Adaptive filtering real-time continuous analyte measurement method and device

Abstract

The invention is suitable for the technical field of analyte monitoring, and provides a self-adaptive filtering real-time continuous analyte measuring method and a self-adaptive filtering real-time continuous analyte measuring device. In the embodiment, a to-be-measured sample is sampled in real time to obtain a real-time sample analyte signal; and determining the analyte signals during ascending, descending and stability according to the ratio K'of the sample analyte signals in the preset time and preset parameters. The method comprises the following steps: acquiring an analyte signal, then performing Savitzky-Golay filtering on the analyte signal according to a filtering parameter group in a preset parameter set so as to find a determined target filtering parameter group conforming to a preset condition, and subsequently performing Savitzky-Golay filtering on the analyte signal through the target filtering parameter group so as to obtain a final filtered detection result. Therefore, details of the filtered data can be reserved. Therefore, the problems that in the prior art, when signals of analytes such as alcohol, lactic acid and blood sugar are filtered, conventional filtering is adopted for processing, so that overall data are too smooth, description of data details is lacked, and the data details are not consistent with true values are solved.

Description

Technical Field

[0001] The present application relates to methods and devices for continuous analyte monitoring, and particularly to an adaptive filtering real-time continuous analyte measurement method and device. Background Art

[0002] Continuous analyte monitoring (CAM), such as continuous glucose monitoring (CGM), has become a routine monitoring operation. CAM can provide real-time analyte analysis (e.g., analyte concentration) of an individual. In the case of CGM, real-time glucose concentration, real-time alcohol concentration, and real-time lactate concentration of an individual can be provided. By providing real-time glucose concentration, real-time alcohol concentration, and real-time lactate concentration, treatment / clinical measures can be applied to the monitored individual more timely, and blood glucose and alcohol consumption conditions can be better controlled.

[0003] The acquisition of continuous analyte monitoring data is achieved by reacting the implanted sensor with the free analyte in the interstitial fluid. A chemical reaction occurs between the analyte in the interstitial fluid and the enzyme layer on the sensor to generate an electrical signal, and then the generated electrical signal is converted to obtain the current analyte value. However, there are few new calculation methods designed for analyte signals currently, and the filtering process of analyte signals mostly uses conventional filtering, resulting in overly smooth overall data, lacking a description of data details and deviating from the true value. Summary of the Invention

[0004] In view of this, the present application provides an adaptive filtering real-time continuous analyte measurement method and device to smooth the analyte signal while retaining the detailed data of the analyte signal and making it more consistent with the true value.

[0005] The first aspect of the present application provides an adaptive filtering real-time continuous analyte measurement method, the method comprising:

[0006] Performing real-time sampling on the sample to be measured in a sliding window manner to obtain a real-time sample analyte signal;

[0007] Determining the ratio K' of the sample analyte signal within a preset time, and determining the analyte signal in different analyte states through a first preset parameter, a second preset parameter, and K', where the analyte states include rising, falling, and stable;

[0008] Performing adaptive Savitzky-Golay real-time filtering on the sample analyte signal according to the filter parameter group in the preset filter parameter set, determining the mean square error mse between the initial filtering result and the actual analyte value, and determining the target filter parameter group according to the mean square error mse, where the filter parameter group includes the filter window size w and the order order;

[0009] Perform adaptive Savitzky-Golay real-time filtering using the target filter parameter set to obtain a target filtering result, and determine the analyte value of the sample to be measured according to the target filtering result.

[0010] Optionally, the determining the analyte signal under different analyte states by the first preset parameter, the second preset parameter, and K' includes:

[0011] Determine the analyte signal based on the in vivo sensitivity and the sample analyte signal. When it is determined that the analyte state is rising, i.e., K'≥α, the determination formula for the analyte signal is When the analyte state is falling, i.e., K'≤β, the determination formula for the analyte signal is When the analyte state is stable, i.e., β < K' < α, the determination formula for the analyte signal is BG t = K t * I t + C 0 , where BG t is the analyte signal at time t, K t is the sensitivity at time t, I t is the sample analyte signal at time t, C, C0, and C1 are preset constants, α is the first preset parameter, and β is the second preset parameter.

[0012] Optionally, after performing real-time sampling on the sample to be measured by means of a sliding window, the method further includes:

[0013] Perform an abnormality determination on the collected sample analyte according to a preset threshold range. When it is determined that the

[0014] sample analyte signal exceeds the threshold range, replace the sample analyte signal with the sample analyte signal at the previous moment.

[0015] After performing real-time sampling on the sample to be measured by means of a sliding window, the method further includes:

[0016] Perform an abnormality determination on the collected sample analyte signal according to a preset threshold range. When the change amount of the sample analyte signals in adjacent sampling windows exceeds the threshold γ, then the sample analyte signal I t is calculated as I t = I t-1 ± γ.

[0017] Optionally, the determining the target filter parameter set according to the mean square error mse includes:

[0018] Determine the filter parameter set corresponding to the minimum mean square error mse as the target filter parameter set.

[0019] Optionally, before performing adaptive Savitzky-Golay real-time filtering, the method further includes:

[0020] Determine the length S of the real-time analyte signal to be filtered, compare the filtering window size w of the current real-time filtering with the length S. When the length S is greater than or equal to the filtering window size w, start the current filtering. When the length S is less than the filtering window size w, perform the current filtering according to the filtering parameter set corresponding to the previous minimum mean square error mse.

[0021] The second aspect of the present application provides an adaptive filtering real-time continuous analyte measurement device, the device includes:

[0022] A signal acquisition unit, configured to perform real-time sampling on the analyte of the sample to be measured in a sliding window manner to obtain a real-time sample analyte signal;

[0023] An analyte signal determination unit, configured to determine the ratio K' of the sample analyte signal within a preset time, and determine the analyte signal in different analyte states through a first preset parameter, a second preset parameter, and K', where the analyte states include rising, falling, and stable;

[0024] A target parameter determination unit, configured to perform adaptive Savitzky-Golay real-time filtering on the analyte signal according to the filtering parameter set in the preset filtering parameter set, determine the mean square error mse between the initial filtering result and the actual analyte value, and determine the target filtering parameter set according to the mean square error mse, where the filtering parameter set includes a filtering window size w and an order order;

[0025] A filtering result determination unit, configured to perform adaptive Savitzky-Golay real-time filtering through the target filtering parameter set to obtain a target filtering result, and determine the analyte value of the sample to be measured according to the target filtering result.

[0026] Optionally, the determining the analyte signal in different analyte states through the first preset parameter, the second preset parameter, and K' in the analyte signal determination unit includes:

[0027] Determine the analyte signal through the in-vivo sensitivity and the sample analyte signal. When it is determined that the analyte state is rising, i.e., K'≥α, the determination formula of the analyte signal is When the analyte state is falling, i.e., K'≤β, the determination formula of the analyte signal is When the analyte state is stable, i.e., β<K'<α, the determination formula of the analyte signal is BG t =Kt *I t +C 0 , where BG t is the analyte signal at time t, K t is the sensitivity at time t, I t is the sample analyte signal at time t, C, C0, and C1 are preset constants, α is the first preset parameter, and β is the second preset parameter.

[0028] Optionally, after the signal acquisition unit performs real-time sampling on the sample to be measured by means of a sliding window, the device further includes:

[0029] A first anomaly determination unit for determining whether the acquired sample analyte signal is anomalous according to a preset threshold range, and when it is determined that the sample analyte signal exceeds the threshold range, replacing the sample analyte signal with the sample analyte signal at the previous moment.

[0030] Optionally, after the signal acquisition unit performs real-time sampling on the sample to be measured by means of a sliding window, the device further includes:

[0031] A second anomaly determination unit for determining whether the acquired sample analyte signal is anomalous according to a preset threshold range, and when the change amount of the sample analyte signal between adjacent sampling windows exceeds the threshold γ, the sample analyte signal I t is calculated as I t = I t-1 ±γ.

[0032] Optionally, the target parameter determination unit determining the target filter parameter group according to the mean square error mse includes:

[0033] Determining the filter parameter group corresponding to the minimum mean square error mse as the target filter parameter group.

[0034] Optionally, before the target parameter determination unit and the filter result determination unit perform adaptive Savitzky-Golay real-time filtering, the device further includes:

[0035] A parameter comparison unit for determining the length S of the real-time analyte signal to be filtered, comparing the filter window size w of the current real-time filtering with the length S, and when the length S is greater than or equal to the filter window size w, starting the current filtering, and when the length S is less than the filter window size w, performing the current filtering according to the filter parameter group corresponding to the previous minimum mean square error mse.

[0036] In the embodiments provided by this application, the sample to be measured is first sampled in real time to obtain a real-time sample analyte signal. Then, the analyte signals at the rising, falling, and stable states are determined based on the ratio K' of the sample analyte signal within a preset time and preset parameters. First, the analyte signal is filtered by the Savitzky-Golay filter according to the filter parameter group in the preset parameter set, and after finding the target filter parameter group that meets the preset conditions, the analyte signal is then filtered by the target filter parameter group using the Savitzky-Golay filter to obtain the final filtered detection result. This enables the data after filtering in this application to retain details. Thus, it solves the problem in the prior art that the filtering of the analyte signal is processed using conventional filtering, resulting in overly smooth overall data, lack of description of details, and deviation from the true value. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 It is a flowchart of the method provided by the embodiment of this application;

[0038] Figure 2 It is a comparison result diagram provided by the embodiment of this application;

[0039] Figure 3 It is a usage scenario diagram provided by the embodiment of this application;

[0040] Figure 4 It is a device structure diagram provided by the embodiment of this application;

[0041] Figure 5 It is a schematic internal structure diagram of a computer device provided by the embodiment of this application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0042] Here, the exemplary embodiments will be described in detail, and the examples are shown in the drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. On the contrary, they are merely examples of devices and methods consistent with some aspects of this application as detailed in the appended claims.

[0043] The terms used in this application are only for the purpose of describing specific embodiments and are not intended to limit this application. The singular forms "a", "the", and "said" used in this application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more of the associated listed items.

[0044] The present application provides an adaptive filtering real-time continuous analyte measurement method and device to solve the problems that existing filtering of analyte signals mostly uses conventional filtering, resulting in overly smooth overall data, lack of description of data details, deviation from the true value, etc.

[0045] The technical solutions of the present application will be described in detail below with specific embodiments. These specific embodiments below can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments.

[0046] As Figure 1 shown, it is a flowchart of an adaptive filtering real-time continuous analyte measurement method provided by the present application. In the following embodiments, the analyte can be alcohol, lactic acid, blood glucose, etc. Taking blood glucose as an example in the present application, the steps of this process include:

[0047] S101, perform real-time sampling on the sample to be measured by means of a sliding window to obtain a real-time sample analyte signal.

[0048] In this embodiment, during the real-time dynamic monitoring of the sample to be measured, sampling is performed by means of a sliding window. The specific method is as follows: sample n times per minute, collect for N minutes as a window, and obtain an initial real-time interstitial fluid glucose signal dataset with a quantity of n*N. Then, screen this dataset, remove one highest value and one lowest value, average the remaining results, and finally output. Then, calculate the next window at intervals of every N minutes. The finally obtained real-time sampling window interstitial fluid glucose signal (i.e., the real-time sample analyte signal) is i tm is the interstitial fluid glucose signal at time t.

[0049] In another embodiment, after performing real-time sampling on the sample to be measured by the above-mentioned sliding window method, the method further includes:

[0050] Perform abnormal determination on the collected sample analyte signal according to a preset threshold range. When it is determined that the sample analyte signal exceeds this threshold range, replace the sample analyte signal with the sample analyte signal at the previous moment, and the previous moment refers to the previous N minutes.

[0051] In this embodiment, according to the linear range of the sensor, such as the linear range is x’~y’ mmol / L, the corresponding interstitial fluid glucose signal range can be obtained. And a threshold range is preset according to this signal range. For example, the current sensitivity S can be calculated through the blood glucose concentration G measured by a fingertip blood paired blood glucose meter and the preset background current I 0 calculate the current sensitivity S, background current I 0It can be preset to 0 or determined by fitting the glucose solution and the corresponding generated current value. The specific formula is I t = S*G + I 0 , I t is the interstitial fluid glucose signal at time t. Then set the above threshold range to [S*x' + I 0 , S*y' + I 0 . The interstitial fluid glucose signal exceeding this threshold range can be considered an outlier, and this value is filled with the interstitial fluid glucose signal at the previous moment, such as the previous N minutes.

[0052] In another embodiment, during the process of collecting the interstitial fluid glucose signal, based on the prior knowledge that the instantaneous change in human blood glucose within 3 minutes <= 1 mmol / L, the instantaneous change can be adjusted according to conditions such as health status and age, and the current change threshold γ is calculated according to the sampling window. For example, based on the prior knowledge that the instantaneous change in human blood glucose within 3 minutes <= 1 mmol / L, the change per minute of the threshold γ is determined to be <= 0.33 mmol / L. Then when the sampling window is 5 minutes, the corresponding γ is determined to be <= 0.33 * 5 = 1.65 mmol / L. If the change in the interstitial fluid glucose signal between adjacent sampling windows exceeds the threshold γ, then the interstitial fluid glucose signal I t is calculated with the interstitial fluid glucose signal I t-1 ±γ. For example, when the interstitial fluid glucose signal rises by more than γ, then I t = I t-1 +γ is used instead. If it drops by more than γ, then I t = I t-1 -γ is used instead.

[0053] S102. Determine the ratio K' of the above analyte signal within a preset time, and determine the analyte signal in different analyte states through the first preset parameter α, the second preset parameter β, and K'.

[0054] In this embodiment, the analyte states include rising blood glucose, falling blood glucose, and stable blood glucose. Two parameters α and β can be set as thresholds. According to the ratio K' = i t / i t-1To determine the change in the interstitial fluid glucose signal, when it is greater than or equal to α, it is determined that the blood glucose is rising rapidly. When it is less than or equal to β, it is determined to be rapidly decreasing. Within the range between the two, it is determined to be in a stable state or a state of small fluctuations. After dividing the blood glucose situation into three parts by the two parameters α and β, the blood glucose signal under different blood glucose states, that is, the analyte signal, is obtained. In this embodiment, the threshold parameter during the rising period is preferably 1.02 - 1.04, and the threshold parameter during the falling period is 0.96 - 0.98. The specific situation is as follows:

[0055] 1. When the blood glucose state is rising, that is, K’≥α, the determination formula for the blood glucose signal is

[0056] 2. When the blood glucose state is falling, that is, K’≤β, the determination formula for the blood glucose signal is

[0057] 3. When the blood glucose state is stable, that is, β < K’ < α, the determination formula for the blood glucose signal is BG t =K t *I t +C 0 。

[0058] In the above formula, BG t represents the blood glucose signal at time t, C 0 represents the initial constant, which can be set to 0 or adjusted according to historical data. K t represents the sensitivity at time t, which can be calculated by the formula K t =θ*K 0 *G(t)*K'*K T +C2. In this formula, K 0 represents the sensitivity obtained from the in vitro test of the sensor, which is obtained before the sensor is implanted into the body. θ is the change coefficient from the in vitro sensitivity to the in vivo sensitivity. G(t) represents the attenuation curve of the sensor. K T is the temperature compensation coefficient, and C2 is a constant, which can be 0. The determination methods of each parameter in this formula are described below:

[0059] 1. After the sensor is finished, the linear relationship between in vitro implantation and in vivo implantation can be obtained, and the change coefficient θ can be obtained by linear regression.

[0060] 2. Since the sensor decays to a certain extent over time after being implanted in the body, the decay curve can be determined based on the characteristics of the sensor and its relationship with time. For the aforementioned sensor decay curve, it can be a logarithmic curve, a quadratic curve, etc. The decay curve method adopted in this article has a specific formula: G(t) = Aln(t) + C3. Here, A is a coefficient obtained through linear regression of the sensor decay with respect to time and sensitivity, and C3 is a constant that can be set to 0.

[0061] 3. For the temperature compensation coefficient K T , due to the influence of temperature on enzyme catalytic activity, physiological homeostasis environment, etc., which will affect the magnitude and instantaneous sensitivity of the current value during the process of the in-vivo sensor monitoring blood glucose, it is necessary to compensate for the sensitivity. It can be determined by the formula K T = K 0 (1 + λ * ΔT). In this formula, ΔT is the difference between the monitored temperature and the standard temperature, such as the difference between the monitored temperature and room temperature of 25°C. The compensation coefficient λ is calculated based on the sensor sensitivities at different temperatures under the same concentration, obtaining the difference in sensitivity between different temperatures, and further obtaining the change in sensitivity per degree Celsius, that is, the compensation coefficient λ.

[0062] Through the above method, the blood glucose signals in different blood glucose states can be determined.

[0063] S103. Perform adaptive Savitzky-Golay real-time filtering on the above analyte signal according to the filter parameter group in the preset filter parameter set. Determine the mean square error mse between the initial filtering result and the actual analyte value, and determine the target filter parameter group according to the mean square error mse.

[0064] Filtering the analyte signal can remove the noise in the signal, and smoothing can reduce the influence of outliers and other irrelevant factors on the signal.

[0065] This filtering method performs local polynomial fitting of a sliding window on the blood glucose signal, and uses the fitted polynomial to replace the blood glucose signal for smoothing. It can adapt to different signals and data sets by adjusting the window size and polynomial order. Therefore, this application selects the Savitzky-Golay filter to filter the blood glucose signal. Since the selection of the window size and order is crucial for this filtering, the grid search method is used to search for the parameters to perform real-time filtering on the blood glucose signal in an adaptive manner to achieve a good filtering effect, which not only smooths the blood glucose signal but also can well protect the detailed information of the blood glucose signal.

[0066] In this embodiment, the filtering parameter set includes the filtering window size w and the order order. A set of filtering parameters can be preset, and the specific steps of Savitzky-Golay filtering are as follows:

[0067] 1. Select the filtering window: Select a filtering window of an appropriate size by means of a sliding window. This filtering window will be used for subsequent data fitting. For example, the filtering window can be set to w = [t - (W - 1),..., t - 3, t - 2, t - 1, t]. w is the parameter for setting the filtering window size, and the size of w is determined according to the needs of the data set or manually, so that w is an odd number. w is used as the filtering window size for sliding window. The larger the filtering window, the smoother the filtering, but some details will be lost.

[0068] 2. Pair the filtering window with the signal point: Pair the filtering window with a certain target moment in the blood glucose signal. Then, at the target moment and its past moments, take the filtering window size as w, then x = (-w + 1, -w + 2,..., 0), and 0 represents the current value at the target moment.

[0069] 3. Perform polynomial fitting: On this data vector, use a polynomial function (usually a low-order polynomial) to fit the data. This polynomial function is obtained by the least squares method. The polynomial function is as follows: (taking a 2nd-order polynomial as an example)

[0070] y = a 0 + a 1 x + a 2 x 2

[0071] 4. Calculate the fitted value: According to the fitted polynomial function, calculate the fitted value at the target moment of the filtering window. This fitted value will be used as the filtered result. For example, through the formula y = a 0 + a 1 *x + a 2 *x 2 +... + a k-1 *x k-1 Calculate. x is the data to be fitted, a is the coefficient, k is the order, that is, the order order in the filtering parameter set, until all data points within this filtering window have been fitted, and y is saved and output in real time.

[0072] 5. Move the filtering window and repeat: Move the window one position to the right (i.e., one filtering window), and then repeat steps 2 to 4.

[0073] 6. Save and output: Finally, save the fitted values at all moments and output them as the filtered signal. Represented in matrix form, when w is 5, it is specifically as follows, This matrix can be abbreviated as Y 5*1 = X 5*k * A k*1 + E 5*1 , which are respectively denoted as matrices Y, X, A, and E. Among them, the Y matrix is The X matrix is The A matrix is The E matrix is The least squares solution of the A matrix is In this formula, X T is the transpose matrix of the X matrix. The filtered value of the Y matrix is Among them, E = X * (X T * X) -1 * X T .

[0074] During the Savitzky-Golay filtering process, each parameter combination in the filtering parameter set can be traversed by an exhaustive method for filtering. And the filtering results are judged by the mean square error mse (mean-square error) to determine the target filtering parameter group. The specific calculation formula is as follows: In this formula, Y i represents the actual analyte value, represents the initial filtering result. Different conditions for the mean square error can be preset according to different usage scenarios to screen the target filtering parameter group. Such as the maximum, minimum, or closest to the average of the mean square error, etc. After determining the target filtering parameter group, combined with the blood glucose signals under different blood glucose states, the filtering results under each blood glucose signal can be obtained. For example, when the preset filtering parameter set is {w: [1, 3, 5, 7, 9, 11, 13], order: [1, 2, 3, 4, 5, 6, 7]}. First, perform Savitzky-Golay real-time filtering on the blood glucose signals during the rising, falling, and stable states respectively. The target filtering parameter groups for the rising and falling states are determined to be {w: [5, 7, 9, 11], order: [2, 3, 4]}, and the target filtering parameter group for the stable state is {w: [5, 7, 9], order: [1, 2]}.

[0075] In another embodiment, the parameter combination with the smallest mse can be selected as the target filtering parameter group.

[0076] In another embodiment, before performing the adaptive Savitzky-Golay real-time filtering, the above method further includes:

[0077] Determine the length S of the real-time analyte signal to be filtered, compare the filter window size w of this real-time filtering with the length S. When the length S is greater than or equal to the filter window size w, start this filtering. When the length S is less than the filter window size w, stop the continuous search, and perform this filtering according to the filter parameter group corresponding to the previous minimum mean square error mse.

[0078] In this embodiment, the length S of the analyte signal refers to the data at the existing time. When the length S is greater than or equal to the filter window size w, start this filtering. When the length S is less than w, stop the continuous search, and perform this filtering with the filter parameter group obtained from the previous minimum mse. This embodiment can select to perform filtering only when the length S of the blood glucose signal is greater than or equal to the filter window size w, so as to avoid filtering failure.

[0079] S104, perform adaptive Savitzky-Golay real-time filtering through the above target filter parameter group to obtain a target filtering result, and determine the analyte value of the above-mentioned sample to be tested according to this target filtering result.

[0080] In this embodiment, taking the target filter parameter group determined in step S103 as an example, through the formula the real-time filtering of the blood glucose signal can be realized. In the above formula, when the blood glucose state is in the rising or falling stage, the filtered result contains more detailed parts. The comparison results of the blood glucose values before and after filtering with the fingertip blood are as Figure 2 shown. It can be seen that the blood glucose value filtered by this method fluctuates significantly at the place where the blood glucose fluctuates, which is consistent with the true value, and solves the problem that the existing conventional filtering method is still too smooth at the place where the blood glucose fluctuates and lacks the detailed processing of the blood glucose fluctuation stage.

[0081] Table 1 shows the comparison results of the filtering results determined by different filtering methods for 14 days with the true values of fingertip blood.

[0082]

[0083] Table 1

[0084] Table 2 shows the comparison results of the real-time adaptive filtering results, unfiltered results and true values of fingertip blood for a certain time period in Table 1.

[0085]

[0086] Table 2

[0087] As can be seen from Table 1 above, the mard value of S-G real-time adaptive filtering and fingertip blood is the lowest, and the performance result is the best. The difference from the mard value of unfiltered is 0.43%, and both are better than wavelet real-time and S-G real-time fixed parameter filtering; as can be seen from Table 2 above, the MSE b average value of S-G real-time adaptive filtering is lower than the MSEa average value of unfiltered, and the MSE corresponding to the rising and falling moments is lower; by performing S-G real-time adaptive filtering through this method, not only is it more consistent with the true value, but also the results of the blood glucose state in the rising or falling stage are processed more accurately and in detail.

[0088] So far, the Figure 1 shown process is completed.

[0089] In the embodiment of the present application, first, a real-time sample is taken for the sample to be measured to obtain a real-time sample analyte signal. Then, the analyte signals at the rising, falling, and stable times are determined through the ratio K' of the sample analyte signal within a preset time and preset parameters. Then, first, the analyte signal is subjected to Savitzky-Golay filtering according to the filtering parameter group in the preset parameter set, so as to find a determined target filtering parameter group that meets the preset conditions. Subsequently, the blood glucose signal is subjected to Savitzky-Golay filtering through the target filtering parameter group to obtain a final filtered detection result. This enables the filtered data of the present application to retain the details of the analyte. Thus, it solves the problem in the prior art that the filtering of the analyte signal is processed by conventional filtering, resulting in the overall data being too smooth, lacking a description of the data details, and not conforming to the true value.

[0090] In this solution, after determining the sample analyte signal, the sample analyte signal, such as the glucose signal of interstitial fluid, can also be wirelessly transmitted to the user's mobile phone through an existing communication method such as uart via a low-power Bluetooth transceiver module, converted into a blood glucose value by a blood glucose calculation formula, and displayed in real time in the APP and historical images are displayed. Then, it is synchronously uploaded to the cloud through the wifi in the mobile phone. The cloud can include parameters such as insulin information, interstitial fluid glucose signal, blood glucose data, or the patient's daily diet, etc., to realize the preservation and output of the interstitial fluid glucose signal. For doctors and users, they can use the client or mobile terminal to obtain the required information from the cloud. Doctors can use some historical data based on the cloud to give appropriate guidance to the user. Users can also use mobile devices to obtain cloud data and obtain corresponding consultations from doctors according to past historical blood glucose values, realizing online consultations, which is convenient for consultations and guidance between users and doctors. The specific usage scenario is as Figure 3 shown.

[0091] The present application also provides an adaptive filtering real-time continuous analyte measurement device, as Figure 4 shown. This device includes:

[0092] The signal acquisition unit 401 is configured to perform real-time sampling on the sample to be measured by means of a sliding window, and obtain a real-time sample analyte signal;

[0093] The analyte signal determination unit 402 is configured to determine the ratio K' of the sample analyte signal within a preset time, and determine the blood glucose signal under different analyte states through a first preset parameter α, a second preset parameter β, and K'. The blood glucose states include rising, falling, and stable;

[0094] The target parameter determination unit 403 is configured to perform adaptive Savitzky-Golay real-time filtering on the analyte signal according to the filter parameter group in the preset filter parameter set, determine the mean square error mse between the initial filtering result and the actual blood glucose value, and determine the target filter parameter group according to the mean square error mse. Among them, the filter parameter group includes the filter window size w and the order order;

[0095] The filtering result determination unit 404 is configured to perform adaptive Savitzky-Golay real-time filtering through the target filter parameter group to obtain a target filtering result, and determine the analyte value of the sample to be measured according to the target filtering result.

[0096] In another embodiment, the determination of the analyte signal under different analyte states through the first preset parameter α, the second preset parameter β, and K' in the analyte signal determination unit includes:

[0097] Determine the analyte signal by the in vivo sensitivity and the sample analyte signal. When it is determined that the analyte state is rising, i.e., K'≥α, the determination formula of the analyte signal is When the analyte state is falling, i.e., K'≤β, the determination formula of the analyte signal is When the analyte state is stable, i.e., β<K'<α, the determination formula of the analyte signal is BG t =K t *I t +C 0 , where BG t is the analyte signal at time t, K t is the sensitivity at time t, I t is the sample analyte signal at time t, C, C0, and C1 are preset constants, α is the first preset parameter, and β is the second preset parameter.

[0098] In another embodiment, after the signal acquisition unit performs real-time sampling on the sample to be measured by means of a sliding window, this device further includes:

[0099] The first anomaly determination unit 405 is configured to perform anomaly determination on the collected sample analyte signal according to a preset threshold range. When it is determined that the sample analyte signal exceeds the threshold range, the sample analyte signal is replaced with the sample analyte signal at the previous moment.

[0100] In another embodiment, this device further includes:

[0101] The second anomaly determination unit 406 is configured to, when determining that the sample analyte signal exceeds the threshold γ, calculate the sample analyte signal as the sample analyte signal ±γ.

[0102] In another embodiment, the determining the target filter parameter group according to the mean square error mse in the target parameter determination unit 403 includes:

[0103] Determining the filter parameter group corresponding to the minimum mean square error mse as the target filter parameter group.

[0104] In another embodiment, before the target parameter determination unit 403 and the filter result determination unit 404 perform adaptive Savitzky-Golay real-time filtering, this device further includes:

[0105] The parameter comparison unit 407 is configured to determine the length S of the real-time analyte signal to be filtered, compare the filter window size w of this real-time filtering with the length S. When the length S is greater than or equal to the filter window size w, this filtering is started. When the length S is less than the filter window size w, this filtering is performed according to the filter parameter group corresponding to the previous minimum mean square error mse.

[0106] In the above embodiments of the present invention, an adaptive filtering real-time continuous analyte measurement method is provided, and based on this method, an adaptive filtering real-time continuous analyte measurement device is provided. Through the above method and device, the problem that in the prior art, the filtering of analyte signals such as alcohol, lactic acid, and blood glucose is processed by a single filtering, resulting in the overall data being too smooth, lacking the description of details and not conforming to the true value can be solved.

[0107] This embodiment also discloses a computer device, as Figure 5 shown, the computer device includes a processor and a memory. At least one instruction is stored in the memory. Of course, y can also be saved and output to the display. At least one instruction is loaded and executed by the processor to implement any of the above adaptive filtering real-time continuous analyte measurement methods.

[0108] The above are only the preferred embodiments of the present application and are not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application shall be included within the scope of protection of the present application.

Claims

1. A method for real-time continuous analyte measurement using adaptive filtering, characterized in that: The method comprises: The sample to be tested is sampled in real time by means of a sliding window to obtain a real-time sample analyte signal; Determine a ratio K' of the sample analyte signal within a preset time, and determine the analyte signal under different analyte states by using the first preset parameter, the second preset parameter and K', wherein the analyte state includes rising, falling and stable; Performing adaptive Savitzky-Golay real-time filtering on the analyte signal according to a filter parameter group in a preset filter parameter set, determining a mean square error mse between an initial filter result and an actual analyte value, and determining a target filter parameter group according to the mean square error mse, wherein the filter parameter group includes a filter window size w and an order order; Adaptive Savitzky-Golay real-time filtering is performed using the target filtering parameter group to obtain a target filtering result, and the analyte value of the sample to be tested is determined according to the target filtering result.

2. The method according to claim 1, characterized in that The determining of the analyte signal under different analyte states by using the first preset parameter, the second preset parameter and K' comprises: Determine the analyte signal based on the in-vivo sensitivity and the sample analyte signal. When it is determined that the analyte state is rising, i.e., K’≥α, the determination formula for the analyte signal is When the analyte state is falling, i.e., K’≤β, the determination formula for the analyte signal is When the analyte state is stable, i.e., β<K’<α, the determination formula for the analyte signal is BG t =K t *I t +C0, where BG t is the analyte signal at time t, K t is the sensitivity at time t, I t is the sample analyte signal at time t, C, C0, and C1 are preset constants, α is the first preset parameter, and β is the second preset parameter.

3. The method according to claim 1, characterized in that After the real-time sampling of the sample to be tested is performed by means of a sliding window, the method further comprises: The collected sample analyte signal is judged to be abnormal according to a preset threshold range, and when it is determined that the sample analyte signal exceeds the threshold range, the sample analyte signal is replaced with the sample analyte signal at the previous moment.

4. The method according to claim 1, characterized in that: After the real-time sampling of the sample to be tested is performed by means of a sliding window, the method further comprises: The sample analyte signal collected is judged to be abnormal according to the preset threshold range. When the change of the sample analyte signal in the adjacent sampling window exceeds the threshold γ, the sample analyte signal I t Take I t =I t-1 ±γ is calculated.

5. The method according to claim 1, characterized in that Determining a target filtering parameter group according to the mean square error MSE comprises: The filtering parameter group corresponding to the minimum mean square error mse is determined as the target filtering parameter group.

6. The method according to claim 1, characterized in that Before performing the adaptive Savitzky-Golay real-time filtering, the method further includes: Determine the length S of the real-time analyte signal to be filtered, and compare the filter window size w of this real-time filtering with the length S. When the length S is greater than or equal to the filter window size w, start this filtering. When the length S is less than the filter window size w, perform this filtering according to the filter parameter group corresponding to the previous minimum mean square error mse.

7. An adaptive filtering real-time continuous analyte measurement device, characterized in that: The device comprises: A signal acquisition unit, used to perform real-time sampling of the analyte of the sample to be tested by means of a sliding window to obtain a real-time sample analyte signal; an analyte signal determination unit, configured to determine a ratio K' of the sample analyte signal within a preset time, and determine the analyte signal under different analyte states by using the first preset parameter, the second preset parameter and K', wherein the analyte state includes rising, falling and stable; a target parameter determination unit, configured to perform adaptive Savitzky-Golay real-time filtering on the analyte signal according to a filter parameter group in a preset filter parameter set, determine a mean square error mse between an initial filter result and an actual analyte value, and determine a target filter parameter group according to the mean square error mse, wherein the filter parameter group includes a filter window size w and an order order; The filtering result determination unit is used to perform adaptive Savitzky-Golay real-time filtering through the target filtering parameter group to obtain a target filtering result, and determine the analyte value according to the target filtering result.

8. The device according to claim 7, characterized in that The analyte signal determination unit determines the analyte signal under different analyte states by using the first preset parameter, the second preset parameter and K', including: Determine the analyte signal based on the in-vivo sensitivity and the sample analyte signal. When it is determined that the analyte state is rising, i.e., K’≥α, the judgment formula for the analyte signal is When the analyte state is falling, i.e., K’≤β, the judgment formula for the analyte signal is When the analyte state is stable, i.e., β<K’<α, the judgment formula for the analyte signal is BG t =K t *I t +C0, where BG t is the analyte signal at time t, K t is the sensitivity at time t, I t is the sample analyte signal at time t, C, C0, and C1 are preset constants, α is the first preset parameter, and β is the second preset parameter.

9. The device according to claim 7, characterized in that After the signal acquisition unit performs real-time sampling of the sample to be tested by means of a sliding window, the device further comprises: The first abnormality determination unit is used to perform abnormality determination on the collected sample analyte signal according to a preset threshold range, and when it is determined that the sample analyte signal exceeds the threshold range, replace the sample analyte signal with the sample analyte signal at the previous moment.

10. The device according to claim 7, characterized in that After the signal acquisition unit performs real-time sampling of the sample to be measured by means of a sliding window, the device further comprises: The second abnormality determination unit is used to determine the abnormality of the collected sample analyte signal according to a preset threshold range. When the change amount of the sample analyte signal of the adjacent sampling window exceeds the threshold value γ, the sample analyte signal I t Take I t =I t-1 ±γ is calculated.

11. The device according to claim 7, characterized in that Determining a target filtering parameter group according to the mean square error mse in the target parameter determination unit includes: The filtering parameter group corresponding to the minimum mean square error mse is determined as the target filtering parameter group.

12. The device according to claim 7, characterized in that Before the target parameter determination unit and the filtering result determination unit perform adaptive Savitzky-Golay real-time filtering, the device further includes: The parameter comparison unit is used to determine the length S of the real-time analyte signal to be filtered, and compare the filter window size w of this real-time filtering with the length S. When the length S is greater than or equal to the filter window size w, the current filtering is started. When the length S is less than the filter window size w, the current filtering is performed according to the filter parameter group corresponding to the previous minimum mean square error mse.

Images