A database-based clinical tissue immunofluorescence image classification method

Abstract

The present invention discloses a clinical tissue immunofluorescence image classification method based on a database, including a database initial construction unit, a unified analysis unit, an independent analysis unit, a similarity comparison unit, and an upload distinction unit. The database initial construction unit presets multiple groups of types, each group of type labels corresponds to a total group, and the type information includes but is not limited to cell type, tissue structure, and disease state. Each group of types includes multiple type labels, and each tissue immunofluorescence image corresponds to multiple different groups of type labels, and a standard initial library composed of multiple standard tissue immunofluorescence images is established. The present invention can perform differential analysis according to the source of the supplemented tissue immunofluorescence images during the database supplement process, so that the database can make more efficient judgments on the standards of the supplemented tissue immunofluorescence images during the expansion process.

Description

Technical Field

[0001] The present invention relates to, in particular, a clinical tissue immunofluorescence image classification method based on a database. Background Art

[0002] In clinical diagnosis, tissue immunofluorescence images are obtained through immunofluorescence technology, which combines immunology and fluorescence microscopy to detect specific antigens or antibodies in tissue samples. In the image, the location and distribution of specific antigens or antibodies in cells or tissues are clearly displayed by using fluorescently labeled antibodies or antigen probes, making the location and properties of the antigens or antibodies clear at a glance. When interpreting tissue immunofluorescence images, attention should be paid to the type of fluorescent markers, distribution pattern, and intensity of the fluorescent signal. This information not only helps to determine the type, stage, and progression of the disease, but also provides a strong basis for the formulation of treatment plans.

[0003] When classifying clinical tissue immunofluorescence images, an image database consisting of different types of images is generally set up based on clinical diagnosis and expert knowledge, and then an algorithm model is used based on the image database to intelligently classify the tissue immunofluorescence images encountered in the clinical diagnosis process to assist in the confirmation of clinical diagnosis results. However, this classification method is limited by the data set size of the database. In the prior art, in the process of database supplementation, the supplemented images are generally compared with the images in the existing database for similarity, and then the images with similarity greater than a certain threshold are included to expand the database. However, this method cannot identify the source of the image, and it is difficult to select different comparison and judgment methods according to different sources, which makes it difficult for the expansion of the database to break through the limitations of the original database, affecting the effect of database construction, and further affecting the efficiency of classification. Summary of the invention

[0004] The purpose of the present invention is to solve the problem that it is difficult to select different comparison and judgment methods according to different sources in the image classification database, which makes it difficult to expand the database and break through the limitations of the original database, and to propose a clinical tissue immunofluorescence image classification method based on the database.

[0005] The purpose of the present invention can be achieved by the following technical solution: A clinical tissue immunofluorescence image classification method based on a database, including a database initial construction unit, a unified analysis unit, an independent analysis unit, a similarity comparison unit, and an upload distinction unit, and the specific steps are as follows:

[0006] Step 1: The database initial construction unit presets multiple groups of types. Each group of type tags corresponds to a total group. Each group of types includes multiple type tags. Each tissue immunofluorescence image corresponds to multiple type tags of different groups. A standard initial library is established. The tissue immunofluorescence images are classified according to the type groups. The user uploads tissue immunofluorescence images into the standard initial library. The standard initial library and the received tissue immunofluorescence images together form the existing standard database;

[0007] Step 2: Denote the standard tissue immunofluorescence images in the standard initial library as standard images. The upload differentiation unit is used to obtain the upload source and uploader information of the uploaded tissue immunofluorescence images. The upload sources are divided into actively supplemented tissue immunofluorescence images and existing extracted tissue immunofluorescence images. The similarity comparison unit comprehensively calculates the similarity between the uploaded tissue immunofluorescence image and each existing tissue immunofluorescence image through multiple image similarity algorithms in the prior art and takes the maximum value as P. Set an approximate threshold A and a similarity threshold B, and A and B satisfy 0.8 < A < B < 1;

[0008] Step 3: When 0 < P ≤ A, the independent analysis unit analyzes the uploaded tissue immunofluorescence image differently according to whether the upload source is actively supplemented or existing extracted. When A < P ≤ B, the unified analysis unit marks the uploaded tissue immunofluorescence image as a similar image, conducts a type analysis on the similar image, and determines whether the uploaded tissue immunofluorescence image is a non-compliant standard image or a supplementary image. When B < P ≤ 1, the unified analysis unit marks the uploaded tissue immunofluorescence image as an existing image, rejects the reception, and generates an image already exists signal to send to the uploading user;

[0009] Step 4: Reject the reception of non-compliant standard images and generate an image not up to standard signal to send to the uploading user. The existing standard database receives supplementary images, and the standard processing unit conducts standardized preprocessing on the supplementary images.

[0010] As a preferred implementation manner of the present invention, the similarity calculation process between the uploaded tissue immunofluorescence image and the existing tissue immunofluorescence image is as follows:

[0011] Calculate the color similarity value between the two images through the color histogram comparison algorithm and denote it as Calculate the texture similarity value between the two images through the texture histogram comparison algorithm and denote it as Extract all standard images in the current standard database with the same type information as the uploaded tissue immunofluorescence image to form an algorithm learning data set. Train a convolutional neural network model through the algorithm learning data set and calculate the feature similarity value between the two images through the convolutional neural network model Substitute it into the formula for calculation to obtain its similarity , where is a preset weight factor.

[0012] As a preferred embodiment of the present invention, when the upload source is an actively supplemented tissue immunofluorescence image and 0 < P ≤ A, the independent analysis unit marks the uploaded tissue immunofluorescence image as a new image, obtains the type information, image information, and uploader information of the new image, analyzes and calculates the standard reference value of the new image. When the standard reference value is greater than or equal to the preset standard threshold, the new image is marked as a supplementary image; when the standard reference value is less than the preset standard threshold, the new image is marked as an image not meeting the standard.

[0013] As a preferred embodiment of the present invention, the calculation process of the standard reference value of the new image is as follows:

[0014] Obtain the fluorescence labeling color, fluorescence intensity, fluorescence distribution, and cell morphology of the new image, extract the standard tissue immunofluorescence image with the same type information in the standard database as the comparison image, calculate the chromaticity difference between the fluorescence labeling colors of the new image and the comparison image and divide it by the chromaticity value of the fluorescence labeling color in the comparison image as the chromaticity difference ratio , calculate the fluorescence brightness difference between the fluorescence labels of the new image and the comparison image and divide it by the brightness value of the fluorescence label in the comparison image as the brightness difference ratio , calculate the difference in the fluorescence area occupancy ratio between the new image and the comparison image as the area difference , count the number and total occupied area of the fluorescence-labeled cells in the new image and the comparison image, calculate the total occupied area divided by the number of fluorescence-labeled cells to obtain the cell morphology size value, calculate the ratio of the cell morphology size values of the fluorescence-labeled cells in the new image and the comparison image as the morphology difference , and substitute the chromaticity difference ratio , brightness difference ratio , area difference , and morphology difference into the formula for calculation to obtain the image judgment value ;

[0015] Obtain the uploader information, which includes the number of uploaders N for obtaining the new image and the number of standard tissue immunofluorescence images uploaded by each uploader , the number of times of uploading images , substitute into the formula for calculation to obtain the parameter influence value corresponding to each uploader , where n represents the label of each uploader, and is a preset weight factor, and substitute the parameter influence value Substitute the number of uploaders N into the preset model for calculation to obtain the uploader judgment value corresponding to each newly added picture , where is a preset weight factor. Normalize the image judgment value and the uploader judgment value , and then substitute them into the formula for calculation to obtain the standard reference value of the newly added picture .

[0016] As a preferred embodiment of the present invention, when the upload source is an existing tissue immunofluorescence picture that has been extracted, and 0 < P ≤ A, the independent analysis unit obtains the source location data and the specific source data of the uploaded tissue immunofluorescence picture. There are multiple preset standard source locations. When the source location data is the same as any one of the multiple standard source locations, the retrieval system retrieves the relevant files to which the tissue immunofluorescence picture belongs according to the specific source data. When there is a picture in the retrieved files whose similarity to the uploaded tissue immunofluorescence picture is greater than the similarity threshold B, the uploaded tissue immunofluorescence picture is marked as a supplementary picture; otherwise, the newly added picture is marked as a picture that does not meet the standard.

[0017] As a preferred embodiment of the present invention, when A < P ≤ B, obtain the type information of the similar picture and the standard picture with a similarity value of P corresponding to it, calculate the type similarity value of the two pictures. The type similarity value is equal to the number of the same type information divided by the total number of type information categories. When the type similarity value is greater than the preset type similarity threshold, it is marked as a supplementary picture and a picture type error signal is sent to the uploading user. When the type similarity value is less than or equal to the preset type similarity threshold, the newly added picture is marked as a picture that does not meet the standard.

[0018] As a preferred embodiment of the present invention, the standard processing unit's standardization process is as follows:

[0019] Set a standard value set, which contains multiple different types of standard intervals. The types of the standard intervals include size, data size, format, resolution, etc. Perform standardization processing on the supplementary pictures so that the corresponding values of each type are within the standard intervals to obtain standard supplementary pictures. Perform quality preprocessing on the standard supplementary pictures, including denoising, enhancing contrast, color correction, etc., to improve the image quality, and supplement the processed standard supplementary pictures into the standard database.

[0020] Compared with the prior art, the beneficial effects of the present invention are:

[0021] 1. The present invention can perform differential analysis according to the source of the supplemented tissue immunofluorescence images during the database supplement process, so that the comparison conditions for judging whether the tissue immunofluorescence images are standard are more three-dimensional. By distinguishing the upload sources and selecting different comparison methods for judgment and decision-making according to the different upload sources, the database can more efficiently judge whether the supplemented tissue immunofluorescence images are standard during the expansion process.

[0022] 2. The present invention can obtain standard tissue immunofluorescence images of corresponding types and fields through screening, and further refine comparison images for targeted comparison and judgment. Compared with the basic algorithm similarity comparison, the comparison applied to the subdivided field focuses more on the judgment of fluorescent marker color, fluorescence intensity, fluorescence distribution, and cell morphology rather than the comparison with the similarity of existing images. The effect of this decision is that it can further judge those images that do not meet the similarity requirements with the existing ones, and extract images that can be supplemented into the standard database, thereby supplementing the database and overcoming the defect of determining whether the image is standard or not only by similarity judgment. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to facilitate understanding by those skilled in the art, the present invention is further described below with reference to the accompanying drawings.

[0024] Figure 1 It is a principle block diagram of the present invention. DETAILED DESCRIPTION

[0025] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0026] See also Figure 1 As shown, a clinical tissue immunofluorescence image classification method based on a database includes a database initialization unit, an upload distinction unit, a similarity comparison unit, a unified analysis unit, an independent analysis unit, and a standard processing unit.

[0027] Step 1: The database initialization unit is preset with multiple groups of types, each group of type labels corresponds to a general group, and the type information includes but is not limited to cell type, tissue structure, and disease status. Each group of types includes multiple type labels, and each tissue immunofluorescence image corresponds to multiple different groups of type labels. A standard initial library consisting of multiple standard tissue immunofluorescence images is established. The number of standard tissue immunofluorescence images in the standard initial library is greater than or equal to the preset inventory threshold, and each tissue immunofluorescence image in the standard initial library is classified according to the type group.

[0028] Step 2: The upload and differentiation unit obtains the tissue immunofluorescence images uploaded by the user. When uploading the tissue immunofluorescence images, the user inputs the upload source of the tissue immunofluorescence images and the source-related information. There are multiple preset upload sources for the user to choose from, including but not limited to personal supplementation, joint supplementation, textbook extraction, example extraction, and other source methods. The uploaded sources are divided into tissue immunofluorescence images supplemented actively and tissue immunofluorescence images extracted from existing ones;

[0029] It should be noted that active supplementation means that the tissue immunofluorescence images submitted by the user are newly obtained and not included in the existing database, textbooks, magazines, and books, while existing extraction means that the tissue immunofluorescence images submitted by the user are intercepted and extracted from existing known public documents.

[0030] The similarity comparison unit compares the uploaded tissue immunofluorescence images with the tissue immunofluorescence images in the existing standard database (the existing standard database) one by one, and comprehensively calculates the similarity (percentage) between the uploaded tissue immunofluorescence image and each existing tissue immunofluorescence image through various image similarity algorithms in the prior art, and takes the maximum value as P. An approximate threshold A and a similarity threshold B are set, and A and B satisfy 0.8 < A < B < 1;

[0031] Step 3: When the upload source is an actively supplemented tissue immunofluorescence image and 0 < P ≤ A, the independent analysis unit marks the uploaded tissue immunofluorescence image as a new image, obtains the type information (group and label), image information, and uploader information of the image (one image may be uploaded by a single user or jointly uploaded by multiple users), analyzes and calculates the standard reference value of the new image. When the standard reference value is greater than or equal to the preset standard threshold, the new image is marked as a supplementary image; when the standard reference value is less than the preset standard threshold, the new image is marked as an image that does not meet the standard;

[0032] Specifically, the calculation process of the standard reference value of the new image is as follows:

[0033] Obtain the image information of the new image. The image information includes fluorescence marker color, fluorescence intensity, fluorescence distribution, and cell morphology. Extract the standard tissue immunofluorescence images with the same corresponding type information (same disease state, same tissue structure, same cell type...) in the standard database as the comparison images. Calculate the chromaticity difference between the fluorescence marker colors of the new image and the comparison images and divide it by the chromaticity value of the fluorescence marker color in the comparison images as the chromaticity difference ratio , calculate the fluorescence brightness difference between the fluorescence markers of the new image and the comparison images and divide it by the brightness value of the fluorescence marker in the comparison images as the brightness difference ratio , calculate the difference between the fluorescence area ratio in the new image and the comparison image and record it as the area difference , count the number of fluorescently labeled cells and the total area in the new pictures and the comparison pictures, calculate the total area divided by the number of fluorescently labeled cells to get the cell morphology size value, calculate the ratio of the cell morphology size values ​​of fluorescently labeled cells in the new pictures and the comparison pictures and record it as the morphology difference , the color difference ratio , brightness difference ratio , Area Difference and shape difference, Substitute into the formula Calculate the image judgment value ;

[0034] By screening, standard tissue immunofluorescence images of corresponding types and fields are obtained, and comparison images are further refined for targeted comparison and judgment. Compared with the basic algorithm similarity comparison, the comparison applied to the subdivided field focuses more on the judgment of fluorescent marker color, fluorescence intensity, fluorescence distribution, and cell morphology rather than the comparison with the similarity of existing images. The effect of this decision is that it can further judge those images that do not meet the similarity requirements with the existing ones, and extract images that can be supplemented to the standard database, thereby supplementing the database and overcoming the defect of determining whether the image is standard or not only by similarity judgment.

[0035] Get the uploader information, which includes the number of uploaders N who have uploaded the newly added images and the number of standard tissue immunofluorescence images uploaded by each uploader , the number of times you uploaded pictures , substitute into the formula Calculate the parameter impact value corresponding to each uploader , the greater the parameter influence value, the more reference the uploader has for the judgment result of the newly added pictures, where n represents the label of each uploader, and is the preset weight factor, which is the parameter influence value corresponding to each uploader Substitute the number of uploaders N into the preset model Calculate in and get the uploader judgment value corresponding to each newly added picture ,in is the preset weight factor, and the image judgment value and uploader judgment value After normalization, enter the formula Calculate in and get the standard reference value of the newly added picture . The standard reference value can indirectly reflect the degree to which the newly added picture meets the requirements for inclusion of standard pictures. The larger the standard reference value, the more the newly added picture meets the requirements for standard immunohistofluorescence pictures of tissues.

[0036] When the upload source is an existing extracted immunohistofluorescence picture of tissue and 0 < P ≤ A, obtain the source location data and specific source data of the uploaded immunohistofluorescence picture of tissue. There are multiple preset standard source locations. When the source location data is the same as any one of the multiple standard source locations, the retrieval system retrieves the relevant files to which the immunohistofluorescence picture of tissue belongs according to the specific source data. When there is a picture in the retrieved files with a similarity greater than B to the uploaded immunohistofluorescence picture of tissue, mark the uploaded immunohistofluorescence picture of tissue as a supplementary picture; otherwise, mark the newly added picture as a picture that does not meet the standards and reject it.

[0037] It should be noted that when the user uploads an existing extracted immunohistofluorescence picture of tissue, the user needs to fill in the source location data and specific source data of the picture. The source location data includes the journal or periodical to which it belongs, the thesis website, etc., and the specific source data includes the specific journal number, website, etc.

[0038] By performing source verification on the uploaded existing extracted immunohistofluorescence pictures of tissues, ensure that the standard pictures supplemented into the standard database have a clear and standard-compliant source, thereby enhancing the reference value and authority of the standard database.

[0039] When A < P ≤ B, mark the uploaded immunohistofluorescence picture of tissue as a similar picture, obtain the type information of the similar picture, obtain the type information of the standard picture with a similarity value of P corresponding to the similar picture, calculate the type similarity value between the two pictures. The type similarity value is equal to the number of identical type information divided by the total number of type information categories. (For example, in a specific embodiment, the type information includes cell type, tissue structure, and disease state. Among them, the identical type information is tissue structure and disease state, so the type similarity value is 2 / 3). When the type similarity value is greater than the preset type similarity threshold, mark it as a supplementary picture and generate a picture type error signal to be sent to the uploading user. When the type similarity value is less than or equal to the preset type similarity threshold, mark the newly added picture as a picture that does not meet the standards.

[0040] It should be noted that when A < P ≤ B, it indicates that the uploaded immunohistofluorescence picture of tissue is somewhat similar to the existing standard picture. By further comparing the type information of the pictures, clarify whether there is a close type connection between the similar picture and the standard picture. When the type similarity degree between the two exceeds a certain threshold, it is reasonable to consider that the similar picture is a standard picture with some differences. By supplementing the similar picture into the database, the richness of the pictures in the database can be supplemented, and the comprehensiveness of the database can be enhanced.

[0041] When B < P ≤ 1, mark the uploaded tissue immunofluorescence image as an existing image, reject it, and generate an image already exists signal to send to the uploading user;

[0042] It should be noted that when B < P ≤ 1, it indicates that the uploaded tissue immunofluorescence image is significantly the same as one or more of the existing standard images. Therefore, its supplementary effect on the database is small, and it is rejected to avoid data bloat and duplication in the database.

[0043] Step 4: Reject the images that do not meet the standards and generate an image non-compliance signal to send to the uploading user. The standard processing unit obtains the supplementary images and performs standardization processing. Set a set of standardization values, which contains multiple different types of standard intervals. The types of standard intervals include size, data size, format, resolution, etc. Perform standardization processing on the supplementary images so that the corresponding values of each type are within the standard intervals to obtain standard supplementary images. Perform quality preprocessing on the standard supplementary images, including denoising, enhancing contrast, color correction, etc., to improve the image quality. Supplement the processed standard supplementary images into the standard database.

[0044] Through operations such as noise elimination and contrast enhancement in the standardization process, the key features in the image can be made more prominent, thereby improving the accuracy and reliability of image matching. Limiting the standard intervals, such as size, can remove redundant and irrelevant information in the image, reduce the data volume, make subsequent processing more efficient, reduce the complexity of the image and the amount of calculation, and significantly improve the speed of image processing.

[0045] More specifically, the similarity calculation process between the uploaded tissue immunofluorescence image and the existing tissue immunofluorescence images is as follows:

[0046] Calculate the color similarity value between the two images through the color histogram comparison algorithm and denote it as ; calculate the texture similarity value between the two images through the texture histogram comparison algorithm and denote it as ; extract all the standard images in the current standard database with the same type information as the uploaded tissue immunofluorescence image to form an algorithm learning data set. Train a convolutional neural network (CNN) model through the algorithm learning data set and calculate the feature similarity value between the two images through the convolutional neural network model ; substitute it into the formula for calculation to obtain its similarity where is a preset weight factor.

[0047] It should be noted that by combining different algorithms to calculate the similarity between the uploaded tissue immunofluorescence image and the existing tissue immunofluorescence image, a more accurate similarity value can be obtained to determine the degree of similarity between the two images.

[0048] The preferred embodiments of the present invention disclosed above are only used to help explain the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to only specific implementation methods. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. A database-based clinical tissue immunofluorescence image classification method, characterized in that: It includes a database initial construction unit, an upload classification unit, a similarity comparison unit, a unified analysis unit, an independent analysis unit, and a standard processing unit. The specific steps are as follows: Step 1: The database initial construction unit presets multiple groups of types. Each group of type labels corresponds to a general group. Each group of types includes multiple type labels. Each tissue immunofluorescence image corresponds to multiple type labels of different groups. A standard initial library is established, and the tissue immunofluorescence images are classified according to the type groups. The user uploads the tissue immunofluorescence images into the standard initial library, and the standard initial library and the received tissue immunofluorescence images together form the existing standard database; Step 2: The standard tissue immunofluorescence images in the standard initial library are denoted as standard images. The upload classification unit is used to obtain the upload source and uploader information of the uploaded tissue immunofluorescence images. The upload sources are divided into actively supplemented tissue immunofluorescence images and existing extracted tissue immunofluorescence images. The similarity comparison unit comprehensively calculates the similarity between the uploaded tissue immunofluorescence image and each existing tissue immunofluorescence image through various image similarity algorithms in the prior art and takes the maximum value denoted as P. An approximate threshold A and a similarity threshold B are set, and A and B satisfy 0.8 < A < B < 1; Step 3: When 0 < P ≤ A, the independent analysis unit distinguishes and analyzes the uploaded tissue immunofluorescence image according to whether the upload source is actively supplemented or existing extracted. When A < P ≤ B, the unified analysis unit marks the uploaded tissue immunofluorescence image as a similar image, conducts a type analysis on the similar images, and determines whether the uploaded tissue immunofluorescence image is a non-compliant standard image or a supplementary image. When B < P ≤ 1, the unified analysis unit marks the uploaded tissue immunofluorescence image as an existing image, rejects the reception, generates an image already exists signal and sends it to the uploading user; When the upload source is an actively supplemented tissue immunofluorescence image and 0 < P ≤ A, the independent analysis unit marks the uploaded tissue immunofluorescence image as a new image, obtains the type information, image information, and uploader information of the new image, analyzes and calculates to obtain the standard reference value of the new image. When the standard reference value is greater than or equal to the preset standard threshold, the new image is marked as a supplementary image. When the standard reference value is less than the preset standard threshold, the new image is marked as a non-compliant standard image; The calculation process of the standard reference value of the new image is as follows: Obtain the fluorescent marker color, fluorescence intensity, fluorescence distribution, and cell morphology of the newly added image, extract the standard tissue immunofluorescence image with the same type information in the standard database as the comparison image, calculate the chromaticity difference of the fluorescent marker color in the newly added image and the comparison image, and divide it by the chromaticity value of the fluorescent marker color in the comparison image to record it as the color difference ratio Calculate the brightness difference between the fluorescent markers in the new image and the comparison image and divide it by the brightness value of the fluorescent marker in the comparison image to record it as the brightness difference ratio , calculate the difference between the fluorescence area ratio in the new image and the comparison image and record it as the area difference , count the number of fluorescently labeled cells and the total area in the newly added pictures, calculate the total area divided by the number of fluorescently labeled cells to get the cell morphology size value of the newly added pictures, count the number of fluorescently labeled cells and the total area in the comparison pictures, calculate the total area divided by the number of fluorescently labeled cells to get the cell morphology size value of the comparison pictures, calculate the ratio of the cell morphology size value of the newly added pictures to the cell morphology size value of the comparison pictures and record it as the morphology difference , the color difference ratio , brightness difference ratio , Area Difference and morphological difference Substitute into the formula Calculate the image judgment value ; Get the uploader information, which includes the number of uploaders N who have uploaded the newly added images and the number of standard tissue immunofluorescence images uploaded by each uploader , the number of times you uploaded pictures , substitute into the formula Calculate the parameter impact value corresponding to each uploader , where n represents the label of each uploader, and is the preset weight factor, which is the parameter influence value corresponding to each uploader Substitute the number of uploaders N into the preset model Calculate in and get the uploader judgment value corresponding to each newly added picture ,in is the preset weight factor, and the image judgment value and uploader judgment value After normalization, enter the formula Calculate in and get the standard reference value of the newly added picture ; When the upload source is an existing extracted tissue immunofluorescence image and 0 < P ≤ A, the independent analysis unit obtains the source location data and source specific data of the uploaded tissue immunofluorescence image. Multiple standard source locations are preset. When the source location data is the same as any one of the multiple standard source locations, the retrieval system retrieves the relevant files to which the tissue immunofluorescence image belongs according to the source specific data. When there is an image in the retrieved files whose similarity to the uploaded tissue immunofluorescence image is greater than the similarity threshold B, the uploaded tissue immunofluorescence image is marked as a supplementary image. Otherwise, the new image is marked as a non-compliant standard image; Step 4: Reject the pictures that do not meet the standards and generate a picture non-compliance signal to be sent to the uploading user. The existing standard database receives the supplementary pictures, and the standard processing unit performs standardized preprocessing on the supplementary pictures.

2. A database-based clinical tissue immunofluorescence image classification method according to claim 1, characterized in that: The process of calculating the similarity between the uploaded tissue immunofluorescence picture and the existing tissue immunofluorescence pictures is as follows: The color similarity value between two images is calculated by the color histogram comparison algorithm and recorded as , the texture similarity value between two images is calculated by the texture histogram comparison algorithm and recorded as , extract all standard images in the current standard database that have the same type information as the uploaded tissue immunofluorescence image to form an algorithm learning data set, train the convolutional neural network model through the algorithm learning data set, and calculate the feature similarity value between the two images through the convolutional neural network model , substitute into the formula Calculate the similarity ,in is the preset weight factor.

3. A database-based clinical tissue immunofluorescence image classification method according to claim 2, characterized in that: When A < P ≤ B, obtain the type information of the similar picture and the standard picture with the corresponding similarity value of P, and calculate the type similarity value between the two pictures. The type similarity value is equal to the number of the same type information divided by the total number of type information categories. When the type similarity value is greater than the preset type similarity threshold, it is marked as a supplementary picture and a picture type error signal is sent to the uploading user. When the type similarity value is less than or equal to the preset type similarity threshold, the new picture is marked as a non-compliant picture.

4. A database-based clinical tissue immunofluorescence image classification method according to claim 3, characterized in that: The standardized processing process of the standard processing unit is as follows: Set a set of standardized values, which contains multiple different types of standard intervals. Perform standardized processing on the supplementary pictures so that the corresponding values of each type are within the standard intervals to obtain the standard supplementary pictures. Perform quality preprocessing on the standard supplementary pictures and supplement the processed standard supplementary pictures into the standard database.

Images