Pathological image classification method

Abstract

The invention relates to the technical field of image processing, in particular to a pathological image classification method, which comprises the following steps: acquiring a historical medical pathological image; obtaining a historical medical pathology image data set; based on the historical medical pathology image data set, a mixed learning model based on the Colmogov-Arnod representation theorem is trained to obtain an image classification model, the mixed learning model comprises a feature extraction assembly and a classification assembly, the feature extraction assembly is used for outputting an image feature set based on the historical medical pathology image data set, and the classification assembly is used for classifying the image feature set based on the historical medical pathology image data set; the classification component is used for outputting a classification result based on an image feature set wavelet transform formula, and the wavelet transform formula is determined based on significant medical features of the historical pathological image data set; and determining the classification result of the to-be-identified pathological image based on the to-be-identified pathological image and the image classification model, and by adopting the Colmogorov-Arnod theorem network, the complexity and parameter quantity of the network structure can be reduced, the risk of overfitting can be reduced, and the generalization performance can be improved.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and in particular, to a method for classifying pathological images. Background Art

[0002] Pathological image classification is an important field of medical image analysis. Pathological images are mainly tissue section images taken by microscopes for medical diagnosis and research, especially for identifying and classifying various diseases to assist doctors in making quick and accurate diagnoses. Pathological diagnosis is considered one of the most reliable diagnostic methods for diseases and is known as the gold standard for disease diagnosis. With the development of computational pathology and the progress of pathological diagnosis equipment, the resolution of whole slide pathological images (WSI) is getting higher and higher, and the image files are getting larger and larger. Using machine learning and deep learning for pathological image classification has become the mainstream method. In particular, the introduction of deep learning has greatly improved the accuracy and efficiency of classification.

[0003] Traditional pathological image classification in the deep learning framework includes two main types of networks: feature extraction networks and classification networks. Feature extraction networks include ResNet based on CNN (Convolutional Neural Network), Transformer with self-attention mechanism, etc. These networks extract and learn local features of images through components. Classification networks include MLP (Multi-Layer Perceptron), etc. These networks calculate classification probabilities based on image features and determine the final classification. It is a feedforward neural network composed of an input layer, one or more hidden layers, and an output layer. Neurons in each layer are connected to all neurons in the previous layer, forming a fully connected structure. The important theoretical basis of the MLP is the Universal Approximation Theory (UAT). It simulates complex functional relationships through the connection of multiple layers of neurons. The MLP network uses a fixed activation function at the nodes, that is, the same activation function is used throughout the network, rather than dynamically selecting different activation functions according to different layers or tasks. This method has the advantages of simplicity, ease of use, and network consistency, but it also has disadvantages such as poor flexibility, gradient problems, and performance limitations, which in turn affect the final classification performance.

[0004] Deep learning models have achieved remarkable performance in medical image classification tasks with the development of large models. Large models have significant advantages in terms of performance and functionality, being able to handle complex tasks and achieve excellent performance. However, large models have high storage requirements and intensive computational resources due to their large number of parameters. At the same time, large models have problems such as high data requirements in order to reduce overfitting. These problems make large models face challenges and limitations such as high computational and storage resource requirements, large data requirements, high model complexity, and poor interpretability in high-resolution pathological image classification tasks. These limitations not only affect the performance of the model but also limit its wide adoption in practical applications.

[0005] Therefore, optimizing the model structure, reducing computational and storage costs, improving the interpretability and security of the model, etc. can further promote the application and development of large models. Summary of the Invention

[0006] In view of the above problems, the present invention provides a pathological image classification method that overcomes or at least partially solves the above problems.

[0007] In a first aspect, the present invention provides a pathological image classification method, including: Obtaining historical medical pathological images, where the historical medical pathological images have classification labels; Performing data augmentation processing on the historical medical pathological images to obtain a historical medical pathological image dataset; Based on the historical medical pathological image dataset, training a hybrid learning model based on the Kolmogorov - Arnold representation theorem to obtain an image classification model. The hybrid learning model includes a feature extraction component and a classification component. The feature extraction component is used to output an image feature set based on the historical medical pathological image dataset, and the classification component is used to output a classification result based on the image feature set and a wavelet transform formula, where the wavelet transform formula is determined based on the significant medical features of the historical pathological image dataset; Obtaining a pathological image to be recognized; Based on the pathological image to be recognized and the image classification model, determining the classification result of the pathological image to be recognized.

[0008] Preferably, performing data augmentation processing on the historical medical pathological images to obtain a historical medical pathological image dataset includes: Performing shearing and flipping operations on the historical medical pathological images to generate modified historical medical pathological images, and forming a historical medical pathological image dataset with the unmodified historical medical pathological images and the modified historical medical pathological images.

[0009] Preferably, the feature extraction component is used to output an image feature set based on the historical medical pathological image dataset, including: The feature extraction component is used to extract features of corresponding receptive fields based on a historical medical pathological image dataset by using different convolutional kernels, and perform a non-linear transformation on the features to obtain a first target feature; Based on the first target feature, a second target feature is extracted through pooling; Based on the first target feature and the second target feature, the semantic feature information of the historical medical pathological image is extracted through continuous iteration, and an image feature set is output.

[0010] Preferably, the feature extraction component is used to extract features of corresponding receptive fields based on a historical medical pathological image dataset by using different convolutional kernels, and perform a non-linear transformation on the features to obtain a first target feature, which is specifically implemented by the following formula:

[0011] wherein, is the number of hidden layers, is the convolutional kernel size, is the weight matrix of the feature extraction component, is the bias vector, is the activation function, is the number of channels, is the historical medical pathological image, is the first target feature.

[0012] Preferably, based on the first target feature, a second target feature is extracted through a pooling layer, which is specifically implemented by the following formula: ; is the pooling function, is the second target feature.

[0013] Preferably, based on the first target feature and the second target feature, the semantic feature information of the historical medical pathological image is extracted through continuous iteration, and an image feature set is output, which is specifically implemented by the following formula: ; is the iteration symbol, is the image feature set.

[0014] Preferably, the classification component is used to output a classification result based on the image feature set and a wavelet transform formula, including: The classification component is used to perform a translation transformation and a scaling transformation on the image feature set to obtain a medical significant feature; The wavelet transform formula is applied to the medical significant feature to obtain a classification result.

[0015] Preferably, wavelet transform is applied to the medically significant features to obtain a classification result, including: Wavelet transform is applied to the medically significant features according to the following formula to obtain a classification result: ; wherein, is the Mexican hat function, which is the second derivative of the Gauss function, is the scale parameter, which is used to control the width of the wavelet, is the medically significant feature.

[0016] Preferably, wavelet transform is applied to the medically significant features to obtain a classification result, including: Wavelet transform is applied to the medically significant features according to the following formula to obtain a classification result: ; wherein, is the Gaussian wavelet function, which is the first derivative of the Gauss function, is the medically significant feature.

[0017] Preferably, based on the pathological image to be recognized and the image classification model, the classification result of the pathological image to be recognized is determined, including: The pathological image to be recognized is input into the image classification model to obtain the classification result of the pathological image to be recognized.

[0018] In a second aspect, the present invention further provides a pathological image classification device, including: A first acquisition module, configured to acquire historical medical pathological images, where the historical medical pathological images have classification labels; An enhancement module, configured to perform data enhancement processing on the historical medical pathological images to obtain a historical medical pathological image data set; A training module, configured to train a hybrid learning model based on the Kolmogorov - Arnold representation theorem based on the historical medical pathological image data set to obtain an image classification model. The hybrid learning model includes a feature extraction component and a classification component. The feature extraction component is configured to output an image feature set based on the historical medical pathological image data set, and the classification component is configured to output a classification result based on the image feature set and a wavelet transform formula, where the wavelet transform formula is determined based on the significant medical features of the historical pathological image data set; A second acquisition module, configured to acquire a pathological image to be recognized; A determination module, configured to determine the classification result of the pathological image to be recognized based on the pathological image to be recognized and the image classification model.

[0019] In a third aspect, the present invention further provides a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, the method described in the first aspect is implemented.

[0020] In a fourth aspect, the present invention further provides a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, the method described in the first aspect is implemented.

[0021] One or more technical solutions in the embodiments of the present invention have at least the following technical effects or advantages: The present invention provides a method for classifying pathological images, including: obtaining historical medical pathological images with classification labels; performing data augmentation processing on the historical medical pathological images to obtain a historical medical pathological image dataset; training a hybrid learning model based on the Kolmogorov - Arnold representation theorem on the basis of the historical medical pathological image dataset to obtain an image classification model. The hybrid learning model includes a feature extraction component and a classification component. The feature extraction component is used to output an image feature set based on the historical medical pathological image dataset, and the classification component is used to output a classification result based on the wavelet transform formula of the image feature set, and the wavelet transform formula is determined based on the significant medical features of the historical pathological image dataset; obtaining a pathological image to be recognized; determining the classification result of the pathological image to be recognized based on the pathological image to be recognized and the image classification model. By adopting the Kolmogorov - Arnold theorem network, the complexity of the network structure and the number of parameters can be reduced, the risk of overfitting can be reduced, the generalization performance can be improved, and the classification result can be made more accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] By reading the detailed description of the preferred embodiments below, various other advantages and benefits will become clear to those of ordinary skill in the art. The drawings are only for the purpose of showing the preferred embodiments and are not considered to be a limitation of the present invention. Moreover, throughout the drawings, the same reference numerals are used to represent the same components. In the drawings: Figure 1 The schematic diagram of the step flow of the pathological image classification method in the embodiment of the present invention is shown; Figure 2 The schematic diagram of the hybrid learning model based on the Kolmogorov - Arnold representation theorem in the embodiment of the present invention is shown; Figure 3 The schematic diagram of the structure of the pathological image classification device in the embodiment of the present invention is shown; Figure 4 The schematic diagram of the computer device for implementing the pathological image classification method in the embodiment of the present invention is shown. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0023] Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided so that the present disclosure can be more thoroughly understood and the scope of the present disclosure can be fully conveyed to those skilled in the art.

[0024] Embodiment 1

[0025] An embodiment of the present invention provides a method for classifying pathological images, as Figure 1 shown, including: S101, obtaining historical medical pathological images, where the historical medical pathological images have classification labels; S102, performing data augmentation processing on the historical medical pathological images to obtain a historical medical pathological image dataset; S103, training a hybrid learning model based on the Kolmogorov - Arnold representation theorem on the historical medical image dataset to obtain an image classification model. The hybrid learning model includes a feature extraction component and a classification component. The feature extraction component is used to output an image feature set based on the historical medical pathological image dataset, and the classification component is used to output a classification result based on the image feature set wavelet transform formula, where the wavelet transform formula is determined based on the significant medical features of the historical pathological image dataset; S104, obtaining a pathological image to be recognized; S105, determining the classification result of the pathological image to be recognized based on the pathological image to be recognized and the image classification model.

[0026] First, historical medical pathological images of skin cancer or breast cancer with different resolutions can be obtained, as well as corresponding classification labels. The classification labels are marked with labels belonging to breast cancer or skin cancer. Of course, for normal skin tissue pictures, normal skin tissue is also correspondingly marked, and for normal breast tissue atlases, normal breast tissue is also correspondingly marked, and so on.

[0027] Since the number of these historical medical pathological images of skin cancer or breast cancer is limited, more historical medical pathological images can be obtained through data augmentation processing.

[0028] Specifically, perform shearing and flipping operations on the historical medical pathological images to generate modified historical medical pathological images, and form a historical medical pathological image dataset with the unmodified historical medical pathological images and the modified historical medical pathological images.

[0029] Of course, it can also be a rotation operation, etc. Through the above data augmentation processing, the historical medical pathological images are amplified, making the corresponding dataset richer and the trained model more generalizable.

[0030] Before performing data augmentation, the size of these historical medical pathology images can be adjusted first. After data augmentation, it is then converted into a type suitable for model training.

[0031] Next, execute S103 to train a hybrid learning model based on the Kolmogorov - Arnold representation theorem on the historical medical pathology image dataset to obtain an image classification model.

[0032] First, the Kolmogorov - Arnold representation theorem assumes that any continuous function can be represented as a combination of continuous univariate functions of a finite number of variables. This provides a theoretical basis for constructing a general neural network model, the Kolmogorov - Arnold network (KAN, Kolmogorov - Arnold Network). The core idea of KAN is to transform the approximation problem of multivariate functions into the problem of learning a set of univariate functions. In principle, KAN only contains two layers, consisting of an internal function and an external function, corresponding to different input and output difficulties respectively. Thus, it significantly reduces the parameter scale of the classification network, reduces storage and computing requirements, and improves learning efficiency. Of course, KAN can also be extended to consist of a series of cascaded Kolmogorov - Arnold layers. Each layer contains a set of learnable one - dimensional activation functions, and the rotation of the activation function is selected according to the characteristics of the problem domain and the application scenario. This design method of stacking KAN layers not only extends the depth of the KAN network but also maintains the interpretability and expressive power of the network because each layer is composed of univariate functions, and the internal mechanism is relatively transparent, allowing the functions to be learned and understood separately. At the same time, KAN has local plasticity and can use learnable activation functions such as spline functions and limitations to avoid the catastrophic forgetting effect of MLP. KAN can also use a smaller layer width, reducing the complexity of the network structure and the number of parameters, reducing the risk of overfitting, and improving generalization performance.

[0033] As Figure 2 shown, it is a schematic diagram of a hybrid learning model based on the Kolmogorov - Arnold representation theorem. The hybrid learning module includes a feature extraction component CNN and a classification component KAN. Among them, the CNN component can be a custom simple convolutional neural network or a classic CNN network such as the ResNet network, etc. As a general framework, the KAN component can select univariate functions according to the application scenario.

[0034] The feature extraction component is used to output an image feature set based on the historical medical pathology image dataset. Specifically: The feature extraction component is used to, based on the historical medical image dataset, use different convolutions to extract features of corresponding receptive fields and perform a non - linear transformation on the features to obtain the first target feature; Extract the second target feature through pooling based on the first target feature; Based on the first target feature and the second target feature, extract the semantic feature information of the historical medical pathological image through continuous iteration, and output an image feature set.

[0035] Among them, in a convolutional neural network, the receptive field refers to the area of the input image that a certain point on the feature map can see, that is, the point on the feature map is calculated from the area of the receptive field size in the input image. The larger the value of the neuron receptive field, the larger the range of the original image it can touch, which also means that it may contain more global and higher-level semantic features; the smaller the value, the more local and detailed the features it contains. Therefore, the value of the receptive field can be used to roughly judge the abstraction level of each layer.

[0036] The process of obtaining the first target feature is specifically implemented using the following formula: ; Among them, is the number of hidden layers, is the convolutional kernel size, is the weight matrix of the feature extraction component, is the bias vector, is the activation function, is the number of channels, which refers to the number of dimensions in the data group and is used to store and process multiple features. is the historical medical pathological image. is the calculation formula for the first hidden layer in the image extraction component; is the calculation formula for other hidden layers (L-1 hidden layers) in the image extraction component. Thus, the first target feature is obtained, that is, .

[0037] Next, the process of obtaining the second target feature is specifically implemented using the following formula: ; is the pooling function, is the first target feature, is the second target feature.

[0038] Then, based on the first target feature and the second target feature, extract the semantic feature information of the historical medical pathological image through continuous iteration, and output an image feature set. The iterative process is specifically implemented using the following formula: ; is the iteration symbol, is the image feature set.

[0039] Through the above process of feature extraction, an image feature set is obtained.

[0040] After the feature component extracts features, a classification component is required for classification, including: The classification component is used to perform translation transformation and scaling transformation on the image feature set to obtain medically significant features; The wavelet transform formula is applied to the medically significant features to obtain the classification result.

[0041] The translation transformation and scaling transformation of the image feature set to obtain medically significant features are specifically reflected by the following formula:

[0042] Among them, the translation transformation is to add a constant to each eigenvalue to change the central position of the data.

[0043] The scaling transformation includes: standardization, normalization, and interval scaling. Applying these transformations can improve the model accuracy and accelerate gradient descent.

[0044] Next is the wavelet transform processing: The wavelet transform formula used here is specifically determined based on the significant medical features of the historical pathological image dataset.

[0045] In the present invention, two wavelet transform formulas are provided. The first wavelet transform formula: ; Among them, is the Mexican Hat function, which is the second derivative of the Gauss function, is the scale parameter, which is used to control the width of the wavelet, is the medically significant feature.

[0046] The second wavelet transform formula: ; Among them, is the Gaussian wavelet function, which is the first derivative of the Gauss function, is the medically significant feature.

[0047] In addition, other wavelet transform formulas can be selected according to the feature differences of historical medical pathological images, which are not limited here.

[0048] Since the Mexican Hat and DOG basis functions of the wavelet transform are sensitive to the geometric features of the image, therefore, the wavelet transform is used to classify the significant medical features of the historical pathological image data. The medical features here are geometric features, such as features like shape, edge, texture, etc. Therefore, it is necessary to first perform translation transformation and scaling transformation on the geometric features of the historical pathological image to obtain significant medical features, and then perform classification through wavelet transform.

[0049] Through any of the above wavelet transform formulas, the classification result can be obtained.

[0050] Train the hybrid learning model in the above manner to obtain an image classification model.

[0051] The image classification model can identify whether there is an image of breast cancer in the pathological image of breast tissue, or can identify whether there is an image of skin cancer in the pathological image of skin tissue.

[0052] After obtaining the image classification model, execute S104 to obtain the pathological image to be recognized; S105, based on the pathological image to be recognized and the image classification model, determine the classification result of the pathological image to be recognized.

[0053] Specifically, input the pathological image to be recognized into the image classification model to obtain the classification result of the pathological image to be recognized.

[0054] Of course, when using the image classification model for classification, specifically, the wavelet transform formula is used for classification. This will not be elaborated here.

[0055] By adopting such a classification method, the storage capacity is effectively reduced, and thus the classification result is more accurate.

[0056] One or more technical solutions in the embodiments of the present invention have at least the following technical effects or advantages: The present invention provides a method for classifying pathological images, including: obtaining historical medical pathological images, the historical medical pathological images having classification labels; performing data augmentation processing on the historical medical pathological images to obtain a historical medical pathological image data set; training a hybrid learning model based on the Kolmogorov - Arnold representation theorem based on the historical medical pathological image data set to obtain an image classification model, the hybrid learning model including a feature extraction component and a classification component, the feature extraction component being used to output an image feature set based on the historical medical pathological image data set, the classification component being used to output a classification result based on the image feature set and a wavelet transform formula, the wavelet transform formula being determined based on the significant medical features of the historical pathological image data set; obtaining the pathological image to be recognized; based on the pathological image to be recognized and the image classification model, determining the classification result of the pathological image to be recognized. By adopting the Kolmogorov - Arnold theorem network, the complexity of the network structure and the number of parameters can be reduced, the risk of overfitting can be reduced, the generalization performance can be improved, and the classification result can be made more accurate.

[0057] Embodiment 2 Based on the same inventive concept, the embodiments of the present invention further provide a pathological image classification device, as Figure 3 shown, including: The first acquisition module 301 is configured to acquire historical medical pathological images, and the historical medical pathological images have classification labels; The enhancement module 302 is configured to perform data enhancement processing on the historical medical pathological images to obtain a historical medical pathological image dataset; The training module 303 is configured to train a hybrid learning model based on the Kolmogorov - Arnold representation theorem based on the historical medical pathological image dataset to obtain an image classification model. The hybrid learning model includes a feature extraction component and a classification component. The feature extraction component is configured to output an image feature set based on the historical medical pathological image dataset, and the classification component is configured to output a classification result based on the image feature set and a wavelet transform formula, and the wavelet transform formula is determined based on the significant medical features of the historical pathological image dataset; The second acquisition module 304 is configured to acquire a pathological image to be recognized; The determination module 305 is configured to determine the classification result of the pathological image to be recognized based on the pathological image to be recognized and the image classification model.

[0058] In an alternative embodiment, the enhancement module 302 is configured to: Perform shearing and flipping operations on the historical medical pathological images to generate modified historical medical pathological images, and form a historical medical pathological image dataset with the unmodified historical medical pathological images.

[0059] In an alternative embodiment, the feature extraction component is configured to: Based on the historical medical pathological image dataset, extract features of corresponding receptive fields using different convolutional kernels, and perform non - linear transformation on the features to obtain first target features; Based on the first target features, extract second target features through pooling; Based on the first target features and the second target features, through continuous iteration, extract semantic feature information of the historical medical pathological images, and output an image feature set.

[0060] In an alternative embodiment, the feature extraction component is specifically implemented using the following formula:

[0061] where, is the number of hidden layers, is the size of the convolutional kernel, is the weight matrix of the feature extraction component, is the bias vector, is the activation function, is the number of channels, It is a historical medical pathological image, which is the first target feature.

[0062] In an alternative embodiment, the feature extraction component is specifically implemented using the following formula: ; is a pooling function, which is the second target feature.

[0063] In an alternative embodiment, the feature extraction component is specifically implemented using the following formula: ; is an iteration symbol, which is the image feature set.

[0064] In an alternative embodiment, the classification component is used to: perform a translation transformation and a scaling transformation on the image feature set to obtain a medical significant feature; apply a wavelet transform formula to the medical significant feature to obtain a classification result.

[0065] In an alternative embodiment, the classification component is used to: perform a wavelet transform on the medical significant feature using the following formula to obtain a classification result: ; where, is the Mexican hat function, which is the second derivative of the Gauss function, is the scale parameter, which is used to control the width of the wavelet, is the medical significant feature.

[0066] In an alternative embodiment, the classification component is used to: perform a wavelet transform on the medical significant feature using the following formula to obtain a classification result: ; where, is the Gaussian wavelet function, which is the first derivative of the Gauss function, is the medical significant feature.

[0067] In an alternative embodiment, the determination module 305 is used to: input the pathological image to be recognized into the image classification model to obtain the classification result of the pathological image to be recognized.

[0068] Embodiment III Based on the same inventive concept, an embodiment of the present invention provides a computer device, such as Figure 4As shown, it includes a memory 404, a processor 402, and a computer program stored on the memory 404 and executable on the processor 402. When the processor 402 executes the program, it implements the steps of the above-mentioned pathological image classification method.

[0069] Among them, in Figure 4 the bus architecture (represented by bus 400), bus 400 can include any number of interconnected buses and bridges. Bus 400 links various circuits including one or more processors represented by processor 402 and memory represented by memory 404 together. Bus 400 can also link various other circuits such as peripheral devices, voltage regulators, and power management circuits, etc. These are well known in the art, so they will not be further described herein. Bus interface 406 provides an interface between bus 400 and receiver 401 and transmitter 403. Receiver 401 and transmitter 403 can be the same element, that is, a transceiver, which provides a unit for communicating with various other devices on the transmission medium. Processor 402 is responsible for managing bus 400 and general processing, while memory 404 can be used to store data used by processor 402 when performing operations.

[0070] Embodiment 4 Based on the same inventive concept, an embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, it implements the steps of the above-mentioned pathological image classification method.

[0071] The algorithms and displays provided herein are not inherently related to any particular computer, virtual system, or other device. Various general-purpose systems can also be used in conjunction with the teachings herein. Based on the above description, the structure required to construct such systems is obvious. In addition, the present invention is not directed to any particular programming language. It should be understood that the content of the present invention described herein can be implemented using various programming languages, and the description of a particular language above is to disclose the best implementation manner of the present invention.

[0072] In the specification provided herein, a large number of specific details are described. However, it can be understood that the embodiments of the present invention can be practiced without these specific details. In some instances, well-known methods, structures, and technologies are not shown in detail so as not to obscure the understanding of this specification.

[0073] Similarly, it should be understood that, for the purpose of streamlining the present disclosure and assisting in understanding one or more of the various inventive aspects, in the foregoing description of the exemplary embodiments of the present invention, the various features of the present invention are sometimes grouped together into a single embodiment, figure, or description thereof. However, the disclosed method should not be construed as reflecting an intention that the claimed invention requires more features than are expressly recited in each embodiment. Rather, as reflected in each embodiment, the inventive aspects lie in less than all the features of the single foregoing disclosed embodiment. Thus, the claims following the detailed description are hereby expressly incorporated into the detailed description, with each claim standing on its own as a separate embodiment of the present invention.

[0074] Those skilled in the art will appreciate that the modules in the devices in the embodiments can be adaptively changed and disposed in one or more devices different from those of the embodiments. The modules or units or components in the embodiments can be combined into one module or unit or component, and in addition, they can be divided into multiple sub-modules or sub-units or sub-components. Except that at least some of such features and / or processes or units are mutually exclusive, any combination can be used to combine all the features disclosed in this specification (including the accompanying claims, abstract, and drawings) and all the processes or units of any method or device so disclosed. Unless otherwise expressly stated, each feature disclosed in this specification (including the accompanying claims, abstract, and drawings) can be replaced by an alternative feature that provides the same, equivalent, or similar purpose.

[0075] In addition, those skilled in the art will be able to understand that, although some of the embodiments herein include certain features included in other embodiments rather than other features, the combination of the features of different embodiments means that it is within the scope of the present invention and forms different embodiments. For example, in the detailed description, any one of the claimed embodiments can be used in any combination.

[0076] Each component embodiment of the present invention can be implemented in hardware, or in software modules running on one or more processors, or in a combination thereof. Those skilled in the art should understand that a microprocessor or a digital signal processor (DSP) can be used in practice to implement some or all of the functions of some or all of the components of the pathological image classification device and the computer device according to the embodiments of the present invention. The present invention can also be implemented as a device or device program (e.g., a computer program and a computer program product) for executing part or all of the methods described herein. Such a program for implementing the present invention can be stored on a computer-readable medium, or can be in the form of one or more signals. Such signals can be downloaded from an Internet website, or provided on a carrier signal, or in any other form.

[0077] It should be noted that the above embodiments illustrate the present invention rather than limit the present invention, and those skilled in the art can design alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses shall not be construed as limiting the claim. The word "comprising" does not exclude the presence of elements or steps not listed in the claim. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The present invention can be implemented by means of hardware including several different elements and by means of a suitably programmed computer. In the unit claims listing several devices, several of these devices can be embodied by the same item of hardware. The use of the words first, second, and third, etc. does not denote any order. These words can be interpreted as names.

Claims

1. A pathological image classification method, characterized in that: include: Acquire a historical medical pathology image, wherein the historical medical pathology image has a classification label; Performing data enhancement processing on the historical medical pathology images to obtain a historical medical pathology image dataset; Based on the historical medical pathology image dataset, a hybrid learning model based on the Kolmogorov-Arnold representation theorem is trained to obtain an image classification model, wherein the hybrid learning model includes a feature extraction component and a classification component, wherein the feature extraction component is used to output an image feature set based on the historical medical pathology image dataset, and the classification component is used to output a classification result based on the image feature set and a wavelet transform formula, wherein the wavelet transform formula is determined based on significant medical features of the historical pathology image dataset; Acquire a pathological image to be identified; Based on the pathological image to be identified and the image classification model, a classification result of the pathological image to be identified is determined.

2. The method according to claim 1, characterized in that Performing data enhancement processing on the historical medical pathology images to obtain a historical medical pathology image dataset, including: The historical medical pathology image is sheared and flipped to generate a modified historical medical pathology image, and the unchanged historical medical pathology image and the modified historical medical pathology image are formed into a historical medical pathology image data set.

3. The method according to claim 1, characterized in that The feature extraction component is used to output an image feature set based on the historical medical pathology image dataset, including: The feature extraction component is used to extract features of corresponding receptive fields based on historical medical pathology image datasets using different convolution kernels, and perform nonlinear transformation on the features to obtain first target features; Based on the first target feature, extracting a second target feature by pooling; Based on the first target feature and the second target feature, semantic feature information of historical medical pathology images is extracted through continuous iteration, and an image feature set is output.

4. The method according to claim 3, characterized in that The feature extraction component is used to extract the features of the corresponding receptive field based on the historical medical pathology image dataset using different convolution kernels, and perform nonlinear transformation on the features to obtain the first target feature, which is specifically implemented using the following formula: ; ； in, is the number of hidden layers, is the convolution kernel size, is the weight matrix of the feature extraction component, is the bias vector, is the activation function, is the number of channels, For historical medical pathology images, is the first target feature.

5. The method according to claim 4, characterized in that Based on the first target feature, the second target feature is extracted through the pooling layer, which is specifically implemented using the following formula: ; is the pooling function, is the second target feature.

6. The method according to claim 5, characterized in that Based on the first target feature and the second target feature, the semantic feature information of the historical medical pathology image is extracted through continuous iteration, and the image feature set is output, which is specifically implemented by the following formula: ; is the iteration symbol, is the image feature set.

7. The method according to claim 1, characterized in that The classification component is used to output a classification result based on the image feature set and the wavelet transform formula, including: The classification component is used to perform translation transformation and scaling transformation processing on the image feature set to obtain medically significant features; The wavelet transform formula is used on the medically significant features to obtain classification results.

8. The method according to claim 7, characterized in that The medically significant features are subjected to wavelet transformation to obtain classification results, including: The medically significant features are subjected to wavelet transformation using the following formula to obtain the classification result: ; in, is the Mexican function, which is the second-order derivative of the Gauss function. is the scale parameter, which is used to control the width of the wavelet. It is a medically significant feature.

9. The method according to claim 7, characterized in that The medically significant features are subjected to wavelet transformation to obtain classification results, including: The medically significant features are subjected to wavelet transformation using the following formula to obtain the classification result: ; in, is the Gaussian wavelet function, which is the first-order derivative of the Gauss function. It is a medically significant feature.

10. The method according to claim 1, characterized in that Determining a classification result of the pathological image to be identified based on the pathological image to be identified and the image classification model includes: The pathological image to be identified is input into the image classification model to obtain a classification result of the pathological image to be identified.

Images