Method of Determining Parameters of a Test Fluid

Abstract

Determining first and second parameters of a fluid sample includes obtaining a first data set including data from output signals as a function of pluralities of the first and second parameters. The method includes applying an autocorrelation function to the output signals set so as to obtain a second data set including data from a plurality of autocorrelation signals as a function of the pluralities of the first and second parameters. The method includes generating a test output signal at a device by reacting the device with the fluid sample, applying the autocorrelation function to the test output signal so as to obtain a test autocorrelation signal, identifying in the first and second data sets an intersection of data from the test output signal with corresponding data from the test autocorrelation signal, and determining the first and second parameters of the sample based on the intersection.

Claims

1. A method of determining first and second parameters of a test sample of a test fluid, comprising: obtaining a first data set, the first data set comprising data from a plurality of output signals as a function of pluralities of the first and second parameters, wherein each output signal is representative of an output signal generated at a test device reacting with a corresponding sample of the test fluid; applying an autocorrelation function to the plurality of output signals set so as to obtain a second data set, the second data set comprising data from a plurality of autocorrelation signals as a function of the pluralities of the first and second parameters; generating a test output signal at a test device by reacting the test device with the test sample of the test fluid; applying the autocorrelation function to the test output signal so as to obtain a test autocorrelation signal; identifying in the first and second data sets an intersection of data from the test output signal with corresponding data from the test autocorrelation signal; and determining the first and second parameters of the test sample based on the intersection.

2. The method of claim 1, wherein each output signal comprises a plurality of output values as a function of time, and wherein the first data set comprises a plurality of output values at a specific time as a function of the pluralities of the first and second parameters.

3. The method of claim 2, wherein each autocorrelation signal comprises a plurality of autocorrelation values as a function of lag, and wherein the second data set comprises a plurality of autocorrelation values at a specific lag as a function of the pluralities of the first and second parameters.

4. The method of claim 3, wherein identifying the intersection comprises identifying an intersection of the plurality of output values at the specific time with the plurality of autocorrelation values at the specific lag.

5. The method of claim 2, wherein the specific time is approximately 5 seconds from when the output signal is first generated at the test device reacting with the corresponding sample of the test fluid.

6. The method of claim 3, wherein the specific lag is selected based on a dissimilarity between the plurality of output values at the specific time and the plurality of autocorrelation values at the specific lag.

7. The method of claim 6, wherein the dissimilarity comprises a dissimilarity between a variation of the plurality of output values at the specific time as a function of the pluralities of the first and second parameters, and a variation of the plurality of autocorrelation values at the specific lag as a function of the pluralities of the first and second parameters.

8. The method of claim 1, wherein the test device is an electrochemical test device.

9. The method of claim 1, wherein the first parameter is a concentration of an analyte in the test sample.

10. The method of claim 9, wherein the analyte is any one of: glucose, ketone, lactate, glycerol and cholesterol.

11. The method of claim 1, wherein the test fluid is blood and wherein the second parameter is the haematocrit of the test sample.

12. The method of claim 1, wherein the first data set is obtained by modelling the reactions of the test device with the plurality of samples of the test fluid.

13. The method of claim 1, wherein the first data set is obtained by reacting each of the plurality of samples of the test fluid with the test device.

14. The method of claim 1, wherein identifying the intersection comprises using numerical analysis to solve equations representing the data from the first and second data sets.

15. The method of claim 1, wherein the test output signal comprises a current generated at the test device.

16. An apparatus, comprising: one or more memories storing: a first data set comprising data from a plurality of output signals as a function of pluralities of first and second parameters, wherein each output signal is representative of an output signal generated at a test device reacting with a corresponding sample of a test fluid; and a second data set comprising data from a plurality of autocorrelation signals as a function of the pluralities of the first and second parameters; means for reading a test output signal generated at a test device by reacting the test device with a test sample of the test fluid; and one or more processors arranged to: apply an autocorrelation function to the test output signal so as to obtain a test autocorrelation signal; identify in the first and second data sets an intersection of data from the test output signal with corresponding data from the test autocorrelation signal; and determine the first and second parameters of the test sample based on the intersection.

17. A computer-readable medium having instructions stored thereon, wherein the instructions are configured when executed to cause a computer to: obtain a first data set, the first data set comprising data from a plurality of output signals as a function of pluralities of the first and second parameters, wherein each output signal is representative of an output signal generated at a test device reacting with a corresponding sample of the test fluid; apply an autocorrelation function to the plurality of output signals set so as to obtain a second data set, the second data set comprising data from a plurality of autocorrelation signals as a function of the pluralities of the first and second parameters; generate a test output signal at a test device by reacting the test device with the test sample of the test fluid; apply the autocorrelation function to the test output signal so as to obtain a test autocorrelation signal; identify in the first and second data sets an intersection of data from the test output signal with corresponding data from the test autocorrelation signal; and determine the first and second parameters of the test sample based on the intersection.

18. (canceled)

Description

BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Specific embodiments will now be described in connection with the accompanying drawings, of which:

[0033] FIG. 1 is a schematic representation of a meter arranged to read an electrochemical test strip, in accordance with an embodiment;

[0034] FIG. 2 shows a method of determining parameters of a sample of a test fluid, in accordance with an embodiment;

[0035] FIG. 3 shows a first data set showing end current as a function of plasma glucose and haematocrit;

[0036] FIG. 4 is a plot of an autocorrelation signal as a function of lag k;

[0037] FIG. 5 is a plot of a current transient as a function of time;

[0038] FIG. 6 shows a second data set showing autocorrelation values for a lag of 25 as a function of plasma glucose and haematocrit;

[0039] FIG. 7 is a plot of the data of FIG. 3 combined with the data of FIG. 6;

[0040] FIG. 8 shows a surface representation of the first data set of FIG. 3;

[0041] FIG. 9 shows a surface representation of the second data set of FIG. 6;

[0042] FIG. 10 is a plot showing end current measurements as a function of plasma glucose;

[0043] FIG. 11 is a plot of mean percent bias from glucose reference measurement, as a function of haematocrit;

[0044] FIG. 12 is a surface representation of mean current as a function of plasma glucose and haematocrit;

[0045] FIG. 13 is a plot of an autocorrelation signal as a function of lag k;

[0046] FIG. 14 is a plot of autocorrelation values for a lag of 50 as a function of plasma glucose and haematocrit;

[0047] FIG. 15 is a plot of end current as a function of plasma glucose and haematocrit; and

[0048] FIG. 16 is a plot of mean percent bias from glucose reference measurement, as a function of haematocrit.

DETAILED DESCRIPTION OF SPECIFIC EMBODIMENTS

[0049] The presently disclosed embodiments seek to provide an improved method of determining parameters of a test fluid. Whilst various embodiments are described below, the contemplated embodiments are not limited to these embodiments, and variations of these embodiments may well fall within the scope of the appended claims.

[0050] FIG. 1 shows a strip-meter system 10 according to an embodiment. System 10 comprises a meter 12 for reading an electrochemical test strip 14. Electrochemical test strip 14 comprises one or more working electrodes (not shown) and a counter/reference electrode, each of the working electrodes having a reagent coated thereon for reacting with a sample of test fluid to be applied to electrochemical test strip 14. The counter/reference electrode may also have a reagent coated thereon. Meter 12 comprises receiving means 13 for receiving test strip 14 and applying a potential difference between the working electrode(s) and the counter/reference electrode.

[0051] Meter 12 further comprises processing circuitry 15 for carrying various functions relating to the operation of meter 12. For example, processing circuitry 15: controls operation of receiving means 13 so to control application of a potential difference between the working electrode(s) and the counter/reference electrode; processes transients generated at test strip 14; controls the display of messages on display 18; etc. Meter 12 further comprises a memory storage 16 and a display 18 for displaying readouts of measurements taken by meter 12.

[0052] FIG. 2 shows a method of determining parameters of a test fluid, in accordance with an embodiment. It should be noted that FIG. 2 shows an example method, and the order of the steps may be changed (for example the point in time at which the strip is inserted in the meter) without departing from the scope of the disclosed embodiments. The method may also comprise a fewer or greater number of steps.

[0053] At step 21, a first data set is obtained. The first data set comprises data (e.g. end current) from a plurality of current transients representing current responses generated at an electrochemical test device. The end current is shown as a function of both plasma glucose (e.g. the concentration of glucose within the plasma portion of blood) and haematocrit. In other embodiments, meter 12 could of course be configured to determine the glucose concentration in the whole blood sample (i.e. the glucose content of both the plasma and red blood cells). An example of the first data set is illustrated in FIG. 3, which shows contours of constant end current (μA) as a function of haematocrit and plasma glucose.

[0054] It is clear from FIG. 3 that the higher the haematocrit, the lower the end current response for a given plasma glucose, and vice versa. Thus, when a given end current value is obtained, it could have been generated by any combination of glucose and haematocrit along the appropriate contour. For example, a measured end current of 30 μA could mean a glucose concentration between 270 mg/dL and 440 mg/dL. If another quantity or parameter can be measured which has a different sensitivity to haematocrit and plasma glucose than end current, then it is possible to obtain simultaneous estimates of plasma glucose and haematocrit.

[0055] The data from FIG. 3 was obtained using a simulated system. A system of reaction-diffusion equations modelling the chemical and physical properties of an electrochemical test strip was developed. Using the model, a set of transients was obtained, with haematocrit and glucose values on a grid over the stated range. An algorithm for determining a concentration of glucose within each test sample was then applied to each current transient. Examples of such algorithms are disclosed in co-pending UK patent application no. 1419799.0, which is incorporated herein by reference.

[0056] The 5-second current value (known as the end current) from each transient was recorded, in addition to the glucose and haematocrit values used to create the transient. In the present embodiment, the data may be pre-stored in memory 16 of meter 10 for use in the method described in more detail below. However, in other embodiments the data may be ‘real’ data, i.e. data obtained from multiple tests carried out on test samples of various different known haematocrit and plasma glucose levels. Examples of such real data are disclosed below in connection with FIGS. 10-16.

[0057] The sample autocorrelation function is a well-known means of measuring the degree of correlation between values in a signal, based on the separation in time between the values. Without loss of generality, assume that the signal has N sequential readings equally spaced in time: x(1), x(2), . . . , x(N). Then, the autocorrelation coefficient r(k) for lag of length k is defined as:

[00001]$r\left(k\right)=\frac{c\left(k\right)}{c\left(0\right)}$$c\left(k\right)=\frac{1}{N}.Math.\underset{t=1}{\overset{N-k}{.Math.}}.Math..Math.\left(x\left(t\right)-\stackrel{\_}{x}\right).Math.\left(x\left(t+k\right)-\stackrel{\_}{x}\right),.Math.k=0,1,2,\mathrm{.Math.}.Math.,K$

r(k) is the autocorrelation coefficient for lag k, c(k) is the autocovariance function of the lag k, K is a maximum lag less than N, and x is the mean of the signal readings. A particular c(k) describes the covariance between points in the transient k sample time points apart (lag of k). Not all the c(k) values need to be calculated—only the ones of interest depending on the lag. When scaled by c(0), r(k) is obtained—the autocorrelation coefficient between points k samples apart.

[0058] An example autocorrelation plot applied to a transient (real or virtual) is shown in FIG. 4. As can be seen, the r(k) function decreases steadily with lag length and shows some more complicated behaviour towards the end of the plot, rising and then falling again. In the present embodiment, the current/time transient from which the autocorrelation plot of FIG. 4 was obtained is shown in FIG. 5. The transient was obtained for a sample where the plasma glucose was 45 mg/dL and the haematocrit was 30%. 50 measurements of current were taken in a space of 5 seconds.

[0059] By applying the autocorrelation function to each transient used to obtain the first data set, a second data set is obtained (step 22). An example of the second data set can be seen in FIG. 6 and shows contours of constant autocorrelation coefficient, r, for a specific lag, as a function of haematocrit and plasma glucose. In the present embodiment a specific lag of k=25 was chosen to give a ‘landscape’ (FIG. 6) that is sufficiently distinguished to the 5-second end current contour map of FIG. 3. It is clear that the shape of the autocorrelation coefficient response to haematocrit and plasma glucose is quite different to that of end current as in FIG. 3. In particular, in the end current map of FIG. 3, end current increases as plasma glucose increases, but decreases as haematocrit increases. On the other hand, in the |r(25)| map of FIG. 6, |r(25)| increases as both plasma glucose and haematocrit increase. The intersection of the two data sets can be seen in FIG. 7.

[0060] Using the first and second data sets, the method is able to simultaneously determine or at least estimate the plasma glucose and haematocrit for a given test sample of blood. At step 23, electrochemical test strip 14 is inserted into receiving means 13 of meter 12, in a reading position. In the reading position, receiving means 13 is positioned relative to the working electrode(s) of strip 14 so as to be able to apply a potential difference across the working electrode(s) and the counter/reference electrode, as known in the art. Receiving means 13, under control of processor 15, then applies a potential difference across the working electrode(s) and the counter-reference electrode. At step 24, a test sample of blood having unknown plasma glucose and haematocrit is applied to strip 14. As known in the art, a current/time transient is generated as the blood flows into contact with the working electrode(s) and the counter/reference electrode. The glucose in the blood reacts with the reagent on the working electrode(s), and causes a current to flow between the working electrode(s) and the counter/reference electrode. At step 25 the current response is measured by the meter using processor 15. It should be understood that other analytes in the blood may be measured, such as ketones, lactate, glycerol or cholesterol, and that in the present embodiment glucose is merely used as an example.

[0061] Once collected, at step 26 the autocorrelation function is applied to the test transient as explained above, thereby obtaining an autocorrelation coefficient r(k). The end current and r surfaces can be approximated by suitable functions. By way of example, the surfaces may be represented in polynomial form as per the below:

[00002]$C=\underset{j,k}{.Math.}.Math..Math.p_{j,k}.Math.X^j.Math.Y^k$$R=\underset{j,k}{.Math.}.Math.q_{j,k}.Math.X^j.Math.Y^k$

[0062] Here, X denotes haematocrit, Y denotes glucose concentration, R=|r(25)| and C denotes end current. X and Y may then be obtained simultaneously using the actual measurements of C and R for the test transient. The C and R contours are characterised as two surfaces in FIGS. 8 and 9, using polynomials with up to second order terms. Numerical procedures for solving simultaneous polynomials are plentiful, and may be implemented in a hand-held device such as meter 12. The general procedure for implementing a Newton-Raphson solver iterates until a solution is known as precisely as it is desired, and may be written compactly in matrix notation as:

X.sub.n+1=X.sub.n−J.sup.−1(X.sub.n)F(X.sub.n)

n is the index of the iteration, X is the vector of the two values to be sought (haematocrit and plasma glucose), F is the vector of equations to be solved, and J.sup.−1 is the inverse of the Jacobian matrix of F.

[0063] At step 27, processor 15 applies such an iterative method to the C and R surfaces to identify where they intersect, and obtains estimates of the plasma glucose and haematocrit of the blood sample (step 28). Of course, other methods of solving two simultaneous equations with two unknowns may be used. The plasma glucose concentration may be displayed to the user on display 18.

[0064] FIGS. 10-16 relate to data obtained from transients generated from real tests on blood samples, as opposed to a simulated system. The figures build on the model-based concept described above by looking at data from physical test strips measuring a particular analyte in blood.

[0065] FIG. 10 shows 572 end current values, each taken at the 5-second point from the start of the test, for a range of glucose concentration (50-500 mg/dL) and haematocrit values (20%-60%). In general, a measurement frequency of at least 10 measurements per second, and at least 50 measurements in total, yields appropriate current transients. The measurements should be equally spaced in time. It is clear from FIG. 10 that there is a relatively strong glucose signal as current is substantially proportional to glucose concentration. However, there is also much variation at any given value of glucose concentration.

[0066] FIG. 11 shows by way of illustration the mean percent bias from glucose reference measurement for the 500 mg/dL plasma glucose concentration and each of the haematocrit levels (circles), as well as the linear regression line of best fit. This graph also clearly shows a pronounced sensitivity to haematocrit in the signal, albeit with noise.

[0067] Combining the data from FIG. 10 with FIG. 11 gives FIG. 12, which shows the mean current at the different combinations of glucose and haematocrit. In order to reduce the sensitivity to haematocrit, the autocorrelation function at various lags can be calculated for this data, as described above.

[0068] FIG. 13 plots the dynamic range of autocorrelation values obtained at each lag. A possible means of choosing the best lag is to use one from the region of highest range (between 40 to 60 in the example of FIG. 13), and in this case lag k=50 is chosen. This corresponds to a correlation timescale of 0.76 seconds.

[0069] The contour plots of r(50) and 5-second end current (μA) for this data are seen in FIGS. 14 and 15 respectively. Clearly, there exists a portion of the plasma glucose/haematocrit range where a contrary slope to that of the 5-second end current is seen. This dissimilarity between the two contour plots may therefore be exploited in much the same fashion as described above in connection with the simulated system of FIGS. 3-9.

[0070] FIG. 16 shows the result of using the two contour plots of FIGS. 14 and 15 to obtain a plot of mean percentage bias against haematocrit, for a 500 mg/dL plasma glucose concentration. In comparison to the plot of uncorrected data of FIG. 11, it is clear that the haematocrit sensitivity has been reduced by two thirds, thereby enhancing the accuracy of the glucose estimation.

[0071] Whilst described in connection with specific embodiments, it is to be understood that the contemplated embodiments are not limited to those described, and that alterations, modifications, and variations of these embodiments may be carried out by the skilled person without departing from the scope of the contemplated embodiments. For instance, whilst described primarily in the context of determining parameters of a test fluid, with particular reference to medical devices for measuring glucose in people with diabetes, the contemplated improvements may equally well be used in other fields, for example in health and fitness, food, drink, bio-security applications, environmental sample monitoring, veterinary devices, etc. Thus, instead of using a meter as used in electrochemical assays, it is envisaged that the method could be used with general scientific apparatus suitable for fluid samples.

[0072] Furthermore, whilst primarily described in the context of its use with electrochemical test strips, the contemplated improvements may extend to other electrochemical devices, such as wearable devices that actively acquire a fluid sample (such as interstitial fluid) from a user and cause an electrochemical reaction to occur with the sample. Examples of such are continuous (or semi-continuous) glucose monitoring devices used for controlling glucose concentrations (and insulin dosing) by users with diabetes.