A medical system visualization device and label marking method thereof

Abstract

The present invention relates to a medical system visualization device and a labeling method thereof, wherein the medical system visualization device establishes a communication connection with an image labeling system; the medical system visualization device at least includes an imaging unit, an image display unit, an image analysis unit and an image labeling unit, the image labeling system matches the label in at least one label ontology library based on the keyword of the labeling request, and individually allocates the medical image to be labeled to the labeling operator; the imaging unit converts the image information sent by the medical image labeling unit into a medical image and presents it to the labeling operator in the image display unit, the image analysis unit records the labeling information of at least one labeling operator and counts the intersection labels of the same medical image; the image labeling unit locally completes the matching of the medical image and the label according to the received medical image, the labeling content matching the medical image and the matching relationship code of the medical image. The present invention automatically recommends labels for users and supports multiple users to work collaboratively.

Description

[0001] This invention is a divisional application of the invention patent with the application number 201580084598.X, the application title of "A Medical Image Annotation Method and System", and the application date of December 24, 2015. Technical Field

[0002] The present invention relates to the technical field of image processing, and particularly to a medical system visualization device and its label annotation method. Background Art

[0003] So far, various medical academic institutions in the world have accumulated a large number of medical images. These images are huge in quantity and diverse in type, and the management, retrieval, and reuse of these medical images have always been a difficult problem.

[0004] Currently, in terms of image retrieval, Content-Based Image Retrieval (CBIR) is a commonly used solution. CBIR performs retrieval by comparing the visual features of an image with the visual features of the retrieval condition (such as a picture) input by the user. However, due to the problem of the "semantic gap" between the underlying visual features and the high-level semantics of human visual cognition, in the medical field, the retrieval results of CBIR are often not ideal. Therefore, the main method of image retrieval still needs to be based on the text information of the image, and the text label annotation of the image becomes crucial.

[0005] The current image text label annotation methods are mainly manual annotation or pure automatic annotation. Manual annotation has low efficiency, high labor costs, and is completely dependent on the professional knowledge of the annotation operators. The quality of the labels annotated for a long time cannot be guaranteed. Although automatic annotation is efficient, there is currently no label recommendation method that can completely guarantee the quality.

[0006] Chinese Patent (CN 104462738 A) discloses a method for annotating medical images. The method includes: dividing an unannotated medical image set into at least two unannotated medical image subsets; allocating the at least two unannotated medical image subsets to at least two annotation terminals for each annotation terminal to annotate the medical images in the unannotated medical image subset assigned to it; receiving the annotation information uploaded by each annotation terminal. Although this patent enables users to perform collaborative annotation of medical images at any time and place, the pure manual annotation is completely dependent on the professional knowledge of the annotation operators, unable to guarantee the quality of the labels, and the pure manual label annotation is slow and inefficient. Therefore, there is a need in the market for a semi-automatic medical image annotation method to improve the quality and efficiency of annotation.

[0007] In the prior art, no technical solution discloses the method of adjusting the label recommendation order based on the change of the annotation weight of the label in the present invention. Summary of the Invention

[0008] Aiming at the deficiencies of the prior art, the present invention provides a medical image annotation method, which is characterized in that the method includes:

[0009] Recommending at least two selectable labels for the annotation operator to choose from to at least two annotation terminals based on the matching value between the keywords of the medical image to be annotated and the labels in the label ontology library;

[0010] The annotation operators of at least two annotation terminals independently annotate the medical image to be annotated based on the selectable labels;

[0011] Confirming the annotation label of the medical image to be annotated based on the comparison result of the annotation contents of at least two annotation operators.

[0012] According to a preferred embodiment, the method further includes:

[0013] Allocating the medical image to be annotated to at least two annotation operators for independent annotation, and confirming the annotation result based on the intersection of the labels annotated by at least two annotation operators.

[0014] According to a preferred embodiment, the method further includes:

[0015] Allocating the medical image to be annotated to at least two annotation operators for independent annotation, and obtaining the confirmed annotation result by fusing the weights of the labels annotated by at least two annotation operators.

[0016] According to a preferred embodiment, the method further includes:

[0017] Allocating the medical image to be annotated to at least two annotation operators for independent annotation, comparing the annotation labels of the at least two annotation operators, respectively displaying the annotation results with larger differences to the at least two annotation operators, and having the at least two annotation operators negotiate and confirm the annotation result.

[0018] According to a preferred embodiment, the method further includes:

[0019] Based on the medical images from authoritative journals and books, automatically finding out the sentences related to the medical images from the full texts of the authoritative journals and books by using full-text indexing, so as to generate at least two labels for the annotation operator to choose from based on the sentences.

[0020] According to a preferred embodiment, the method further includes:

[0021] The label is displayed on the annotation terminal device of the annotation operator in the form of a selectable button.

[0022] According to a preferred embodiment, the method further includes:

[0023] The label is displayed on the terminal device of the annotation operator in a manner combined with statements related to medical images.

[0024] According to a preferred embodiment, the method further includes:

[0025] Sort the matching values of keywords and labels, and select at least two labels greater than the matching threshold as the recommended selection labels.

[0026] According to a preferred embodiment, the method further includes:

[0027] Generate a dynamically changing label queue based on the annotation content of public annotators, and add the labels less than the sequence threshold to the label ontology library after confirmation by experts; wherein,

[0028] The order of the labels in the label queue is based on the change of annotation weights,

[0029] The annotation ability value of the public annotator changes based on the order of the labels in the label queue,

[0030] The number of labels added to the label ontology library correspondingly increases the annotation ability value of the corresponding public annotator.

[0031] A medical image annotation system, characterized in that the annotation system includes an import unit for importing medical images, a first storage server for storing medical images and their keywords, a second storage server for storing labels, a matching unit, an allocation unit for allocating the medical images to be annotated and the corresponding selection labels to at least two annotation terminals, a confirmation unit for confirming the annotation results of at least two annotation operators, and at least two annotation terminals;

[0032] The import unit imports and stores the medical images to be annotated into the first storage unit;

[0033] The matching unit extracts the keywords of the medical images to be annotated stored in the first storage server and at least one label stored in the second storage unit for matching,

[0034] The allocation unit recommends at least two selection labels for the annotation operator to select based on the matching value between the keyword of the medical image to be annotated and the label in the label ontology library;

[0035] The confirmation unit confirms the annotation label of the medical image to be annotated based on the comparison result of the annotation content of at least two annotation operators.

[0036] According to an independent aspect of the present invention, a multi-person collaborative semi-automatic medical image annotation method is disclosed. The method is characterized in that the medical image to be annotated is assigned to at least two annotation terminals for annotation operations. The annotation operations are completed based on at least two tags recommended by the multi-person collaborative semi-automatic medical image system for the annotation operators of the annotation terminals, and the multi-person collaborative semi-automatic medical image system fuses or compares the annotation results independently completed by the annotation operators of the at least two annotation terminals to determine the annotation tags for the medical image to be annotated.

[0037] According to a preferred embodiment, the multi-person collaborative semi-automatic medical image system determines the annotation tags for the medical image to be annotated in the following manner:

[0038] The annotation tags for the medical image to be annotated are the intersection of the annotation results of the multi-person collaborative semi-automatic medical image system for the at least two annotation terminals; for the medical image to be annotated with an empty intersection, the multi-person collaborative semi-automatic medical image system re-sends the medical image to be annotated to the at least two annotation terminals for annotation operations until the multi-person collaborative semi-automatic medical image system determines the annotation tags for the medical image to be annotated; or

[0039] For the medical image to be annotated with an empty intersection, the multi-person collaborative semi-automatic medical image system simultaneously displays the annotation results of the at least two annotation terminals to the annotation operators of the at least two annotation terminals, and the at least two annotation operators determine the annotation tags for the medical image to be annotated after negotiation.

[0040] According to a preferred embodiment, the multi-person collaborative semi-automatic medical image system determines the annotation tags for the medical image to be annotated in the following manner:

[0041] The multi-person collaborative semi-automatic medical image system compares the annotation results of the at least two annotation terminals. For the annotations with differences in the comparison results, the multi-person collaborative semi-automatic medical image system simultaneously displays the annotation results of the at least two annotation terminals to the annotation operators of the at least two annotation terminals, and the at least two annotation operators determine the annotation tags for the medical image to be annotated after negotiation.

[0042] According to a preferred embodiment, the multi-person collaborative semi-automatic medical image system recommends at least two selectable tags for the annotation operators of the at least two annotation terminals in the following manner:

[0043] The multi - person collaborative semi - automatic medical image system is based on medical images from authoritative journals and books. By using full - text indexing, relevant statements related to the medical images are automatically found from the full texts of the authoritative journals and books, so as to generate at least two tags for the annotation operator to choose from. And the tags are displayed on the annotation terminal of the annotation operator in the form of selectable buttons and / or the tags are displayed on the annotation terminal of the annotation operator in a way combined with the relevant statements.

[0044] According to a preferred embodiment, keywords are extracted from the automatically found statements related to the medical images, and the keywords are matched with the tags in the tag ontology library. At least two tags for the annotation operator to choose from are generated according to the matching degree between the keywords and the tags.

[0045] According to a preferred embodiment, the annotation operators of the at least two annotation terminals mark the boundaries of the regions of interest in the medical image to be annotated based on the selected tags and the statements related to the tags.

[0046] According to a preferred embodiment, the multi - person collaborative semi - automatic medical image annotation system distributes the medical image to be annotated to at least two annotation terminals in the following ways:

[0047] Distribute the medical image to be annotated to the at least two annotation terminals based on the pre - set priority order of the annotation terminals; or

[0048] Distribute the medical image to be annotated to the at least two annotation terminals based on the order of the terminal processing capabilities of the at least two annotation terminals; or

[0049] Distribute the medical image to be annotated to the at least two annotation terminals based on the principle of load balancing.

[0050] According to a preferred embodiment, the medical image is stored on the first server, the annotation matching the medical image is stored on the second server, and the matching relationship between the medical image and the tag is stored as a data record on the second server; when the user extracts the annotated medical image, the first server and the second server send data concurrently, and then the user matches the tag with the medical image according to the data record from the second server and displays it locally.

[0051] According to a preferred embodiment, the annotation terminal is a feature phone, a smart phone, a personal digital assistant, a personal computer, a tablet computer or a personal digital assistant.

[0052] According to a preferred embodiment, the method includes the following steps:

[0053] The multi-person collaborative semi-automatic medical image annotation system assigns the medical image to be annotated to at least two annotation terminals for annotation operations;

[0054] The multi-person collaborative semi-automatic medical image system is based on medical images from authoritative journals and books, and automatically finds statements related to the medical images from the full texts of the authoritative journals and books by using full-text indexing, so as to generate at least two labels for the annotation operator to select based on the related statements, and the system displays the labels to the at least two annotation terminals;

[0055] The annotation operators of the at least two annotation terminals independently complete the annotation of the medical image to be annotated based on the labels recommended by the system;

[0056] The system uses a fusion or comparison method for the annotation results independently completed by the annotation operators of the at least two annotation terminals to determine the annotation label for the medical image to be annotated.

[0057] According to another independent aspect of the present invention, the present invention discloses a multi-person collaborative semi-automatic medical image annotation system, characterized in that the system matches based on the text of the medical image with the labels and / or articles in the label ontology library unit of the medical image, and automatically recommends at least two matching labels and / or related statements to at least two annotation terminals.

[0058] According to a preferred embodiment, the label ontology library unit includes an annotated label unit and a medical journal and book data unit. When the operator of the annotation terminal imports an unannotated image into the system, it retrieves in the annotated label unit and / or the medical journal and book data unit according to the image text information, and generates at least two labels and / or related statements based on the retrieval result matching score.

[0059] According to a preferred embodiment, the at least two generated labels are displayed on the operator's annotation terminal in the form of selectable buttons.

[0060] According to a preferred embodiment, the at least two generated labels are displayed on the operator's annotation terminal in a manner combined with the related statements.

[0061] According to a preferred embodiment, the system further includes a manual input label unit, and the manual input label unit allows the operator to manually input accurate labels based on the at least two labels and / or related statements.

[0062] According to a preferred embodiment, the system further includes an allocation unit and a comparison unit. The allocation unit is configured to allocate unlabeled medical images to at least two labeled terminal operators, and the at least two operators independently complete the labeling respectively; the comparison unit is configured to compare and analyze the labeling results of the at least two operators. If the comparison results for the same medical image are different, the comparison unit sends the labeling results of the at least two operators to the at least two operators simultaneously after comparison and analysis, and the at least two operators negotiate to determine the accurate label.

[0063] According to a preferred embodiment, the system further includes a high-speed remote server unit and a labeled content server unit. The medical images are stored in the high-speed remote server unit, the labels matching the medical images are stored in the labeled content server unit, and the matching relationship data records between the medical images and the matching labels are stored in the labeled content server unit.

[0064] According to a preferred embodiment, after receiving the command to extract the labeled medical images, the high-speed remote server unit and the labeled content server unit respectively send relevant data from different locations and display them on the labeling terminal. The operator matches the label with the medical image according to the matching relationship data record from the labeled content server and displays it on the labeling terminal.

[0065] According to a preferred embodiment, the labeled content server unit has an encryption system. According to a preferred embodiment, the system further includes an import and export unit, which is configured to import unlabeled medical images and export the labeled medical images into local files.

[0066] According to another independent aspect of the present invention, the present invention discloses a method for labeling medical images, characterized in that the steps of the method include:

[0067] Respond to the request of at least one labeling terminal and extract the keywords of the description information of the medical image to be labeled;

[0068] Match the keywords with the labels in at least one label ontology library;

[0069] Separate and allocate unlabeled medical images to at least one labeling terminal;

[0070] Recommend at least one selectable label to the labeling operator of the corresponding labeling terminal based on the keywords and the recommended values based on at least one label;

[0071] Retrieve the statements associated with the medical image / or the selectable label in a full-text search manner and mark and display them to the labeling operator;

[0072] Record the annotation information of at least one annotating operator and count the intersection labels of the same medical image.

[0073] According to a preferred embodiment, the medical image to be annotated is stored in a medical image database and is at least divided into at least two subsets of unannotated medical images according to its description information, and each subset includes at least one medical image to be annotated.

[0074] According to a preferred embodiment, the subsets of unannotated medical images are divided according to the description information of the medical images to be annotated and based on the biological anatomical structure or the biological physiological system or a combination of the biological anatomical structure and the biological physiological system.

[0075] According to a preferred embodiment, the medical image database includes an unannotated medical image database and an annotated medical image database. The annotated medical image database is divided into at least two subsets of annotated medical images based on the labels or annotation information of the annotated images, and each subset includes at least one medical image to be annotated.

[0076] According to a preferred embodiment, the subsets of annotated medical images are divided according to the labels or annotation information of the annotated medical images and based on the biological anatomical structure and / or the biological physiological system.

[0077] According to a preferred embodiment, the selected labels include the selected labels generated by matching keywords with the labels in at least one label ontology library and the labels generated by matching keywords with medical images from authoritative journals and books. By using full-text indexing, relevant statements related to the keywords and medical images are automatically found from the full text of the authoritative journals and books to generate at least two labels for the annotating operator to select.

[0078] According to a preferred embodiment, the at least two generated labels for the annotating operator to select are displayed on the annotating terminal of the annotating operator in the form of selectable buttons. At the same time, the statements related to the labels are displayed on the annotating terminal of the annotating operator, and the annotating operator marks the boundary of the region of interest in the medical image to be annotated based on the selected labels and the statements related to the labels.

[0079] According to a preferred embodiment, the method further includes: actively sending unannotated medical images to at least one annotating terminal, including sending the medical images to be annotated to at least two annotating operators at the same time, and the at least two annotating operators independently complete the annotation respectively.

[0080] According to a preferred embodiment, for the annotations independently completed by the at least two annotation operators, a comparison is made; if the comparison results have a large difference, the annotation contents of both parties are simultaneously displayed to the two people, and the two parties negotiate to determine the most accurate annotation label.

[0081] A visualization device for a medical system, characterized in that the visualization device includes an image display unit, an image analysis unit, and an image annotation unit, and the visualization device is connected to a medical image annotation system.

[0082] As an image annotation terminal, the visualization device sends an image annotation request to the image annotation system.

[0083] The image annotation system matches the keyword of the annotation request with the tags in at least one tag ontology library, and separately allocates the medical image to be annotated to the visualization device.

[0084] The image annotation system recommends at least one selection tag to the annotation operator of the visualization device based on the matching value of the keyword of the annotation request and at least one tag.

[0085] The image annotation system retrieves the statements associated with the selection tag in a full-text search manner through the visualization device and marks and displays them to the annotation operator.

[0086] Based on the selection tag and the statements associated with the selection tag displayed by the image display unit, the annotation operator completes the annotation of the medical image to be annotated in the image annotation unit, and the image annotation unit sends the annotation content to the image annotation system.

[0087] The image analysis unit records the annotation information of at least one annotation operator and counts the intersection tags of the same medical image.

[0088] The present invention also provides another visualization device for a medical system. The visualization device for the medical system establishes a communication connection with the image annotation system; the visualization device for the medical system at least includes an imaging unit, an image display unit, an image analysis unit, and an image annotation unit. The image annotation system matches the keyword of the annotation request with the tags in at least one tag ontology library and separately allocates the medical image to be annotated to the annotation operator. The imaging unit converts the image information sent by the medical image annotation unit into a medical image and presents it to the annotation operator in the image display unit. The image analysis unit records the annotation information of at least one annotation operator and counts the intersection tags of the same medical image. The image annotation unit completes the matching of the medical image and the annotation locally according to the received medical image, the annotation content matching the medical image, and the matching relationship code of the medical image.

[0089] Preferably, the image annotation system includes an annotation content server unit. After the preliminary annotation of the medical image to be annotated, the annotation content server unit generates a dynamically changing tag queue based on the annotation content of the public annotators, and adds the tags smaller than the order threshold to the tag ontology library after being confirmed by experts; the order of the tags in the tag queue is based on the change of the annotation weight, and the annotation ability value of the public annotator changes based on the order of the tags in the tag queue.

[0090] Preferably, the annotation content server unit evaluates the annotation ability value of the corresponding public annotator based on the dynamic change of the tag in the tag queue; if the order of the tag keeps moving forward, the annotation ability value of its public annotator will increase; if the order of the tag keeps moving backward, the annotation ability value of its public annotator will decrease.

[0091] Preferably, after the tag is incorporated into the tag entity library, the annotation ability value of the public annotator corresponding to the tag will increase, and the increased score is set by the administrator; the more tags are incorporated into the tag entity library, the more the annotation ability value of the public annotator corresponding to the tag increases.

[0092] Preferably, the matching unit in the image annotation system recommends at least one selectable tag to the annotation operator of the visualization device based on the matching value between the keyword of the annotation request of the annotation operator and at least one tag.

[0093] Preferably, the annotation operator completes the annotation of the medical image to be annotated based on the selectable tag displayed by the image display unit and the statement associated with the selectable tag, and sends it to the image annotation system.

[0094] Preferably, the annotation weight is obtained by weighting the annotation content based on the qualifications and annotation history of the public annotators.

[0095] The present invention also provides a tag annotation method for a visualization device of a medical system, and the method at least includes: matching the keyword of the annotation request with at least one tag in the tag ontology library, and separately allocating the medical image to be annotated to the annotation operator; converting the image information sent by the medical image annotation unit into a medical image and presenting it to the annotation operator in the image display unit, recording the annotation information of at least one annotation operator and counting the intersection tags of the same medical image; completing the matching of the medical image and the annotation locally according to the received medical image, the annotation content matching the medical image, and the matching relationship code of the medical image.

[0096] The method further includes: after the preliminary annotation of the medical image to be annotated, generating a dynamically changing tag queue based on the annotation content of the public annotators, and adding the tags smaller than the sequential threshold to the tag ontology library after being confirmed by experts; the order of the tags in the tag queue changes based on the annotation weights, and the annotation ability value of the public annotators changes based on the order of the tags in the tag queue.

[0097] The method further includes: evaluating the annotation ability value of the corresponding public annotator based on the dynamic change situation of the tag in the tag queue; if the order of the tag continuously moves forward, the annotation ability value of its public annotator will increase; if the order of the tag continuously moves backward, the annotation ability value of its public annotator will decrease.

[0098] Advantageous technical effects of the present invention:

[0099] First of all, the present invention can automatically recommend tags for users and support multiple users to work collaboratively, which is the most important function.

[0100] Secondly, the present invention provides a user management function. The administrator can easily manage the information of the annotation users using the management tool and allocate the images to be annotated to the annotators.

[0101] In addition, the present invention provides a data import and export function. The annotator can log in to upload their own pictures, and the administrator is responsible for allocating the pictures. In addition to saving the data in the system database, the user can also export the specified data into a local file, and the local file supports two formats of.csv and.xml.

[0102] Finally, since the pictures all come from some articles, the present invention also supports viewing the articles associated with the images and automatically finding the sentences related to the tags and highlighting them. Description of the Drawings

[0103] Figure 1 is a schematic diagram of a preferred medical image annotation method of the present invention;

[0104] Figure 2 is a schematic diagram of a multi-person collaborative semi-automatic medical image annotation method of the present invention;

[0105] Figure 3 is a schematic diagram of another preferred medical image annotation method of the present invention;

[0106] Figure 4 is a schematic diagram of a preferred medical image annotation system of the present invention;

[0107] Figure 5 is a schematic diagram of a multi-person collaborative semi-automatic medical image annotation system of the present invention;

[0108] Figure 6 It is a schematic diagram of a visualization device of a medical system according to the present invention; and

[0109] Figure 7 It is a schematic diagram of the framework of a medical image annotation system according to the present invention. Specific embodiments

[0110] The following will be described in detail with reference to the accompanying drawings.

[0111] In the present invention, a medical picture, also known as a medical image, refers to a picture or image of the internal tissues of an animal body, a human body, or a part of the human body obtained in a non-invasive manner for medical treatment or medical research.

[0112] Embodiment 1

[0113] This embodiment provides a medical image annotation method, characterized in that the method includes recommending at least two selectable tags to at least two annotation terminals based on the matching value between the keyword of the medical image to be annotated and the tags in the tag ontology library, and the annotation operators of at least two annotation terminals independently annotate the medical image to be annotated based on the tags selected by the annotation operators. Alternatively, sort the matching values between the keyword and at least one tag, select at least two tags according to a certain rule, send them to the corresponding annotation terminals and display them as selectable tags for the annotation operators to select.

[0114] As Figure 1 shown, at least one annotation operator inputs an annotation request at at least one annotation terminal. In response to the annotation requirements of the annotation operator, at least one medical image to be annotated is allocated to at least two annotation terminals, so that the annotation operator independently annotates at the annotation terminal. Preferably, the medical image to be annotated is allocated to at least two annotation terminals based on the preset priority order of the annotation terminals. Alternatively, the subset of unannotated medical images is allocated to at least two annotation terminals based on the order of the terminal processing capabilities of the annotation terminals. Alternatively, the subset of unannotated medical images is allocated to at least two annotation terminals based on the principle of load balancing. Alternatively, the subset of unannotated medical images is allocated to at least two annotation terminals based on the annotation terminal priority order, the terminal processing capability order, and the load balancing principle at the same time.

[0115] According to a preferred embodiment, the unlabeled medical image subset is assigned to multiple labeling terminals based on a preset priority order of the labeling terminals. For example, suppose there are three labeling terminals, namely labeling terminal 1, labeling terminal 2, and labeling terminal 3, and each labeling terminal can process 10 medical images. The priority of labeling terminal 1 is the highest, the priority of labeling terminal 2 is the second highest, and the priority of labeling terminal 3 is the lowest. Suppose there are two unlabeled medical image subsets to be assigned, namely unlabeled medical image subset 1 and unlabeled medical image subset 2, and each unlabeled medical image subset contains 10 images. Then, the task assignment result can be: send 10 images in unlabeled medical image subset 1 (or unlabeled medical image subset 2) to labeling terminal 1, send 10 images in the remaining unlabeled medical image subset to labeling terminal 1, and do not send unlabeled images to labeling terminal 3.

[0116] According to a preferred embodiment, the unlabeled medical image subset is assigned to the labeling terminals based on the order of the terminal processing capabilities of the labeling terminals. For example, suppose there are three labeling terminals, namely labeling terminal 1, labeling terminal 2, and labeling terminal 3. Labeling terminal 1 can process 10 medical images, labeling terminal 2 can process 10 medical images, and labeling terminal 3 can process 5 medical images. Suppose there are two unlabeled medical image subsets to be assigned, namely unlabeled medical image subset 1 and unlabeled medical image subset 2, and each unlabeled medical image subset contains 10 images. Then, the task assignment result can be: send 10 images in unlabeled medical image subset 1 (or unlabeled medical image subset 2) to labeling terminal 1, send 10 images in the remaining unlabeled medical image subset to labeling terminal 1, and do not send unlabeled images to labeling terminal 3.

[0117] Extract keywords from the description text of the medical image to be labeled, and match the keywords with at least one label in the label ontology library. Record the matching values of the keywords and the at least one label. Recommend at least two labels with matching values greater than a preset matching threshold to the corresponding labeling terminal for the labeling operator to select. The labels are displayed on the labeling terminal of the labeling operator in the form of selectable buttons. Or, the labels are displayed on the labeling terminal of the labeling operator in a manner combined with statements related to the medical image.

[0118] Or, based on medical images from authoritative journals and books, use full-text indexing to automatically find statements related to the medical images from the full texts of authoritative journals and books. Generate labels for the labeling operator to select from the relevant statements for labeling.

[0119] After at least two annotation operators annotate the same medical image, compare the annotation contents of the at least two annotation operators. Confirm the annotation result based on the intersection of the at least two annotation contents.

[0120] If the annotation contents of the at least two annotation operators are quite different, display the annotation contents or selected tags of other annotation operators to the annotation operators respectively on the annotation terminals. Request the annotation operators to re-annotate the medical image to be annotated. Alternatively, in the case where the annotation contents of the at least two annotation operators are quite different, establish a communication connection or an instant messaging connection for the at least two annotation operators annotating the same medical image. Have the at least two annotation operators confirm the final annotation tag through negotiation.

[0121] Alternatively, after at least two annotation operators annotate the same medical image, confirm the annotation result based on the weight fusion of the tags selected by the at least two annotation operators. Thus, obtain the final annotation tag.

[0122] Embodiment 2

[0123] This embodiment provides a medical image annotation method, which is characterized in that the method includes recommending at least two selectable tags to at least two annotation terminals based on the matching value between the keyword of the medical image to be annotated and the tags in the tag ontology library, and having the annotation operators of the at least two annotation terminals independently annotate the medical image to be annotated based on the tags selected by the annotation operators.

[0124] First, classify and store a large number of keywords related to the annotated medical images and the annotated tags as samples. Establish mapping relationships for the annotated medical images, keywords, and tags respectively, and store the mapping relationships in a database.

[0125] Then, calculate the image similarity between each annotated medical image in the database and the medical image to be annotated respectively. Among them, the image similarity is mainly used to calculate the similarity degree of the image contents of two pictures, and an image similarity value is obtained. The higher the image similarity value, the more similar the contents of the two pictures are. The image similarity can be calculated through the visual features of the two pictures. The visual features can specifically be color RGB (Red Green Blue, three primary colors) features, texture features, histogram features, and SIFT (Scale-invariant feature transform) features, etc.

[0126] Select the labeled medical images with an image similarity greater than the first threshold to the medical image to be labeled to form a picture group. Here, the first threshold is a preset similarity value. Specifically, the first threshold can be set by the labeling operator himself. Relatively speaking, the higher the first threshold is set, the more similar the labeled images found in the database are to the image to be labeled, but the number of labeled pictures found will be relatively small.

[0127] Extract the labels corresponding to each labeled image in the medical image group to form a label phrase group. The corresponding keywords form a keyword group. Extract keywords and labels according to the mapping relationship of the labeled images. If the labeled images with printed labels are stored in the database, first identify the keywords in the description text of the labeled images, and then extract the labels.

[0128] Output at least one label in the label phrase group as the selected label for the image to be labeled. Since there may be many labels in the label phrase group, but the user may not want to output too many labels and only hopes to output a preset number of labels, the following steps can be used to achieve this: Determine whether the number of labels in the label phrase group is greater than the third threshold. When the number of labels in the label phrase group is greater than the third threshold, output the preset number of labels in the label phrase group as the selected label for the image to be labeled, and the preset number is less than or equal to the third threshold. Specifically, the preset number and the third threshold are the number of labels set by the labeling operator himself.

[0129] Display the corresponding at least one label in the form of selectable buttons on the labeling terminal.

[0130] According to a preferred embodiment, the method for extracting keywords from the description text of the medical image to be labeled includes: performing word segmentation on the text information to obtain at least one word segment, and obtaining the semantic content and semantic type of at least one word segment. The semantic content is the semantic information with meaning corresponding to the word segment, and the semantic type is the type of semantic information, for example, the part of speech of the word segment, the meaning represented by the word segment, etc.

[0131] Filter at least one word segment in the corresponding keyword group according to the semantic content and semantic type to filter out the keywords related to the medical image to be labeled. The labeled medical images with extremely high similarity to the medical image to be labeled correspond to multiple keywords. Select the keyword with the highest semantic similarity to the word segment in the description text in the keyword group as the keyword of the medical image to be labeled. The semantic similarity mainly calculates the similarity degree of the semantics of two words to obtain a semantic similarity value. The higher the semantic similarity value, the more similar the semantics of these two words are. The second threshold can specifically be the semantic similarity value preset by the user. Obtain at least one label according to the mapping relationship between the keyword and the label phrase group.

[0132] Similarly, the label can be a relevant statement related to the medical image to be labeled. The relevant statements are obtained by performing a full-text index on authoritative journals or books. If the labeled medical images are from authoritative journals and books, the articles in the authoritative journals and books are established as relevant statement groups. For the descriptive text of the medical image to be labeled, relevant statements related to the medical image are automatically found from the full text of authoritative journals and books by using the full-text index as recommended labels. The relevant statements are displayed in the form of selectable buttons on the labeling terminal of the labeling operator.

[0133] Embodiment III

[0134] This embodiment provides a multi-person collaborative semi-automatic medical image labeling method. By allocating the medical image to be labeled to at least two labeling terminals for labeling operations, the labeling operations are completed based on at least two labels recommended by the multi-person collaborative semi-automatic medical image system for the labeling operators of the labeling terminals to select, and the multi-person collaborative semi-automatic medical image system fuses or compares the labeling results independently completed by the labeling operators of at least two labeling terminals to determine the labeling label for the medical image to be labeled.

[0135] As Figure 2 shown, the labeling operator issues a labeling request on the labeling terminal. In response to the labeling request of the labeling terminal, at least one medical image to be labeled is allocated to at least two labeling terminals. The allocation methods include: allocating the medical image to be labeled to at least two labeling terminals of the labeling operator based on the pre-set priority order of the labeling terminals; or, allocating the medical image to be labeled to at least two labeling terminals based on the terminal processing capacity order of at least two labeling terminals; or, allocating the medical image to be labeled to at least two labeling terminals based on the principle of load balancing.

[0136] Based on the medical images from authoritative journals and books, relevant statements related to the medical images are automatically found from the full text of authoritative journals and books by using the full-text index, so as to generate at least two labels for the labeling operator to select based on the relevant statements, and the labels are displayed in the form of selectable buttons on the labeling terminal of the labeling operator and / or the labels are displayed on the labeling terminal of the labeling operator in a manner combined with the relevant statements of the labeling operator.

[0137] Alternatively, extract keywords from the automatically found relevant statements related to the medical image, and match the keywords with the labels in the label ontology library, and generate at least two labels for the labeling operator to select according to the matching degree between the keywords and the labels.

[0138] The annotation label of the medical image to be annotated is the intersection of the annotation results of at least two annotation terminals by the multi-person collaborative semi-automatic medical image system. For the medical image to be annotated with an empty intersection, the multi-person collaborative semi-automatic medical image system resends the medical image to be annotated to at least two annotation terminals for annotation operations until the multi-person collaborative semi-automatic medical image system determines the annotation label of the medical image to be annotated.

[0139] Alternatively, for the medical image to be annotated with an empty intersection, the multi-person collaborative semi-automatic medical image system simultaneously displays the annotation results of at least two annotation terminals to the annotation operators of at least two annotation terminals, and the annotation label of the medical image to be annotated is determined after consultation by at least two annotation operators.

[0140] Alternatively, the multi-person collaborative semi-automatic medical image system compares the annotation results of at least two annotation terminals. For the annotations with differences in the comparison results, the multi-person collaborative semi-automatic medical image system simultaneously displays the annotation results of at least two annotation terminals to the annotation operators of at least two annotation terminals, and the annotation label of the medical image to be annotated is determined after consultation by at least two annotation operators.

[0141] The annotation operators of at least two annotation terminals mark the boundaries of the region of interest (ROI) in the medical image to be annotated based on the selected label and the statement related to the label.

[0142] According to a preferred embodiment, the medical image is stored on the first server. The label matching the medical image is stored on the second server. The matching relationship between the medical image and the label is stored as a data record on the second server. When the user extracts the annotated medical image, the first server and the second server concurrently send data respectively, and the user matches the label with the medical image according to the data record from the second server and displays it locally.

[0143] Example 4

[0144] This embodiment provides a medical image annotation method. The steps of the method include;

[0145] As Figure 3 shown, in response to a request from at least one annotation terminal, extract the keywords of the description information of the medical image to be annotated. Match the keywords with the labels in at least one label ontology library. Separately allocate unannotated medical images to at least one annotation terminal. Recommend at least one selectable label to the annotation operator of the corresponding annotation terminal based on the matching value between the keyword and at least one label. Retrieve the statements associated with the medical image / or selectable label in a full-text search manner and mark and display them to the annotation operator. Record the annotation information of at least one annotation operator and count the intersection of the same medical image.

[0146] Specifically, the steps of a medical image annotation method include:

[0147] S01: In response to a request from at least one annotation terminal, extract keywords from the description information of the medical image to be annotated.

[0148] At least one annotation operator issues an annotation request on the annotation terminal. In response to the request from at least one annotation terminal, extract keywords from the description information of the medical image to be annotated. The medical image to be annotated is accompanied by descriptive text, and the descriptive text contains keywords. Extract the keywords from the descriptive text.

[0149] S02: Match the keywords with the labels in at least one label ontology library.

[0150] S03: Recommend at least one selectable label to the annotation operator of the corresponding annotation terminal based on the keywords and the matching values based on at least one label.

[0151] Match the keywords with the labels in at least one label ontology library and calculate the matching values. According to the order of the magnitudes of the matching values, recommend at least one selectable label for the annotation operator to choose from to the annotation operator of the corresponding annotation terminal.

[0152] Proactively send the unannotated medical images to at least one annotation terminal, including sending the medical images to be annotated to at least two annotation operators simultaneously. Let at least two annotation operators complete the annotation independently. The selectable labels for the annotation operator are displayed on the annotation terminal simultaneously with the corresponding medical images to be annotated. Meanwhile, the annotation terminal also displays a manual annotation input field. When the annotation operator is not satisfied with any of the displayed selectable labels, the operator can enter manual annotations in the manual annotation input field.

[0153] S04: Retrieve the statements associated with the medical image and / or the selectable labels in a full-text search manner and mark and display them to the annotation operator.

[0154] Based on the medical images from authoritative journals and books, automatically find the statements related to the medical images from the full texts of the authoritative journals and books by using full-text indexing, so as to generate at least two selectable labels for the annotation operator based on the statements.

[0155] S05: Record the annotation information of at least one annotation operator and count the intersection of the same medical image.

[0156] Record the annotation information of at least one annotation operator. Count the intersection of the labels annotated by at least one annotation operator for the same medical image. Use the labels in the intersection as the final annotation labels for the medical image to be annotated.

[0157] According to a preferred embodiment, the medical images to be labeled are stored in a medical image database and are divided into at least two unlabeled medical image subsets according to the description information of the medical images to be labeled, and each subset includes at least one medical image to be labeled.

[0158] The medical image database stores an unlabeled medical image set and a labeled medical image set. The labeled medical image set contains medical images that have been labeled. The unlabeled medical image set contains medical images that have not been labeled. The medical image database can have a central structure or a distributed structure. Moreover, the storage capacity of the medical image database can also be correspondingly expanded as the number of medical images increases.

[0159] After receiving the task of labeling the unlabeled medical image set in the medical image database, the unlabeled medical image set can be divided into multiple (at least two) unlabeled medical image subsets, and each unlabeled medical image subset can include one or more medical images.

[0160] According to a preferred embodiment, the unlabeled medical image subsets are divided based on the biological (such as human) anatomical structure. For example, according to the anatomical structure of the human body, the unlabeled medical image set can be specifically divided into: image subsets such as the brain, chest, heart, abdomen, upper limbs, and lower limbs.

[0161] According to a preferred embodiment, the unlabeled medical image subsets are divided according to the biological physiological system structure. For example, the unlabeled medical image set is specifically divided into image subsets such as the digestive system, nervous system, musculoskeletal system, endocrine system, urinary system, reproductive system, circulatory system, respiratory system, and immune system.

[0162] In fact, the unlabeled medical image set can also be refined and distinguished at multiple levels. For example, for brain images, they can also be divided into image subsets such as the Central Core, Limbic System, and Cerebral Cortex.

[0163] According to a preferred embodiment, the relevant heart images are combined into image subsets such as the aorta, left atrium, left ventricle, right atrium, and right ventricle.

[0164] According to a preferred embodiment, in order to facilitate the subsequent combination of medical image subsets, an identifier can be assigned to each medical image subset. All medical images in the same medical image subset share the same identifier.

[0165] According to a preferred embodiment, the labeled medical image database is divided into at least two labeled medical image subsets based on the labels or annotation information of the labeled images, and each subset includes at least one medical image to be labeled.

[0166] According to a preferred embodiment, the subset of labeled medical images is divided according to the labels or annotation information of the labeled medical images and based on biological anatomical structures and / or biological physiological systems.

[0167] Example Five

[0168] This embodiment provides a medical annotation system. As Figure 4 shown, the annotation system includes an import unit for importing medical images, a first storage server for storing medical images, a second storage server for storing labels and / or statements related to medical images, a matching unit, an allocation unit for allocating the medical images to be annotated and the corresponding selected labels to at least two annotation terminals, a confirmation unit for confirming at least two annotation results, and at least two annotation terminals.

[0169] The import unit imports and stores the medical images generated by the visible medical image device into the first storage unit.

[0170] The first storage server divides the medical images into at least two subsets of medical images and stores the medical images and their keyword information classified.

[0171] The matching unit extracts the keywords of the medical images to be annotated stored in the first storage server and at least one label and / or statement related to the medical image stored in the second storage unit, matches and calculates the matching value locally, and assigns at least one label and / or statement related to the medical image that meets the conditions as the selected label to the corresponding annotation terminal.

[0172] The allocation unit allocates the medical images to be annotated to at least two annotation terminals based on the preset priority order of the annotation terminals, or the allocation unit allocates the medical images to be annotated to at least two annotation terminals based on the terminal processing capacity order / based on the load balancing principle.

[0173] The confirmation unit confirms the annotation result of the medical image to be annotated based on the annotation contents of at least two annotation operators.

[0174] Example Six

[0175] This embodiment provides a multi - person collaborative semi - automatic medical image annotation system. As Figure 5 shown, the multi - person collaborative semi - automatic medical image annotation system at least includes a label ontology library unit for storing labels and / or articles / statements related to medical images, a manual input label unit, an allocation unit, a comparison unit, a high - speed remote server unit, and an annotation content server unit.

[0176] The multi - person collaborative semi - automatic medical image annotation system matches tags and / or articles in the text and tag ontology library unit of medical images, and automatically recommends at least two matching tags and / or related statements to at least two annotation terminals.

[0177] The tag ontology library unit includes an annotated tag unit and a medical journal and book data unit. When the annotation operator of the annotation terminal imports an unannotated image into the system, it retrieves in the annotated tag unit and / or the medical journal and book data unit according to the image text information, and generates at least two tags and / or related statements based on the retrieval result matching scores. The at least two generated tags are displayed on the annotation terminal of the annotation operator in the form of selectable buttons. Alternatively, the at least two generated tags are displayed on the annotation terminal of the annotation operator in a combined manner with the related statements.

[0178] The manual input tag unit is used for the annotation operator to manually input accurate tags based on at least two tags and / or related statements.

[0179] The allocation unit is used to allocate unannotated medical images to at least two annotation terminal operators, and at least two operators independently complete the annotation respectively. The comparison unit is used to compare and analyze the annotation results of at least two operators. If the comparison results for the same medical image show differences, the comparison unit sends the annotation results of at least two operators to at least two operators after comparison and analysis, and at least two operators negotiate to determine the accurate tags.

[0180] According to a preferred embodiment, the medical images are stored in the high - speed remote server unit. The tags matching the medical images are stored in the annotation content server unit. The matching relationship data records between the medical images and the matching tags are stored in the annotation content server unit. After receiving the command to extract the annotated medical images, the high - speed remote server unit and the annotation content server unit send relevant data from different locations and display them on the annotation terminal. The annotation operator matches the tags with the medical images according to the matching relationship data records from the annotation content server and displays them on the annotation terminal. Among them, the annotation content server unit has an encryption system.

[0181] According to a preferred embodiment, the multi - person collaborative semi - automatic medical image annotation system further includes an import and export unit. The import and export unit is used to import unannotated medical images and to export the annotated medical images to generate local files.

[0182] Example Seven

[0183] This embodiment provides a visualization device for a medical system. As Figure 6As shown in the figure, the visualization device includes an imaging unit, an image display unit, an image analysis unit, and an image annotation unit. The image annotation unit completes all image annotations based on the images generated by the imaging unit in combination with biological anatomical structures or biological physiological systems.

[0184] Meanwhile, as an image annotation terminal, the visualization device sends an image annotation request to the image annotation system.

[0185] The image annotation system matches the keywords in the annotation request with the tags in at least one tag ontology library, and separately allocates the medical images to be annotated to the visualization device. The imaging unit converts the image information sent by the medical image annotation system into medical images and presents them to the annotation operator on the image display unit.

[0186] The image annotation system recommends at least one selection tag to the visualization device based on the matching values of the keywords in the annotation request and at least one tag.

[0187] Alternatively, the image annotation system retrieves the statements associated with the selection tags in a full-text search manner through the visualization device and marks and displays them to the annotation operator.

[0188] The annotation operator completes the annotation of the medical images to be annotated based on the selection tags displayed on the image display unit and the statements associated with the selection tags, and sends them to the image annotation system.

[0189] The image analysis unit can process and analyze the annotation content annotated by the imaging annotation unit in combination with the images already annotated in the image annotation system.

[0190] The image analysis unit receives, through the network, the medical images stored on the high-speed remote server, the annotation content matching the medical images stored on another annotation content server, and the matching relationship code between the annotation content and the medical images.

[0191] The image annotation unit completes the matching of the medical images and the annotations locally according to the received medical images, the annotation content matching the medical images, and the matching relationship code between the annotation content and the medical images, and records the annotation information of at least one annotation operator, and counts the intersection tags of the same medical image.

[0192] As Figure 7 shown, the medical image annotation system of the present invention, MITagger adopts a B / S architecture, is developed using Python + Django in the background, adopts a standard Django MVC framework, and provides a unified HTTP interface. All front-end services are completed by calling the HTTP interface provided by the background server. The background basically adopts a layered and modular architecture, divides the system into several layers, and each layer consists of a certain number of modules.

[0193] The Http Interface encapsulates all functions into interfaces for front - end use. Request Auth is responsible for verifying the identity of the requesting user, and most services can only be accessed by legitimate users. The Network Service completes various functional processes. The Tag Recommdation layer encapsulates a tag recommendation engine based on an entity library. Data Storage is implemented based on DjangoModel and Elasticsearch interfaces and is responsible for data storage management.

[0194] Example Eight

[0195] This embodiment is an improvement on any of the foregoing embodiments.

[0196] After the medical images of the present invention are labeled, crowdsourcing means are used to update the rediscovered and newly discovered information related to the medical images. After the preliminary labeling of the medical images to be labeled, the labeling information is opened to the public users. After the public users register their personal information in the medical image labeling system, they become public labelers and can label the medical images. The labeling content of the public labelers is saved and added to the label queue corresponding to the medical images. The labels in the label queue include the preliminary labeling labels and the labels generated by the public labelers. Based on the qualifications and labeling history of the public labelers, the labeling content is weighted. Under the influence of the weights, the labels labeled by multiple public labelers present a dynamically sorted label queue. When the label labeled by a public labeler leads in the queue and the sorting order is less than a preset order threshold, the label labeled by the public labeler will be verified and confirmed by experts and then added to the label ontology library. In this way, the present invention incorporates new knowledge related to medical images into the label ontology library, thus continuously updating the label ontology library. For example, if the preset order threshold is 5, the labels ranked in the top five in the label queue will be verified and confirmed by experts and then added to the label ontology library.

[0197] According to a preferred embodiment, the labeling ability value of the corresponding public labeler is evaluated based on the dynamic change of the label in the label queue. If the order of the label continuously moves forward, the labeling ability value of the public labeler will increase. If the order of the label continuously moves backward, the labeling ability value of the public labeler will decrease.

[0198] According to a preferred embodiment, the labeling ability value of the public labeler whose at least one label is incorporated into the label ontology library will increase accordingly. After the label is incorporated into the label entity library, the labeling ability value of the public labeler corresponding to the label will increase. The increased score is set by the administrator. The more labels are incorporated into the label entity library, the more the labeling ability value of the public labeler increases.

[0199] The method steps in each embodiment of the present invention can be combined and used with each other. In any embodiment of the present invention, the labeled terminal is any entity with computing and processing capabilities, such as a feature phone, a smart phone, a personal digital assistant, a personal computer, a tablet computer, or a personal digital assistant. The labeled terminal also has a communication function with the network, so that it can receive the medical images to be labeled provided by the medical image labeling system through the network and return the labeling information to the medical image labeling system. The labeling operator can also view the labeled medical images through the labeled terminal. Preferably, the labeling function can be implemented by installing a plug-in in the browser of the intelligent processing device. The browser can include, for example, Internet Explorer, Firefox, Safari, Opera, Google Chrome, GreenBrowser, etc. The labeled terminals can be distributed in different geographical regions. The embodiments of the present invention are not limited to these browsers, but can be applied to any application (APP) that can be used to display files in a web server or a file system and allow users to interact with the files. These applications can be various common browsers currently, or other application programs with web browsing functions. The medical images of the present invention include medical images.

[0200] It should be noted that the above specific embodiments are exemplary. Those skilled in the art can come up with various solutions inspired by the disclosed content of the present invention, and these solutions also fall within the scope of the disclosure of the present invention and within the protection scope of the present invention. Those skilled in the art should understand that the description and drawings of the present invention are illustrative and do not constitute a limitation on the claims. The protection scope of the present invention is defined by the claims and their equivalents.

Claims

1. A medical system visualization device, wherein the medical system visualization device establishes a communication connection with an image annotation system; characterized in that: The medical system visualization device at least includes an imaging unit, an image display unit, an image analysis unit and an image annotation unit. The image annotation system matches the tags in at least one tag ontology library based on the keywords of the annotation request issued by the medical system visualization device, and individually assigns the medical images to be annotated to the annotation operators; The imaging unit converts the image information sent by the medical image annotation unit into a medical image and presents it to the annotation operator in the image display unit. The annotation operator completes the annotation of the medical image to be annotated based on the selection tags and the statements associated with the selection tags displayed by the image display unit, and sends it to the image annotation system. The image analysis unit records the annotation information of at least one annotation operator and counts the intersection labels of the same medical image; The image annotation unit locally completes the matching of the medical image and the annotation according to the received medical image, the annotation content matching the medical image and the matching relationship code of the medical image; The image annotation system comprises an annotation content server unit, After the medical image to be annotated is initially annotated, the annotation content server unit generates a dynamically changing tag queue based on the annotation content of the public annotators, and adds tags that are smaller than the order threshold to the tag ontology library after being confirmed by experts; the order of tags in the tag queue changes based on the tag annotation weights, and the annotation capability values ​​of the public annotators change based on the order of tags in the tag queue; The number of tags added to the tag ontology library will increase the tagging ability value of the corresponding public tagger accordingly.

2. The medical system visualization device according to claim 1, characterized in that: The annotation content server unit evaluates the annotation capability value of the corresponding public annotator based on the dynamic change of the tag in the tag queue; If the order of labels keeps changing forward, the annotation ability value of the public annotators will increase; If the order of labels keeps changing backwards, the annotation ability value of the public annotators will decrease.

3. The medical system visualization device according to claim 1, characterized in that: After the label is included in the label ontology library, the labeling ability value of the public labeler corresponding to the label will increase, and the increased score is set by the management personnel; The more tags included in the tag ontology library, the more the tagging ability value of the public tagger corresponding to the tag increases.

4. The medical system visualization device according to claim 1, characterized in that: The matching unit in the image annotation system recommends at least one selected tag to the annotation operator of the visualization device based on a matching value between a keyword in the annotation request of the annotation operator and at least one tag.

5. The medical system visualization device according to claim 1, characterized in that: The annotation weight is obtained by weighting the annotation content based on the qualifications and annotation history of the public annotators.

6. A labeling method for a medical system visualization device, characterized in that: Matching the keywords in the annotation request with the tags in at least one tag ontology library, and individually assigning the medical images to be annotated to the annotation operators; The image information sent by the medical image annotation unit is converted into a medical image and presented to the annotation operator in the image display unit. The annotation operator completes the annotation of the medical image to be annotated based on the selection tags displayed by the image display unit and the statements associated with the selection tags, and sends it to the image annotation system. Record the annotation information of at least one annotation operator and count the intersection labels of the same medical image; Matching the medical image with the annotation locally according to the received medical image, the annotation content matched with the medical image, and the matching relationship code of the medical image; After the medical images to be annotated are initially annotated, a dynamically changing label queue is generated based on the annotation content of the public annotators, and labels smaller than the order threshold are confirmed by experts and added to the label ontology library; the order of labels in the label queue changes based on the label annotation weights, and the annotation ability values ​​of the public annotators change based on the order of labels in the label queue; The number of tags added to the tag ontology library will increase the tagging ability value of the corresponding public tagger accordingly.

7. The label marking method according to claim 6, characterized in that: Evaluate the annotation capability of the corresponding public annotators based on the dynamic changes of the labels in the label queue; If the order of labels keeps changing forward, the annotation ability value of the public annotators will increase; If the order of labels keeps changing backwards, the annotation ability value of the public annotators will decrease.

Images