Method for determining current glucose concentration in a carrier fluid - Patents.com

Abstract

The present invention relates to a method for continuously measuring a current glucose concentration in a carrier-fluid, particularly in blood, of a living organism, comprising: a) detecting a series of measurements of tissue glucose concentrations in tissue surrounding the carrier-fluid using a sensor device, the series of measurements including at least two measurements separated by a time interval; b) calculating the tissue glucose concentration using the detected series of measurements based on a sensor model, where the sensor model correlates the measurements of the sensor device with the tissue glucose concentration taking into account measurement noise; c) providing a state transition model, where the state transition model correlates at least one glucose concentration in the carrier-fluid with the calculated tissue glucose concentration taking into account process noise; and d) calculating the current glucose concentration based on the provided state transition model and the calculated tissue glucose concentration, wherein at least step d), particularly steps b) to d), are performed using at least one moving horizon estimation method.

Description

[Technical field]

[0001] The present invention relates to a method for measuring, preferably continuously, the current glucose concentration in a carrier fluid of a biological organism, in particular in blood.

[0002] The invention also relates to a device for measuring, preferably continuously, the current glucose concentration in a carrier fluid of a biological organism, in particular in blood.

[0003] The invention also relates to an estimation device for measuring, preferably continuously, the current glucose concentration in a transport fluid of a living organism, in particular in the blood.

[0004] The present invention also relates to a non-transitory machine-readable medium storing instructions that, when executed on a computer, cause the implementation of a method for measuring, preferably continuously, the current glucose concentration in a carrier fluid of a biological body, in particular in blood.

[0005] Although the invention is generally applicable to any method of measuring the current glucose concentration in a carrier fluid, the invention will be described with reference to blood glucose concentration in an organism. [Background technology]

[0006] 2. Description of the Related Art Continuous glucose monitoring systems, also called CGM (continuous glucose monitoring), are known for measuring blood glucose concentration BG in a living organism, particularly a human body. Generally, in a CGM system, interstitial tissue glucose concentration (IG) is measured automatically, for example, every 1 to 5 minutes. Diabetics in particular benefit from CGM systems because measurements can be taken significantly more frequently than with self-monitoring, where patients manually measure their blood glucose levels four to ten times a day. This allows for automatic estimation and warning signals to patients even while they are asleep, helping to prevent critical health conditions.

[0007] On the one hand, known CGM systems are based on electrochemical methods. Such a CGM system is described, for example, in Patent Document 1. Also, an optical CGM system in which glucose concentration-dependent fluorescence is used is known, for example from US Pat. No. 6,399,323, which is incorporated herein by reference. Both types of CGM systems measure glucose concentrations in interstitial tissue.

[0008] Furthermore, the glucose concentration in tissue or the glucose concentration in interstitial fluid (IG) deviates from the glucose concentration in blood (hereinafter abbreviated as BG). As described in Non-Patent Document 1, large influences on blood glucose concentration can result in large deviations, for example, from the ingestion of food or nutrients or the injection of insulin. As described in Non-Patent Document 2, this separation occurs due to diffusion processes in the blood and surrounding tissues, so that IG concentrations follow BG concentrations in a time-delayed and attenuated manner. [Prior art documents] [Patent documents]

[0009] [Patent Document 1] International Publication No. 2006 / 017358 A1 [Patent Document 2] German Patent Application Publication No. 10 2015 101 847 B4 [Non-patent literature]

[0010] [Non-Patent Document 1] Basu, Ananda et al. “Time lag of Glucose from intravascular to interstitial compartment in humans.” (Diabetes(2013): DB-131132) [Non-Patent Document 2] Rebrin, Kerstin et al. “Subcutaneous Glucose predicts plasma Glucose independent of insulin: implications for continuous monitoring” (American Journal of Physiology-Endocrinology and Metabolism 277.3 (1999): E561-E571) [Non-Patent Document 3] Keenan, D. Barry et al. “Delays in minimally invasive continuous Glucose monitoring devices: a review of current technology.” (Journal of diabetes science and technology 3.5 (2009): 1207-1214) [Non-Patent Document 4] Knobbe, Edward J. and Bruce Buckingham “The extended Kalman filter for continuous Glukose monitoring.” (Diabetes technology & therapeutics 7.1 (2005):15-27) [Non-Patent Document 5] Grace Wahba “Bayesian 'Confidence Intervals' for the Cross-Validated Smoothing Spline” (Journal of the Royal Statistical Society: Series B (Methodological) 45, (1983), 133-150) [Non-Patent Document 6] Rios und Sahinidis “Derivative-free optimization: a review of algorithms and comparison of software implementations” (Journal of Global Optimization 56,3 (2013), 1247-1293) [Non-Patent Document 7] Kuester, Nikolaus et al. “First Clinical Evaluation of a New Percutaneous Optical Fiber Glukose Sensor for Continuous Glukose Monitoring in Diabetes” (Journal of Diabetes Science and Technology 7, 1 (2014), 13-23) Summary of the Invention [Problem to be solved by the invention]

[0011] Due to the above-mentioned attenuation and time delay between the two glucose concentrations, BG in blood on the one hand and IG in the surrounding tissues on the other hand, calibration of CGM systems performed by manual measurement of blood glucose concentration, e.g., by taking a drop of blood from a finger and measuring the glucose concentration in this drop of blood using an external measurement device, results in significant inaccuracies.

[0012] However, to achieve accurate calibration of a CGM system, the above-mentioned discrepancy between tissue and blood glucose concentrations must be taken into account, or at least estimated. Various methods are known for doing this. From Non-Patent Document 3 it is known to use a time-delayed glucose signal for calibration. Also, from Non-Patent Document 4, it is known to use a Kalman filter to compensate for the attenuation and time delay of the glucose diffusion process between blood and tissue. [Means for solving the problem]

[0013] It is therefore an object of the present invention to provide a more accurate device for measuring blood glucose concentration and an apparatus for estimating same, which is highly flexible when considering additional parameters and simpler implementation. It is a further object of the present invention to provide an alternative method, an alternative apparatus and an alternative estimation apparatus. It is a further object of the present invention to provide an improved method, apparatus and estimator for the measurement of blood glucose concentration in an organism based on the measurement of glucose concentration in interstitial tissue.

[0014] A method for measuring, preferably continuously, the current glucose concentration in a transport fluid, in particular in blood, of a biological organism, comprising: a) detecting a series of measurements of tissue glucose concentration in tissue surrounding a carrier fluid using a sensor device, the series of measurements including at least two measurements separated in time; b) calculating a tissue glucose concentration using the series of detected measurements based on a sensor model, where the sensor model correlates the measurements of the sensor device with the tissue glucose concentration while taking into account measurement noise; c) providing a state transition model, whereby at least one glucose concentration in the transport fluid is correlated with the calculated tissue glucose concentration while taking into account process noise; d) calculating a current glucose concentration based on the provided state transition model and the calculated tissue glucose concentration; The invention solves the above problem by a method, wherein at least step d), in particular steps b) to d), are performed using at least one moving horizon estimation method.

[0015] Furthermore, a device for measuring, preferably continuously, the current glucose concentration in a transport fluid, in particular in blood, of a biological body, which performs the method according to any one of claims 1 to 12, a sensor device for measuring fluorescence in tissue surrounding a carrier fluid by means of an optical fiber probe, the sensor device being configured to detect a series of measurements of tissue glucose concentration in the tissue surrounding the carrier fluid, the series including at least two measurements spaced apart in time; a providing device configured to provide a state transition model, whereby at least one glucose concentration in a transport fluid is correlated with a calculated tissue glucose concentration while taking into account process noise, and a providing device configured to provide a sensor model, whereby a measurement value of the sensor device is correlated with a tissue glucose concentration while taking into account measurement noise; an estimator configured to use the series of detected measurements to calculate a tissue glucose concentration based on the sensor model, and to calculate a current glucose concentration using a moving horizon estimation method based on the provided state transition model and the calculated tissue glucose concentration; The present invention solves the above problems by providing an apparatus comprising:

[0016] In a further embodiment, an estimation device for measuring, preferably continuously, the current glucose concentration in a transport fluid, in particular in blood, of a biological body, the estimation device performing the method according to any one of claims 1 to 12, at least one interface for connecting the sensor device to detect a series of measurements of tissue glucose concentration in tissue surrounding a carrier fluid, the series of measurements including at least two measurements spaced apart in time; at least one memory storing a state transition model, whereby at least one glucose concentration in the carrier fluid is correlated with a tissue glucose concentration, taking into account at least one process noise value, and at least one memory storing a sensor model, whereby a measurement value of the sensor device is correlated with a tissue glucose concentration, taking into account at least one measurement noise value; a computing device configured to use the set of detected measurements to calculate a tissue glucose concentration based on the sensor model, and to calculate a current glucose concentration using at least one moving horizon estimation method based on the stored state transition model and the calculated tissue glucose concentration; The present invention solves the above-mentioned problems by providing an estimation device comprising:

[0017] In a further embodiment, a non-transitory machine-readable medium storing instructions which, when executed on a computer, cause the execution of a method for measuring, preferably continuously, a current glucose concentration in a carrier fluid, in particular in blood, of a biological body, the non-transitory machine-readable medium executing the method according to any one of claims 1 to 12, a) detecting a series of measurements of tissue glucose concentration in tissue surrounding a carrier fluid using a sensor device, the series of measurements including at least two measurements spaced apart in time; b) calculating a tissue glucose concentration using the series of detected measurements based on a sensor model, where the sensor model correlates the measurements of the sensor device with the tissue glucose concentration while taking into account measurement noise; c) providing a state transition model, whereby at least one glucose concentration in the transport fluid is correlated with the calculated tissue glucose concentration while taking into account process noise; d) calculating a current glucose concentration based on the provided state transition model and the calculated tissue glucose concentration; The present invention solves the above problem by a non-transitory machine-readable medium, in which at least step d), in particular steps b) to d), are performed using at least one moving horizon estimation method.

[0018] Thus, a method for measuring blood glucose concentration in an organism is provided. This includes the following methods: In a first method step, a series of measurements is made by the sensor device, comprising at least two temporally separated sensor measurements of tissue glucose concentration in tissue of the organism. In a further step, a sensor model of the relationship between the sensor measurements and the tissue glucose concentration is provided, and a state transition model is provided which includes a model of the relationship between the tissue glucose concentration and the blood glucose concentration, and in a further method step, the blood glucose concentration of the organism is quantified by the sensor model and the state transition model which depend on the sensor measurements, and it is important here to use a moving horizon estimation method.

[0019] The moving horizon estimation method, abbreviated as MHE, is known to estimate a measurement signal that exists in a time series. Applicant's research has shown that, in contrast to previously used methods, the use of moving horizon estimation techniques offers significant advantages for estimating sensor measurements to quantify blood glucose concentrations, both in terms of estimation accuracy, flexibility with respect to model predictions, and also in terms of speed at which blood glucose concentrations can be detected. The moving horizon estimation method provides a current blood glucose concentration and a retrospective blood glucose concentration, the latter being calculated taking into account at least one past blood glucose concentration (a previously calculated blood glucose concentration in the measurement value progression (concentration progression)). In this way, the retrospective blood glucose concentration allows a better reconstruction of the current blood glucose measurement signal than in terms of the current blood glucose concentration alone.

[0020] Here, the estimator may be a computer, an integrated circuit or the like, which is particularly configured for optimally calculating, for example, the trace of the matrix. The device and / or the estimator may be configured as a portable device having an independent energy source, e.g. a battery or the like, allowing for efficient operation. Thus, energy consumption when performing the method of the invention is kept as low as possible, thereby allowing the battery to operate for as long as possible, improving the user experience. For this purpose, energy-efficient processors, circuitry, circuits, interfaces, especially wireless interfaces, etc. may be used. In this case, the implementation of the method may be adapted with respect to its parameters to the underlying device, e.g. an estimator, as described below, e.g. with respect to the estimation horizon and / or the noise horizon, so that on the one hand sufficient accuracy and on the other hand long running times are achieved.

[0021] As a result, the estimation of the current glucose concentration in the carrier fluid, and in particular in the blood, is performed efficiently in terms of time and computational resources. Also, the restriction of predefined sensor and / or state transition models is removed, thus providing a significant increase in flexibility compared to known methods. Furthermore, not only is the accuracy of the current glucose concentration improved, but at the same time, the previous glucose concentration is improved as well.

[0022] Additional features, advantages, and embodiments of the invention are described below or disclosed herein.

[0023] A moving horizon estimation method is performed to calculate the current glucose concentration in step d) and is applied to the previously calculated glucose concentration and to at least one previous tissue glucose concentration. This effectively allows a current glucose concentration to be determined based on previously measured glucose concentrations.

[0024] In a further embodiment, the sensor model is provided in the form of a linear function of the measurement and the tissue glucose concentration. This allows the glucose concentration in the interstitial tissue to be calculated efficiently and quickly, while still being sufficiently accurate, based on the measured values ​​of a series of measurements.

[0025] According to another embodiment, the sensor model is provided in the form of a non-linear function of the measurements and the tissue glucose concentration. Here, for example, the following sensor model may be provided, where y is the measured value, I is the glucose concentration, and a, b, c or A, b, c are sensor parameters: y=ca*b / (IG+b)

[0026] Alternatively, the following nonlinear sensor model is possible: y=(A*b+c*IG) / (IG+b)

[0027] In particular, in the case of sensor devices having binding affinity sensors or optical sensors, the accuracy of the glucose concentration in the interstitial tissue calculated based on the measured values ​​of a series of measurements is higher.

[0028] In a further embodiment, the value of the horizon in the moving horizon estimation method for calculating the current glucose concentration is selected to be 10 or less. This allows the glucose concentration in the interstitial tissue to be calculated efficiently, quickly and with sufficient accuracy based on the measured values ​​of a series of measurements.

[0029] In a further embodiment, the variance of the measurement noise and / or the process noise is estimated, in particular at least at certain time intervals. This allows a noise value to be calculated easily and quickly, and overall provides an accurate measurement of the current glucose concentration.

[0030] In a further embodiment, the variance of the measurement noise and / or the variance of the process noise is estimated, in particular interpolated and / or weighted, using exponential smoothing methods. In this case, any time-dependent varying noise values ​​may be adapted or updated, thereby further improving the overall accuracy of the current glucose concentration measurement.

[0031] In a further embodiment, measurements that are only partially used to calculate an estimate of the measurement noise and / or process noise may be temporarily stored, and the required measurements may be interpolated using the stored measurements without being temporarily stored. Thus, for example, the computationally intensive measurement values ​​required to calculate the measurement noise values ​​and / or process noise values ​​can be at least partially stored temporarily and made available for further measurements. Overall, this reduces the amount of calculations required to determine the current glucose concentration without significantly reducing its accuracy.

[0032] In a further embodiment, the number of measurements is chosen to be larger than the value of the horizon in the moving horizon estimation method, in particular at least two times larger, preferably at least five times larger. This ensures accuracy in estimating process and / or measurement noise with respect to the current glucose concentration, improving the overall accuracy of the measurement or quantification of the current glucose concentration.

[0033] In a further embodiment, the measurement noise variance and / or process noise variance are adjusted at regular time intervals based on the horizon sum value of the moving horizon estimation method and the number of measurements for calculating the measurement noise and / or process noise estimate. This ensures that at any given time, the noise value is efficiently adjusted at regular time intervals, on the one hand to achieve sufficient accuracy of the current glucose concentration, and on the other hand to prevent unnecessary adjustments or updates that do not or barely contribute to improving the accuracy of the current glucose concentration.

[0034] In a further embodiment, the calculated values ​​are filtered by a filter function, by means of which errors of the sensor arrangement, in particular measurement errors, are suppressed. The filter function allows for simple removal of erroneous measurements, e.g. sensor errors or outliers in the measurements, i.e. the erroneous measurements are not taken into account when further calculating the current glucose concentration.

[0035] In a further embodiment, the measurement noise value is weighted by a filter function. This prevents underestimation and overestimation of the measurement noise value. That is, underestimation will result in too large an error in the signal or measurement, and overestimation will cause the measurement progression over a series of measurements to be overly smooth. This further improves accuracy overall.

[0036] In a further embodiment, the slope of the increase in the current tissue glucose concentration and / or the current tissue glucose concentration is estimated to determine the error of the sensor device. Alternatively or additionally, this may be done using the current glucose concentration of the carrier-fluid and / or the gradient of increase of the current glucose concentration of the carrier-fluid. This allows errors in the sensor arrangement to be determined simply, reliably and efficiently.

[0037] In a further embodiment, calculated values ​​that are below a predefined low threshold and / or above a predefined high threshold are removed by a filter function, in particular the low and high thresholds corresponding to physiological limit values, preferably the low threshold indicating a value between 10 mg / dL and 50 mg / dL, in particular 30 mg / dL, and the high threshold indicating a value between 100 mg / dL and 600 mg / dL, in particular 450 mg / dL. By appropriate thresholds, erroneous sensor measurements, particularly those containing both blood glucose and tissue glucose values ​​in the carrier fluid, are removed for further calculations using a weighting matrix, preferably a diagonal weighting matrix. The weighting matrix may be such that the sensor device's erroneous measurements are weighted by a factor of 0 and all other measurements are weighted by a factor of 1. Erroneous measurements, for example outliers in sensor measurements, are determined by the absolute glucose concentration in the carrier fluid, particularly in blood, and its rate of change or its slope. In the former case, a physiological limit for blood glucose concentration is introduced, and blood glucose concentrations in the range of 10 mg / dL to 600 mg / dL, preferably 20 mg / dL to 500 mg / dL, and most preferably 30 mg / dL to 450 mg / dL are considered. Measurements outside these ranges are weighted with a factor of 0 as erroneous sensor measurements. The gradient of the glucose concentration in the carrier fluid, in particular in blood, or the rate of change of the glucose concentration in the carrier fluid, in particular in blood, may be determined in a similar manner and the value may be compared with a physiologically realistic rate of change. Therefore, the quantitative rate of change of the glucose concentration in the carrier fluid, particularly in blood, is between 0.1 mg / dL per minute and 15 mg / dL per minute, preferably between 0.5 mg / dL per minute and 10 mg / dL per minute, and most preferably between 1 mg / dL per minute and 3 mg / dL per minute.

[0038] In a further embodiment, a calibration of the measurements of the sensor device is performed after step d). This allows for the use of uncalibrated tissue glucose concentrations, which has the advantage that calibration does not necessarily have to be performed before the moving horizon estimation method is performed, and even more so, calibration may be performed after the moving horizon estimation method as well. A further advantage of using uncalibrated tissue glucose concentrations is that uncalibrated tissue glucose concentrations exhibit a higher correlation to self-monitored blood glucose concentrations, which may advantageously allow, for example, sensor model parameters to be more easily and accurately determined.

[0039] In a further embodiment, the state transition model includes a diffusion model for time-dependent modeling of the diffusion process of glucose from the carrier fluid to the surrounding tissue. The diffusion model provides a simple and less computationally intensive modeling of the attenuation and time delay in the carrier-fluid, particularly between glucose concentrations in blood and tissue, based on diffusion constants.

[0040] In a further embodiment, sensor model parameters of the sensor model and / or state transition parameters of the state transition model are estimated and / or updated at least at regular time intervals. The advantage of this is that overall the accuracy of the measurement of the current glucose concentration is thereby improved, as well as the parameters of these models can be flexibly adjusted to changing conditions or influences.

[0041] Other features and advantages of the invention emerge from the dependent claims, the drawings and the corresponding figure description.

[0042] The features mentioned above, and those described below, may not only be used in the respective combinations described, but also in other combinations or alone by themselves, without departing from the scope of the invention.

[0043] Configurations and embodiments of the invention are illustrated in the drawings and are further explained in the following description. All remodeling steps, including formulas, hypotheses, and methods for solving problems, may be used individually without departing from the scope of the present invention. [Brief description of the drawings]

[0044] [Figure 1] FIG. 1 illustrates the steps of a method according to one embodiment of the present invention. [Diagram 2] FIG. 2 illustrates the steps of a method according to one embodiment of the invention. [Diagram 3] FIG. 3 shows a comparison of a method according to one embodiment of the present invention with known methods. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0045] FIG. 1 details the steps for measuring blood glucose concentration based on a moving horizon estimation method in which the variances of the process noise and the measurement noise are adjusted.

[0046] In the initial phase T1, the method is initialized by steps S1 to S3, which will be described below. After initialization, in the second phase T2, the blood glucose concentration is calculated in discrete time steps based on the moving horizon estimation method in steps S4 to S6 and a decision step E1, which will be described below. In parallel with this, in the third phase T3, measurement noise and process noise are adjusted using steps V1 to V3 described below.

[0047] First, before going into the details of each of the phases T1 to T3 and their steps, the principle of executing the moving horizon estimation method will be explained below. The moving horizon estimation method used below is a method for estimating states by minimizing a cost function, where the cost function is run over a moving time window of n discrete time steps. Here, a discrete-time system is defined. TIFF0007635996000001.tif1236Here, ▲x K ▼ is the vector of state variables, ▲y K ▼ is the measurement vector. The cost function is the measurement noise ▲v K ▼ and process noise ▲ w k ▼ includes weighted norms of

[0048] This results in an optimization problem of the form:

number

[0049] For modeling the diffusion process between blood and surrounding tissue, the following relationship is predicted:

number

[0050] However, for example, f(i k )=(p 1 *i k ) / (p 0 +i k It is also possible to use non-linear models of the form

[0051] Given an ideal and calibrated blood glucose signal x, with the expectation that the output signal is linear in i(t), b =p 0 b(t)+p 1 and an ideal and calibrated tissue glucose signal x i =p 0 i(t)+p 1 may be introduced. Sensor parameter p 0 and p 1 The spreading is also valid for uncalibrated signals, with the further expectation that varies only slowly.

number

[0052] This equation is modeled at time step Δt to obtain the blood glucose concentration x b When is modeled using an autoregressive model, it results in the following discrete state space representation:

number

[0053] As an alternative nonlinear model to the above, the following state space representation would be appropriate: TIFF0007635996000008.tif5053

[0054] In the following, the linear sensor model is applied again. TIFF0007635996000009.tif8169 Finally, the blood glucose concentration can be calculated for the linear model by:

number

[0055] Here, the calibrated, noise-free blood glucose concentration By optimizing TIFF0007635996000011.tif933, the moving horizon problem, formalized in equation (2), is solved. In the following, a matrix representation of the previous state variables or the N-dimensional vector of the sensor measurement values ​​at time point k will be used. TIFF0007635996000012.tif836

[0056] Therefore, the tissue glucose signal is expressed as:

number

number

number

[0057] In this case, the formulation of the moving horizon problem for a linear sensor model is given by:

number

[0058] In the following, the linear sensor model is again applied. Substituting equations (8) and (9) into the moving horizon problem for the linear model, we obtain the following quadratic problem:

number

number

[0059] Initial state z k is calculated according to the solution of the previous estimation step.

number

[0060] In the case of initialization with k+N, the initial states need to be estimated. TIFF0007635996000024.tif8126So, in this case, A init = [BA] and C init =[DC] and B init and D. init The optimization problem can be rewritten by replacing with a zero matrix:

[0061] In this case, the solution to the modified optimization problem is:

number

[0062] TIFF0007635996000026.tif29170 and updating the blood glucose concentration or signal, we get:

number

[0063] In general, CGM systems are sensitive to mechanical disturbances that can result in erroneous measurements, and such erroneous sensor measurements may be taken into account by weighting the measurement noise with a weighting matrix W.

[0064] In the above equations (11) and (12), a weighted inverse covariance matrix may be introduced.

number

number

[0065] The gradient in the corresponding tissue glucose concentration and the current glucose concentration may be used to detect sensor errors or measurement outliers. Alternatively or additionally, high and low thresholds may be applied to the measurement signal or sensor measurements, and measurements that fall outside the low and high thresholds may be classified as erroneous.

[0066] TIFF0007635996000032.tif14170 In general, these two parameters are unknown and need to be estimated. Furthermore, these parameters change over time. Therefore, adjusting or updating the variance will directly result in changes that will improve the quality of the estimation by the moving horizon estimation method. On the other hand, underestimating the measurement noise will result in a very noisy measurement signal, which will lead to erroneous measurements. On the other hand, overestimating the measurement noise or underestimating the process noise will result in a time-delayed estimation, which may also reduce the accuracy of the measurement of the current glucose concentration.

[0067] In the following, a method for predicting the measurement noise variance of the measurement noise and the process noise variance of the process noise will be described. Also described below is a method for easily adjusting or updating the variance in any case.

[0068] The principle of this method, as explained for example in [5], is that every degree of freedom is equal to every other degree of freedom.

[0069] Process noise j-1|k and the measurement noise v j|k But in a horizon of length n, j=k-n+1,...k TIFF0007635996000033.tif8170, covariance matrix R k and the covariance matrix Q k teeth, TIFF0007635996000034.tif11102, which simplifies equation (12) to: TIFF0007635996000035.tif1058 and TIFF0007635996000036.tif20141

[0070] In the next step, equations (7) and (12) are substituted into the measurement noise definition equation (8) and the process noise definition equation (9). TIFF0007635996000037.tif1667

[0071] initial point z k If the variance of is zero, then the covariance matrix takes the form

number

[0072] TIFF0007635996000039.tif17170Since the variance matrix of the measurement signal consists of the covariance matrix of the measurement noise and the covariance matrix of the uncalibrated, noise-free tissue glucose signal, we obtain:

number

[0073] Here, by substituting equation (19) into equation (18) and calculating the inverse matrix, the covariance of the process noise and the covariance of the measurement noise are obtained as follows:

number

number

[0074] TIFF0007635996000043.tif7155TIFF0007635996000044.tif1293

[0075] In this case the variance of the measurement noise is: TIFF0007635996000045.tif1537Here, s(γ k )=tr(AH(γ k )).

[0076] TIFF0007635996000046.tif7111

number

[0077] In a similar manner to the measurement noise, a consistent estimate of the variance of the process noise can be calculated.

number

[0078] TIFF0007635996000049.tif7167. Value gamma k is expressed as follows: TIFF0007635996000050.tif19170 is estimated using a free optimization method.

number

[0079] TIFF0007635996000052.tif10166 and the measurement noise are updated taking into account the smoothing factor η.

number

[0080] The length n of the noise adjustment horizon does not have to match the estimation horizon N. As the estimation horizon N becomes longer, the computational load increases but the accuracy of the estimation improves only slightly. Since the accuracy of the variance estimation and the number of data points are highly correlated, the noise adjustment horizon n is chosen to be significantly larger than the estimation horizon N.

[0081] TIFF0007635996000054.tif8168 = sum of squared process errors, or, correspondingly, sum of squared measurement errors

number

[0082] In the case of erroneous sensor measurements, the noise variance is not updated, but if it is updated, the process TIFF0007635996000056.tif8155

[0083] s(γ k )=tr(AH(γ k In order to be able to estimate , a higher computational load is required. Since matrix A depends only on the predefined matrices A and C and γ, one can create a table of γ and associated ratio ranges and interpolate the values ​​in the table to calculate s for the current value of γ. It is therefore possible to achieve satisfactory and sufficiently accurate estimations in a short time and with a low computational load, even on a computer or tablet.

[0084] In summary, in a first step S1, the matrix W can be estimated from equation (17) during an initialization phase T1. Based on the matrix W, in a second step S2, the ratio of the process variance and the measurement noise variance is estimated from equation (24). In a third step S3, an initial value is estimated from equation (14).

[0085] Based on the estimated initial values, in a fourth step S4 during the initial phase T2, an initial state is calculated from equation (13), and using the initial state, the matrix W is again estimated from equation (17) in a fifth step S5, similar to step S1. In a sixth step S6, an initial value is estimated from equation (12). Then, in step E1, it is determined whether the ratio of the instant k to the sum of the estimation horizon N and the noise adjustment horizon n is an integer greater than or equal to 1. If No, the time index k is incremented by 1, and then steps S4 to S6 and step E1 are executed again. On the other hand, if the answer is Yes, noise adjustment T3 having steps V1 to V3 is executed. In the noise adjustment T3, in the first step V1, the sum of squared process errors or the sum of squared measurement errors is calculated from equation (26). These are then used to calculate, in a second step V2, the corresponding variances of the measurement noise and process noise from equations (22) and (23), and the value of γ is updated from equation (25). After that, steps S4 to S6 and step E1 are executed again.

[0086] Based on the estimated initial values, in a fourth step S4 during the initialization phase T2, the initial values ​​are calculated from equation (13), and using the initial state, the matrix W is again estimated from equation (17) in a fifth step S5, similar to step S1. In a sixth step S6, an initial value is calculated from equation (12). Then, in step E1, it is determined whether the ratio of the instant k to the sum of the estimation horizon N and the noise adjustment horizon n is an integer greater than or equal to 1. If No, the time index k is incremented by 1, and then steps S4 to S6 and step E1 are executed again. On the other hand, if the answer is Yes, noise adjustment T3 having steps V1 to V3 is executed. Here, in a first step V1, the sum of squared process errors or the sum of squared measurement errors is calculated from equation (26). These are then used to calculate, in a second step V2, the corresponding variances of the measurement noise and process noise from equations (22) and (23), and the value of γ is updated from equation (25). After that, steps S4 to S6 and step E1 are executed again.

[0087] In FIG. 2, the steps of a method according to one embodiment of the invention are shown.

[0088] In particular, FIG. 2 shows a method for measuring, preferably continuously, the current glucose concentration in a carrier fluid of an organism, in particular in the blood.

[0089] In step a), a series of measurements of tissue glucose concentration in tissue surrounding the carrier fluid is detected using the sensor device, the series of measurements comprising at least two measurements spaced apart in time.

[0090] In a further step b), the set of detected measurements is used to calculate the tissue glucose concentration based on a sensor model, by means of which the measurements of the sensor device are correlated with the tissue glucose concentration while taking into account the measurement noise.

[0091] In a further step c), a state transition model is provided, by means of which at least one glucose concentration in the transport fluid is correlated with the calculated tissue glucose concentration, taking into account process noise.

[0092] In a further step d), a current glucose concentration is calculated based on the provided state transition model and the calculated tissue glucose concentration, and at least step d), in particular steps b) to d), are performed using at least one moving horizon estimation method.

[0093] FIG. 3 shows a comparison of a method according to one embodiment of the present invention with known methods.

[0094] In the following, a comparison of different blood glucose concentration estimation methods is described and the results are shown in FIG. In detail, the estimation method is carried out using the same sensor device, namely a fiber sensor, known from [7], detecting the blood glucose concentration based on fluorescence measurements, with a sampling rate of every 2 minutes. Estimation of blood glucose concentration by the moving horizon estimation method according to one embodiment of the present invention is now compared with two other estimation methods, namely, the Kalman filter KF and the moving average filter MA using the smoothed sensor signal. We also explain the effect of different parameters on blood glucose concentration estimation. Here, data was collected from eight type 1 and eight type 2 diabetes patients who wore the corresponding CGM sensors of the sensor device over a 28-day period. On days 1, 7, 15, and 28, reference data were determined every 10 minutes for 4 hours using a Yellow Springs Instruments (YSI) 2300STAT Plus Glucose Analyzer (YSI Life Sciences, Yellow Springs, OH).

[0095] Also, the horizon for the moving horizon estimation method was set to N=10, and the horizon for the noise adjustment was set to n=50. The Kalman filter here is based on equation (5). A diffusion constant, i.e., a time constant τ = 6 min, was applied for both methods respectively. The signals generated by the TIFF0007635996000057.tif7167 and the Kalman filter and smoothed by the moving average method are compared, and the comparison is based on their agreement with reference data measured during clinical monitoring at any given time point. To evaluate the three different estimation methods, three evaluation parameters, namely, mean absolute relative difference (MARD), root mean square error (RMSE), and maximum relative absolute difference (maxRAD), were used for the four measurements of all 16 patients.

[0096] To estimate the median values, in the following table the 25th percentile (1st quartile) and 75th percentile (3rd quartile) of the three evaluation parameters are presented.

[0097] [Table 1]

[0098] It can be seen that for the three evaluation parameters, the best results are obtained by the moving horizon estimation method, followed by the Kalman filter signal KF. Signal MA, which shows the tissue glucose signal smoothed and filtered by the moving average method, does not take into account the diffusion process between the blood glucose concentration and the tissue glucose concentration, and therefore gives poor results.

[0099] To illustrate the impact of sensor calibration, different calibration methods are described below and compared with each other.

[0100] TIFF0007635996000059.tif17168 contains the results and provides smaller parameters compared to the results of the moving average signal. TIFF0007635996000060.tif9168g / dL, and median maxRAD=14.2mg / dL).

[0101] TIFF0007635996000061.tif8167 Improve. The increased accuracy results from better accounting for time delays. "Overshoot" due to sudden changes in blood glucose concentration, or noise, is also reduced.

[0102] The effect of the accuracy of blood glucose concentration estimation using CGM signals on calibration error is described below. For this purpose, a two-point calibration method is used. Two reference measurements b 1 and b 2 and the corresponding blood glucose concentration over time. TIFF0007635996000062.tif20136

[0103] For each reference combination and each estimated, uncalibrated blood glucose signal (MHE, pMHE, KF, and MA), sensor parameters are identified and blood glucose concentrations are calculated.

[0104] [Table 2]

[0105] Table 2 above shows the medians and quartiles for the MARD and RMSE methods for all possible calibrations. From Table 2, it can be seen that the median and interquartile difference are smallest for pMHE for the MARD and RMSE methods.

[0106] In summary, embodiments of the present invention have the following advantages and / or features: Compensating for time delays by modeling the diffusion process and estimating the blood glucose levels over a moving horizon (moving horizon estimation method). - Physiological range limits ensure robustness against outliers. · Adaptive estimation of the measurement noise and adaptive estimation of the state noise are combined by adaptively determining the control parameters of the problem. It is now even possible to accommodate slowly changing model parameters. -Efficient implementation ensures high accuracy while using limited computational load. - Efficiently calculate and estimate blood glucose levels over time for past horizons. · Model parameter adaptation. -Improved robustness of estimation by introducing limit values. Flexibility regarding sensor models, e.g. nonlinear sensor models can also be used. Lower computational load conserves limited battery life. Physiological range limits ensure a physiologically reasonable solution. Improved blood glucose estimation for calibration by optimizing past horizons.

[0107] Although the present invention has been described using the embodiments, the present invention is not limited to these and may be modified in various ways.

Claims

1. A method for continuously measuring a current glucose concentration in a transport fluid of a biological organism, comprising: a) detecting a series of measurements of tissue glucose concentration in tissue surrounding the carrier-fluid using a sensor device, the series of measurements including at least two measurements spaced apart in time; b) calculating the tissue glucose concentration based on a sensor model using the detected series of measurements, the sensor model correlating the measurements of the sensor device with the tissue glucose concentration while taking into account measurement noise; c) providing a state transition model, said state transition model correlating at least one glucose concentration in said transport-fluid with said calculated tissue glucose concentration while taking into account process noise; d) calculating the current glucose concentration based on the provided state transition model and the calculated tissue glucose concentration; A method of measuring, wherein at least step d) is performed using at least one moving horizon estimation method, which is performed to calculate the current glucose concentration in step d) and applied to a previously calculated glucose concentration and at least one previous tissue glucose concentration.

2. The method of claim 1 , wherein the sensor model is provided in the form of a linear or non-linear function of the measurement value and the tissue glucose concentration.

3. 3. The method of claim 1, wherein a horizon value in the moving horizon estimation method for calculating the current glucose concentration is selected to be 10 or less.

4. Method for measuring according to any one of claims 1 to 3, wherein the variance of the measurement noise and / or the variance of the process noise is estimated at least at regular time intervals.

5. 5. The method of claim 4, wherein measurement values ​​that are only partially used to calculate the estimate of the measurement noise and / or the process noise can be temporarily stored, and the required measurement values ​​can be interpolated using the stored values ​​without being temporarily stored.

6. 5. A method of measuring according to claim 3 or claim 4, wherein the number of measurements is chosen to be greater than the value of the horizon in the moving horizon estimation method.

7. 5. The method of claim 4, wherein the measurement noise variance and / or the process noise variance are adjusted at regular time intervals based on a horizon sum value of the moving horizon estimation method and a number of measurements for calculating the measurement noise and / or process noise estimate.

8. The method according to any one of the preceding claims, further comprising filtering the calculated value by a filter function, the filter function suppressing errors of the sensor arrangement.

9. The method of claim 8, wherein the gradient of increase of the current tissue glucose concentration and / or the current tissue glucose concentration is estimated to determine the error of the sensor device.

10. 10. The method of claim 8 or 9, wherein calculated values ​​that are below a predefinable low threshold and / or above a predefinable high threshold are eliminated by the filter function.

11. The method according to any one of the preceding claims, wherein a calibration of the measurements of the sensor device is performed after step d).

12. A measuring method according to any one of claims 1 to 11, wherein the state transition model includes a diffusion model for time-dependent modeling of the diffusion process of glucose from the carrier fluid to the surrounding tissue, and / or sensor model parameters of the sensor model and / or state transition parameters of the state transition model are estimated and / or updated at least at regular time intervals.

13. An apparatus for carrying out the method of any one of claims 1 to 12, which continuously measures a current glucose concentration in a transport fluid of a biological body, comprising: a sensor device for measuring fluorescence in tissue surrounding the carrier fluid by means of an optical fiber probe, the sensor device being configured to detect a series of measurements including at least two temporally separated measurements of tissue glucose concentration in the tissue surrounding the carrier fluid; a providing device configured to provide the state transition model, where the state transition model correlates at least one glucose concentration in the transport fluid with the calculated tissue glucose concentration while taking into account process noise, and a providing device configured to provide a sensor model, where the sensor model correlates a measurement value of the sensor device with the tissue glucose concentration while taking into account measurement noise; an estimation device configured to calculate the tissue glucose concentration based on the sensor model using the series of detected measurements, and to calculate the current glucose concentration using a moving horizon estimation method based on the provided state transition model and the calculated tissue glucose concentration; An apparatus comprising:

14. An estimation device for performing the measurement method according to any one of claims 1 to 12, which continuously measures a current glucose concentration in a transport fluid of a biological body, at least one interface for connecting a sensor device to detect a series of measurements of tissue glucose concentration in tissue surrounding the carrier fluid, the series of measurements including at least two measurements spaced apart in time; at least one memory storing a state transition model, where at least one glucose concentration in the carrier fluid is correlated with the tissue glucose concentration by the state transition model while taking into account at least one process noise value, and at least one memory storing a sensor model, where the sensor model correlates a measurement value of the sensor device with the tissue glucose concentration while taking into account at least one measurement noise value; a computing device configured to use the set of detected measurements to calculate the tissue glucose concentration based on the sensor model, and to calculate the current glucose concentration using at least one moving horizon estimation method based on the stored state transition model and the calculated tissue glucose concentration; An estimation device comprising:

15. A non-transitory machine-readable medium for carrying out the method of any one of claims 1 to 12, storing instructions which, when executed on a computer, cause the method to be carried out for continuously measuring a current glucose concentration in a carrier fluid of a biological body, comprising: a) detecting a series of measurements of tissue glucose concentration in tissue surrounding the carrier-fluid using the sensor device, the series of measurements including at least two measurements spaced apart in time; b) calculating the tissue glucose concentration using the series of detected measurements based on a sensor model, the sensor model correlating the measurements of the sensor device with the tissue glucose concentration while taking into account measurement noise; c) providing the state transition model, the state transition model correlating at least one glucose concentration in the transport fluid with the calculated tissue glucose concentration while taking into account process noise; d) calculating the current glucose concentration based on the provided state transition model and the calculated tissue glucose concentration; A non-transitory machine-readable medium, wherein at least step d) is performed using at least one moving horizon estimation method.

Images