Diabetes risk early warning system and method

By obtaining and analyzing information, signs and facial images of diabetic patients, using machine learning and deep learning technology to establish a diabetes risk warning model, solving the problem of insufficient early warning for unoccupied populations, realizing early warning and effective defense for diabetic patients, and improving The timeliness and accuracy of monitoring and early warning are provided.

CN119028596AActive Publication Date: 2024-11-26BAIZHITANG TRADITIONAL CHINESE MEDICINE HOSPITAL (CHENGDU) CO LTD

Patent Information

Application Number
CN202411525703.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-10-30
Publication Date
2024-11-26
Estimated Expiration
2044-10-30

Smart Images

  • Figure CN119028596A_ABST
    Figure CN119028596A_ABST
Patent Text Reader

Abstract

The invention discloses a diabetes risk early warning system and method, and the method comprises the steps: obtaining known patient information, sign information, face images and case information, generating a patient database, dividing the patient database into a first database and a second database, the patients in the second database are non-diabetic patients; acquiring patient information, case information and sign information of diabetic patients from the first database, outputting disease control decision information, acquiring patient information, case information and sign information of non-diabetic patients from the second database, and outputting prevention control decision information; monitoring physical sign information and facial images of the patient in real time, and outputting illness state control feedback information; receiving the illness state control feedback information, and updating illness state control decision information and prevention control decision information in real time; illness state control feedback information of the patient is obtained in real time through the early warning model, and timeliness and accuracy of illness state monitoring are improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of diabetes data processing, and in particular to a diabetes risk early warning system and method. Background Art

[0002] In recent years, diabetes has become a major public health problem that seriously endangers the health of the nation. It also brings a heavy economic burden to society. Prevention and treatment work is imminent, and it is of great significance to conduct individualized data monitoring of the population to prevent diabetes.

[0003] At present, a Chinese invention patent with application publication number CN116189896A discloses a cloud-based diabetes health data early warning method, which retrieves the onset symptom characteristics and reference physical sign data of different types of diabetes at various stages from the database and stores them in the cloud server; detects the physical sign parameters of the target person in real time through the portable wearable device of the target person, and uploads the real-time physical sign parameters to the cloud server; determines whether the physical sign parameters are abnormal, and sets a physical sign monitoring schedule for the target person according to the judgment result; reminds the target person to perform periodic physical sign measurements according to the physical sign monitoring schedule, and obtains the physical sign measurement results; compares the physical sign measurement results with the reference physical sign data of different types of diabetes, determines the diabetes type and stage of the target person according to the comparison results and issues an early warning, and can detect physical sign indicators for the patient without affecting the patient's own time arrangement and make accurate judgments based on the measurement results; however, there is no early warning of diabetes for the non-ill-patient population in the related technology, which is not conducive to early prevention, obtaining the case characteristics of the existing ill population, and performing periodic monitoring and analysis. The monitoring lacks timeliness and accuracy, and the analysis lacks comprehensiveness and comprehensiveness. Summary of the invention

[0004] The technical problem solved by the present invention is that the related technology does not provide early warning for diabetes in the non-ill population, which is not conducive to early prevention, obtaining case characteristics of the ill population, and conducting periodic monitoring and analysis. The monitoring lacks timeliness and accuracy, and the analysis lacks comprehensiveness and comprehensiveness.

[0005] In order to overcome the shortcomings of the prior art, the present invention provides a diabetes risk early warning method, comprising the following steps: Step S1, obtaining known patient information, physical sign information, facial images and case information, generating a patient database, and dividing the patient database into a first database and a second database, wherein the patients in the first database are diabetic patients, and the patients in the second database are non-diabetic patients; Step S2, acquiring patient information, case information and physical sign information of diabetic patients according to the first database, and outputting disease control decision information according to the patient information, case information and physical sign information of diabetic patients, acquiring patient information, case information and physical sign information of non-diabetic patients according to the second database, and outputting prevention and control decision information according to the patient information, case information and physical sign information of non-diabetic patients; Step S3, real-time monitoring of the patient's vital signs and facial images, and output of disease control feedback information; Step S4, receiving the disease control feedback information, and updating the disease control decision information and the prevention control decision information in real time; As a preferred embodiment of the diabetes risk early warning method of the present invention, wherein: The patient information includes name, age, gender, height, blood pressure, blood sugar, weight, eating habits, existing diseases and genetic diseases; The case information includes symptoms, causes, types of illness and stages of illness; The physical sign information includes blood sugar information, heart rate information, activity information and sleep information; As a preferred embodiment of the diabetes risk early warning method of the present invention, wherein: The step S1 specifically includes: Obtain known patient information, vital signs, facial images and case information; Determine whether the patient is diabetic based on key indicators in the physical sign information; If the patient is a diabetic patient, storing the patient information, physical sign information, facial image and case information of the patient in the first database; If the patient is a non-diabetic patient, storing the patient information, physical sign information, facial image and case information of the patient in a second database; As a preferred embodiment of the diabetes risk early warning method of the present invention, wherein: The step S2 specifically includes: Inputting patient information of diabetic patients into a first database, matching case information and physical sign information of diabetic patients in the first database, using case information and physical sign information as input of a first machine learning model, using condition control decision information as output of the first machine learning model, training the first machine learning model until the condition control decision information is diagnosed by an expert with 100% accuracy, and completing the training of the first machine learning model; Inputting patient information of non-diabetic patients into a second database, matching case information and physical sign information of non-diabetic patients in the second database, using case information and physical sign information as input of a second machine learning model, using prevention and control decision information as output of the second machine learning model, and training the second machine learning model until the prevention and control decision information is diagnosed by an expert with 100% accuracy, thereby completing the training of the second machine learning model; As a preferred embodiment of the diabetes risk early warning method of the present invention, wherein: The step S3 specifically includes: Calling a first database, obtaining patient information, physical sign information, facial images and case information corresponding to known diabetic patients, obtaining the type of diabetes corresponding to the corresponding case information, identifying and extracting keywords of the case information corresponding to the diabetic patients, building a keyword joint model based on deep learning and updating joint parameters in real time, wherein the input of the keyword joint model is the keyword, and the output is the first comprehensive phrase; Calling the second database to obtain case information, patient information and case information corresponding to non-diabetic patients, building a keyword joint model based on deep learning and updating joint parameters in real time, the input of the keyword joint model is the keyword, and the output is the keyword related to the keyword; Acquire case information, patient information and physical sign information of diabetic patients from the first database, and case information, patient information and physical sign information of non-diabetic patients from the second database; Calling the first database and the second database, taking the patient information, physical sign information and case information of the diabetic patients and non-diabetic patients as data sets respectively, training the data sets using a machine learning model, wherein the input of the training is sentences in the patient information and physical sign information, and the output is case information, taking the case information as a first-level sentence, wherein the first-level sentence includes contextual semantic information and sentence-level features, encoding and annotating the first-level sentence, generating a second-level sentence, and converting the second-level sentence into a third-level sentence with the same length as the first-level sentence and a fixed and settable dimension; Retrieve the third-level sentence corresponding to the second database, extract the keywords in the third-level sentence, calculate the similarity between the corresponding keywords in the second database and the corresponding keywords in the data set, and compare the similarity with the similarity threshold: If the similarity is greater than the similarity threshold, the keywords are merged to generate a keyword group; If the similarity is less than the similarity threshold, the keyword is returned to the third-level sentence to generate a third database; As a preferred embodiment of the diabetes risk early warning method of the present invention, wherein: The third database includes patient information and corresponding keyword groups, keywords and facial images; Matching the patient information in the third database with the facial image one by one; As a preferred embodiment of the diabetes risk early warning method of the present invention, wherein: The step S3 further comprises: Real-time monitoring of patients' vital signs and facial images; The real-time collected blood sugar information, heart rate information, activity information and sleep information are used as keywords in the first database and the second database, input into the keyword joint model, and the keywords related to the keywords are output; Input the keywords and keywords related to the keywords into the machine learning model to establish the first diabetes risk warning model; Retrieving the facial image in the third database, matching the facial image of the patient using face recognition technology, and obtaining the patient information corresponding to the facial image; Using image recognition technology, extract and analyze the features of the patient's facial image to obtain keywords and keyword groups corresponding to the features of the patient's facial image; Input keywords and keyword groups corresponding to the patient's facial image features into the machine learning model to establish a second diabetes risk warning model; The first diabetes risk warning model is fused with the second diabetes risk warning model by using model fusion technology according to the patient information to obtain a diabetes risk warning model; Input the patient's physical information and facial image to obtain feedback on the patient's condition control; The disease control feedback information includes symptoms, disease type and disease stage; As a preferred embodiment of the diabetes risk early warning method of the present invention, wherein: The real-time monitoring of the patient's vital signs and facial images includes: Use a blood glucose meter to monitor the patient's blood glucose and obtain the patient's blood glucose information; Use a heart rate monitoring bracelet to monitor the patient's heart rate and obtain the patient's heart rate information; Use the accelerometer and pedometer in the smart bracelet to monitor the patient's physical activity and obtain the patient's activity information; Use smart bracelets to monitor patients' sleep and obtain their sleep information; As a preferred embodiment of the diabetes risk early warning method of the present invention, wherein: The step S4 specifically includes: Obtain the patient's disease control feedback information, and determine whether the patient is a diabetic patient based on the key disease indicators in the disease control feedback information; If the patient is diabetic, the disease control feedback information is input into the machine learning model to obtain disease control decision information. After the disease control decision information is confirmed by experts, the disease control decision information is output and updated; If the patient is not a diabetic, the disease control feedback information is input into the machine learning model to obtain prevention and control decision information. After confirmation by experts, the prevention and control decision information is output and updated.

[0006] In a second aspect, a diabetes risk early warning system includes a monitoring module, an early warning module and a decision-making module; The monitoring module is used to monitor and obtain the patient's vital sign information and facial image; The early warning module is used to establish an early warning model based on known patient information, physical sign information, facial image and case information, and use the early warning model to obtain patient condition control feedback information based on the patient's physical sign information and facial image; The decision-making module is used to comprehensively analyze the patient's condition control feedback to obtain the condition control decision information of diabetic patients and the prevention and control decision information of non-diabetic patients.

[0007] Beneficial effects of the present invention: The present invention establishes an early warning model based on known patient information, physical sign information, case information and facial images, monitors the patient's physical sign information and facial images in real time, and uses the early warning model to obtain the patient's condition control feedback information, thereby improving the timeliness and accuracy of condition monitoring.

[0008] By establishing the first diabetes risk warning model and the second diabetes risk warning model, and fusing the two models to obtain the warning model, the patient's condition control feedback information is obtained based on the patient's blood sugar information, heart rate information, activity information, sleep information and facial images, and by conducting a comprehensive analysis of the patient's various physical data, the accuracy and comprehensiveness of the diabetes risk warning are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Figure 1 A basic flow chart of a diabetes risk early warning method provided by one embodiment of the present invention; Figure 2 A basic flow chart of a diabetes risk warning system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0010] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, but not all of the embodiments.

[0011] Example 1, reference Figure 1, as an embodiment of the present invention, provides a diabetes risk early warning method, comprising the following steps: Step S1, obtaining known patient information, physical sign information, facial images and case information, generating a patient database, and dividing the patient database into a first database and a second database, wherein the patients in the first database are diabetic patients, and the patients in the second database are non-diabetic patients; Step S2, acquiring patient information, case information and physical sign information of diabetic patients according to the first database, and outputting disease control decision information according to the patient information, case information and physical sign information of diabetic patients, acquiring patient information, case information and physical sign information of non-diabetic patients according to the second database, and outputting prevention and control decision information according to the patient information, case information and physical sign information of non-diabetic patients; Step S3, real-time monitoring of the patient's vital signs and facial images, and output of disease control feedback information; Step S4, receiving the disease control feedback information, and updating the disease control decision information and the prevention control decision information in real time; The acquiring of patient information, case information and physical sign information of non-diabetic patients according to the second database comprises: The patient's vital sign information is collected, and whether the patient is a diabetic patient is determined based on key indicators of the patient's vital sign information. If the patient is a non-diabetic patient, the patient information, case information and physical sign information of the corresponding non-diabetic patient are matched in the second database according to the physical sign identifier of the characteristic information. If the corresponding non-diabetic patient is not matched in the second database, the patient information, case information and physical sign information of the patient are collected and stored together in the second database as new patient data of the non-diabetic patient. The physical sign identifier is a unique identifier used to internally distinguish each diabetic patient.

[0012] In this embodiment, an early warning model is established based on known patient information, physical sign information, case information and facial images, the patient's physical sign information and facial images are monitored in real time, and the patient's condition control feedback information is obtained using the early warning model, thereby improving the timeliness and accuracy of condition monitoring; By establishing the first diabetes risk warning model and the second diabetes risk warning model, and fusing the two models to obtain the warning model, the patient's condition control feedback information is obtained based on the patient's blood sugar information, heart rate information, activity information, sleep information and facial image. By comprehensively analyzing the patient's various physical conditions, the accuracy and comprehensiveness of the diabetes risk warning are improved. The patient information includes name, age, gender, height, blood pressure, blood sugar, weight, eating habits, existing diseases and genetic diseases; The case information includes symptoms, causes, types of illness and stages of illness; The physical sign information includes blood sugar information, heart rate information, activity information and sleep information; In this embodiment, accurate and comprehensive data support is provided for establishing a machine learning model by obtaining the patient's patient information, case information, and physical sign information; The step S1 specifically includes: Obtain known patient information, vital signs, facial images and case information; Determine whether the patient is diabetic based on key indicators in the physical sign information; If the patient is a diabetic patient, storing the patient information, physical sign information, facial image and case information of the patient in the first database; If the patient is a non-diabetic patient, storing the patient information, physical sign information, facial image and case information of the patient in a second database; In this embodiment, the patient data is divided into diabetic patient data and non-diabetic patient data according to the key indicators in the physical sign information, and stored in the first database and the second database respectively, which provides accurate data support for establishing the disease control decision-making machine learning model and the prevention and control decision-making machine learning model; The step S2 specifically includes: Inputting patient information of diabetic patients into a first database, matching case information and physical sign information of diabetic patients in the first database, using case information and physical sign information as input of a first machine learning model, using condition control decision information as output of the first machine learning model, training the first machine learning model until the condition control decision information is diagnosed by an expert with 100% accuracy, and completing the training of the first machine learning model; Inputting patient information of non-diabetic patients into a second database, matching case information and physical sign information of non-diabetic patients in the second database, using case information and physical sign information as input of a second machine learning model, using prevention and control decision information as output of the second machine learning model, and training the second machine learning model until the prevention and control decision information is diagnosed by an expert with 100% accuracy, thereby completing the training of the second machine learning model; The machine learning model training includes: Data preparation: Prepare the diabetic patient data set and the non-diabetic patient data set that need to be trained, clean and preprocess the data, including missing value processing, outlier processing and duplicate value processing, and divide the cleaned and processed data set into training data set and test data set; Model selection: Select appropriate machine learning algorithms and models. Commonly used algorithms include linear regression, logistic regression, decision tree, random forest, support vector machine, and neural network. Model training: Use the training data set to train the selected machine learning model, and use optimization algorithms such as gradient descent to iteratively update the machine learning model parameters to minimize the loss function; Model evaluation: During the machine learning model training process, the machine learning model is evaluated to determine the performance of the machine learning model. Evaluation indicators usually include precision, recall rate, and F1 value. Model parameter adjustment: According to the results of machine learning model evaluation, the machine learning model parameters are adjusted to further improve the performance of the machine learning model; Model preservation and deployment: After the machine learning model training is completed, the trained machine learning model is saved to make decisions on disease control decision information and prevention and control decision information; In this embodiment, by performing machine learning training based on the data of diabetic patients and non-diabetic patients, the condition control decision information of diabetic patients and the prevention and control decision information of non-diabetic patients are obtained through the machine learning model, and the diagnosis and confirmation are performed by experts, thereby improving the accuracy of diabetes condition control and prevention and control and saving medical resources; The step S3 specifically includes: Calling a first database, obtaining patient information, physical sign information, facial images and case information corresponding to known diabetic patients, obtaining the type of diabetes corresponding to the corresponding case information, identifying and extracting keywords of the case information corresponding to the diabetic patients, building a keyword joint model based on deep learning and updating joint parameters in real time, wherein the input of the keyword joint model is the keyword, and the output is the first comprehensive phrase; Calling the second database to obtain case information, patient information and case information corresponding to non-diabetic patients, building a keyword joint model based on deep learning and updating joint parameters in real time, the input of the keyword joint model is the keyword, and the output is the keyword related to the keyword; Acquire case information, patient information and physical sign information of diabetic patients from the first database, and case information, patient information and physical sign information of non-diabetic patients from the second database; Calling the first database and the second database, taking the patient information, physical sign information and case information of the diabetic patients and non-diabetic patients as data sets respectively, training the data sets using a machine learning model, wherein the input of the training is sentences in the patient information and physical sign information, and the output is case information, taking the case information as a first-level sentence, wherein the first-level sentence includes contextual semantic information and sentence-level features, encoding and annotating the first-level sentence, generating a second-level sentence, and converting the second-level sentence into a third-level sentence with the same length as the first-level sentence and a fixed and settable dimension; Retrieve the third-level sentence corresponding to the second database, extract the keywords in the third-level sentence, calculate the similarity between the corresponding keywords in the second database and the corresponding keywords in the data set, and compare the similarity with the similarity threshold: If the similarity is greater than the similarity threshold, the keywords are merged to generate a keyword group; If the similarity is less than the similarity threshold, the keyword is returned to the third-level sentence to generate a third database; In this embodiment, a keyword joint model is constructed to provide a comprehensive and accurate model input for constructing a first diabetes risk warning model, and a third database is established to provide an accurate and comprehensive model input for constructing a second diabetes risk warning model; The third database includes patient information and corresponding keyword groups, keywords and facial images; Matching the patient information in the third database with the facial image one by one; In this embodiment, by making a one-to-one correspondence between the patient information and the facial image, corresponding fusion parameters are provided for fusing the first diabetes risk warning model and the second diabetes warning model to obtain a warning model; The step S3 further comprises: Real-time monitoring of patients' vital signs and facial images; The real-time collected blood sugar information, heart rate information, activity information and sleep information are used as keywords in the first database and the second database, input into the keyword joint model, and the keywords related to the keywords are output; Input the keywords and keywords related to the keywords into the machine learning model to establish the first diabetes risk warning model; Retrieving the facial image in the third database, matching the facial image of the patient using face recognition technology, and obtaining the patient information corresponding to the facial image; Using image recognition technology, extract and analyze the features of the patient's facial image to obtain keywords and keyword groups corresponding to the features of the patient's facial image; Input keywords and keyword groups corresponding to the patient's facial image features into the machine learning model to establish a second diabetes risk warning model; The first diabetes risk warning model is fused with the second diabetes risk warning model by using model fusion technology according to the patient information to obtain a diabetes risk warning model; Input the patient's physical information and facial image to obtain feedback on the patient's condition control; The disease control feedback information includes symptoms, disease type and disease stage; The face recognition technology includes: Image quality assessment: After acquiring the image, the image quality is assessed to filter out low-quality images and reduce noise and errors in subsequent processing; Face detection: Use face detection algorithms to automatically locate the face area in the image, ensuring that subsequent processing is focused on the face area.

[0013] Face alignment: The acquisition of face images may vary due to different shooting angles and postures. In the preprocessing process, a common method is to align the face so that the positions of feature points such as eyes, nose and mouth in the image remain consistent, thereby reducing the difficulty of subsequent recognition; Image enhancement: Enhance the image, such as image denoising, image contrast enhancement, histogram equalization, etc., to improve image quality and enhance facial features; Feature extraction: After preprocessing, a feature extraction algorithm is used to convert facial features in the image into mathematical vector representations to facilitate subsequent recognition and comparison; The method for extracting and analyzing features of the patient's facial image includes a HOG facial feature extraction method and a DeepFace facial feature extraction method; The HOG facial feature extraction method comprises: Image preprocessing: grayscale conversion, size adjustment, background removal and other operations are performed on the input image to facilitate subsequent feature extraction; Calculate gradient: perform gradient calculation on the preprocessed image to obtain the gradient map of the image; Partitioning: Divide the gradient map into multiple unit areas, each of which contains a certain number of gradients; Calculate the direction histogram: perform directional statistics on the gradient in each unit area to obtain a direction histogram; Normalization: Normalize the direction histogram to facilitate subsequent feature matching and comparison; The DeepFace facial feature extraction method includes: Data preprocessing: grayscale conversion, size adjustment, background removal and other operations are performed on the input image to facilitate subsequent feature extraction; Convolutional layers: Use convolutional layers to learn low-level features of images; Pooling layer: Use the pooling layer to reduce the spatial resolution of the image, reduce the number of parameters and computational complexity; Fully connected layer: Use fully connected layers to learn high-level features; Output layer: Use the output layer to output image features; In this embodiment, the warning model is obtained by integrating the first diabetes risk warning model and the second diabetes risk warning model, so that the disease control feedback information obtained by using the warning model is more comprehensive and accurate; The real-time monitoring of the patient's vital signs and facial images includes: Use a blood glucose meter to monitor the patient's blood glucose and obtain the patient's blood glucose information; Use a heart rate monitoring bracelet to monitor the patient's heart rate and obtain the patient's heart rate information; Use the accelerometer and pedometer in the smart bracelet to monitor the patient's physical activity and obtain the patient's activity information; Use smart bracelets to monitor patients' sleep and obtain their sleep information; The step S4 specifically includes: Obtain the patient's disease control feedback information, and determine whether the patient is a diabetic patient based on the key disease indicators in the disease control feedback information; If the patient is diabetic, the disease control feedback information is input into the machine learning model to obtain disease control decision information. After the disease control decision information is confirmed by experts, the disease control decision information is output and updated; If the patient is not a diabetic, the disease control feedback information is input into the machine learning model to obtain the prevention and control decision information. After confirmation by experts, the prevention and control decision information is output and updated. In this embodiment, a machine learning model is used to obtain condition control decision information and preventive control decision information based on condition control feedback information, thereby improving the timeliness and comprehensiveness of diabetes risk warning.

[0014] Example 2, reference Figure 2 , which is another embodiment of the present invention, and which is different from the first embodiment in that it provides a urology remote control method, including the above-mentioned diabetes risk early warning method, including a monitoring module, an early warning module and a decision module; The monitoring module is used to monitor and obtain the patient's vital sign information and facial image; The early warning module is used to establish an early warning model based on known patient information, physical sign information, facial image and case information, and use the early warning model to obtain patient condition control feedback information based on the patient's physical sign information and facial image; The decision module is used to comprehensively analyze the patient's disease control feedback, and obtain disease control decision information for diabetic patients and prevention and control decision information for non-diabetic patients; In this embodiment, the monitoring module can be used to monitor and obtain the patient's vital signs information and facial images in real time, the early warning module can be used to obtain the patient's condition control feedback information based on the patient's vital signs information and facial images, and the decision-making module can be used to update the condition control decision information and prevention and control decision information based on the patient's condition control feedback information, thereby improving the accuracy, timeliness and comprehensiveness of diabetes risk warning.

[0015] It should be appreciated that embodiments of the present invention may be implemented or enforced by a combination of computer hardware and software, or by computer instructions stored in a non-transitory computer-readable memory. The method may be implemented in a computer program using standard programming techniques - including a non-transitory computer-readable storage medium configured with a computer program, wherein the storage medium so configured causes the computer to operate in a specific and predefined manner - according to the methods and drawings described in the specific embodiments. Each program may be implemented in a high-level procedural or object-oriented programming language to communicate with a computer system. However, if desired, the program may be implemented in an assembly or machine language. In any case, the language may be a compiled or interpreted language. In addition, the program may be run on a programmed ASIC for this purpose.

[0016] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A diabetes risk early warning method, characterized in that: The steps include: Step S1, obtaining known patient information, physical sign information, facial images and case information, generating a patient database, and dividing the patient database into a first database and a second database, wherein the patients in the first database are diabetic patients, and the patients in the second database are non-diabetic patients; Step S2, acquiring patient information, case information and physical sign information of diabetic patients according to the first database, and outputting disease control decision information according to the patient information, case information and physical sign information of diabetic patients, acquiring patient information, case information and physical sign information of non-diabetic patients according to the second database, and outputting prevention and control decision information according to the patient information, case information and physical sign information of non-diabetic patients; Step S3, real-time monitoring of the patient's vital signs and facial images, and output of disease control feedback information; Step S4, receiving the disease control feedback information, and updating the disease control decision information and the prevention control decision information in real time.

2. A diabetes risk early warning method according to claim 1, characterized in that: The patient information includes name, age, gender, height, blood pressure, blood sugar, weight, eating habits, existing diseases and genetic diseases; The case information includes symptoms, causes, types of illness and stages of illness; The vital sign information includes blood sugar information, heart rate information, activity information and sleep information.

3. A diabetes risk early warning method according to claim 1, characterized in that: The step S1 specifically includes: Obtain known patient information, vital signs, facial images and case information; Determine whether the patient is diabetic based on key indicators in the physical sign information; If the patient is a diabetic patient, storing the patient information, physical sign information, facial image and case information of the patient in the first database; If the patient is a non-diabetic patient, the patient information, physical sign information, facial image and case information of the patient are stored in the second database.

4. A diabetes risk early warning method according to claim 1, characterized in that: The step S2 specifically includes: Inputting patient information of diabetic patients into a first database, matching case information and physical sign information of diabetic patients in the first database, using case information and physical sign information as input of a first machine learning model, using condition control decision information as output of the first machine learning model, training the first machine learning model until the condition control decision information is diagnosed by an expert with 100% accuracy, and completing the training of the first machine learning model; Input the patient information of non-diabetic patients into the second database, match the case information and physical sign information of non-diabetic patients in the second database, use the case information and physical sign information as the input of the second machine learning model, use the prevention and control decision information as the output of the second machine learning model, and train the second machine learning model until the prevention and control decision information is diagnosed by experts with 100% accuracy, thereby completing the training of the second machine learning model.

5. A diabetes risk early warning method according to claim 1, characterized in that: The step S3 specifically includes: Calling the first database, obtaining patient information, physical sign information, facial images and case information corresponding to known diabetic patients, obtaining the type of diabetes corresponding to the case information, identifying and extracting keywords of the case information corresponding to the diabetic patients, building a keyword joint model based on deep learning and updating the joint parameters in real time, the input of the keyword joint model is the keyword, and the output is the first comprehensive phrase; Calling the second database to obtain case information, patient information and case information corresponding to non-diabetic patients, building a keyword joint model based on deep learning and updating joint parameters in real time, the input of the keyword joint model is the keyword, and the output is the keyword related to the keyword; Acquire case information, patient information and physical sign information of diabetic patients from the first database, and case information, patient information and physical sign information of non-diabetic patients from the second database; Calling the first database and the second database, taking the patient information, physical sign information and case information of the diabetic patients and non-diabetic patients as data sets respectively, training the data sets using a machine learning model, wherein the input of the training is sentences in the patient information and physical sign information, and the output is case information, taking the case information as a first-level sentence, wherein the first-level sentence includes contextual semantic information and sentence-level features, encoding and annotating the first-level sentence, generating a second-level sentence, and converting the second-level sentence into a third-level sentence with the same length as the first-level sentence and a fixed and settable dimension; Retrieve the third-level sentence corresponding to the second database, extract the keywords in the third-level sentence, calculate the similarity between the corresponding keywords in the second database and the corresponding keywords in the data set, and compare the similarity with the similarity threshold: If the similarity is greater than the similarity threshold, the keywords are merged to generate a keyword group; If the similarity is less than the similarity threshold, the keyword is returned to the third-level sentence to generate a third database.

6. A diabetes risk early warning method according to claim 5, characterized in that: The third database includes patient information and corresponding keyword groups, keywords and facial images; The patient information in the third database is matched with the facial image one by one.

7. A diabetes risk early warning method according to claim 1, characterized in that: The step S3 further comprises: Real-time monitoring of patients' vital signs and facial images; The real-time collected blood sugar information, heart rate information, activity information and sleep information are used as keywords in the first database and the second database, input into the keyword joint model, and the keywords related to the keywords are output; Input the keywords and keywords related to the keywords into the machine learning model to establish the first diabetes risk warning model; Retrieving the facial image in the third database, matching the facial image of the patient using face recognition technology, and obtaining the patient information corresponding to the facial image; Using image recognition technology, extract and analyze the features of the patient's facial image to obtain keywords and keyword groups corresponding to the features of the patient's facial image; Input keywords and keyword groups corresponding to the patient's facial image features into the machine learning model to establish a second diabetes risk warning model; The first diabetes risk warning model is fused with the second diabetes risk warning model by using model fusion technology according to the patient information to obtain a diabetes risk warning model; Input the patient's physical information and facial image to obtain feedback on the patient's condition control; The disease control feedback information includes symptoms, disease type and disease stage.

8. A diabetes risk early warning method according to claim 1, characterized in that: The real-time monitoring of the patient's vital signs and facial images includes: Use a blood glucose meter to monitor the patient's blood glucose and obtain the patient's blood glucose information; Use a heart rate monitoring bracelet to monitor the patient's heart rate and obtain the patient's heart rate information; Use the accelerometer and pedometer in the smart bracelet to monitor the patient's physical activity and obtain the patient's activity information; Use smart bracelets to monitor patients' sleep and obtain their sleep information.

9. A diabetes risk early warning method according to claim 1, characterized in that: The step S4 specifically includes: Obtain the patient's disease control feedback information, and determine whether the patient is a diabetic patient based on the key disease indicators in the disease control feedback information; If the patient is diabetic, the disease control feedback information is input into the machine learning model to obtain disease control decision information. After the disease control decision information is confirmed by experts, the disease control decision information is output and updated; If the patient is not a diabetic, the disease control feedback information is input into the machine learning model to obtain prevention and control decision information. After confirmation by experts, the prevention and control decision information is output and updated.

10. A diabetes risk early warning system, comprising a diabetes risk early warning method according to any one of claims 1 to 9, characterized in that: Including monitoring module, early warning module and decision-making module; The monitoring module is used to monitor and obtain the patient's vital sign information and facial image; The early warning module is used to establish an early warning model based on known patient information, physical sign information, facial image and case information, and use the early warning model to obtain patient condition control feedback information based on the patient's physical sign information and facial image; The decision-making module is used to comprehensively analyze the patient's condition control feedback to obtain the condition control decision information of diabetic patients and the prevention and control decision information of non-diabetic patients.

Citation Information

Patent Citations

  • Cloud-based diabetes health data early warning method and system

    CN116189896A

  • Personalized device recommendations for proactive health monitoring and management

    CN112204671A

  • DSS-based gestational diabetes patient management method and system, terminal and medium

    CN116130093A

  • Remote health management service method based on big data

    CN117958805A

  • Syndrome monitoring and early warning method and apparatus, computer device, and storage medium

    WO2022217713A1

Cited By

  • Diabetes risk assessment method and system

    CN120977592A

  • A method and system for assessing risk of diabetes

    CN120977592B