WO2021000674A1

WO2021000674A1 - Cell image recognition method and system, computer device, and readable storage medium

Info

Publication number: WO2021000674A1
Application number: PCT/CN2020/093042
Authority: WO
Inventors: 庞烨; 王义文; 王健宗
Original assignee: 平安科技（深圳）有限公司
Priority date: 2019-07-03
Filing date: 2020-05-28
Publication date: 2021-01-07
Also published as: CN110502970A

Abstract

An embodiment of the present application provides a cell image recognition method employing a CapsNet capsule network. The method comprises: inputting a cell image to a CapsNet capsule network; performing a convolution operation on the cell image by means of a convolution layer to acquire A convolution feature maps; performing a convolution operation on the A convolution feature maps by means of a first capsule layer to acquire B multidimensional vectors; performing a full connection operation on the B multidimensional vectors by means of a second capsule layer to acquire C activation vectors, wherein the C activation vectors indicate that C cell categories are present; and calculating, according to the C activation vectors outputted by the second capsule layer, the probability that a cell in the cell image corresponds to each cell category. The CapsNet capsule network structure of the embodiment of the present application identifies a cell category, and performs a cell recognition operation and maintains a high recognition accuracy rate in cases where a biomedical image data set is small, cells are closely spaced, and multiple cells are present in a single image.

Description

Cell picture recognition method, system, computer equipment and readable storage medium

This application affirms the priority of the Chinese patent application filed with the Chinese Patent Office on July 3, 2019, the application number is 201910596892.9, and the name is "Cell image recognition method, system, computer equipment and readable storage medium", the entire content of which is incorporated by reference Incorporated in this application.

Technical field

The embodiments of the present application relate to the field of artificial intelligence, and in particular to a method, system, computer equipment, and computer-readable storage medium for cell image recognition based on CapsNet capsule network.

Background technique

Cell classification usually requires the help of biochemical and light path equipment. Currently, researchers mostly use microscope images to collect patient serum images. Doctors use the serum images to check for the presence of antibodies to determine whether there are autoimmune diseases. Relying on the doctor's judgment is a very subjective method and highly dependent on the doctor's experience, so it is not easy to achieve an efficient and accurate diagnosis. Therefore, the field of medical diagnosis urgently needs automatic microscope image processing technology and cell classification technology to assist doctors in realizing convenient and efficient medical diagnosis. Deep learning has attracted much attention since its introduction, and deep learning technology has begun to be applied to cell classification technology. Using deep learning to classify cells can help researchers better understand cell characteristics, establish genome connections, accelerate drug development, and even discover details that humans cannot perceive.

In recent years, the CNN model has become more and more mature, and researchers continue to try to use the CNN model to classify cells, but the CNN model still has many problems. First, training the CNN model requires a large amount of data, but for many problems, a large number of data sets cannot be obtained. For example, for many unpopular classifications, there is no perfect data set to use; secondly, the CNN model cannot solve the image well. In the case of overlapping objects, in the microscopic world, the density of cells may be very high. If they are close together, CNN cannot identify their characteristics. At the same time, CNN uses a large-scale pooling layer, although the pooling layer can Help extract the main information and quickly reduce the size of the image, but the information is lost. It is very likely that the detailed features around the main feature will be lost. Finally, CNN can only extract the features of the image, but cannot learn between the features When CNN learns the classification task, it may only recognize the eyes and nose and then judge it as a human face, but the eyes may be under the nose, which will lead to misjudgment.

On the issue of cell classification, the above shortcomings of CNN are very obvious. First of all, there is a lack of complete and mature data sets for cell types; cells may be very close to each other, and it is difficult to separate them individually. At the same time, it will also cause multiple cells in the same picture, and CNN cannot learn from one picture. Features of multiple categories; at the same time, in multi-label classification, CNN cannot learn the relationship between cells, resulting in misclassification.

Therefore, the inventor realized that it is necessary to provide a new cell identification technology to improve the accuracy of cell identification.

Summary of the invention

In view of this, the purpose of the embodiments of the present application is to provide a method, system, computer equipment, and computer-readable storage medium for cell image recognition based on the CapsNet capsule network to solve the problems of cell recognition errors and low recognition accuracy.

In order to achieve the foregoing objective, an embodiment of the present application provides a method for recognizing a cell image based on a CapsNet capsule network, which includes the following steps:

Input the cell picture into the CapsNet capsule network;

Performing a convolution operation on the cell image through a convolution layer to obtain A convolution feature maps;

Performing a convolution operation on the A convolution feature maps through the first capsule layer to obtain B multidimensional vectors;

Perform a full connection operation on the B multi-dimensional vectors through the second capsule layer to obtain C activation vectors, where the C activation vectors indicate that there are C cell types; and

According to the C activation vectors output by the second capsule layer, the probability that the cell image in the cell picture corresponds to each cell category is calculated.

In order to achieve the above objective, the embodiment of the present application also provides a cell picture recognition system based on CapsNet capsule network, including:

Input module, used to input cell pictures into CapsNet capsule network;

The first convolution module is configured to perform a convolution operation on the cell image through a convolution layer to obtain A convolution feature maps;

The second convolution module is configured to perform a convolution operation on the A convolution feature maps through the first capsule layer to obtain B multidimensional vectors;

The fully connected module is used to perform a fully connected operation on the B multi-dimensional vectors through the second capsule layer to obtain C activation vectors, where the C activation vectors indicate that there are C cell types; and

The calculation module is used to calculate the probability that the cell image in the cell picture corresponds to each cell category according to the C activation vectors output by the second capsule layer.

In order to achieve the foregoing objective, an embodiment of the present application further provides a computer device, the computer device including a memory, a processor, and computer-readable instructions stored on the memory and running on the processor, the When the computer-readable instructions are executed by the processor, the following steps are implemented:

Input the cell picture into the CapsNet capsule network;

In order to achieve the foregoing objective, an embodiment of the present application also provides a computer-readable storage medium having computer-readable instructions stored in the computer-readable storage medium, and the computer-readable instructions may be executed by at least one processor, So that the at least one processor executes the following steps:

Input the cell picture into the CapsNet capsule network;

The method, system, computer equipment, and computer-readable storage medium for cell image recognition based on the CapsNet capsule network provided by the embodiments of this application can obtain the characteristic parameters of different cells (for example, position, size, etc.) through the first capsule layer and the second capsule layer. Direction, texture, color, etc.), as well as the relationship between the cells, so that the cells squeeze and block each other, and still be able to identify the cell image in the cell picture according to the above characteristic parameters, and output multiple cells on the cell image Label, output the cell type of the cell image in the cell picture. It can be seen that the CapsNet capsule network structure of the embodiment of this application can identify cell types, and can perform identification operations on cells when the biomedical image data set is small, the cells are close to each other, and there are multiple cells in the same image. Maintain a high recognition accuracy rate.

Description of the drawings

FIG. 1 is a schematic flowchart of Embodiment 1 of a method for recognizing a cell image based on a CapsNet capsule network according to an embodiment of this application.

FIG. 2 is a schematic diagram of program modules of Embodiment 2 of a cell image recognition system based on CapsNet capsule network according to an embodiment of the application.

FIG. 3 is a schematic diagram of the hardware structure of the third embodiment of the computer equipment of this application.

Detailed ways

In order to make the purpose, technical solutions, and advantages of this application clearer, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the application, and not used to limit the application. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative work shall fall within the protection scope of this application.

It should be noted that the descriptions related to "first", "second", etc. in this application are only for descriptive purposes, and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of technical features indicated. . Therefore, the features defined with "first" and "second" may explicitly or implicitly include at least one of the features. In addition, the technical solutions between the various embodiments can be combined with each other, but it must be based on what can be achieved by a person of ordinary skill in the art. When the combination of technical solutions is contradictory or cannot be achieved, it should be considered that such a combination of technical solutions does not exist. , Not within the scope of protection required by this application.

The following embodiments will exemplarily describe the computer device 2 as the execution subject.

Example one

Referring to FIG. 1, it shows a flowchart of the steps of a method for recognizing a cell image based on a CapsNet capsule network in the first embodiment of the present application. It can be understood that the flowchart in this method embodiment is not used to limit the order of execution of the steps. details as follows.

Step S100, input the cell picture into the CapsNet capsule network.

The computer device 2 can directly or indirectly control the electron microscope to collect cell pictures of the sample cells, and input the cell pictures collected by the electron microscope into the CapsNet capsule network.

For example, the length and width of the cell image are both 28 pixels. In this cell picture, there may be differences in the shape of each cell image, and multiple cell images may overlap. It should be noted that the cell image refers to an image area representing the shape of the cell in the cell image.

Step S102, performing a convolution operation on the cell image through the convolution layer to obtain A convolution feature maps.

For example, the convolutional layer includes 256 9*9 convolution kernels with a step length of 1, and a convolution operation is performed on the cell image through the convolutional layer to obtain 256 20*20 convolutions Feature map.

Step S104, performing a convolution operation on the A convolution feature maps through the first capsule layer to obtain B multidimensional vectors.

Exemplarily, the step S104 further includes: dividing the A convolution feature maps into B convolution feature map combinations; performing a convolution operation on the B convolution feature map combinations to obtain B X* Y convolution feature map, each X*Y convolution feature map includes X*Y grids corresponding to each grid, and each grid corresponds to a multi-dimensional vector; wherein, the first capsule layer does not include an activation function.

It should be noted that the A convolution feature map can be divided equally, such as segmenting into B convolution feature map combinations, and each convolution feature map combination includes

Convolution feature maps, where,

Is a positive integer greater than or equal to 2.

It can be preferably 8, which means that 8 different forms (position, movement, rotation, size change, etc.) of the cell image are packaged together. of course,

You can also choose other numbers.

For example, the first capsule layer divides 256 convolutional feature maps into 32 parts, and performs a convolution operation on these 256 convolutional feature maps through a 9*9 convolution kernel and a step size of 1, to obtain 32 A 6*6 convolution feature map. The grid of the 6*6 convolution feature map includes 6*6 grids, and each grid is an 8-dimensional vector. Therefore, the first capsule layer outputs 32*6*6 8-dimensional vectors (also called first-level capsules).

The length of each vector represents the estimated probability of whether the object exists, and its direction (for example, in an 8-dimensional space) records the posture parameters of the object (for example, precise position, rotation, etc.). If the object changes slightly (for example, movement, rotation, size change, etc.), the first capsule layer outputs a vector with the same length but slightly changed direction.

Based on the characteristics of small morphological differences between various types of cells, the first capsule layer does not use the activation function ReLU.

Step S106: Perform a full connection operation on the B multi-dimensional vectors through the second capsule layer to obtain C activation vectors, where the C activation vectors indicate that there are C cell types.

It can be understood that in the CapsNet capsule network, the convolutional layer is to extract the picture features in the cell picture. The first capsule layer contains B first-level capsules, and is to receive the image features extracted by the convolutional layer to form multiple feature combinations (ie, multi-dimensional vectors). A one-dimensional capsule corresponds to a multi-dimensional vector, which is used for centralized representation Multiple graphical attributes of a cell, and its length characterizes the probability of cell existence. The second capsule layer, including C second-level capsules, each second-level capsule is based on multiple feature combinations and the fully connected layer in the second capsule layer to obtain the corresponding activation vector, the length of the activation vector represents the probability of the corresponding cell category V _j .

Exemplarily, C activation vectors can be obtained by the following formula:

Among them, u _i represents the i-th multidimensional vector in the B X*Y convolution feature maps output by the first capsule layer, 1≤i≤B; W _ij is the transformation matrix; c _ij is the coupling coefficient, which is used to measure a capsule activation stage i j is the probability of two capsules, 1≤j; V _j is the activation of cell class j is a vector, the vector V _j for each active cell as two capsules of class j.

Exemplarily, it further includes a training step of obtaining the coupling coefficient c _ij :

Step (1), initialize the temporary variable b _ij , and the initial value of b _ij is 0;

Step (2), through the formula

Calculate c _ij , where the initial value of i is 1, and the initial value of j is 1;

Step (3), through the formula

with

Calculate s _j ;

Step (4), input s _j into the formula

Calculate V _j ;

Step (5), through the formula

Update b _ij ;

Repeat steps (2) to (5) to get the value corresponding to c _ij .

For example, the layer may include a second capsule 10 two capsules, two capsules each cell corresponds to a category, which the transformation matrix W _ij 16x8 dimension converting 8 capsules a capsule 16 through a two-dimensional shape , Each second-level capsule corresponds to a cell category j.

The prediction vector is:

The activation vector V _j of cell category _j is:

Each activation vector V _j acts as a second-level capsule of cell category j. The probability that the cell image in the cell picture is classified as j can be obtained by calculating ||V _j ||.

Where u _i represents the i-th 8-dimensional vector in the 32 8-dimensional 6*6 convolutional feature maps output by the first capsule layer, and _Wij is the transformation matrix (16*8): the 8-dimensional first-level capsule is converted to 16 Dimensional second-level capsule; c _ij is the coupling coefficient, which is used to measure how likely the first-level capsule i is to activate the second-level capsule j, and the sum is 1 (ie, the coupling coefficient of the second-level capsule j and all the first-level capsules in the upper layer The sum is 1, and is determined by "routing softmax").

Step S108: Calculate the probability that the cell image in the cell picture corresponds to each cell category according to the C activation vectors output by the second capsule layer.

For example, for each activation vector mod ||V _j ||, the larger the value of ||V _j ||, the higher the probability of the corresponding cell category j, therefore, the maximum ||V _j || The cell type corresponding to the value is determined as the cell type of the cell image in the cell image.

Example two

Please continue to refer to FIG. 2, which shows a schematic diagram of the program modules of the second embodiment of the cell image recognition system based on the CapsNet capsule network in the embodiment of the present application. In this embodiment, the cell picture recognition system 20 may include or be divided into one or more program modules, and the one or more program modules are stored in a storage medium and executed by one or more processors to complete This application can also implement the above-mentioned method for recognizing cell pictures based on the CapsNet capsule network. The program module referred to in the embodiments of the present application refers to a series of computer-readable instruction segments capable of completing specific functions, and is more suitable for describing the execution process of the cell picture recognition system 20 in the storage medium than the program itself. The following description will specifically introduce the functions of each program module in this embodiment:

The input module 200 is used to input cell pictures into the CapsNet capsule network.

For example, the length and width of the cell image are both 28 pixels. In this cell picture, there may be differences in the shape of each cell image, and multiple cell images may overlap. These cell images may correspond to different types of cells.

The first convolution module 202 is configured to perform a convolution operation on the cell image through a convolution layer to obtain A convolution feature maps.

For example, the convolutional layer includes 256 9*9 convolution kernels with a step length of 1, and a convolution operation is performed on the cell through the convolutional layer to obtain 256 20*20 convolution features Figure.

The second convolution module 204 is configured to perform a convolution operation on the A convolution feature maps through the first capsule layer to obtain B multidimensional vectors. In an exemplary embodiment, the second convolution module 204 is configured to: divide the A convolution feature maps into B convolution feature map combinations; perform a convolution operation on the B convolution feature map combinations , Obtain B X*Y convolution feature maps, each X*Y convolution feature map includes X*Y grids corresponding to each grid, and each grid corresponds to a multi-dimensional vector; wherein, the first capsule layer does not include activation function.

For example, the first capsule layer divides 256 convolution feature maps into 32 parts, and performs convolution operation on these 256 convolution feature maps through a 9*9 convolution kernel to obtain 32 6*6 convolutions Product feature map. The grid of the 6*6 convolution feature map includes 6*6 grids, and each grid is an 8-dimensional vector. Therefore, the first capsule layer outputs 32*6*6 8-dimensional vectors (also called first-level capsules).

The fully connected module 206 is configured to perform a fully connected operation on the B multi-dimensional vectors through the second capsule layer to obtain C activation vectors. The C activation vectors indicate that there are C cell types.

Exemplarily, C activation vectors can be obtained by the following formula:

In an exemplary embodiment, a coupling coefficient acquisition module is further included, configured to:

Step (1), initialize the temporary variable b _ij , the initial value of b _ij is 0; step (2), through the formula

Calculate c _ij , where the initial value of i is 1, and the initial value of j is 1; step (3), through the formula

with

Calculate s _j ; step (4), enter s _j into the formula

Calculate V _j ; step (5), through the formula

Update b _ij ; repeat steps (2) to (5) to obtain the value corresponding to c _ij .

The calculation module 208 is configured to calculate the probability that the cell image in the cell picture corresponds to each cell category according to the C activation vectors output by the second capsule layer.

This application obtains the characteristic parameters of different cells (for example, position, size, direction, texture, color, etc.) through the first capsule layer and the second capsule layer, as well as the relationship between the cells, so that the cells squeeze each other to block , It can still identify each cell according to the above-mentioned characteristic parameters, and output multiple tags on the cell image, and output the cell type of each cell on the cell image. It can be seen that the CapsNet capsule network structure of the embodiment of this application can identify cell types, and can perform identification operations on cells when the biomedical image data set is small, the cells are close to each other, and there are multiple cells in the same image. Maintain a high recognition accuracy rate.

Example three

Refer to FIG. 3, which is a schematic diagram of the hardware architecture of the computer device in the third embodiment of the present application. In this embodiment, the computer device 2 is a device that can automatically perform numerical calculation and/or information processing according to pre-set or stored instructions. The computer device 2 may be a PC, a rack server, a blade server, a tower server, or a cabinet server (including an independent server, or a server cluster composed of multiple servers). As shown in the figure, the computer device 2 at least includes, but is not limited to, a memory 21, a processor 22, a network interface 23, and a cell image recognition system 20 that can be communicated with each other through a system bus. among them:

In this embodiment, the memory 21 includes at least one type of computer-readable storage medium. The readable storage medium includes flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), random access memory ( RAM), static random access memory (SRAM), read only memory (ROM), electrically erasable programmable read only memory (EEPROM), programmable read only memory (PROM), magnetic memory, magnetic disks, optical disks, etc. In some embodiments, the memory 21 may be an internal storage unit of the computer device 2, such as a hard disk or memory of the computer device 2. In other embodiments, the memory 21 may also be an external storage device of the computer device 2, for example, a plug-in hard disk, a smart media card (SMC), and a secure digital (Secure Digital, SD card, Flash Card, etc. Of course, the memory 21 may also include both the internal storage unit of the computer device 2 and its external storage device. In this embodiment, the memory 21 is generally used to store the operating system and various application software installed in the computer device 2, such as the program code of the cell image recognition system 20 in the second embodiment. In addition, the memory 21 can also be used to temporarily store various types of data that have been output or will be output.

The processor 22 may be a central processing unit (Central Processing Unit, CPU), a controller, a microcontroller, a microprocessor, or other data processing chips in some embodiments. The processor 22 is generally used to control the overall operation of the computer device 2. In this embodiment, the processor 22 is used to run the program code or process data stored in the memory 21, for example, to run the cell image recognition system 20, so as to implement the CapsNet capsule network-based cell image recognition method of the first embodiment.

The network interface 23 may include a wireless network interface or a wired network interface, and the network interface 23 is generally used to establish a communication connection between the computer device 2 and other electronic devices. For example, the network interface 23 is used to connect the computer device 2 with an external terminal through a network, and establish a data transmission channel and a communication connection between the computer device 2 and the external terminal. The network may be Intranet, Internet, Global System of Mobile Communication (GSM), Wideband Code Division Multiple Access (WCDMA), 4G network, 5G Network, Bluetooth (Bluetooth), Wi-Fi and other wireless or wired networks.

It should be pointed out that FIG. 3 only shows the computer device 2 with components 20-23, but it should be understood that it is not required to implement all the components shown, and more or fewer components may be implemented instead.

In this embodiment, the cell picture recognition system 20 stored in the memory 21 can also be divided into one or more program modules, and the one or more program modules are stored in the memory 21 and are composed of one or more program modules. Multiple processors (in this embodiment, the processor 22) are executed to complete the application.

For example, FIG. 2 shows a schematic diagram of program modules for implementing the second embodiment of the cell picture recognition system 20. In this embodiment, the cell picture recognition system 20 can be divided into an input module 200 and a first convolution module 202. , The second convolution module 204, the fully connected module 206, and the calculation module 208. Among them, the program module referred to in the present application refers to a series of computer-readable instruction segments that can complete specific functions, and is more suitable than a program to describe the execution process of the cell image recognition system 20 in the computer device 2. The specific functions of the program modules 200-208 have been described in detail in the second embodiment, and will not be repeated here.

Example four

This embodiment also provides a computer-readable storage medium. The computer-readable storage medium may be non-volatile or volatile, such as flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX). Memory, etc.), random access memory (RAM), static random access memory (SRAM), read only memory (ROM), electrically erasable programmable read only memory (EEPROM), programmable read only memory (PROM), magnetic memory , Magnetic disks, optical disks, servers, App application malls, etc., on which computer-readable instructions are stored, and corresponding functions are realized when the programs are executed by the processor. The computer-readable storage medium of this embodiment stores computer-readable instructions in the computer-readable storage medium, and the computer-readable instructions can be executed by at least one processor, so that the at least one processor performs the following steps :

Input the cell picture into the CapsNet capsule network;

The serial numbers of the foregoing embodiments of the present application are only for description, and do not represent the advantages and disadvantages of the embodiments.

Through the description of the above embodiments, those skilled in the art can clearly understand that the method of the above embodiments can be implemented by means of software plus the necessary general hardware platform. Of course, it can also be implemented by hardware, but in many cases the former is better.的实施方式。

The above are only preferred embodiments of this application, and do not limit the scope of this application. Any equivalent structure or equivalent process transformation made using the content of the description and drawings of this application, or directly or indirectly used in other related technical fields , The same reason is included in the scope of patent protection of this application.

Claims

A cell picture recognition method based on CapsNet capsule network, wherein the method includes:

Input the cell picture into the CapsNet capsule network;

Performing a convolution operation on the cell image through a convolution layer to obtain A convolution feature maps;

Performing a convolution operation on the A convolution feature maps through the first capsule layer to obtain B multidimensional vectors;

Perform a full connection operation on the B multi-dimensional vectors through the second capsule layer to obtain C activation vectors, where the C activation vectors indicate that there are C cell types; and

According to the C activation vectors output by the second capsule layer, the probability that the cell image in the cell picture corresponds to each cell category is calculated.
The method for recognizing cell pictures based on the CapsNet capsule network according to claim 1, wherein the step of performing a convolution operation on the A convolution feature maps through the first capsule layer to obtain B multidimensional vectors comprises:

Dividing the A convolution feature maps into B convolution feature map combinations;

Perform a convolution operation on the combination of the B convolution feature maps to obtain B X*Y convolution feature maps, each X*Y convolution feature map includes X*Y grids, and each grid corresponds to one Multidimensional vector

Wherein, the first capsule layer does not include an activation function.
The method for cell image recognition based on CapsNet capsule network according to claim 2, wherein C activation vectors are obtained by the following formula:

Among them, u i represents the i-th multidimensional vector in the B X*Y convolution feature maps output by the first capsule layer, 1≤i≤B; W ij is the transformation matrix; c ij is the coupling coefficient, which is used to measure a capsule activation stage i j is the probability of two capsules, 1≤j; V j is the activation of cell class j is a vector, the vector V j for each active cell as two capsules of class j.
The method for recognizing cell pictures based on the CapsNet capsule network according to claim 3, further comprising a training step of obtaining the coupling coefficient c ij :

Step (1), initialize the temporary variable b ij , and the initial value of b ij is 0;

Step (2), through the formula
Calculate c ij , where the initial value of i is 1, and the initial value of j is 1;

Step (3), through the formula
with
Calculate s j ;

Step (4), input s j into the formula
Calculate V j ;

Step (5), through the formula
Update b ij ;

Repeat steps (2) to (5) to get the value corresponding to c ij .
The method for recognizing cell pictures based on CapsNet capsule network according to claim 3, wherein the second capsule layer includes 10 second-level capsules.
A cell picture recognition system based on CapsNet capsule network, wherein the system includes:

Input module, used to input cell pictures into CapsNet capsule network;

The first convolution module is configured to perform a convolution operation on the cell image through a convolution layer to obtain A convolution feature maps;

The second convolution module is configured to perform a convolution operation on the A convolution feature maps through the first capsule layer to obtain B multidimensional vectors;

The fully connected module is used to perform a fully connected operation on the B multi-dimensional vectors through the second capsule layer to obtain C activation vectors, where the C activation vectors indicate that there are C cell types; and

The calculation module is used to calculate the probability that the cell image in the cell picture corresponds to each cell category according to the C activation vectors output by the second capsule layer.
The cell picture recognition system based on CapsNet capsule network according to claim 6, wherein the second convolution module is used for:

Dividing the A convolution feature maps into B convolution feature map combinations;

Perform a convolution operation on the combination of the B convolution feature maps to obtain B X*Y convolution feature maps, each X*Y convolution feature map includes X*Y grids, and each grid corresponds to one Multidimensional vector

Wherein, the first capsule layer does not include an activation function.
The cell picture recognition system based on the CapsNet capsule network of claim 7, wherein C activation vectors are obtained by the following formula:

Among them, u i represents the i-th multidimensional vector in the B X*Y convolutional feature maps output by the first capsule layer; Wij is the transformation matrix; c ij is the coupling coefficient, which is used to measure the activation of the first-level capsule i The probability of capsule j, the total is 1; V j is the activation vector of cell type j, and each activation vector V j is used as the second-level capsule of cell type j.
The cell picture recognition system based on CapsNet capsule network according to claim 8, further comprising a coupling coefficient acquisition module for:

Step (1), initialize the temporary variable b ij , and the initial value of b ij is 0;

Step (2), through the formula
Calculate c ij , where the initial value of i is 1, and the initial value of j is 1;

Step (3), through the formula
with
Calculate s j ;

Step (4), input s j into the formula
Calculate V j ;

Step (5), through the formula
Update b ij ;

Repeat steps (2) to (5) to get the value corresponding to c ij .
8. The cell picture recognition system based on CapsNet capsule network according to claim 8, wherein the second capsule layer includes 10 secondary capsules.
A computer device, wherein the computer device includes a memory, a processor, and computer-readable instructions stored in the memory and capable of running on the processor, and the computer-readable instructions are implemented when the processor is executed The following steps:

Input the cell picture into the CapsNet capsule network;

Performing a convolution operation on the cell image through a convolution layer to obtain A convolution feature maps;

Performing a convolution operation on the A convolution feature maps through the first capsule layer to obtain B multidimensional vectors;

Perform a full connection operation on the B multi-dimensional vectors through the second capsule layer to obtain C activation vectors, where the C activation vectors indicate that there are C cell types; and

According to the C activation vectors output by the second capsule layer, the probability that the cell image in the cell picture corresponds to each cell category is calculated.
The computer device according to claim 11, wherein the computer-readable instructions are also executable by at least one processor, so that the at least one processor executes the following steps:

Dividing the A convolution feature maps into B convolution feature map combinations;

Perform a convolution operation on the combination of the B convolution feature maps to obtain B X*Y convolution feature maps, each X*Y convolution feature map includes X*Y grids, and each grid corresponds to one Multidimensional vector

Wherein, the first capsule layer does not include an activation function.
The computer device according to claim 12, wherein the C activation vectors are obtained by the following formula:

Among them, u i represents the i-th multidimensional vector in the B X*Y convolution feature maps output by the first capsule layer, 1≤i≤B; W ij is the transformation matrix; c ij is the coupling coefficient, which is used to measure a capsule activation stage i j is the probability of two capsules, 1≤j; V j is the activation of cell class j is a vector, the vector V j for each active cell as two capsules of class j.
The computer device according to claim 13, wherein the computer-readable instructions are also executable by at least one processor, so that the at least one processor executes the following steps:

Training steps to obtain the coupling coefficient c ij :

Step (1), initialize the temporary variable b ij , and the initial value of b ij is 0;

Step (2), through the formula
Calculate c ij , where the initial value of i is 1, and the initial value of j is 1;

Step (3), through the formula
with
Calculate s j ;

Step (4), input s j into the formula
Calculate V j ;

Step (5), through the formula
Update b ij ;

Repeat steps (2) to (5) to get the value corresponding to c ij .
The computer device according to claim 13, wherein the second capsule layer includes 10 secondary capsules.
A computer-readable storage medium, wherein computer-readable instructions are stored in the computer-readable storage medium, and the computer-readable instructions can be executed by at least one processor, so that the at least one processor executes the following step:

Input the cell picture into the CapsNet capsule network;

Performing a convolution operation on the cell image through a convolution layer to obtain A convolution feature maps;

Performing a convolution operation on the A convolution feature maps through the first capsule layer to obtain B multidimensional vectors;

Perform a full connection operation on the B multi-dimensional vectors through the second capsule layer to obtain C activation vectors, where the C activation vectors indicate that there are C cell types; and

According to the C activation vectors output by the second capsule layer, the probability that the cell image in the cell picture corresponds to each cell category is calculated.
The computer-readable storage medium according to claim 16, wherein the computer-readable instructions are also executable by at least one processor, so that the at least one processor executes the following steps:

Dividing the A convolution feature maps into B convolution feature map combinations;

Perform a convolution operation on the combination of the B convolution feature maps to obtain B X*Y convolution feature maps, each X*Y convolution feature map includes X*Y grids, and each grid corresponds to one Multidimensional vector

Wherein, the first capsule layer does not include an activation function.
The computer-readable storage medium according to claim 17, wherein the C activation vectors are obtained by the following formula:

Among them, u i represents the i-th multidimensional vector in the B X*Y convolution feature maps output by the first capsule layer, 1≤i≤B; W ij is the transformation matrix; c ij is the coupling coefficient, which is used to measure a capsule activation stage i j is the probability of two capsules, 1≤j; V j is the activation of cell class j is a vector, the vector V j for each active cell as two capsules of class j.
The computer-readable storage medium according to claim 18, wherein the computer-readable instructions are also executable by at least one processor, so that the at least one processor executes the following steps:

Training steps to obtain the coupling coefficient c ij :

Step (1), initialize the temporary variable b ij , and the initial value of b ij is 0;

Step (2), through the formula
Calculate c ij , where the initial value of i is 1, and the initial value of j is 1;

Step (3), through the formula
with
Calculate s j ;

Step (4), input s j into the formula
Calculate V j ;

Step (5), through the formula
Update b ij ;

Repeat steps (2) to (5) to get the value corresponding to c ij .
The computer readable storage medium of claim 18, wherein the second capsule layer includes 10 secondary capsules.