Multi-wavelength noninvasive blood glucose system calibration method, storage medium and blood glucose meter

Abstract

The invention discloses a multi-wavelength noninvasive blood glucose system calibration method which comprises the following steps: step 1, continuously detecting and recording pulse wave signals by a user, and taking obtained pulse wave signal measurement values as a data set A; 2, during the period of continuously detecting the pulse wave signals, part of time nodes are selected for one-time invasive blood glucose measurement, blood glucose measurement values are obtained, all the blood glucose measurement values are calculated based on an interpolation prediction method, and the user selects the same-period blood glucose prediction values within the duration time range of the time nodes and serves as a data set B; and step 3, based on the data set A and the data set B, establishing an individualized noninvasive blood glucose detection model of the user. According to the invention, the pricking frequency is reduced by about 10 times, so that the physiological discomfort and psychological burden of a testee are greatly reduced; or compared with a method for acquiring a data set through a minimally invasive sensor, the method does not need additional application type sensor consumables, and the detection cost is greatly reduced. The invention provides a more convenient and economical solution for calibrating the noninvasive blood glucose monitoring device for a testee, and is a beneficial innovation in the technical field of noninvasive blood glucose detection.

Description

Technical Field

[0001] The present invention relates to a non-invasive blood sugar measurement technology, and in particular to a non-invasive blood sugar calibration method. Background Art

[0002] A non-invasive blood glucose meter is a device that uses dual-wavelength near-infrared light to irradiate the measurement site to obtain the patient's blood glucose value. The biggest advantage of this device is that it can obtain blood glucose non-invasively and reduce the patient's pain; for example, the publication number is CN116942153A, the publication date is October 27, 2023, and the patent name is "A dual-wavelength near-infrared measurement method for non-destructive measurement of blood glucose". The disclosed dual-wavelength near-infrared measurement method for non-destructive measurement of blood glucose includes a reference step: obtaining blood glucose related parameters corresponding to the reference time t0; a measurement step: obtaining the relevant blood glucose parameters corresponding to the measurement time t1; based on the blood glucose related parameters corresponding to the reference time t0 and the relevant blood glucose parameters corresponding to the measurement time t1, the blood glucose value G1 at time t1 is calculated through a statistical method model.

[0003] However, such devices also suffer from the problem of unreliable measurement accuracy. Therefore, before using the device, each patient must undergo pre-calibration if they need to improve the device's measurement accuracy. This generally requires multiple invasive blood glucose measurements. Non-invasive blood glucose testing requires multiple invasive blood glucose measurements because everyone's body is very different, and the pattern and response of blood glucose fluctuations vary from person to person. To ensure the accuracy of the results calculated by this non-invasive blood glucose device, it must be calibrated once the subject receives the device to establish a personalized model. However, current sampling and modeling methods require users to use it multiple times, which invisibly increases the user's pain. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to realize a method suitable for non-invasive blood glucose detection, especially the initialization calibration and measurement of fingertip multi-wavelength non-invasive blood glucose detection devices. Through calibration, the measurement accuracy of the non-invasive blood glucose measurement equipment can be improved and the number of sampling times can be reduced.

[0005] In order to achieve the above object, the technical solution adopted by the present invention is: a multi-wavelength non-invasive blood glucose system calibration method, comprising the following steps:

[0006] Step 1: The user continuously detects and records the pulse wave signal, and the obtained pulse wave signal measurement value is used as data set A;

[0007] Step 2: During the continuous detection of the pulse wave signal, select some time nodes to perform an invasive blood glucose measurement and obtain the blood glucose measurement value. All blood glucose measurement values are calculated based on the interpolation prediction method. The blood glucose prediction values for the same period within the duration of the user's selected time node are used as data set B.

[0008] Step 3: Based on dataset A and dataset B, establish an individualized non-invasive blood glucose detection model for this user.

[0009] In step 2, 3 - 5 time nodes are selected;

[0010] If 3 time nodes are selected: 10 minutes, 40 minutes, and 70 minutes after the end of eating;

[0011] If 4 time nodes are selected: 10 minutes, 30 minutes, 50 minutes, and 70 minutes after the end of eating;

[0012] If 5 time nodes are selected: 10 minutes, 25 minutes, 40 minutes, 55 minutes, and 70 minutes after the end of eating;

[0013] Among them:

[0014] The deviation between the actual time of each time node and the set value is within 3 minutes;

[0015] The deviation between the actual time of the time interval between each time node and the set value is within 3 minutes;

[0016] The time interval between the first time node and the last time node for each selection method is within 60 - 65 minutes.

[0017] In step 2, the eating time is carried out between 7:00 and 9:30 in the morning. Before eating, the user is in a fasting state. The food ingested by the user is liquid and has a high glycemic index, and the eating time does not exceed 5 minutes.

[0018] At the first time node T1, the blood glucose measurement value obtained by invasive blood glucose measurement is G1, the pulse wave signal measurement value obtained at the same time is PPG1, and the corresponding test maintenance time information is t1. By analogy, T1, G2, PPG2 and T2, t2, T3, G3, PPG3 and t3... are obtained.

[0019]

PPG1, PPG2, PPG3...

t1, t2, t3...

[0020]

G1, G2, G3...

T1, T2, T3...

[0021] In step 2, with time as the abscissa and blood glucose measurement value as the ordinate, taking each time node T1, T2, T3... and its corresponding invasive blood glucose measurement values G1, G2, G3... as reference data, using the interpolation prediction algorithm to obtain coordinate points (T1, G1), (T2, G2), (T3, G3)... The curve formed by the coordinate points is the synchronous blood glucose prediction value curve;

[0022] In step 3, for each measured value of the pulse wave signal recorded at each time point in dataset A, the predicted blood glucose value corresponding to this time point can be found in the curve of the predicted blood glucose value at the same period. Then, the curve of the predicted blood glucose value at the same period is the mathematical relationship model between the pulse wave signal and the blood glucose value of this user.

[0023] In step 1, the processing steps of the pulse wave signal include:

[0024] 1) Preprocess the pulse wave signal by filtering, including filtering and removing baseline drift;

[0025] 2) Through the signal validity detection method of the dynamic threshold curve, use the intersection points of the detected dynamic threshold curve and the pulse wave signal to identify and reject the signal segments with excessive interference, and perform data screening on the pulse wave signal;

[0026] 3) Align and stack the pulse wave signal cycles, and perform vertical median filtering on the stacked signal to obtain the standard average pulse wave signal;

[0027] 4) Perform energy operator analysis on the pulse wave signal, extract the characteristics of variance, standard deviation, skewness, DC and AC components, and obtain the absorbance parameters of different blood glucose levels.

[0028] It further includes step 4, optimizing and validating the individualized non-invasive blood glucose detection model by means of cross-validation and / or error analysis, and verifying that when reaching the final input of the newly collected pulse wave feature quantity, the root mean square error and the mean absolute error between the blood glucose estimated value output by the model and the true value reach the preset standard.

[0029] A storage medium, which is a computer-readable storage medium for storing software program codes, and the software program codes are used to execute the calibration method of the multi-wavelength non-invasive blood glucose system.

[0030] A blood glucose meter, which is a multi-wavelength non-invasive blood glucose meter, includes a blood glucose meter body, and the blood glucose meter adopts the calibration method of the multi-wavelength non-invasive blood glucose system before each user uses it.

[0031] The present invention reduces the number of blood samplings by about 10 times, greatly reducing the physical discomfort and psychological burden of the subjects; or compared with the method of obtaining datasets by minimally invasive sensors, this method does not require additional patch-type sensor consumables, greatly reducing the detection cost. The present invention provides a more convenient and economical solution for calibrating non-invasive blood glucose monitoring devices for subjects, and is a beneficial innovation in the field of non-invasive blood glucose detection technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Briefly describe the content expressed in each drawing in the specification of the present invention as follows:

[0033] Figure 1 It is a flowchart of the calibration method for a multi-wavelength non-invasive blood glucose system. Specific implementation manners

[0034] The following is a further detailed description of the specific implementation manners of the present invention in terms of the shapes, structures of the various components involved, the mutual positions and connection relationships between the various parts, the functions and working principles of the various parts, the manufacturing processes, and the operation and usage methods, etc., with reference to the accompanying drawings, so as to help those skilled in the art have a more complete, accurate, and in-depth understanding of the inventive concept and technical solution of the present invention.

[0035] The reason why three to five invasive blood glucose measurements are required for the non-invasive blood glucose detection technology of the present invention is that the physical differences among individuals are very large, and the patterns and responses of blood glucose fluctuations vary from person to person. To ensure the accuracy of the calculation results of this non-invasive blood glucose device, a calibration must be performed after the subject receives the device to establish an individualized model. And this process of establishing the model requires the blood glucose values and pulse wave signals of the subject as the data set, so as to be able to construct the mathematical mapping relationship between the two through the linear regression method used in the present invention, that is, the least squares fitting, and establish the individualized non-invasive blood glucose detection model of the subject. Only in this way can the blood glucose levels of different subjects with individual differences be accurately predicted.

[0036] In the calibration method of the multi-wavelength non-invasive blood glucose system, there is a preferred scheme for the time of the calibration process. It is preferably arranged between 7:00 and 9:30 in the morning, and the subject must be in a fasting state before the calibration starts. After confirming that the above information is correct, immediately enter the calibration process. The subject must eat, and the food eaten should be liquid and have a high glycemic index, such as white rice porridge, oatmeal and other foods. In addition, the eating time should be controlled within five minutes. After the subject finishes eating, record the time point T0 when eating ends.

[0037] The above-mentioned pre-calibration preparation work such as the time period to be selected for the calibration process, the subject should remain in a fasting state, the subject must eat, and the content of the food eaten by the subject, etc., are all essential steps of this calibration method and an important part of this method. This part is related to the biological law of human blood glucose changes and is a necessary condition for the blood glucose prediction and the calibration method of the non-invasive blood glucose system in this method to achieve the best effect;

[0038] The calibration method of the multi-wavelength non-invasive blood glucose system includes the following steps:

[0039] First: During the calibration process of the subject using the non-invasive blood glucose detection device, continuously detect and record the pulse wave signal using this device, that is, obtain the measured value of the pulse wave signal of the subject as the data set A;

[0040] For the signal processing of the data set A, it includes the following steps:

[0041] 1. Filtering, baseline drift removal, and validity detection of the photoplethysmogram (PPG) signal;

[0042] Preprocess the PPG signal filtering, including filtering and baseline drift removal. After the preprocessing steps, a signal validity detection method based on a dynamic threshold curve is added. By detecting the intersection points of the dynamic threshold curve and the PPG signal, accurately identify and reject signal segments with excessive interference, and achieve automatic data screening;

[0043] 2. Align and stack the PPG signal periods, and perform vertical median filtering on the stacked signal to obtain a standard average PPG signal;

[0044] To obtain a PPG signal that can reflect the blood glucose level at the measurement moment, a method of aligning and stacking the PPG signal periods and performing vertical median filtering on the stacked signal is proposed.

[0045] 3. Perform energy operator analysis on the PPG signal, and extract features such as variance, standard deviation, skewness, DC / AC components, etc.;

[0046] Blood with a higher glucose concentration will have a different absorbance from that of blood with a glucose level, which will affect the shape of the PPG signal. Based on this principle, perform energy operator analysis on the PPG signal and extract relevant features such as variance, skewness, DC / AC components, etc.;

[0047] Secondly: During the continuous detection process of the PPG signal, select three to five specific time nodes, and perform an invasive blood glucose measurement on the subject at each time node; Based on the invasive blood glucose measurement values at all time nodes, calculate the synchronous blood glucose prediction value of the subject based on the interpolation prediction method as dataset B.

[0048] Dataset A: Dataset of PPG signal measurement values;

[0049] Dataset B: Dataset of blood glucose prediction values;

[0050] The selection of specific time nodes must follow the following rules:

[0051] When selecting 3 time points, the time interval between two adjacent time points should be 30 minutes. They are: 10 minutes, 40 minutes, and 70 minutes after the end of eating, which are T1, T2, and T3 respectively;

[0052] When selecting 4 time points, the time interval between two adjacent time points should be 20 minutes. They are: 10 minutes, 30 minutes, 50 minutes, and 70 minutes after the end of eating, which are T1, T2, T3, and T4 respectively;

[0053] When selecting 5 time points, the time interval between two adjacent time points should be 15 minutes. They are: 10 minutes, 25 minutes, 40 minutes, 55 minutes, and 70 minutes after the end of eating, which are T1, T2, T3, T4, and T5 respectively;

[0054] However, the following three points need to be noted:

[0055] (1) The deviation between the actual time and the theoretical value at each time point should be controlled within 3 minutes;

[0056] (2) The deviation between the actual time interval and the theoretical time interval between adjacent time points should be controlled within 3 minutes;

[0057] (3) The time interval between the first and the last time points should be controlled within the range of 60 - 65 minutes.

[0058] The preferred interpolation methods include, but are not limited to, quadratic spline, cubic spline and other interpolation prediction methods. According to the invasive blood glucose measurement values at three to five specific time points, calculate the blood glucose prediction value data within the duration range from the first time node to the last time node as dataset A.

[0059] During the calibration process of this non-invasive blood glucose measurement device, it is necessary to cooperate with an additional invasive blood glucose measurement tool that meets the national standard in terms of accuracy, such as a complete set of invasive blood glucose measurement tools including a blood collection pen, a tester, and test strips.

[0060] When reaching the first specific time node T1, use the invasive blood glucose measurement tool to perform the first invasive blood glucose measurement and record the blood glucose value G1 at that time. At the same time, start the multi-wavelength near-infrared non-invasive blood glucose measurement device to continuously detect the fingertip pulse wave signal (PPG signal). The duration of each group of pulse wave signals (PPG signals) is 40 seconds - 60 seconds, and record several groups of pulse wave signals (PPG signals) and their corresponding time point information. They are respectively recorded as PPG1, PPG2,... PPG n And t1, t2,... tn.

[0061] During the entire data acquisition process from starting to detect the pulse wave signal by the device to ending the detection of the pulse wave signal by closing the device, the subject should be in a sitting position and keep breathing smoothly, so that there is no large movement in various parts of the body, especially the hands. The subject should continuously place the finger in the device to ensure the signal consistency during the entire detection process to the greatest extent.

[0062] During the above process, that is, during the process where the subject continuously detects the fingertip pulse wave signal, when the time reaches each selected specific time node, an invasive blood glucose measurement should be performed at each time node respectively and the blood glucose value at that time should be recorded.

[0063] When selecting 3 time points, the blood glucose values G2 and G3 corresponding to the moments T2 and T3 should be recorded;

[0064] When selecting 4 time points, the blood glucose values G2, G3, and G4 corresponding to the moments T2, T3, and T4 should be recorded;

[0065] When selecting 5 time points, the blood glucose values G2, G3, G4, and G5 corresponding to the moments T2, T3, T4, and T5 should be recorded;

[0066] It should be noted that the finger or arm used for invasive blood glucose measurement should be different from the hand used for detecting the pulse wave signal (for example, if the subject uses the left hand finger to detect the pulse wave signal, then invasive blood glucose measurement should be performed on the right hand or right arm;

[0067] When the last invasive blood glucose measurement is completed at the last time node, after recording the blood glucose value at the last time node, turn off the device, complete the acquisition of the pulse wave signal, and organize all the collected pulse wave signals and their corresponding time point information, that is, the aforementioned [several groups of PPG signals: PPG1, PPG2,... PPG n and [the time point information t1, t2,... tn corresponding to each group of PPG signals], and denote it as dataset A.

[0068] When selecting 3 time points, the last time node is T3, and the corresponding blood glucose value is G3;

[0069] When selecting 4 time points, the last time node is T4, and the corresponding blood glucose value is G4;

[0070] When selecting 5 time points, the last time node is T5, and the corresponding blood glucose value is G5;

[0071] Finally, based on dataset A and dataset B, establish an individualized non-invasive blood glucose detection model for the subject to complete the calibration of the device described in the present invention.

[0072] Based on the subject's dataset A and dataset B, adopt regression algorithms including but not limited to least squares fitting, etc., to establish an individualized non-invasive blood glucose detection model for the subject, thereby completing the calibration of the non-invasive blood glucose measurement device described in the present invention.

[0073] In the above steps, we obtained: several groups of PPG signals between the first time point T1 and the last time point T3 / T4 / T5 (varying according to the number of specific time nodes selected) (with a total duration of 60 - 65 minutes) of the subject and the time point information corresponding to each group of PPG signals. We regard these data as dataset A;

[0074] Among them, regardless of whether the number of selected time nodes is 3, 4, or 5, the time interval between the first time node and the last time node is the same, that is, the total duration is 60 - 65 minutes;

[0075] Therefore, based on the total duration of 60 - 65 minutes and the duration of each group of pulse wave signals (PPG signals) being 40 seconds - 60 seconds, we expect to obtain 60 - 100 groups of PPG signals and their corresponding time point information;

[0076] The invasive blood glucose measurement values G1 / G2 / G3 / G4 / G5 (varying according to the number of specific time nodes selected) corresponding to each time node T1 / T2 / T3 / T4 / T5 (varying according to the number of specific time nodes selected) of the subject;

[0077] The above content includes all the operations that the subject needs to perform. The steps in the following content will be independently completed by the device.

[0078] Calculate the predicted blood glucose value within 60 - 65 minutes;

[0079] After the measurement, based on the invasive blood glucose measurement values G1, G2, G3, G4, G5 of the selected three to five specific time nodes and their corresponding time node information T1, T2, T3, T4, T5, the blood glucose prediction value of the subject between 60 and 65 minutes from the first time node T1 to the last time node (T3 or T4 or T5, varying according to the number of specific time nodes selected) is calculated through the interpolation prediction algorithm, which is used as dataset B. The experimental results show that within the prediction time range of 60 minutes to 65 minutes, compared with invasive blood glucose measurement, the prediction error rate of this method can be controlled within 5%.

[0080] The following are the specific operation steps of the interpolation prediction algorithm: (Taking the selection of 4 points as an example, the following calculations in this paragraph are based on the number of selected time nodes being 4)

[0081] Taking time (unit: minute) as the abscissa and blood glucose measurement value (unit: mmol / L) as the ordinate, and using the 4 specific time nodes T1, T2, T3, T4 and their corresponding invasive blood glucose measurement values G1, G2, G3, G4 as reference data, the interpolation prediction algorithm (taking the cubic spline interpolation method as an example) will calculate a continuous curve passing through the selected time nodes, that is, the points with coordinates (T1, G1), (T2, G2), (T3, G3), (T4, G4). The starting point of this curve is the first time node (T1, G1), and the ending point is (T4, G4). Between the starting point and the ending point (60 minutes to 65 minutes), for each time point on the abscissa, the predicted blood glucose value corresponding to this point can be found on the prediction curve.

[0082] Note: The specific curve equation varies from person to person and is jointly determined by the selected specific time nodes and their corresponding blood glucose values.

[0083] Through this method, we have achieved: By only selecting 3 to 5 specific time nodes to collect the invasive blood glucose data of the subjects, we can still obtain a large number of blood glucose prediction value data with the error controllable within 5% through interpolation prediction, which can be used for the establishment of the subsequent non-invasive blood glucose individualized model.

[0084] After the above interpolation prediction steps are completed, we obtain all the data for establishing the non-invasive blood glucose individualized model. One is the pulse wave (PPG) signal between the first time node T1 and the last time node (T3 or T4 or T5, depending on the number of selected specific time nodes) described in step 3 during the whole process of calibration, and its corresponding time point information, which is dataset A; the other is the blood glucose prediction values within 60 to 65 minutes between the first time node T1 and the last time node (T3 or T4 or T5, depending on the number of selected specific time nodes) described in (IV. Obtaining the predicted blood glucose values within 60 - 65 minutes), which is dataset B.

[0085] It can be seen that the pulse wave (PPG) signal of the subject is synchronous with the predicted blood glucose value, both within 60 to 65 minutes between the first time node T1 and the last time node. Therefore, in the subsequent modeling steps, for each PPG signal recorded at each time point in dataset A, we can find the corresponding predicted blood glucose value at this time point in dataset B, so as to establish a mathematical relationship model between the pulse wave (PPG) signal and the blood glucose value of the subject.

[0086] Precisely because of this, the blood glucose value prediction method described in this invention solves a major problem in the previous non-invasive blood glucose detection devices, that is, if the previous non-invasive blood glucose detection devices need to perform individualized modeling, they must obtain no less than 40 groups of invasive blood glucose measurement values of the subjects.

[0087] This method can reduce the number of finger pricks by about 10 times, greatly reducing the physiological discomfort and psychological burden of the subjects; or compared with the method of obtaining the dataset by minimally invasive sensors, this method does not require additional patch sensor consumables, greatly reducing the detection cost. This invention provides a more convenient and economical solution for calibrating non-invasive blood glucose monitoring devices for the subjects, and is a beneficial innovation in the field of non-invasive blood glucose detection technology.

[0088] The features of the extracted photoplethysmogram (PPG) signal are used as the input vector. This step utilizes the features extracted from the signal processing of dataset A, and the predicted blood glucose value is used as the output. This step utilizes the blood glucose prediction value data of the subjects in dataset B. A relevant mathematical model (such as partial least squares regression, support vector machine, etc.) is established. In addition, the model is optimized and verified through means such as cross-validation and error analysis to ensure that when finally inputting the newly collected PPG feature quantities, evaluation indicators such as the root mean square error (RMSE) and mean absolute error (MAE) between the blood glucose estimated value output by the model and the true value reach the clinically acceptable standard, thereby achieving accurate and reliable non-invasive blood glucose monitoring.

[0089] The establishment of dataset B for individual modeling of subjects is invented. Through the method in this patent, the number of invasive blood glucose measurements can be significantly reduced, and a reconstructed blood glucose value dataset is constructed for modeling. Therefore, some of the methods and means described above, such as the selection of specific time points, the regulation of the subject's state before the calibration stage, the selection of the time period during which the calibration process should be carried out, and the food content eaten by the subject, are all for obtaining more accurate blood glucose prediction values, that is, dataset B. Then, by using dataset A and dataset B for modeling, a more accurate individual non-invasive blood glucose detection model is obtained, rather than directly predicting the blood glucose value of a subject at a certain day, a certain time period, or any specific moment. Therefore, this method is not a method directly related to diagnosis and treatment, but a more superior method for obtaining the dataset required for modeling in the field of non-invasive blood glucose detection, so as to enable the subject to complete the calibration process of the non-invasive blood glucose detection device more economically and conveniently.

[0090] During the calibration process, the subject's PPG signal dataset (dataset A) is used as the input feature, and the corresponding blood glucose value dataset (in this invention, it is dataset B: blood glucose prediction value dataset, and in the previous invention, it is the blood glucose measurement value dataset) is used as the output feature. A mathematical mapping relationship between the two is constructed using a regression algorithm (such as least squares fitting) to establish an individual non-invasive blood glucose detection model for the subject. In addition, the model is optimized and verified through means such as cross-validation and error analysis to ensure that when finally inputting the newly collected PPG feature quantities, evaluation indicators such as the root mean square error (RMSE) and mean absolute error (MAE) between the blood glucose estimated value output by the model and the true value reach the clinically acceptable standard, thereby achieving accurate and reliable non-invasive blood glucose monitoring.

[0091] The reason why three to five invasive blood glucose measurements are required for non-invasive blood glucose detection technology is that the physical differences among individuals are very large, and the patterns and responses of blood glucose fluctuations vary from person to person. To ensure the accuracy of the calculation results of this non-invasive blood glucose device, a calibration must be performed after the subject receives the device to establish an individualized model. And this process of establishing the model requires using the subject's blood glucose value and pulse wave signal as a data set, so that through the linear regression method used in the present invention, namely the least squares fitting, a mathematical mapping relationship between the two can be constructed, and an individualized non-invasive blood glucose detection model for the subject can be established. Only in this way can the blood glucose levels of different subjects with individual differences be accurately predicted.

[0092] The present invention has been described above in an exemplary manner with reference to the accompanying drawings. Obviously, the specific implementation of the present invention is not limited by the above methods. As long as various non-substantial improvements are made using the method concept and technical solution of the present invention, or the concept and technical solution of the present invention are directly applied to other occasions without improvement, they are all within the protection scope of the present invention.

Claims

1. A calibration method for a multi-wavelength non-invasive blood glucose system, characterized in that, It includes the following steps: Step 1: The user continuously detects and records the pulse wave signal, and takes the obtained pulse wave signal measurement value as dataset A; Step 2: During the continuous detection of the pulse wave signal, select some time nodes to perform an invasive blood glucose measurement once and obtain the blood glucose measurement value. Based on the interpolation prediction method, calculate the synchronous blood glucose prediction value within the duration range of the selected time nodes of this user for all blood glucose measurement values, and take it as dataset B; Step 3: Based on dataset A and dataset B, establish an individualized non-invasive blood glucose detection model for this user.

2. The calibration method of the multi-wavelength non-invasive blood glucose system according to claim 1, wherein In step 2, 3 - 5 time nodes are selected; If 3 time nodes are selected: 10 minutes, 40 minutes, and 70 minutes after the end of eating; If 4 time nodes are selected: 10 minutes, 30 minutes, 50 minutes, and 70 minutes after the end of eating; If 5 time nodes are selected: 10 minutes, 25 minutes, 40 minutes, 55 minutes, and 70 minutes after the end of eating; Wherein: The deviation between the actual time of each time node and the set value is within 3 minutes; The deviation between the actual time of the time interval between each time node and the set value is within 3 minutes; The time interval between the first time node and the last time node for each selection method is within 60 - 65 minutes.

3. The calibration method of the multi-wavelength non-invasive blood glucose system according to claim 2, characterized in that, In step 2, the eating time is carried out between 7:00 and 9:30 in the morning. Before eating, the user is in a fasting state. The food ingested by the user is liquid and has a high glycemic index, and the eating time does not exceed 5 minutes.

4. The calibration method for the multi-wavelength non-invasive blood glucose system according to claim 3, wherein At the first time node T1, the blood glucose measurement value obtained by performing an invasive blood glucose measurement is G1, the pulse wave signal measurement value obtained at the same time is PPG1, and the corresponding test maintenance time information is t1. By analogy, T1, G2, PPG2 and T2, t2, T3, G3, PPG3 and t3... are obtained 【PPG1, PPG2, PPG3...】and【t1, t2, t3...】constitute dataset A; 【G1, G2, G3...】and【T1, T2, T3...】constitute dataset B.

5. The calibration method of the multi-wavelength non-invasive blood glucose system according to claim 4, characterized in that In step 2, with time as the abscissa and the blood glucose measurement value as the ordinate, taking each time node T1, T2, T3... and its corresponding invasive blood glucose measurement values G1, G2, G3... as reference data, use the interpolation prediction algorithm to obtain the coordinate points (T1, G1), (T2, G2), (T3, G3)... The curve formed by the coordinate points is the synchronous blood glucose prediction value curve.

6. The calibration method of the multi-wavelength non-invasive blood glucose system according to claim 5, characterized in that In step 3, for the pulse wave signal measurement value recorded at each time point in dataset A, the predicted blood glucose value corresponding to this time point can be found in the synchronous blood glucose prediction value curve. Then the synchronous blood glucose prediction value curve is the mathematical relationship model between the pulse wave signal and the blood glucose value of this user.

7. The calibration method of the multi-wavelength non-invasive blood glucose system according to any one of claims 1-6, characterized in that In step 1, the pulse wave signal processing steps include: 1) Perform preprocessing on the pulse wave signal by filtering, including filtering and removing baseline drift; 2) Through the signal validity detection method of the dynamic threshold curve, use the intersection points of the detected dynamic threshold curve and the pulse wave signal to identify and reject the signal segments with excessive interference, and perform data screening on the pulse wave signal; 3) Align and stack the pulse wave signal cycles, and perform vertical median filtering on the stacked signal to obtain a standard average pulse wave signal; 4) Perform energy operator analysis on the pulse wave signal, extract the characteristics of variance, standard deviation, skewness, DC and AC components, and obtain the absorbance parameters of different blood glucose levels.

8. The calibration method of the multi-wavelength non-invasive blood glucose system according to claim 7, characterized in that It further includes step 4, which optimizes and validates the individualized non-invasive blood glucose detection model through cross-validation and / or error analysis methods, and verifies that when the newly collected pulse wave feature quantities are finally input, the root mean square error and mean absolute error between the blood glucose estimated value output by the model and the true value reach the preset standard.

9. A storage medium, which is a computer-readable storage medium for storing software program codes, and is characterized in that: The software program code is used to execute the multi-wavelength non-invasive blood glucose system calibration method described in any one of claims 1-8.

10. A blood glucose meter, the blood glucose meter being a multi-wavelength non-invasive blood glucose non-invasive blood glucose meter, including a blood glucose meter body, characterized in that: Before each user uses the blood glucose meter, the multi-wavelength non-invasive blood glucose system calibration method described in any one of claims 1-8 is adopted.

Images