Microprocessors, devices, and methods for use in analyte monitoring systems

Abstract

The present invention comprises one or more microprocessors programmed to execute methods for improving the performance of an analyte monitoring device including prediction of glucose levels in a subject by utilizing a predicted slower-time constant (1/k 2 ). In another aspect of the invention, pre-exponential terms (1/c 2 ) can be used to provide a correction for signal decay (e.g., a Gain Factor). In other aspects, the present invention relates to one or more microprocessors comprising programming to control execution of (i) methods for conditional screening of data points to reduce skipped measurements, (ii) methods for qualifying interpolated/extrapolated analyte measurement values, (iii) various integration methods to obtain maximum integrals of analyte-related signals, as well as analyte monitoring devices comprising such microprocessors. Further, the present invention relates to algorithms for improved optimization of parameters for use in prediction models that require optimization of adjustable parameters.

What is claimed is:  
     
         1 . One or more microprocessors, comprising programming to control 
 obtaining a measured charge signal over time, comprising a measured charge signal response curve specifically related to the amount or concentration of the glucose extracted from the subject, wherein said measured charge signal response curve comprises a kinetic region;    using (i) a mathematical model as presented in Eq. (3A)                    Q        (   t   )       =       S   o     +         c   1       k   1            (     1   -            -     k   1          t         )       +         c   2       k   2            (     1   -            -     k   2          t         )                 (       Eq   .              3        A     )                           wherein “Q” represents the charge, “t” represents the elapsed time, “S o ” is a fitted parameter, “c 1 ” and “c 2 ” are pre-exponential terms that correspond to the electric current contribution at t=0 for first and second reactions, respectively, “k 1 ” and “k 2 ” are rate constants for the first and second reactions, respectively, and (ii) an error minimization method, to iteratively estimate values of parameters S o , c 1 , c 2 , k 1 , and k 2  using said model and error minimization method to fit a predicted response curve to said kinetic region of said measured charge signal response curve, wherein (a) the error minimization method provides a calculated error based on differences between kinetic regions of said predicted and measured charge signal response curves and (b) said estimating is iteratively performed until the calculated error between the predicted and measured charge signal response curves is minimized or until no further statistically significant change is seen in the calculated error, at which time iterative estimation of the parameters is stopped, said iterative estimation and error minimization results in estimated values of said parameters; and    correlating 1/k 2  with a glucose amount or concentration to provide a measurement of the amount or concentration of the glucose in the subject.    
     
     
         2 . The one or more microprocessors of  claim 1 , further programmed to control operating a sensing device for obtaining a measured charge signal over time.  
     
     
         3 . The one or more microprocessors of  claim 1 , further programmed to control a measurement cycle comprising (a) operating a sampling device for extracting a sample from the subject, said sample comprising glucose, and (b) operating a sensing device for obtaining a measured charge signal over time.  
     
     
         4 . The one or more microprocessors of  claim 3 , further programmed to perform a series of measurement cycles resulting in a series of measured charge signal response curves.  
     
     
         5 . The one or more microprocessors of  claim 4 , wherein after estimation of each predicted response curve for each measured charge signal response curve in the series of measurements an amount or concentration of the glucose is determined based on each estimated parameter 1/k 2 .  
     
     
         6 . The one or more microprocessors of  claim 1 , wherein correlating 1/k 2  with a glucose amount or concentration to provide a measurement of the amount or concentration of glucose is performed by a method comprising applying a calibration value.  
     
     
         7 . The one or more microprocessors of  claim 6 , wherein said correlating is carried out as follows:  
       
         
           
             
               
                 
                   [ 
                   Glu 
                   ] 
                 
                 t 
               
               = 
               
                 
                   
                     
                       [ 
                       Glu 
                       ] 
                     
                     
                       c 
                        
                       
                           
                       
                        
                       al 
                     
                   
                   
                     
                       ( 
                       
                         1 
                         / 
                         
                           k 
                           2 
                         
                       
                       ) 
                     
                     
                       c 
                        
                       
                           
                       
                        
                       al 
                     
                   
                 
                  
                 
                   
                     ( 
                     
                       1 
                       / 
                       
                         k 
                         2 
                       
                     
                     ) 
                   
                   t 
                 
               
             
           
           
           
               
           
         
         where Glu t  is glucose concentration at time t, Glu cal  is glucose concentration at a time of calibration that corresponds to an estimated 1/k 2  at the time of calibration, and (1/k 2 ) t  is the estimated 1/k 2  at time t.  
       
     
     
         8 . The one or more microprocessors of  claim 1 , wherein said measured charge signal response curve was obtained by integration of a measured current signal response curve, and said one or more microprocessors are further programmed to control said integration.  
     
     
         9 . The one or more microprocessors of  claim 8 , wherein before said integration is performed, said one or more microprocessors are further programmed to control a background subtraction correction of the measured current signal response curve.  
     
     
         10 . The one or more microprocessors of  claim 9 , wherein said obtaining comprises extracting a sample comprising said glucose from the subject into a collection reservoir using a sampling device to obtain a concentration of the glucose in said reservoir and said one or more microprocessors are programmed to control operation of said sampling device.  
     
     
         11 . The one or more microprocessors of  claim 10 , wherein the collection reservoir is in contact with a skin or mucosal surface of the subject and the glucose is extracted across said skin or mucosal surface.  
     
     
         12 . The one or more microprocessors of  claim 11 , wherein glucose is extracted using an iontophoretic current applied to said skin or mucosal surface.  
     
     
         13 . The one or more microprocessors of  claim 12 , wherein the collection reservoir comprises an enzyme that reacts with the extracted glucose to produce an electrochemically detectable signal.  
     
     
         14 . The one or more microprocessors of  claim 13 , wherein the enzyme comprises glucose oxidase.  
     
     
         15 . The one or more microprocessors of  claim 13 , wherein said electrochemically detectable signal is peroxide, said signal is detected at a reactive surface of a biosensor electrode, said detecting is accomplished using a sensing device, and said one or more microprocessors are further programmed to control operation of said sensing device.  
     
     
         16 . The one or more microprocessors of  claim 15 , wherein said kinetic region of said measured charge signal response curve corresponds to a measurement time period of 0 to about 180 seconds.  
     
     
         17 . An analyte monitoring system, comprising 
 the one or more microprocessors of  claim 2;  and    the sensing device used to obtain said measured charge signal response curve.    
     
     
         18 . An analyte monitoring system, comprising 
 the one or more microprocessors of  claim 3;     the sampling device; and    the sensing device used to obtain said measured charge signal response curve.    
     
     
         19 . The monitoring system of  claim 18 , wherein the sampling device comprises a laser device.  
     
     
         20 . The monitoring system of  claim 18 , wherein the sampling device comprises a sonophoretic device.  
     
     
         21 . The monitoring system of  claim 18 , wherein the sampling device comprises an iontophoretic device.  
     
     
         22 . A method of providing a glucose amount or concentration in a subject, comprising 
 obtaining a measured charge signal over time, comprising a measured charge signal response curve specifically related to the amount or concentration of the glucose extracted from the subject, wherein said measured charge signal response curve comprises a kinetic region;    using (i) a mathematical model as presented in Eq. (3A)                    Q        (   t   )       =       S   o     +         c   1       k   1            (     1   -            -     k   1          t         )       +         c   2       k   2            (     1   -            -     k   2          t         )                 (       Eq   .              3        A     )                           wherein “Q” represents the charge, “t” represents the elapsed time, “S o ” is a fitted parameter, “c 1 ” and “c 2 ” are pre-exponential terms that correspond to the electric current contribution at t=0 for first and second reactions, respectively, “k 1 ” and “k 2 ” are rate constants for the first and second reactions, respectively, and (ii) an error minimization method, to iteratively estimate values of parameters S o , c 1 , c 2 , k 1 , and k 2  using said model and error minimization method to fit a predicted response curve to said kinetic region of said measured charge signal response curve, wherein (a) the error minimization method provides a calculated error based on differences between kinetic regions of said predicted and measured charge signal response curves, and (b) said estimating is iteratively performed until the calculated error between the predicted and measured charge signal response curves is minimized or until no further statistically significant change is seen in the calculated error, at which time iterative estimation of the parameters is stopped, said iterative estimation and error minimization results in estimated values of said parameters; and    correlating 1/k 2  with a glucose amount or concentration to provide a measurement of the amount or concentration of the glucose in the subject.    
     
     
         23 . One or more microprocessors, comprising programming to control 
 obtaining a measured charge signal over time using an electrochemical sensor, said measured charge signal comprising a measured charge signal response curve specifically related to an amount or concentration of glucose extracted from a subject, wherein said measured charge signal response curve comprises a kinetic region;    using (i) a mathematical model as presented in Eq. (3A)                    Q        (   t   )       =       S   o     +         c   1       k   1            (     1   -            -     k   1          t         )       +         c   2       k   2            (     1   -            -     k   2          t         )                 (       Eq   .              3        A     )                           wherein “Q” represents the charge, “t” represents the elapsed time, “S o ” is a fitted parameter, “c 1 ” and “c 2 ” are pre-exponential terms that correspond to the electric current contribution at t=0 for first and second reactions, respectively, “k 1 ” and “k 2 ” are rate constants for the first and second reactions, respectively, and (ii) an error minimization method, to iteratively estimate values of parameters S o , c 1 , c 2 , k 1 , and k 2  using said model and error minimization method to fit a predicted response curve to said kinetic region of said measured charge signal response curve, wherein (a) the error minimization method provides a calculated error based on differences between kinetic regions of said predicted and measured charge signal response curves, and (b) said estimating is iteratively performed until the calculated error between the predicted and measured charge signal response curves is minimized or until no further statistically significant change is seen in the calculated error, at which time iterative estimation of the parameters is stopped, said iterative estimation and error minimization results in estimated values of said parameters; and    correcting for signal decay of the electrochemical sensor by multiplying the measured charge signal by a gain factor estimated from 1/c 2 .    
     
     
         24 . The one or more microprocessors of  claim 23 , further programmed to control operating a sensing device for obtaining a measured charge signal over time.  
     
     
         25 . The one or more microprocessors of  claim 23 , further programmed to control a measurement cycle comprising (a) operating a sampling device for extracting a sample from the subject, said sample comprising glucose and (b) operating a sensing device for obtaining a measured charge signal over time.  
     
     
         26 . The one or more microprocessors of  claim 25 , further programmed to perform a series of measurement cycles resulting in a series of measured charge signal response curves.  
     
     
         27 . The one or more microprocessors of  claim 26 , wherein after estimation of each predicted response curve for each measured charge signal response curve in the series of measurements, said one or more microprocessors are further programmed to determine a gain factor on each estimated parameter 1/c 2  and multiply each gain factor by the measured charge signal corresponding to the predicted response curve from which the gain factor was estimated.  
     
     
         28 . The one or more microprocessor of  claim 27 , wherein said series of measurements comprise measured charge signal response curves at times t, t-1, t-2, etc.  
     
     
         29 . The one or more microprocessor of  claim 28 , further programmed to normalize and/or smooth two or more gain factors from the series of measurements to obtain a normalized and/or smoothed gain factor, and correct for signal decay of the electrochemical sensor by multiplying the measured charge signal at time t by said normalized and/or smoothed gain factor.  
     
     
         30 . The one or more microprocessor of  claim 29 , wherein the series comprises at least five measured charge signal response curves, and said normalized and/or smoothed gain factor is calculated based on (1/c 2 ) t , (1/c 2 ) t-1 , (1/c 2 ) t-2 , (1/c 2 ) t-3 , and (1/c 2 ) t-4 .  
     
     
         31 . The one or more microprocessors of  claim 23 , wherein said measured charge signal response curve was obtained by integration of a measured current signal response curve, and said one or more microprocessors are further programmed to control said integration.  
     
     
         32 . The one or more microprocessors of  claim 31 , wherein before said integration is performed said one or more microprocessors are further programmed to control a background subtraction correction of the measured current signal response curve.  
     
     
         33 . The one or more microprocessors of  claim 32 , wherein said obtaining comprises extracting a sample comprising said glucose from the subject into a collection reservoir using a sampling device to obtain a concentration of the glucose in said reservoir and said one or more microprocessors are programmed to control operation of said sampling device.  
     
     
         34 . The one or more microprocessors of  claim 33 , wherein the collection reservoir is in contact with a skin or mucosal surface of the subject and the glucose is extracted across said skin or mucosal surface.  
     
     
         35 . The one or more microprocessors of  claim 34 , wherein glucose is extracted using an iontophoretic current applied to said skin or mucosal surface.  
     
     
         36 . The one or more microprocessors of  claim 35 , wherein the collection reservoir comprises an enzyme that reacts with the extracted glucose to produce an electrochemically detectable signal.  
     
     
         37 . The one or more microprocessors of  claim 36 , wherein the enzyme comprises glucose oxidase.  
     
     
         38 . The one or more microprocessors of  claim 37 , wherein said electrochemically detectable signal is peroxide, said signal is detected at a reactive surface of the electrochemical sensor, said detecting is accomplished using a sensing device, and said one or more microprocessors are further programmed to control operation of said sensing device.  
     
     
         39 . The one or more microprocessors of  claim 38 , wherein said kinetic region of said measured charge signal response curve corresponds to a measurement time period of 0 to about 180 seconds.  
     
     
         40 . An analyte monitoring system, comprising 
 the one or more microprocessors of  claim 24;  and    the sensing device used to obtain said measured charge signal response curve.    
     
     
         41 . An analyte monitoring system, comprising 
 the one or more microprocessors of  claim 25;     the sampling device; and    the sensing device used to obtain said measured charge signal response curve.    
     
     
         42 . The monitoring system of  claim 41 , wherein the sampling device comprises a laser device.  
     
     
         43 . The monitoring system of  claim 41 , wherein the sampling device comprises a sonophoretic device.  
     
     
         44 . The monitoring system of  claim 41 , wherein the sampling device comprises an iontophoretic device.  
     
     
         45 . A method of correcting for signal decay of an electrochemical sensor used for the detection of an amount or concentration of glucose in a subject, said method comprising 
 obtaining a measured charge signal over time using said electrochemical sensor, said measured charge signal comprising a measured charge signal response curve specifically related to the amount or concentration of glucose extracted from the subject, wherein said measured charge signal response curve comprises a kinetic region;    using (i) a mathematical model as presented in Eq. (3A)                    Q        (   t   )       =       S   o     +         c   1       k   1            (     1   -            -     k   1          t         )       +         c   2       k   2            (     1   -            -     k   2          t         )                 (       Eq   .              3        A     )                           wherein “Q” represents the charge, “t” represents the elapsed time, “S o ” is a fitted parameter, “c 1 ” and “c 2 ” are pre-exponential terms that correspond to the electric current contribution at t=0 for first and second reactions, respectively, “k 1 ” and “k 2 ” are rate constants for the first and second reactions, respectively, and (ii) an error minimization method, to iteratively estimate values of parameters S o , c 1 , c 2 , k 1 , and k 2  using said model and error minimization method to fit a predicted response curve to said kinetic region of said measured charge signal response curve, wherein (a) the error minimization method provides a calculated error based on differences between kinetic regions of said predicted and measured charge signal response curves, and (b) said estimating is iteratively performed until the calculated error between the predicted and measured charge signal response curves is minimized or until no, further statistically significant change is seen in the calculated error, at which time iterative estimation of the parameters is stopped, said iterative estimation and error minimization results in estimated values of said parameters; and    correcting for signal decay of the electrochemical sensor by multiplying the measured charge signal by a gain factor estimated from 1/c 2 .    
     
     
         46 . One or more microprocessors, comprising programming to control 
 providing a measurement value related to glucose amount or concentration in a subject, a skin conductance reading associated in time with said glucose measurement value, and one or more further data integrity screens associated with said glucose measurement value; and    accepting said measurement value when either (i) said skin conductance reading and said one or more further data integrity screens fall within predetermined acceptable ranges or within predetermined threshold values, or (ii) said skin conductance reading falls outside of predetermined acceptable range or beyond predetermined threshold value and said one or more further data integrity screens fall within predetermined acceptable ranges or with predetermined threshold values, or    skipping said measurement value when said skin conductance reading falls outside of predetermined acceptable range or beyond predetermined threshold value and one or more of said one or more further data integrity screens fall outside of predetermined acceptable ranges or beyond predetermined threshold values.    
     
     
         47 . The one or more microprocessors of  claim 46 , wherein said one or more further data integrity screens is selected from the group consisting of peak sensor current and background current.  
     
     
         48 . An analyte monitoring system comprising, 
 said one or more microprocessors of  claim 46;     a sensing device used to provide said measurement value related to glucose amount or concentration; and    a skin conductance measurement device used to provide said skin conductance reading, wherein said one or more microprocessors are further programmed to control operation of said sensing device and said skin conductance measurement device.    
     
     
         49 . One or more microprocessors, comprising programming to control 
 providing a measurement signal, comprising data points, related to glucose amount or concentration in a subject, wherein said data points typically have a monotonic trend; and    evaluating said data points for one or more non-monotonic event, wherein 
 (i) if the data points have an acceptable monotonic trend the measurement signal is accepted for further processing, or  
 (ii) if the data points comprise one or more non-monotonic events then a percent contribution of said one or more non-monotonic events relative to total measurement signal is further evaluated, wherein if the percent contribution is less than a predetermined threshold value or falls within a predetermined range relative to total measurement signal, then the measurement signal is accepted for further processing; however, if the percent contribution is greater than a predetermined threshold value or falls outside a predetermined range relative to total measurement signal, then the measurement signal is not accepted for further processing and the measurement signal is skipped.  
   
     
     
         50 . The one or more microprocessors of  claim 49 , wherein said measurement signal comprising data points that typically have a monotonic trend is either a current measurement or a charge measurement.  
     
     
         51 . An analyte monitoring system, comprising 
 said one or more microprocessors of  claim 49;  and    a sensing device used to provide said measurement signal related to glucose amount or concentration.    
     
     
         52 . One or more microprocessors, comprising programming to control: 
 qualifying whether an unusable analyte-related electrochemical current signal from a given measurement cycle should be replaced by interpolation or extrapolation by applying one or more of the following criteria: (i) if a sensor consistency check value for the measurement cycle falls within a predetermined acceptable range or within a predetermined threshold then the corresponding analyte-related signal may be replaced; (ii) if a change in background current for the measurement cycle falls within a predetermined acceptable range or within a predetermined threshold then the corresponding analyte-related signal may be replaced; (iii) if a change in temperatures falls within a predetermined acceptable range or within a predetermined threshold then the corresponding analyte-related signal may be replaced; and    replacing, in a series of analyte-related signals, an unusable analyte-related signal with an estimated signal by either: 
 (A) if one or more analyte-related signals previous to the unusable analyte-related signal and one or more analyte-related signals subsequent to the unusable analyte related signal are available, then interpolation is used to estimate the unusable, intervening analyte-related signal, or  
 (B) if two or more analyte-related signals previous to the unusable analyte-related signal are available, then extrapolation is used to estimate the unusable, subsequent analyte-related signal;  
   wherein said series of analyte-related signals is obtained from an analyte monitoring device over time, and each analyte-related signal is related to an amount or concentration of analyte in a subject being monitored with the analyte monitoring device.    
     
     
         53 . The one or more microprocessors of  claim 52 , wherein said analyte monitoring device comprises a sensing device and said one or more microprocessors are further programmed to control operation of said sensing device.  
     
     
         54 . The one or more microprocessors of  claim 53 , wherein said analyte monitoring device further comprises a sampling device and said one or more microprocessors are further programmed to control operation of said sampling device.  
     
     
         55 . An analyte monitoring system, comprising 
 said one or more microprocessors of  claim 53;  and    the sensing device used to provide said analyte-related signals.    
     
     
         56 . An analyte monitoring system, comprising 
 said one or more microprocessors of  claim 54;     the sampling device; and    the sensing device used to provide said analyte-related signals.    
     
     
         57 . One or more microprocessors, comprising programming to control 
 selecting a current integration method for an analyte-related current signal, wherein said analyte-related current signal comprises data points, a two sensor system is used for detecting said analyte-related current signal, each of said two sensors are electrochemical sensors, each sensor alternately acts as cathode and anode, a current signal, comprising data points, is detected in a half-measurement cycle from the anode and the cathode, and the analyte-related current signal is obtained from the cathode;    determining a background baseline for a given sensor when acting as cathode is determined from the last two data points of the current signal detected for the same sensor in a previous half-cycle when the sensor acted as an anode; and    subtracting the background baseline from the analyte-related current signal and if over-subtraction of the analyte-related current signal occurs, employing one of the following integration methods to determine an analyte-related charge signal based on the analyte-related current signal: (i) stopping integration when the maximum integral is reached and using the maximum integral as the analyte-related charge signal; or (ii) recalculating a background baseline based on the last two data points from the analyte-related current signal at the cathode, subtracting the recalculated background baseline from the analyte-related current signal, and integrating the background subtracted analyte-related current signal to obtain the analyte-related charge signal.    
     
     
         58 . An analyte monitoring system, comprising 
 said one or more microprocessors of  claim 57;  and    a sensing device comprising said two sensor system.    
     
     
         59 . One or more algorithms to optimize parameters for use in a model that requires optimization of adjustable parameters, said one or more algorithms comprising 
 dividing a data set into a training set and a validation set;    training the model to determine the adjustable parameters using said training set;    stopping said training before the model parameters fully converged; and    validating the parameters using the validation set, wherein said validated parameters are optimized parameters for use in the model.    
     
     
         60 . The one or more algorithms of  claim 59 , wherein the model is a Mixtures of Experts (MOE) model.  
     
     
         61 . One or more algorithms to optimize parameters for use in a prediction model used by an analyte monitoring device, wherein said prediction model requires optimization of adjustable parameters, said one or more algorithms comprising 
 optimizing the parameters based on multiple analyte readings that quantify two or more regions corresponding to various levels of accuracy for the prediction model used by the analyte monitoring device, wherein one or more of the regions have an associated higher risk relative to one or more other regions, such that optimization of the parameters is carried out until error associated with the prediction model is minimized in the regions associated with higher risk.    
     
     
         62 . The one or more algorithms of  claim 61 , wherein said optimizing comprises optimizing a distribution of paired points by (a) constructing an x-y plane of paired points representing (i) a target analyte amount or concentration measured independently as the x coordinate, and (ii) a corresponding model prediction of target analyte amount or concentration as a paired y coordinate, wherein the model is employed by an analyte monitoring device; (b) dividing the x-y plane divided into two or more regions corresponding to various levels of accuracy for the model prediction of the analyte monitoring device; (c) constructing individual mathematical risk functions (F) that assign a numerical value to each paired point (pp) for a particular region; (d) summing the individual risk functions to provide a total risk function; and (e) minimizing the total risk function resulting in optimized parameters for the model.  
     
     
         63 . The one or more algorithms of  claim 61 , wherein the model is a Mixtures of Experts (MOE) model.  
     
     
         64 . The one or more algorithms of  claim 61 , wherein the analyte is glucose.  
     
     
         65 . The one or more algorithms of  claim 64 , wherein (i) said two or more regions corresponding to various levels of accuracy for the prediction model comprise a hypoglycemic region, a glucose target range, and a hyperglycemic region, and (ii) said one or more of the regions that have an associated higher risk relative to one or more other regions comprise said hypoglycemic region and said hyperglycemic region.