Real-time blood glucose monitor and high-precision blood glucose monitoring method thereof

Abstract

The invention relates to the technical field of blood glucose monitoring, in particular to a real-time blood glucose monitor and a high-precision blood glucose monitoring method thereof. The method comprises the following steps: acquiring an analysis curve peak of a daytime period every day in a monitoring period, acquiring a normal diet period according to an overlapping degree of time intervals corresponding to the analysis curve peak in different days, and determining the normal diet period according to the overlapping condition of the time intervals corresponding to the analysis curve peak and the normal diet period. Selecting diet peaks in the analysis curve peaks according to morphological characteristics of different analysis curve peaks; and determining an abnormal index according to the peak value difference and morphological characteristic difference between the analysis curve peak and the diet peak, training a neural network according to the analysis curve peak and the abnormal index of each day except the current day in the monitoring period, and predicting the blood glucose condition in the night time period in the current day by using the trained neural network. According to the method, the attention condition of the analysis curve peak in the neural network process is adaptively set based on the abnormal indexes, and the monitoring precision of the blood sugar of the patient is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of blood glucose monitoring, and particularly relates to a real-time blood glucose monitor and a high-precision blood glucose monitoring method thereof. Background Art

[0002] Blood glucose monitoring is an essential part of diabetes management and can help patients understand their own blood glucose levels. Diabetes allows individuals or caregivers to take action before such health conditions occur to mitigate potential adverse health disorders closely related to changes in glucose levels. It is estimated that nearly half of hypoglycemic episodes and over half of severe episodes occur during nighttime sleep. Hyperglycemia also occurs during nighttime sleep, and patients cannot perceive symptoms in a timely manner at night. Therefore, it is of great significance to predict whether hypoglycemia or diabetes will occur at night based on the patient's blood glucose during the day, so as to prevent it in advance for the patient's physical health.

[0003] A real-time blood glucose monitor uses Continuous Glucose Monitoring (CGM) technology to monitor the blood glucose content in a patient's body and statistically monitor the changes in glucose concentration in subcutaneous interstitial fluid. The CGM device can detect hidden hyperglycemia, hypoglycemia and other phenomena that are not easily detected by traditional blood glucose monitoring methods, especially the blood glucose situation at night.

[0004] Existing methods predict the patient's blood glucose at night based on the patient's blood glucose situation during the day through a neural network. However, due to the different changes in the patient's blood glucose situation during the day, especially the obvious changes in blood glucose before and after meals, the neural network will pay too much attention to the blood glucose changes before and after meals during training, resulting in insufficient attention to the blood glucose changes caused by other factors such as exercise level, mood fluctuations and insulin sensitivity. As a result, there is a deviation between the predicted blood glucose changes of the patient at night by the neural network and the actual blood glucose changes, reducing the accuracy of blood glucose monitoring for the patient. Summary of the Invention

[0005] In order to solve the technical problem that the neural network pays inappropriate attention to the blood glucose changes caused by diet and other factors, resulting in low accuracy in predicting the blood glucose changes at night, the purpose of the present invention is to provide a real-time blood glucose monitor and a high-precision blood glucose monitoring method thereof. The specific technical solutions adopted are as follows:

[0006] The present invention proposes a high-precision blood glucose monitoring method for a real-time blood glucose monitor, and the method includes:

[0007] Obtain the analysis curve peaks in the blood glucose curve of the patient during the daytime period of each day during the monitoring period;

[0008] Obtain the normal diet period according to the overlapping degree of the time intervals corresponding to the peaks of the analysis curves on different days during the monitoring period;

[0009] Select the diet peaks from the analysis curve peaks during the monitoring period according to the overlapping situation of the time intervals corresponding to each analysis curve peak with the normal diet period, and the differences in the morphological characteristics between each analysis curve peak and the other analysis curve peaks during the monitoring period;

[0010] Obtain the abnormal index of each analysis curve peak according to the peak value difference and morphological characteristic difference between each analysis curve peak and the diet peaks during the monitoring period;

[0011] The last day during the monitoring period is the current day. Train a neural network based on the analysis curve peaks of each day except the current day during the monitoring period and their abnormal indexes; Use the trained neural network to predict the blood glucose situation in the night time period of the current day.

[0012] Further, the obtaining of the normal diet period includes:

[0013] The analysis curve peaks include periodic peaks; Denote the central moment of the time interval corresponding to each periodic peak as the analysis moment of the corresponding periodic peak;

[0014] Count the total number of periodic peaks in the analysis curve peaks of each day, denoted as the analysis number of each day; Cluster the analysis moments of all the analysis periodic peaks during the monitoring period to obtain several clustering clusters, and the number of the clustering clusters is equal to the mode of the analysis numbers of all the days during the monitoring period;

[0015] For each clustering cluster, obtain the overlapping interval of the time intervals corresponding to the periodic peaks with the analysis moments within the clustering cluster, count the total number of periodic peaks corresponding to the time intervals constituting each overlapping interval, denoted as the overlapping number of the corresponding overlapping interval; For the overlapping interval of the clustering cluster, denote the overlapping interval corresponding to the largest overlapping number as the normal diet period.

[0016] Further, the selection of the diet peaks from the analysis curve peaks during the monitoring period includes:

[0017] If the time interval corresponding to each periodic peak only contains any one normal diet period, then regard the corresponding periodic peak as the first diet peak;

[0018] The analysis curve peaks also include aperiodic peaks; Select the second diet peaks from the aperiodic peaks during the monitoring period according to the overlapping situation of the time intervals corresponding to each aperiodic peak with the normal diet period, and the differences in the morphological characteristics between each aperiodic peak and the first diet peaks during the monitoring period;

[0019] Regard both the first diet peaks and the second diet peaks as diet peaks.

[0020] Further, selecting the second dietary peak from the aperiodic peaks during the monitoring period includes:

[0021] Dividing each analytical curve peak into different sub-curves, and obtaining the morphological index of each analytical curve peak according to the inclination degree of the sub-curves of each analytical curve peak;

[0022] If the time interval corresponding to each aperiodic peak only overlaps with any one normal dietary period, the corresponding aperiodic peak is recorded as a peak to be tested;

[0023] Arrange the first dietary peaks of each day in chronological order to obtain the peak sequence of each day; obtain the time intervals between all the first dietary peaks of each day and the analysis moments of each peak to be tested, and record the subscript value of the first dietary peak corresponding to the shortest time interval in its peak sequence as the analysis subscript value of each peak to be tested;

[0024] Take the mean value of the morphological indexes of the first dietary peaks with the analysis subscript value of each peak to be tested in the peak sequences of all days during the monitoring period as the overall morphological value of each peak to be tested; normalize the absolute value of the difference between the morphological index of each peak to be tested and the overall morphological value to obtain the morphological difference value of each peak to be tested.

[0025] For all the peaks to be tested during the monitoring period, select the peaks to be tested corresponding to the morphological difference values less than the preset difference threshold as the second dietary peaks.

[0026] Further, obtaining the morphological index of each analytical curve peak includes:

[0027] Based on the peak point, divide each analytical curve peak into two sub-curves; the sub-curves include: an ascending sub-curve and a descending sub-curve;

[0028] Take the ratio of the absolute value of the difference between the amplitudes of the two end points of each sub-curve to the time interval as the blood glucose change rate of the corresponding sub-curve; take the ratio of the blood glucose change rates of the ascending sub-curve and the descending sub-curve of each analytical curve peak as the morphological index of each analytical curve peak.

[0029] Further, obtaining the abnormal index of each analytical curve peak includes:

[0030] Obtain the mean value of the peak values of all dietary peaks during the monitoring period and record it as the reference value;

[0031] For each analytical curve peak during the monitoring period, when the peak value of the analytical curve peak is less than or equal to the reference value, take the morphological difference value of the analytical curve peak as the abnormal index of the analytical curve peak;

[0032] When the peak value of the analysis curve peak is greater than the reference value, obtain the ratio of the difference between the peak value of the analysis curve peak and the reference value to the peak value of the analysis curve peak, and use the sum value of the ratio and the constant 1 as the adjustment coefficient of the analysis curve peak; use the adjustment coefficient to perform weighted processing on the morphological difference value of the analysis curve peak to obtain the anomaly index of the analysis curve peak.

[0033] Further, the training of the neural network based on the analysis curve peaks of the remaining days except the current day during the monitoring period includes:

[0034] Record the analysis curve peaks except the diet peaks during the monitoring period as anomaly factor peaks; the blood glucose curve also includes a baseline curve. According to the number and the anomaly index of the baseline curve, anomaly factor peaks, and diet peaks in the blood glucose curve of each day during the monitoring period, obtain the weight indexes of the baseline curve, anomaly factor peaks, and diet peaks of each day.

[0035] Obtain the blood glucose curve in the night time period of each day; use the weight indexes of the baseline curve, anomaly factor peaks, and diet peaks of each day as the initial weights of the Attention model, and use the blood glucose curve in the day time period of each day as the input data of the training sample and the blood glucose curve in the night time period as the output data of the training sample respectively; the training samples of the remaining days except the current day during the monitoring period constitute the training set of the Attention model; use the training set to train the Attention model.

[0036] Further, the obtaining of the weight indexes of the baseline curve, anomaly factor peaks, and diet peaks of each day includes:

[0037] Set the reference weight factor as w; set the weight index of the baseline curve in the blood glucose curve of each day during the monitoring period as w, the weight index of the anomaly factor peak as 2w, and the weight index of the i-th diet peak as w×(1 + p i ) ; where p i is the anomaly index of the i-th diet peak in the blood glucose curve of each day during the monitoring period.

[0038] The method for obtaining the reference weight factor w is:

[0039] In the formula, n1 is the total number of baseline curves in the blood glucose curve of each day during the monitoring period; n2 is the total number of anomaly factor peaks in the blood glucose curve of each day during the monitoring period; n3 is the total number of diet peaks in the blood glucose curve of each day during the monitoring period.

[0040] Further, the prediction of the blood glucose situation in the night time period of the current day by using the trained neural network includes:

[0041] Input the blood glucose curve of the patient during the daytime period of the current day into the trained Attention model, and the Attention model outputs the blood glucose curve during the nighttime period of the current day;

[0042] If the amplitude of the data point on the blood glucose curve during the nighttime period of the current day is less than the first preset threshold or greater than the second preset threshold, the patient will have abnormal blood glucose during the nighttime period of the current day.

[0043] A real-time blood glucose monitor, the monitor includes: a data acquisition module, a diet period analysis module, a diet peak screening module, an abnormality analysis module, and a blood glucose monitoring module;

[0044] The data acquisition module is used to obtain the analysis curve peak in the blood glucose curve of the patient during the daytime period of each day during the monitoring period;

[0045] The diet period analysis module is used to obtain the normal diet period according to the overlapping degree of the time intervals corresponding to the analysis curve peaks on different days during the monitoring period;

[0046] The diet peak screening module is used to select diet peaks from the analysis curve peaks during the monitoring period according to the overlapping situation between the time interval corresponding to each analysis curve peak and the normal diet period, and the difference in the morphological characteristics between each analysis curve peak and the other analysis curve peaks during the monitoring period;

[0047] The abnormality analysis module is used to obtain the abnormality index of each analysis curve peak according to the peak difference and morphological characteristic difference between each analysis curve peak and the diet peak during the monitoring period;

[0048] The blood glucose monitoring module is used to use the analysis curve peaks of each day except the current day during the monitoring period and their abnormality indexes to train a neural network when the last day during the monitoring period is the current day; use the trained neural network to predict the blood glucose situation during the nighttime period of the current day.

[0049] The present invention has the following beneficial effects:

[0050] In the embodiments of the present invention, due to the regularity of people's eating habits, the normal eating period is obtained by monitoring the overlapping degree of the peaks of the analysis curves on different days during the monitoring period; the morphological characteristics of the peaks of the analysis curves reflect the increase and decrease of the blood glucose concentration they represent, and the peaks of the analysis curves overlapping with the normal eating period have a greater possibility of blood glucose changes caused by diet. Therefore, the overlapping situation between the time interval corresponding to the peak of the analysis curve and the normal eating period, as well as the differences in the morphological characteristics between the peak of the analysis curve and the other peaks of the analysis curve, are used to screen the diet peaks that cause blood glucose changes due to diet; the diet peaks are inevitable blood glucose fluctuations in normal human life. The peak value of the analysis curve peak represents the highest blood glucose fluctuation level, and the morphological characteristics represent the relative changes in the increase and decrease rates of blood glucose. Both can represent the fluctuation situation of the analysis curve peak. Therefore, the differences in the peak values and morphological characteristics between the analysis curve peak and the diet peak during the monitoring period can reflect the differences between the analysis curve peak and the blood glucose fluctuations in normal human life, obtain abnormal indicators, and measure the abnormal degree of the blood glucose change of the analysis curve peak; based on the abnormal indicators, the attention of the neural network to the analysis curve peak during the training process is adaptively adjusted, which helps the model learn the complex relationship between blood glucose during the day and at night more accurately, thereby improving the accuracy of the neural network in predicting blood glucose changes at night, making the predicted blood glucose changes at night closer to the actual blood glucose changes, and improving the monitoring accuracy of the patient's blood glucose. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0052] Figure 1 It is a flowchart of the steps of a high-precision blood glucose monitoring method for a blood glucose real-time monitor provided by an embodiment of the present invention;

[0053] Figure 2 It is a flowchart of the steps of a method for selecting diet peaks provided by an embodiment of the present invention;

[0054] Figure 3 It is a structural block diagram of a blood glucose real-time monitor provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0055] To further elaborate on the technical means and effects adopted by the present invention to achieve the intended invention purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details a real-time blood glucose monitor and its high-precision blood glucose monitoring method proposed according to the present invention, including its specific implementation manner, structure, features, and effects. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0056] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which this invention belongs.

[0057] The specific scenario targeted by the present invention: When using a real-time blood glucose monitor to monitor a patient's blood glucose, since the patient cannot timely perceive symptoms at night, the neural network can predict the patient's blood glucose situation at night in advance based on the blood glucose data during the day and give an early warning of abnormalities. The neural network monitors the patient's blood glucose situation at night through the blood glucose data during the day. Because diet can cause changes in the patient's blood glucose, it leads to excessive attention to the blood glucose changes caused by diet during the neural network training, thereby affecting the training results of the neural network.

[0058] The following specifically describes the specific solutions of a real-time blood glucose monitor and its high-precision blood glucose monitoring method provided by the present invention in conjunction with the accompanying drawings.

[0059] Embodiment 1:

[0060] The present invention proposes a high-precision blood glucose monitoring method for a real-time blood glucose monitor. Please refer to Figure 1 , which shows the step flowchart of a high-precision blood glucose monitoring method for a real-time blood glucose monitor provided by an embodiment of the present invention. The method includes:

[0061] Step S1: Obtain the analysis curve peaks in the blood glucose curves during the daytime of each day within the monitoring period of the patient.

[0062] This solution uses a real-time blood glucose monitor to monitor the blood glucose content in the patient's body. The real-time blood glucose monitor uses continuous glucose monitoring technology to monitor the blood glucose content in the patient's body. The continuous glucose monitoring technology continuously monitors the glucose concentration in the interstitial fluid of the subcutaneous tissue through a glucose sensor, and can continuously record the trend and characteristics of blood glucose fluctuations, helping patients understand the blood glucose changes caused by diet, exercise, stress, sleep, and hypoglycemic drugs.

[0063] The patient implants a glucose sensor under the subcutaneous tissue to record the electrical signals generated by the glucose oxidation reaction in the interstitial fluid, and the electrical signals reflect the blood glucose concentration level; the glucose sensor communicates with a transmitter on the skin to send the blood glucose information to a receiver that can receive the information. After the receiver receives the data, it displays the glucose concentration of the patient's interstitial fluid on the display screen. Obtain the blood glucose curves during the daytime and nighttime periods of each day during the monitoring period of the patient.

[0064] It should be noted that the blood glucose data is displayed on the display screen through a two-dimensional coordinate system. The horizontal axis of the two-dimensional coordinate system is time, and the vertical axis is the blood glucose concentration. The unit of the blood glucose concentration is millimoles per liter. The duration of the monitoring period is 7 days. The last day during the monitoring period is the current day. The daytime period is from 6 o'clock to 20 o'clock, and the nighttime period is from 20 o'clock to 6 o'clock the next day. The implementer can set it according to the specific situation.

[0065] In the embodiment of the present invention, the blood glucose data at all times during the daytime period are connected by a smooth curve to obtain the blood glucose curve during the daytime period. In other embodiments, the data points of the blood glucose data at all times during the daytime period can also be curve-fitted, and the obtained fitted curve is used as the blood glucose curve during the daytime period. Among them, the method of curve fitting is the least squares method, which is a well-known technique to those skilled in the art and will not be elaborated here.

[0066] It should be noted that the method for obtaining the blood glucose curves during the nighttime period is the same as that during the daytime period; the amplitude of each data point on the blood glucose curve is the blood glucose concentration corresponding to the time of that data point.

[0067] People's eating habits are regular. In order to identify whether the blood glucose changes during the daytime period are caused by diet or other factors, the blood glucose curve during the daytime period is divided to obtain a baseline curve, periodic peaks, and non-periodic peaks; the periodic peaks and non-periodic peaks are recorded as analysis curve peaks.

[0068] In this embodiment, the method for dividing the blood glucose curve is as follows: Use the Automatic Multiscale-based Peak Detection (AMPD) algorithm to obtain the periodic peak points and non-periodic peak points on the blood glucose curve during the daytime period of each day, and record the two as analysis points; on the blood glucose curve, the curve between the first data points smaller than the reference threshold on the left and right sides of each analysis point is used as the curve peak corresponding to the analysis point. The curve peaks of the periodic peak points are respectively recorded as periodic peaks, and the curve peaks of the non-periodic peak points are recorded as non-periodic peaks. On the blood glucose curve, the continuous data points with amplitudes smaller than the reference threshold between the right endpoint of any analysis curve peak and the left endpoint of the next adjacent analysis curve peak are smoothly connected to obtain the baseline curve. Among them, the AMPD algorithm is a well-known technique to those skilled in the art and will not be elaborated here.

[0069] It should be noted that on the blood glucose curve, the amplitude of the data points between each analysis point and the first data point smaller than the reference threshold on its left or right is greater than the reference threshold. Since the blood glucose value when a person is fasting is between 3.9 and 6.1 mmol / L, the reference threshold takes the empirical value of 6.1 mmol / L, and the implementer can set it according to the specific situation.

[0070] The distribution of periodic peaks on the blood glucose curve is regular. Since people's eating habits are regular, it is more likely that the periodic peaks are caused by blood glucose changes due to diet; non-periodic peaks are generally caused by blood glucose changes due to other reasons, such as exercise level, mood swings, and insulin sensitivity, etc.; the baseline curve is the normal blood glucose fluctuation of the human body and is relatively stable.

[0071] Step S2: Obtain the normal eating period according to the overlapping degree of the time intervals corresponding to the analysis curve peaks on different days during the monitoring period.

[0072] The time interval corresponding to the analysis curve peak is the time period between the corresponding moments of the two end points of the analysis curve peak.

[0073] Preferably, in some possible implementation manners of the embodiments of the present invention, the method for obtaining the normal eating period includes: the analysis curve peaks include periodic peaks; record the central moment of the time interval corresponding to each periodic peak as the analysis moment of the corresponding periodic peak; count the total number of periodic peaks in the analysis curve peaks of each day, denoted as the analysis number of each day; perform clustering on the analysis moments of all the analysis periodic peaks during the monitoring period to obtain several clustering clusters, and the number of clustering clusters is equal to the mode of the analysis numbers of all days during the monitoring period; for each clustering cluster, obtain the overlapping interval of the time intervals corresponding to the periodic peaks with the analysis moments within the clustering cluster, and count the total number of periodic peaks corresponding to the time intervals constituting each overlapping interval, denoted as the overlapping number of the corresponding overlapping interval; for the overlapping intervals of the clustering clusters, denote the overlapping interval corresponding to the largest overlapping number as the normal eating period.

[0074] Since people's eating habits are regular and it is more likely that the periodic peaks are caused by blood glucose changes due to diet, the normal eating period can be obtained according to the overlapping degree of the periodic peaks on different days during the monitoring period. Since most of the time of the patient during the monitoring period has regular eating habits, and the eating regularity may be disrupted for some reasons during part of the time, the analysis number is equivalent to the number of eating times of the patient each day, so the number of clustering clusters, that is, the number of normal eating periods, should be equal to the mode of the analysis numbers of all days during the monitoring period.

[0075] It should be noted that in this embodiment, the K-means clustering algorithm is selected for clustering, and the value of K is equal to the mode of the analysis numbers of all days during the monitoring period. Among them, the K-means clustering algorithm is a well-known technology to those skilled in the art and will not be elaborated here.

[0076] Since people's eating periods are relatively concentrated and there will be overlaps in eating periods, the eating periods can be determined by the degree of overlap of eating times. Therefore, by obtaining the overlapping intervals of the time intervals corresponding to the periodic peaks at the analysis moments within the clustering clusters, the greater the total number of periodic peaks corresponding to the time intervals of each overlapping interval, the greater the likelihood that the overlapping interval is the normal eating period of the patient. Thus, the overlapping interval corresponding to the largest number of overlaps is recorded as the normal eating period.

[0077] As an example, 6:00 to 8:00 and 7:00 to 8:00 are two time intervals, and the overlapping interval of these two time intervals is 7:00 to 8:00. The time intervals constituting this overlapping interval include: 6:00 to 8:00 and 7:00 to 8:00.

[0078] Step S3: Select the eating peaks from the analysis curve peaks during the monitoring period according to the overlapping situation of the time intervals corresponding to each analysis curve peak and the normal eating period, and the differences in the morphological characteristics between each analysis curve peak and the other analysis curve peaks during the monitoring period.

[0079] Please refer to Figure 2 , which shows the flowchart of the steps of a method for selecting eating peaks provided by an embodiment of the present invention. The method includes:

[0080] Step S310: If the time interval corresponding to each periodic peak only contains any one normal eating period, then take the corresponding periodic peak as the first eating peak.

[0081] When the patient is eating, they may advance or delay part of the time. If the time interval corresponding to the periodic peak contains any one normal eating period, then the time interval corresponding to the periodic peak is within the normal eating time, and the periodic peak is the blood glucose change caused by eating.

[0082] The analysis curve peaks also include non-periodic peaks; select the second eating peaks from the non-periodic peaks during the monitoring period according to the overlapping situation of the time intervals corresponding to each non-periodic peak and the normal eating period, and the differences in the morphological characteristics between each non-periodic peak and the first eating peaks during the monitoring period.

[0083] The first sub-step: Divide each analysis curve peak into different sub-curves, and obtain the morphological index of each analysis curve peak according to the inclination degree of the sub-curves of each analysis curve peak.

[0084] The analysis curve peak shows the rising and falling situations of blood glucose. The blood glucose elevation caused by eating, exercise level, and emotional changes is closely related to the patient's physical condition; in the short-term dietary intake, the absorption rate and metabolism rate of blood glucose in the human body are approximately the same, and the blood glucose elevation rate and decline rate of the analysis curve peak show abnormal changes in blood glucose.

[0085] The inclination degree of the sub-curve of the analysis curve peak reflects the rising or falling rate of blood glucose, can describe the morphological characteristics of the analysis curve peak, and obtain morphological indicators.

[0086] Preferably, in some possible implementation manners of the embodiments of the present invention, the method for obtaining morphological indicators includes: dividing each analysis curve peak into two sub-curves based on the peak point; the sub-curves include: a rising sub-curve and a falling sub-curve; taking the ratio of the absolute value of the difference between the amplitudes of the two end points of each sub-curve to the time interval as the blood glucose change rate of the corresponding sub-curve; taking the ratio of the blood glucose change rates of the rising sub-curve and the falling sub-curve of each analysis curve peak as the morphological indicator of each analysis curve peak.

[0087] The blood glucose change rate reflects the rising or falling rate of blood glucose in the patient's body during the time period corresponding to the sub-curve; since the blood glucose usually has a faster rising rate and a slower falling rate, the morphological indicator reflects the relative situation of blood glucose change within the corresponding time of the analysis curve peak.

[0088] It should be noted that the curve between the left end point and the peak point of the analysis curve peak is the rising sub-curve, and the curve between the peak point and the right end point of the analysis curve peak is the falling sub-curve. Since the corresponding times of the two end points of the sub-curve must be unequal, the blood glucose change rate of the sub-curve and the morphological indicator of the analysis curve peak must be meaningful.

[0089] The second sub-step: If the time interval corresponding to each non-periodic peak only overlaps with any one normal diet period, then mark the corresponding non-periodic peak as a to-be-detected peak.

[0090] Part of the time may, for various reasons, disrupt the diet regularity, resulting in the blood glucose change caused by diet being recognized as a non-periodic peak, that is, there are some differences between the time intervals corresponding to the non-periodic peak and the periodic peak. When the time interval corresponding to the non-periodic peak only overlaps with any one normal diet period, it indicates that the non-periodic peak is still within the diet time, and the non-periodic peak is more likely to be a blood glucose change caused by diet.

[0091] The third sub-step: Screen the second diet peak from the to-be-detected peaks.

[0092] Since the blood glucose level of the human body may vary in different time periods, in order to improve the accuracy of screening the second diet peak from non-periodic peaks, select the first diet peak closest to the non-periodic peak during the monitoring period, and screen the second diet peak by comparing the difference degrees of the morphological characteristics of these first diet peaks and the non-periodic peak. The specific operation is as follows:

[0093] Arrange the first dietary peaks of each day in chronological order to obtain the peak sequence of each day; obtain the time intervals between all the first dietary peaks of each day and the analysis time of each peak to be measured, and record the subscript value of the first dietary peak corresponding to the shortest time interval in its peak sequence as the analysis subscript value of each peak to be measured; take the mean value of the morphological indexes of the first dietary peaks with the analysis subscript value of each peak to be measured in the peak sequences of all days during the monitoring period as the overall morphological value of each peak to be measured; normalize the absolute value of the difference between the morphological index of each peak to be measured and the overall morphological value to obtain the morphological difference value of each peak to be measured.

[0094] It should be noted that the blood glucose changes of the first dietary peak corresponding to the shortest time interval and each peak to be measured are relatively close. The overall morphological value of the peak to be measured reflects the overall situation of the blood glucose changes caused by dietary changes during the normal diet time of the peak to be measured; as an example, if two normal diet periods are from 8:00 to 9:00 and from 12:00 to 13:00, and the time interval corresponding to the peak to be measured is from 9:00 to 10:00, and the time interval from 8:00 to 9:00 is the closest to the time interval corresponding to the peak to be measured, then the overall morphological value of the peak to be measured is obtained according to the morphological indexes of the first dietary peaks in the morning during all days of the monitoring period. The method for obtaining the analysis time of aperiodic peaks and periodic peaks is the same.

[0095] The morphological difference value of each peak to be measured is expressed by the formula:

[0096]

[0097] In the formula, q is the morphological difference value of each peak to be measured; r is the morphological index of each peak to be measured; U is the total number of peak sequences, that is, the total number of days during the monitoring period; b is the analysis subscript value of each peak to be measured; r u,b is the morphological index of the first dietary peak with the subscript b in the u-th peak sequence; is the overall morphological value of each peak to be measured; || is the absolute value function; Norm is the normalization function.

[0098] If the morphological difference value of the peak to be measured is smaller, it indicates that the blood glucose change reflected by the peak to be measured is closer to the blood glucose change caused by diet at the corresponding time, and the peak to be measured is more in line with the blood glucose change caused by diet. Then the possibility that the peak to be measured is caused by the blood glucose change due to diet is greater. For all peaks to be measured during the monitoring period, select the peaks to be measured corresponding to the morphological difference values less than the preset difference threshold as the second dietary peaks.

[0099] It should be noted that in the embodiment of the present invention, the preset difference threshold takes an empirical value of 0.2, and the implementer can set it according to the specific situation. The morphological difference value of each peak to be measured during the monitoring period is obtained by the above method.

[0100] Step S330: Record both the first dietary peaks and the second dietary peaks as dietary peaks.

[0101] The first dietary peak is selected from the periodic peaks, and the second dietary peak is selected from the aperiodic peaks; the dietary peak is the blood glucose change caused by diet.

[0102] Step S4: Obtain the abnormality index of each analysis curve peak according to the peak difference and morphological feature difference between each analysis curve peak and the dietary peak during the monitoring period.

[0103] The dietary peak is an inevitable blood glucose fluctuation in normal human life. The peak value of the analysis curve peak represents the highest blood glucose fluctuation level, and the morphological feature represents the relative change of the rising and falling rates of blood glucose. Both can represent the fluctuation of the analysis curve peak. Therefore, the peak difference and morphological feature difference between the analysis curve peak and the dietary peak during the monitoring period can reflect the difference between the analysis curve peak and the blood glucose fluctuation in normal human life, so as to obtain the abnormality index and measure the abnormality degree of the blood glucose change of the analysis curve peak.

[0104] Preferably, in some possible implementation manners of the embodiment of the present invention, the method for obtaining the abnormality index includes: obtaining the average value of the peak values of all dietary peaks during the monitoring period and recording it as the reference value; for each analysis curve peak during the monitoring period, when the peak value of the analysis curve peak is less than or equal to the reference value, taking the morphological difference value of the analysis curve peak as the abnormality index of the analysis curve peak; when the peak value of the analysis curve peak is greater than the reference value, obtaining the ratio of the difference between the peak value of the analysis curve peak and the reference value to the peak value of the analysis curve peak, and taking the sum value of the ratio and the constant 1 as the adjustment coefficient of the analysis curve peak; using the adjustment coefficient to perform weighted processing on the morphological difference value of the analysis curve peak to obtain the abnormality index of the analysis curve peak.

[0105] The dietary peak is the blood glucose change caused by diet and cannot be avoided. The reference value is the highest level of blood glucose change in normal human life. The morphological difference value presents the degree to which the blood glucose change represented by the analysis curve peak deviates from the normal blood glucose change of the human body, reflecting the abnormality degree of the blood glucose change represented by the analysis curve peak.

[0106] When the peak value of the analysis curve peak is less than or equal to the reference value, the blood glucose change represented by the analysis curve peak does not exceed the blood glucose change level in normal human life, and the morphological difference value is sufficient to represent the abnormality degree of the blood glucose change of the analysis curve peak. When the peak value of the analysis curve peak is greater than the reference value, the blood glucose change represented by the analysis curve peak exceeds the normal blood glucose change level of the human body. The greater the abnormality degree of the blood glucose change represented by the analysis curve peak, the morphological difference value is adjusted based on the difference between the peak value of the analysis curve peak and the reference value to obtain the abnormality index of the analysis curve peak.

[0107] It should be noted that if the peak of the analysis curve is a non-periodic peak, the method for obtaining the morphological difference value between the peak of the analysis curve and the peak to be measured is exactly the same; if the peak of the analysis curve is a periodic peak, the subscript value of the peak of the analysis curve in the peak sequence to which it belongs is recorded as the analysis subscript value of each peak of the analysis curve, and the other acquisition steps are the same.

[0108] In a specific implementation manner of the embodiment of the present invention, the abnormal index of the peak of the analysis curve is expressed by the formula:

[0109]

[0110] In the formula, p v is the abnormal index of the v-th peak of the analysis curve during the monitoring period; H v is the peak value of the v-th peak of the analysis curve during the monitoring period; is the reference value; y v is the morphological difference value of the v-th peak of the analysis curve during the monitoring period; is the adjustment coefficient of the v-th peak of the analysis curve during the monitoring period.

[0111] Step S5: The last day during the monitoring period is the current day. Train the neural network according to the peaks of the analysis curve and their abnormal indexes of each day except the current day during the monitoring period; use the trained neural network to predict the blood glucose situation in the night time period of the current day.

[0112] The peaks of the analysis curve except the diet peak during the monitoring period are recorded as abnormal factor peaks; the blood glucose curve also includes a baseline curve. According to the quantity and abnormal indexes of the baseline curve, abnormal factor peaks and diet peaks in the blood glucose curve of each day during the monitoring period, obtain the weight indexes of the baseline curve, abnormal factor peaks and diet peaks of each day. The specific operation is as follows:

[0113] Set the reference weight factor as w; set the weight index of the baseline curve in the blood glucose curve of each day during the monitoring period as w, the weight index of the abnormal factor peak as 2w, and the weight index of the i-th diet peak as w×(1 + p i ); where p i is the abnormal index of the i-th diet peak in the blood glucose curve of each day during the monitoring period.

[0114] Record the baseline curve, abnormal factor peaks and diet peaks as the peaks to be analyzed; the greater the abnormal index of the peak to be analyzed, the greater the abnormal degree of the blood glucose change represented by it. When training the neural network, it is necessary to increase the weight of this peak of the analysis curve to improve the attention degree of the neural network to this peak of the analysis curve and improve the accuracy of predicting the blood glucose curve in the night time period.

[0115] It should be noted that the baseline curve represents the normal blood glucose fluctuations in the human body, the dietary peak represents the blood glucose fluctuations caused by diet, and the abnormal factor peak represents the blood glucose fluctuations caused by other factors such as exercise level, emotional fluctuations, and insulin sensitivity. When training the neural network, compared with the dietary peak of normal blood glucose fluctuations caused by diet, the abnormal factor peak is the key object of concern. Therefore, the weight of the baseline curve is the smallest, the weight of the abnormal factor peak is the largest, and the weight of the dietary peak is between the two.

[0116] Since the sum of the weights of the baseline curve, dietary peak, and abnormal factor peak in the blood glucose curve during the daytime of each day is a constant 1, the method for obtaining the reference weight factor w is as follows:

[0117]

[0118] In the formula, n1 is the total number of baseline curves in the blood glucose curves of each day during the monitoring period; n2 is the total number of abnormal factor peaks in the blood glucose curves of each day during the monitoring period; n3 is the total number of dietary peaks in the blood glucose curves of each day during the monitoring period.

[0119] Obtain the blood glucose curve during the night time period of each day; use the weight indicators of the baseline curve, abnormal factor peak, and dietary peak of each day as the initial weights of the Attention model. Respectively, use the blood glucose curves during the daytime of each day as the input data of the training samples, and the blood glucose curves during the night time period as the output data of the training samples; use the training samples of the remaining days except the current day during the monitoring period to form the training set of the Attention model; use the training set to train the Attention model.

[0120] It should be noted that the loss function of the Attention model is the mean square error function. Training the neural network using the training set is a well-known technique for those skilled in the art and will not be elaborated here. If the weight indicator of the peak to be analyzed is larger, it means that the Attention model will give priority consideration when processing the peak to be analyzed, thereby better predicting the blood glucose changes during the night time period.

[0121] Input the blood glucose curve during the daytime of the current day of the patient into the trained Attention model, and the Attention model outputs the blood glucose curve during the night time period of the current day; if the amplitude of the data points on the blood glucose curve during the night time period of the current day is less than the first preset threshold or greater than the second preset threshold, then the patient will have abnormal blood glucose during the night time period of the current day.

[0122] The specific preventive measures are as follows: If the amplitude of the data points on the blood glucose curve during the night period of the current day is less than the first preset threshold, it indicates that the patient may experience hypoglycemic symptoms during the night period of the current day, and the patient needs to appropriately ingest sugary foods before going to bed; if the amplitude is greater than the second preset threshold, it indicates that the patient may experience hyperglycemic symptoms during the night period of the current day, and the patient needs to use drugs such as insulin injection and metformin hydrochloride tablets before going to bed to reduce the blood glucose level at night.

[0123] It should be noted that the human body is in a fasting state at night, and the normal blood glucose value range at night is 3.9 to 6.1 millimoles per liter. Therefore, the first preset threshold takes the empirical value of 3.9 millimoles per liter, and the second preset threshold takes the empirical value of 6.1 millimoles per liter. The implementer can set it according to the specific situation.

[0124] Taking the weight index of the peak to be analyzed as the initial weight of the Attention model means that the model will give priority to these peaks to be analyzed with high weight indexes when analyzing the blood glucose curve during the daytime period, helping the model to determine the importance of the blood glucose curves in different time periods at the initial stage of training. During the training process, these weights will be adjusted through the self-optimization mechanism of the model, and finally help the model to more accurately learn the complex relationship between daytime and nighttime blood glucose, thereby improving the prediction accuracy.

[0125] So far, the present invention is completed.

[0126] Embodiment 2:

[0127] As Figure 3 shown, based on the same concept as in Embodiment 1 above, this embodiment also proposes a blood glucose real-time monitor, which includes: a data acquisition module, a diet time period analysis module, a diet peak screening module, an abnormal analysis module, and a blood glucose monitoring module, including:

[0128] A data acquisition module, configured to obtain the analysis curve peaks in the blood glucose curve during the daytime period of each day during the monitoring period of the patient;

[0129] A diet time period analysis module, configured to obtain the normal diet time period according to the overlapping degree of the time intervals corresponding to the analysis curve peaks on different days during the monitoring period;

[0130] A diet peak screening module, configured to select diet peaks from the analysis curve peaks during the monitoring period according to the overlapping situation between the time interval corresponding to each analysis curve peak and the normal diet time period, and the morphological feature differences between each analysis curve peak and the other analysis curve peaks during the monitoring period;

[0131] An abnormal analysis module, configured to obtain the abnormal index of each analysis curve peak according to the peak value difference and morphological feature difference between each analysis curve peak and the diet peaks during the monitoring period;

[0132] The blood glucose monitoring module is used to set the last day during the monitoring period as the current day, and train a neural network based on the peak of the analysis curve and its abnormal indicators for each day except the current day during the monitoring period; use the trained neural network to predict the blood glucose situation during the night period of the current day.

[0133] It should be understood that the device provided in this embodiment is used to execute the high-precision blood glucose monitoring method of the above-mentioned blood glucose real-time monitor, so the same effect as the above implementation method can be achieved.

[0134] This embodiment also provides an electronic device, including a processor, a memory, and a program or instruction stored on the memory and running on the processor. When the program or instruction is executed by the processor, the steps of the high-precision blood glucose monitoring method of the above-mentioned blood glucose real-time monitor are implemented.

[0135] This embodiment also provides a readable storage medium, on which a program or instruction is stored. When the program or instruction is executed by a processor, the steps of the high-precision blood glucose monitoring method of the above-mentioned blood glucose real-time monitor are implemented.

[0136] The above are only the preferred embodiments of the present invention, and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A high-precision blood glucose monitoring method for a real-time blood glucose monitor, characterized in that, The method includes: Obtaining the analysis curve peaks in the blood glucose curve during the daytime of each day within the monitoring period of the patient; Obtaining the normal diet period according to the overlapping degree of the time intervals corresponding to the analysis curve peaks on different days within the monitoring period; Selecting diet peaks from the analysis curve peaks within the monitoring period according to the overlapping situation between the time interval corresponding to each analysis curve peak and the normal diet period, and the difference in the morphological characteristics between each analysis curve peak and the remaining analysis curve peaks within the monitoring period; Obtaining the abnormal index of each analysis curve peak according to the peak value difference and morphological characteristic difference between each analysis curve peak and the diet peaks within the monitoring period; Taking the last day within the monitoring period as the current day, training a neural network according to the analysis curve peaks of each day except the current day within the monitoring period and their abnormal indexes; predicting the blood glucose situation during the night time period of the current day by using the trained neural network.

2. The high-precision blood glucose monitoring method of a real-time blood glucose monitor according to claim 1, characterized in that, The obtaining of the normal diet period includes: The analysis curve peaks include periodic peaks; recording the central moment of the time interval corresponding to each periodic peak as the analysis moment of the corresponding periodic peak; Counting the total number of periodic peaks in the analysis curve peaks of each day, which is denoted as the analysis number of each day; clustering the analysis moments of all the analysis periodic peaks within the monitoring period to obtain several clustering clusters, and the number of the clustering clusters is equal to the mode of the analysis numbers of all days within the monitoring period; For each clustering cluster, obtaining the overlapping interval of the time intervals corresponding to the periodic peaks with analysis moments within the clustering cluster, counting the total number of periodic peaks corresponding to the time intervals constituting each overlapping interval, which is denoted as the overlapping number of the corresponding overlapping interval; for the overlapping interval of the clustering cluster, denoting the overlapping interval corresponding to the largest overlapping number as the normal diet period.

3. A high-precision blood glucose monitoring method for a real-time blood glucose monitor according to claim 2, characterized in that, The selecting of the diet peaks from the analysis curve peaks within the monitoring period includes: If the time interval corresponding to each periodic peak only contains any one normal diet period, taking the corresponding periodic peak as the first diet peak; The analysis curve peaks further include non-periodic peaks; selecting the second diet peaks from the non-periodic peaks within the monitoring period according to the overlapping situation between the time interval corresponding to each non-periodic peak and the normal diet period, and the difference in the morphological characteristics between each non-periodic peak and the first diet peaks within the monitoring period; Denoting both the first diet peaks and the second diet peaks as diet peaks.

4. A high-precision blood glucose monitoring method for a real-time blood glucose monitor according to claim 3, characterized in that, The selecting of the second diet peaks from the non-periodic peaks within the monitoring period includes: Dividing each analysis curve peak into different sub-curves, and obtaining the morphological index of each analysis curve peak according to the inclination degree of the sub-curves of each analysis curve peak; If the time interval corresponding to each non-periodic peak only overlaps with any one normal diet period, denoting the corresponding non-periodic peak as a to-be-tested peak; Arranging the first diet peaks of each day in chronological order to obtain the peak sequence of each day; obtaining the time intervals between all the first diet peaks of each day and the analysis moments of each to-be-tested peak respectively, and denoting the subscript value of the first diet peak corresponding to the shortest time interval in its peak sequence as the analysis subscript value of each to-be-tested peak; The mean of the morphological indices of the first dietary peak with the analytical subscript value of each peak to be measured in the peak sequence for all days during the monitoring period is taken as the overall morphological value of each peak to be measured; the absolute value of the difference between the morphological index of each peak to be measured and the overall morphological value is normalized to obtain the morphological difference value of each peak to be measured. For all peaks to be measured during the monitoring period, the peaks to be measured corresponding to the morphological difference values less than the preset difference threshold are selected as the second dietary peaks.

5. A high-precision blood glucose monitoring method for a real-time blood glucose monitor according to claim 4, characterized in that, The obtaining of the morphological index of each analytical curve peak includes: Each analytical curve peak is divided into two sub-curves based on the peak point; the sub-curves include: an ascending sub-curve and a descending sub-curve; The ratio of the absolute value of the difference between the amplitudes of the two end points of each sub-curve to the time interval is taken as the blood glucose change rate of the corresponding sub-curve; the ratio of the blood glucose change rates of the ascending sub-curve and the descending sub-curve of each analytical curve peak is taken as the morphological index of each analytical curve peak.

6. A high-precision blood glucose monitoring method for a real-time blood glucose monitor according to claim 4, characterized in that The obtaining of the abnormal index of each analytical curve peak includes: The mean of the peak values of all dietary peaks during the monitoring period is obtained and denoted as the reference value; For each analytical curve peak during the monitoring period, when the peak value of the analytical curve peak is less than or equal to the reference value, the morphological difference value of the analytical curve peak is taken as the abnormal index of the analytical curve peak; When the peak value of the analytical curve peak is greater than the reference value, the ratio of the difference between the peak value of the analytical curve peak and the reference value to the peak value of the analytical curve peak is obtained, and the sum value of the ratio and the constant 1 is taken as the adjustment coefficient of the analytical curve peak; the morphological difference value of the analytical curve peak is weighted by using the adjustment coefficient to obtain the abnormal index of the analytical curve peak.

7. A high-precision blood glucose monitoring method for a real-time blood glucose monitor according to claim 3, characterized in that, The training of the neural network according to the analytical curve peaks and their abnormal indices of the other days except the current day during the monitoring period includes: The analytical curve peaks except the dietary peaks during the monitoring period are denoted as abnormal factor peaks; there is also a baseline curve in the blood glucose curve. According to the number and the abnormal indices of the baseline curve, abnormal factor peaks and dietary peaks in the blood glucose curve of each day during the monitoring period, the weight indices of the baseline curve, abnormal factor peaks and dietary peaks of each day are obtained; The blood glucose curve in the night time period of each day is obtained; the weight indices of the baseline curve, abnormal factor peaks and dietary peaks of each day are used as the initial weights of the Attention model. The blood glucose curves in the day time period of each day are respectively used as the input data of the training samples, and the blood glucose curves in the night time period are used as the output data of the training samples; the training samples of the other days except the current day during the monitoring period constitute the training set of the Attention model; the Attention model is trained by using the training set.

8. A high-precision blood glucose monitoring method for a real-time blood glucose monitor according to claim 7, characterized in that The obtaining of the weight indices of the baseline curve, abnormal factor peaks and dietary peaks of each day includes: Set the reference weight factor to w; set the weight index of the baseline curve in the blood glucose curve for each day during the monitoring period to w, the weight index of the abnormal factor peak to 2w, and the weight index of the i-th diet peak to w×(1 + p i ); where p i is the abnormal index of the i-th diet peak in the blood glucose curve for each day during the monitoring period; The method for obtaining the reference weight factor w is: Where n1 is the total number of baseline curves in the blood glucose curves for each day during the monitoring period; n2 is the total number of abnormal factor peaks in the blood glucose curves for each day during the monitoring period; n3 is the total number of diet peaks in the blood glucose curves for each day during the monitoring period.

9. A high-precision blood glucose monitoring method for a real-time blood glucose monitor according to claim 8, characterized in that, The prediction of the blood glucose situation in the night time period of the current day by using the trained neural network includes: The blood glucose curve of the patient in the day time period of the current day is input into the trained Attention model, and the Attention model outputs the blood glucose curve in the night time period of the current day; If there are data points on the blood glucose curve during the night period of the current day whose amplitudes are less than the first preset threshold or greater than the second preset threshold, then the patient will experience abnormal blood glucose during the night period of the current day.

10. A real-time blood glucose monitor, characterized in that, The monitor includes: a data acquisition module, a diet period analysis module, a diet peak screening module, an anomaly analysis module, and a blood glucose monitoring module; The data acquisition module is used to obtain the analysis curve peaks in the blood glucose curve during the daytime period of each day within the monitoring period of the patient; The diet period analysis module is used to obtain the normal diet period according to the overlapping degree of the time intervals corresponding to the analysis curve peaks on different days within the monitoring period; The diet peak screening module is used to select diet peaks from the analysis curve peaks within the monitoring period according to the overlapping situation between the time interval corresponding to each analysis curve peak and the normal diet period, and the morphological feature differences between each analysis curve peak and the other analysis curve peaks within the monitoring period; The anomaly analysis module is used to obtain the anomaly index of each analysis curve peak according to the peak value difference and morphological feature difference between each analysis curve peak and the diet peaks within the monitoring period; The blood glucose monitoring module uses the last day within the monitoring period as the current day, and trains a neural network based on the analysis curve peaks of each day except the current day within the monitoring period and their anomaly indices; uses the trained neural network to predict the blood glucose situation during the night period of the current day.

Images