Blood glucose prediction method and device, wearable equipment, storage medium and program product
The PPG signal was obtained through photovoltaic pulse wave schema, combined with reference blood glucose values and training models, and the problem of low efficiency of traditional blood glucose monitoring is solved, achieving efficient and safe blood glucose prediction.
Patent Information
- Application Number
- CN202510326238.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-19
- Publication Date
- 2025-07-04
AI Technical Summary
Traditional blood sugar monitoring methods have low blood collection efficiency through acupuncture through blood glucose meter, making it difficult to achieve efficient blood sugar monitoring.
The PPG signal was obtained by photovoltaic pulse wave gramography (PPG), combined with the pre-configured reference blood glucose value and the trained blood glucose prediction model, multiple target characteristic values were extracted, and the blood glucose prediction value was calculated through the model.
It improves the accuracy and convenience of blood sugar monitoring, avoids acupuncture and blood collection, and enhances monitoring efficiency.
Smart Images
Figure CN120241053A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of blood glucose prediction, and particularly to a blood glucose prediction method, device, wearable device, storage medium and program product. Background Art
[0002] Diabetes is a metabolic disorder disease that endangers human health, characterized by long-term abnormal blood glucose concentration. As more and more people understand the harm of diabetes, blood glucose monitoring has become an important means of health management. For example, the general population can understand their own blood glucose levels in a timely manner through blood glucose monitoring, so as to manage their lifestyle and avoid inducing diabetes. Another example is that diabetic patients can monitor their blood glucose levels in a timely manner to delay complications. In traditional methods, blood glucose monitoring is usually achieved by using a blood glucose meter to puncture and collect blood from parts such as fingertips to detect blood glucose.
[0003] However, the method of detecting blood glucose by puncturing and collecting blood with a blood glucose meter has low blood glucose monitoring efficiency. Summary of the Invention
[0004] Based on this, in view of the above technical problems, it is necessary to provide a blood glucose prediction method, device, wearable device, storage medium and program product that can improve the blood glucose monitoring efficiency.
[0005] In a first aspect, the present application provides a blood glucose prediction method, including:
[0006] Obtaining the collected PPG signal, and obtaining the pre-configured reference blood glucose value for the target object to which the PPG signal belongs;
[0007] According to the reference blood glucose value, determining the target object type to which the target object belongs among different pre-configured preset object types;
[0008] Determining the trained target blood glucose prediction model applicable to the target object type, and determining a plurality of target features applicable to the target blood glucose prediction model;
[0009] Based on the PPG signal, extracting the feature values of the respective target features;
[0010] Using the target blood glucose prediction model, determining the blood glucose prediction value of the target object according to the feature values of the respective target features.
[0011] In a second aspect, the present application further provides a blood glucose prediction device, including:
[0012] An acquisition module, configured to obtain the collected PPG signal, and obtain the pre-configured reference blood glucose value for the target object to which the PPG signal belongs;
[0013] An object type determination module, configured to determine the target object type to which the target object belongs from different pre-configured preset object types according to the reference blood glucose value;
[0014] A prediction module, configured to determine a trained target blood glucose prediction model applicable to the target object type, and determine a plurality of target features applicable to the target blood glucose prediction model; extract the feature values of the respective target features based on the PPG signal; and determine the blood glucose prediction value of the target object according to the feature values of the respective target features through the target blood glucose prediction model.
[0015] In a third aspect, the present application further provides a wearable device, including a memory and a processor, where the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:
[0016] Obtain the collected PPG signal, and obtain the reference blood glucose value pre-configured for the target object to which the PPG signal belongs;
[0017] Determine the target object type to which the target object belongs from different pre-configured preset object types according to the reference blood glucose value;
[0018] Determine a trained target blood glucose prediction model applicable to the target object type, and determine a plurality of target features applicable to the target blood glucose prediction model;
[0019] Extract the feature values of the respective target features based on the PPG signal;
[0020] Determine the blood glucose prediction value of the target object according to the feature values of the respective target features through the target blood glucose prediction model.
[0021] In a fourth aspect, the present application further provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:
[0022] Obtain the collected PPG signal, and obtain the reference blood glucose value pre-configured for the target object to which the PPG signal belongs;
[0023] Determine the target object type to which the target object belongs from different pre-configured preset object types according to the reference blood glucose value;
[0024] Determine a trained target blood glucose prediction model applicable to the target object type, and determine a plurality of target features applicable to the target blood glucose prediction model;
[0025] Extract the feature values of the respective target features based on the PPG signal;
[0026] Based on the respective feature values of the multiple target features by means of the target blood glucose prediction model, determine the blood glucose prediction value of the target object.
[0027] In a fifth aspect, the present application further provides a computer program product, including a computer program which, when executed by a processor, implements the following steps:
[0028] Obtain the collected PPG signal, and obtain the reference blood glucose value pre-configured for the target object to which the PPG signal belongs;
[0029] Based on the reference blood glucose value, determine the target object type to which the target object belongs among different pre-configured preset object types;
[0030] Determine the trained target blood glucose prediction model applicable to the target object type, and determine the multiple target features applicable to the target blood glucose prediction model;
[0031] Based on the PPG signal, extract the respective feature values of the multiple target features;
[0032] Based on the respective feature values of the multiple target features by means of the target blood glucose prediction model, determine the blood glucose prediction value of the target object.
[0033] For the above blood glucose prediction method, device, wearable device, storage medium and program product, by obtaining the collected PPG signal, based on the PPG signal, extracting the respective feature values of the multiple target features applicable to the target blood glucose prediction model, and by means of the target blood glucose prediction model, based on the respective feature values of the multiple target features, determining the blood glucose prediction value of the target object to which the PPG signal belongs. Since the multiple target features are specifically applicable to the target blood glucose prediction model, the target blood glucose prediction model is specifically applicable to the target object type to which the target object belongs, and the target object type is obtained based on the reference blood glucose value pre-configured for the target object, the accuracy of the blood glucose prediction value of the target object can be improved. Moreover, the PPG signal is collected based on the optoelectronic principle. Compared with the method of using a blood glucose meter to prick the finger for blood glucose detection, for blood glucose prediction by collecting the PPG signal based on the optoelectronic principle, there is no need to prick the finger of the target object for blood collection, the collection method is safer and does not require manual operation, improving the convenience of blood glucose prediction, and thus the efficiency of blood glucose monitoring can be improved. Description of the Drawings
[0034] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following will briefly introduce the drawings required to be used in the description of the embodiments of the present application or related technologies. Obviously, the following drawings are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other related drawings can also be obtained based on these drawings.
[0035] Figure 1 is a schematic flowchart of a blood glucose prediction method in an embodiment;
[0036] Figure 2 is a schematic flowchart of the steps for training a blood glucose prediction model in an embodiment;
[0037] Figure 3 is a structural block diagram of a blood glucose prediction device in an embodiment;
[0038] Figure 4 is an internal structure diagram of a wearable device in an embodiment. Detailed implementation manners
[0039] In order to make the objectives, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.
[0040] In one embodiment, as Figure 1 shown, a blood glucose prediction method is provided. In this embodiment, the application of this method to a wearable device is taken as an example for illustration. The wearable device may be a smart watch or a smart bracelet. In this embodiment, the method includes the following steps:
[0041] Step 102, obtain the collected PPG signal, and obtain the pre-configured reference blood glucose value for the target object to which the PPG signal belongs.
[0042] Among them, the PPG signal is a physiological signal collected by the PPG method (Photoplethysmography). The PPG method detects the periodic changes in blood volume in tissues through photoelectric principles. This periodic change is driven by the pulsation of arterial blood caused by heart beats. The PPG signal can be collected by a wearable device. Specifically, the wearable device can be equipped with a PPG sensor. The PPG sensor can include a light source and a photodetector. The light source is used to irradiate light of a specific wavelength onto the skin of the person whose PPG signal needs to be collected. When the light passes through the skin, the light will be absorbed by the glucose molecules in the blood in the subcutaneous tissue, and reflection and scattering will be generated. Then, the intensity change of the reflected light is detected by the photodetector to capture the change in blood volume and obtain the PPG signal. Among them, glucose molecules have characteristic absorption peaks in the near-infrared band or the mid-infrared band (such as 900-1000 nm, 1500-1700 nm). The PPG sensor irradiates the specific wavelength of light in this band onto the skin for detection. The collected PPG signal can reflect the concentration information of glucose molecules in the blood to a certain extent. The skin may be the skin at the fingertips or the wrist. The light generated by the light source may include green light and red light.
[0043] The target object is the object for which blood glucose prediction is required. The target object may specifically be the name, electronic account or other electronic identity bound to the person collected when the PPG signal is collected. It can be understood that the collected person refers to the person for whom the PPG signal is collected. The reference blood glucose value may be a historical blood glucose value of the target object in the recent historical period; it may also be determined by using multiple historical blood glucose values of the target object in the recent historical period, for example, taking the average of the historical blood glucose values of at least three days in the past week as the blood glucose reference value. The historical blood glucose value may be measured by the target object using a blood glucose meter. The reference blood glucose value may also be the blood glucose value estimated by the target object. The reference blood glucose value may also be a standard blood glucose value determined based on information of a preset information type of the target object. The preset information types include age, height, weight, family history of diabetes, eating habits, exercise habits, etc. The standard blood glucose value may be obtained based on a large amount of statistical data, and the statistical data may include the actual blood glucose value of each statistical object, as well as information of the corresponding preset information type.
[0044] Exemplarily, the wearable device can collect PPG signals, determine the target object bound to the wearable device, determine the target object to which the PPG signal belongs, and obtain the reference blood glucose value preconfigured for the target object from the configuration information pre-stored for the target object.
[0045] Among them, the wearable device can collect PPG signals for a preset duration in the case of a blood glucose prediction trigger event. The preset duration can be, for example, 30 seconds, one minute, or others. The blood glucose prediction trigger event can be an automatic trigger event. For example, it is automatically triggered at preset time intervals. The preset time interval can be, for example, 1 hour, 2 hours, or others. The blood glucose prediction trigger event can also be a manual trigger operation. For example, the wearable device can be provided with a virtual or physical blood glucose prediction trigger control, and the blood glucose prediction trigger event is triggered by operating the blood glucose prediction trigger control.
[0046] The configuration information can be stored in the wearable device. The configuration information can also be stored in a computer device, which can be a terminal or a server. Specifically, the terminal can be bound to the wearable device, and a client of the application can run on the terminal. The server can run a server side of the application. The configuration information is stored through the server side or the client of the application, so as to store the reference blood glucose value. The terminal can be a mobile phone, a laptop, a desktop computer, or a tablet computer. The reference blood glucose value can be input by the user. Specifically, the user can use the terminal bound to the wearable device to run the application and input the reference blood glucose value in the application. Then, the reference blood glucose value is stored in the configuration information and synchronized to the wearable device. It can be understood that other health data can also be input in this way, such as age, weight, etc.
[0047] The wearable device can record the position measured by the PPG sensor when the user inputs the reference blood glucose value. During the user's use of the wearable device, when it is detected that the position measured by the PPG sensor changes significantly (for example, switching from the left wrist to the right wrist), the wearable device can prompt the user to re-enter the reference blood glucose value. Specifically, the wearable device can push a message to the bound terminal, and the message can be a prompt message for the user to re-enter the reference blood glucose value.
[0048] Step 104: According to the reference blood glucose value, determine the target object type to which the target object belongs among different pre-configured preset object types.
[0049] Among them, different preset object types can respectively correspond to different preset blood glucose value ranges. Since people with different preset blood glucose value ranges have differences in aspects such as blood flow characteristics, blood vessel status, and autoregulation, when predicting blood glucose based on PPG signals, different preset object types respectively apply different blood glucose prediction models, which can improve the accuracy of blood glucose prediction.
[0050] The different preset object types may include a first preset object type, a second preset object type, and a third preset object type. The first preset object type may correspond to a first preset blood glucose value range, the second preset object type may correspond to a second preset blood glucose value range, and the third preset object type may correspond to a third preset blood glucose value range. The minimum value in the first preset blood glucose value range may be greater than the maximum value in the second preset blood glucose value range, and the minimum value in the second preset blood glucose value range may be greater than the maximum value in the third preset blood glucose value range. In some specific scenarios, the second preset blood glucose value range may be the normal blood glucose value range corresponding to the fasting state, such as 3.9~6.1 mmol / L (millimoles per liter), the first preset blood glucose value range may be the range higher than 6.1 mmol / L, and the third preset blood glucose value range may be the range lower than 3.9 mmol / L.
[0051] Exemplarily, the wearable device may determine the preset blood glucose value range where the reference blood glucose value is located, and determine the preset object type that matches the preset blood glucose value range where the reference blood glucose value is located among the pre-configured different preset object types as the target object type to which the target object belongs.
[0052] Step 106, determine the trained target blood glucose prediction model applicable to the target object type, and determine multiple target features applicable to the target blood glucose prediction model.
[0053] Among them, different preset object types may respectively be applicable to different blood glucose prediction models, and each preset object type may be applicable to one blood glucose prediction model. Different blood glucose prediction models may be obtained by training on a computer device. The wearable device may pre-load the blood glucose prediction models. Among different blood glucose prediction models, the same multiple target features may be applicable, or different multiple target features may be applicable. The blood glucose prediction model may be a multiple linear regression model, a random forest model, a neural network, or others. The multiple target features applicable to the target blood glucose prediction model may be screened from multiple preset features during the process of training the target blood glucose prediction model. The multiple preset features may include time-domain features, frequency-domain features, and non-linear features.
[0054] Exemplarily, the wearable device may obtain the pre-stored model configuration information, and from the model configuration information, according to the identification information of the target object type, query the identification information of the blood glucose prediction model corresponding to the identification information of the target object type, and the identification information of each of the multiple target features. According to the queried identification information of the blood glucose prediction model and the identification information of each of the multiple target features, determine the trained target blood glucose prediction model applicable to the target object type, and the multiple target features applicable to the target blood glucose prediction model.
[0055] Among them, the model configuration information may include the correspondence relationship among the identification information of the preset object type, the identification information of the applicable blood glucose prediction model, and the identification information of each of the applicable multiple target features. The identification information can have a unique identification function. The identification information can be a number, a name, or others.
[0056] Step 108: Based on the PPG signal, extract the eigenvalue of each of the multiple target features.
[0057] Exemplarily, the wearable device can extract the eigenvalue of each of the multiple target features based on the PPG signal according to the preconfigured algorithm corresponding to each of the multiple target features.
[0058] Among them, the preconfigured algorithm can be an eigenvalue extraction algorithm configured based on the meaning of the target feature for the PPG signal. For example, the target feature can be the peak in the time-domain features. For the PPG signal, the meaning of the peak is the peak value of the PPG signal within each cardiac cycle. Therefore, the preconfigured algorithm corresponding to the peak can be to first divide the PPG signal into different cardiac cycles, and then identify the peak value within each cardiac cycle. A cardiac cycle is from the start of one heartbeat to the start of the next heartbeat. A PPG signal can include multiple cardiac cycles.
[0059] In one embodiment, the wearable device can preprocess the PPG signal to obtain the preprocessed PPG signal, and extract the eigenvalue of each of the multiple target features from the preprocessed PPG signal. Among them, the preprocessing can include filtering, noise reduction, baseline correction, or others. The filtering can adopt digital filtering algorithms, wavelet transform algorithms, or others.
[0060] Step 110: Through the target blood glucose prediction model, determine the blood glucose prediction value of the target object according to the eigenvalue of each of the multiple target features.
[0061] Among them, the blood glucose prediction value can be used to provide lifestyle suggestions, such as providing suggestions on diet, exercise, work and rest, etc. For example, by comparing the blood glucose prediction values of the user before and after consuming a certain food, the food that causes the user's blood glucose to rise above the specified value can be found, and a dietary suggestion to reduce the consumption of this food can be given. The blood glucose prediction value can be displayed on the interface of the wearable device. At the same time, the waveform of the PPG signal can be displayed graphically, and the eigenvalue of the target feature can be presented in the waveform.
[0062] Exemplarily, the wearable device inputs the feature values of multiple target features into the target blood glucose prediction model to obtain the blood glucose prediction value of the target object output by the target blood glucose prediction model. For example, the target blood glucose prediction model can be a multiple linear regression model, which can be specifically expressed as a relational expression with blood glucose as the dependent variable and multiple applicable target features as the independent variables, and the coefficient of each target feature is a value determined during the training process of the target blood glucose prediction model. Therefore, by substituting the feature values of multiple target features into the corresponding independent variables of this relational expression respectively, the blood glucose prediction value can be calculated.
[0063] In the above blood glucose prediction method, by acquiring the collected PPG signal, based on the PPG signal, the feature values of multiple target features applicable to the target blood glucose prediction model are extracted, and through the target blood glucose prediction model, according to the feature values of multiple target features, the blood glucose prediction value of the target object to which the PPG signal belongs is determined. Since multiple target features are specifically applicable to the target blood glucose prediction model, the target blood glucose prediction model is specifically applicable to the target object type to which the target object belongs, and the target object type is obtained based on the pre-configured reference blood glucose value of the target object, the accuracy of the blood glucose prediction value of the target object can be improved. Moreover, the PPG signal is collected based on the optoelectronic principle. Compared with the method of using a blood glucose meter to prick the finger for blood sampling to detect blood glucose, predicting blood glucose by collecting the PPG signal based on the optoelectronic principle does not require pricking the finger of the target object for blood sampling, and the collection method is safer and does not require manual operation, improving the convenience of blood glucose prediction, thereby improving the efficiency of blood glucose monitoring.
[0064] In an exemplary embodiment, the trained target blood glucose prediction model is one of the trained blood glucose prediction models applicable to different preset object types. As Figure 2 shown, the steps of training the blood glucose prediction model may include the following steps:
[0065] Step 202, acquire a plurality of sampled PPG signals that are collected, the actual blood glucose value corresponding to each sampled PPG signal, and the sampled reference blood glucose value pre-configured for the sampled object from which each sampled PPG signal is sourced.
[0066] Step 204, based on each sampled PPG signal, extract the sample feature values of multiple preset features, and obtain the sample feature values of each sampled PPG signal corresponding to multiple preset features respectively.
[0067] Step 206, according to the sampled reference blood glucose values of the sampled objects from which multiple sampled PPG signals are respectively sourced, associate multiple sampled PPG signals with the preset object types to which the sampled objects from which they are respectively sourced belong among different preset object types.
[0068] Step 208: For each preset object type, based on the sample feature values corresponding to multiple preset features of each sample PPG signal associated with the preset object type, and the actual blood glucose values corresponding to each sample PPG signal associated with the preset object type, perform model training to obtain a blood glucose prediction model applicable to the preset object type. The blood glucose prediction model represents the relationship between blood glucose and at least two preset features among the multiple preset features that are the applicable target features.
[0069] Among them, the step of training the blood glucose prediction model in this embodiment can be executed by a computer device. The computer device can obtain multiple collected sample PPG signals from one or more wearable devices, and then execute the subsequent steps. The sample PPG signal is the PPG signal collected for training the blood glucose prediction model. The actual blood glucose value corresponding to the sample PPG signal is collected for the same person in the same time period as the sample PPG signal. For example, within a given time duration in a preset scenario among multiple preset scenarios, the sample PPG signal and the actual blood glucose value can be successively collected for the same person, and the sample PPG signal and the actual blood glucose value correspond to each other. The actual blood glucose value can be detected by a blood glucose meter. It can be understood that an accurate blood glucose value can be detected by the blood glucose meter.
[0070] The multiple preset scenarios can include a standard scenario, such as a fasting scenario; it can also include different scenarios that cause blood glucose changes, for example, including scenarios before night sleep, before three meals a day, after three meals a day, before eating desserts, after eating desserts, before exercise, after exercise, etc. By collecting the sample PPG signal and the actual blood glucose value under multiple preset scenarios, the sample PPG signal and the actual blood glucose value will change under different scenarios, which can form a comparison and improve the diversity of the samples.
[0071] The sample object is the object targeted for collecting the information required for training during the stage of training the blood glucose prediction model. The information required for training can include the sample PPG signal, the actual blood glucose value, and the sample reference blood glucose value. The sample object can be the name, electronic account, or other electronic identities bound to the person when the sample PPG signal is collected. Each of the multiple sample PPG signals can have the sample object from which it is sourced. The sample reference blood glucose value is the reference blood glucose value configured for the sample object during the stage of training the blood glucose prediction model. The sample feature values of each of the multiple preset features can be extracted from the sample PPG signal or from the signal after preprocessing the sample PPG signal. The method for determining the preset object type to which the sample object belongs can be the same as the method for determining the target object type of the target object in step 104.
[0072] The blood glucose prediction model can be a multiple linear regression model. During the model training process, the stepwise regression method can be used to train this multiple linear regression model. Specifically, multiple preset features can be used as independent variables, and blood glucose can be used as the dependent variable. The stepwise regression method can, through an iterative training process, gradually add or delete independent variables to automatically select the independent variables that have a significant impact on the dependent variable until the preset number of features or statistical indicators no longer improve significantly, thereby determining the relationship formula between the selected independent variables and the dependent variable and obtaining the trained blood glucose prediction model. Among them, the selected independent variables are the selected preset features, which will be used as the target features applicable to the blood glucose prediction model. The preset features in the relationship formula are used to substitute corresponding feature values, and blood glucose is used to substitute corresponding actual blood glucose values. The statistical indicator can be the error between the predicted blood glucose value and the actual blood glucose value during the training process.
[0073] Exemplarily, the multiple preset features can include time-domain features, frequency-domain features, and non-linear features; the time-domain features include at least one of peak features, trough features, peak-trough change features, fluctuation features, dicrotic wave position features, dicrotic wave amplitude features, and dicrotic wave area features. In this embodiment, the multiple preset features involve multiple feature types, improving the feature richness and creating conditions for accurately training the model.
[0074] Among them, the time-domain features are features extracted based on the PPG signal in the time domain. The peak feature can include the peak and the peak position. The peak is the peak value of each cardiac cycle in the PPG signal. The peak position is the time point at which the peak value is generated in the PPG signal. The trough feature can include the trough and the trough position. The trough is the trough value of each cardiac cycle in the PPG signal. The trough position is the time point at which the trough value is generated in the PPG signal. The peak-trough change feature can include the rise time and the fall time within each cardiac cycle of the PPG signal. The rise time is the time length from the trough to the peak. The fall time is the time length from the peak to the trough.
[0075] The fluctuation feature includes the time interval between the fluctuation points of adjacent cardiac cycles and the fluctuation amplitude between the fluctuation points of adjacent cardiac cycles in the PPG signal. Within each cardiac cycle, the fluctuation point can be a point with an amplitude equal to the product of the peak and a preset ratio, or a point with an amplitude equal to the product of the trough and a preset ratio. The preset ratio is greater than 0 and less than 1, and can be, for example, 15%, 10%, or others. The fluctuation amplitude is the difference between the amplitudes of the fluctuation points of adjacent cardiac cycles.
[0076] The dicrotic wave position feature is the time point of the peak value of the dicrotic wave within each cardiac cycle in the PPG signal. The dicrotic wave is a phenomenon where the heart contracts again within a short period after a single contraction. Each cardiac cycle in the PPG signal may include a main wave (corresponding to the cardiac systolic phase) and a descending branch (corresponding to the cardiac diastolic phase). A dicrotic wave may appear in the descending branch. The dicrotic wave is a secondary wave that rises and then falls again in the descending branch, and the peak value of the dicrotic wave is smaller than the peak of the main wave. The dicrotic wave amplitude feature may be the peak value of the dicrotic wave. The dicrotic wave area feature may be the area of the region formed by the envelope line of the dicrotic wave and the coordinate axes when the PPG signal is plotted in a rectangular coordinate system.
[0077] Frequency domain features are features extracted after converting the PPG signal to the frequency domain. The PPG signal in the time domain can be converted to the frequency domain through Fourier transform or power spectral density. Frequency domain features may include DC amplitude, total power, dominant frequency, high-frequency power, low-frequency power, or others. Among them, the DC amplitude can characterize the amplitude value of the DC component in the PPG signal. The total power is the total energy of the PPG signal in the entire frequency band. The dominant frequency is the frequency with the highest energy in the power spectrum of the PPG signal. The dominant frequency can reflect the heart rate. The high-frequency power is the integrated energy within the high-frequency band in the power spectrum of the PPG signal. The high-frequency band is, for example, 0.15 Hz (Hertz) to 0.4 Hz. The low-frequency power is the integrated energy within the low-frequency band in the power spectrum of the PPG signal. The low-frequency band is, for example, 0.04 Hz to 0.15 Hz.
[0078] Nonlinear features are used to characterize the nonlinear changes in blood pressure flow in the PPG signal. Nonlinear features include, for example, sample entropy, approximate entropy, fractal dimension, or others. Among them, sample entropy is used to measure the complexity of the PPG signal, and a lower value indicates a stronger regularity of the PPG signal waveform. Approximate entropy is used to measure the irregularity of the PPG signal. The fractal dimension can be used to describe the self-similarity of the PPG signal. Multiple preset features may also include features extracted from the PPG signal through an autoregressive model. Specifically, for example, AR coefficients (autoregressive coefficients), residual variance (used to evaluate the fitting degree of the autoregressive model).
[0079] Multiple preset features may also include the eigenvalue offset of the amplitude-related features among the above features. Amplitude-related features are features related to amplitude and may include wave peaks, wave troughs, the peak of the dicrotic wave, DC amplitude, etc. The eigenvalue offset is the difference value between the eigenvalues of the acquired PPG signal and the PPG signal collected during the calibration phase on the same feature. The calibration phase is the phase when information is collected when the wearable device is first used.
[0080] In this embodiment, by separately training models for different preset object types, a blood glucose prediction model for each preset object type can be obtained, creating conditions for accurately predicting blood glucose subsequently. Moreover, during the process of training the blood glucose prediction model for each preset object type, by using the sample reference blood glucose values to associate multiple sample PPG signals with the preset object types to which the sample objects from their respective sources belong, and based on the sample feature values corresponding to multiple preset features for each associated sample PPG signal respectively, as well as the corresponding actual blood glucose values for training, the relationship between multiple target features and blood glucose can be found according to the actual blood glucose values, such that the trained blood glucose prediction model has a relatively accurate blood glucose prediction ability.
[0081] In an exemplary embodiment, the step of obtaining the reference blood glucose value pre-configured for the target object to which the PPG signal belongs in step 102 may include: obtaining the configuration information for the target object to which the PPG signal belongs; the configuration information includes the historical blood glucose values configured historically and the corresponding configuration times; determining the target configuration time for the most recently configured historical blood glucose value from the configuration information; and when the time difference between the target configuration time and the acquisition time of the PPG signal does not exceed a preset time difference, determining the most recently configured historical blood glucose value in the configuration information as the reference blood glucose value pre-configured for the target object to which the PPG signal belongs.
[0082] Among them, the historical blood glucose value is the actual blood glucose value detected by a blood glucose meter before the acquisition time of the PPG signal, and can be the actual blood glucose value collected under a standard scenario, such as a fasting scenario. It can be understood that the historical blood glucose value can be used as a static blood glucose value to assist in determining the target object type to which the target object belongs. Different object types apply different blood glucose prediction models. Compared with the blood glucose value detected by a blood glucose meter, the PPG signal is easier to collect, enabling convenient and dynamic use of the PPG signal for relatively accurate blood glucose prediction.
[0083] For example, the configuration time interval between two historical blood glucose values at adjacent configuration times can be greater than the acquisition time interval between two PPG signals at adjacent acquisition times. For example, the configuration time interval can be greater than half a year, and the acquisition time interval can be greater than 1 hour and less than 24 hours. In this way, obtaining historical blood glucose values using a blood glucose meter is carried out at a low frequency. And using the historical blood glucose value as the reference blood glucose value can accurately determine the target object type to which the target object belongs, thus creating conditions for relatively accurate blood glucose prediction based on the PPG signal. The characteristic that the PPG signal is easy to collect enables high-frequency and continuous blood glucose prediction. In this way, on the basis of using a blood glucose meter at a low frequency, high-frequency and continuous relatively accurate blood glucose prediction can be carried out based on the PPG signal.
[0084] In the historical time before obtaining the PPG signal, there may be one or more historically configured blood glucose values. For example, when a user first uses a wearable device and enters the calibration phase, the user can enter the blood glucose value detected by a blood glucose meter. The server can record the blood glucose value entered by the user into the configuration information of the user as the historically configured blood glucose value and record the corresponding entry time as the configuration time. Another example is that after the calibration phase, during the user's use of the wearable device, a new blood glucose value can be re-entered, and the server can also record it into the configuration information.
[0085] The preset time difference can be 6 months, 1 year, or others. When the time difference between the target configuration time and the acquisition time of the PPG signal does not exceed the preset time difference, it can be characterized that the historically configured blood glucose value of the most recent configuration can still effectively represent the actual blood glucose level of the person represented by the target object. As time goes by, the actual blood glucose level of the person represented by the target object may change, and the historically configured blood glucose value needs to be updated to improve the accuracy of determining the target object type to which the target object belongs. A prompt message for prompting the target object to update the configured blood glucose value can be sent when the time difference between the target configuration time and the acquisition time of the PPG signal exceeds the preset time difference.
[0086] In this embodiment, when the time difference between the target configuration time and the acquisition time of the PPG signal does not exceed the preset time difference, it can be characterized that the historically configured blood glucose value of the most recent configuration can still effectively represent the actual blood glucose level of the person represented by the target object. By determining the historically configured blood glucose value of the most recent configuration as the reference blood glucose value, the target object type to which the target object belongs can be determined more accurately.
[0087] In an exemplary embodiment, step 104 may include: obtaining the correspondence between different preset object types and different preset blood glucose value ranges that are pre-configured; determining the preset blood glucose value range to which the reference blood glucose value belongs among different preset blood glucose value ranges; and determining the target object type to which the target object belongs according to the correspondence and the preset blood glucose value range to which the reference blood glucose value belongs.
[0088] Among them, the correspondence can describe the corresponding preset object types and preset blood glucose value ranges. For example, the first preset object type and the first preset blood glucose value range can correspond to each other, the second preset object type and the second preset blood glucose value range can correspond to each other, and the third preset object type and the third preset blood glucose value range can correspond to each other. The preset object type corresponding to the preset blood glucose value range to which the reference blood glucose value belongs can be used as the target object type to which the target object belongs.
[0089] In this embodiment, by pre-configuring the correspondence between different preset object types and different preset blood glucose value ranges, the classification methods of different preset object types can be quantified. Thus, based on the correspondence and the reference blood glucose value, the target object type to which the target object belongs can be accurately determined, creating conditions for accurately performing blood glucose prediction subsequently.
[0090] In an exemplary embodiment, step 108 may include: obtaining a target preprocessing method pre-configured for the target object type; preprocessing the PPG signal according to the target preprocessing method to obtain a preprocessed PPG signal; and extracting the feature values of multiple target features from the preprocessed PPG signal.
[0091] Among them, different preset object types may correspond to different preprocessing methods. For example, the minimum value in the first preset blood glucose value range corresponding to the first preset object type may be greater than the maximum value in the second preset blood glucose value range corresponding to the second preset object type, and the minimum value in the second preset blood glucose value range may be greater than the maximum value in the third preset blood glucose value range corresponding to the third preset object type. In the above scenario, the PPG signal corresponding to the first preset object type may have problems such as dicrotic wave attenuation and significant baseline low-frequency drift. Then the corresponding preprocessing method may include baseline correction through morphological filtering, estimating and repairing the dicrotic wave waveform based on a standard waveform template, and wavelet threshold denoising. The PPG signal corresponding to the second preset object type has relatively high quality, and a standard preprocessing process can be adopted. Then the corresponding preprocessing method may include using band-pass filtering to remove baseline drift and high-frequency noise, and wavelet threshold denoising. The PPG signal corresponding to the third preset object type may have problems such as low signal amplitude and motion artifacts. Then the corresponding preprocessing method may include removing motion artifacts, performing dynamic normalization processing based on a sliding window to reduce the impact of low signal amplitude, and processing high-frequency noise through zero-phase filtering.
[0092] In this embodiment, by using the target preprocessing method pre-configured for the target object type to preprocess the PPG signal, and the target preprocessing method is specifically adapted to the target object type, the quality of the preprocessed PPG signal can be improved, so that the feature values of multiple target features can be accurately extracted from the preprocessed PPG signal.
[0093] In a specific embodiment, the above blood glucose prediction method may include the following steps.
[0094] The computer device can obtain multiple sampled PPG signals that are collected, the actual blood glucose values corresponding to each sampled PPG signal, and the sampled reference blood glucose values pre-configured for the sampled objects from which each sampled PPG signal is sourced; based on each sampled PPG signal, extract the sampled feature values of each of multiple preset features, and obtain the sampled feature values of each sampled PPG signal corresponding to multiple preset features respectively; and can associate each of the multiple sampled PPG signals with the preset object type to which the sampled object from which each sampled PPG signal is sourced belongs in different preset object types according to the sampled reference blood glucose values of the sampled objects from which the multiple sampled PPG signals are respectively sourced.
[0095] For each preset object type, the computer device can perform model training according to the sampled feature values of each of the sampled PPG signals associated with the preset object type corresponding to multiple preset features respectively, and the actual blood glucose values corresponding to each of the sampled PPG signals associated with the preset object type, to obtain a blood glucose prediction model applicable to the preset object type, where the blood glucose prediction model represents the relationship between blood glucose and at least two preset features among the multiple preset features that are the applicable target features.
[0096] The wearable device can obtain the collected PPG signal; obtain the configuration information for the target object to which the PPG signal belongs; determine, from the configuration information, the target configuration time when the historical blood glucose value was last configured; and when the time difference between the target configuration time and the collection time of the PPG signal does not exceed a preset time difference, determine the historical blood glucose value that was last configured in the configuration information as the reference blood glucose value pre-configured for the target object to which the PPG signal belongs.
[0097] The wearable device can obtain the corresponding relationship between different preset object types and different preset blood glucose value ranges pre-configured; determine the preset blood glucose value range to which the reference blood glucose value belongs among different preset blood glucose value ranges; and determine the target object type to which the target object belongs according to the corresponding relationship and the preset blood glucose value range to which the reference blood glucose value belongs.
[0098] The wearable device can determine the trained target blood glucose prediction model applicable to the target object type, and determine the multiple target features applicable to the target blood glucose prediction model; obtain the target preprocessing method pre-configured for the target object type; preprocess the PPG signal according to the target preprocessing method to obtain the preprocessed PPG signal; extract the feature values of each of the multiple target features from the preprocessed PPG signal; and determine the blood glucose prediction value of the target object according to the feature values of each of the multiple target features through the target blood glucose prediction model.
[0099] It should be understood that although the steps in the flowcharts involved in the above-described embodiments are shown in sequence according to the indications of the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless there is a clear indication in this article, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-described embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily executed at the same moment, but can be executed at different moments, and the execution order of these steps or stages is not necessarily sequential, but can be executed alternately or in turn with at least a part of other steps or steps or stages in other steps.
[0100] Based on the same inventive concept, an embodiment of the present application further provides a blood glucose prediction device for implementing the above-mentioned blood glucose prediction method. The implementation solutions provided by this device to solve problems are similar to the implementation solutions described in the above method. Therefore, the specific limitations in one or more embodiments of the following blood glucose prediction device can refer to the limitations on the blood glucose prediction method in the above text, and will not be repeated here.
[0101] In an exemplary embodiment, as Figure 3 shown, a blood glucose prediction device 300 is provided, including: an acquisition module 310, an object type determination module 320, and a prediction module 330, where:
[0102] The acquisition module 310 is configured to acquire the collected PPG signal and acquire the pre-configured reference blood glucose value for the target object to which the PPG signal belongs.
[0103] The object type determination module 320 is configured to determine the target object type to which the target object belongs among different pre-configured preset object types according to the reference blood glucose value.
[0104] The prediction module 330 is configured to determine the trained target blood glucose prediction model applicable to the target object type and determine multiple target features applicable to the target blood glucose prediction model; extract the feature values of each of the multiple target features based on the PPG signal; and determine the blood glucose prediction value of the target object through the target blood glucose prediction model according to the feature values of each of the multiple target features.
[0105] In an exemplary embodiment, the trained target blood glucose prediction model is one of the trained blood glucose prediction models applicable to different preset object types. The blood glucose prediction device 300 further includes a training module. The training module is configured to obtain a plurality of collected sample PPG signals, the actual blood glucose value corresponding to each sample PPG signal, and the sample reference blood glucose value pre-configured for the sample object from which each sample PPG signal is sourced; based on each sample PPG signal, extract the sample feature values of each of the plurality of preset features, and obtain the sample feature values of each sample PPG signal corresponding to the plurality of preset features respectively; according to the sample reference blood glucose values of the sample objects from which the plurality of sample PPG signals are sourced respectively, associate the plurality of sample PPG signals with the preset object type to which the sample objects from which they are sourced belong among different preset object types; for each preset object type, perform model training according to the sample feature values of each of the sample PPG signals associated with the preset object type corresponding to the plurality of preset features, and the actual blood glucose value corresponding to each of the sample PPG signals associated with the preset object type, to obtain a blood glucose prediction model applicable to the preset object type, where the blood glucose prediction model represents the relationship between blood glucose and at least two preset features among the plurality of preset features that are used as the applicable target features.
[0106] In an exemplary embodiment, the plurality of preset features include time-domain features, frequency-domain features, and non-linear features; the time-domain features include at least one of peak feature, valley feature, peak-valley change feature, fluctuation feature, dicrotic wave position feature, dicrotic wave amplitude feature, and dicrotic wave area feature.
[0107] In an exemplary embodiment, the acquisition module 310 is further configured to obtain the configuration information for the target object to which the PPG signal belongs; the configuration information includes the historical blood glucose value configured historically and the corresponding configuration time; from the configuration information, determine the target configuration time when the historical blood glucose value was configured most recently; in the case where the time difference between the target configuration time and the acquisition time of the PPG signal does not exceed a preset time difference, determine the historical blood glucose value configured most recently in the configuration information as the reference blood glucose value pre-configured for the target object to which the PPG signal belongs.
[0108] In an exemplary embodiment, the object type determination module 320 is further configured to obtain the correspondence between different preset object types and different preset blood glucose value ranges pre-configured; determine the preset blood glucose value range to which the reference blood glucose value belongs among different preset blood glucose value ranges; according to the correspondence and the preset blood glucose value range to which the reference blood glucose value belongs, determine the target object type to which the target object belongs.
[0109] In an exemplary embodiment, the prediction module 330 is further configured to obtain a target preprocessing method preconfigured for the target object type; preprocess the PPG signal according to the target preprocessing method to obtain a preprocessed PPG signal; and extract the feature values of multiple target features from the preprocessed PPG signal.
[0110] Each module in the above blood glucose prediction device can be implemented in whole or in part by software, hardware, and their combination. The above modules can be embedded in or independent of the processor in the wearable device in the form of hardware, or stored in the memory of the wearable device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.
[0111] In an exemplary embodiment, a wearable device is provided, and its internal structure diagram can be as Figure 4 shown. The wearable device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O), a communication interface, a display unit, and an input device. Among them, the processor, the memory, and the input / output interface are connected through a system bus, and the communication interface, the display unit, and the input device are connected to the system bus through the input / output interface. Among them, the processor of the wearable device is used to provide computing and control capabilities. The memory of the wearable device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the wearable device is used to exchange information between the processor and external devices. The communication interface of the wearable device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a mobile cellular network, near field communication (Near Field Communication, NFC), or other technologies. The computer program, when executed by the processor, implements a blood glucose prediction method. The display unit of the wearable device is used to form a visually visible picture, which can be a display screen. The display screen can be a liquid crystal display screen. The input device of the wearable device can be a touch layer covering the display screen, or a button, a trackball, etc. provided on the shell of the wearable device.
[0112] Those skilled in the art can understand that Figure 4 the structure shown in
[0113] In one embodiment, a wearable device is further provided, which includes a memory and a processor. A computer program is stored in the memory, and when the processor executes the computer program, the steps in the foregoing method embodiments are implemented.
[0114] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the foregoing method embodiments are implemented.
[0115] In one embodiment, a computer program product is provided, which includes a computer program. When the computer program is executed by a processor, the steps in the foregoing method embodiments are implemented.
[0116] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use, and processing of relevant data need to comply with relevant regulations.
[0117] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, database, or other medium used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc. The databases involved in the embodiments provided in this application can include at least one of relational databases and non-relational databases. Non-relational databases can include distributed databases based on blockchain, etc., without limitation. The processors involved in the embodiments provided in this application can be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, data processing logics based on quantum computing, artificial intelligence (AI) processors, etc., without limitation.
[0118] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope recorded in this application.
[0119] The above-described embodiments merely represent several implementation manners of the present application. The description thereof is relatively specific and detailed, but it should not be construed as a limitation to the patent scope of the present application. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all fall within the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the appended claims.
Claims
1. A blood glucose prediction method, characterized in that, The method includes: Obtaining the collected PPG signal and obtaining a reference blood glucose value pre-configured for a target object to which the PPG signal belongs; Determining, according to the reference blood glucose value, a target object type to which the target object belongs among different pre-configured preset object types; Determining a trained target blood glucose prediction model applicable to the target object type and determining a plurality of target features applicable to the target blood glucose prediction model; Extracting the feature value of each of the plurality of target features based on the PPG signal; Determining a blood glucose prediction value of the target object through the target blood glucose prediction model according to the feature value of each of the plurality of target features.
2. The method according to claim 1, characterized in that, The trained target blood glucose prediction model is one of the trained blood glucose prediction models applicable to different preset object types respectively. The steps of training the blood glucose prediction model include: Obtaining a plurality of collected sample PPG signals, the actual blood glucose value corresponding to each sample PPG signal, and a sample reference blood glucose value pre-configured for a sample object from which each sample PPG signal is sourced; Extracting the sample feature value of each of a plurality of preset features based on each sample PPG signal, and obtaining the sample feature value of each sample PPG signal corresponding to the plurality of preset features respectively; According to the sample reference blood glucose value of the sample object from which each of the plurality of sample PPG signals is sourced, associating each of the plurality of sample PPG signals with the preset object type to which the sample object from which it is sourced belongs among different preset object types; For each preset object type, performing model training according to the sample feature value of each of the plurality of preset features corresponding to each sample PPG signal associated with the preset object type and the actual blood glucose value corresponding to each sample PPG signal associated with the preset object type, to obtain a blood glucose prediction model applicable to the preset object type, where the blood glucose prediction model represents the relationship between blood glucose and at least two preset features among the plurality of preset features that are used as applicable target features.
3. The method according to claim 2, wherein The plurality of preset features include time domain features, frequency domain features, and non-linear features; the time domain features include at least one of peak feature, trough feature, peak-trough change feature, fluctuation feature, dicrotic wave position feature, dicrotic wave amplitude feature, and dicrotic wave area feature.
4. The method according to claim 1, wherein The obtaining a reference blood glucose value pre-configured for a target object to which the PPG signal belongs includes: Obtaining configuration information of the target object to which the PPG signal belongs; the configuration information includes a historical blood glucose value configured historically and a corresponding configuration time; Determining a target configuration time for the most recently configured historical blood glucose value from the configuration information; When the time difference between the target configuration time and the acquisition time of the PPG signal does not exceed a preset time difference, determining the most recently configured historical blood glucose value in the configuration information as the reference blood glucose value pre-configured for the target object to which the PPG signal belongs.
5. The method according to claim 1, wherein The determining, according to the reference blood glucose value, a target object type to which the target object belongs among different pre-configured preset object types includes: Obtain the corresponding relationship between different pre-configured preset object types and different preset blood glucose value ranges; Determine the preset blood glucose value range to which the reference blood glucose value belongs among the different preset blood glucose value ranges; According to the corresponding relationship and the preset blood glucose value range to which the reference blood glucose value belongs, determine the target object type to which the target object belongs.
6. The method according to any one of claims 1-5, characterized in that, The extracting the eigenvalue of each of the plurality of target features based on the PPG signal includes: Obtain the target preprocessing method pre-configured for the target object type; Preprocess the PPG signal according to the target preprocessing method to obtain a preprocessed PPG signal; Extract the eigenvalue of each of the plurality of target features from the preprocessed PPG signal.
7. A blood glucose prediction device, characterized in that, The device includes: An acquisition module, configured to acquire the collected PPG signal and acquire the reference blood glucose value pre-configured for the target object to which the PPG signal belongs; An object type determination module, configured to determine the target object type to which the target object belongs among different pre-configured preset object types according to the reference blood glucose value; A prediction module, configured to determine the trained target blood glucose prediction model applicable to the target object type and determine the plurality of target features applicable to the target blood glucose prediction model; extract the eigenvalue of each of the plurality of target features based on the PPG signal; and determine the blood glucose prediction value of the target object according to the eigenvalue of each of the plurality of target features through the target blood glucose prediction model.
8. A wearable device, comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.
9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, the steps of the method according to any one of claims 1 to 6 are implemented.
10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, the steps of the method according to any one of claims 1 to 6 are implemented.