WO2021169101A1

WO2021169101A1 - Method and apparatus for generating medical image recognition model, computer device and medium

Info

Publication number: WO2021169101A1
Application number: PCT/CN2020/097964
Authority: WO
Inventors: 丁思雨; 丁文静
Original assignee: 平安国际智慧城市科技股份有限公司
Priority date: 2020-02-28
Filing date: 2020-06-24
Publication date: 2021-09-02
Also published as: CN111462865A

Abstract

A method for generating a medical image recognition model, which relates to the technical field of artificial intelligence. The method comprises: acquiring medical data; extracting a medical image in the medical data, and performing feature extraction on the medical image to obtain image features; extracting a medical file in the medical data, analyzing the medical file to generate annotation information, and searching within the annotation information for image annotation information corresponding to the medical image; establishing a mapping relationship according to the image features and the image annotation information, and generating an image annotation file according to the mapping relationship; and inputting the image annotation file into a machine learning model so as to learn the feature relationship between the image features in the image annotation file and the image annotation information by means of the machine learning model, and generating a medical image recognition model according to the feature relationship.

Description

Medical image recognition model generation method, device, computer equipment and medium

Cross-references to related applications

This application claims the priority of a Chinese patent application filed with the Chinese Patent Office on February 28, 2020, the application number is 2020101284650, and the application name is "Medical Image Recognition Model Generation Method, Apparatus, Computer Equipment, and Medium". The entire content of the application is approved The reference is incorporated in this application.

Technical field

This application relates to a method, device, computer equipment and medium for generating a medical image recognition model.

Background technique

With the rapid economic development and the improvement of living standards, people are paying more and more attention to health problems. Nowadays, the use of medical equipment to collect images of human organs has been widely used.

However, the inventor realizes that medical images taken through medical equipment must be recognized by medical professionals, and manual recognition cannot guarantee the accuracy of information recognition in medical images. Although there are medical records established based on medical documents in traditional technology. The file library, medical imaging information can be obtained by querying the medical file library, but the current methods for constructing the medical file library are mostly manual establishment, which makes the model establishment inefficient.

Summary of the invention

According to various embodiments disclosed in the present application, a method, device, computer device, and medium for generating a medical image recognition model are provided.

A method for generating a medical image recognition model, including:

Access to medical data;

Extracting medical images in the medical data, and performing feature extraction on the medical images to obtain image features;

Extracting medical documents in the medical data, analyzing the medical documents to generate annotation information, and searching for image annotation information corresponding to the medical image in the annotation information;

Establishing a mapping relationship according to the image characteristics and the image annotation information, and generating an image annotation file according to the mapping relationship; and

The image annotation file is input into a machine learning model to learn the feature relationship between the image features in the image annotation file and the image annotation information through the machine learning model, and a medical image recognition model is generated according to the feature relationship.

A medical image recognition model generation device, including:

Data acquisition module for acquiring medical data;

The feature extraction module is used to extract medical images in the medical data, and perform feature extraction on the medical images to obtain image features;

An annotation generation module, configured to extract medical documents in the medical data, analyze the medical documents to generate annotation information, and search for image annotation information corresponding to the medical image in the annotation information;

An image file generating module, configured to establish a mapping relationship according to the image characteristics and the image annotation information, and generate an image annotation file according to the mapping relationship; and

The model generation module is configured to input the image annotation file into a machine learning model to learn the feature relationship between the image feature in the image annotation file and the image annotation information through the machine learning model, and according to the feature relationship Generate medical image recognition model.

A computer device, including a memory and one or more processors, the memory stores computer readable instructions, and when the computer readable instructions are executed by the processor, the one or more processors execute The following steps:

Access to medical data;

One or more computer-readable storage media storing computer-readable instructions. When the computer-readable instructions are executed by one or more processors, the one or more processors perform the following steps:

Access to medical data;

The details of one or more embodiments of the present application are set forth in the following drawings and description. Other features and advantages of this application will become apparent from the description, drawings and claims.

Description of the drawings

In order to more clearly describe the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings needed in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present application. A person of ordinary skill in the art can obtain other drawings based on these drawings without creative work.

Fig. 1 is an application scenario diagram of a method for generating a medical image recognition model according to one or more embodiments.

Fig. 2 is a schematic flowchart of a method for generating a medical image recognition model according to one or more embodiments.

FIG. 3 is a schematic flowchart of a method for generating and using a medical image recognition model according to one or more embodiments.

Fig. 4 is a schematic flowchart of a method for generating a medical image recognition model in another embodiment.

Fig. 5 is a structural block diagram of a device for generating a medical image recognition model according to one or more embodiments.

Figure 6 is a block diagram of a computer device according to one or more embodiments.

Detailed ways

In order to make the technical solutions and advantages of the present application clearer, the following further describes the present application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application, and are not used to limit the present application.

The medical image recognition model generation method provided in this application can be applied to the application environment as shown in FIG. 1. The user terminal 102 communicates with the server 104 through the network. The server 104 obtains medical data; the server 104 extracts medical images in the medical data, and extracts features from the medical images to obtain image features; the server 104 extracts medical files in the medical data, analyzes the medical files to generate annotation information, and the server 104 searches for the annotation information The image annotation information corresponding to the medical image in the image; the server 104 establishes a mapping relationship according to the image characteristics and the image annotation information, and generates an image annotation file according to the mapping relationship; inputs the image annotation file into the machine learning model to learn the image annotation file through the machine learning model Based on the feature relationship between the image feature and the image annotation information, a medical image recognition model is generated according to the feature relationship. Furthermore, the medical image recognition model can also be pushed to the user terminal 102. The user terminal 102 may be, but is not limited to, various personal computers, notebook computers, smart phones, tablet computers, and portable wearable devices. The server 104 may be implemented by an independent server or a server cluster composed of multiple servers. When the server 104 is an independent server, multiple databases can be deployed in the server 104, and each database can store specific medical files; when the server 104 is a server cluster composed of multiple servers, the database deployed in each server Can store specific medical file data tables.

In one of the embodiments, as shown in FIG. 2, a schematic flow chart of a method for generating a medical image recognition model is provided. The method is applied to the server 104 in FIG. 1 as an example for description. In other embodiments, the The method can also be applied to the terminal, and the method includes the following steps:

Step 210: Obtain medical data.

Medical data includes at least medical images and medical documents generated during medical procedures. Specifically, the user terminal uploads medical data with the help of the client, and the server obtains the medical data uploaded by the user terminal to perform data analysis on the medical data.

Step 220: Extract medical images from the medical data, and perform feature extraction on the medical images to obtain image features.

Medical images can be human body structure diagrams collected by medical equipment. Medical images not only include the medical image itself, but may also have attribute information associated with the medical image. For example, the attribute information can be the label information, diagnosis description information, and annotations corresponding to the medical image. Auxiliary information of medical images such as result information. And medical images can also have multiple storage formats.

Specifically, the server may use an image feature extraction algorithm to extract image features corresponding to the medical image. For example, use the Histogram of Oriented Gradient (HOG) feature extraction algorithm, Local Binary Pattern (LBP) feature extraction algorithm, or Haar feature extraction algorithm to extract image features corresponding to medical images, such as image features It can be "raised", "dark tone" and so on.

Further, before the feature extraction of the medical image, it also includes: the server preprocesses the acquired medical image, such as performing enhancement processing or filtering processing on the medical image, so as to improve the accuracy of the feature extraction of the medical image by the server.

Step 230: Extract medical files in the medical data, analyze the medical files to generate annotation information, and search for image annotation information corresponding to the medical image in the annotation information.

Medical documents can be textual information generated during medical treatment, etc., and medical documents can have multiple formats, such as medical paper medical records or medical electronic medical record documents, etc., more specifically, medical electronic medical record documents can be excel, xml And other file formats. Specifically, the server extracts medical documents from the medical data, analyzes the medical documents to generate annotation information, and the annotation information may be diagnostic information given by medical professionals. For example, the medical file may contain labeling information for the medical image, and the labeling information may be the diagnostic information for the medical image given by the medical professional.

Specifically, the server may use keyword extraction technology to extract annotation information from medical documents, and when the acquired medical document is a medical paper medical record, the server also includes electronic processing of the paper medical record to obtain an electronic medical record document, Then the required annotation information is parsed from the electronic medical record document. Parsing refers to the process of extracting the required information from the medical file. The annotation information can be the cause of the disease, diagnosis suggestions, or diagnosis results. For example, the server can extract the annotation information of the medical image in the medical text according to the keyword matching technology, for example, the annotation information corresponding to the diagnosis result of the medical image may be "lumps", "shadows", and so on.

Step 240: Establish a mapping relationship according to the image characteristics and the image annotation information, and generate an image annotation file according to the mapping relationship.

The image features are associated with the image annotation information, a mapping relationship is established, and the image annotation file is obtained. For example, an image annotation file can be: shadow-pulmonary cyst, indicating that when the medical image feature extracted by the server is "shadow", the corresponding image annotation information should be "pulmonary cyst" at this time, so that the annotation information can be automatically obtained from the medical image Function.

Further, the server can also perform data processing on the acquired image annotation files, such as data cleaning, data management, data verification, etc., to make the acquired data more standard.

Step 250: Input the image annotation file into the machine learning model to learn the feature relationship between the image features in the image annotation file and the image annotation information through the machine learning model, and generate a medical image recognition model based on the feature relationship.

Taking pre-acquired image annotation files as a sample set, the server uses the sample set to train the machine learning model to obtain the association relationship between the image features and the image annotation information, and obtain the medical image recognition model according to the association relationship. Since the medical image recognition model stores the correspondence between image features and image annotation information, the medical image recognition model can be used to realize automatic acquisition of medical image features or image annotation information.

In this embodiment, the server extracts medical images and medical texts from medical data, extracts image features corresponding to medical images and annotation information in medical texts, realizes automatic extraction of medical information from medical data, and improves the processing of medical data. Efficiency and accuracy. According to the image features and annotation information, an image annotation file is established, and the image annotation file is used as a sample set to train the machine learning model to obtain a medical image recognition model. The medical image recognition model can then be used to automatically extract image annotation information, which improves The recognition accuracy and efficiency of medical data.

Moreover, in this embodiment, not only can the automatic processing of medical texts be realized, but also the automatic processing of medical images can be realized, which improves the ability to process medical data in different formats, and the medical images have more medical feature information. , Recognizing medical image data also makes the acquisition of medical data more comprehensive.

In one of the embodiments, establishing a mapping relationship based on image features and image annotation information, and generating an image annotation file based on the mapping relationship includes: obtaining an image identifier corresponding to the medical image; searching for an image analysis method corresponding to the image identifier; The attribute information associated with the medical image is analyzed to generate image label information; the mapping relationship is established according to the image characteristics, the image annotation information, and the image label information, and the image annotation file is generated according to the mapping relationship.

The image identifier is used to uniquely identify the category of the medical image, and the category of the medical image can be the storage format of the image. Different medical images can be stored in different medical image formats. The information contained in different formats of medical images is different, and the usage scenarios are different. In traditional technology, they are read and viewed in different formats through corresponding client tools. Medical imaging requires users to download multiple tools and know the instructions for use of the corresponding tools, which brings great inconvenience to users and also makes the efficiency of viewing medical images inefficient.

Specifically, different formats of medical images require different analysis methods to realize the analysis of medical images. The server extracts the medical image from the medical data, obtains the image identifier corresponding to the medical image, and calls the image analysis method associated with the image identifier to automatically analyze the attribute information associated with the medical image according to the image analysis method to generate image tag information. For example, the storage format can be DICOM, ndpi, tiff, jpg, bmp, png, etc. For example, for medical images in DICOM format, C++-based DCMTK, Java-based dcm4che, and python-based pydicom can be used for analysis. Another example is the use of dcm4che to analyze the tag information of DICOM files, and the use of opencv to analyze and process medical images such as pathology and fundus.

The image analysis method is used to extract and analyze the attribute information corresponding to the medical image to obtain the required image tag information. For example, the attribute information associated with a medical image can be encoded information, and the server parses and converts the encoded information according to the image analysis method found to obtain image tag information. For example, the parsed image tag information is: Tag. Tag.PatientName--patient name, Tag.PatientAge--patient age, Tag.PatientSex--patient gender, Tag.StudyID--examination ID, Tag.StudyDate--examination date, etc.

In this embodiment, by identifying the format of the medical image, and selecting the corresponding medical image analysis method according to different formats, different medical image analysis methods are used to identify medical images in different formats, which improves the recognition of medical images. The scope and adaptability to medical image formats have improved the processing capabilities of medical images. To solve the problem that the viewing process of different formats of medical images in the traditional technology needs to obtain other medical document auxiliary instructions, such as labeling results, diagnosis results, electronic medical records, etc., and the format and content of these auxiliary files are also very different, making users feel Storage, management, viewing, and comparison have brought great inconvenience.

In one of the embodiments, performing feature extraction on medical images to obtain image features includes: extracting annotation information in the medical image; obtaining location coordinates corresponding to the annotation information; extracting regional medical images corresponding to the location coordinates from the medical images; using Image feature extraction algorithm extracts image features of regional medical images.

The annotation information in the medical image may be information for identifying areas that need to be focused on in the medical image. Specifically, the annotation information may be parsed based on files such as xml and json generated by an internal annotation platform that has been desensitized.

The location coordinates can be the coordinates of the labeled information in the medical image, and the server can obtain the regional medical image corresponding to the area that needs to be focused on through the coordinates.

Specifically, the server may use HOG feature extraction algorithm, LBP feature extraction algorithm, Haar feature extraction algorithm, etc. to automatically extract image features corresponding to regional medical images. For example, the extracted image features can be "bulge" or "dark tone". The server can extract the labeling information of the regional medical image in the medical text according to the keyword matching technology, for example, the labeling information can be "lumps", "shadows", and so on.

In this embodiment, the regional medical image in the medical image is automatically extracted according to the coordinates in the medical image, and the key areas in the medical image and the regional features of the key areas are automatically extracted, so that the operation steps of the regional medical image can be realized. Improve the efficiency of medical image data processing.

In one of the embodiments, after parsing the medical file to generate the annotation information, it further includes: dividing the annotation information into unit annotation information according to preset rules; matching the unit annotation information with the information in the annotation information database, and matching The information with the highest degree is extracted as label information.

The labeling information generated by the server parsing the medical documents can be professionals, such as doctors’ diagnosis suggestions for patients. Since different doctors may have different expressions for the same disease, the labeling information obtained at this time may exist. Inconsistent textual representations of the same matter.

In order to standardize the labeling information, the server can use Chinese natural language processing word segmentation methods, such as HanLP for Chinese word segmentation, and segment and adjust the obtained labeling information according to certain rules to obtain multiple unit labeling information, and then label the unit The information is matched with words in an annotation information database such as a dictionary, and the information with the largest matching degree in the dictionary is extracted as annotation information. Since the information stored in the dictionary is standardized information, the use of dictionary matching technology can realize the standardized processing of labeling information.

For example, Chinese word segmentation methods include word segmentation methods based on string matching, word segmentation methods based on understanding, and word segmentation methods based on statistics.

In this embodiment, Chinese word segmentation is performed based on HanLP to generate labeling information, which realizes the standardization of labeling information and facilitates subsequent data retrieval.

In one of the embodiments, after generating the image annotation file according to the mapping relationship, the method further includes: receiving the query request sent by the terminal, extracting the query identifier carried in the query request; extracting the image identifier corresponding to the image annotation file; calculating the query identifier and each image The matching degree of the logo; the image annotation file associated with the image logo with the largest matching degree is pushed to the terminal.

Further, after the image annotation file is established according to each image feature and corresponding image annotation information, it also includes associating each image annotation file with the corresponding image identifier to generate an image annotation file library, that is, the image annotation file library stores different images Mark the corresponding image annotation file, such as "lung-lung image annotation file, stomach-stomach image annotation file, etc." are stored.

When the user terminal wants to find medical image diagnosis suggestions from the image annotation file library, the user terminal sends a query request to the server, and the server matches the corresponding image annotation file from the image annotation file library according to the obtained query request, and annotates the image The file is pushed to the user terminal, and then the user can perform operations based on the queried image annotation file. For example, users can match their own medical images with the medical images in the acquired image annotation files, and then query the annotation information of the medical images in the image annotation file library to obtain diagnosis suggestions for personal pathological conditions.

Specifically, the user terminal may send a query request to the server, for example, it may be an image query request. The server extracts the query identifier carried in the query request, for example, extracts the image query identifier carried in the image query request. Then the server matches the obtained query identifier with the image identifier corresponding to the image annotation file, and pushes the image annotation file associated with the successfully matched image identifier to the user terminal, so as to realize the function that the user can obtain the corresponding image annotation file through the query request .

For another example, if the user wants to obtain the image annotation file corresponding to a certain type of disease, he can obtain the corresponding file only according to the query request. Specifically, the user can use the obtained image annotation file as a reference file, and can refer to the medical image of his concern. Compare with the medical images in the acquired reference image annotation files, and then obtain the reference image features and reference image annotation information of the medical images that they are concerned about, and help users obtain professional medical image suggestions in time.

In this embodiment, the server obtains the user's query request, and then can push the corresponding image annotation file to the user, which realizes the automatic acquisition of the image annotation file and improves the efficiency of file acquisition.

As shown in Figure 3, a flow diagram of a method for generating and using a medical image recognition model is provided, including:

Step 310: Obtain medical data.

Step 320: Extract medical images from the medical data, and perform feature extraction on the medical images to obtain image features.

Step 330: Extract medical files in the medical data, analyze the medical files to generate annotation information, and search for image annotation information corresponding to the medical image in the annotation information.

Step 340: Establish a mapping relationship according to the image characteristics and the image annotation information, and generate an image annotation file according to the mapping relationship.

Step 350: Input the image annotation file into the machine learning model to learn the feature relationship between the image features in the image annotation file and the image annotation information through the machine learning model, and generate a medical image recognition model based on the feature relationship.

For specific embodiments of step 310 to step 350, reference may be made to the description of step 210 to step 250 in the foregoing, which is not repeated here.

In one of the embodiments, after generating the medical image recognition model, the method further includes:

Step 360: Receive the medical image sent by the terminal.

Specifically, the user terminal may send medical images to the server, and there is no restriction on the format of the medical images sent by the user terminal. The server receives the medical images sent by the user terminal.

Step 370: Input the medical image into the medical image recognition model to extract image features corresponding to the medical image through the medical image recognition model, and obtain image annotation information based on the medical image features.

Since the medical image recognition model has the function of recognizing medical image characteristics and medical image annotation information, the server recognizes the acquired medical images according to the trained medical image recognition model, and then obtains the image annotation information corresponding to the medical image, and realizes the medical image Automatic recognition.

Step 380: Push the image annotation information to the terminal.

The server pushes the image annotation information recognized according to the medical image recognition model to the user terminal.

In this embodiment, the trained medical image recognition model can be used to recognize medical images in any format, realize automatic recognition and automatic labeling of medical images, and use the medical image recognition model obtained by automatic training of big data. The recognition result is more accurate and the recognition efficiency is higher.

As shown in Fig. 4, a schematic diagram of the process of generating a medical image recognition model in another embodiment is provided, including:

Step 400: Obtain files in DICOM, ndpi, tiff, jpg, bmp, png and other formats.

Specifically, the server can obtain medical files in multiple formats, and the server can obtain files uploaded by the user terminal.

Step 410: Obtain a diagnosis file, an annotation file, and an electronic medical record.

Specifically, the server can obtain the diagnosis file, the annotation file, the electronic medical record, etc. in the medical file. The server can obtain the files uploaded by the user terminal.

Step 420, upload the client.

Specifically, the medical files obtained in step 400 and step 410 are uploaded to the client. Then the server can call the corresponding analysis algorithm for analysis based on the obtained medical files in multiple formats, and obtain the medical image in step 430 and the metadata information in step 440. The metadata information can be label information, including tag information and diagnosis. Information and other label information, etc.

Step 450: Obtain a medical tag library.

This includes operations such as cleaning and management of the metadata information obtained in step 440 by the server to obtain standardized metadata information, and generating a medical tag library based on the standardized metadata information. The medical tag library may be an image annotation file.

Step 460, the medical image management platform.

The server uploads the data acquired in step 430 and step 450 to the medical image management platform. The medical image management platform may be an image annotation file library, which stores multiple types of image annotation files. Then, the medical image management platform can be used to perform the medical image retrieval in step 470, the medical image export in step 480, and the intelligent labeling function in step 490.

It should be understood that although the various steps in the flowcharts of FIGS. 2-4 are displayed in sequence as indicated by the arrows, these steps are not necessarily performed in sequence in the order indicated by the arrows. Unless specifically stated in this article, the execution of these steps is not strictly limited in order, and these steps can be executed in other orders. Moreover, at least part of the steps in Figures 2-4 may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. These sub-steps or stages The execution order of is not necessarily performed sequentially, but may be performed alternately or alternately with at least a part of other steps or sub-steps or stages of other steps.

In one of the embodiments, as shown in FIG. 5, a medical image recognition model generation device is provided, including: a data acquisition module 510, a feature extraction module 520, an annotation generation module 530, an image file generation module 540, and a model generation module 550, of which:

The data acquisition module 510 is used to acquire medical data.

The feature extraction module 520 is configured to extract medical images in the medical data, and perform feature extraction on the medical images to obtain image features.

The annotation generating module 530 is configured to extract medical documents in the medical data, analyze the medical documents to generate annotation information, and search for image annotation information corresponding to the medical image in the annotation information.

The image file generating module 540 is configured to establish a mapping relationship according to the image characteristics and the image annotation information, and generate an image annotation file according to the mapping relationship.

The model generation module 550 is configured to input the image annotation file into a machine learning model to learn the feature relationship between the image feature in the image annotation file and the image annotation information through the machine learning model, and according to the feature Relationship generation medical image recognition model.

In one of the embodiments, the feature extraction module 520 includes:

The image annotation information extraction unit is used to extract the annotation information in the medical image.

The coordinate acquisition unit is used to acquire the position coordinates corresponding to the label information.

The regional image extraction unit is used to extract regional medical images corresponding to the position coordinates from the medical images.

The regional feature extraction unit is used to extract image features of the regional medical image by using an image feature extraction algorithm.

In one of the embodiments, the device further includes:

The segmentation module is configured to segment the label information into unit label information according to preset rules.

The information extraction module is configured to extract and match the unit label information with the information in the label information database, and extract the information with the greatest degree of matching as label information.

In one of the embodiments, the image file generating module 540 includes:

The identification acquiring unit is used to acquire the image identification corresponding to the medical image.

The method search unit is used to search for the image analysis method corresponding to the image identifier.

The information generating unit is configured to analyze the attribute information associated with the medical image by using the image analysis method to generate image tag information.

The file generating unit is configured to establish a mapping relationship according to the image characteristics, the image annotation information, and the image tag information, and generate an image annotation file according to the mapping relationship.

In one of the embodiments, the device further includes:

The query identifier extraction module is used to receive the query request sent by the terminal, and extract the query identifier carried in the query request.

The identification extraction module is used to extract the image identification corresponding to the image annotation file.

The calculation module is used to calculate the matching degree between the query identifier and each of the image identifiers.

The first pushing module is configured to push the image annotation file associated with the image identifier with the greatest matching degree to the terminal.

In one of the embodiments, the device further includes:

The receiving module is used to receive medical images sent by the terminal.

The annotation information recognition module is used to input the medical image into the medical image recognition model to extract the image feature corresponding to the medical image through the medical image recognition model, and obtain the image according to the medical image feature Annotate information.

The second push module is used to push the image annotation information to the terminal.

Regarding the specific limitations of the medical image recognition model generating device, please refer to the above limitation on the medical image recognition model generating method, which will not be repeated here. Each module in the above-mentioned medical image recognition model generating device can be implemented in whole or in part by software, hardware, and a combination thereof. The above-mentioned modules may be embedded in the form of hardware or independent of the processor in the computer equipment, or may be stored in the memory of the computer equipment in the form of software, so that the processor can call and execute the operations corresponding to the above-mentioned modules.

In one of the embodiments, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as shown in FIG. 6. The computer equipment includes a processor, a memory, a network interface, and a database connected through a system bus. The processor of the computer device is used to provide calculation and control capabilities. The memory of the computer device includes non-volatile or volatile storage media and internal memory. The non-volatile or volatile storage medium stores an operating system, computer readable instructions, and a database. The internal memory provides an environment for the operation of the operating system and computer-readable instructions in the non-volatile or volatile storage medium. The computer equipment database is used for medical data processing related data. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer-readable instructions are executed by the processor, a method for generating a medical image recognition model is realized.

Those skilled in the art can understand that the structure shown in FIG. 6 is only a block diagram of part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may Including more or fewer parts than shown in the figure, or combining some parts, or having a different arrangement of parts.

A computer device includes a memory and one or more processors. The memory stores computer-readable instructions. When the computer-readable instructions are executed by the processor, the one or more processors perform the following steps: acquiring medical data; extracting For the medical images in the medical data, feature extraction is performed on the medical images to obtain image features; the medical files in the medical data are extracted, the medical files are parsed to generate annotation information, and the annotation information is searched for and the The image annotation information corresponding to the medical image; establish a mapping relationship according to the image characteristics and the image annotation information, and generate an image annotation file according to the mapping relationship; and input the image annotation file into a machine learning model to pass machine learning The model learns the feature relationship between the image feature and the image tag information in the image tag file, and generates a medical image recognition model according to the feature relationship.

In one of the embodiments, when the processor executes the computer-readable instructions to implement the step of extracting features from the medical image to obtain image features, it is also used to: extract the annotation information in the medical image; and obtain the annotation information Corresponding location coordinates; extracting regional medical images corresponding to the location coordinates from the medical images; and extracting image features of the regional medical images using an image feature extraction algorithm.

In one of the embodiments, when the processor executes the computer-readable instructions, after the processor executes the parsing of the medical file to generate annotation information, it is further used to: divide the annotation information into unit annotation information according to a preset rule; And matching the unit labeling information with the information in the labeling information database, and extracting the information with the greatest degree of matching as labeling information.

In one of the embodiments, when the processor executes the computer-readable instructions, the step of establishing a mapping relationship based on the image characteristics and the image annotation information, and generating an image annotation file based on the mapping relationship is further used for: Obtain the image identifier corresponding to the medical image; find the image analysis method corresponding to the image identifier; use the image analysis method to analyze the attribute information associated with the medical image to generate image tag information; and according to the image feature Establishing a mapping relationship between the image annotation information and the image tag information, and generating an image annotation file according to the mapping relationship.

In one of the embodiments, when the processor executes the computer-readable instruction, after the image annotation file is generated according to the mapping relationship, the processor is further used to: receive the query request sent by the terminal, and extract the query identifier carried in the query request Extracting the image identification corresponding to the image annotation file; calculating the matching degree between the query identification and each of the image identification; and pushing the image annotation file associated with the image identification with the largest matching degree to the terminal.

In one of the embodiments, when the processor executes the computer-readable instructions to realize the generation of the medical image recognition model, it is further used to: receive the medical image sent by the terminal; input the medical image into the medical image recognition model to Extracting the image feature corresponding to the medical image through the medical image recognition model, obtaining the image annotation information according to the medical image feature; and pushing the image annotation information to the terminal.

One or more computer-readable storage media storing computer-readable instructions. When the computer-readable instructions are executed by one or more processors, the one or more processors perform the following steps: acquiring medical data; extracting the medical The medical image in the data is extracted from the medical image to obtain the image feature; the medical file in the medical data is extracted, the medical file is parsed to generate annotation information, and the annotation information is searched for in the annotation information and the medical image Corresponding image annotation information; establish a mapping relationship according to the image features and the image annotation information, and generate an image annotation file according to the mapping relationship; and input the image annotation file into a machine learning model to learn all the information through the machine learning model According to the feature relationship between the image feature and the image tag information in the image tag file, a medical image recognition model is generated according to the feature relationship.

The computer-readable storage medium may be non-volatile or volatile.

In one of the embodiments, when the computer-readable instruction is executed by the processor to implement the step of extracting features from the medical image to obtain image features, it is also used to: extract the annotation information in the medical image; and obtain the annotation information Corresponding location coordinates; extracting regional medical images corresponding to the location coordinates from the medical images; and extracting image features of the regional medical images using an image feature extraction algorithm.

In one of the embodiments, when the computer-readable instruction is executed by the processor, after realizing the parsing of the medical file to generate annotation information, it is further used to: divide the annotation information into unit annotation information according to a preset rule; And matching the unit labeling information with the information in the labeling information database, and extracting the information with the greatest degree of matching as labeling information.

In one of the embodiments, when the computer-readable instructions are executed by the processor, the step of establishing a mapping relationship based on the image characteristics and the image annotation information, and generating an image annotation file based on the mapping relationship is further used to: Obtain the image identifier corresponding to the medical image; find the image analysis method corresponding to the image identifier; use the image analysis method to analyze the attribute information associated with the medical image to generate image tag information; and according to the image feature Establishing a mapping relationship between the image annotation information and the image tag information, and generating an image annotation file according to the mapping relationship.

In one of the embodiments, when the computer-readable instruction is executed by the processor, after the image annotation file is generated according to the mapping relationship, it is further used to: receive a query request sent by the terminal, and extract the query identifier carried in the query request Extracting the image identification corresponding to the image annotation file; calculating the matching degree between the query identification and each of the image identification; and pushing the image annotation file associated with the image identification with the largest matching degree to the terminal.

In one of the embodiments, when the computer-readable instruction is executed by the processor to realize the generation of the medical image recognition model, it is further used to: receive the medical image sent by the terminal; input the medical image into the medical image recognition model to Extracting the image feature corresponding to the medical image through the medical image recognition model, obtaining the image annotation information according to the medical image feature; and pushing the image annotation information to the terminal.

A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing relevant hardware through computer-readable instructions. The computer-readable instructions can be stored in a computer-readable storage. In the medium, when the computer-readable instructions are executed, they may include the processes of the above-mentioned method embodiments. Any reference to memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and/or volatile memory. Non-volatile memory may include read only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not a limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous chain Channel (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

The technical features of the above embodiments can be combined arbitrarily. In order to make the description concise, all possible combinations of the technical features in the above embodiments are not described. However, as long as there is no contradiction in the combination of these technical features, they should be It is considered as the range described in this specification.

The above-mentioned embodiments only express several implementation manners of the present application, and the description is relatively specific and detailed, but it should not be understood as a limitation on the scope of the invention patent. It should be pointed out that for those of ordinary skill in the art, without departing from the concept of this application, several modifications and improvements can be made, and these all fall within the protection scope of this application. Therefore, the scope of protection of the patent of this application shall be subject to the appended claims.

Claims

A method for generating a medical image recognition model, including:

Access to medical data;

Extracting medical images in the medical data, and performing feature extraction on the medical images to obtain image features;

Extracting medical documents in the medical data, analyzing the medical documents to generate annotation information, and searching for image annotation information corresponding to the medical image in the annotation information;

Establishing a mapping relationship according to the image characteristics and the image annotation information, and generating an image annotation file according to the mapping relationship; and

The image annotation file is input into a machine learning model to learn the feature relationship between the image features in the image annotation file and the image annotation information through the machine learning model, and a medical image recognition model is generated according to the feature relationship.
The method according to claim 1, wherein said performing feature extraction on said medical image to obtain image features comprises:

Extract the annotation information in the medical image;

Acquiring position coordinates corresponding to the label information;

Extracting regional medical images corresponding to the position coordinates from the medical images; and

The image feature extraction algorithm is used to extract the image feature of the regional medical image.
The method according to claim 1, wherein after the analysis of the medical file to generate annotation information, the method further comprises:

Dividing the label information into unit label information according to preset rules; and

The unit labeling information is matched with the information in the labeling information database, and the information with the largest matching degree is extracted as labeling information.
The method according to claim 1, wherein the establishing a mapping relationship according to the image characteristics and the image annotation information, and generating an image annotation file according to the mapping relationship, comprises:

Acquiring an image identifier corresponding to the medical image;

Searching for an image analysis method corresponding to the image identifier;

Using the image analysis method to analyze the attribute information associated with the medical image to generate image tag information; and

A mapping relationship is established according to the image feature, the image annotation information, and the image tag information, and an image annotation file is generated according to the mapping relationship.
The method according to claim 1 or 4, wherein, after the image annotation file is generated according to the mapping relationship, the method further comprises:

Receiving the query request sent by the terminal, and extracting the query identifier carried in the query request;

Extracting the image identifier corresponding to the image annotation file;

Calculating the degree of matching between the query identifier and each of the image identifiers; and

Push the image annotation file associated with the image identifier with the greatest degree of matching to the terminal.
The method according to claim 1, wherein after said generating the medical image recognition model, the method further comprises:

Receive medical images sent by the terminal;

Inputting the medical image into the medical image recognition model to extract the image feature corresponding to the medical image through the medical image recognition model, and obtain the image annotation information according to the medical image feature; and

Push the image annotation information to the terminal.
A medical image recognition model generation device, including:

Data acquisition module for acquiring medical data;

The feature extraction module is used to extract medical images in the medical data, and perform feature extraction on the medical images to obtain image features;

An annotation generation module, configured to extract medical documents in the medical data, analyze the medical documents to generate annotation information, and search for image annotation information corresponding to the medical image in the annotation information;

An image file generating module, configured to establish a mapping relationship according to the image characteristics and the image annotation information, and generate an image annotation file according to the mapping relationship; and

The model generation module is configured to input the image annotation file into a machine learning model to learn the feature relationship between the image feature in the image annotation file and the image annotation information through the machine learning model, and according to the feature relationship Generate medical image recognition model.
A computer device includes a memory and one or more processors. The memory stores computer-readable instructions. When the computer-readable instructions are executed by the one or more processors, the one or more Each processor performs the following steps:

Access to medical data;

Extracting medical images in the medical data, and performing feature extraction on the medical images to obtain image features;

Extracting medical documents in the medical data, analyzing the medical documents to generate annotation information, and searching for image annotation information corresponding to the medical image in the annotation information;

Establishing a mapping relationship according to the image characteristics and the image annotation information, and generating an image annotation file according to the mapping relationship; and

The image annotation file is input into a machine learning model to learn the feature relationship between the image features in the image annotation file and the image annotation information through the machine learning model, and a medical image recognition model is generated according to the feature relationship.
8. The computer device according to claim 8, wherein the feature extraction of the medical image to obtain the image feature, which is implemented when the processor executes the computer readable instruction, comprises:

Extract the annotation information in the medical image;

Acquiring position coordinates corresponding to the label information;

Extracting regional medical images corresponding to the position coordinates from the medical images; and

The image feature extraction algorithm is used to extract the image feature of the regional medical image.
The computer device according to claim 8, wherein, after the analysis of the medical file to generate annotation information, which is realized when the processor executes the computer-readable instruction, further comprises:

Dividing the label information into unit label information according to preset rules; and

The unit labeling information is matched with the information in the labeling information database, and the information with the largest matching degree is extracted as labeling information.
8. The computer device according to claim 8, wherein the establishment of a mapping relationship based on the image feature and the image annotation information, which is implemented when the processor executes the computer-readable instruction, is generated based on the mapping relationship Image annotation files, including:

Acquiring an image identifier corresponding to the medical image;

Searching for an image analysis method corresponding to the image identifier;

Using the image analysis method to analyze the attribute information associated with the medical image to generate image tag information; and

A mapping relationship is established according to the image feature, the image annotation information, and the image tag information, and an image annotation file is generated according to the mapping relationship.
The computer device according to claim 8 or 11, wherein after the generating of the image annotation file according to the mapping relationship, which is implemented when the processor executes the computer-readable instruction, further comprises:

Receiving the query request sent by the terminal, and extracting the query identifier carried in the query request;

Extracting the image identifier corresponding to the image annotation file;

Calculating the degree of matching between the query identifier and each of the image identifiers; and

Push the image annotation file associated with the image identifier with the greatest degree of matching to the terminal.
8. The computer device according to claim 8, wherein after the generating of the medical image recognition model realized when the processor executes the computer-readable instruction, the method further comprises:

Receive medical images sent by the terminal;

Inputting the medical image into the medical image recognition model to extract the image feature corresponding to the medical image through the medical image recognition model, and obtain the image annotation information according to the medical image feature; and

Push the image annotation information to the terminal.
One or more computer-readable storage media storing computer-readable instructions, which when executed by one or more processors, cause the one or more processors to perform the following steps:

Access to medical data;

Extracting medical images in the medical data, and performing feature extraction on the medical images to obtain image features;

Extracting medical documents in the medical data, analyzing the medical documents to generate annotation information, and searching for image annotation information corresponding to the medical image in the annotation information;

Establishing a mapping relationship according to the image characteristics and the image annotation information, and generating an image annotation file according to the mapping relationship; and

The image annotation file is input into a machine learning model to learn the feature relationship between the image features in the image annotation file and the image annotation information through the machine learning model, and a medical image recognition model is generated according to the feature relationship.
The storage medium according to claim 14, wherein the feature extraction of the medical image to obtain the image feature, which is realized when the computer-readable instruction is executed by the processor, comprises:

Extract the annotation information in the medical image;

Acquiring position coordinates corresponding to the label information;

Extracting regional medical images corresponding to the position coordinates from the medical images; and

The image feature extraction algorithm is used to extract the image feature of the regional medical image.
The storage medium according to claim 14, wherein, after the parsing of the medical file to generate annotation information realized when the computer-readable instructions are executed by the processor, the method further comprises:

Dividing the label information into unit label information according to preset rules; and

The unit labeling information is matched with the information in the labeling information database, and the information with the largest matching degree is extracted as labeling information.
The storage medium according to claim 14, wherein the establishment of a mapping relationship based on the image feature and the image annotation information, which is implemented when the computer-readable instruction is executed by the processor, is based on the mapping relationship Generate image annotation files, including:

Acquiring an image identifier corresponding to the medical image;

Searching for an image analysis method corresponding to the image identifier;

Using the image analysis method to analyze the attribute information associated with the medical image to generate image tag information; and

A mapping relationship is established according to the image feature, the image annotation information, and the image tag information, and an image annotation file is generated according to the mapping relationship.
The storage medium according to claim 14 or 17, wherein, after the generating of the image annotation file according to the mapping relationship, which is realized when the computer-readable instruction is executed by the processor, further comprises:

Receiving the query request sent by the terminal, and extracting the query identifier carried in the query request;

Extracting the image identifier corresponding to the image annotation file;

Calculating the degree of matching between the query identifier and each of the image identifiers; and

Push the image annotation file associated with the image identifier with the greatest degree of matching to the terminal.
14. The storage medium according to claim 14, wherein after the generating of the medical image recognition model realized when the computer-readable instructions are executed by the processor, the method further comprises:

Receive medical images sent by the terminal;

Inputting the medical image into the medical image recognition model to extract the image feature corresponding to the medical image through the medical image recognition model, and obtain the image annotation information according to the medical image feature; and

Push the image annotation information to the terminal.