Medical image classification method, device, terminal equipment and storage medium

Abstract

The present application is applicable to the field of image processing technology, and provides a medical image classification method, apparatus, terminal device and storage medium, including: obtaining a medical image to be classified; obtaining a tensor of the medical image to be classified; performing a tensor rank-one decomposition on the tensor of the medical image to be classified to obtain the N-order component of the tensor of the medical image to be classified, wherein N refers to the order of the tensor of the medical image to be classified; inputting the N-order component of the tensor of the medical image to be classified into a pre-constructed decision function, classifying the medical image to be classified through the decision function, and outputting the classification result of the medical image to be classified. The present application can improve the classification accuracy of medical images.

Description

Technical Field

[0001] The present application belongs to the field of image processing technology, and in particular, relates to a medical image classification method, apparatus, terminal device and storage medium. Background Art

[0002] Image classification has become an important research topic in the field of image processing, such as the classification of medical images. Existing medical image classification algorithms need to vectorize medical images, and the vectorization process easily leads to the loss of spatial information of medical images, which reduces the classification accuracy. Summary of the invention

[0003] The present application provides a medical image classification method, apparatus, terminal device and storage medium to improve the classification accuracy of medical images.

[0004] In a first aspect, an embodiment of the present application provides a medical image classification method, the medical image classification method comprising:

[0005] Acquire a medical image to be classified;

[0006] Acquire the tensor of the medical image to be classified;

[0007] Performing a tensor rank-one decomposition on the tensor of the medical image to be classified to obtain an N-order component of the tensor of the medical image to be classified, wherein N refers to the order of the tensor of the medical image to be classified;

[0008] The N-order component of the tensor of the medical image to be classified is input into a pre-constructed decision function, the medical image to be classified is classified by the decision function, and the classification result of the medical image to be classified is output.

[0009] In a second aspect, an embodiment of the present application provides a medical image classification device, the medical image classification device comprising:

[0010] An image acquisition module, used for acquiring medical images to be classified;

[0011] A tensor acquisition module, used to acquire the tensor of the medical image to be classified;

[0012] A tensor decomposition module, used for performing a tensor rank-one decomposition on the tensor of the medical image to be classified to obtain an N-order component of the tensor of the medical image to be classified, wherein N is the order of the tensor of the medical image to be classified;

[0013] The component processing module is used to input the N-order components of the tensor of the medical image to be classified into a pre-constructed decision function, classify the medical image to be classified through the decision function, and output the classification result of the medical image to be classified.

[0014] In a third aspect, an embodiment of the present application provides a terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the steps of the medical image classification method described in the first aspect above are implemented.

[0015] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the medical image classification method described in the first aspect above are implemented.

[0016] In a fifth aspect, an embodiment of the present application provides a computer program product, which, when executed on a terminal device, enables the terminal device to execute the steps of the medical image classification method described in the first aspect above.

[0017] As can be seen from the above, the present application can effectively retain the spatial information of the medical image to be classified by using tensors to represent the medical image to be classified, and after performing a tensor rank-one decomposition on the tensor of the medical image to be classified and inputting the decimal point into the decision function for classification, it can make full use of the structural information and correlation of the tensor of the medical image to be classified, and extract the discriminant information hidden in the tensor of the medical image to be classified, thereby improving the classification accuracy of the medical image to be classified. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0019] Figure 1 This is a schematic diagram of the implementation process of the medical image classification method provided in Example 1 of the present application;

[0020] Figure 2 This is a schematic diagram of the implementation process of the medical image classification method provided in Example 2 of the present application;

[0021] Figure 3 is a schematic diagram of the structure of a medical image classification device provided in Embodiment 3 of the present application;

[0022] Figure 4 It is a structural diagram of the terminal device provided in Example 4 of the present application. DETAILED DESCRIPTION

[0023] In the following description, specific details such as specific system structures, technologies, etc. are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present application. However, it should be clear to those skilled in the art that the present application may also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present application.

[0024] It should be understood that when used in this specification and the appended claims, the term "comprising" indicates the presence of described features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof.

[0025] In a specific implementation, the terminal device described in the embodiments of the present application includes, but is not limited to, other portable devices such as mobile phones, laptop computers, or tablet computers with touch-sensitive surfaces (e.g., touch screen displays and / or touch pads). It should also be understood that in some embodiments, the device is not a portable communication device, but a desktop computer with a touch-sensitive surface (e.g., touch screen displays and / or touch pads).

[0026] In the following discussion, a terminal device including a display and a touch-sensitive surface is described. However, it should be understood that the terminal device may include one or more other physical user interface devices such as a physical keyboard, a mouse and / or a joystick.

[0027] The terminal device supports various applications, such as one or more of the following: a drawing application, a presentation application, a word processing application, a website creation application, a disk burning application, a spreadsheet application, a game application, a telephone application, a video conferencing application, an email application, an instant messaging application, a workout support application, a photo management application, a digital camera application, a digital camcorder application, a web browsing application, a digital music player application, and / or a digital video player application.

[0028] Various applications that can be executed on the terminal device can use at least one common physical user interface device such as a touch-sensitive surface. One or more functions of the touch-sensitive surface and corresponding information displayed on the terminal can be adjusted and / or changed between applications and / or within corresponding applications. In this way, the common physical architecture of the terminal (e.g., the touch-sensitive surface) can support various applications with user interfaces that are intuitive and transparent to the user.

[0029] It should be understood that the size of the serial numbers of the steps in this embodiment does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present application.

[0030] In order to illustrate the technical solution described in this application, a specific embodiment is provided below for illustration.

[0031] See also Figure 1 , is a schematic diagram of the implementation flow of the medical image classification method provided in the first embodiment of the present application. The medical image classification method is applied to a terminal device. As shown in the figure, the medical image classification method may include the following steps:

[0032] Step 101: Obtain a medical image to be classified.

[0033] The medical image to be classified may refer to a medical image to be classified, and the dimension of the medical image to be classified is at least two-dimensional. For example, if the medical image to be classified is a grayscale medical image, then the dimension of the medical image to be classified is two-dimensional; if the medical image to be classified is a color medical image, then the dimension of the medical image to be classified is three-dimensional.

[0034] It should be noted that the medical images to be classified in this embodiment can be acquired through the image acquisition device in the terminal device, or can be acquired by a server or other device and then sent to the terminal device. The method of acquiring the medical images to be classified is not limited in this embodiment.

[0035] Step 102: Obtain a tensor of a medical image to be classified.

[0036] Tensors are high-order extensions of vectors and can be considered as natural expressions of images. Images of different dimensions can be represented by different tensors. The order of the image tensor is related to the image dimension. If the image dimension is N, then the order of the image tensor is N. For example, a two-dimensional medical image can be represented by a second-order tensor, a three-dimensional medical image can be represented by a third-order tensor, and a four-dimensional medical image can be represented by a fourth-order tensor. Taking a third-order tensor as an example, the first order represents the horizontal coordinate of the medical image (i.e., the width of the image), the second order represents the vertical coordinate of the medical image (i.e., the length of the image), and the third order represents the vector dimension of the medical image. The first and second orders jointly describe the space of the medical image and contain the spatial information of the medical image. The third order describes the vector dimension of the medical image and contains the sagittal plane spatial structure. Therefore, by using tensors to represent the medical image to be classified, the spatial information of the medical image to be classified can be effectively retained and the loss of spatial information of the medical image to be classified can be reduced.

[0037] Step 103 , performing a tensor rank-one decomposition on the tensor of the medical image to be classified to obtain the N-th order component of the tensor of the medical image to be classified.

[0038] Where N refers to the order of the tensor of the medical image to be classified.

[0039] Performing a tensor rank-one decomposition on the tensor of the medical image to be classified may refer to performing a tensor rank-one decomposition on the tensor of the medical image to be classified with a length of 1. Performing a tensor rank-one decomposition on the tensor of the medical image to be classified with a length of 1 may refer to decomposing the tensor of the medical image to be classified into the outer product of components of order N. For example, the tensor of the medical image to be classified is X. After performing a tensor rank-one decomposition on X with a length of 1, X can be expressed as x (1) is the first-order component of X, x (2) is the second-order component of X, x (N) is the Nth-order component of X, (x (1) ,x (2) ,…,x (N) ) is the component of X at order N, Represents the outer product operator.

[0040] Step 104 , inputting the N-order component of the tensor of the medical image to be classified into a pre-constructed decision function, classifying the medical image to be classified through the decision function, and outputting the classification result of the medical image to be classified.

[0041] In this embodiment, before classifying the medical image to be classified, a decision function can be constructed based on a large number of image training sets, so that the constructed decision function can classify the medical image to be classified more accurately, predict the category of the medical image to be classified, and output the category of the medical image to be classified predicted by the decision function to prompt the user of the category of the medical image to be classified.

[0042] Among them, the number of categories that the decision function can predict is at least two. For example, the number of categories that the decision function can predict is two, and the two categories are target categories and non-target categories. Then, when the medical image to be classified is classified by the decision function, the classification result of the medical image to be classified is that the category of the medical image to be classified is the target category or the category of the medical image to be classified is not the target category.

[0043] In one application scenario, a user's brain magnetic resonance imaging (MRI) image is used as a medical image to be classified. The dimension of the brain MRI image is three-dimensional. The tensor of the brain MRI image can be firstly decomposed into a tensor rank one to obtain the third-order components of the tensor of the brain MRI image. The third-order components of the tensor of the brain MRI image are input into a pre-constructed decision function. The classification result of the above-mentioned brain MRI image can be determined by the decision function. The classification result includes but is not limited to mild cognitive impairment, Alzheimer's disease, normal elderly people, etc. If the classification result is mild cognitive impairment, it is determined that the user is in the mild cognitive impairment stage. If the classification result is Alzheimer's disease, it is determined that the user suffers from Alzheimer's disease. If the classification result is normal elderly people, it is determined that the user is normal, does not suffer from Alzheimer's disease, and is not in the mild cognitive impairment stage.

[0044] Optionally, outputting the classification result of the medical image to be classified includes:

[0045] Output the classification results of the medical images to be classified in a preset manner.

[0046] The preset method may refer to a preset output method for the classification result, and the preset method may be at least one of a plurality of output methods such as screen display, voice broadcast, and text message of a terminal device.

[0047] The embodiment of the present application uses tensors to represent the medical images to be classified, which can effectively retain the spatial information of the medical images to be classified, and after performing a tensor rank-one decomposition on the tensor of the medical images to be classified and inputting the decimal point into a decision function for classification, it can make full use of the structural information and correlation of the tensor of the medical images to be classified, and extract the discriminant information hidden in the tensor of the medical images to be classified, thereby improving the classification accuracy of the medical images to be classified, for example, improving the recognition accuracy of Alzheimer's disease.

[0048] See also Figure 2 , is a schematic diagram of the implementation flow of the medical image classification method provided in the second embodiment of the present application. The medical image classification method is applied to a terminal device. As shown in the figure, the medical image classification method may include the following steps:

[0049] Step 201, obtaining a sample training set.

[0050] In this embodiment, the sample training set is used to construct a decision function, so that the constructed decision function can more accurately classify the medical images to be classified (i.e., perform category prediction on the medical images to be classified). The sample training set includes a tensor of a sample image set and a category label of the sample image set, the sample image set includes M sample images, the tensor of the sample image set includes the tensors of the M sample images, the category labels of the sample image set include the category labels of the M sample images, the order of the tensor of the sample image set is N, M is an integer greater than 1, and the M sample images belong to medical images. Among them, the category label of a sample image is the true category of the sample image, and the order of the tensor of the sample image set is N, which means that the order of the tensors of the M sample images is N.

[0051] In one application scenario, MRI images in the Alzheimer's disease neuroimaging database can be used as a sample image set. The categories of the sample image set include mild cognitive impairment, Alzheimer's disease, and normal elderly people. The category of each sample image in the sample image set (that is, the category label of each sample image) can be pre-labeled, and a decision function can be constructed based on the tensor of each sample image and its category label, so that the constructed decision function can identify the three categories of mild cognitive impairment, Alzheimer's disease, and normal elderly people.

[0052] In order to reduce the loss of spatial information of the medical image to be classified, this embodiment can store data of M sample images in the form of tensors. For example, a three-dimensional brain MRI image is represented in the form of a tensor, which is conducive to retaining the spatial information of the brain MRI image. The tensors of the M sample images and the category labels of the M sample images are all in the form of column vectors to form a sample training set, and the tensor of each sample image in the sample training set corresponds to its own category label one by one. For example, the image training set is {X m ,y m |m=1,2,…,M}, where X m Represents the tensor of the mth sample image, y m Represents the category label of the mth sample image, and the value of m ranges from 1 to M.

[0053] In this embodiment, when constructing a decision function, it is necessary to first obtain the parameters in the decision function (i.e., the target offset scalar and the target projection vector of the tensor of the sample image set at the Nth order, the target projection vector of the tensor of the sample image set at the Nth order can also be called the Nth order weight vector, which is used to construct the decision function). When obtaining the parameters in the decision function, the tensor of the sample image set is used as input, which enables the terminal device to process data in the tensor field. Processing data in the tensor field does not require dimensionality reduction or vectorization processing, thereby reducing the computational complexity and storage cost during data processing.

[0054] Step 202, obtaining the intra-class scatter matrix of the tensor of the sample image set at each order.

[0055] The intra-class scatter matrix of the tensor of the sample image set at each order may refer to the intra-class scatter matrix estimated after the tensor of the sample image set is expanded along each order.

[0056] Since the tensor of the sample image set includes the tensors of the M sample images, and one sample image corresponds to one tensor, the tensor of the sample image set includes M tensors in total. Through step 202, the intra-class scatter matrix of the M tensors at each order can be obtained, thereby obtaining the intra-class scatter matrix of the M tensors at the Nth order, that is, N intra-class scatter matrices can be obtained. Among them, the intra-class scatter matrix of the M tensors at each order can refer to the intra-class scatter matrix estimated after the M tensors are expanded along each order, for example, the intra-class scatter matrix of the M tensors at the nth order refers to the intra-class scatter matrix estimated after the M tensors are expanded along the nth order.

[0057] The intra-class scatter matrix represents the scatter of the tensor of the sample image set around the mean. Taking the intra-class scatter matrix of the tensor of the sample image set at the nth order as an example, the method of obtaining the intra-class scatter matrix is ​​explained. The intra-class scatter matrix of the tensor of the sample image set at the nth order is S w represents the within-class scatter matrix, Represents the expansion matrix of the tensor of the mth sample image at the nth order, Represents the mean of the expanded matrix of the tensor of the sample image set at the nth order, The expanded matrix of the tensor of the m-th sample image at the n-th order refers to a matrix obtained by rearranging the elements in the tensor of the m-th sample image.

[0058] Step 203, constructing a quadratic programming problem of the tensor of the sample image set at each order according to the intra-class scatter matrix of the tensor of the sample training set and the sample image set at each order.

[0059] Among them, the order of the tensor of the sample image set is N, each order corresponds to a quadratic programming problem, so the N order corresponds to N quadratic programming problems.

[0060] In this embodiment, the first objective function of the tensor of the sample image set in the quadratic programming problem at each order can be constructed according to the intra-class scatter matrix of the tensor of the sample image set at each order; the first constraint condition of the tensor of the sample image set in the quadratic programming problem at each order is constructed according to the sample training set, and the first objective function and the first constraint condition of the tensor of the sample image set in the quadratic programming problem at each order constitute the quadratic programming problem of the tensor of the sample image set at each order. Among them, by introducing the intra-class scatter matrix of the tensor of the sample image set at each order into the first objective function of the tensor of the sample image set in the quadratic programming problem at each order, the first objective function can reflect the Fisher criterion, thereby achieving the maximization of the inter-class scatter while minimizing the intra-class scatter, achieving better classification performance, and thus improving the classification accuracy of the image.

[0061] Among them, for the nth quadratic programming problem, the nth quadratic programming problem is the quadratic programming problem of the tensor of the sample image set at the nth order, the nth order is any order among the N orders, and the first objective function in the nth quadratic programming problem is:

[0062]

[0063] The tensor representing the sample image set is the within-class scatter matrix of order n, The tensor representing the sample image set is the within-class scatter matrix of the i-th order, w (n) The target projection vector of the tensor representing the sample image set at the nth order, w (i) Represents the target projection vector of the tensor of the sample image set at the i-th order, and E represents the target projection vector with w (n) and w (i) Unit vectors of the same order, ε represents the balance factor, used to balance [w (n) ] T w (n) and 0<ε<1, Indicates that (n) The corresponding slack variable is used to measure the degree to which the m-th sample image is misclassified or the m-th sample image relative to w (n) The error, C represents the penalty factor, b (n) A scalar representing the offset of the tensor of the sample image set at the nth order;

[0064] The first constraint in the nth quadratic programming problem is:

[0065]

[0066] X m Represents the tensor of the mth sample image, y m Represents the category label of the mth sample image.

[0067] It should be noted that the first objective function and the first constraint condition in the above quadratic programming problem are both linear optimization problems, so the above quadratic programming problem can also be called a quadratic linear programming problem.

[0068] Step 204 , solving N quadratic programming problems to obtain a target offset scalar and a target projection vector of the tensor of the sample image set at order N.

[0069] The target offset scalar refers to the offset scalar obtained when solving N quadratic programming problems. The target projection vector of the tensor of the sample image set at the Nth order includes w (1) ,…,w (N) , for example, if N is three, then the target projection vector of the tensor of the sample image set at the third order includes the target projection vector of the tensor of the sample image set at the first order (i.e., w (1) ), the target projection vector at the second order (i.e. w (2) ) and the target projection vector at the third order (i.e. w (3) ).

[0070] Optionally, solving N quadratic programming problems to obtain a target offset scalar and a target projection vector of the tensor of the sample image set at order N includes:

[0071] make Convert N quadratic programming problems into a single N-fold quadratic programming problem, which includes the second objective function and the second constraint. Represents the initial projection vector of the tensor of the sample image set at the nth order, × i represents the i-modular product operator, Λ (n) and Λ (i) represents a diagonal matrix, p (n) and p (i) represents an orthogonal matrix, Λ (n) and p (n) satisfy Λ (i) and p (i) satisfy The Nth order corresponds to N initial projection vectors;

[0072] Among them, the second objective function is:

[0073] W * represents the initial projection tensor of the sample image set, which is the outer product of N initial projection vectors, and b represents the offset scalar;

[0074] The second constraint is: m ( <W * ,V m>+b)≥1-ξ m ;

[0075] Solve a single N-fold quadratic programming problem to obtain the target offset scalar and the target projection vector of the tensor of the sample image set at the Nth order.

[0076] In this embodiment, due to is a symmetric matrix, so there exists an orthogonal matrix p (n) ,satisfy W * is the outer product of N initial projection vectors, specifically represents the initial projection vector of the first order of the tensor of the sample image set, represents the initial projection vector of the second-order tensor of the sample image set, Represents the initial projection vector of the tensor of the sample image set at the Nth order. For example, if N is 4 (that is, the order of the tensor of the sample image set is four), then

[0077] It should be noted that the above-mentioned single N-fold quadratic programming problem can also be called a single N-fold quadratic programming problem in tensor mode, and belongs to a convex optimization problem, which can avoid the time-consuming alternating iteration process and speed up the classification of images (for example, speeding up the recognition process of Alzheimer's disease).

[0078] Optionally, a single N-fold quadratic programming problem is solved to obtain a target offset scalar and a target projection vector of the tensor of the sample image set at the Nth order, including:

[0079] According to a single N-fold quadratic programming problem, construct the Lagrangian function:

[0080]

[0081] Among them, α represents the first Lagrange multiplier, α m represents the first Lagrange multiplier corresponding to the mth sample image, β represents the second Lagrange multiplier, β m represents the second Lagrange multiplier corresponding to the mth sample image;

[0082] According to the Lagrange multiplier method, the dual problem of the Lagrangian function is obtained, and the dual problem of the Lagrangian function includes a third objective function and a third constraint condition;

[0083] Among them, the third objective function is:

[0084] α d represents the first Lagrange multiplier corresponding to the d-th sample image, α e represents the first Lagrange multiplier corresponding to the e-th sample image, y dRepresents the category label of the dth sample image, y e represents the category label of the e-th sample image,

[0085] The third constraint is: 0≤α m ≤C;

[0086] Solve the dual problem of the Lagrangian function to obtain the target offset scalar and the target projection vector of the tensor of the sample image set at order N.

[0087] In this embodiment, according to the Lagrange multiplier method, the variable W in the Lagrange function can be * And b find the partial derivative to obtain the dual problem of the Lagrangian function.

[0088] Optionally, solving the dual problem of the Lagrangian function to obtain the target offset scalar and the target projection vector of the tensor of the sample image set at order N includes:

[0089] V d Decompose into the sum of R first-rank tensors, each of which is represented as the outer product of its components at order N;

[0090] V e Decomposed into the sum of R second-rank tensors, each of which is represented as the outer product of its components at order N;

[0091] According to each component of the first rank tensor at the Nth order and each component of the second rank tensor at the Nth order, modify the third objective function, and determine the modified third objective function as the fourth objective function, and the fourth objective function and the third constraint condition constitute a modified dual problem;

[0092] Among them, the fourth objective function is:

[0093]

[0094] represents the component of the kth first-rank tensor at the nth order, represents the component of the qth second-rank tensor at the nth order;

[0095] The modified dual problem is solved by using a sequential minimum optimization algorithm to obtain a target offset scalar and target Lagrangian multipliers corresponding to each of the M sample images. The target Lagrangian multipliers corresponding to each of the M sample images refer to the first Lagrangian multipliers corresponding to each of the M sample images obtained when the modified dual problem is solved.

[0096] According to the target Lagrange multipliers and Get the projection tensor of the sample image set represents the target Lagrange multiplier corresponding to the mth sample image;

[0097] Performing a tensor rank one decomposition on the projection tensor of the sample image set to obtain an initial projection vector of the tensor of the sample image set at order N;

[0098] The initial projection vector of the sample image set at the Nth order is inversely projected to obtain the target projection vector of the sample image set at the Nth order.

[0099] In this embodiment, by d Perform CP decomposition to convert V d Decomposed into the sum of R first-rank tensors, each of which is represented by the outer product of its components at order N, i.e. Among them, V dk represents the kth first rank tensor, represents the first-order component of the kth first-rank tensor, represents the second-order component of the kth first-rank tensor, represents the component of the kth first-rank tensor at the Nth order; it can be obtained by e Perform CP decomposition and convert V e Decomposed into the sum of R second-rank tensors, each of which is represented by the outer product of its components at order N, i.e. Among them, V eq represents the qth second rank tensor, represents the first-order component of the qth second-rank tensor, represents the second-order component of the qth second-rank tensor, represents the component of the qth second rank tensor at the Nth order. By d and V e CP decomposition can obtain a more compact and meaningful expression of the tensor, thereby effectively obtaining the complete information of the sample image set without storing all tensors in memory space. While saving memory space, it can avoid overfitting of the decision function and enhance the generalization ability of the decision function. In addition, this application is more suitable for small sample learning. When the number of sample images is small, a decision function with higher accuracy can also be obtained, which reduces the image classification time and improves the image classification efficiency.

[0100] In this embodiment, the projection tensor W of the sample image set is * Performing a tensor rank one decomposition, W * It is expressed as the outer product of the tensor of the sample image set and the initial projection vector of order N, that is, In this expression This is the initial projection vector of the sample image set at the Nth order. The initial projection vector of the sample image set at the Nth order is calculated by the formula By performing back projection, the target projection vector of the sample image set at order N can be obtained, where n ranges from 1 to N.

[0101] Sequential Minimal Optimization (SMO) is an algorithm for solving optimization problems generated during the training of tensor machines. The present application can solve the modified dual problem through SMO to solve the target offset scalar and the target Lagrange multiplier corresponding to each of the M sample images.

[0102] It should be noted that the parameters in the tensor learning decision function based on the sample image set in this embodiment can reduce the input parameters of the sample image set, thereby effectively solving problems such as information loss, overfitting, dimensionality disaster, and small samples.

[0103] Step 205 , constructing a decision function according to the target offset scalar and the target projection vector of the tensor of the sample image set at the Nth order.

[0104] Building a decision function X represents the tensor of the medical image to be classified, x (n) represents the n-th order component of the tensor of the medical image to be classified, b * represents the target offset scalar, <·> represents the inner product operator, and sign represents the sign function.

[0105] In this embodiment, the target offset scalar can be called the optimal offset scalar, and the target projection vector of the tensor of the sample image set at the Nth order can be called the optimal projection vector of the tensor of the sample image set at the Nth order. Then, the decision function constructed according to the target offset scalar and the target projection vector of the tensor of the sample image set at the Nth order can also be called the optimized projection based high-order support tensor machine (OPHSTM).

[0106] Step 206: Obtain a medical image to be classified.

[0107] This step is the same as step 101. For details, please refer to the relevant description of step 101, which will not be repeated here.

[0108] It should be noted that the M sample images in the sample image set belong to the same type of images as the medical image to be classified. For example, if the medical image to be classified is a brain MRI image, then the M sample images are also required to be brain MRI images, thereby ensuring that the decision function constructed based on the M sample images can accurately classify the medical image to be classified.

[0109] Optionally, the present application may not limit the image to be classified and the M sample images to be medical images, that is, the present application may also classify images that are not medical images.

[0110] Step 207, obtaining the tensor of the medical image to be classified.

[0111] This step is the same as step 102. For details, please refer to the relevant description of step 102, which will not be repeated here.

[0112] Step 208 , performing a tensor rank-one decomposition on the tensor of the medical image to be classified to obtain the N-th order component of the tensor of the medical image to be classified.

[0113] This step is the same as step 103. For details, please refer to the relevant description of step 103, which will not be repeated here.

[0114] Step 209 , inputting the N-order component of the tensor of the medical image to be classified into the constructed decision function, classifying the medical image to be classified through the decision function, and outputting the classification result of the medical image to be classified.

[0115] This step is the same as step 104. For details, please refer to the relevant description of step 104, which will not be repeated here.

[0116] The embodiment of the present application obtains the parameters of the decision function by constructing a quadratic programming problem, which can reduce the time spent on parameter iteration in the parameter acquisition process. While ensuring efficient processing capabilities, it highlights better classification effects and improves the accuracy of medical image classification, and has strong practicality and popularization.

[0117] See also Figure 3 , is a structural diagram of the medical image classification device provided in Example 3 of the present application. For the sake of convenience of explanation, only the parts related to the embodiment of the present application are shown.

[0118] The medical image classification device includes:

[0119] An image acquisition module 31 is used to acquire a medical image to be classified;

[0120] A tensor acquisition module 32, used to acquire the tensor of the medical image to be classified;

[0121] A tensor decomposition module 33, used for performing a tensor rank-one decomposition on the tensor of the medical image to be classified to obtain a component of the tensor of the medical image to be classified at an Nth order, wherein N is the order of the tensor of the medical image to be classified;

[0122] The component processing module 34 is used to input the N-order components of the tensor of the medical image to be classified into a pre-constructed decision function, classify the medical image to be classified through the decision function, and output the classification result of the medical image to be classified.

[0123] Optionally, the medical image classification device further includes:

[0124] A training set acquisition module is used to acquire a sample training set, wherein the sample training set includes a tensor of a sample image set and a category label of the sample image set, wherein the sample image set includes M sample images, the tensor of the sample image set includes respective tensors of the M sample images, the category label of the sample image set includes respective category labels of the M sample images, the order of the tensor of the sample image set is N, M is an integer greater than 1, and the M sample images are medical images;

[0125] A matrix acquisition module is used to obtain the intra-class scatter matrix of the tensor of the sample image set at each order;

[0126] A problem construction module is used to construct a quadratic programming problem of the tensor of the sample image set at each order according to the intra-class scatter matrices of the tensors of the sample training set and the sample image set at each order. The quadratic programming problem of the tensor of the sample image set at each order includes a first objective function and a first constraint condition. The first objective function in the quadratic programming problem of the tensor of the sample image set at each order minimizes the intra-class scatter and maximizes the inter-class scatter. N orders correspond to N quadratic programming problems.

[0127] Among them, for the nth quadratic programming problem, the nth quadratic programming problem is the quadratic programming problem of the tensor of the sample image set at the nth order, the nth order is any order among the N orders, and the first objective function in the nth quadratic programming problem is:

[0128]

[0129] The tensor representing the sample image set is the within-class scatter matrix of order n, The tensor representing the sample image set is the within-class scatter matrix of the i-th order, w (n) The target projection vector of the tensor representing the sample image set at the nth order, w (i) Represents the target projection vector of the tensor of the sample image set at the i-th order, and E represents the target projection vector with w (n) and w (i) Unit vectors of the same order, ε represents the balance factor, Indicates that w (n) The corresponding slack variable, C represents the penalty factor, b (n) The offset scalar of the tensor representing the sample image set at the nth order;

[0130] The first constraint in the nth quadratic programming problem is:

[0131]

[0132] Xm Represents the tensor of the mth sample image, y m Represents the category label of the mth sample image;

[0133] A problem solving module is used to solve N quadratic programming problems to obtain a target offset scalar and a target projection vector of the tensor of the sample image set at the Nth order. The target offset scalar refers to the offset scalar obtained when solving the N quadratic programming problems.

[0134] Function building module, used to build a decision function based on the target offset scalar and the target projection vector of the tensor of the sample image set at order N X represents the tensor of the medical image to be classified, x (n) Represents the nth-order component of the tensor of the medical image to be classified, b * represents the target offset scalar, <·> represents the inner product operator, and sign represents the sign function.

[0135] Optionally, the problem solving module includes:

[0136] The problem conversion submodule is used to × i represents the i-modular product operator, which transforms N quadratic programming problems into a single N-fold quadratic programming problem, which includes the second objective function and the second constraint. Represents the initial projection vector of the tensor of the sample image set at the nth order, Λ (n) and Λ (i) represents a diagonal matrix, p (n) and p (i) represents an orthogonal matrix, Λ (n) and p (n) satisfy Λ (i) and p (i) satisfy The Nth order corresponds to N initial projection vectors;

[0137] Among them, the second objective function is:

[0138] W * represents the initial projection tensor of the sample image set, which is the outer product of N initial projection vectors, and b represents the offset scalar;

[0139] The second constraint is: m ( <W * ,V m >+b)≥1-ξ m ;

[0140] The problem solving submodule is used to solve a single N-fold quadratic programming problem and obtain the target offset scalar and the target projection vector of the tensor of the sample image set at the Nth order.

[0141] Optionally, the problem solving submodule includes:

[0142] Function construction unit, used to construct Lagrangian function according to a single N-fold quadratic programming problem:

[0143]

[0144] Among them, α represents the first Lagrange multiplier, α m represents the first Lagrange multiplier corresponding to the mth sample image, β represents the second Lagrange multiplier, β m represents the second Lagrange multiplier corresponding to the mth sample image;

[0145] A problem obtaining unit, used for obtaining the dual problem of the Lagrangian function according to the Lagrangian multiplier method, wherein the dual problem of the Lagrangian function includes a third objective function and a third constraint condition;

[0146] Among them, the third objective function is:

[0147] α d represents the first Lagrange multiplier corresponding to the d-th sample image, α e represents the first Lagrange multiplier corresponding to the e-th sample image, y d Represents the category label of the dth sample image, y e represents the category label of the e-th sample image,

[0148] The third constraint is: 0≤α m ≤C;

[0149] The problem solving unit is used to solve the dual problem of the Lagrangian function to obtain the target offset scalar and the target projection vector of the tensor of the sample image set at the Nth order.

[0150] Optionally, the problem solving unit is specifically used for:

[0151] V d Decompose into the sum of R first-rank tensors, each of which is represented as the outer product of its components at order N;

[0152] V e Decomposed into the sum of R second-rank tensors, each of which is represented as the outer product of its components at order N;

[0153] According to each component of the first rank tensor at the Nth order and each component of the second rank tensor at the Nth order, modify the third objective function, and determine the modified third objective function as the fourth objective function, and the fourth objective function and the third constraint condition constitute a modified dual problem;

[0154] Among them, the fourth objective function is:

[0155]

[0156] represents the component of the kth first-rank tensor at the nth order, represents the component of the qth second-rank tensor at the nth order;

[0157] The modified dual problem is solved by using a sequential minimum optimization algorithm to obtain a target offset scalar and target Lagrangian multipliers corresponding to each of the M sample images. The target Lagrangian multipliers corresponding to each of the M sample images refer to the first Lagrangian multipliers corresponding to each of the M sample images obtained when the modified dual problem is solved.

[0158] According to the target Lagrange multipliers and Get the projection tensor of the sample image set represents the target Lagrange multiplier corresponding to the mth sample image;

[0159] Performing a tensor rank one decomposition on the projection tensor of the sample image set to obtain an initial projection vector of the tensor of the sample image set at order N;

[0160] The initial projection vector of the sample image set at the Nth order is inversely projected to obtain the target projection vector of the sample image set at the Nth order.

[0161] Optionally, the component processing module 34 is specifically used for:

[0162] Output the classification results of the medical images to be classified in a preset manner.

[0163] The medical image classification device provided in the embodiment of the present application can be applied in the aforementioned method embodiments 1 and 2. For details, please refer to the description of the aforementioned method embodiments 1 and 2, which will not be repeated here.

[0164] Figure 4 Schematic diagram of the structure of the terminal device provided in the fourth embodiment of the present application. Figure 4 As shown, the terminal device 4 of this embodiment includes: one or more processors 40 (only one is shown in the figure), a memory 41, and a computer program 42 stored in the memory 41 and executable on the processor 40. When the processor 40 executes the computer program 42, the steps in the above-mentioned various medical image classification method embodiments are implemented.

[0165] The terminal device 4 may be a computing device such as a desktop computer, a notebook, a PDA, or a cloud server. The terminal device may include, but is not limited to, a processor 40 and a memory 41. Those skilled in the art will appreciate that Figure 4 It is only an example of the terminal device 4 and does not constitute a limitation on the terminal device 4. It may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, the terminal device may also include input and output devices, network access devices, buses, etc.

[0166] The processor 40 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc.

[0167] The memory 41 may be an internal storage unit of the terminal device 4, such as a hard disk or memory of the terminal device 4. The memory 41 may also be an external storage device of the terminal device 4, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the terminal device 4. Further, the memory 41 may also include both an internal storage unit and an external storage device of the terminal device 4. The memory 41 is used to store the computer program and other programs and data required by the terminal device. The memory 41 may also be used to temporarily store data that has been output or is to be output.

[0168] The technicians in the relevant field can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In practical applications, the above-mentioned function allocation can be completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated in a processing unit, or each unit can exist physically separately, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned device can refer to the corresponding process in the aforementioned method embodiment, which will not be repeated here.

[0169] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0170] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0171] In the embodiments provided in the present application, it should be understood that the disclosed devices / terminal equipment and methods can be implemented in other ways. For example, the device / terminal equipment embodiments described above are only schematic. For example, the division of the modules or units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0172] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0173] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0174] If the integrated module / unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present application implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0175] The present application implements all or part of the processes in the above-mentioned embodiment method, and may also be completed through a computer program product. When the computer program product runs on a terminal device, the terminal device can implement the steps in the above-mentioned method embodiments when executing the computer program product.

[0176] The embodiments described above are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. A medical image classification method, characterized in that: The medical image classification method comprises: Acquire a medical image to be classified; Acquire the tensor of the medical image to be classified; Performing a tensor rank-one decomposition on the tensor of the medical image to be classified to obtain an N-order component of the tensor of the medical image to be classified, wherein N refers to the order of the tensor of the medical image to be classified; Inputting the N-order component of the tensor of the medical image to be classified into a pre-constructed decision function, classifying the medical image to be classified by the decision function, and outputting the classification result of the medical image to be classified; Before obtaining the medical image to be classified, it also includes: Acquire a sample training set, the sample training set comprising a tensor of a sample image set and a category label of the sample image set, the sample image set comprising M sample images, the tensor of the sample image set comprising respective tensors of the M sample images, the category label of the sample image set comprising respective category labels of the M sample images, the order of the tensor of the sample image set being N, M being an integer greater than 1, and the M sample images being medical images; Obtaining a scatter matrix of the tensor of the sample image set within each class; According to the intra-class scatter matrices of the tensors of the sample training set and the sample image set at each order, a quadratic programming problem of the tensor of the sample image set at each order is constructed, wherein the quadratic programming problem of the tensor of the sample image set at each order includes a first objective function and a first constraint condition, and the first objective function of the quadratic programming problem of the tensor of the sample image set at each order minimizes the intra-class scatter and maximizes the inter-class scatter, and N orders correspond to N quadratic programming problems; Wherein, for the nth quadratic programming problem, the nth quadratic programming problem is a quadratic programming problem of the tensor of the sample image set at the nth order, the nth order is any order among the N orders, and the first objective function in the nth quadratic programming problem is: The tensor representing the sample image set is the within-class scatter matrix of the nth order, The tensor representing the sample image set is the scatter matrix of the i-th order within the class, w (n) represents the target projection vector of the tensor of the sample image set at the nth order, w (i) represents the target projection vector of the tensor of the sample image set at the i-th order, and E represents the target projection vector of w (n) and w (i) Unit vectors of the same order, ε represents the balance factor, Indicates that w (n) The corresponding slack variable, C represents the penalty factor, b (n) A scalar representing the offset of the tensor of the sample image set at the nth order; The first constraint in the nth quadratic programming problem is: X m Represents the tensor of the mth sample image, y m Represents the category label of the mth sample image; Solving the N quadratic programming problems to obtain a target offset scalar and a target projection vector of the tensor of the sample image set at order N, wherein the target offset scalar refers to an offset scalar obtained when solving the N quadratic programming problems; The decision function is constructed according to the target offset scalar and the target projection vector of the tensor of the sample image set at the Nth order. X represents the tensor of the medical image to be classified, x (n) represents the nth-order component of the tensor of the medical image to be classified, b * represents the target offset scalar, <·> represents the inner product operator, and sign represents the sign function.

2. The medical image classification method according to claim 1, characterized in that: The step of solving the N quadratic programming problems to obtain a target offset scalar and a target projection vector of the tensor of the sample image set at the Nth order includes: make converting the N quadratic programming problems into a single N-fold quadratic programming problem, wherein the single N-fold quadratic programming problem includes a second objective function and a second constraint condition, represents the initial projection vector of the tensor of the sample image set at the nth order, V m Represents the tensor of the mth sample image after projection along each order, × i represents the i-modular product operator, Λ (n) and Λ (i) represents a diagonal matrix, p (n) and p (i) represents an orthogonal matrix, Λ (n) and p (n) satisfy Λ (i) and p (i) satisfy N order corresponds to N initial projection vectors; Wherein, the second objective function is: W * represents the initial projection tensor of the sample image set, which is the outer product of N initial projection vectors, and b represents the offset scalar; The second constraint condition is: m ( <W * ,V m >+b)≥1-ξ m ; The single N-fold quadratic programming problem is solved to obtain the target offset scalar and the target projection vector of the tensor of the sample image set at the Nth order.

3. The medical image classification method according to claim 2, characterized in that: The step of solving the single N-fold quadratic programming problem to obtain the target offset scalar and the target projection vector of the tensor of the sample image set at the Nth order includes: According to the single N-fold quadratic programming problem, a Lagrangian function is constructed: Among them, α represents the first Lagrange multiplier, α m represents the first Lagrange multiplier corresponding to the mth sample image, β represents the second Lagrange multiplier, β m represents the second Lagrange multiplier corresponding to the mth sample image; According to the Lagrange multiplier method, a dual problem of the Lagrangian function is obtained, wherein the dual problem of the Lagrangian function includes a third objective function and a third constraint condition; Wherein, the third objective function is: α d represents the first Lagrange multiplier corresponding to the d-th sample image, α e represents the first Lagrange multiplier corresponding to the e-th sample image, y d Represents the category label of the dth sample image, y e represents the category label of the e-th sample image, V d Represents the tensor of the d-th sample image after projection along each order, V e Represents the tensor of the e-th sample image after projection along each order; The third constraint condition is: The dual problem of the Lagrangian function is solved to obtain the target offset scalar and the target projection vector of the tensor of the sample image set at order N.

4. The medical image classification method according to claim 3, characterized in that: The dual problem of solving the Lagrangian function to obtain the target offset scalar and the target projection vector of the tensor of the sample image set at order N includes: V d Decompose into the sum of R first-rank tensors, each of which is represented as the outer product of its components at order N; V e Decomposed into the sum of R second-rank tensors, each of which is represented as the outer product of its components at order N; Modify the third objective function according to the components of each first-rank tensor at the Nth order and the components of each second-rank tensor at the Nth order, and determine the modified third objective function as the fourth objective function, wherein the fourth objective function and the third constraint condition constitute a modified dual problem; Wherein, the fourth objective function is: represents the component of the kth first-rank tensor at the nth order, represents the component of the qth second-rank tensor at the nth order; Using a sequential minimum optimization algorithm to solve the modified dual problem, the target offset scalar and the target Lagrangian multipliers corresponding to each of the M sample images are obtained, wherein the target Lagrangian multipliers corresponding to each of the M sample images refer to the first Lagrangian multipliers corresponding to each of the M sample images obtained when the modified dual problem is solved; According to the target Lagrange multipliers and Get the projection tensor of the sample image set represents the target Lagrange multiplier corresponding to the mth sample image; Performing a tensor rank-one decomposition on the projection tensor of the sample image set to obtain an initial projection vector of the tensor of the sample image set at order N; The initial projection vector of the sample image set at the Nth order is back-projected to obtain the target projection vector of the sample image set at the Nth order.

5. The medical image classification method according to any one of claims 1 to 4, characterized in that: The outputting the classification result of the medical image to be classified comprises: The classification result of the medical image to be classified is output in a preset manner.

6. A medical image classification device, characterized in that: The medical image classification device comprises: An image acquisition module, used for acquiring medical images to be classified; A tensor acquisition module, used to acquire the tensor of the medical image to be classified; A tensor decomposition module, used for performing a tensor rank-one decomposition on the tensor of the medical image to be classified to obtain an N-order component of the tensor of the medical image to be classified, wherein N is the order of the tensor of the medical image to be classified; A component processing module, used for inputting the N-order component of the tensor of the medical image to be classified into a pre-constructed decision function, classifying the medical image to be classified through the decision function, and outputting the classification result of the medical image to be classified; The medical image classification device also includes: A training set acquisition module, used to acquire a sample training set, wherein the sample training set includes a tensor of a sample image set and a category label of the sample image set, wherein the sample image set includes M sample images, the tensor of the sample image set includes respective tensors of the M sample images, the category label of the sample image set includes respective category labels of the M sample images, the order of the tensor of the sample image set is N, M is an integer greater than 1, and the M sample images are medical images; A matrix acquisition module, used for acquiring the intra-class scatter matrix of the tensor of the sample image set at each order; A problem construction module, used for constructing a quadratic programming problem of the tensor of the sample image set at each order according to the intra-class scatter matrices of the tensors of the sample training set and the sample image set at each order, wherein the quadratic programming problem of the tensor of the sample image set at each order includes a first objective function and a first constraint condition, and the first objective function in the quadratic programming problem of the tensor of the sample image set at each order minimizes the intra-class scatter and maximizes the inter-class scatter, and N orders correspond to N quadratic programming problems; Wherein, for the nth quadratic programming problem, the nth quadratic programming problem is a quadratic programming problem of the tensor of the sample image set at the nth order, the nth order is any order among the N orders, and the first objective function in the nth quadratic programming problem is: The tensor representing the sample image set is the within-class scatter matrix of the nth order, The tensor representing the sample image set is the scatter matrix of the i-th order within the class, w (n) represents the target projection vector of the tensor of the sample image set at the nth order, w (i) represents the target projection vector of the tensor of the sample image set at the i-th order, and E represents the target projection vector of w (n) and w (i) Unit vectors of the same order, ε represents the balance factor, Indicates that w (n) The corresponding slack variable, C represents the penalty factor, b (n) A scalar representing the offset of the tensor of the sample image set at the nth order; The first constraint in the nth quadratic programming problem is: X m Represents the tensor of the mth sample image, y m Represents the category label of the mth sample image; A problem solving module, used for solving the N quadratic programming problems, obtaining a target offset scalar and a target projection vector of the tensor of the sample image set at the Nth order, wherein the target offset scalar refers to an offset scalar obtained when solving the N quadratic programming problems; A function construction module, used to construct the decision function according to the target offset scalar and the target projection vector of the tensor of the sample image set at the Nth order X represents the tensor of the medical image to be classified, x (n) represents the nth-order component of the tensor of the medical image to be classified, b * represents the target offset scalar, <·> represents the inner product operator, and sign represents the sign function.

7. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the medical image classification method according to any one of claims 1 to 5 are implemented.

8. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the medical image classification method according to any one of claims 1 to 5 are implemented.

Images