WO2017101142A1 - 一种医学图像标注方法及*** - Google Patents

一种医学图像标注方法及*** Download PDF

Info

Publication number
WO2017101142A1
WO2017101142A1 PCT/CN2015/098710 CN2015098710W WO2017101142A1 WO 2017101142 A1 WO2017101142 A1 WO 2017101142A1 CN 2015098710 W CN2015098710 W CN 2015098710W WO 2017101142 A1 WO2017101142 A1 WO 2017101142A1
Authority
WO
WIPO (PCT)
Prior art keywords
labeling
medical image
labeled
label
operator
Prior art date
Application number
PCT/CN2015/098710
Other languages
English (en)
French (fr)
Chinese (zh)
Inventor
安宁
金柳颀
杨矫云
巩博文
段荣
冯逢
Original Assignee
安宁
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 安宁 filed Critical 安宁
Priority to DE212015000240.3U priority Critical patent/DE212015000240U1/de
Priority to CN202210164256.0A priority patent/CN114297429A/zh
Priority to CN201580084598.XA priority patent/CN108463814B/zh
Priority to CN202210115900.5A priority patent/CN114398511A/zh
Publication of WO2017101142A1 publication Critical patent/WO2017101142A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
    • G06F16/58—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
    • G06F16/5866—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using information manually generated, e.g. tags, keywords, comments, manually generated location and time information
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00—Handling natural language data
    • G06F40/10—Text processing
    • G06F40/166—Editing, e.g. inserting or deleting
    • G06F40/169—Annotation, e.g. comment data or footnotes
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H30/00—ICT specially adapted for the handling or processing of medical images
    • G16H30/20—ICT specially adapted for the handling or processing of medical images for handling medical images, e.g. DICOM, HL7 or PACS
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H30/00—ICT specially adapted for the handling or processing of medical images
    • G16H30/40—ICT specially adapted for the handling or processing of medical images for processing medical images, e.g. editing

Definitions

  • the present invention relates to the field of image processing technologies, and in particular, to a medical image annotation method and system.
  • CBIR Content-Based Image Retrieval
  • the current image text labeling method is mainly manual labeling, or pure automatic labeling.
  • Manual labeling is inefficient, labor costs are high, and it relies entirely on the expertise of the labeling operator.
  • the quality of labels that are marked for long periods of time cannot be guaranteed.
  • automatic labeling is highly efficient, there is currently no label recommendation method that can fully guarantee quality.
  • Chinese patent discloses a method for labeling a medical image, the method comprising: dividing an unlabeled medical image set into at least two unlabeled medical image subsets; and the at least two unlabeled medical image sub- The set is allocated to at least two labeling terminals, so that each labeling terminal labels the medical images in the unlabeled medical image subsets that are respectively assigned; and receives the labeling information uploaded by each labeling terminal.
  • the patent enables users to collaboratively mark medical images at any time and place, pure manual labeling relies entirely on the expertise of the labeling operator, and the quality of the label cannot be guaranteed. The manual labeling 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.
  • the present invention provides a medical image annotation method, the method comprising:
  • the label of the medical image to be labeled is confirmed based on the comparison result of the label contents of at least two labeling operators.
  • the method further comprises:
  • the medical image to be labeled is assigned to at least two labeling operators to independently mark, and the labeling result is confirmed based on the intersection of at least two labeling labels marked by the operator.
  • the method further comprises:
  • the medical images to be labeled are assigned to at least two labeling operators to be independently labeled, and the result of the labeling is confirmed based on the weights of the labels labeled by at least two labeling operators.
  • the method further comprises:
  • the method further comprises:
  • the medical image-related sentences are automatically found from the full text in the authoritative journals and books by using full-text indexing to generate at least two annotation operations based on the statements.
  • the label selected by the person.
  • the method further comprises:
  • the label is displayed in the form of a selectable button on the labeling terminal device labeling the operator.
  • the method further comprises:
  • the label is displayed on the terminal device labeled the operator in a manner combined with the medical image related statement.
  • the method further comprises:
  • the matching values of the keywords and tags are sorted, and at least two tags larger than the matching threshold are selected as the recommended selection tags.
  • the method further comprises:
  • the label order in the label queue is based on the change in label weight.
  • the tagging ability value of the public tagging personnel varies based on the tag order in the tag queue.
  • the number of tags added to the tag body library correspondingly increases the tagging ability value of the corresponding public tagging personnel.
  • a medical image annotation system characterized in that the annotation system comprises an import unit for importing a medical image, a first storage server storing a medical image and its keywords, a second storage server storing a label, a matching unit, to be labeled
  • the medical image and the corresponding selection label are assigned to at least two allocation units labeling the terminal, confirming at least two confirmation units labeling the result of the operator, and at least two labeling terminals;
  • the import unit imports and stores the medical image to be labeled into the first storage unit
  • the matching unit extracts a keyword of the medical image to be labeled stored by the first storage server and at least one tag stored by the second storage unit to match,
  • the allocating unit recommends at least two selection labels that can be selected by the labeling operator to the at least two labeling terminals based on the matching values of the keywords of the medical image to be labeled and the labels in the label body library;
  • the confirmation unit confirms the label of the medical image to be labeled based on the comparison result of the at least two labeled contents of the operator.
  • a multi-person collaborative semi-automatic medical image annotation method performs labeling operation by assigning a medical image to be labeled to at least two labeled terminals.
  • the labeling operation is performed based on at least two labels selected by the labeling operator of the labeling terminal recommended by the multi-person collaborative semi-automatic medical image system, and the multi-person collaborative semi-automatic medical image system pairs the at least two Labeling the terminal.
  • the labeling results independently performed by the performer are fused or compared to determine the labeling of the medical image to be labeled.
  • the multi-person collaborative semi-automatic medical image system determines the labeling of the medical image to be labeled by:
  • the label of the medical image to be labeled is an intersection of the labeling result of the at least two labeled terminals by the multi-person collaborative semi-automatic medical image system; and the medical image to be labeled that is empty is the a multi-person collaborative semi-automatic medical image system resending the to-be-labeled medical image to the at least two annotated terminals to perform an annotation operation until the multi-person collaborative semi-automatic medical image system determines an annotation label of the medical image to be labeled; or
  • the multi-person collaborative semi-automatic medical image system simultaneously displays the labeling results of the at least two labeled terminals to the labeling operator of the at least two labeled terminals.
  • the at least two labeling operators determine the label of the medical image to be labeled after negotiation.
  • the multi-person collaborative semi-automatic medical image system determines the labeling of the medical image to be labeled by:
  • the labeling results of the at least two labeled terminals are compared by the multi-person collaborative semi-automatic medical image system, and the at least two labels are marked by the multi-person collaborative semi-automatic medical image system in comparison with the labeling of the gaps in the results.
  • the labeling result of the terminal is simultaneously displayed to the labeling operator of the at least two labeling terminals, and the labeling label of the medical image to be labeled is determined by the at least two labeling operators.
  • the multi-person collaborative semi-automatic medical image system recommends at least two selectable labels for the labeling operator of the at least two labeled terminals by:
  • the multi-person collaborative semi-automatic medical image system automatically finds a sentence related to a medical image from the full text in the authoritative journal and the book by using a full-text index based on medical images derived from authoritative journals and books, based on the Corresponding statements to generate at least two labels that are selectable by the labeling operator, and the labels are displayed on the labeling terminal of the labeling operator and/or the label in the form of a selectable button to The way the related statements are combined is displayed on the labeling terminal of the labeling operator.
  • the automatically found statement related to the medical image extracts a keyword and matches the keyword with a tag in the tag body library, according to the keyword
  • the degree of matching with the tag generates at least two tags that can be selected by the labeling operator.
  • the at least two labeling operators of the labeling terminal mark the boundaries of the region of interest in the medical image to be labeled based on the selected label and the statement associated with the label.
  • the multi-person collaborative semi-automatic medical image annotation system assigns the medical image to be labeled to at least two labeled terminals by:
  • the medical image is stored on the first server
  • the annotation matching the medical image is stored in the second server
  • the matching relationship between the medical image and the label is stored as a data record in the On the second server; when the user extracts the annotated medical image, the first server and the second server respectively transmit data concurrently and are locally determined by the user according to the data record from the second server The tag is matched to the medical image and displayed locally.
  • the tagging terminal is a feature phone, a smartphone, a palmtop, a personal computer, a tablet or a personal digital assistant.
  • the method comprises the steps of:
  • the medical image to be labeled to at least two labeled terminals to perform an annotation operation
  • the multi-person collaborative semi-automatic medical image system is based on medical images derived from authoritative journals and books, and automatically finds medical document-related sentences from the full text in the authoritative journals and books by using full-text indexing, so as to be based on The related statement to generate at least two tags selectable by the labeling operator, and the system displays the tags to the at least two labeled terminals;
  • the labeling operator of the at least two labeling terminals independently completes the labeling of the medical image to be labeled based on the label recommended by the system;
  • the result of the labeling performed independently by the labeling operator of the at least two labeled terminals is fused or compared by the system to determine the labeling of the medical image to be labeled.
  • a multi-person collaborative semi-automated medical image annotation system wherein the system matches a text of a medical image with a label and/or article in a tag ontology library unit. , at least two matching tags and/or related statements are automatically recommended to at least two labeled terminals.
  • the label body library unit includes an labeled label unit and a medical journal and book data unit.
  • the labeling terminal operator introduces an unlabeled image into the system
  • the labeled text information is in the labeled
  • the tag unit and/or the medical journal and the book data unit perform a search and generate at least two tags and/or related statements based on the search result matching scores.
  • the generated at least two tags are displayed on the operator's tag terminal in the form of selectable buttons.
  • the generated at least two tags are displayed on the operator's tag terminal in a manner combined with the related statements.
  • the system further comprises a manual input tag unit, the manual input tag unit manually inputting an accurate tag by the operator based on the at least two tags and/or related statements.
  • the system further comprises an allocation unit and a comparison unit for assigning unlabeled medical images to at least two labeled terminal operators and independently of the at least two operators Completing the annotation;
  • the comparison unit is configured to compare and analyze the annotation results of the at least two operators, and if the comparison results are different for the same medical image, the comparison unit performs annotation of at least two operators after comparative analysis The result is simultaneously sent to the at least two operators, and the at least two operators negotiate to determine an accurate label.
  • the system further includes a high speed remote server unit and an annotated content server unit, wherein the medical image is stored in the high speed remote server unit, and a label matching the medical image is stored in the label content server unit.
  • a matching relationship data record between the medical image and the tag that matches the tag is stored in the tagged content server unit.
  • the operator matches the label with the medical image according to the matching relationship data record from the labeling content server and displays the label on the labeling terminal.
  • the annotated content server unit has an encryption system.
  • the system further comprises an import and export unit for importing the unlabeled medical image and for exporting the labeled medical image to a local file.
  • the present invention discloses a medical image annotation method, characterized in that the steps of the method include:
  • the medical image to be labeled is stored in a medical image database and is divided according to at least two unlabeled medical image subsets according to its description information, each subset comprising at least one medical image to be labeled.
  • the unlabeled medical image subset is divided according to the medical image description information to be labeled and based on the biological anatomy or the biophysical system or the biological anatomy and the biophysical system.
  • the medical image database comprises an unlabeled medical image database and an annotated medical image database, the labeled medical image database being divided into at least two labeled medical images based on the labeled or labeled information of the annotated image. Set, each subset includes at least one medical image to be labeled.
  • the labeled medical image subset is based on an annotated medical image Label or label information and divide it based on biological anatomy and/or biophysiology systems.
  • the selection tag comprises a selection tag generated based on matching keywords with tags in at least one tag ontology library and a medical image based on keywords and derived from authoritative journals and books, by using full-text indexing
  • the full text in the authoritative journals and books automatically finds statements related to keywords and medical images, and generates at least two labels that can be selected by the operator.
  • the generated at least two labels that can be selected by the labeling operator are displayed on the labeling terminal of the labeling operator in the form of a selectable button, and the statement related to the label is displayed in the label.
  • the labeling operator marks the boundary of the region of interest in the medical image to be labeled based on the selected label and the statement associated with the label.
  • the method further comprises: actively transmitting the unlabeled medical image to the at least one labeling terminal, comprising simultaneously transmitting the medical image to be labeled to at least two labeling operators, wherein the at least two labeling operators respectively Complete the annotations independently.
  • the at least two labels respectively marked by the operator are separately compared and compared; if the comparison result is very different, the label contents of both parties are simultaneously displayed to the two persons, and the two parties negotiate Determine the most accurate labeling label.
  • a medical system visualization device comprising: an image display unit, an image analysis unit, and an image labeling unit, wherein the visualization device is connected to a medical image annotation system,
  • the visualization device acts as an image labeling terminal and issues an image labeling request to the image labeling system
  • the image annotation system matches the keywords in the labeling request with the labels in the at least one label body library, and separately assigns the medical images to be labeled to the visualization device;
  • the image tagging system recommends at least one selection tag to the tagging operator of the visualization device based on the matching value of the keyword of the tag request and the matching value of the at least one tag;
  • the image annotation system retrieves, by the visualization device, a statement associated with the selection label in a full-text search manner and marks the display to the labeling operator;
  • the labeling operator completes the labeling of the medical image to be labeled in the image labeling portion based on the selection label displayed by the image display portion and the statement associated with the selection label, the image
  • the labeling unit sends the label content to the image labeling system
  • the image analysis unit records at least one annotation information labeling the operator and counts the intersection label of the same medical image.
  • the present invention can automatically recommend tags for users and support multiple users to work together. This is the most important function.
  • the present invention provides a user management function that allows an administrator to easily manage the information of the labeled user using the management tool and assign the image to be labeled to the caller.
  • the present invention provides a data import and export function.
  • the caller can log in to upload his or her own image, and the administrator is responsible for assigning the image.
  • users can also export specified data to local files. Local files support both .csv and .xml formats.
  • the present invention also supports viewing the articles associated with the images and automatically finding the sentences associated with the tags, highlighting them.
  • FIG. 1 is a schematic view of a preferred medical image annotation method of the present invention
  • FIG. 2 is a schematic diagram of a multi-person collaborative semi-automatic medical image annotation method according to the present invention
  • FIG. 3 is a schematic view of another preferred medical image annotation method of the present invention.
  • FIG. 4 is a schematic view of a preferred medical image annotation system of the present invention.
  • FIG. 5 is a schematic diagram of a multi-person collaborative semi-automatic medical image annotation system of the present invention.
  • Figure 6 is a schematic illustration of a medical system visualization device of the present invention.
  • Figure 7 is a schematic block diagram of a medical image annotation system of the present invention.
  • a medical picture also referred to as a medical image, refers to a picture or image of an internal tissue that is obtained non-invasively to an animal's body, the human body, or a part of the human body for medical or medical research.
  • the embodiment provides a medical image annotation method, which is characterized in that the method includes
  • the matching value of the keyword labeling the medical image and the label in the label body library is recommended to at least two labeling terminals to at least two labeling terminals, and the labeling operator of the at least two labeling terminals respectively treats the medical image to be labeled based on the labeling operator selecting the label Label it independently.
  • the matching value of the keyword and the at least one label is sorted, and at least two labels are selected according to a certain rule, and sent to the corresponding labeling terminal and displayed as a selection label that can be selected by the labeling operator.
  • At least one tag operator inputs a tag request at at least one tag terminal.
  • at least one medical image to be labeled is assigned to at least two labeling terminals, so that the labeling operator independently marks the labeling terminal.
  • the medical images to be labeled are assigned to at least two labeled terminals based on a preset order of priority of the labeled terminals.
  • the unlabeled medical image subset is assigned to at least two labeled terminals based on the terminal processing capability order of the labeling terminal.
  • the unlabeled medical image subset is assigned to at least two labeled terminals based on a load balancing principle.
  • the terminal processing capability sequence and the load balancing principle assign the unlabeled medical image subset to at least two labeled terminals.
  • the unlabeled medical image subset is assigned to a plurality of labeled terminals based on a preset set of terminal priority orders. For example, suppose there are three labeled terminals, namely labeled terminal 1, labeled terminal 2 and labeled terminal 3, and each labeled terminal can process 10 medical images. The labeling terminal 1 has the highest priority, the labeling terminal 2 has the lower priority, and the labeling terminal 3 has the lowest priority. Assuming that there are two unlabeled medical image subsets to be assigned, there are two unlabeled medical image subsets 1 and unlabeled medical image subsets 2, and each unlabeled medical image subset contains 10 images.
  • the task assignment result may be: sending 10 images in the unlabeled medical image subset 1 (or unlabeled medical image subset 2) to the labeling terminal 1 and transmitting the remaining unlabeled medical image subsets to the labeling terminal 2. 10 images are not sent to the tag terminal 3 without an unlabeled image.
  • the unlabeled medical image subset is assigned to the labeling terminal based on the terminal processing capability order of the labeling terminal. For example, suppose there are three labeled terminals, namely labeled terminal 1, labeled terminal 2, and labeled terminal 3. The number of medical images that can be processed by the labeling terminal 1 is ten, the number of medical images that the labeling terminal 2 can process is ten, and the number of medical images that the labeling terminal 3 can process is five. Assuming that there are two unlabeled medical image subsets to be assigned, there are two unlabeled medical image subsets 1 and unlabeled medical image subsets 2, and each unlabeled medical image subset contains 10 images.
  • the task assignment result may be: sending the unlabeled medical image to the labeling terminal 1
  • Ten images in episode 1 (or unlabeled medical image subset 2) are sent to the tag terminal 2 for 10 images in the remaining unlabeled medical image subset, without sending unlabeled images to the tag terminal 3.
  • the keyword is extracted from the descriptive text of the medical image to be labeled, and the keyword is matched with at least one tag in the tag body library. Record the match value of the keyword with at least one tag. At least two tags whose matching value is greater than the preset matching threshold are recommended to the corresponding labeling terminal for the labeling operator to select.
  • the label is displayed in the form of a selectable button on the labeling terminal of the labeling operator. Alternatively, the label is displayed on the labeling terminal of the labeling operator in a manner combined with the medical image related statement
  • statements related to medical images are automatically found from the full text of authoritative journals and books by using full-text indexing.
  • the relevant statement generates a label that can be selected by the labeling operator and is selected by the labeling operator for labeling.
  • the labeling terminal respectively displays the labeling contents or the selected labels of the labeling operators. Ask the operator to re-mark the labeled medical image for labeling.
  • a communication connection or an instant communication connection is established for at least two labeling operators that mark the same medical image. The final label is confirmed by negotiation by at least two labeling operators.
  • the labeling result is confirmed based on at least two weights of the label selected by the operator.
  • the resulting label is thus obtained.
  • the embodiment provides a medical image annotation method, which comprises: recommending at least two selection labels to at least two labeling terminals based on a matching value of a keyword of the medical image to be labeled and a label in the label body library, At least two labeling operators of the labeling terminals independently mark the medical images to be labeled based on the labeling operator selection label.
  • the image similarity of each labeled medical image and the medical image to be labeled in the database is separately calculated.
  • the image similarity is mainly calculated for the similarity degree of the image content of the two pictures, and an image similarity value is obtained.
  • the image similarity can be calculated by the visual features of the two images.
  • the visual features can be RGB (Red Green Blue) features, texture features and histogram features and SIFT (Scale-invariant feature transform). ) Features, etc.
  • the labeled medical image having a similarity to the image of the medical image to be labeled is greater than the first threshold is selected to form a group of pictures.
  • the first threshold is a preset similarity value. Specifically, the first threshold can be set by the labeling operator. In contrast, the higher the first threshold is set, the more similar the labeled image found in the database is to the image to be labeled, but the number of labeled images found will be relatively small.
  • the labels corresponding to each labeled image in the extracted medical image group constitute a tag phrase.
  • the corresponding keywords constitute a keyword group. Extract keywords and tags based on the mapping of the annotated images. If an annotated image with a label printed in the database is stored, the keyword of the descriptive text of the annotated image is first identified, and then the label is extracted.
  • At least one tag in the output tag phrase is used as the selection tag of the picture to be labeled. Since there may be many tags in the tag phrase, the user may not want to output too many tags, and only wants to output a preset number of tags, so it can be implemented by determining whether the number of tags in the tag phrase is greater than The third threshold. When the number of labels in the tag phrase is greater than the third threshold, the preset number of tags in the tag phrase is output as the selection tag of the to-be-labeled picture, and the preset number is less than or equal to the third threshold. Specifically, the preset number and the third threshold are labels indicating the number of labels set by the operator.
  • the corresponding at least one tag is displayed on the tag terminal in the form of a selectable button.
  • the method for extracting a keyword from a description text of a medical image to be annotated includes segmenting the text information to obtain at least one word segment, and acquiring semantic content and semantic type of at least one word segment.
  • Semantic content is the semantic information with meaning corresponding to the participle.
  • the semantic type is the type of semantic information, for example, the part of speech of the participle and the meaning of the participle.
  • At least one word segment is filtered according to the semantic content and the semantic type in the corresponding keyword group to filter out keywords related to the medical image to be labeled.
  • the labeled medical image corresponds to multiple keywords.
  • the keyword with the highest degree of similarity between the word and the description word in the keyword group is selected as the keyword of the medical image to be labeled.
  • Semantic similarity is mainly to calculate the similarity degree of semantics of two words, and obtain a semantic similarity value. The higher the semantic similarity value, the more similar the semantics of these two words are.
  • the second threshold may specifically be a semantic similarity value preset by the user.
  • At least one tag is obtained according to the mapping relationship between the keyword and the tag phrase.
  • the tag can be a related statement related to the medical image to be labeled.
  • Related statements are obtained by full-text indexing of authoritative journals or books. If the medical image has been labeled from authoritative journals and books, the articles of authoritative journals and books are established as related sentence groups. For the descriptive text of the medical image to be labeled, the document related to the medical image is automatically found as the recommended label from the full text in the authoritative journal and the book by using the full-text index.
  • the relevant statement is displayed in the form of a selectable button on the labeling terminal of the labeling operator.
  • This embodiment provides a multi-person collaborative semi-automatic medical image annotation method.
  • the labeling operation by assigning the medical image to be labeled to at least two labeling terminals, the labeling operation is performed based on at least two labels selected by the labeling operator for marking the terminal recommended by the multi-person collaborative semi-automatic medical image system, and
  • the multi-person collaborative semi-automatic medical image system fuses or compares the annotation results independently performed by the labeling operators of at least two labeled terminals to determine the labeling labels of the medical images to be labeled.
  • the labeling operator issues a labeling request at the labeling terminal.
  • at least one medical image to be labeled is assigned to at least two labeling terminals.
  • the method for allocating includes: assigning the labeling operator to the labeled medical image to the labeling operator at least two labeling terminals based on the priority order of the labeling terminals preset; or, based on the labeling operator, the terminal processing capability sequence of at least two labeled terminals
  • the labeling operator assigns the labeled medical image to the labeling operator at least two labeled terminals; or, based on the load balancing principle, assigns the labeling operator the labeled medical image to the labeling operator at least two labeled terminals.
  • the use of full-text indexing automatically finds statements related to medical images from the full text of the authoritative journals and books in the operator, so as to generate at least two based on the statements associated with the operator.
  • Labels that can be labeled by the operator, and the labeling operator labels are displayed in the form of buttons that can be selected in the labeling operator.
  • the operator's labeling terminal and/or the labeling operator label are displayed on the labeling operator's labeling terminal in a manner combined with the labeling operator related statement.
  • the label of the medical image to be labeled is the intersection of the results of the multi-person collaborative semi-automatic medical image system for the annotation of at least two labeled terminals.
  • the medical image to be labeled is re-issued to at least two labeled terminals to perform the labeling operation by the multi-person collaborative semi-automatic medical image system until the multi-person collaborative semi-automatic medical image system determines the annotation of the medical image to be labeled. label.
  • the multi-person collaborative semi-automatic medical image system simultaneously displays the labeling results of at least two labeled terminals to at least two labeled operators of the labeled terminal, and is negotiated by at least two labeled operators. The label of the medical image to be labeled is then determined.
  • the labeling results of the at least two labeled terminals are compared by a multi-person collaborative semi-automatic medical image system, and the annotation results of the gaps in the results are compared, and the labeling results of at least two labeled terminals are simultaneously displayed by the multi-person collaborative semi-automatic medical image system.
  • the labeling operator of at least two labeling terminals is determined by at least two labeling operators to determine the labeling label of the medical image to be labeled.
  • At least two labeling operators of the labeling terminal mark the boundaries of the region of interest (ROI) in the medical image to be labeled by the operator indicia based on the selected label and the statement associated with the labeling operator label.
  • ROI region of interest
  • the medical image is stored on the first server.
  • a tag that matches the medical image of the operator is stored on the second server.
  • the matching relationship between the medical image of the operator and the label of the operator is marked as a data record stored on the second server of the labeling operator.
  • the operator label is marked by the labeling operator first server and the labeling operator second server respectively, and the user locally marks the data according to the data record from the labeling operator second server. Match the operator's medical image and display it locally.
  • This embodiment provides a medical image annotation method.
  • the steps of the method include:
  • a keyword of the description information of the medical image to be labeled is extracted. Match the label operator keyword to the label in at least one label ontology library. An unlabeled medical image is individually assigned to at least one of the labeled terminals. At least one selection tag is recommended to the tagging operator of the corresponding tag terminal based on the keyword and the matching value based on the at least one tag. The statement associated with the medical image/or selection tag is retrieved in a full-text search and the display is marked to the labeling operator. Record at least one annotation information labeling the operator and count the intersection of the same medical image.
  • the steps of a medical image annotation method include:
  • S01 Extract a keyword of the description information of the medical image to be labeled in response to the request of the at least one labeling terminal.
  • At least one labeling operator issues a labeling request at the labeling terminal.
  • the keyword in the description information of the medical image to be labeled is extracted in response to the at least one request to mark the terminal.
  • the medical image to be labeled is accompanied by a description text containing keywords. Extract the keywords in the description text.
  • S03 recommend at least one selection label to the labeling operator of the corresponding labeling terminal based on the keyword and the matching value based on the at least one label.
  • At least one selection label that can be selected by the labeling operator is recommended to the labeling operator of the corresponding labeling terminal.
  • the unlabeled medical image is actively sent to the at least one labeling terminal, including simultaneously transmitting the medical image to be labeled to at least two labeling operators. At least two labeling operators are marked separately by the labeling operator.
  • the selection label for the labeling operator is displayed at the same time as the corresponding medical image to be labeled. At the same time, the label terminal also displays a manual label input field. When the labeling operator is not satisfied with the displayed selection label, you can enter the manual label in the manual label input field.
  • S04 retrieve the statement associated with the medical image/or selection tag in a full-text search manner and mark the display to the labeling operator.
  • full-text indexing Based on medical images from authoritative journals and books, the use of full-text indexing automatically finds statements related to medical images from the full text of the authoritative journals and books in the operator, in order to generate at least two based on the annotation operator statements. A label for the operator to select.
  • S05 Record at least one annotation information labeling the operator and count the intersection of the same medical image.
  • the medical image to be labeled is stored in the medical image database, and is divided into at least two unlabeled medical image subsets according to the medical image description information to be labeled, each subset including at least one medical image to be labeled.
  • the medical image database stores an unlabeled medical image set and an annotated medical image set.
  • the annotated medical image set contains medical images that have been labeled.
  • Unlabeled medical image sets contain 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 expanded correspondingly as the number of medical images increases.
  • the unlabeled medical image set After receiving the task of annotating the unlabeled medical image set in the medical image database, the unlabeled medical image set may be divided into a plurality (at least two) of unlabeled medical image subsets, each unlabeled medical image subset One or more medical images may be included.
  • the unlabeled medical image subset is divided based on the biological (e.g., human) anatomy.
  • the unlabeled medical image set can be specifically divided into: a subset of images such as a brain, a chest, a heart, an abdomen, an upper limb, and a lower limb.
  • the unlabeled medical image subset is divided according to the biophysiological system structure.
  • the unlabeled medical image set is specifically divided into a subset of images of the digestive system, the nervous system, the exercise system, the endocrine system, the urinary system, the reproductive system, the circulatory system, the respiratory system, and the immune system.
  • the unlabeled medical image set it is also possible to make a multi-level refinement of the unlabeled medical image set.
  • it can also be divided into subsets of images such as the Central Core, the Limbic System, and the Cerebral Cortex.
  • the associated cardiac image is divided into a subset of images of the aorta, left atrium, left ventricle, right atrium, right ventricle, and the like.
  • an identifier may be assigned to each subset of medical images in order to facilitate subsequent combining of the subset of medical images. All medical images in the same medical image subset Share the same identifier.
  • the annotated medical image database is divided into at least two labeled medical image subsets based on the tagged or labeled information of the annotated image, each subset comprising at least one medical image to be labeled.
  • the subset of medical images has been labeled according to the label or annotation information of the labeled medical image and is based on or based on the biological anatomy and/or biophysiological system.
  • the labeling operator labeling system includes an import unit for importing a medical image, a first storage server storing a medical image, a second storage server storing a label and/or a medical image related sentence, a matching unit, a medicine to be labeled
  • the image and the corresponding selection label are assigned to at least two allocation units labeling the terminal, confirming at least two confirmation units labeling the operator's labeling results, and at least two labeling terminals.
  • the labeling operator introduction unit imports and stores the medical image generated by the visual medical image device to the first storage unit of the labeling operator.
  • the labeling operator first storage server divides the medical image into at least two medical image subsets and classifies the medical image and its keyword information.
  • the annotation operator matching unit extracts a keyword labeling the medical image to be labeled stored by the operator first storage server and at least one label and/or medical image related statement stored by the operator second storage unit and performs local matching and calculation matching A value that assigns at least one tag and/or medical image related statement that meets the criteria as a selection tag to the corresponding tag terminal.
  • the labeling operator assigning unit assigns the medical image to be labeled to at least two labeling terminals based on the preset labeling terminal priority order, or the labeling operator assigning unit assigns the medical image to be labeled based on the terminal processing capability sequence/based on the load balancing principle Give at least two labeled terminals.
  • the labeling operator confirmation unit confirms the labeling result of the medical image to be labeled based on the labeling contents of the at least two labeling operators.
  • the human collaborative semi-automated medical image annotation system includes at least a tag body library unit that stores tags and/or articles/related statements related to medical images, a manual input tag unit, an allocation unit, a comparison unit, a high speed remote server unit, and an annotated content server unit.
  • the multi-person collaborative semi-automatic medical image annotation system matches the text of the medical image with the tags and/or articles in the tag ontology library unit, and automatically matches the matched at least two tags and/or related statements to at least two tagged terminals.
  • the tag ontology library unit includes the tagged tag unit and the medical journal and the book data unit.
  • the tag operator of the tag terminal introduces the unlabeled image into the system
  • the tagged unit and/or the medical journal and the book data are marked according to the image text information.
  • the unit performs a search and generates at least two tags and/or related statements based on the search result matching scores.
  • the generated at least two labels are displayed in the form of selectable buttons on the labeling terminal of the labeling operator.
  • the generated at least two tags are displayed on the tagging terminal of the tag operator in a manner combined with the related statements.
  • the manual input label unit manually enters the accurate label by the labeling operator based on at least two labels and/or related statements.
  • the allocation unit is configured to assign an unlabeled medical image to at least two labeled terminal operators, and the annotation is performed independently by at least two operators.
  • the comparison unit is configured to compare and analyze the annotation results of at least two operators. If the comparison results are different for the same medical image, the comparison unit sends the annotation results of at least two operators to at least two operators simultaneously after comparative analysis. The exact label is determined by at least two operators.
  • the medical image is stored in a high speed remote server unit.
  • a tag that matches the medical image is stored in the tagged content server unit.
  • the matching relationship data record between the medical image and the matching tag is stored in the tag content server unit.
  • the high-speed remote server unit and the annotated content server unit respectively send relevant data from different locations and display them on the labeling terminal, and the labeling operator records according to the matching relationship data from the labeling content server.
  • the tag is matched to the medical image and displayed at the tag terminal.
  • the tagged content server unit has an encryption system.
  • the multi-person collaborative semi-automated medical image annotation system further comprises an import and export unit for importing unlabeled medical images and for exporting the annotated medical images to generate local files.
  • the visualization device includes an imaging portion, an image display portion, an image analysis portion, and an image labeling portion.
  • the image labeling unit completes the image annotation based on the image generated by the imaging unit in combination with the biological anatomy or the biophysical system.
  • the visualization device acts as an image annotation terminal and issues an image annotation request to the image annotation system.
  • the image annotation system matches the keywords in the at least one tag body library based on the keywords tagged by the operator tagging request, and separately assigns the medical images to be tagged to the tag operator visualization device.
  • the imaging unit converts the image information transmitted by the medical image annotation system into a medical image and presents it to the labeling operator at the image display portion.
  • the image annotation system recommends at least one selection label to the annotation operator of the visualization device based on the matching value of the keyword labeled with the operator's annotation request and the at least one label.
  • the image annotation system retrieves the statement associated with the selection label in a full-text search by annotating the operator visualization device and marks the display to the labeling operator.
  • the labeling operator marks the operator based on the labeling operator's selection label displayed by the operator's image display part and the statement associated with the labeling operator's selection label to complete the marking of the medical image annotation to be marked by the operator and sent to the labeling operator image labeling system.
  • the labeling operator image analysis unit can perform processing analysis on the labeling content marked by the labeling operator's imaging labeling unit in combination with the labeled image in the operator image labeling system.
  • the labeling operator image analysis unit receives the medical image stored on the high-speed remote server through the network, the labeling content matched with the labeling operator medical image stored on another labeling content server, marking the operator labeling content and labeling the operator medical The matching relationship code of the image.
  • the labeling operator image labeling unit locally marks the operator's medical image according to the received medical image of the operator and the labeling content matching the medical image of the operator and the matching relationship code between the labeling operator and the medical image of the operator.
  • the medical image annotation system of the present invention MITagger adopts B/S architecture, and then Developed in Python+Django, the platform uses the standard Django MVC framework and provides a uniform HTTP interface. All services on the front end are called by the HTTP interface provided by the backend server.
  • the background basically adopts a layered and modular architecture, and the system is divided into several levels, each level consists of a certain number of modules.
  • the Http Interface encapsulates all functions as interfaces for use by the front end.
  • Request Auth is responsible for verifying the identity of the requested user. Most services can only be accessed by legitimate users.
  • the Network Service completes the processing of each function.
  • the Tag Recommdation layer encapsulates a tag recommendation engine based on the entity library. Data Storage is based on the Django Model and Elasticsearch interfaces and is responsible for data storage management.
  • This embodiment is an improvement over any of the foregoing embodiments.
  • the rediscovery of the medical picture related knowledge and the newly discovered information are updated by means of crowdsourcing.
  • the annotation information is opened to the public user.
  • Public users become public labeling personnel after registering personal information in the medical image annotation system, and can mark medical images.
  • the labeling content of the public labeling person is saved and added to the label queue of the corresponding medical picture.
  • the labels in the label queue include the initial label and the label generated by the public labeler.
  • the labeling content is weighted based on the qualifications and labeling history of the publicly marked personnel. Under the influence of weights, the labels marked by multiple public labelers present a dynamically ordered label queue.
  • the present invention incorporates new knowledge related to medical pictures into the tag ontology library, so that the tag ontology library is continuously updated. For example, if the preset order threshold is 5, the order of the label queues will be added to the label body library after the top five labels are verified by the experts.
  • the value of the tagging ability of the corresponding public tagging person is evaluated based on the dynamic change of the tag in the tag queue. If the order of the labels changes continuously, the value of the labeling ability of the public labeling personnel will increase. If the order of the labels changes continuously, the value of the labeling ability of the public labeling personnel will decrease.
  • the value of the tagging ability of the public tagger who is included in the tag body library of at least one tag is correspondingly increased.
  • the tag corresponds to the public tag
  • the value of the person's marking ability will increase.
  • the added score is set by the manager. The more tags that are included in the tag entity library, the more the tagging ability value of the public tagger increases.
  • the tagging terminal is any entity having a computing processing capability, such as a feature phone, a smart phone, a palmtop computer, a personal computer, a tablet computer, or a personal digital assistant.
  • the tagging terminal also has a communication function with the network so that the medical image to be labeled provided by the medical image tagging system can be received over the network and the tagging information can be returned to the medical image tagging system.
  • the labeling operator can also view the marked medical images by labeling the terminal.
  • the annotation function can be implemented by installing a plug-in in a browser of the smart processing device, and the browser can include, for example, Internet Explorer, Firefox, Safari, Opera, Google Chrome, GreenBrowser, and the like. Label terminals can be distributed among different geographical areas.
  • 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 file system and allow users to interact with files, which can be common at present.
  • Medical images of the invention include medical images.

Landscapes

  • Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Theoretical Computer Science (AREA)
  • General Health & Medical Sciences (AREA)
  • Public Health (AREA)
  • Primary Health Care (AREA)
  • Medical Informatics (AREA)
  • Epidemiology (AREA)
  • Radiology & Medical Imaging (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Physics & Mathematics (AREA)
  • Databases & Information Systems (AREA)
  • Data Mining & Analysis (AREA)
  • Library & Information Science (AREA)
  • Computational Linguistics (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Artificial Intelligence (AREA)
  • Medical Treatment And Welfare Office Work (AREA)
PCT/CN2015/098710 2015-12-17 2015-12-24 一种医学图像标注方法及*** WO2017101142A1 (zh)

Priority Applications (4)

Application Number Priority Date Filing Date Title
DE212015000240.3U DE212015000240U1 (de) 2015-12-17 2015-12-24 System zur medizinischen Bildannotation
CN202210164256.0A CN114297429A (zh) 2015-12-17 2015-12-24 一种医学图像标签的半自动标注***及方法
CN201580084598.XA CN108463814B (zh) 2015-12-17 2015-12-24 一种医学图像标注方法及***
CN202210115900.5A CN114398511A (zh) 2015-12-17 2015-12-24 一种医疗***可视化设备及其标签标注方法

Applications Claiming Priority (8)

Application Number Priority Date Filing Date Title
CN201510937924.9 2015-12-17
CN201510937922.X 2015-12-17
CN201510937924 2015-12-17
CN201510937921 2015-12-17
CN201510937925 2015-12-17
CN201510937925.3 2015-12-17
CN201510937921.5 2015-12-17
CN201510937922 2015-12-17

Publications (1)

Publication Number Publication Date
WO2017101142A1 true WO2017101142A1 (zh) 2017-06-22

Family

ID=59055552

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2015/098710 WO2017101142A1 (zh) 2015-12-17 2015-12-24 一种医学图像标注方法及***

Country Status (3)

Country Link
CN (3) CN114398511A (de)
DE (1) DE212015000240U1 (de)
WO (1) WO2017101142A1 (de)

Cited By (17)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107516005A (zh) * 2017-07-14 2017-12-26 上海交通大学 一种数字病理图像标注的方法和***
CN109446370A (zh) * 2018-10-26 2019-03-08 广州金域医学检验中心有限公司 医学图像的病理标注方法及装置、计算机可读存储介质
CN109461147A (zh) * 2018-10-26 2019-03-12 广州金域医学检验中心有限公司 应用于移动终端的fov图片的病理标注方法及装置
CN109523535A (zh) * 2018-11-15 2019-03-26 首都医科大学附属北京友谊医院 一种病变图像的预处理方法
CN109544526A (zh) * 2018-11-15 2019-03-29 首都医科大学附属北京友谊医院 一种慢性萎缩性胃炎图像识别***、装置和方法
CN109584218A (zh) * 2018-11-15 2019-04-05 首都医科大学附属北京友谊医院 一种胃癌图像识别模型的构建方法及其应用
CN109671053A (zh) * 2018-11-15 2019-04-23 首都医科大学附属北京友谊医院 一种胃癌图像识别***、装置及其应用
CN109670530A (zh) * 2018-11-15 2019-04-23 首都医科大学附属北京友谊医院 一种萎缩性胃炎图像识别模型的构建方法及其应用
CN109686423A (zh) * 2018-11-06 2019-04-26 众安信息技术服务有限公司 一种医疗影像标注方法及***
CN110752015A (zh) * 2018-07-24 2020-02-04 由昉信息科技（上海）有限公司 一种运用于医学领域的智能分类与标记***及方法
CN110796011A (zh) * 2019-09-29 2020-02-14 湖北工程学院 一种基于深度学习的稻穗识别方法、***、装置及介质
CN110858327A (zh) * 2018-08-24 2020-03-03 宏达国际电子股份有限公司 验证训练数据的方法、训练***以及计算机程序产品
CN112418263A (zh) * 2020-10-10 2021-02-26 上海鹰瞳医疗科技有限公司 一种医学图像病灶分割标注方法及***
CN112951353A (zh) * 2019-11-26 2021-06-11 广州知汇云科技有限公司 一种医疗病历标注平台及其操作方法
US11094411B2 (en) 2018-10-26 2021-08-17 Guangzhou Kingmed Center for Clinical Laboratory Co., Ltd. Methods and devices for pathologically labeling medical images, methods and devices for issuing reports based on medical images, and computer-readable storage media
US11501165B2 (en) 2020-03-04 2022-11-15 International Business Machines Corporation Contrastive neural network training in an active learning environment
US11562587B2 (en) * 2017-12-04 2023-01-24 Merative Us L.P. Systems and user interfaces for enhancement of data utilized in machine-learning based medical image review

Families Citing this family (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109461495B (zh) * 2018-11-01 2023-04-14 腾讯科技（深圳）有限公司 一种医学图像的识别方法、模型训练的方法及服务器
CN110096480A (zh) * 2019-03-28 2019-08-06 厦门快商通信息咨询有限公司 一种文本标注***、方法及存储介质
CN110222709B (zh) * 2019-04-29 2022-01-25 上海暖哇科技有限公司 一种多标签智能打标方法及***
CN110751629A (zh) * 2019-09-29 2020-02-04 中国科学院深圳先进技术研究院 一种心肌影像分析装置及设备
CN111062255A (zh) * 2019-11-18 2020-04-24 苏州智加科技有限公司 三维点云的标注方法、装置、设备及存储介质
CN111340131B (zh) * 2020-03-09 2023-07-14 北京字节跳动网络技术有限公司 图像的标注方法、装置、可读介质和电子设备
CN111784284B (zh) * 2020-06-15 2023-09-22 杭州思柏信息技术有限公司 一种宫颈图像多人协同标记云服务***及云服务方法
CN112989087B (zh) * 2021-01-26 2023-01-31 腾讯科技（深圳）有限公司 一种图像处理方法、设备以及计算机可读存储介质
CN112860416A (zh) * 2021-04-25 2021-05-28 城云科技（中国）有限公司 标注任务分派策略方法、装置
CN113326890B (zh) * 2021-06-17 2023-07-28 北京百度网讯科技有限公司 标注数据处理方法、相关装置及计算机程序产品
CN113592981B (zh) * 2021-07-01 2022-10-11 北京百度网讯科技有限公司 图片标注方法、装置、电子设备和存储介质
CN113571162B (zh) * 2021-07-19 2024-02-06 蓝网科技股份有限公司 一种多用户协同操作医学影像的实现方法、装置及***
CN115795076B (zh) * 2023-01-09 2023-07-14 北京阿丘科技有限公司 图像数据的交叉标注方法、装置、设备及存储介质

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101467145A (zh) * 2006-07-24 2009-06-24 谷歌公司 用于自动注释图像的方法和装置
US20100171682A1 (en) * 2006-04-14 2010-07-08 Konica Minolta Medical & Graphic, Inc. Medical image display apparatus and computer readable medium
CN104021222A (zh) * 2014-06-26 2014-09-03 深圳信息职业技术学院 一种基于隐形狄利克雷模型的生物医学图像的标注算法
CN104239359A (zh) * 2013-06-24 2014-12-24 富士通株式会社 基于多模态的图像标注装置以及方法
CN104462738A (zh) * 2013-09-24 2015-03-25 西门子公司 一种标注医学图像的方法、装置和***

Family Cites Families (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2004051516A1 (en) * 2002-11-27 2004-06-17 Amirsys, Inc. An electronic clinical reference and education system and method of use
CN101765843A (zh) * 2007-08-01 2010-06-30 皇家飞利浦电子股份有限公司 使用医学相关术语访问医学图像数据库
US8600771B2 (en) * 2007-11-21 2013-12-03 General Electric Company Systems and methods for generating a teaching file message
US8774515B2 (en) * 2011-04-20 2014-07-08 Xerox Corporation Learning structured prediction models for interactive image labeling
WO2012177192A1 (en) * 2011-06-21 2012-12-27 Telefonaktiebolaget L M Ericsson (Publ) Electronic device and method for handling tags
CN102306298B (zh) * 2011-07-19 2012-12-12 北京航空航天大学 一种基于Wiki的图像分类体系动态演化方法
IN2014CN03530A (de) * 2011-11-08 2015-07-03 Vidinoti Sa
US9747600B2 (en) * 2012-03-30 2017-08-29 United State Poastal Service Item status tracking
CN103020293B (zh) * 2012-12-28 2019-02-05 百度在线网络技术（北京）有限公司 一种移动应用的本体库的构建方法及***
CN104572735B (zh) * 2013-10-23 2018-02-23 华为技术有限公司 一种图片标注词推荐方法及装置
CN104809113B (zh) * 2014-01-23 2019-08-09 腾讯科技（深圳）有限公司 网页信息的显示方法及装置
US9619470B2 (en) * 2014-02-04 2017-04-11 Google Inc. Adaptive music and video recommendations
CN105094760B (zh) * 2014-04-28 2019-10-29 小米科技有限责任公司 一种图片标记方法及装置
CN105118068B (zh) * 2015-09-29 2017-12-05 常熟理工学院 一种小样本条件下医学图像自动标注方法

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20100171682A1 (en) * 2006-04-14 2010-07-08 Konica Minolta Medical & Graphic, Inc. Medical image display apparatus and computer readable medium
CN101467145A (zh) * 2006-07-24 2009-06-24 谷歌公司 用于自动注释图像的方法和装置
CN104239359A (zh) * 2013-06-24 2014-12-24 富士通株式会社 基于多模态的图像标注装置以及方法
CN104462738A (zh) * 2013-09-24 2015-03-25 西门子公司 一种标注医学图像的方法、装置和***
CN104021222A (zh) * 2014-06-26 2014-09-03 深圳信息职业技术学院 一种基于隐形狄利克雷模型的生物医学图像的标注算法

Cited By (19)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107516005A (zh) * 2017-07-14 2017-12-26 上海交通大学 一种数字病理图像标注的方法和***
US11562587B2 (en) * 2017-12-04 2023-01-24 Merative Us L.P. Systems and user interfaces for enhancement of data utilized in machine-learning based medical image review
CN110752015A (zh) * 2018-07-24 2020-02-04 由昉信息科技（上海）有限公司 一种运用于医学领域的智能分类与标记***及方法
CN110858327A (zh) * 2018-08-24 2020-03-03 宏达国际电子股份有限公司 验证训练数据的方法、训练***以及计算机程序产品
CN109446370A (zh) * 2018-10-26 2019-03-08 广州金域医学检验中心有限公司 医学图像的病理标注方法及装置、计算机可读存储介质
CN109461147A (zh) * 2018-10-26 2019-03-12 广州金域医学检验中心有限公司 应用于移动终端的fov图片的病理标注方法及装置
US11094411B2 (en) 2018-10-26 2021-08-17 Guangzhou Kingmed Center for Clinical Laboratory Co., Ltd. Methods and devices for pathologically labeling medical images, methods and devices for issuing reports based on medical images, and computer-readable storage media
CN109686423A (zh) * 2018-11-06 2019-04-26 众安信息技术服务有限公司 一种医疗影像标注方法及***
CN109544526A (zh) * 2018-11-15 2019-03-29 首都医科大学附属北京友谊医院 一种慢性萎缩性胃炎图像识别***、装置和方法
CN109670530A (zh) * 2018-11-15 2019-04-23 首都医科大学附属北京友谊医院 一种萎缩性胃炎图像识别模型的构建方法及其应用
CN109671053A (zh) * 2018-11-15 2019-04-23 首都医科大学附属北京友谊医院 一种胃癌图像识别***、装置及其应用
CN109584218A (zh) * 2018-11-15 2019-04-05 首都医科大学附属北京友谊医院 一种胃癌图像识别模型的构建方法及其应用
CN109523535A (zh) * 2018-11-15 2019-03-26 首都医科大学附属北京友谊医院 一种病变图像的预处理方法
CN109523535B (zh) * 2018-11-15 2023-11-17 首都医科大学附属北京友谊医院 一种病变图像的预处理方法
CN110796011A (zh) * 2019-09-29 2020-02-14 湖北工程学院 一种基于深度学习的稻穗识别方法、***、装置及介质
CN110796011B (zh) * 2019-09-29 2022-04-12 湖北工程学院 一种基于深度学习的稻穗识别方法、***、装置及介质
CN112951353A (zh) * 2019-11-26 2021-06-11 广州知汇云科技有限公司 一种医疗病历标注平台及其操作方法
US11501165B2 (en) 2020-03-04 2022-11-15 International Business Machines Corporation Contrastive neural network training in an active learning environment
CN112418263A (zh) * 2020-10-10 2021-02-26 上海鹰瞳医疗科技有限公司 一种医学图像病灶分割标注方法及***

Also Published As

Publication number Publication date
CN108463814A (zh) 2018-08-28
CN114297429A (zh) 2022-04-08
CN114398511A (zh) 2022-04-26
DE212015000240U1 (de) 2017-05-24
CN108463814B (zh) 2022-02-18
DE212015000240U8 (de) 2017-07-13

Similar Documents

Publication Publication Date Title
WO2017101142A1 (zh) 一种医学图像标注方法及***
CN104462738B (zh) 一种标注医学图像的方法、装置和***
Clough et al. The CLEF 2005 cross–language image retrieval track
KR101618735B1 (ko) 디지털 영상 컬렉션에 자동 얼굴 인식 기능을 통합하는 방법 및 장치
CN104750737B (zh) 一种相册管理方法及装置
CN102156715A (zh) 面向医学影像数据库的基于多病灶区域特征的检索***
Liu et al. Associating textual features with visual ones to improve affective image classification
US11521013B2 (en) Systems and methods for providing personalized product recommendations using deep learning
KR102258420B1 (ko) 지능정보기술기반 애니메이션 콘텐츠 리소스 서비스 시스템 및 방법
CN107679121B (zh) 分类体系的映射方法及装置、存储介质、计算设备
JP2015166978A (ja) 画像検索装置及び画像検索プログラム
Wajid et al. Neutrosophic-CNN-based image and text fusion for multimodal classification
US20140067349A1 (en) Clustering of personal electronically encoded items
CN113159186A (zh) 一种基于半监督学习的医学图像自动标注方法
CN117235362A (zh) 一种基于智慧文旅大数据的分析***
CN113972009A (zh) 一种基于临床检验医学大数据的医学检验咨询***
JP7255585B2 (ja) 情報処理装置、情報処理方法、および、プログラム
US9836548B2 (en) Migration of tags across entities in management of personal electronically encoded items
Abirami et al. CONTENT BASED IMAGE RETRIEVAL TECHNIQUES FOR RETRIEVAL OF MEDICAL IMAGES FROM LARGE MEDICAL DATASETS–A SURVEY
JP2015097036A (ja) 推薦画像提示装置及びプログラム
Shen et al. Photo selection for family album using deep neural networks
EP2704028A1 (de) Migration von Etiketten zwischen Entitäten bei der Verwaltung persönlicher elektronisch codierter Elemente
EP2704026A9 (de) Bündelung von persönlichen, elektronisch codierten Elementen
US20140067878A1 (en) Analysis and proposal creation for management of personal electronically encoded items
Zare et al. Comparative analysis of image retrieval approaches

Legal Events

Date Code Title Description
WWE Wipo information: entry into national phase

Ref document number: 212015000240

Country of ref document: DE

121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 15910611

Country of ref document: EP

Kind code of ref document: A1

122 Ep: pct application non-entry in european phase

Ref document number: 15910611

Country of ref document: EP

Kind code of ref document: A1