal positions distribution map of energy diagram, in turn, crystal positions look-up table is determined according to the crystal positions distribution map.Since crystal positions distributed model is to pass through machine learning algorithm, the neural network obtained based on the corresponding calibration crystal positions distribution map training of energy diagram and energy diagram, during carrying out signature analysis to energy diagram using the neural network, it can be automatically to because of energy diagram noise, the error that the factors such as artifact and image grayscale are uneven generate correspondingly is corrected, to guarantee to be based on neural network crystal positions distribution map accuracy with higher generated, accuracy to guarantee the crystal positions look-up table determined based on the crystal positions distribution map is higher.

Description

Crystal position lookup table generation method, device, equipment and medium

Technical Field

The application relates to the technical field of nuclear medicine imaging, in particular to a crystal position lookup table generation method, a crystal position lookup table generation device, crystal position lookup table generation equipment and a crystal position lookup table generation medium.

Background

Positron Emission Tomography (PET) is the currently advanced imaging device for clinical examination in the field of nuclear medicine, wherein the detector portion is the portion that most directly affects the image resolution. The PET detector is generally composed of a scintillation crystal array coupled with a photoelectric conversion device, and the positron annihilation event is decoded by the PET detector to generate a position distribution of the event, which is statistically generated to be a two-dimensional histogram (2d flow histogram image), also called an energy map, for reflecting the distribution of the scintillation crystal array.

Since the energy map generation process is affected by many nonlinear factors, such as compton scattering effect, nonlinear response of electronic devices, and non-uniform physical characteristics of the scintillation crystal, the correspondence between the finally generated energy map and the actual physical position of the scintillation crystal is non-linear, and the energy map usually has the problems of overall deformation of the array profile, non-uniform energy distribution, adhesion of crystal spots, crystal loss, crystal artifacts, and the like. In order to ensure that the PET can obtain a high-resolution image, it is necessary to accurately find the corresponding relationship between the actual physical position of the scintillation crystal and the energy map, i.e. determine the crystal position lookup table, so as to correct and rectify the image based on the crystal position lookup table in the following.

In the prior art, a crystal position lookup table is generated by adopting the following two ways; in the first mode, after an energy map is generated based on acquired radiation data, operations such as image filtering or image enhancement and the like which can reduce image noise are performed on the energy map, image segmentation or clustering operations are further performed on the energy map obtained after noise reduction processing to obtain a crystal center position distribution map, and a crystal position lookup table is generated based on the crystal center position distribution map; in the second mode, after an energy map is generated based on the acquired radiation data, the energy map is subjected to gray level inversion, then the energy map subjected to gray level inversion is segmented by methods such as a watershed algorithm and the like to obtain a crystal contour line distribution map, and a crystal position lookup table is generated based on the crystal contour line distribution map.

However, the two methods for generating the crystal position lookup table have certain limitations in practical applications, and are affected by factors such as energy map noise, artifacts, uneven image gray scale, and the like, so that problems such as a crystal center position lookup error, an incomplete crystal center lookup, an inaccurate crystal contour line lookup, and the like often occur, and thus accurate generation of the crystal position lookup table is affected.

Disclosure of Invention

The embodiment of the application provides a method, a device, equipment and a medium for generating a crystal position lookup table, which can effectively improve the accuracy of the generated crystal position lookup table.

In view of the above, a first aspect of the present application provides a method for generating a crystal position lookup table, the method comprising:

acquiring radiation data, and generating a target energy map according to the radiation data;

generating a crystal position distribution diagram corresponding to the target energy diagram by using a crystal position distribution model to serve as a first crystal position distribution diagram; the crystal position distribution model is a neural network, is obtained based on an energy map and a calibration crystal position distribution map corresponding to the energy map through training, and takes the energy map as input and the crystal position distribution map corresponding to the energy map as output;

and determining a crystal position lookup table according to the first crystal position distribution diagram.

A second aspect of the present application provides a model training method, the method comprising:

determining a training sample set, wherein each training sample in the training sample set comprises an energy map and a calibration crystal position distribution map corresponding to the energy map;

training a neural network by using the training sample set to obtain a crystal position distribution model through training; the crystal position distribution model takes an energy map as input and takes a crystal position distribution map corresponding to the energy map as output.

A third aspect of the present application provides an apparatus for generating a crystal position lookup table, the apparatus comprising:

the acquisition module is used for acquiring radiation data and generating a target energy map according to the radiation data;

the first processing module is used for generating a crystal position distribution diagram corresponding to the target energy diagram by using a crystal position distribution model to serve as a first crystal position distribution diagram; the crystal position distribution model is a neural network, is obtained based on an energy map and a calibration crystal position distribution map corresponding to the energy map through training, and takes the energy map as input and the crystal position distribution map corresponding to the energy map as output;

and the determining module is used for determining a crystal position lookup table according to the first crystal position distribution diagram.

A fourth aspect of the present application provides a model training apparatus, the apparatus comprising:

the system comprises a sample determining module, a calibration module and a calibration module, wherein the sample determining module is used for determining a training sample set, and each training sample in the training sample set comprises an energy map and a calibration crystal position distribution map corresponding to the energy map;

the training module is used for training the neural network by utilizing the training sample set so as to obtain a crystal position distribution model through training; the crystal position distribution model takes an energy map as input and takes a crystal position distribution map corresponding to the energy map as output.

A fifth aspect of the present application provides an apparatus comprising a processor and a memory:

the memory is used for storing program codes and transmitting the program codes to the processor;

the processor is configured to execute the method for generating a crystal position lookup table according to the first aspect or execute the model training method according to the second aspect according to instructions in the program code.

A sixth aspect of the present application provides a computer-readable storage medium comprising instructions which, when executed on a computer, cause the computer to perform the method for generating a crystal position lookup table according to the first aspect or the method for training a model according to the second aspect.

According to the technical scheme, the embodiment of the application has the following advantages:

the embodiment of the application provides a method for generating a crystal position lookup table, which provides a new method for generating a crystal position lookup table, and utilizes a crystal position distribution model to generate a crystal position distribution diagram corresponding to an energy diagram, and further determines the crystal position lookup table according to the crystal position distribution diagram. Because the crystal position distribution model is a neural network obtained by training based on the energy map and a calibration crystal position distribution diagram corresponding to the energy map through a machine learning algorithm, errors caused by factors such as noise of the energy map, artifacts and uneven image gray scale can be automatically and correspondingly corrected in the process of performing feature analysis on the energy map by using the neural network, so that the crystal position distribution diagram generated based on the neural network has high accuracy, and the accuracy of a crystal position lookup table determined based on the crystal position distribution diagram is high.

Drawings

FIG. 1 is a schematic diagram of a scintillation crystal array, an energy map, and a crystal position look-up table in a PET detector;

FIG. 2 is a schematic flow chart illustrating a method for generating a crystal position lookup table according to the present application;

FIG. 3 is a schematic flow chart of another method for generating a crystal position lookup table according to the present application;

FIG. 4 is a schematic flow chart of a model training method provided herein;

FIG. 5 is a schematic diagram of the overall structure of the crystal position lookup table generation process provided in the present application;

FIG. 6 is a schematic structural diagram of an apparatus for generating a crystal position lookup table according to the present application;

FIG. 7 is a schematic structural diagram of a model training device provided in the present application;

fig. 8 is a schematic structural diagram of a server according to an embodiment of the present application;

fig. 9 is a schematic structural diagram of a terminal device according to an embodiment of the present application.

Detailed Description

In order to make the technical solutions of the present application better understood, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.

The terms "first," "second," "third," "fourth," and the like in the description and in the claims of the present application and in the drawings described above, if any, are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used is interchangeable under appropriate circumstances such that the embodiments of the application described herein are capable of operation in sequences other than those illustrated or described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.

In order to ensure that the PET can obtain a high-resolution image, it is usually necessary to accurately determine a corresponding relationship between an actual physical position of a scintillation crystal in a PET detector and an energy map, and in the specific implementation, a radiation source is placed inside the PET detector, radiation data (gamma photons) generated by decay of the radiation source is collected, the energy map is generated according to the collected radiation data, a crystal position lookup table is further determined according to the energy map, and the crystal position lookup table is used to represent the corresponding relationship between the actual physical position of the scintillation crystal in the PET detector and the energy map.

It should be noted that, with the application of the scintillation crystal containing the lutetium isotope and capable of generating the background radiation, in the process of generating the crystal position lookup table, the energy map can be generated directly based on the background radiation data generated by the scintillation crystal itself, and no additional radioactive source is needed, so that the operation is relatively simpler, and the operation experience is better for the operator.

In the prior art, after an energy map is generated based on collected radiation data, the distribution position of the crystal center or the distribution position of the crystal contour line is usually determined according to the distribution characteristics of light spots in the energy map, and then a crystal position lookup table is determined according to the distribution position of the crystal center or the distribution position of the crystal contour line. However, the generation method of the crystal position lookup table is affected by factors such as energy map noise, artifacts, uneven image gray and the like, and the problems of searching error of the crystal center position, incomplete searching of the crystal center, inaccurate searching of the crystal contour line and the like often occur, so that the accuracy of the finally generated crystal position lookup table is usually low; in addition, compared with an energy map generated based on radiation data generated by radioactive source decay, the energy map generated based on background radiation data is poorer in effect, the energy map generated based on background radiation data is processed by adopting the existing crystal position lookup table generation method, and the generated crystal position lookup table is generally lower in accuracy.

Taking a PET detector composed of 11 × 11 scintillation crystals as an example, as shown in fig. 1, an image 101 is an actual position distribution diagram of the scintillation crystals in the PET detector, both an image 102 and an image 103 are energy diagrams generated based on a position distribution condition of the scintillation crystals in the PET detector shown in the image 101, and an image 104 and an image 105 are crystal position lookup tables generated based on the image 102 and the image 103 respectively by using an existing crystal position lookup table generation method.

It can be found through observation that the correspondence between the actual positions of 11 × 11 scintillation crystals in the image 101 and the distribution positions of light spots in the energy map 102 and the energy map 103 is non-linear, the image noise, deformation and uniformity at different positions in the energy map 102 and the energy map 103 are inconsistent, and compared with the energy map 102, the display effect of the energy map 103 is worse, and the problems of overall profile deformation, non-uniformity of energy distribution, light spot adhesion and the like are more serious.

By adopting the existing crystal lookup table generation method, the distribution positions of the crystal centers can be determined according to the energy diagram 102 and the energy diagram 103, namely the distribution positions of points in the image 104 and the image 105 are determined, and then the crystal position lookup table is obtained by respectively segmenting the energy diagram 102 and the energy diagram 103 according to the distribution positions of the crystal centers; the distribution positions of the crystal contour lines, that is, the distribution positions of the lines in the image 104 (since all crystal centers are not identified in the image 105, the distribution positions of the contour lines of the crystals cannot be determined from the image 105) may also be determined from the energy maps 102 and 103, and the image including the distribution positions of the crystal contour lines is actually a crystal position lookup table.

It can be found by observation that the distribution of each crystal in the crystal position distribution diagrams shown in the images 104 and 105 and the actual distribution of the crystal shown in the image 101 have large differences in both the crystal shape and the crystal distribution position; moreover, the image 104 and the image 105 are compared to find that the display effect of the crystal position distribution diagram is greatly influenced by the actual distribution of the energy diagram, and only under the condition of better display effect of the energy diagram, the determined crystal position distribution diagram can be ensured to have better display effect, so that the crystal position with higher accuracy can be determined to be searched; accordingly, a method for accurately determining a crystal position lookup table suitable for various energy maps is needed.

In order to improve the accuracy of the generated crystal position lookup table, a new method for generating the crystal position lookup table is provided, a crystal position distribution diagram corresponding to an energy diagram is generated by using a crystal position distribution model obtained through machine learning, and the crystal position lookup table is determined based on the crystal position distribution diagram. When the crystal position distribution model is used for carrying out characteristic analysis on the energy map, the crystal position distribution model can automatically correct errors caused by factors such as noise of the energy map, artifacts and uneven image gray scale, so that the crystal position distribution map generated based on the crystal position distribution model has high accuracy, and the crystal position lookup table determined based on the crystal position distribution map has high accuracy.

The method for generating the crystal position lookup table provided in the present application is described below by way of example.

Referring to fig. 2, fig. 2 is a schematic flowchart of a method for generating a crystal position lookup table according to an embodiment of the present application. As shown in fig. 2, the method for generating the crystal position lookup table includes the following steps:

step 201: and acquiring radiation data, and generating a target energy map according to the radiation data.

The radioactive source is placed in the PET detector for scanning, radiation data generated by decay of the radioactive source is collected, an energy diagram is generated according to the collected radiation data, and the energy diagram is used as a target energy diagram.

It should be noted that, in order to simplify the operation process of the operator, the scintillation crystal containing lutetium isotope that can generate background radiation can be directly used as a radioactive source, the background radiation data generated by the scintillation crystal itself is collected, an energy map is generated according to the background radiation data, and the energy map is used as a target energy map. Therefore, the operator can avoid using an additional radioactive source, thereby reducing the operation difficulty caused by using the additional radioactive source.

Step 202: and generating a crystal position distribution diagram corresponding to the target energy diagram by using a crystal position distribution model to serve as a first crystal position distribution diagram.

The generated target energy map is input to a crystal position distribution model, the crystal position distribution model performs characteristic analysis processing on the target energy map, a crystal position distribution map corresponding to the target energy map is output, and the crystal position distribution map is used as a first crystal position distribution map.

The crystal position distribution model is a neural network trained based on an energy map and a calibration crystal position distribution map corresponding to the energy map, and the crystal position distribution model takes the energy map as an input and takes the crystal position distribution map corresponding to the energy map as an output.

Specifically, the crystal position distribution model is a neural network trained based on an energy map and a calibrated crystal position distribution map corresponding to the energy map through a machine learning algorithm, after the energy map is input into the crystal position distribution model, the crystal position distribution model automatically performs feature analysis on the energy map to determine the crystal distribution condition represented by the energy map, and automatically corrects errors caused by factors such as energy map noise, artifacts and image gray scale unevenness while determining the crystal distribution condition represented by the energy map, so that the finally output crystal position distribution map has high accuracy.

In a possible implementation manner, the crystal position distribution model is obtained by training based on an energy map and a crystal center position distribution map corresponding to the energy map, wherein the crystal center position distribution map belongs to one of the crystal position distribution maps and comprises the distribution positions of all crystal centers; accordingly, after the target energy map is input into the crystal position distribution model obtained by training in this way, the crystal position distribution model performs characteristic analysis on the target energy map, and outputs a crystal center position distribution map corresponding to the target energy map, which is the first crystal position distribution map including the distribution positions of the respective crystal centers.

In another possible implementation manner, the crystal position distribution model is obtained by training based on an energy map and a crystal contour line distribution map corresponding to the energy map, the crystal contour line distribution map belongs to one of the crystal position distribution maps, wherein the crystal contour line distribution map comprises contour line distribution positions of each crystal, and a region framed by a contour line of one crystal is a region where the crystal is located; accordingly, after the target energy map is input into the crystal position distribution model obtained by the training, the crystal position distribution model outputs a crystal contour line distribution map corresponding to the target energy map through characteristic analysis of the target energy map, wherein the crystal contour line distribution map is the first crystal position distribution map and comprises contour line distribution positions of all crystals.

It should be understood that in practical applications, other types of crystal position distribution maps can also be used to train the crystal position distribution model, and accordingly, after the target energy map is input into the crystal position distribution model, the crystal position distribution model will output the type of crystal position distribution map as the first crystal position distribution map; the type of the crystal position distribution pattern output by the crystal position distribution model is not limited in any way in this application.

Step 203: and determining a crystal position lookup table according to the first crystal position distribution diagram.

And after the first crystal position distribution diagram output by the crystal position distribution model is obtained, a crystal position lookup table capable of reflecting the corresponding relation between the energy diagram and the actual distribution position of the scintillation crystal is further determined according to the first crystal position distribution diagram.

In particular, different methods can be selected according to the type of the first crystal position distribution map to determine the crystal position lookup table.

In a possible implementation manner, when the first crystal position distribution map is a crystal center position distribution map, that is, when the first crystal position distribution map includes distribution positions of respective crystal centers, the first crystal position distribution map may be divided according to the distribution positions of the respective crystal centers in the first crystal position distribution map, so as to generate the crystal position lookup table.

Specifically, when determining the distribution position of the crystal to which a certain crystal center belongs in the first crystal position distribution map, each crystal center adjacent to the crystal center may be determined in the first crystal position distribution map, and further, a dividing line between the crystal center and each crystal center adjacent to the crystal center may be determined, where the dividing line between the crystal center and each adjacent crystal center is actually the contour line of the crystal to which the crystal center belongs, so as to determine the contour line of each crystal in the first crystal position distribution map, that is, determine the crystal position lookup table corresponding to the first crystal position distribution map.

It should be understood that the above dividing manner is only an example, and in practical applications, other dividing manners may also be adopted to generate the crystal position lookup table according to the first crystal position distribution diagram including the distribution positions of the centers of the respective crystals, and the manner of dividing the first crystal position distribution diagram is not specifically limited herein.

In another possible implementation, when the first crystal position distribution map is a crystal contour distribution map, that is, when the first crystal position distribution map includes contour distribution positions of respective crystals, the first crystal position distribution map may be directly used as a crystal position lookup table. Since the essence of the crystal position lookup table is an image capable of representing the actual region position of each crystal, and the first crystal position distribution diagram including the contour line distribution positions of each crystal can already represent the actual region position of each crystal, when the first crystal position distribution diagram includes the contour line distribution positions of each crystal, the first crystal position distribution diagram can be directly used as the crystal position lookup table.

It should be understood that when the first crystal position distribution map is a crystal position distribution map of another type, another way of determining the crystal position lookup table based on the first crystal position distribution map may be correspondingly adopted, and no limitation is made to the way of determining the crystal position lookup table.

It should be noted that, in order to ensure the accuracy of the finally generated crystal position lookup table, after the first crystal position distribution map output by the crystal position distribution model is obtained, the first crystal position distribution map may be further determined to determine whether the first crystal position distribution map is accurate, and further, determine whether the crystal position lookup table can be generated according to the first crystal position distribution map.

Specifically, whether the number of crystals in the first crystal position distribution map is a preset number or not can be judged, if so, a crystal position lookup table is further determined according to the first crystal position distribution map, otherwise, an operator can be prompted to manually adjust the first crystal position distribution map, or the operator can be prompted to detect the energy map so as to determine whether the energy map has a problem or not; or, whether the crystal position in the first crystal position distribution diagram meets the position reasonableness condition can be judged, if yes, a crystal position lookup table is further determined according to the first crystal position distribution diagram, and if not, the related prompt can be given to an operator; or, whether the number of crystals in the first crystal position distribution map is a preset number or not and whether the crystal positions in the first crystal position distribution map meet the position rationality condition or not can be judged, if so, the crystal position lookup table is further determined according to the first crystal position distribution map, and otherwise, the related prompt can be performed on an operator.

When the first crystal position distribution diagram includes the distribution positions of the crystal centers, whether the number of the crystal centers in the first crystal position distribution diagram is a preset number or not can be judged, and/or whether the distribution positions of the crystal centers in the first crystal position distribution diagram meet a position rationality condition or not can be judged; when the first crystal position distribution diagram includes the contour line distribution positions of the respective crystals, it may be determined whether the regions framed by the contour lines in the first crystal position distribution diagram are a predetermined number, and/or whether the distribution positions of the contour lines in the first crystal position distribution diagram meet a position rationality condition.

It should be understood that the above-mentioned preset number is determined according to the number of scintillation crystals in the PET detector, and the preset number is generally equal to the number of scintillation crystals in the PET detector; the above-mentioned position rationality condition is determined on the basis of the actual distribution position of the scintillation crystals in the PET detector.

The method for generating the crystal position lookup table provides a new mode for generating the crystal position lookup table, and generates a crystal position distribution diagram corresponding to the energy diagram by using the crystal position distribution model, and further determines the crystal position lookup table according to the crystal position distribution diagram. Because the crystal position distribution model is a neural network obtained by training based on the energy map and a calibration crystal position distribution diagram corresponding to the energy map through a machine learning algorithm, errors caused by factors such as noise of the energy map, artifacts and uneven image gray scale can be automatically and correspondingly corrected in the process of performing feature analysis on the energy map by using the neural network, so that the crystal position distribution diagram generated based on the neural network has high accuracy, and the accuracy of a crystal position lookup table determined based on the crystal position distribution diagram is high.

It should be noted that, with the continuous use of the PET detector, the geometric distribution and intensity distribution of the energy diagram generated based on the PET detector will change to some extent under the influence of temperature and humidity changes in the environment, the aging of the scintillation crystal, the electronic circuit, and other factors; in general, most of training samples used in training the crystal position distribution model are data acquired at the initial stage of PET detector usage, and after the PET detector is used for a long time, the crystal position distribution map corresponding to the energy map may not be accurately determined by using the crystal position distribution model.

In view of the above situation, the embodiment of the present application further provides a method for generating a crystal position lookup table, so as to ensure that the crystal position lookup table can be accurately generated even if the PET detector is used for a long time. Referring to fig. 3, fig. 3 is a schematic flowchart of another method for generating a crystal position lookup table according to an embodiment of the present application; as shown in fig. 3, the method for generating the crystal position lookup table includes the following steps:

step 301: and acquiring radiation data, and generating a target energy map according to the radiation data.

Step 302: and generating a crystal position distribution diagram corresponding to the target energy diagram by using a crystal position distribution model to serve as a first crystal position distribution diagram.

The specific implementation manners of step 301 and step 302 are similar to the specific implementation manners of step 201 and step 202 in the embodiment shown in fig. 2, and refer to the related descriptions of step 201 and step 202 in detail, which is not described herein again.

Step 303: generating a crystal position distribution diagram as a second crystal position distribution diagram according to the light spot distribution characteristics in the target energy diagram; the second crystal position profile has the same crystal position attributes as in the first crystal position profile.

After the target energy map is generated, a crystal position distribution map can be correspondingly generated according to the light spot distribution characteristics in the target energy map, and the crystal position distribution map is used as a second crystal position distribution map; the second crystal position distribution map has the same crystal position attributes as those of the first crystal position distribution map generated in step 302, that is, if the first crystal position distribution map includes the distribution positions of the respective crystal centers, the second crystal position distribution map also includes the distribution positions of the respective crystal centers; if the first crystal position distribution map includes the contour line distribution positions of the respective crystals, the second crystal position distribution map also includes the contour line distribution positions of the respective crystals, and so on.

When the crystal position attribute of the first crystal position distribution map identifies that the crystal positions are distribution positions of the crystal centers, that is, when the first crystal position distribution map includes the distribution positions of the respective crystal centers, the second crystal position distribution map also needs to include the distribution positions of the respective crystal centers; at this time, image denoising processing may be performed on the target energy map to obtain a first reference energy map; then, the distribution positions of the respective crystal centers in the first reference energy map are determined by dividing or clustering the first reference energy map, thereby generating a crystal center position distribution map as a second crystal position distribution map.

When the crystal position attribute of the first crystal position distribution diagram identifies that the crystal position is the contour line distribution position of the crystal, namely when the first crystal position distribution diagram comprises the contour line distribution position of each crystal, the second crystal position distribution diagram also needs to comprise the contour line distribution position of each crystal; at this time, the target energy map may be subjected to gray inversion processing to obtain a second reference energy map; then, the second reference energy map is segmented by a watershed algorithm to generate a crystal contour distribution map as a second crystal position distribution map. When the second reference energy map is subjected to the segmentation processing, in addition to the watershed algorithm, other algorithms may also be used to perform the segmentation processing on the second reference energy map, and the algorithm used in the segmentation processing is not limited at all.

It should be understood that the crystal position attribute of the first crystal position distribution map may be a position attribute for identifying a distribution position where the crystal position is the center of the crystal or a contour line distribution position of the crystal, and may also be a position attribute for identifying other position attributes capable of characterizing the crystal position distribution characteristics, and accordingly, when the content identified by the crystal position attribute of the first crystal position distribution map is other crystal position distribution characteristics, the second crystal position distribution map may be generated according to the target energy map in a manner corresponding to the crystal position distribution characteristics, and the manner of generating the second crystal position distribution map is not limited in any way.

It should be noted that, in practical applications, step 302 may be executed first and then step 303 is executed, step 303 may be executed first and then step 302 is executed, and step 302 and step 303 may also be executed simultaneously, where the execution order of step 302 and step 303 is not limited at all. It will be appreciated that whatever execution order is employed, it is necessary to ensure that the crystal position attributes of the first crystal position profile generated via step 302 are the same as the second crystal position profile generated via step 303.

Step 304: calculating a target confidence matrix according to the first crystal position distribution diagram and the second crystal position distribution diagram; the target confidence matrix is used to characterize the confidence of the first crystal position distribution map.

After the first and second crystal position distribution maps are generated, a target confidence matrix is further calculated according to the first and second crystal position distribution maps, wherein the target confidence matrix is used for representing the confidence degree of the first crystal position distribution map, namely the accuracy degree of the first crystal position distribution map.

When the first crystal position distribution diagram includes the distribution positions of the crystal centers, the second crystal position distribution diagram correspondingly includes the distribution positions of the crystal centers, and at this time, the target confidence matrix needs to be calculated according to the distribution positions of the crystal centers in the first crystal position distribution diagram and the second crystal position distribution diagram; the formula for specifically calculating the target confidence matrix is shown as formula (1):

wherein (M)_ConfidenceM_atrix)_i，jFor the element in the ith row and the jth column in the target confidence matrix,is the distribution position of the ith row and jth column crystal center in the first crystal position distribution diagram,the distribution position of the ith row and the jth column crystal center in the second crystal position distribution diagram, and D is the maximum value used for limiting the distribution position difference of the crystal centers.

When the first crystal position distribution diagram includes contour line distribution positions of the crystals, the second crystal position distribution diagram correspondingly includes contour line distribution positions of the crystals, and at this time, a target confidence matrix is calculated according to the first crystal position distribution diagram and the region positions of the crystals framed by the contour lines in the second crystal position distribution diagram; the formula for specifically calculating the target confidence matrix is shown in formula (2):

wherein (M)_ConfidenceM_atrix)_i，jFor the element in the ith row and jth column of the target confidence matrix,is the area position of the ith row and jth column crystal in the first crystal position distribution diagram,is the area position of the ith row and jth column crystal in the second crystal position distribution diagram,is composed ofAndthe intersection area of (a).

It should be understood that the values of i in the equations (1) and (2) are both dependent on the number of rows of the crystal center included in the PET detector, i.e., the value of i ranges from 1 to the number of rows of the crystal center in the PET detector, i is an integer; similarly, the values of j in equations (1) and (2) depend on the number of columns of crystal centers included in the PET detector, i.e., j ranges from 1 to the number of columns of crystal centers in the PET detector, and j is an integer.

It should be understood that when the crystal position attributes of the first and second crystal position distribution maps are used to identify other position attributes that characterize the crystal position distribution, the target confidence matrix may be calculated in other manners, and the manner of calculating the target confidence matrix is not limited in any way.

Step 305: and generating a target crystal position distribution diagram according to the first crystal position distribution diagram, the second crystal position distribution diagram and the target confidence matrix.

After the target confidence matrix is determined, generating a target crystal position distribution diagram further according to the first crystal position distribution diagram, the second crystal position distribution diagram and the target confidence matrix; the first crystal position distribution diagram generated by the crystal position distribution model is corrected by utilizing the second crystal position distribution diagram and the target confidence matrix, so that a more accurate target crystal position distribution diagram is obtained.

When the first crystal position distribution diagram and the second crystal position distribution diagram both include the distribution positions of the respective crystal centers, the distribution positions of the respective crystal centers in the target crystal position distribution diagram may be determined according to the distribution positions of the respective crystal centers in the first crystal position distribution diagram, the distribution positions of the respective crystal centers in the second crystal position distribution diagram, and the target confidence matrix calculated by equation (1), and then the target crystal position distribution diagram may be generated according to the distribution positions of the respective crystal centers in the target crystal position distribution diagram; specifically, when the distribution position of each crystal center in the target crystal position distribution map is calculated, the formula shown in formula (3) may be adopted:

wherein,the distribution position of the ith row and jth column crystal center in the target crystal position distribution diagram; (M)_ConfidenceM_atrix)_i，jThe element of the ith row and the jth column in the target confidence matrix is calculated by the formula (1);is the distribution position of the ith row and jth column crystal center in the first crystal position distribution diagramPlacing;the distribution position of the ith row and the jth column crystal center in the second crystal position distribution diagram, D is the maximum value used for limiting the distribution position difference of the crystal centers, and α is a parameter used for adjusting the weight occupied by the first crystal position distribution diagram.

When the first crystal position distribution diagram and the second crystal position distribution diagram both include the contour line distribution positions of the respective crystals, determining the region positions of the respective crystals in the target crystal position distribution diagram according to the region positions of the respective crystals in the first crystal position distribution diagram, the region positions of the respective crystals in the second crystal position distribution diagram, and the target confidence matrix calculated by the formula (2), and further generating the target crystal position distribution diagram according to the region positions of the respective crystals in the target crystal position distribution diagram; specifically, when calculating the region position of each crystal in the target crystal position distribution diagram, the formula shown in formula (4) may be adopted:

wherein,the target area position of the ith row and the jth column crystal in the target crystal position distribution diagram is obtained; (M)_ConfidenceM_atrix)_i，jThe element of the ith row and the jth column in the target confidence matrix is calculated by the formula (2);is the area position of the ith row and jth column crystal in the first crystal position distribution diagram,is the area position of the ith row and jth column crystal in the second crystal position distribution diagram,is composed ofAndα is a parameter used to adjust the weight taken by the first crystal position profile.

It should be understood that the values of i in the equations (3) and (4) are both dependent on the number of rows of crystal centers included in the PET detector, i.e., i ranges from 1 to the number of rows of crystal centers in the PET detector, and i is an integer; similarly, the values of j in equations (3) and (4) depend on the number of columns of crystal centers included in the PET detector, i.e., j ranges from 1 to the number of columns of crystal centers in the PET detector, and j is an integer.

It should be understood that when the crystal position attributes of the first and second crystal position distribution maps are used to identify other position attributes that can characterize the crystal position distribution, other manners of determining the target crystal position distribution map based on the first and second crystal position distribution maps and the target confidence matrix may be correspondingly employed, without any limitation on the manner of determining the target crystal position distribution map.

Step 306: and determining the crystal position lookup table according to the target crystal position distribution diagram.

And further determining a crystal position lookup table according to the target crystal position distribution diagram after generating the target crystal position distribution diagram according to the first crystal position distribution diagram, the second crystal position distribution diagram and the target confidence matrix. In specific implementation, different methods can be correspondingly selected according to the crystal position attribute in the target crystal position distribution diagram to determine the crystal position lookup table.

In a possible implementation manner, when the crystal position attribute of the target crystal position distribution map identifies that the crystal position is a distribution position of the crystal center, that is, when the target crystal position distribution map includes distribution positions of the respective crystal centers, the target crystal position distribution map may be divided according to the distribution positions of the respective crystal centers in the target crystal position distribution map, so as to generate the crystal position lookup table.

In another possible implementation, when the crystal position attribute of the target crystal position distribution map identifies that the crystal position is the contour line distribution position of the crystal, that is, when the target crystal position distribution map includes the contour line distribution position of each crystal, the target crystal position distribution map may be directly used as the crystal position lookup table.

It should be understood that when the target crystal position distribution map is used to identify other position attributes that can characterize the crystal position distribution, other manners of determining the crystal position lookup table from the target crystal position distribution map may be correspondingly employed, and no limitation is made on the manner of determining the crystal position lookup table.

It should be noted that, in order to ensure the accuracy of the finally generated crystal position lookup table, after the target crystal position distribution map is obtained, the target crystal position distribution map may be further determined to determine whether the target crystal position distribution map is accurate, and further, determine whether the crystal position lookup table can be generated according to the target crystal position distribution map.

Specifically, whether the number of crystals in the target crystal position distribution map is a preset number or not can be judged, if so, a crystal position lookup table is further determined according to the target crystal position distribution map, otherwise, an operator can be prompted to manually adjust the first crystal position distribution map, or the operator can be prompted to detect the energy map so as to determine whether the energy map has problems or not; or, whether the crystal position in the target crystal position distribution diagram meets the position reasonableness condition can be judged, if yes, a crystal position lookup table is further determined according to the target crystal position distribution diagram, and if not, the relevant prompt can be given to an operator; or, whether the number of crystals in the target crystal position distribution map is a preset number or not and whether the crystal positions in the target crystal position distribution map meet the position rationality condition or not can be judged, if so, a crystal position lookup table is further determined according to the target crystal position distribution map, and otherwise, the related prompt can be performed on an operator.

When the target crystal position distribution diagram includes the distribution positions of the crystal centers, whether the number of the crystal centers in the target crystal position distribution diagram is a preset number or not can be judged, and/or whether the distribution positions of the crystal centers in the target crystal position distribution diagram meet a position rationality condition or not can be judged; when the target crystal position distribution diagram includes the contour line distribution positions of the crystals, whether the regions framed by the contour lines in the target crystal position distribution diagram are in the preset number or not can be judged, and/or whether the distribution positions of the contour lines in the target crystal position distribution diagram meet the position rationality condition or not can be judged.

It should be noted that, the crystal position distribution model may determine the corresponding crystal position distribution map according to the energy map, and may also determine a confidence matrix corresponding to the crystal position distribution map, where the confidence matrix is used to characterize the credibility of the crystal position distribution map output by the crystal position distribution model; i.e. the output of the model of the crystal position distribution also includes the confidence matrix.

Accordingly, after the target crystal position distribution map is determined, the target energy map obtained in step 301, the target confidence matrix obtained in step 304, and the target crystal position distribution map obtained in step 305 may be obtained, and an optimization sample is formed by using the target energy map, the target confidence matrix, and the target crystal position distribution map and added to the optimization sample set; so as to further optimize and train the crystal position distribution model by using the optimized sample set.

Therefore, the corrected data are obtained to further optimize and train the crystal position model, and the model performance of the crystal position distribution model can be continuously optimized, so that the crystal position distribution model can be suitable for various conditions, and the accuracy of the crystal position distribution map generated by the crystal position distribution model under various conditions is high.

In the method for generating the crystal position lookup table, a crystal position distribution model is used for generating a first crystal position distribution map according to a target energy map, a second crystal position distribution map is generated according to light spot characteristics in the target energy map, the target crystal position distribution map is determined according to the first crystal position distribution map and the second crystal position distribution map, and then the crystal position lookup table is determined according to the target crystal position distribution map. In the case of a long-term use of the PET detector, the first crystal position distribution map may be corrected using the second crystal position distribution map, thereby ensuring a high accuracy of the determined target crystal position distribution map and thus of the crystal position look-up table determined based on the target crystal position distribution map.

It should be understood that whether the crystal position distribution model can accurately determine the crystal position distribution map corresponding to the energy map depends on the model performance of the crystal position distribution model, and the model performance of the crystal position distribution model depends on the training process of the crystal position distribution model.

A model training method of the crystal position distribution model is described below, referring to fig. 4, where fig. 4 is a schematic flow chart of the model training method provided in the embodiment of the present application; as shown in fig. 4, the model training method includes the following steps:

step 401: and determining a training sample set, wherein each sample in the training sample set comprises an energy map and a calibration crystal position distribution map corresponding to the energy map.

When training the crystal position distribution model, a training sample set for training the crystal position distribution model needs to be obtained first, the training sample set usually includes a large number of training samples, and each training sample is composed of an energy map and a calibration crystal position distribution map corresponding to the energy map.

It should be understood that the energy map in the training sample may be generated from radiation data generated by decay of the radioactive source, or from radiation data generated by background radiation, and the data source of the energy map in the training sample is not limited in any way.

Specifically, when the calibration crystal position distribution map in the training sample is obtained, the crystal position distribution map is generated according to the energy map by adopting a mode of generating the crystal position distribution map in the prior art, and then the crystal position distribution map is further corrected according to artificial experience to obtain the calibration crystal position distribution map, so that the accuracy of the calibration crystal position distribution map is ensured.

It should be noted that the calibration crystal position distribution diagram may include distribution positions of the centers of the respective crystals, may also include distribution positions of the contour lines of the respective crystals, and may also include other crystal distribution position features, where the crystal distribution position features included in the calibration crystal position distribution diagram are not limited at all.

It should be noted that, the training sample may include an energy map and a calibration crystal position distribution map corresponding to the energy map, and may further include a calibration confidence matrix, where the calibration confidence matrix is used to characterize the confidence of the calibration crystal position distribution map; since the calibration crystal position distribution map is usually calibrated manually, the initial value of each element in the calibration confidence matrix is usually set to 1 to characterize that the calibration crystal position distribution map is absolutely reliable.

Of course, the initial values of the elements in the calibration confidence matrix may also be set to other values according to the actual confidence level of the calibration crystal position distribution map, and no limitation is made on the values of the elements in the calibration confidence matrix.

Step 402: training a neural network by using the training sample set to obtain a crystal position distribution model through training; the crystal position distribution model takes an energy map as input and takes a crystal position distribution map corresponding to the energy map as output.

When training the crystal position distribution model, a neural network model needs to be constructed in advance as the trained crystal position distribution model, and the structure of the neural network model is the same as that of the crystal position distribution model put into practical application; in the specific training, deep learning algorithms such as clustering and convolutional neural network can be adopted, the training samples in the training sample set determined in step 401 are utilized to train the pre-constructed neural network model, the model parameters of the neural network model are continuously optimized, when the neural network model meets the training end condition, a crystal position distribution model which can be put into practical application can be constructed according to the neural network model structure and the model parameters at the moment, the crystal position distribution model takes the energy map as input, and the crystal position distribution map corresponding to the energy map as output.

In a possible implementation manner, a neural network model can be correspondingly constructed in advance according to the size of the input energy map and the number of crystals in the calibration crystal position distribution map, and the neural network model generally comprises a plurality of neural network layers; for example, assuming that the input of the neural network model is an energy map of 256 × 256 (pixels), and the nominal crystal position distribution map corresponding to the energy map includes 11 × 11 crystals, a neural network model including 5 neural network layers may be constructed accordingly, where the input layer of the neural network model includes 256 × 256 network nodes, the second layer, the third layer, and the fourth layer include M, M/2 and M/4 network nodes, respectively, and the output layer includes 11 × 11 network nodes, where M may be set according to actual requirements, and no limitation is made on the specific value of M.

After the neural network model is built, initialization weight parameters can be given to each network node in each neural network layer in the neural network model in a random initialization mode, and in the process of training the neural network model by using a training sample, the respective winning weight parameters of each network node in the neural network model are continuously adjusted according to an objective function so as to continuously optimize the performance of the neural network model, so that the light spot distribution characteristics in the energy diagram can be better learned, and the crystal position distribution diagram corresponding to the energy diagram can be more accurately determined. When judging whether the neural network model meets the training end condition, verifying a first model by using a test sample, wherein the first model is obtained by performing first round training optimization on the neural network model by using training samples in a training sample set; specifically, an energy map in a test sample is input into the first model, and the input energy map is correspondingly processed by using the first model to obtain a predicted crystal position distribution map; and then, calculating the prediction accuracy according to the calibration crystal position distribution diagram in the test sample and the predicted crystal position distribution diagram output by the first model, and when the prediction accuracy is greater than a preset threshold value, determining that the model performance of the first model is better and can meet the requirement, and generating a crystal position distribution model according to the model parameters and the model structure of the first model.

It should be noted that the preset threshold may be set according to actual situations, and the preset threshold is not specifically limited herein.

In addition, when judging whether the neural network model meets the training end condition, whether the model is continuously trained or not can be determined according to a plurality of models obtained through a plurality of rounds of training so as to obtain a crystal position distribution model with optimal model performance. Specifically, a plurality of neural network models obtained through a plurality of rounds of training can be verified respectively by using test samples, if the difference between the prediction accuracy rates of the models obtained through each round of training is judged to be small, the performance of the neural network model is considered to have not improved a space, the neural network model with the highest prediction accuracy rate can be selected, and a crystal position distribution model is determined according to model parameters and a model structure of the neural network model; if the prediction accuracy rates of the neural network models obtained through the training of the rounds have larger differences, the performance of the neural network model is considered to have a space for improvement, and the neural network model can be continuously trained until the crystal position distribution model with the most stable and optimal model performance is obtained.

It should be noted that, when the training sample includes the calibration confidence matrix, it may be determined whether the neural network model satisfies the training end condition based on the crystal position distribution map and the confidence matrix output by the neural network model, and the specific determination method needs to calculate the prediction accuracy of the confidence matrix based on the determination method and further based on the confidence matrix and the calibration confidence matrix output by the neural network model, and determine whether the neural network model satisfies the training end condition by combining the prediction accuracy of the crystal position distribution map and the prediction accuracy of the confidence matrix.

It should be understood that when the calibration confidence matrix is included in the training sample, the output of the crystal position distribution model finally trained based on the training sample also includes a confidence matrix that is used to characterize the confidence of the crystal position distribution map output by the crystal position distribution model.

It should be noted that, after the crystal position distribution model obtained through the training by the above training method is put into practical application, the crystal position distribution model can be further optimized and trained by using the data obtained in practical application.

Specifically, in the process of generating the crystal position lookup table by using the method shown in fig. 3, a target confidence matrix and a target crystal position distribution map are generated, and an optimization sample may be formed by using the target energy map, the target confidence matrix and the target crystal position distribution map, and is added to the optimization sample set; furthermore, when the crystal position distribution model needs to be further optimally trained, the crystal position model can be optimally trained by utilizing the optimal sample set.

In the model training method, a training sample set is obtained first, and training samples in the training sample set all comprise an energy diagram and a calibration crystal position distribution diagram corresponding to the energy diagram; further, the neural network is trained using the training sample set to obtain a crystal position distribution model having the energy map as an input and the crystal position distribution map corresponding to the energy map as an output. Because the crystal position distribution model is a neural network obtained by training based on the energy diagram and the calibration crystal position distribution diagram corresponding to the energy diagram through a machine learning algorithm, errors caused by factors such as noise of the energy diagram, artifacts and uneven image gray scale can be automatically and correspondingly corrected in the process of carrying out feature analysis on the energy diagram by using the neural network, so that the crystal position distribution diagram generated based on the neural network is ensured to have higher accuracy, and the accuracy of the crystal position lookup table determined based on the crystal position distribution diagram is ensured to be higher.

In order to further understand the model training method and the crystal position lookup table generation method provided in the above embodiments, the method provided in the embodiments of the present application is described below in conjunction with fig. 5.

The method provided by the embodiment of the application mainly comprises two parts: a training part and a recognition part.

The training part is mainly used for training to obtain a crystal position distribution model capable of identifying a corresponding crystal position distribution diagram according to an energy diagram; when a crystal position distribution model is trained, a training sample set is required to be obtained first, wherein training samples in the training sample set comprise an energy diagram, a calibration crystal position distribution diagram corresponding to the energy diagram and a calibration confidence matrix; and training the neural network model by using the training sample set so as to continuously optimize the model parameters of the neural network model, and generating a crystal position distribution model which can be put into practical application according to the structure and the model parameters of the neural network model when the model performance of the neural network model meets the training finishing condition.

Wherein the main purpose of the identification portion is to generate a crystal position look-up table from the collected radiation data; in specific implementation, a target energy map can be generated according to the acquired radiation data; then, inputting the target energy diagram into a crystal position distribution model obtained through training in a training process, and obtaining a first crystal position distribution diagram output by the crystal position distribution model; generating a second crystal position distribution diagram according to the distribution characteristics of the light spots in the target energy diagram; then, generating a target confidence matrix according to the first crystal position distribution diagram and the second crystal position distribution diagram, and generating a target crystal position distribution diagram by using the first crystal position distribution diagram, the second crystal position distribution diagram and the target confidence matrix; finally, a crystal position look-up table is determined according to the target crystal position distribution diagram.

In addition, after the target crystal position distribution diagram is determined, an optimization sample can be formed by using the target energy diagram, the target confidence matrix and the target crystal position distribution diagram, the optimization sample is added into an optimization sample set, and the optimization sample set can be used for further optimization training of the crystal position distribution model.

In view of the above-described method for generating a crystal position lookup table, the present application also provides a corresponding apparatus for generating a crystal position lookup table, so that the method for generating a crystal position lookup table can be applied and implemented in practice.

Referring to fig. 6, fig. 6 is a schematic structural diagram of a crystal position look-up table generating device 600 corresponding to the crystal position look-up table generating method shown in fig. 2, the crystal position look-up table generating device 600 includes:

the acquisition module 601 is configured to acquire radiation data and generate a target energy map according to the radiation data;

a first processing module 602, configured to generate a crystal position distribution map corresponding to the target energy map by using a crystal position distribution model, as a first crystal position distribution map; the crystal position distribution model is a neural network, is obtained based on an energy map and a calibration crystal position distribution map corresponding to the energy map through training, and takes the energy map as input and the crystal position distribution map corresponding to the energy map as output;

a determining module 603 configured to determine a crystal position look-up table according to the first crystal position distribution map.

Optionally, the first crystal position distribution map includes distribution positions of respective crystal centers;

the determining module 603 is specifically configured to:

according to the distribution position of each crystal center in the first crystal position distribution diagram, carrying out segmentation processing on the first crystal position distribution diagram to generate the crystal position lookup table;

or,

the first crystal position distribution diagram comprises contour line distribution positions of all crystals;

the determining module 603 is specifically configured to:

using the first crystal position distribution map as the crystal position lookup table.

Optionally, the apparatus further comprises:

the first verification module is used for judging whether the number of the crystals in the first crystal position distribution diagram is a preset number or not; if yes, triggering the determining module 603 to determine the crystal position lookup table;

or,

a second verification module, configured to determine whether a crystal position in the first crystal position distribution map meets a position reasonableness condition, and if yes, trigger the determination module 603 to determine the crystal position lookup table;

or,

a third verification module, configured to determine whether the number of crystals in the first crystal position distribution map is a preset number, and determine whether the crystal positions in the first crystal position distribution map meet a position rationality condition, if both are true, trigger the determination module 603 to determine the crystal position lookup table.

Optionally, the apparatus further comprises:

the second processing module is used for generating a crystal position distribution diagram according to the light spot distribution characteristics in the target energy diagram, and the crystal position distribution diagram is used as a second crystal position distribution diagram; the second crystal position profile has the same crystal position attributes as in the first crystal position profile;

the determining module 603 comprises:

the confidence coefficient calculation submodule is used for calculating a target confidence matrix according to the first crystal position distribution diagram and the second crystal position distribution diagram; the target confidence matrix is used for characterizing the confidence of the first crystal position distribution map;

a crystal position distribution diagram generation submodule, configured to generate a target crystal position distribution diagram according to the first crystal position distribution diagram, the second crystal position distribution diagram, and the target confidence matrix;

and the determining submodule is used for determining the crystal position lookup table according to the target crystal position distribution diagram.

Optionally, the crystal position attribute of the first crystal position distribution map identifies the crystal position as a distribution position of the crystal center;

the second processing module is specifically configured to:

performing image noise reduction processing on the target energy map to obtain a first reference energy map;

determining the distribution position of each crystal center in the first reference energy map by carrying out segmentation or clustering processing on the first reference energy map;

generating a crystal center position distribution diagram as a second crystal position distribution diagram according to the distribution positions of the crystal centers in the first reference energy diagram;

the determination submodule is specifically configured to:

according to the distribution position of each crystal center in the target crystal position distribution diagram, carrying out segmentation processing on the target crystal position distribution diagram to generate the crystal position lookup table;

or,

the crystal position attribute of the first crystal position distribution diagram identifies the crystal position as a contour line distribution position of the crystal;

the second processing module is specifically configured to:

carrying out gray level inversion on the target energy diagram to obtain a second reference energy diagram;

performing segmentation processing on the second reference energy map by using a watershed algorithm to generate a crystal contour distribution map serving as a second crystal position distribution map;

the determination submodule is specifically configured to:

and taking the target crystal position distribution map as the crystal position lookup table.

Optionally, the determining module 603 further includes:

the first verification submodule is used for judging whether the number of crystals in the target crystal position distribution diagram is a preset number or not, and if yes, the determination submodule is triggered to generate a crystal position lookup table;

or,

the second verification submodule is used for judging whether the crystal position in the target crystal position distribution diagram meets the position reasonability condition or not, and if so, the determination submodule is triggered to generate a crystal position lookup table;

or,

and the third verification submodule is used for judging whether the number of crystals in the target crystal position distribution diagram is a preset number or not and judging whether the crystal positions in the target crystal position distribution diagram meet a position rationality condition or not, and if so, triggering the determination submodule to generate a crystal position lookup table.

Optionally, the output of the crystal position distribution model further includes a confidence matrix;

the apparatus further comprises:

the sample construction module is used for forming an optimized sample by utilizing the target energy map, the target crystal position distribution map and the target confidence matrix and adding the optimized sample into an optimized sample set; and the optimization sample set is used for performing optimization training on the crystal position distribution model.

The crystal position lookup table generation device provides a new mode for generating the crystal position lookup table, and utilizes the crystal position distribution model to generate the crystal position distribution diagram corresponding to the energy diagram, and further determines the crystal position lookup table according to the crystal position distribution diagram. Because the crystal position distribution model is a neural network obtained by training based on the energy map and a calibration crystal position distribution diagram corresponding to the energy map through a machine learning algorithm, errors caused by factors such as noise of the energy map, artifacts and uneven image gray scale can be automatically and correspondingly corrected in the process of performing feature analysis on the energy map by using the neural network, so that the crystal position distribution diagram generated based on the neural network has high accuracy, and the accuracy of a crystal position lookup table determined based on the crystal position distribution diagram is high.

Aiming at the model training method described above, the present application also provides a corresponding model training device, so that the above model training method can be applied and implemented in practice.

Referring to fig. 7, fig. 7 is a schematic structural diagram of a model training apparatus 700 corresponding to the model training method shown in fig. 4, where the model training apparatus 700 includes:

a sample determining module 701, configured to determine a training sample set, where each training sample in the training sample set includes an energy map and a calibration crystal position distribution map corresponding to the energy map;

a training module 702, configured to train a neural network by using the training sample set to obtain a crystal position distribution model through training; the crystal position distribution model takes an energy map as input and takes a crystal position distribution map corresponding to the energy map as output.

Optionally, the calibration crystal position distribution map includes distribution positions of respective crystal centers;

or the calibration crystal position distribution diagram comprises contour line distribution positions of all crystals.

Optionally, the training sample set further includes a calibration confidence matrix; the calibration confidence matrix is used for representing the confidence of the calibration crystal position distribution diagram;

the output of the crystal position distribution model further comprises a confidence matrix; the confidence matrix is used to characterize the confidence of the output crystal position distribution map.

Optionally, the apparatus further comprises:

the optimization training module is used for performing optimization training on the crystal position distribution model by adopting an optimization sample set; the optimized sample set is from the sample construction module shown in fig. 6.

In the model training device, a training sample set is obtained first, and training samples in the training sample set all comprise an energy diagram and a calibration crystal position distribution diagram corresponding to the energy diagram; further, the neural network is trained using the training sample set to obtain a crystal position distribution model having the energy map as an input and the crystal position distribution map corresponding to the energy map as an output. Because the crystal position distribution model is a neural network obtained by training based on the energy diagram and the calibration crystal position distribution diagram corresponding to the energy diagram through a machine learning algorithm, errors caused by factors such as noise of the energy diagram, artifacts and uneven image gray scale can be automatically and correspondingly corrected in the process of carrying out feature analysis on the energy diagram by using the neural network, so that the crystal position distribution diagram generated based on the neural network is ensured to have higher accuracy, and the accuracy of the crystal position lookup table determined based on the crystal position distribution diagram is ensured to be higher.

The application also provides a device for generating the crystal position lookup table, which can be a server; referring to fig. 8, fig. 8 is a schematic diagram of a server structure for generating a crystal position lookup table according to an embodiment of the present application, where the server 800 may have a relatively large difference due to different configurations or performances, and may include one or more Central Processing Units (CPUs) 822 (e.g., one or more processors) and a memory 832, and one or more storage media 830 (e.g., one or more mass storage devices) storing an application 842 or data 844. Memory 832 and storage medium 830 may be, among other things, transient or persistent storage. The program stored in the storage medium 830 may include one or more modules (not shown), each of which may include a series of instruction operations for the server. Still further, a central processor 822 may be provided in communication with the storage medium 830 for executing a series of instruction operations in the storage medium 830 on the server 800.

The server 800 may also include one or more power supplies 826, one or more wired or wireless network interfaces 850, one or more input-output interfaces 858, and/or one or more operating systems 841, such as Windows Server, Mac OS XTM, UnixTM, LinuxTM, FreeBSDTM, and so forth.

The steps performed by the server in the above embodiments may be based on the server structure shown in fig. 8.

The CPU822 is configured to execute the following steps:

Optionally, CPU822 may also execute the method steps of any specific implementation of the method for generating a crystal position lookup table in this embodiment of the present application.

The application also provides equipment for training the model, and the equipment can be specifically a server; the structure of the device is similar to that of the server shown in fig. 8, and is not described herein again; wherein, CPU is used for carrying out the following step:

Optionally, the CPU may further execute the method steps of any specific implementation of the model training method in the embodiment of the present application.

The embodiment of the present application further provides another apparatus for generating a crystal position lookup table, where the apparatus may be a terminal apparatus, as shown in fig. 9, for convenience of description, only a portion related to the embodiment of the present application is shown, and details of the specific technology are not disclosed, please refer to the method portion of the embodiment of the present application. The terminal may be a terminal device including a computer, taking the terminal as the computer as an example:

fig. 9 is a block diagram showing a partial structure of a computer related to a terminal provided in an embodiment of the present application. Referring to fig. 9, the computer includes: radio Frequency (RF) circuit 910, memory 920, input unit 930, display unit 940, sensor 950, audio circuit 960, wireless fidelity (WiFi) module 970, processor 980, and power supply 990; the input unit 930 includes an input panel 931 and other input devices 932, the display unit 940 includes a display panel 941, and the audio circuit 960 includes a speaker 961 and a microphone 962.

Those skilled in the art will appreciate that the computer architecture shown in FIG. 9 is not limiting and may include more or fewer components than those shown, or some components may be combined, or a different arrangement of components.

In the embodiment of the present application, the terminal includes a processor 980 having the following functions:

Optionally, the processor 980 may further perform the method steps of any specific implementation of the method for generating the crystal position lookup table in the embodiment of the present application.

The application also provides equipment for training the model, which can be terminal equipment; the structure of the device is similar to that of the terminal device shown in fig. 9, and is not described herein again; wherein the processor is configured to perform the steps of:

Optionally, the processor may further execute the method steps of any specific implementation of the model training method in the embodiment of the present application.

The present application further provides a computer-readable storage medium for storing a program code for executing any one of the embodiments of the crystal position lookup table generation method or any one of the embodiments of the model training method described in the foregoing embodiments.

The present embodiments also provide a computer program product including instructions, which when run on a computer, cause the computer to execute any one of the embodiments of the crystal position lookup table generation method or any one of the model training methods described in the foregoing embodiments.

It is clear to those skilled in the art that, for convenience and brevity of description, the specific working processes of the above-described systems, apparatuses and units may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.

In the several embodiments provided in the present application, it should be understood that the disclosed system, apparatus and method may be implemented in other manners. For example, the above-described apparatus embodiments are merely illustrative, and for example, the division of the units is only one logical division, and other divisions may be realized in practice, for example, a plurality of units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or units, and may be in an electrical, mechanical or other form.

The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.

In addition, functional units in the embodiments of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.

The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present application may be substantially implemented or contributed to by the prior art, or all or part of the technical solution may be embodied in a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present application. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.

The above embodiments are only used for illustrating the technical solutions of the present application, and not for limiting the same; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions in the embodiments of the present application.

Claims

1. A method for generating a crystal position look-up table, the method comprising:

2. The method of claim 1, wherein the first crystal position profile comprises distribution positions of respective crystal centers;

determining a crystal position look-up table from the first crystal position distribution map, comprising:

or,

3. The method according to claim 1 or 2, wherein prior to said determining a crystal position look-up table from said first crystal position distribution map, the method further comprises:

judging whether the number of crystals in the first crystal position distribution diagram is a preset number, if so, executing the step of determining a crystal position lookup table according to the first crystal position distribution diagram;

or,

judging whether the crystal positions in the first crystal position distribution diagram meet position reasonableness conditions or not, if so, executing the step of determining a crystal position lookup table according to the first crystal position distribution diagram;

or,

and judging whether the number of the crystals in the first crystal position distribution diagram is a preset number or not, judging whether the crystal positions in the first crystal position distribution diagram meet position rationality conditions or not, and if so, executing the step of determining a crystal position lookup table according to the first crystal position distribution diagram.

4. The method of claim 1, wherein prior to the determining a crystal position look-up table from the first crystal position distribution map, the method further comprises:

generating a crystal position distribution diagram as a second crystal position distribution diagram according to the light spot distribution characteristics in the target energy diagram; the second crystal position profile has the same crystal position attributes as in the first crystal position profile;

calculating a target confidence matrix according to the first crystal position distribution diagram and the second crystal position distribution diagram; the target confidence matrix is used for characterizing the confidence of the first crystal position distribution map;

generating a target crystal position distribution diagram according to the first crystal position distribution diagram, the second crystal position distribution diagram and the target confidence matrix;

and determining the crystal position lookup table according to the target crystal position distribution diagram.

5. The method of claim 4, wherein the crystal position attribute of the first crystal position profile identifies a distribution position where a crystal position is a crystal center;

generating a crystal position distribution diagram according to the light spot distribution characteristics in the target energy diagram, wherein the crystal position distribution diagram is used as a second crystal position distribution diagram and comprises the following steps:

determining the crystal position look-up table according to the target crystal position distribution map, including:

or,

6. The method of claim 4 or 5, wherein after the generating the target crystal position profile, the method further comprises:

judging whether the number of crystals in the target crystal position distribution diagram is a preset number or not, if so, executing the step of generating a crystal position lookup table according to the target crystal position distribution diagram;

or,

judging whether the crystal position in the target crystal position distribution diagram meets a position reasonableness condition or not, if so, executing the step of generating a crystal position lookup table according to the target crystal position distribution diagram;

or,

and judging whether the number of crystals in the target crystal position distribution diagram is a preset number or not, judging whether the crystal positions in the target crystal position distribution diagram meet position rationality conditions or not, and if so, executing the step of generating a crystal position lookup table according to the target crystal position distribution diagram.

7. The method of claim 4 or 5, wherein the output of the crystal position distribution model further comprises a confidence matrix;

the method further comprises:

forming an optimization sample by using the target energy map, the target crystal position distribution map and the target confidence matrix, and adding the optimization sample into an optimization sample set; and the optimization sample set is used for performing optimization training on the crystal position distribution model.

8. A method of model training, the method comprising:

9. The method of claim 8, wherein the calibration crystal position profile comprises distribution positions of respective crystal centers;

10. The method according to claim 8 or 9, wherein the set of training samples further comprises a calibration confidence matrix; the calibration confidence matrix is used for representing the confidence of the calibration crystal position distribution diagram;

11. The method of claim 10, further comprising:

carrying out optimization training on the crystal position distribution model by adopting an optimization sample set; the optimized sample set is generated by the method of claim 7.

12. An apparatus for generating a crystal position lookup table, the apparatus comprising:

13. A model training apparatus, the apparatus comprising:

14. An apparatus, comprising a processor and a memory:

the processor is configured to execute the method for generating a crystal position lookup table according to any one of claims 1 to 7 or the method for training a model according to any one of claims 8 to 11 according to instructions in the program code.

15. A computer-readable storage medium comprising instructions which, when run on a computer, cause the computer to perform the method of generating a crystal position lookup table of any of claims 1 to 7 or the method of model training of any of claims 8 to 11.