CN107403061A - User's medical assessment model building method and medical assessment server - Google Patents

Abstract

The present invention provides a user medical evaluation model construction method and a medical evaluation server, the model includes: Step 101: Construct sample data U and sample input data V, the sample data U is fuzzy transformed to obtain the sample input data V, V =Ω.U, where Ω is the weight vector of fuzzy quantization; Step 102: Based on the clinical diagnosis results of the sample data U, construct the sample output result y, Step 103: Establish the neural network mapping between the sample input data V and the sample output result y relationship; step 104: train the neural network of the model; use the sample input data V as the neural network input of the model, and the sample output result y as the neural network output of the model; learn through sample training , to determine the weights in the neural network of the model. The user medical evaluation model construction method and the medical evaluation server of the present invention, combined with the data collected by the smart terminal, can provide users with portable and reliable medical diagnosis services.

Description

User medical evaluation model construction method and medical evaluation server

technical field

The invention relates to the field of computers, in particular to a method for constructing a medical evaluation model and a medical evaluation server.

Background technique

With the improvement of people's living standards, chronic diseases and complications caused by unreasonable living and eating patterns are gradually affecting people's health. In addition, with the intensification of aging and the increase of the elderly population, the health of the elderly group is also affected by the chronic diseases of the elderly. Medicine shows that the exacerbation of these chronic diseases can be detected in advance by monitoring the physiological data indicators of the human body, so it is very meaningful to use the intelligent medical evaluation system to collect and evaluate physiological data.

There are mainly two technical solutions for the construction of the existing intelligent medical evaluation system. They are a single wearable device monitoring system, and an evaluation monitoring system that combines wearable devices with smart phones and other devices.

The first system technical solution is mainly wearable devices such as intelligent postoperative monitoring patches, smart watches or bracelets, whose peripheral sensors collect the user's physiological sign data and display them on the product screen. The cost of this technical solution is lower, but the medical evaluation is not high and the accuracy is poor.

The second system technical solution is mainly that several sensors on the smart watch or bracelet collect the user's physiological sign data, and send the collected data to the user's smart phone and other devices through Bluetooth or wirelessly. The application on the smart phone adopts Simple algorithmic analysis evaluates the results of each data acquisition. Users can call to view recent collected data and corresponding evaluation results. This technical solution improves the accuracy of medical evaluation and analysis results, and can also store data for a period of time as a reference data change trend. However, it is not possible to comprehensively analyze and evaluate a large amount of collected data for a long time, and the computing power is not enough to conduct information fusion analysis on the collected data, and the evaluation results lack the accuracy of medical application level.

The above schemes show that the current smart terminal cannot be used for medical diagnosis. For patients with chronic diseases or potential chronic disease patients, how to combine smart terminals to enjoy reliable medical services conveniently has not yet proposed an effective solution.

Contents of the invention

The invention provides a method for constructing a user's medical evaluation model and a medical evaluation server, which can provide users with portable and reliable medical diagnosis services in combination with data collected by an intelligent terminal.

The present invention provides a method for constructing a user medical assessment model, comprising:

Step 101: Construct sample data U and sample input data V. U includes ECG eigenvalue data, blood oxygen data SpO2, heart rate data HR, and blood pressure data BP. ECG eigenvalue data includes: RR interval, QRS Wave interval, PR interval, QT interval, ST segment, R wave amplitude, P wave amplitude and T wave amplitude; the sample data U is fuzzy transformed to obtain the sample input data V, V=Ω.U, where Ω is fuzzy quantization weight vector of

Step 102: Based on the clinical diagnosis results of the sample data U, construct a sample output result y,

Step 103: Establish a neural network mapping relationship between the sample input data V and the sample output result y;

Step 104: Train the neural network of the model; use the sample input data V as the input of the neural network of the model, and the output result y of the sample as the output of the neural network of the model; determine the weights in the neural network of the model through sample training and learning.

The present invention also provides a medical evaluation server, including: a data module and a user medical evaluation module.

The data module regularly collects and saves the physiological sign data collected by the user terminal. The physiological sign data includes at least the electrocardiographic signal ECG, blood oxygen data SpO2, heart rate data HR and blood pressure data BP;

User medical assessment module: use the user medical assessment model of the present invention to analyze the user data saved by the data module, and judge the user's risk of disease according to the output results of the model.

The user medical evaluation model construction method provided by the present invention can combine the physiological sign data collected by the user terminal to output reliable medical diagnosis results, and the model can self-learn to improve the accuracy of the output results. The medical evaluation server adopting the medical evaluation model can provide users with accurate, reliable and convenient medical services. Through the output of the model, it can automatically warn of risks and seek medical treatment or the server actively notifies doctors to provide medical services, instead of seeking medical treatment after existing physical discomfort, so that patients can receive timely treatment in the early stage or acute stage of disease, avoiding delaying the best treatment period and delaying the condition. , reduce the cost of patient treatment, and improve the experience and timeliness of existing medical services.

Description of drawings

Fig. 1 is a flow chart of the method for constructing the user's medical evaluation model of the present invention;

Fig. 2 is the flow chart of real-time revision of the user's medical evaluation model of the present invention;

Fig. 3 is a schematic structural diagram of the medical evaluation server of the present invention.

detailed description

In order to make the object, technical solution and advantages of the present invention clearer, the present invention will be described in detail below in conjunction with the accompanying drawings and specific embodiments.

The user medical evaluation model construction method of the present invention, as shown in Figure 1, at least includes:

Step 101: Construct sample data U and sample input data V.

Sample data U includes ECG eigenvalue data, blood oxygen data SpO2, heart rate data HR, blood pressure data BP, ECG eigenvalue data includes: RR interval, QRS wave interval, PR interval, QT interval , ST segment, R wave amplitude, P wave amplitude and T wave amplitude; the sample data U is fuzzy transformed to obtain the sample input data V, V=Ω.U, where Ω is the weight vector of fuzzy quantization.

Step 102: Based on the clinical diagnosis results of the sample data U, construct a sample output result y.

For example, the empirical judgment risk y of user sample data can be defined as three types of normal, low risk and high risk, represented by y ₀ , y ₁ , and y ₂ respectively, or the risk level can be further refined.

Step 103: and establish a neural network mapping relationship between the sample input data V and the sample output result y.

Step 104: Train the neural network model of the medical evaluation model; use the sample input data V as the neural network input of the user's medical evaluation model, and the sample output result y as the neural network output of the user's medical evaluation model; through sample training and learning, determine the user Weights in a neural network for a medical assessment model.

The user medical evaluation model construction method of this application can train and build different user medical evaluation models for each user’s sample data, because each user’s physiological signs have their own characteristics, and a certain value of the same physiological sign index , for some users, it is a healthy value, while for other users, it is a dangerous value. Therefore, the user medical evaluation model of this application can realize personalized customization, so that the accuracy of the model is higher than that of general statistical models. high.

After the user medical evaluation model is established based on steps 101 to 103, the model can be used for real-time evaluation; the newly collected user data U' is fuzzy converted into V', and after inputting into the model, the model outputs the result y', According to the comparison between the output result y' and the sample output result y, the risk of the user's disease is judged.

Furthermore, the user medical model of this application also has a self-learning function, which can continuously modify and improve the model based on the real-time evaluation results of the model, as shown in FIG. 2 .

Model correction: compare the model output result y' of the newly collected user data with the clinical diagnosis result, if the output result y' does not match the clinical diagnosis result, calculate the output error as the numerical difference between the output result y' and the diagnosis result, The output error is passed back layer by layer to the input layer through the hidden layer of the neural network of the user's medical evaluation model, and the output error is distributed to all units of each layer of the neural network of the medical evaluation algorithm.

Fig. 3 is another embodiment of the present application, a medical assessment server at least including a data module and a medical assessment module.

The data module regularly collects and saves the physiological sign data collected by the user terminal. The physiological sign data includes at least the electrocardiographic signal ECG, blood oxygen data SpO2, heart rate data HR and blood pressure data BP;

User medical evaluation module: use the user medical evaluation model of any embodiment of the application to analyze the user data saved by the data module to determine the user's risk of disease.

In order to improve the accuracy of the data stored in the data module, before the data is saved, the physiological sign data collected by the user is preprocessed, including the elimination of abnormal data and non-user physiological sign data, so as to ensure input into the user medical evaluation model data consistency and stability.

The medical server of this application saves the user's physiological sign data, packs them into data packets in a certain format, and stores them in the data module. The data packet format is encoded according to the format of packet header plus data, where the packet header includes start bit, device number, time stamp, physiological signs and status data name and check digit. The start bit is used to distinguish each data packet; the device number is used to distinguish different devices that display the user; the timestamp is used to record the collection time; the physiological sign data name is mainly used to distinguish the physiological sign data collected by different sensors and the different states of the user Information, such as distinguishing whether the collected data is ECG data or heart rate data; the check digit is used to verify whether the data transmission is wrong.

The medical evaluation server of this application also includes an online doctor. When the user data saved in the data module is small, the online doctor reviews the output results of the user medical evaluation model, and feeds back the audit results to the user medical evaluation model for use in the user medical evaluation model. real-time corrections.

Or the online doctor provides relevant medical diagnosis suggestions to the user according to the output results of the user's medical evaluation model and the data stored in the data module, and sends the diagnosis suggestions to the user terminal. When the user's medical assessment model can independently output reliable judgment results, the server can also directly send the output result of the user's medical assessment model to the user terminal.

If the output result of the user's medical evaluation model is "high risk", the server automatically turns on the real-time monitoring module, the user terminal uploads the user's physiological sign data in real time, and the server monitors the user data in real time. Measures to provide users with medical services.

The medical evaluation model construction method provided by the present invention can combine the physiological sign data collected by the user terminal to output reliable medical diagnosis results, and the model can self-learn to improve the accuracy of the output results. The medical evaluation server adopting the medical evaluation model can provide users with accurate, reliable and convenient medical services. Through the output of the model, it can automatically warn of risks and seek medical treatment or the server actively notifies doctors to provide medical services, instead of seeking medical treatment after existing physical discomfort, so that patients can receive timely treatment in the early stage or acute stage of disease, avoiding delaying the best treatment period and delaying the condition. , reduce the cost of patient treatment, and improve the experience and timeliness of existing medical services.

The above descriptions are only preferred embodiments of the present invention, and are not intended to limit the scope of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the technical solutions of the present invention are Should be included within the protection scope of the present invention.

Claims (7)

1. A user medical assessment model construction method, characterized in that the construction of the model at least includes: Step 101: Construct sample data U and sample input data V, the U includes ECG eigenvalue data, blood oxygen data SpO2, heart rate data HR, blood pressure data BP, and the ECG eigenvalue data includes: RR Interval, QRS interval, PR interval, QT interval, ST segment, R wave amplitude, P wave amplitude and T wave amplitude; the sample data U is fuzzy transformed to obtain the sample input data V, V=Ω .U, where Ω is the weight vector of fuzzy quantization; Step 102: Based on the clinical diagnosis results of the sample data U, construct a sample output result y, Step 103: Establish a neural network mapping relationship between the sample input data V and the sample output result y; Step 104: train the neural network of the model; use the sample input data V as the input of the neural network of the model, and the output result y of the sample as the output of the neural network of the model; determine through sample training and learning The weights in the neural network for the model. 2. The model according to claim 1, wherein the model further comprises: Model correction: Fuzzy transform the newly collected user data U' into V', input it into the model trained in step 103, and output the result y', if the output result y' does not match the clinical diagnosis result, calculate The output error is the numerical difference between the output result y' and the clinical diagnosis result, and the output error is passed back layer by layer to the input layer through the hidden layer of the neural network of the model, and the output error is apportioned to all All units of each layer of the neural network of the above model. 3. A medical evaluation server, characterized in that it at least includes a data module and a user medical evaluation module; The data module regularly collects and saves the physiological sign data collected by the user terminal, and the physiological sign data includes at least the electrocardiographic signal ECG, blood oxygen data SpO2, heart rate data HR and blood pressure data BP; User medical evaluation module: using the user medical evaluation model described in any one of claims 1-2, analyzing the user data saved by the data module, and judging the user's risk of disease according to the output results of the model. 4. The server according to claim 3, wherein when the user data saved by the data module is less, the server sends the output result of the user medical assessment model to a doctor for review, and sends the The results of the doctor's review are fed back to the user's medical evaluation model. 5. The server according to claim 3, wherein the server sends the output result of the user's medical evaluation model and the data stored by the data module to a doctor for diagnosis, provides relevant medical diagnosis suggestions to the user, and sending the suggestion; or the server directly sending the output result of the user's medical evaluation model. 6. The server according to claim 3, wherein the server also includes a real-time monitoring module, When the output result of the user's medical evaluation model is "high risk", the server automatically turns on the real-time monitoring module, the user terminal uploads the user's physiological sign data in real time, and the server monitors the user data in real time. The above server sends a notification immediately. 7. The server according to claim 3, wherein when the data module saves data, it is stored in a preset data packet format; The preset data packet format is encoded according to the format of header plus data, and the header includes start bit, device number, time stamp, physiological signs and status data name and check digit; the start bit is used to distinguish each segment data packet; the device number is used to distinguish and display different devices of the user; the timestamp is used to record the collection time; the physiological sign data name is used to distinguish the physiological sign data collected by different sensors and the different status information of the user, so The above parity bit is used to check whether there is an error in data transmission.