Image comparison training method and apparatus based on registration, computer device

Abstract

The application relates to the technical field of image processing, and is applied to the medical field, in particular to an image comparison training method and device based on registration, equipment and a medium. The method comprises the following steps: acquiring a standard image and a to-be-processed image; performing registration on the to-be-processed image according to the standard image to obtain a registration image; performing feature extraction on the standard image and the registration image based on self-supervised learning to obtain corresponding standard image features and registration image features; comparing the standard image features and the registration image features to determine the mapping relationship between the standard image features and the registration image features; taking the parameters corresponding to the mapping relationship as fixed parameters of an image comparison model to obtain the image comparison model, thereby optimizing the image comparison technology, and subsequently analyzing and comparing a to-be-analyzed image through the image comparison model to obtain a high-accuracy image comparison result.

Description

Technical Field

[0001] This application relates to the field of image processing technology, and in particular to a registration-based image comparison training method, apparatus, computer equipment, and storage medium. Background Technology

[0002] In the medical field, image-assisted diagnosis is frequently required. Specifically, when an image to be analyzed exists, image comparison is used to obtain a standard image similar to the image to be analyzed. Then, the data information of the image to be analyzed is determined based on the data information of the standard image. In image comparison analysis, self-supervised methods are often used to compare the image to be analyzed with the standard image to identify the corresponding standard image, thereby determining the data information of the image to be analyzed. The self-supervised method not only has the advantage of accurately analyzing unlabeled large datasets but also has good adaptability. However, current self-supervised methods for image comparison analysis are often limited to the comparison of images of the same type. When multiple types of images exist, the image comparison effect is poor and the accuracy is low. Therefore, how to optimize image comparison technology has become an urgent problem to be solved. Summary of the Invention

[0003] The main objective of this application is to provide a registration-based image comparison training method, apparatus, computer equipment, and storage medium, aiming to optimize image comparison technology to obtain highly accurate image comparison results.

[0004] To achieve the aforementioned objectives, this application proposes a registration-based image comparison training method, the method comprising:

[0005] Obtain a training image set, which includes standard images and images to be processed;

[0006] The image to be processed is registered based on the standard image to obtain a registered image;

[0007] Based on self-supervised learning, feature extraction is performed on the standard image and the registered image to obtain the corresponding standard image features and registered image features;

[0008] The standard image features and the registered image features are compared to determine the mapping relationship between the standard image features and the registered image features;

[0009] The parameters corresponding to the mapping relationship are used as fixed parameters of the image comparison model to obtain the image comparison model.

[0010] Further, the step of registering the image to be processed based on the standard image to obtain a registered image includes:

[0011] Obtain the data information of the standard image;

[0012] Based on the data information and a preset registration algorithm, a registration operation is performed on the image to be processed to obtain a registered image.

[0013] Further, the step of extracting features from the standard image and the registered image based on self-supervised learning to obtain corresponding standard image features and registered image features includes:

[0014] The standard image and the registered image are respectively divided into multiple sub-standard images and multiple sub-registered images;

[0015] Based on self-supervised learning, the corresponding initial standard image features and initial registration image features are extracted by combining the multiple sub-standard images and the multiple sub-registration images;

[0016] The initial standard image features and the initial registration image features are adjusted according to preset parameters to obtain the corresponding standard image features and registration image features.

[0017] Further, the step of comparing the standard image features and the registered image features to determine the mapping relationship between the standard image features and the registered image features includes:

[0018] The standard image features and the registered image features are compared, and a first loss value between the standard image features and the registered image features is calculated using a preset first loss function;

[0019] Determine whether the first loss value is greater than a first preset threshold;

[0020] If so, it is determined that there is no mapping between the standard image features and the registered image features, and the corresponding first parameter is recorded.

[0021] Furthermore, after recording its corresponding first parameter, it also includes:

[0022] A second loss value between the standard image features and the registered image features is calculated using a preset second loss function;

[0023] Determine whether the second loss value is less than the second preset threshold;

[0024] If so, it is determined that there is a mapping between the standard image features and the registered image features, and the corresponding second parameter is recorded.

[0025] Further, the step of using the parameters corresponding to the mapping relationship as fixed parameters of the pathological image comparison model to obtain the pathological image comparison model includes:

[0026] The first parameter and the second parameter are used as fixed parameters of the image comparison model to obtain the image comparison model.

[0027] Furthermore, after obtaining the image comparison model, the process further includes:

[0028] Acquire the image to be analyzed;

[0029] Feature extraction is performed on the image to be analyzed to obtain the image features;

[0030] The features of the image to be analyzed are input into the image comparison model to obtain a standard image that is similar to the features of the image to be analyzed.

[0031] This application also provides a registration-based image comparison training device, the device comprising:

[0032] The acquisition module is used to acquire a training image set, which includes standard images and images to be processed;

[0033] The registration module is used to register the image to be processed based on the standard image to obtain a registered image;

[0034] The extraction module is used to extract features from the standard image and the registration image based on self-supervised learning, so as to obtain the corresponding standard image features and registration image features;

[0035] The comparison module is used to compare the standard image features and the registered image features to determine the mapping relationship between the standard image features and the registered image features;

[0036] The determining module is used to use the parameters corresponding to the mapping relationship as fixed parameters of the image comparison model to obtain the image comparison model.

[0037] This application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, characterized in that the processor executes the computer program to implement the steps of the registration-based image comparison training method described in any of the above claims.

[0038] This application also provides a computer-readable storage medium storing a computer program thereon, characterized in that, when the computer program is executed by a processor, it implements the steps of the registration-based image comparison training method described in any of the above claims.

[0039] This application provides a registration-based image comparison training method. It involves acquiring a training image set, which includes a standard image and an image to be processed. The image to be processed is registered with the standard image to obtain a registered image. Features are extracted from the standard image and the registered image based on self-supervised learning to obtain corresponding standard image features and registered image features. The standard image features and the registered image features are compared to determine the mapping relationship between them. The parameters corresponding to this mapping relationship are used as fixed parameters of the image comparison model to obtain the image comparison model. This optimizes the image comparison technology, and the image to be analyzed can then be compared using the image comparison model to obtain highly accurate image comparison results. Attached Figure Description

[0040] Figure 1 This is a schematic flowchart of an embodiment of the image alignment training method based on registration in this application;

[0041] Figure 2 This is a schematic flowchart of another embodiment of the image alignment training method based on registration in this application;

[0042] Figure 3 This is a schematic flowchart of another embodiment of the image alignment training method based on registration in this application;

[0043] Figure 4 This is a schematic flowchart of another embodiment of the image alignment training method based on registration in this application;

[0044] Figure 5 This is a schematic flowchart of another embodiment of the image alignment training method based on registration in this application;

[0045] Figure 6 This is a schematic flowchart of another embodiment of the image alignment training method based on registration in this application;

[0046] Figure 7 This is a schematic diagram of an embodiment of the image comparison training device based on registration according to this application;

[0047] Figure 8 This is a schematic block diagram of an embodiment of the computer device of this application.

[0048] The realization of the purpose, functional features and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0049] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0050] Reference Figure 1 This application provides a registration-based image comparison training method, including steps S10-S50. The detailed description of each step of the registration-based image comparison training method is as follows.

[0051] S10. Obtain a training image set, which includes standard images and images to be processed.

[0052] This embodiment applies to image comparison in the medical field, specifically to the comparison training of pathological images. The training image set is preferably a set of pathological images. Pathological images refer to images converted into digital pathological images by pathologists through sampling, staining, and scanning of patient tissues. These pathological images exist in various forms depending on the staining method, specifically including H&E staining and immunohistochemical (IHC) staining, such as ER, PR, HER2, Ki67, etc. Pathological images can be obtained from publicly available and legally accessible pathological images on third-party websites, or from pathological images uploaded by pathologists. A training image set is constructed based on these acquired pathological images. H&E-stained pathological images are used as standard images, and immunohistochemical (IHC)-stained pathological images are used as images to be processed. The acquisition of these training images provides training samples for subsequent image comparison training.

[0053] S20. Register the image to be processed according to the standard image to obtain a registered image.

[0054] In this embodiment, after acquiring the training image set, the image to be processed is registered based on the standard image to obtain a registered image. Specifically, the data information of the standard image is acquired, and the image to be processed is registered based on the data information and a preset registration algorithm to obtain a registered image. Registering the image to be processed corrects the image and provides a valid basis for subsequently acquiring accurate image features.

[0055] S30. Based on self-supervised learning, feature extraction is performed on the standard image and the registration image to obtain the corresponding standard image features and registration image features.

[0056] In this embodiment, after registering the image to be processed based on the standard image to obtain a registered image, feature extraction is performed on the standard image and the registered image based on self-supervised learning to obtain corresponding standard image features and registered image features. Specifically, the standard image and the registered image are respectively divided into multiple sub-standard images and multiple sub-registered images. Image segmentation facilitates comprehensive extraction of image features in the subsequent process, thereby avoiding the problem of incomplete feature extraction. Then, based on self-supervised learning, the corresponding initial standard image features and initial registered image features are extracted from the multiple sub-standard images and multiple sub-registered images. The initial standard image features and initial registered image features are adjusted according to preset parameters to obtain the corresponding standard image features and registered image features. The acquisition of the standard image features and registered image features provides a valid basis for subsequent image feature comparison.

[0057] S40. Compare the standard image features and the registered image features to determine the mapping relationship between the standard image features and the registered image features.

[0058] In this embodiment, after extracting features from the standard image and the registered image based on self-supervised learning to obtain corresponding standard image features and registered image features, the standard image features and the registered image features are compared to determine the mapping relationship between the standard image features and the registered image features. The mapping relationship includes two cases: the existence of a mapping relationship and the absence of a mapping relationship. Specifically, the standard image features and the registered image features are compared, and a first loss value between the standard image features and the registered image features is calculated using a preset first loss function. It is then determined whether the first loss value is greater than a first preset threshold. If so, it is determined that there is no mapping between the standard image features and the registered image features, and the corresponding first parameter is recorded. The first loss function is used to calculate the dissimilarity between the standard image features and the registered image features, and the obtained first parameter is used to filter out standard images with low relevance to the image to be analyzed. After obtaining the first parameter, a second loss value between the standard image features and the registered image features is calculated using a preset second loss function, and it is determined whether the second loss value is less than a second preset threshold. If so, it is determined that there is a mapping between the standard image features and the registered image features, and the corresponding second parameter is recorded. The second loss function is used to calculate the similarity between the standard image features and the registered image features, and the obtained second parameter is used to quickly obtain standard images with high relevance to the image to be analyzed, thereby accelerating the analysis of the image to be analyzed and the determination of relevant data. By comparing the standard image features and the registered image features, the mapping relationship between the standard image features and the registered image features and their corresponding first and second parameters are determined, thus determining the fixed parameters of the image comparison model and providing a valid basis for obtaining the optimal image comparison model.

[0059] S50. The parameters corresponding to the mapping relationship are used as fixed parameters of the image comparison model to obtain the image comparison model.

[0060] In this embodiment, after comparing the standard image features and the registered image features to determine the mapping relationship between them, the parameters corresponding to the mapping relationship are used as fixed parameters of the image comparison model to obtain the image comparison model. Specifically, the first parameter and the second parameter are used as fixed parameters of the image comparison model to obtain the image comparison model. The first parameter serves as the model's filtering rule, filtering out standard images with low similarity to the image to be analyzed; the second parameter serves as the model's acquisition filtering rule, screening out standard images with high similarity to the image to be analyzed, thus accelerating the determination of the data corresponding to the image to be analyzed when the user performs image analysis. By using the parameters corresponding to the mapping relationship as fixed parameters of the image comparison model, the image comparison model is generated.

[0061] This embodiment provides a registration-based image comparison training method. It involves acquiring a training image set, which includes a standard image and an image to be processed. The image to be processed is registered with the standard image to obtain a registered image. Based on self-supervised learning, features are extracted from the standard image and the registered image to obtain corresponding standard image features and registered image features. The standard image features and the registered image features are compared to determine the mapping relationship between them. The parameters corresponding to this mapping relationship are used as fixed parameters of the image comparison model to obtain the image comparison model. This optimizes the image comparison technology, and the image to be analyzed can then be compared using the image comparison model to obtain highly accurate image comparison results.

[0062] Reference Figure 2 In one embodiment, registering the image to be processed based on the standard image to obtain a registered image includes the following steps S21-S22:

[0063] S21. Obtain the data information of the standard image;

[0064] S22. Based on the data information and a preset registration algorithm, perform a registration operation on the image to be processed to obtain a registered image.

[0065] In this embodiment, registration refers to the matching of geographic coordinates of different images obtained using different imaging methods within the same area. This includes three aspects: geometric correction, projection transformation, and scaling. The process involves loading data from a standard image to obtain its data information. Based on this data information and a pre-defined registration algorithm, a registration operation is performed on the image to be processed to obtain a registered image. The registration algorithm can be any one of linear registration, quadratic polynomial registration, or rectangular registration. If the registration algorithm is linear registration, control points are pre-selected from the data information. These control points are points with latitude and longitude coordinates, and there should be at least four control points. The image to be processed is then corrected based on these control points to obtain the registered image. If the registration algorithm is quadratic polynomial registration, control points are selected from the data information, and there should be at least seven control points. The image to be processed is then corrected based on these control points to obtain the registered image. If the registration algorithm is rectangular registration, control points are selected from the data information, and there should be at least two control points. The image to be processed is then corrected based on these control points to obtain the registered image. Registering the image to be processed with the standard image corrects the image and helps reduce training errors during subsequent image comparisons.

[0066] Reference Figure 3 In one embodiment, the step of extracting features from the standard image and the registered image based on self-supervised learning to obtain corresponding standard image features and registered image features includes the following steps S31-S33:

[0067] S31. Divide the standard image and the registration image into multiple sub-standard images and multiple sub-registration images respectively;

[0068] S32. Based on self-supervised learning, extract the corresponding initial standard image features and initial registration image features by combining the multiple sub-standard images and the multiple sub-registration images;

[0069] S33. Adjust the initial standard image features and the initial registration image features according to preset parameters to obtain the corresponding standard image features and registration image features.

[0070] In this embodiment, self-supervised learning refers to extracting a large amount of information about the image itself from unsupervised data using an auxiliary task. Using 256*256 pixels as the segmentation standard, the standard image and the registration image are divided into multiple sub-standard images and multiple sub-registration images. Initial standard image features are extracted from these sub-standard images based on self-supervised learning, and initial registration image features are extracted from these sub-registration images based on self-supervised learning. Then, the initial standard image features and the initial registration image features are adjusted sequentially using preset parameters, while invalid features are filtered out, thereby obtaining the corresponding standard image features and registration image features. The preset parameters are parameters obtained in advance based on the self-supervised model to correct and adjust the initial features. Extracting the standard image features and registration image features provides a valid basis for subsequent feature comparison.

[0071] refer to Figure 4 In one embodiment, comparing the standard image features and the registered image features to determine the mapping relationship between the standard image features and the registered image features includes the following steps S41-S43:

[0072] S41. Compare the standard image features and the registered image features, and calculate the first loss value between the standard image features and the registered image features using a preset first loss function;

[0073] S42. Determine whether the first loss value is greater than the first preset threshold;

[0074] S43. If so, determine that there is no mapping between the standard image features and the registered image features, and record the corresponding first parameter.

[0075] In this embodiment, the standard image features and the registered image features are compared, and a first loss value between the standard image features and the registered image features is calculated using a preset first loss function. The first loss function is used to calculate the dissimilarity information between the standard image features and the registered image features; the more dissimilar the standard image features and the registered image features are, the greater their loss value. The first loss value is compared with a first preset threshold to determine whether the first loss value is greater than the first preset threshold. If so, it is determined that there is no mapping between the standard image features and the registered image features, and the corresponding first parameter is recorded. Obtaining the first parameter provides a valid basis for subsequently determining the fixed parameters of the image comparison model.

[0076] Reference Figure 5 In one embodiment, after recording the corresponding first parameter, the method further includes the following steps S431-S433:

[0077] S431. Calculate the second loss value between the standard image features and the registered image features using a preset second loss function;

[0078] S432. Determine whether the second loss value is less than the second preset threshold;

[0079] S433. If so, it is determined that there is a mapping between the standard image features and the registered image features, and the corresponding second parameter is recorded.

[0080] In this embodiment, after obtaining the first parameter, a second loss value between the standard image feature and the registered image feature is calculated using a preset second loss function. The second loss function is used to calculate the similarity information between the standard image feature and the registered image feature. The more similar the standard image feature and the registered image feature are, the smaller the loss value is. The second loss value is compared with a second preset threshold to determine whether the second loss value is less than the second preset threshold. If so, it is determined that there is a mapping between the standard image feature and the registered image feature, and the corresponding second parameter is recorded. Obtaining the second parameter provides a valid basis for subsequently determining the fixed parameters of the image comparison model.

[0081] In one embodiment, the step of using the parameters corresponding to the mapping relationship as fixed parameters of the image comparison model to obtain the image comparison model includes:

[0082] The first parameter and the second parameter are used as fixed parameters of the image comparison model to obtain the image comparison model.

[0083] In this embodiment, the first parameter and the second parameter are used as fixed parameters of the image comparison model to obtain the image comparison model. The first parameter and the second parameter are equivalent to the model's screening rules. For example, when an image to be analyzed exists, standard images with low similarity to the image to be analyzed are screened out based on the image features of the image to be analyzed using the first parameter. This reduces the subsequent computational burden while improving the model's ability to acquire similar standard images. Then, the second parameter is used to screen out standard images with high similarity to the image to be analyzed. By using the first parameter and the second parameter as fixed parameters of the image comparison model, the image comparison model is generated.

[0084] Reference Figure 6 In one embodiment, after obtaining the image comparison model, the method further includes the following steps S51-S53:

[0085] S51. Obtain the image to be analyzed;

[0086] S52. Extract features from the image to be analyzed to obtain the features of the image to be analyzed;

[0087] S53. Input the features of the image to be analyzed into the image comparison model to obtain a standard image similar to the features of the image to be analyzed.

[0088] In this embodiment, after obtaining the image comparison model, it can be used as an auxiliary means for pathological analysis. Specifically, when there is a pathological image to be analyzed, the pathological image to be analyzed is preprocessed to obtain its features. The features are then input into the image comparison model, which analyzes and compares them to obtain a standard pathological image similar to the features of the pathological image to be analyzed. The standard pathological image is then output, allowing the user to analyze the pathological image to be analyzed based on the data information of the standard pathological image, thereby quickly determining the condition corresponding to the pathological image and gaining more treatment time for critically ill patients.

[0089] Reference Figure 7 This application provides a registration-based image comparison training device, the device comprising:

[0090] The acquisition module 10 is used to acquire a training image set, which includes standard images and images to be processed;

[0091] Registration module 20 is used to register the image to be processed according to the standard image to obtain a registered image;

[0092] Extraction module 30 is used to extract features from the standard image and the registration image based on self-supervised learning to obtain corresponding standard image features and registration image features;

[0093] The comparison module 40 is used to compare the standard image features and the registered image features to determine the mapping relationship between the standard image features and the registered image features;

[0094] The determining module 50 is used to use the parameters corresponding to the mapping relationship as fixed parameters of the image comparison model to obtain the image comparison model.

[0095] As described above, it is understood that each component of the registration-based image comparison training apparatus proposed in this application can realize the function of any of the registration-based image comparison training methods described above.

[0096] In one embodiment, the image matting module 20 is further configured to perform:

[0097] Obtain the data information of the standard image;

[0098] Based on the data information and a preset registration algorithm, a registration operation is performed on the image to be processed to obtain a registered image.

[0099] In one embodiment, the extraction module 30 is further configured to perform:

[0100] The standard image and the registered image are respectively divided into multiple sub-standard images and multiple sub-registered images;

[0101] Based on self-supervised learning, the corresponding initial standard image features and initial registration image features are extracted by combining the multiple sub-standard images and the multiple sub-registration images;

[0102] The initial standard image features and the initial registration image features are adjusted according to preset parameters to obtain the corresponding standard image features and registration image features.

[0103] In one embodiment, the determination module 40 is further configured to perform:

[0104] The standard image features and the registered image features are compared, and a first loss value between the standard image features and the registered image features is calculated using a preset first loss function;

[0105] Determine whether the first loss value is greater than a first preset threshold;

[0106] If so, it is determined that there is no mapping between the standard image features and the registered image features, and the corresponding first parameter is recorded.

[0107] In one embodiment, the determination module 40 is further configured to perform:

[0108] A second loss value between the standard image features and the registered image features is calculated using a preset second loss function;

[0109] Determine whether the second loss value is less than the second preset threshold;

[0110] If so, it is determined that there is a mapping between the standard image features and the registered image features, and the corresponding second parameter is recorded.

[0111] In one embodiment, the determination module 40 is further configured to perform:

[0112] The first parameter and the second parameter are used as fixed parameters of the image comparison model to obtain the image comparison model.

[0113] In one embodiment, the alarm module 50 is further configured to perform:

[0114] Acquire the image to be analyzed;

[0115] Feature extraction is performed on the image to be analyzed to obtain the image features;

[0116] The features of the image to be analyzed are input into the image comparison model to obtain a standard image that is similar to the features of the image to be analyzed.

[0117] Reference Figure 8 This application embodiment also provides a computer device, the internal structure of which can be as follows: Figure 8 As shown, the computer device includes a processor, memory, network interface, display device, and input device connected via a system bus. The network interface is used for communication with external terminals via a network connection. The display device is used to display interactive pages. The input device is used to receive user input. The processor is designed to provide computing and control capabilities. The memory includes non-volatile storage media. This non-volatile storage media stores the operating system, computer programs, and a database. The database stores raw data. When the computer program is executed by the processor, it implements a registration-based image comparison training method.

[0118] The processor described above executes the registration-based image comparison training method, comprising: acquiring a training image set, the training image set including a standard image and an image to be processed; registering the image to be processed according to the standard image to obtain a registered image; extracting features from the standard image and the registered image based on self-supervised learning to obtain corresponding standard image features and registered image features; comparing the standard image features and the registered image features to determine the mapping relationship between the standard image features and the registered image features; and using the parameters corresponding to the mapping relationship as fixed parameters of the image comparison model to obtain the image comparison model. The computer provides a registration-based image comparison training method. It acquires a training image set, which includes a standard image and an image to be processed. The image to be processed is registered based on the standard image to obtain a registered image. Features are extracted from the standard image and the registered image based on self-supervised learning to obtain corresponding standard image features and registered image features. The standard image features and the registered image features are compared to determine the mapping relationship between them. The parameters corresponding to the mapping relationship are used as fixed parameters of the image comparison model to obtain the image comparison model. This optimizes the image comparison technology. Subsequently, the image to be analyzed and compared using the image comparison model can yield highly accurate image comparison results.

[0119] This application also provides a computer-readable storage medium storing a computer program thereon. When executed by the processor, the computer program implements a registration-based image comparison training method, comprising the steps of: acquiring a training image set, the training image set including a standard image and an image to be processed; registering the image to be processed according to the standard image to obtain a registered image; extracting features from the standard image and the registered image based on self-supervised learning to obtain corresponding standard image features and registered image features; comparing the standard image features and the registered image features to determine the mapping relationship between the standard image features and the registered image features; and using the parameters corresponding to the mapping relationship as fixed parameters of the image comparison model to obtain the image comparison model. The computer-readable storage medium provides a registration-based image comparison training method. This method involves acquiring a training image set, which includes a standard image and an image to be processed. The image to be processed is registered based on the standard image to obtain a registered image. Features are extracted from the standard image and the registered image using self-supervised learning to obtain corresponding standard image features and registered image features. The standard image features and the registered image features are compared to determine the mapping relationship between them. The parameters corresponding to this mapping relationship are used as fixed parameters of the image comparison model to obtain the image comparison model. This optimizes the image comparison technology, and the image to be analyzed can then be compared using the image comparison model to obtain highly accurate image comparison results.

[0120] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media provided in this application and in the embodiments may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual-speed SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0121] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, apparatus, article, or method. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, apparatus, article, or method that includes that element.

[0122] The above description is only a preferred embodiment of this application and does not limit the patent scope of this application. Any equivalent structural or procedural changes made based on the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

Claims

1. A registration-based image co-registration training method, characterized by, The method comprises: acquiring a training image set, the training image set comprising standard images and to-be-processed images, the training image set being a pathological image set; registering the to-be-processed images according to the standard images to obtain registered images; extracting features of the standard images and the registered images based on self-supervised learning to obtain corresponding standard image features and registered image features; comparing the standard image features and the registered image features to determine a mapping relationship between the standard image features and the registered image features; using parameters corresponding to the mapping relationship as fixed parameters of an image comparison model to obtain the image comparison model; wherein the feature extraction of the standard images and the registered images based on self-supervised learning to obtain corresponding standard image features and registered image features comprises: cutting the standard images and the registered images into multiple sub-standard images and multiple sub-registered images respectively; extracting corresponding initial standard image features and initial registered image features based on self-supervised learning in combination with the multiple sub-standard images and the multiple sub-registered images; adjusting the initial standard image features and the initial registered image features according to preset parameters to obtain corresponding standard image features and registered image features; wherein after obtaining the image comparison model, the method further comprises: acquiring a to-be-analyzed image; extracting features of the to-be-analyzed image to obtain to-be-analyzed image features; inputting the to-be-analyzed image features into the image comparison model to obtain a standard image similar to the to-be-analyzed image features; wherein when there is a to-be-analyzed pathological image, the to-be-analyzed pathological image is pre-processed to obtain to-be-analyzed pathological image features, the to-be-analyzed pathological image features are input into the image comparison model, and the to-be-analyzed pathological image features are analyzed and compared based on the image comparison model to obtain a standard pathological image similar to the to-be-analyzed pathological image features, and the standard pathological image is output.

2. The registration-based image comparison training method of claim 1, wherein, The registration of the to-be-processed images according to the standard images to obtain registered images comprises: acquiring data information of the standard images; performing a registration operation on the to-be-processed images according to the data information in combination with a preset registration algorithm to obtain registered images.

3. The registration-based image comparison training method of claim 1, wherein, The comparison of the standard image features and the registered image features to determine a mapping relationship between the standard image features and the registered image features comprises: comparing the standard image features and the registered image features, and calculating a first loss value between the standard image features and the registered image features through a preset first loss function; determining whether the first loss value is greater than a first preset threshold value; if yes, it is determined that there is no mapping between the standard image features and the registered image features, and a corresponding first parameter is recorded.

4. The registration-based image comparison training method of claim 3, wherein, After recording the corresponding first parameter, the method further comprises: calculating a second loss value between the standard image features and the registered image features through a preset second loss function; determining whether the second loss value is less than a second preset threshold value; If yes, it is determined that there is a mapping between the standard image feature and the registration image feature, and a corresponding second parameter is recorded.

5. The registration-based image comparison training method of claim 4, wherein, The parameter corresponding to the mapping relationship is taken as a fixed parameter of an image comparison model, and the image comparison model is obtained. The first parameter and the second parameter are taken as fixed parameters of an image comparison model, and the image comparison model is obtained.

6. A registration-based image co-registration training apparatus, characterized by, The device comprises: The acquisition module is configured to acquire a training image set, the training image set comprising a standard image and a to-be-processed image, and the training image set being a pathological image set. The registration module is configured to register the to-be-processed image based on the standard image to obtain a registration image. The extraction module is configured to perform feature extraction on the standard image and the registration image based on self-supervised learning to obtain corresponding standard image features and registration image features. The comparison module is configured to compare the standard image features and the registration image features to determine a mapping relationship between the standard image features and the registration image features. The determination module is configured to take a parameter corresponding to the mapping relationship as a fixed parameter of an image comparison model to obtain the image comparison model. The feature extraction on the standard image and the registration image based on self-supervised learning to obtain corresponding standard image features and registration image features comprises: The standard image and the registration image are respectively divided into a plurality of sub-standard images and a plurality of sub-registration images. Initial standard image features and initial registration image features are extracted based on self-supervised learning in combination with the plurality of sub-standard images and the plurality of sub-registration images. The initial standard image features and the initial registration image features are adjusted according to a preset parameter to obtain corresponding standard image features and registration image features. After the image comparison model is obtained, the method further comprises: Acquiring a to-be-analyzed image; Performing feature extraction on the to-be-analyzed image to obtain to-be-analyzed image features; Inputting the to-be-analyzed image features into the image comparison model to obtain a standard image similar to the to-be-analyzed image features. When there is a to-be-analyzed pathological image, the to-be-analyzed pathological image is pre-processed to obtain to-be-analyzed pathological image features, the to-be-analyzed pathological image features are input into the image comparison model, the to-be-analyzed pathological image features are analyzed and compared based on the image comparison model, a standard pathological image similar to the to-be-analyzed pathological image features is obtained, and the standard pathological image is output. 7.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is configured to perform the method according to any one of claims 1-6 when the computer program is executed by the processor. The processor executes the computer program to implement the steps of the registration-based image comparison training method according to any one of claims 1 to 5.

8. A computer readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the registration-based image comparison training method according to any one of claims 1 to 5.

Images