Cerebral infarction critical and severe identification system
By designing a critical and critical cerebral infarction that combines user data and medical images, the problems of low recognition efficiency and low accuracy in the prior art are solved, and more efficient and accurate identification of critical and critical cerebral infarction are achieved.
Patent Information
- Application Number
- CN202411526187.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-29
- Publication Date
- 2025-02-14
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The existing methods for identifying critical and severe cerebral infarction rely on doctors to observe brain scan images with naked eyes, resulting in low recognition rate, long time-consuming and subjective and low reliability of the results.
A critical and critical cerebral infarction recognition system is designed to upload user basic data, symptom text data, medical history text data and local images through user clients, and combine medical images acquired by data acquisition devices to generate a feature vector matrix of cerebral infarction evaluation data. The system uses feature extraction and fusion modules to predict cerebral infarction levels and probability values, and finally determines the final identification result through weighted calculations or other calculation methods.
It improves the recognition efficiency and accuracy of critical and critical cerebral infarction, reduces the subjectivity of doctors' diagnosis, and enhances the reliability of results.
Smart Images
Figure CN119446432A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of medical identification technology, and in particular to a critical and severe cerebral infarction identification system. Background Art
[0002] As work pressure increases, the incidence of brain diseases among young people is also on a continuous increase. Insufficient blood supply to the brain is a common factor that causes brain diseases. Severe cases of insufficient blood supply to the brain may even endanger life. Existing brain disease examination methods usually use CT (computer tomography) or MR (magnetic resonance imaging) to scan the patient's brain to obtain brain scan images, and then interpret the brain scan images to determine the core area of cerebral infarction.
[0003] Traditionally, the critical and severe disease recognition of cerebral infarction is done by doctors observing the patient's brain scan image with naked eyes. The recognition rate is low, the time consumption is relatively long, and the diagnosis results are often subjective and unreliable. Therefore, there is an urgent need for a critical and severe disease recognition system for cerebral infarction to improve the recognition efficiency and accuracy of critical and severe diseases of cerebral infarction. Summary of the invention
[0004] The embodiment of the present application provides a system for identifying critical and severe cerebral infarction, which is used to improve the recognition efficiency and accuracy of critical and severe cerebral infarction.
[0005] The embodiment of the present invention provides a critical and severe cerebral infarction identification system, the system comprising: a user client, a data acquisition device and a cerebral infarction identification device, the cerebral infarction identification device is connected to the user client and the data acquisition device, and the cerebral infarction identification device is used to perform the following steps: Acquire user basic data, symptom text data, medical history text data and user partial images uploaded by the user through the user client; and the user medical image acquired by the data acquisition device; the user partial image includes the user's facial image and tongue coating image; Performing text extraction on the user basic data, the symptom text data and the medical history text data respectively, and determining a plurality of basic data key-value pairs, symptom key-value pairs and medical history key-value pairs; the basic data key-value pairs include basic data elements and their corresponding basic data values, the symptom key-value pairs include symptom elements and their corresponding symptom values, and the medical history key-value pairs include medical history elements and their corresponding medical history values; Determine the weight values corresponding to each basic data key-value pair, symptom key-value pair, and medical history key-value pair based on a preset mapping table, wherein the preset mapping table stores the weight values corresponding to all basic data elements, symptom elements, and medical history elements; Determine the cerebral infarction evaluation data feature vector matrix through the multiple basic data key-value pairs, symptom key-value pairs, medical history key-value pairs and their corresponding weight values, wherein the cerebral infarction evaluation data feature vector matrix includes 3 columns of data, the data elements in the first column are the basic data values and weight values corresponding to the basic data elements, the data elements in the second column are the symptom values and weight values corresponding to the symptom elements, and the data elements in the third column are the medical history values and weight values corresponding to the medical history elements; Determine a first cerebral infarction recognition result of the user according to the cerebral infarction evaluation data feature vector matrix and the user partial image; Determine a second cerebral infarction recognition result of the user according to the cerebral infarction evaluation data feature vector matrix and the user medical image; The final cerebral infarction recognition result of the user is determined by the first cerebral infarction recognition result and the second cerebral infarction recognition result.
[0006] In an optional embodiment provided by the present invention, the user basic data, the symptom text data and the medical history text data are respectively subjected to text extraction to determine a plurality of basic data key-value pairs, symptom key-value pairs and medical history key-value pairs, including: Performing keyword extraction on the user basic data, the symptom text data, and the medical history text data to obtain a basic keyword set, a symptom keyword set, and a medical history keyword set; Based on the basic data word library, synonym matching is performed on the keywords in the basic keyword set to obtain basic data elements; based on the symptom word library, synonym matching is performed on the keywords in the symptom keyword set to obtain symptom elements; based on the medical history word library, synonym matching is performed on the keywords in the medical history keyword set to obtain medical history elements; the basic data word library stores basic keywords for determining cerebral infarction; the symptom word library stores symptom keywords for determining cerebral infarction, and the medical history word library stores medical history keywords for determining cerebral infarction; Perform text context analysis and matching on the user basic data to obtain the basic data value corresponding to the basic data element; perform text context analysis and matching on the symptom text data to obtain the symptom value corresponding to the symptom element; perform text context analysis and matching on the medical history element to obtain the medical history value corresponding to the medical history element.
[0007] In an optional embodiment provided by the present invention, determining the first cerebral infarction recognition result of the user according to the cerebral infarction evaluation data feature vector matrix and the user partial image includes: Inputting the cerebral infarction evaluation data feature vector matrix and the user's partial image into a first cerebral infarction recognition model, and extracting color features and texture features of the user's partial image through a feature extraction module in the first cerebral infarction recognition model; The color feature and the texture feature are fused by the feature fusion module in the first cerebral infarction recognition model to obtain a local image fusion feature; A first cerebral infarction recognition result of a user is obtained according to the local image fusion feature and the cerebral infarction evaluation data feature vector matrix prediction, and the first cerebral infarction recognition result includes a predicted cerebral infarction grade and a predicted probability value.
[0008] In an optional embodiment provided by the present invention, determining the second cerebral infarction recognition result of the user according to the cerebral infarction evaluation data feature vector matrix and the user medical image includes: Inputting the cerebral infarction evaluation data feature vector matrix and the user medical image into a second cerebral infarction recognition model, and extracting the infarction site features and infarction area features of the user medical image through a feature extraction module in the second cerebral infarction recognition model; fusing the infarction site feature and the infarction area feature through the feature fusion module in the second cerebral infarction recognition model to obtain a medical image fusion feature; A second cerebral infarction recognition result of the user is predicted based on the medical image fusion features and the cerebral infarction evaluation data feature vector matrix, and the second cerebral infarction recognition result includes a predicted cerebral infarction grade and a predicted probability value.
[0009] In an optional embodiment provided by the present invention, determining the final cerebral infarction recognition result of the user by using the first cerebral infarction recognition result and the second cerebral infarction recognition result includes: Determining whether the predicted probability values corresponding to the first cerebral infarction recognition result and the second cerebral infarction recognition result are both greater than a preset probability value; If both are greater than a preset probability value, weighted calculation is performed on the predicted cerebral infarction levels of the first cerebral infarction recognition result and the second cerebral infarction recognition result to obtain a final cerebral infarction recognition result for the user; If there is a predicted probability value greater than the preset probability value, the predicted cerebral infarction level with the predicted probability value greater than the preset probability value is used as the user's final cerebral infarction recognition result.
[0010] In an optional embodiment provided by the present invention, after determining whether the predicted probability values corresponding to the first cerebral infarction recognition result and the second cerebral infarction recognition result are both greater than a preset probability value, the method further includes: If there is no predicted probability value greater than the preset probability value, the final cerebral infarction recognition result of the user is obtained by calculation based on the basic data key-value pair, the symptom key-value pair, and the medical history key-value.
[0011] In an optional embodiment provided by the present invention, the training process of the first cerebral infarction recognition model is: Obtaining a sample feature vector matrix of cerebral infarction evaluation data and a sample local image corresponding to each piece of sample data; the sample local image includes a facial image and a tongue coating image; Inputting the sample feature vector matrix and the sample local image into the first cerebral infarction recognition model, and extracting the color features and texture features corresponding to the facial image and the tongue coating image respectively through the feature extraction module in the first cerebral infarction recognition model; By using the feature fusion module in the first cerebral infarction recognition model, the facial image and the tongue coating image are respectively fused corresponding to the color features and the texture features to obtain a facial image fusion feature and a tongue coating image fusion feature; Obtain a first prediction value by fusing features of the data sample feature vector matrix and the facial image; obtain a second prediction value by fusing features of the data sample feature vector matrix and the tongue coating image; Calculate a loss value according to the first predicted value and the second predicted value; When the loss value is less than a preset value, the training of the first cerebral infarction recognition model is stopped.
[0012] In an optional embodiment provided by the present invention, the training process of the second cerebral infarction recognition model is: Obtain the cerebral infarction evaluation data sample feature vector matrix and sample medical image corresponding to each sample data; Inputting the sample feature vector matrix of the cerebral infarction evaluation data and the sample medical image into a second cerebral infarction recognition model, and extracting the infarction site features and infarction area features of the sample medical image through a feature extraction module in the second cerebral infarction recognition model; fusing the infarction site feature and the infarction area feature through the feature fusion module in the second cerebral infarction recognition model to obtain a medical image fusion feature; Obtaining a prediction value by fusing the sample feature vector matrix and the medical image features; A loss value is calculated according to the predicted value and the label corresponding to the sample data; when the loss value is less than a preset value, the training of the first cerebral infarction recognition model is stopped.
[0013] In an optional embodiment provided by the present invention, before training the first cerebral infarction recognition model or the second cerebral infarction recognition model, the method further includes: Acquire a sample data set for training a cerebral infarction recognition model, wherein each piece of sample data in the sample data set includes user basic data, symptom text data, medical history text data, user local images, and user medical images; Determine a cerebral infarction evaluation data sample feature vector matrix according to the user basic data, symptom text data, and medical history text data; Performing image smoothing based on median filtering on the user local image, and approximating the gradient of each pixel in the image by formulas for the gradients in the horizontal and vertical directions of the image to obtain a gradient image; and using the obtained gradient image as a sample local image; The user medical image is converted into the same size, and image smoothing based on median filtering is performed to obtain a sample medical image.
[0014] An embodiment of the present invention provides a device for identifying cerebral infarction, the device comprising: An acquisition module, used to acquire user basic data, symptom text data, medical history text data and user partial images uploaded by the user through the user client; and the user medical images acquired by the data acquisition device; the user partial images include the user's facial image and tongue coating image; A text extraction module is used to perform text extraction on the user basic data, the symptom text data and the medical history text data respectively, and determine a plurality of basic data key-value pairs, symptom key-value pairs and medical history key-value pairs; the basic data key-value pairs include basic data elements and their corresponding basic data values, the symptom key-value pairs include symptom elements and their corresponding symptom values, and the medical history key-value pairs include medical history elements and their corresponding medical history values; A determination module, used to determine the weight values corresponding to each basic data key-value pair, symptom key-value pair, and medical history key-value pair based on a preset mapping table, wherein the preset mapping table stores the weight values corresponding to all basic data elements, symptom elements, and medical history elements; The determination module is further used to determine the cerebral infarction evaluation data feature vector matrix through the multiple basic data key-value pairs, symptom key-value pairs, medical history key-value pairs and their corresponding weight values, wherein the cerebral infarction evaluation data feature vector matrix includes 3 columns of data, the data elements in the first column are the basic data values and weight values corresponding to the basic data elements, the data elements in the second column are the symptom values and weight values corresponding to the symptom elements, and the data elements in the third column are the medical history values and weight values corresponding to the medical history elements; The determination module is further configured to determine a first cerebral infarction recognition result of the user according to the cerebral infarction evaluation data feature vector matrix and the user partial image; The determination module is further used to determine the second cerebral infarction recognition result of the user according to the cerebral infarction evaluation data feature vector matrix and the user medical image; The determination module is further configured to determine a final cerebral infarction recognition result of the user through the first cerebral infarction recognition result and the second cerebral infarction recognition result.
[0015] The present invention provides a critical and severe cerebral infarction identification system, which comprises: a user client, a data acquisition device and a cerebral infarction identification device, the cerebral infarction identification device is connected to the user client and the data acquisition device, and the cerebral infarction identification device is used to execute: obtaining user basic data, symptom text data, medical history text data and user local images uploaded by the user through the user client and the user medical images acquired by the data acquisition device, performing text extraction on the user basic data, symptom text data and medical history text data respectively, and determining a plurality of basic data key-value pairs, symptom key-value pairs and medical history key-value pairs; determining weight values corresponding to each basic data key-value pair, symptom key-value pair and medical history key-value pair respectively based on a preset mapping table, determining a cerebral infarction evaluation data feature vector matrix through a plurality of basic data key-value pairs, symptom key-value pairs, medical history key-value pairs and their corresponding weight values respectively, then determining a first cerebral infarction identification result of the user according to the cerebral infarction evaluation data feature vector matrix and the user local images, and determining a second cerebral infarction identification result of the user according to the cerebral infarction evaluation data feature vector matrix and the user medical images, and finally determining a final cerebral infarction identification result of the user through the first cerebral infarction identification result and the second cerebral infarction identification result. Compared with the prior art in which doctors diagnose critical and severe cerebral infarction by visually observing the patient's brain scan images, the present application generates a cerebral infarction evaluation data feature vector matrix based on data uploaded by the user client and the data acquisition device, and then determines the user's final cerebral infarction recognition result based on the first cerebral infarction recognition result and the second cerebral infarction recognition result generated by the cerebral infarction evaluation data feature vector matrix and related images. Therefore, the present application can improve the recognition efficiency and accuracy of critical and severe cerebral infarction. BRIEF DESCRIPTION OF THE DRAWINGS
[0016] Figure 1 A structural diagram of a critical and severe cerebral infarction identification system provided in this application; Figure 2 A flowchart of a method for identifying critical and severe cerebral infarction provided in this application; Figure 3 This is a schematic diagram of the structure of a cerebral infarction identification device provided in this application. DETAILED DESCRIPTION
[0017] In order to better understand the above-mentioned technical scheme, the technical scheme of the embodiments of the present application is described in detail below through the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present application and the specific features in the embodiments are detailed descriptions of the technical scheme of the embodiments of the present application, rather than limitations on the technical scheme of the present application. In the absence of conflict, the embodiments of the present application and the technical features in the embodiments may be combined with each other.
[0018] See also Figure 1, a critical and severe cerebral infarction identification system provided in an embodiment of the present invention, the system includes: a user client, a data acquisition device and a cerebral infarction identification device, and the cerebral infarction identification device is connected to the user client and the data acquisition device. Among them, the user client can be a mobile phone terminal, a computer terminal, etc.; the data acquisition device can be a medical device, such as a medical device that can obtain the user's ultrasound image; the cerebral infarction identification device is used to judge whether the user has a cerebral infarction based on the data provided by the user client and the data acquisition device, and specifically provide the level of cerebral infarction, which can be specifically divided into: severe, moderate, common, etc. according to needs, and this embodiment does not make specific limitations on this.
[0019] The present embodiment provides a system for identifying critical and severe cerebral infarction, which determines whether the user has cerebral infarction through data provided by a user client and a data acquisition device, thereby realizing automatic identification of the severity of the user's cerebral infarction. Compared with the prior art in which doctors diagnose critical and severe cerebral infarction by visually observing the patient's brain scan images, the cerebral infarction identification device generates a cerebral infarction evaluation data feature vector matrix based on user basic data, symptom text data, and medical history text data, and then determines the user's final cerebral infarction identification result through the cerebral infarction evaluation data feature vector matrix, user local images, and user medical images. Therefore, the present application can improve the recognition efficiency and accuracy of critical and severe cerebral infarction.
[0020] Specific as Figure 2 As shown, this embodiment provides a method for identifying critical and severe cerebral infarction, which is applied to the cerebral infarction identification device in the above-mentioned critical and severe cerebral infarction identification system. The execution steps of the method are: S201, obtaining user basic data, symptom text data, medical history text data and user local images uploaded by the user through the user client; and the user medical images obtained by the data acquisition device.
[0021] Among them, user basic data is some basic data describing the user, including the user's gender, age, smoking history, drinking history, etc.; symptom text data is data describing the user's cerebral infarction-related symptoms, such as the user's hyperlipidemia, infarction site, infarction area, etc.; medical history text data is a description of the user's medical history, including a history of hypertension, a history of diabetes, and a history of cerebral infarction. This embodiment does not limit the specific data of user basic data, symptom text data, and medical history text data. The user's medical image can be an ultrasound image of the user.
[0022] It should be noted that, based on the diagnosis of cerebral infarction by face diagnosis and tongue diagnosis in traditional Chinese medicine, this embodiment can divide the user's local image into the user's facial image and tongue coating image, that is, the cerebral infarction of the user is jointly diagnosed by the facial image and the tongue coating image, the facial image at least includes the user's eyes, eyebrows and nose, and the tongue coating image at least includes the user's tongue close-up. Among them, the complexion of patients with cerebral infarction often shows abnormal changes such as dullness, cyanosis or dullness. This is because cerebral infarction causes cerebrovascular obstruction, poor blood circulation, resulting in dull or cyanosis; the tongue color of patients with cerebral infarction is mostly dull, purple or cyanosis, and the tongue quality may also be pale, fat, dry or have ecchymosis. This is because cerebral infarction causes cerebrovascular obstruction, resulting in poor blood circulation, resulting in changes in tongue color and tongue quality.
[0023] S202, performing text extraction on the user basic data, the symptom text data and the medical history text data respectively, and determining a plurality of basic data key-value pairs, symptom key-value pairs, and medical history key-value pairs.
[0024] Among them, the basic data key-value pair includes basic data elements and their corresponding basic data values, such as basic data key values can be specifically (gender, female), (age, 62), (weight, 60), etc.; the symptom key-value pair includes symptom elements and their corresponding symptom values, such as symptom key-value pairs can be specifically (infarction site, thalamus), (infarction area, small area), etc.; the medical history key-value pair includes medical history elements and their corresponding medical history values, such as medical history key values can be (history of hypertension, yes), (history of diabetes, no), etc. It should be noted that the symptom key value in this embodiment can be data uploaded by the user, or it can be determined based on historical diagnostic text data, that is, the symptom text data in this implementation includes data input by the user based on his or her own situation, or it can be the user's historical diagnostic data.
[0025] In an optional embodiment provided by the present application, the user basic data, the symptom text data and the medical history text data are respectively subjected to text extraction to determine a plurality of basic data key-value pairs, symptom key-value pairs and medical history key-value pairs, including: S2021, perform keyword extraction on the user basic data, the symptom text data and the medical history text data to obtain a basic keyword set, a symptom keyword set and a medical history keyword set.
[0026] Specifically, this embodiment may first perform word segmentation on the user basic data, symptom text data, and medical history text data, then remove stop words and invalid words, and then perform keyword extraction, extracting a basic keyword set from the user basic data, extracting a symptom keyword set from the symptom text data, and extracting a medical history keyword set from the medical history text data.
[0027] S2022, based on the basic data vocabulary, synonym matching is performed on the keywords in the basic keyword set to obtain basic data elements; based on the symptom vocabulary, synonym matching is performed on the keywords in the symptom keyword set to obtain symptom elements; based on the medical history vocabulary, synonym matching is performed on the keywords in the medical history keyword set to obtain medical history elements.
[0028] The basic data word library stores basic keywords for determining cerebral infarction; the symptom word library stores symptom keywords for determining cerebral infarction; and the medical history word library stores medical history keywords for determining cerebral infarction.
[0029] For example, the basic keyword set is {name, gender, age, place of origin}, and the basic data vocabulary stores basic keywords related to determining cerebral infarction, which are {gender, age,..., weight}. The basic data elements obtained by synonym matching the keywords in the basic keyword set through the basic data vocabulary are {gender, age}, that is, the keywords in the basic keyword set used to determine cerebral infarction are determined through the basic data vocabulary.
[0030] S2023, perform text context analysis and matching on the user basic data to obtain the basic data value corresponding to the basic data element; perform text context analysis and matching on the symptom text data to obtain the symptom value corresponding to the symptom element; perform text context analysis and matching on the medical history element to obtain the medical history value corresponding to the medical history element.
[0031] S203, determining the weight values corresponding to each basic data key-value pair, symptom key-value pair, and medical history key-value pair based on the preset mapping table.
[0032] Among them, the preset mapping table stores the weight values corresponding to all basic data elements, symptom elements, and medical history elements respectively. The weight value is determined according to the degree of influence of each data on cerebral infarction. For example, if "age" in the basic data element has a relatively high influence on the severity of cerebral infarction, the weight value of the basic data element "age" can be set higher.
[0033] S204, determining the eigenvector matrix of cerebral infarction evaluation data through the multiple basic data key-value pairs, symptom key-value pairs, medical history key-value pairs and their corresponding weight values.
[0034] Among them, the cerebral infarction evaluation data characteristic vector matrix includes 3 columns of data, the data elements in the first column are the basic data values and weight values corresponding to the basic data elements, the data elements in the second column are the symptom values and weight values corresponding to the symptom elements, and the data elements in the third column are the medical history values and weight values corresponding to the medical history elements.
[0035] S205: Determine a first cerebral infarction recognition result of the user according to the cerebral infarction evaluation data feature vector matrix and the user local image.
[0036] In an optional embodiment provided in the present application, determining the first cerebral infarction recognition result of the user according to the cerebral infarction evaluation data feature vector matrix and the user partial image includes: S2051, inputting the cerebral infarction evaluation data feature vector matrix and the user's local image into a first cerebral infarction recognition model, and extracting color features and texture features of the user's local image through a feature extraction module in the first cerebral infarction recognition model; S2052, fusing the color feature and the texture feature through the feature fusion module in the first cerebral infarction recognition model to obtain a local image fusion feature; S2053, obtaining a first cerebral infarction recognition result of the user according to the local image fusion feature and the cerebral infarction evaluation data feature vector matrix prediction, wherein the first cerebral infarction recognition result includes a predicted cerebral infarction grade and a predicted probability value.
[0037] Among them, the cerebral infarction grade is used to evaluate the severity of cerebral infarction. The higher the grade, the more severe the cerebral infarction.
[0038] It should be noted that the training process of the first cerebral infarction recognition model in this embodiment is: obtain the sample feature vector matrix and sample local image of cerebral infarction evaluation data corresponding to each sample data; the sample local image includes a facial image and a tongue coating image; input the sample feature vector matrix and the sample local image into the first cerebral infarction recognition model, and extract the color features and texture features corresponding to the facial image and the tongue coating image respectively through the feature extraction module in the first cerebral infarction recognition model; fuse the facial image and the tongue coating image respectively corresponding to the color features and the texture features through the feature fusion module in the first cerebral infarction recognition model to obtain facial image fusion features and tongue coating image fusion features; obtain the first prediction value through the sample feature vector matrix and the facial image fusion features; obtain the second prediction value through the sample feature vector matrix and the tongue coating image fusion features; calculate the loss value according to the first prediction value and the second prediction value; when the loss value is less than the preset value, stop training the first cerebral infarction recognition model. Among them, the preset value is a value set according to actual needs.
[0039] Specifically, calculating the loss value according to the first predicted value and the second predicted value includes: calculating the loss value by the following formula: ; in, is the loss value, is the first loss value obtained by the first prediction value, is the second loss value obtained by the second prediction value, is the observation noise parameter.
[0040] S206: Determine a second cerebral infarction recognition result of the user based on the cerebral infarction evaluation data feature vector matrix and the user medical image.
[0041] In an optional embodiment provided in the present application, determining the second cerebral infarction recognition result of the user according to the cerebral infarction evaluation data feature vector matrix and the user medical image includes: S2061, inputting the cerebral infarction evaluation data feature vector matrix and the user medical image into a second cerebral infarction recognition model, and extracting the infarction site feature and the infarction area feature of the user medical image through the feature extraction module in the second cerebral infarction recognition model; S2062, fusing the infarction site feature and the infarction area feature through the feature fusion module in the second cerebral infarction recognition model to obtain a medical image fusion feature; S2063, predicting a second cerebral infarction recognition result for the user based on the medical image fusion features and the cerebral infarction evaluation data feature vector matrix, wherein the second cerebral infarction recognition result includes a predicted cerebral infarction grade and a predicted probability value.
[0042] It should be noted that the training process of the second cerebral infarction recognition model in the present embodiment is: obtaining the sample feature vector matrix of cerebral infarction evaluation data and the sample medical image corresponding to each sample data; inputting the sample feature vector matrix of cerebral infarction evaluation data and the sample medical image into the second cerebral infarction recognition model, and extracting the infarction site features and the infarction area features of the sample medical image through the feature extraction module in the second cerebral infarction recognition model; fusing the infarction site features and the infarction area features through the feature fusion module in the second cerebral infarction recognition model to obtain the medical image fusion features; obtaining the predicted value through the sample feature vector matrix and the medical image fusion features; calculating the loss value according to the predicted value and the label corresponding to the sample data; when the loss value is less than the preset value, stopping the training of the first cerebral infarction recognition model.
[0043] In an optional embodiment provided in the application, before training the first cerebral infarction recognition model or the second cerebral infarction recognition model, it is necessary to process the sample data in the sample data set, and the processing process of the sample data is as follows: obtaining a sample data set for training a cerebral infarction recognition model, each sample data in the sample data set includes user basic data, symptom text data, medical history text data, user local image and user medical image; determining a sample feature vector matrix of cerebral infarction evaluation data based on the user basic data, symptom text data and medical history text data; performing image smoothing based on median filtering on the user local image, and performing image smoothing based on the formula The gradient of the image in both horizontal and vertical directions is used to approximate the gradient of each pixel in the image to obtain a gradient image. is the horizontal gradient of the image, is the vertical gradient of the image; the obtained gradient image is used as a sample local image; the user medical image is converted to the same size, and the image is smoothed based on median filtering to obtain a sample medical image.
[0044] The method of determining the eigenvector matrix of the cerebral infarction evaluation data samples is the same as that of determining the eigenvector matrix of the cerebral infarction evaluation data, which will not be described in detail in this embodiment.
[0045] S207: Determine a final cerebral infarction recognition result of the user through the first cerebral infarction recognition result and the second cerebral infarction recognition result.
[0046] In an optional embodiment provided in the present application, determining the final cerebral infarction recognition result of the user through the first cerebral infarction recognition result and the second cerebral infarction recognition result includes: S2071, determining whether the predicted probability values corresponding to the first cerebral infarction recognition result and the second cerebral infarction recognition result are both greater than a preset probability value; S2072: If both are greater than a preset probability value, weighted calculation is performed on the predicted cerebral infarction levels of the first cerebral infarction recognition result and the second cerebral infarction recognition result to obtain a final cerebral infarction recognition result for the user; S2073, if there is a predicted probability value greater than the preset probability value, the predicted cerebral infarction level with the predicted probability value greater than the preset probability value is used as the final cerebral infarction recognition result of the user; S2074: If there is no predicted probability value greater than the preset probability value, the final cerebral infarction recognition result of the user is obtained by calculation based on the basic data key-value pair, the symptom key-value pair, and the medical history key-value.
[0047] An embodiment of the present application provides a method for identifying critical and severe cerebral infarction, which obtains user basic data, symptom text data, medical history text data, user local images uploaded by the user through a user client and user medical images obtained through a data acquisition device, performs text extraction on the user basic data, symptom text data and medical history text data respectively, and determines multiple basic data key-value pairs, symptom key-value pairs, and medical history key-value pairs; determines the weight values corresponding to each basic data key-value pair, symptom key-value pair, and medical history key-value pair based on a preset mapping table, determines a cerebral infarction evaluation data feature vector matrix through multiple basic data key-value pairs, symptom key-value pairs, medical history key-value pairs and their corresponding weight values, and then determines a first cerebral infarction recognition result of the user according to the cerebral infarction evaluation data feature vector matrix and the user local image, and determines a second cerebral infarction recognition result of the user according to the cerebral infarction evaluation data feature vector matrix and the user medical image, and finally determines a final cerebral infarction recognition result of the user through the first cerebral infarction recognition result and the second cerebral infarction recognition result. Compared with the prior art in which doctors diagnose critical and severe cerebral infarction by visually observing the patient's brain scan images, the present application generates a cerebral infarction evaluation data feature vector matrix based on data uploaded by the user client and the data acquisition device, and then determines the user's final cerebral infarction recognition result based on the first cerebral infarction recognition result and the second cerebral infarction recognition result generated by the cerebral infarction evaluation data feature vector matrix and related images. Therefore, the present application can improve the recognition efficiency and accuracy of critical and severe cerebral infarction.
[0048] It should be understood that the order of execution of the steps in the above embodiment does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.
[0049] In one embodiment, a cerebral infarction identification device is provided, and the cerebral infarction identification device corresponds one-to-one with the cerebral infarction critical and severe disease identification method in the above embodiment. Figure 3 As shown, the functional modules of the cerebral infarction identification device are described in detail as follows: The acquisition module 31 is used to acquire the user basic data, symptom text data, medical history text data and user partial image uploaded by the user through the user client; and the user medical image acquired by the data acquisition device; the user partial image includes the user's facial image and tongue coating image; The text extraction module 32 is used to perform text extraction on the user basic data, the symptom text data and the medical history text data respectively, and determine a plurality of basic data key-value pairs, symptom key-value pairs and medical history key-value pairs; the basic data key-value pairs include basic data elements and their corresponding basic data values, the symptom key-value pairs include symptom elements and their corresponding symptom values, and the medical history key-value pairs include medical history elements and their corresponding medical history values; A determination module 33 is used to determine the weight values corresponding to each basic data key-value pair, symptom key-value pair, and medical history key-value pair based on a preset mapping table, wherein the preset mapping table stores the weight values corresponding to all basic data elements, symptom elements, and medical history elements; The determination module 33 is further used to determine the cerebral infarction evaluation data feature vector matrix through the multiple basic data key-value pairs, symptom key-value pairs, medical history key-value pairs and their corresponding weight values, wherein the cerebral infarction evaluation data feature vector matrix includes 3 columns of data, the data elements in the first column are the basic data values and weight values corresponding to the basic data elements, the data elements in the second column are the symptom values and weight values corresponding to the symptom elements, and the data elements in the third column are the medical history values and weight values corresponding to the medical history elements; The determination module 33 is further configured to determine a first cerebral infarction recognition result of the user according to the cerebral infarction evaluation data feature vector matrix and the user partial image; The determination module 33 is further used to determine the second cerebral infarction recognition result of the user according to the cerebral infarction evaluation data feature vector matrix and the user medical image; The determination module 33 is further configured to determine a final cerebral infarction recognition result of the user according to the first cerebral infarction recognition result and the second cerebral infarction recognition result.
[0050] In an optional embodiment provided by the present invention, the determination module 33 is specifically configured to: Performing keyword extraction on the user basic data, the symptom text data, and the medical history text data to obtain a basic keyword set, a symptom keyword set, and a medical history keyword set; Based on the basic data word library, synonym matching is performed on the keywords in the basic keyword set to obtain basic data elements; based on the symptom word library, synonym matching is performed on the keywords in the symptom keyword set to obtain symptom elements; based on the medical history word library, synonym matching is performed on the keywords in the medical history keyword set to obtain medical history elements; the basic data word library stores basic keywords for determining cerebral infarction; the symptom word library stores symptom keywords for determining cerebral infarction, and the medical history word library stores medical history keywords for determining cerebral infarction; Perform text context analysis and matching on the user basic data to obtain the basic data value corresponding to the basic data element; perform text context analysis and matching on the symptom text data to obtain the symptom value corresponding to the symptom element; perform text context analysis and matching on the medical history element to obtain the medical history value corresponding to the medical history element.
[0051] In an optional embodiment provided by the present invention, the determination module 33 is specifically configured to: Inputting the cerebral infarction evaluation data feature vector matrix and the user's partial image into a first cerebral infarction recognition model, and extracting color features and texture features of the user's partial image through a feature extraction module in the first cerebral infarction recognition model; The color feature and the texture feature are fused by the feature fusion module in the first cerebral infarction recognition model to obtain a local image fusion feature; A first cerebral infarction recognition result of a user is obtained according to the local image fusion feature and the cerebral infarction evaluation data feature vector matrix prediction, and the first cerebral infarction recognition result includes a predicted cerebral infarction grade and a predicted probability value.
[0052] In an optional embodiment provided by the present invention, the determination module 33 is specifically configured to: Inputting the cerebral infarction evaluation data feature vector matrix and the user medical image into a second cerebral infarction recognition model, and extracting the infarction site features and infarction area features of the user medical image through a feature extraction module in the second cerebral infarction recognition model; fusing the infarction site feature and the infarction area feature through the feature fusion module in the second cerebral infarction recognition model to obtain a medical image fusion feature; A second cerebral infarction recognition result of the user is predicted based on the medical image fusion features and the cerebral infarction evaluation data feature vector matrix, and the second cerebral infarction recognition result includes a predicted cerebral infarction grade and a predicted probability value.
[0053] In an optional embodiment provided by the present invention, the determination module 33 is specifically configured to: Determining whether the predicted probability values corresponding to the first cerebral infarction recognition result and the second cerebral infarction recognition result are both greater than a preset probability value; If both are greater than a preset probability value, weighted calculation is performed on the predicted cerebral infarction levels of the first cerebral infarction recognition result and the second cerebral infarction recognition result to obtain a final cerebral infarction recognition result for the user; If there is a predicted probability value greater than the preset probability value, the predicted cerebral infarction level with the predicted probability value greater than the preset probability value is used as the user's final cerebral infarction recognition result.
[0054] In an optional embodiment provided by the present invention, the determination module 33 is specifically configured to: If there is no predicted probability value greater than the preset probability value, the final cerebral infarction recognition result of the user is obtained by calculation based on the basic data key-value pair, the symptom key-value pair, and the medical history key-value.
[0055] In an optional embodiment provided by the present invention, the cerebral infarction identification device further includes a training module 34 for training a first cerebral infarction identification model. The training module 34 is specifically used for: Obtaining a sample feature vector matrix of cerebral infarction evaluation data and a sample local image corresponding to each piece of sample data; the sample local image includes a facial image and a tongue coating image; Inputting the sample feature vector matrix and the sample local image into the first cerebral infarction recognition model, and extracting the color features and texture features corresponding to the facial image and the tongue coating image respectively through the feature extraction module in the first cerebral infarction recognition model; By using the feature fusion module in the first cerebral infarction recognition model, the facial image and the tongue coating image are respectively fused corresponding to the color features and the texture features to obtain a facial image fusion feature and a tongue coating image fusion feature; Obtain a first prediction value by fusing features of the data sample feature vector matrix and the facial image; obtain a second prediction value by fusing features of the data sample feature vector matrix and the tongue coating image; Calculate a loss value according to the first predicted value and the second predicted value; When the loss value is less than a preset value, the training of the first cerebral infarction recognition model is stopped.
[0056] In an optional embodiment provided by the present invention, the training module 34 is specifically used for: The loss value is calculated by the following formula: ; in, is the loss value, is the first loss value obtained by the first prediction value, is the second loss value obtained by the second prediction value, is the observation noise parameter.
[0057] In an optional embodiment provided by the present invention, the training module 34 is further used to: Obtain the cerebral infarction evaluation data sample feature vector matrix and sample medical image corresponding to each sample data; Inputting the sample feature vector matrix of the cerebral infarction evaluation data and the sample medical image into a second cerebral infarction recognition model, and extracting the infarction site features and infarction area features of the sample medical image through a feature extraction module in the second cerebral infarction recognition model; fusing the infarction site feature and the infarction area feature through the feature fusion module in the second cerebral infarction recognition model to obtain a medical image fusion feature; Obtaining a prediction value by fusing the sample feature vector matrix and the medical image features; A loss value is calculated according to the predicted value and the label corresponding to the sample data; when the loss value is less than a preset value, the training of the first cerebral infarction recognition model is stopped.
[0058] In an optional embodiment provided by the present invention, the cerebral infarction identification device further includes a data processing module 35, and the data processing module 35 is specifically used for: Acquire a sample data set for training a cerebral infarction recognition model, wherein each piece of sample data in the sample data set includes user basic data, symptom text data, medical history text data, user local images, and user medical images; Determine a cerebral infarction evaluation data sample feature vector matrix according to the user basic data, symptom text data, and medical history text data; The user local image is smoothed based on median filtering, and the formula The gradient of the image in both horizontal and vertical directions is used to approximate the gradient of each pixel in the image to obtain a gradient image. is the horizontal gradient of the image, is the vertical gradient of the image; the obtained gradient image is used as the sample local image; The user medical image is converted into the same size, and image smoothing based on median filtering is performed to obtain a sample medical image.
[0059] For the specific limitations of the cerebral infarction identification device, please refer to the limitations of the cerebral infarction critical and severe identification method above, which will not be repeated here. Each module in the above-mentioned device can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.
[0060] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.
[0061] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention.
Claims
1. A system for identifying critical and severe cerebral infarction, characterized in that: The system comprises: a user client, a data acquisition device and a cerebral infarction identification device, wherein the cerebral infarction identification device is connected to the user client and the data acquisition device, and the cerebral infarction identification device is used to perform the following steps: Acquire user basic data, symptom text data, medical history text data and user partial images uploaded by the user through the user client; and the user medical image acquired by the data acquisition device; the user partial image includes the user's facial image and tongue coating image; Performing text extraction on the user basic data, the symptom text data and the medical history text data respectively, and determining a plurality of basic data key-value pairs, symptom key-value pairs and medical history key-value pairs; the basic data key-value pairs include basic data elements and their corresponding basic data values, the symptom key-value pairs include symptom elements and their corresponding symptom values, and the medical history key-value pairs include medical history elements and their corresponding medical history values; Determine the weight values corresponding to each basic data key-value pair, symptom key-value pair, and medical history key-value pair based on a preset mapping table, wherein the preset mapping table stores the weight values corresponding to all basic data elements, symptom elements, and medical history elements; Determine the cerebral infarction evaluation data feature vector matrix through the multiple basic data key-value pairs, symptom key-value pairs, medical history key-value pairs and their corresponding weight values, wherein the cerebral infarction evaluation data feature vector matrix includes 3 columns of data, the data elements in the first column are the basic data values and weight values corresponding to the basic data elements, the data elements in the second column are the symptom values and weight values corresponding to the symptom elements, and the data elements in the third column are the medical history values and weight values corresponding to the medical history elements; Determine a first cerebral infarction recognition result of the user according to the cerebral infarction evaluation data feature vector matrix and the user partial image; Determine a second cerebral infarction recognition result of the user according to the cerebral infarction evaluation data feature vector matrix and the user medical image; The final cerebral infarction recognition result of the user is determined by the first cerebral infarction recognition result and the second cerebral infarction recognition result.
2. The system according to claim 1, characterized in that The step of performing text extraction on the user basic data, the symptom text data, and the medical history text data respectively to determine a plurality of basic data key-value pairs, symptom key-value pairs, and medical history key-value pairs includes: Performing keyword extraction on the user basic data, the symptom text data, and the medical history text data to obtain a basic keyword set, a symptom keyword set, and a medical history keyword set; Based on the basic data word library, synonym matching is performed on the keywords in the basic keyword set to obtain basic data elements; based on the symptom word library, synonym matching is performed on the keywords in the symptom keyword set to obtain symptom elements; based on the medical history word library, synonym matching is performed on the keywords in the medical history keyword set to obtain medical history elements; the basic data word library stores basic keywords for determining cerebral infarction; the symptom word library stores symptom keywords for determining cerebral infarction, and the medical history word library stores medical history keywords for determining cerebral infarction; Perform text context analysis and matching on the user basic data to obtain the basic data value corresponding to the basic data element; perform text context analysis and matching on the symptom text data to obtain the symptom value corresponding to the symptom element; perform text context analysis and matching on the medical history element to obtain the medical history value corresponding to the medical history element.
3. The system according to claim 2, characterized in that The determining the first cerebral infarction recognition result of the user according to the cerebral infarction evaluation data feature vector matrix and the user partial image includes: Inputting the cerebral infarction evaluation data feature vector matrix and the user's partial image into a first cerebral infarction recognition model, and extracting color features and texture features of the user's partial image through a feature extraction module in the first cerebral infarction recognition model; The color feature and the texture feature are fused by the feature fusion module in the first cerebral infarction recognition model to obtain a local image fusion feature; A first cerebral infarction recognition result of a user is obtained according to the local image fusion feature and the cerebral infarction evaluation data feature vector matrix prediction, and the first cerebral infarction recognition result includes a predicted cerebral infarction grade and a predicted probability value.
4. The system according to claim 2, characterized in that The determining the second cerebral infarction recognition result of the user according to the cerebral infarction evaluation data feature vector matrix and the user medical image comprises: Inputting the cerebral infarction evaluation data feature vector matrix and the user medical image into a second cerebral infarction recognition model, and extracting the infarction site features and infarction area features of the user medical image through a feature extraction module in the second cerebral infarction recognition model; fusing the infarction site feature and the infarction area feature through the feature fusion module in the second cerebral infarction recognition model to obtain a medical image fusion feature; A second cerebral infarction recognition result of the user is predicted based on the medical image fusion features and the cerebral infarction evaluation data feature vector matrix, and the second cerebral infarction recognition result includes a predicted cerebral infarction grade and a predicted probability value.
5. The system according to claim 3 or 4, characterized in that: The determining the final cerebral infarction recognition result of the user by using the first cerebral infarction recognition result and the second cerebral infarction recognition result includes: Determining whether the predicted probability values corresponding to the first cerebral infarction recognition result and the second cerebral infarction recognition result are both greater than a preset probability value; If both are greater than a preset probability value, weighted calculation is performed on the predicted cerebral infarction levels of the first cerebral infarction recognition result and the second cerebral infarction recognition result to obtain a final cerebral infarction recognition result for the user; If there is a predicted probability value greater than the preset probability value, the predicted cerebral infarction level with the predicted probability value greater than the preset probability value is used as the user's final cerebral infarction recognition result.
6. The system according to claim 5, characterized in that After determining whether the predicted probability values corresponding to the first cerebral infarction recognition result and the second cerebral infarction recognition result are both greater than the preset probability values, the method further includes: If there is no predicted probability value greater than the preset probability value, the final cerebral infarction recognition result of the user is obtained by calculation based on the basic data key-value pair, the symptom key-value pair, and the medical history key-value.
7. The system according to claim 3, characterized in that The training process of the first cerebral infarction recognition model is as follows: Obtaining a sample feature vector matrix of cerebral infarction evaluation data and a sample local image corresponding to each piece of sample data; the sample local image includes a facial image and a tongue coating image; Inputting the sample feature vector matrix and the sample local image into the first cerebral infarction recognition model, and extracting the color features and texture features corresponding to the facial image and the tongue coating image respectively through the feature extraction module in the first cerebral infarction recognition model; By using the feature fusion module in the first cerebral infarction recognition model, the facial image and the tongue coating image are respectively fused corresponding to the color features and the texture features to obtain a facial image fusion feature and a tongue coating image fusion feature; Obtain a first prediction value by fusing features of the data sample feature vector matrix and the facial image; obtain a second prediction value by fusing features of the data sample feature vector matrix and the tongue coating image; Calculating a loss value according to the first predicted value and the second predicted value; When the loss value is less than a preset value, the training of the first cerebral infarction recognition model is stopped.
8. The system according to claim 4, characterized in that The training process of the second cerebral infarction recognition model is as follows: Obtain the cerebral infarction evaluation data sample feature vector matrix and sample medical image corresponding to each sample data; Inputting the sample feature vector matrix of the cerebral infarction evaluation data and the sample medical image into a second cerebral infarction recognition model, and extracting the infarction site features and infarction area features of the sample medical image through a feature extraction module in the second cerebral infarction recognition model; fusing the infarction site feature and the infarction area feature through the feature fusion module in the second cerebral infarction recognition model to obtain a medical image fusion feature; Obtaining a prediction value by fusing the sample feature vector matrix and the medical image features; A loss value is calculated according to the predicted value and the label corresponding to the sample data; when the loss value is less than a preset value, the training of the first cerebral infarction recognition model is stopped.
9. The system according to claim 8, characterized in that Before training the first cerebral infarction recognition model or the second cerebral infarction recognition model, the method further includes: Acquire a sample data set for training a cerebral infarction recognition model, wherein each piece of sample data in the sample data set includes user basic data, symptom text data, medical history text data, user local images, and user medical images; Determine a cerebral infarction evaluation data sample feature vector matrix according to the user basic data, symptom text data, and medical history text data; Performing image smoothing based on median filtering on the user local image, and approximating the gradient of each pixel in the image by formulas for the gradients in the horizontal and vertical directions of the image to obtain a gradient image; and using the obtained gradient image as a sample local image; The user medical image is converted into the same size, and image smoothing based on median filtering is performed to obtain a sample medical image.