Manufacturing and testing of multivariate analysis models in the measurement of analytes using spectroscopic techniques

Abstract

A method and apparatus for testing a multivariate model for generating analyte concentration values ​​using spectroscopic techniques, the method comprising: a) providing a multivariate model for generating output values ​​of analyte concentration based on spectroscopic measurements; b) inputting spectra of known concentrations of an analyte to be measured into the model; c) inputting spectra of interfering substances; and d) processing the received spectra of the analyte to be measured and the interfering substances to generate output values ​​of the analyte to be measured using the model. A method for generating such a multivariate analysis model, comprising spectrally spiking the input spectra to the model, is further described.

Description

Technical Field

[0001] The present invention relates to methods and devices for testing multivariate models used for non-invasive measurement of analyte concentration using spectroscopy.

Background Art

[0002] In embodiments, the method can be used to investigate the effect of the presence of a particular substance on the performance of a model when quantifying an analyte. Here, the multivariate model can be defined as a processing tool that establishes a connection between a plurality of input variables and output variables.

[0003] For example, the model can be configured to receive as input several variables that can be measured or determined, and then process these and generate as output a value for the interstitial fluid glucose concentration or blood glucose concentration.

[0004] Diabetes mellitus in its various forms is affecting ever more individuals and placing an excessive burden on national healthcare budgets. An estimated 415 million people worldwide are suffering from diabetes in (2015), and this number is predicted to increase to 642 million by 2040.

[0005] For controlling treatment, self-monitoring of blood glucose is recommended, which is usually performed using an invasive finger-prick method. In type 1 diabetes patients, insulin administration is often based on blood glucose measurements 4 to 6 times a day. For these reasons, developing a truly non-invasive technique for measuring the blood glucose level of diabetic patients has been a long-term goal.

[0006] Current clinical trends are favorable for implantable electrochemical sensors that enable continuous glucose monitoring in a minimally invasive manner. However, skin puncture is still required, which is associated with discomfort for the user and an increased risk of infection. These sensors also have a biocompatibility problem that limits their lifespan to a few weeks.

[0007] For decades, the goal has been to develop non-invasive techniques for measuring blood glucose levels in diabetic patients, but a practical solution for general use has yet to be developed. Most approaches have been based on the optical measurement of glucose in tissues such as skin. Among these, spectroscopic techniques such as fluorescence, absorbance, and Raman have attracted considerable attention. Inelastic Raman scattering, despite being a weak process and therefore yielding an insufficient signal, is an attractive spectroscopic choice for measuring glucose in the user's skin, and indeed the concentration of other analytes, due to several factors. These factors include high chemical specificity, minimal interference from tissue water content, and only very little fluorescence background. These factors make this technique one of the most promising candidates for non-invasive glucose monitoring.

[0008] Since the first feasibility study of blood glucose measurement using near-infrared Raman spectroscopy in 1997, several groups have demonstrated the basic effectiveness of this technique through quantitative measurement of glucose levels in vivo. However, these reports can be considered merely proof of concept, in that all measurements were performed in a controlled environment, and the predictive ability of the calibration model was evaluated solely by cross-validation.

[0009] In previous publications and patent applications, the applicant has described the design and development of a benchtop confocal near-infrared Raman spectrometer for intermittent glucose measurement. This spectrometer utilizes the principle of critical depth Raman spectroscopy, with measurements obtained from interstitial fluid within a defined area of ​​skin. In contrast to previous techniques that also utilize confocal instruments to probe biological portions of skin, the applicant's research is the first of its kind to systematically study the relationship between probing depth and the expected performance of a Raman-based glucomometer, and is therefore noteworthy for enabling the definition of the critical depth at which the Raman signal should be acquired.

[0010] The applicant's International Patent Publication No. WO2011 / 83111 (granted in many jurisdictions) describes a method and apparatus for non-invasive in vivo measurement of glucose present in interstitial fluid in skin by Raman spectroscopy. In particular, the apparatus describes an apparatus for non-invasive in vivo measurement of glucose present in interstitial fluid in the skin of a subject by Raman spectroscopy, comprising a light source, an optical component defining an optical path from the light source to the measurement position, a photodetector unit, an optical component defining a return optical path of Raman scattered light from the measurement position to the photodetector unit, and a skin engagement member having a distal surface for defining the position of the optical component defining the return optical path with respect to the surface of the skin in use, wherein the optical component defining the return optical path of Raman scattered light selectively transmits light scattered from near the measurement position to the photodetector unit such that at least 50% of the Raman scattered light received by the photodetector unit occurs beyond the distal surface of the skin engagement member to a depth of 60 to 400 μm.

[0011] The applicant's concurrently pending and granted patent portfolio describes several devices for use in determining blood glucose concentration using Raman spectroscopy.

[0012] To achieve practical utility in non-invasive glucose monitors (NIGMs), it is preferably necessary that several parameters meet user expectations. These include, for example, accuracy, cost, size, ease of use, calibration requirements, and calibration stability.

[0013] Calibration stability has remained a unique selling point for the most successful continuous glucose monitors, while more conventional finger-prick devices have sought to reduce the cost per measurement. Ultimately, NIGM technology is expected to surpass all other technologies in terms of cost per measurement because it requires no consumables, although this can only be truly achieved if calibration requirements are low or nonexistent.

[0014] U.S. Patent No. 7,756,558, U.S. Patent Publication No. 2022 / 317014, U.S. Patent Publication No. 2006 / 167350, U.S. Patent Publication No. 2009 / 268,203, European Patent No. 2,498,092, U.S. Patent No. 8,914,312, and U.S. Patent Publication No. 2021 / 215610 generally disclose systems and methods relating to multivariate data analysis and Raman spectroscopy for the use of spiking in testing. [Prior art documents] [Patent Documents]

[0015] [Patent Document 1] International Patent Application Publication No. WO2011 / 83111 Specification [Patent Document 2] U.S. Patent No. 7756558 [Patent Document 3] U.S. Patent Application Publication No. 2022 / 317014 [Patent Document 4] U.S. Patent Application Publication No. 2006 / 167350 [Patent Document 5] U.S. Patent Application Publication No. 2009 / 268,203 [Patent Document 6] European Patent No. 2498092 [Patent Document 7] U.S. Patent No. 8914312 [Patent Document 8] U.S. Patent Application Publication No. 2021 / 215610 [Overview of the project] [Problems that the invention aims to solve]

[0016] It is desirable to ensure that the use of multivariate models is stable and that the output does not experience unacceptable fluctuations depending on interfering substances. [Means for solving the problem]

[0017] According to a first aspect of the present invention, there is provided a method of testing a multivariate model for generating a value of analyte concentration using spectroscopy, the method comprising inputting a spectrum of an interferent and processing the spectrum in combination with a spectrum of the analyte to be measured to generate an output value of the analyte measured using the model.

[0018] The method preferably comprises: (a) providing a multivariate model for generating an output value of analyte concentration based on spectroscopic measurements; (b) inputting into the model a spectrum of a known concentration of the analyte to be measured; (c) inputting a spectrum of the interferent; and (d) processing the received spectra of the analyte and interferent to be measured and generating an output value of the analyte measured using the model.

[0019] There is provided a method that enables a multivariate model to be tested, which further enables determination of whether the interferent in question may be contraindicated for use in combination with the model. A technical advantage is provided in that a simple and robust method of testing the multivariate model is achieved, thereby enabling identification of contraindicated interferents.

[0020] In one example, the method comprises repeating steps (a) to (d) to generate an indicator of the effect of the spectrum of the interferent on the determined value of the analyte to be measured.

[0021] In one example, the analyte to be measured is glucose.

[0022] In one example, the interferent is a topical substance outside the user's skin.

[0023] In one example, the interferent is an interstitial substance found in the user's interstitial fluid.

[0024] In one example, the interferent is an interstitial substance derived from outside the user's body.

[0025] In one example, the method involves updating the model in accordance with the generated output values ​​of the analyte being measured.

[0026] A second aspect of the present invention provides a method for generating a multivariate analysis model for a device for non-invasive measurement of an analyte using spectroscopic techniques, the method comprising: spectrally spiking an input spectrum to a model during testing of the model; and determining an output from the model in response to the spectrally spiked input.

[0027] A third aspect of the present invention provides a device for non-invasive measurement of analyte concentration using spectroscopic techniques, the device comprising a light source and a spectrometer for receiving a generated spectrum for measurement of analyte concentration, the device receiving the spectrum and inputting the spectrum into a measurement model, the model being tested using the method of the first aspect of the present invention.

[0028] A fourth aspect of the present invention provides a device for testing a model for non-invasive measurement of analyte concentration using spectroscopic techniques, the device configured to (a) receive spectra of known concentrations of the analyte to be measured and spectra of interfering substances, and (b) process the received spectra of the analyte to be measured and interfering substances to generate an output value of the analyte to be measured using the model.

[0029] Accordingly, a device for carrying out the method of the first aspect of the present invention is provided. Typically, the device can be implemented using a computer processor configured to receive input spectra or a single superimposed spectrum of the analyte to be measured and interfering substance and process it as described herein.

[0030] Herein, embodiments of the present invention will be described in detail with reference to the accompanying drawings. [Brief explanation of the drawing]

[0031] [Figure 1] This is a schematic diagram of an optical configuration for non-invasive measurement of glucose concentration. [Figure 2] This is a schematic flowchart illustrating the steps in the spectral spiking process for device and / or model testing. [Figure 3a] This example illustrates measurement bias induced by urea-induced spectral spiking using a glucose measurement model. [Figure 3b] This example illustrates measurement bias induced by urea spectral spiking using a glucose measurement model. The test concentration of urea in this example is 6.5 mM. [Figure 4] The quantitative Raman spectra of glucose and urea at a concentration of 5 mM are shown. [Figure 5] Examples of the thenar eminence spectra and the same spectra spiked with 5 mM glucose and urea concentrations, respectively, are shown. [Figure 6] The figure shows the average change in glucose measurements when the thenar eminence spectrum is spiked with spectra of glucose (i.e., the analyte of interest) and urea at various concentrations. [Figure 7A-C] Examples corresponding to the figures in Figures 4, 5, and 6 are shown, where different interfering substance spectra are used as spiking spectra (alanine in the case of Figure 7). [Figure 8A-C] Examples corresponding to the figures in Figures 4, 5, and 6 are shown where different interfering substance spectra are used as spiking spectra (ibuprofen in the case of Figure 8). [Modes for carrying out the invention]

[0032] Figure 1 is a schematic diagram of a non-limiting example of an optical configuration for non-invasive measurement of glucose concentration using Raman spectroscopy.

[0033] The foundation of a spectrometer is a light source used to illuminate a sample, such as a laser. Light from the light source (incident light) interacts with the sample, often resulting in changes to the light that is transmitted through the sample, emitted by the sample, reflected by the sample, and / or scattered by the sample. By collecting the modified light and analyzing its spectral distribution, information about the interaction between the incident light and the molecular sample can be obtained, and therefore, information about the molecular components can be obtained.

[0034] The spectral distribution is typically measured using a spectrometer. A spectrometer is an optical device that works by separating a light beam directed into different frequency components and then measuring the intensity of these components using, for example, a CCD detector, CCD array, or photodiode.

[0035] Figure 1 shows a first embodiment of an optical configuration that may be contained within the optical probe 201. The probe comprises a first optical fiber 203 for guiding light into the optical probe 201. The light source is typically a laser, and the optical configuration is shown only as an example of an optical configuration suitable for use with the calibration method described herein.

[0036] Exiting the first fiber 203, the incident light 205 is collimated using the first lens 207 and optically filtered by passing through the first filter 209, which blocks any percentage of frequencies / wavelengths outside the laser frequency / wavelength from 0 to 100. Blocking frequencies outside the laser frequency ensures, for example, that fluorescence generated inside the first fiber 203 is removed from the incident light 205. The first filter 209 may also block any percentage of the laser frequency from 0 to 100. This is advantageous when the intensity of the incident light 205 is too high for the requirements of the sample. The first filter 209 is preferably a bandpass filter, a notch filter, an edge filter, etc.

[0037] The optical probe 201 further comprises a dichroic mirror 211 that reflects or transmits any percentage of light between 0 and 100, the percentage of reflected and transmitted light depending on the coating of the dichroic mirror 211, the angle at which the light strikes the dichroic mirror 211, and the frequency of the light. The dichroic mirror 211 can be coated, for example, to reflect the highest percentage of incident light 205 when the dichroic mirror 211 is positioned at a given angle with respect to the direction of the incident light 205. Thus, changing the angle between the dichroic mirror 211 and the incident light 205 reduces the percentage of incident light 205 reflected by the dichroic mirror 211.

[0038] In this example, most of the incident light 205 is reflected by the dichroic mirror 211 and focused into the inside of the subject's skin 213 by the second lens 215. The focal point 217 of the incident light 205 is defined by the focal length 218 of the second lens 215 and the distance to the distal window 219 of the lens, particularly its distal surface that engages with the skin during use. The second lens 215 is preferably convex, but may be aspherical or planar.

[0039] This method and device relate to a system that can understand the influence of interfering substances on results provided by a non-invasive glucose monitoring system and allow them in a calibration model.

[0040] Interference or interfering substance is defined as a physical condition or chemical substance that may affect the performance, such as the safety and / or effectiveness, of a measuring device.

[0041] Interference generally falls into the categories of interference conditions and interfering substances. Interference conditions include both device-dependent conditions that affect the device, which can be introduced by the device itself or the surrounding environment, such as temperature, humidity, and depth of field. User-dependent interference conditions relate to physical conditions introduced by the user, such as disease or skin phototype.

[0042] The interfering substance may be either a topical substance administered externally to the skin, some of which may penetrate into the skin, or an interstitial substance found in interstitial fluid (ISF), including both naturally occurring compounds in ISF and compounds of exogenous origin.

[0043] This method and device primarily relates to potential interfering substances, including the presence or, in some cases, the absence of chemicals on or within a subject's body, that may cause a device used to non-invasively measure glucose concentration to measure falsely high or falsely low readings.

[0044] In systems relying on common electrochemical glucose measurement techniques, univariate regression is used to convert current into glucose concentration. The univariate approach is highly sensitive to spurious chemistry. This requires that glucose specificity and robustness issues be resolved at the hardware level to prevent the generation of interfering signals. For example, the problem of interfering substances can be mitigated by selecting enzymes particularly sensitive to glucose and by coating electrodes with partially selective membranes that limit the drugs that can pass through the membrane.

[0045] According to the ISO 15197 standard for blood glucose self-monitoring systems, a substance is considered an interfering substance if it meets any of the following criteria:

[0046] 1. If the glucose concentration is less than 5.56 mmol / l, the mean difference between the test sample and the control sample must exceed 0.55 mmol / l, and 2. If the glucose concentration exceeds 5.56 mmol / l, the mean difference between the test sample and the control sample must exceed 10%.

[0047] It is recognized that the criteria for classifying substances as interfering substances can differ among regulatory authorities. For example, the U.S. Food and Drug Administration expects interference assessments to follow the guidance document "Self-Monitoring Blood Glucose Test Systems for Over-the-Counter Use."

[0048] This system, which uses non-invasive spectroscopic techniques to measure the concentration of analytes such as ISF glucose, employs multivariate analysis. Multivariate analysis techniques differ fundamentally from univariate techniques. Multivariate analysis techniques utilize information from multiple variables simultaneously. This allows for more sophisticated calibration and improved robustness by enabling the sensor to analyze and consider multiple sources of variation. For example, by utilizing multiple variables simultaneously, the multivariate sensing principle can provide improved robustness in the presence of environmental disturbances and other sources of variation, resulting in more reliable results.

[0049] The multivariate detection principle can provide more robust and accurate results by simultaneously analyzing multiple variables, especially in large datasets containing many naturally occurring changes in the environment and samples. This allows the model to better learn and understand how other substances affect many variables, thus providing more reliable results. Therefore, the multivariate detection principle can take into account many causes that may affect the signal and can distinguish between changes related to the analyte in question and changes related to other molecules in the sample or environment.

[0050] For non-invasive glucose measurements, the hardware is typically a confocal Raman spectrometer designed to acquire Raman spectra of the thenar eminence skin. The thenar spectrum contains general information about the skin and ISF components, including glucose. However, only a small portion of the signal arises from the interaction between the incident laser light and glucose molecules in the ISF. The multivariate nature of the Raman spectrum allows for the separation and quantification of the glucose signal by multivariate regression techniques.

[0051] As is well known, multivariate regression models are a common mathematical tool for establishing connections between multiple input and output variables. Models are generally achieved and designed by training them against paired spectral and baseline glucose concentrations, which can be collected over several days / months, typically for many subjects and devices. An example of such a configuration is described in our concurrently pending applications, UK Patent Application Publication No. 2116869.5 and International Application No. PCT / EP2022 / 078431.

[0052] Using such a multivariate regression model means that spectral variations due to biological, environmental, and device variations can be distinguished from glucose-induced variations. In other words, the model is adapted to measure glucose via human skin measurements.

[0053] Referring to Figure 2, the method for spectral spiking is described here.

[0054] First, glucose model 2 is provided.

[0055] A thenar spectrum 4 is provided as input to glucose model 2. An initial glucose measurement (G) 6 is derived from the glucose model in accordance with the input thenar spectrum 4.

[0056] Next, the thenar eminence spectrum 4 is spiked, or perturbed, by the superposition of the Raman spectra 8 of potential interfering substances. Thus, the spiked spectrum 10 is produced.

[0057] Next, the spiked spectrum 10 is similarly input into glucose model 2, yielding a spiked glucose measurement (G*) 12.

[0058] The influence of potential interfering substances is assessed by calculating the induction bias in the measurement. The influence can be determined simply based on the difference between G* and G, i.e., Δ = |G* - G|.

[0059] This spectral spiking is performed on multiple thenar spectra from different subjects and different reference glucose concentrations, thus yielding a distribution of Δ values ​​for a given potential interfering substance. The worst-case Δ value can be compared to clinically acceptable values, which may typically be included in known standards such as ISO 15197.

[0060] While individual spikings and readings can be used to determine the effect of interfering material spectra on the desired output of a particular model, in practice, a large number of spectra are spiked and data collected.

[0061] The data collected by the applicant included thenar spectra from 11,046 measurement sessions from 139 subjects, which were used in the spiking procedure. See Figure 3 for an example of the Δ distribution.

[0062] Figures 3a and 3b show calculations of urea-induced spiking using a glucose model, where Δ is presented in a box plot with an upper whisker defining the worst-case bias compared to the ISO 15197 threshold, shown as a dashed horizontal line.

[0063] The upper whisker is defined as follows in this example.

number

[0064] Here, Q3 is the third quartile, (Q3-Q1) is the interquartile range, and M is a medcouple parameter that robustly quantifies the skewness of the distribution. It is worth noting that the worst-case bias can be defined in several ways. For example, ISO 15197 or FDA guidance documents recommend comparing the mean bias to a defined threshold.

[0065] The concept of spectral spiking relies on the fact that spontaneous Raman scattering is a linear process, which means that the intensities of Raman signals from a multi-component sample can be read as follows:

number

[0066] In the above formula, I レーザー This represents the intensity of the irradiated laser, σ i This is the Raman scattering cross-section, i.e., the scattering intensity. c i This is the concentration of the i-th component.

[0067] Therefore, the above equation emphasizes the superposition of contributions from individual components.

[0068] Therefore, when a potential interfering substance is introduced into the sample being measured, the resulting Raman signal will superimpose on the original signal. This is the fundamental principle behind spectral spiking, where the spectrum of the potential interfering substance is superimposed on the thenar spectrum as if it were present in the skin or interstitial fluid.

[0069] For proper spiking of the thenar eminence spectrum, it is necessary to know the Raman spectrum, scattering intensity, and physiological concentration of potential interfering substances. Physiological concentrations can generally be found in scientific literature or databases, and quantitative Raman spectra can be measured in the laboratory by properly dissolving the potential interfering substances.

[0070] The superposition principle and glucose specificity of the apparatus used in this example are illustrated by spiking the thenar spectra with various concentrations of glucose and, for example, urea.

[0071] As shown in Figure 4, glucose and urea have very similar Raman scattering intensities. When they are superimposed on the thenar spectrum at physiological concentrations, they represent perturbations of the original spectrum. See, for example, Figure 5. Spectral variations can be seen in the magnified points for glucose and urea.

[0072] However, the situation is quite different when the spiked spectrum is input into a glucose model, such as the one used in the process in Figure 2, because, as the name suggests, the model responds strongly (as expected) to glucose spiking but hardly changes to urea spiking. The sensitivity to glucose (and insensitivity to urea) is illustrated in Figure 6, which illustrates the effect on the mean bias of glucose measurements when glucose or urea spectra are added to the thenar spectrum to represent interfering substances at various concentrations. The spiking concentrations ranged from -5 to 5 mM, and the results shown in Figure 6 are based on 11,046 measurement sessions from 139 subjects. Error bars represent the standard deviation.

[0073] As can be seen from the figure, spiking the input spectrum results in a nearly linear change in glucose measurements when the spiking itself is glucose. The output slope for glucose has a gradient very close to 1 (0.96). However, the gradient for urea spectral spiking is -0.04, i.e., virtually negligible. The gradient represents a measure of sensitivity, and the fact that its glucose gradient is around 1 supports the usefulness of spectral spiking for identifying potential interference. In other words, urea does not cause any significant variation in the expected results, which demonstrates the robustness of the model for use when measuring a patient's glucose levels.

[0074] Therefore, the applicants recognized that the introduction of spectral spiking works well as a method for determining the robustness of a model for non-invasive measurement of glucose concentration using multivariate analysis.

[0075] Figures 7A–7C and 8A–8C show the results of the multivariate spectral spiking process of the glucose prediction model using the spectra of alanine and ibuprofen. In both cases, the relative gradient shift is shown compared to the gradient shift caused by glucose. As can be seen from the figures, different perturbations of the thenar spectra have different effects on glucose measurements. In the examples in Figures 7 and 8, the calibration model is (as expected) more sensitive to glucose spiking than spiking from the selected tested interfering substances, alanine and ibuprofen, in all cases.

[0076] The above describes a method for testing models for the non-invasive measurement of analyte concentrations using spectroscopic techniques. As described above, the method involves providing an initial multivariate model (to be tested) for generating output values ​​of analyte concentrations based on spectroscopic measurements. Typically, this model is a multivariate regression model for determining the concentration of glucose in blood or interstitial fluid.

[0077] The method, in summary, involves inputting the spectra of the analyte to be measured at known concentrations and the spectra of interfering substances into a multivariate model. Typically, a superimposed spectrum can be provided as input consisting of a superposition of the spectrum of the analyte to be measured and the spectrum of the interfering substance. The received spectrum (or superimposed spectrum) is processed to generate an output value of the analyte to be measured using the multivariate model. The effects of spectral perturbations or spiking can be observed in the change in the output value, preferably by repeating the process for various consultation levels of the analyte to be measured and / or interfering substance.

[0078] This process works very well in identifying potential interfering elements and ensuring that the model is robust enough to meet, for example, certain standards.

[0079] Embodiments of the present invention are described with particular reference to illustrative examples. However, it will be understood that modifications and alterations may be made to the described examples within the scope of the invention.

Claims

1. A method for testing a multivariate model for generating analyte concentration values ​​using spectroscopic techniques, wherein the method is: Inputting the spectrum of an interfering substance and processing the spectrum in combination with the spectrum of the analyte to be measured in order to generate an output value of the analyte measured using the model. Methods that include...

2. a) To provide a multivariate model for generating output values ​​of analyte concentrations based on spectral measurements, b) Inputting the spectrum of a known concentration of the analyte to be measured into the model, c) Processing the received spectra of the analyte to be measured and the interfering substance to generate the output value of the analyte to be measured. The method according to claim 1, including the method described in claim 1.

3. The method according to claim 2, comprising repeating steps (a) to (c) to generate an index of the spectral influence of the interfering substance on the determined value of the analyte to be measured.

4. The method according to any one of claims 1 to 3, wherein the analyte to be measured is glucose.

5. The method according to any one of claims 1 to 4, wherein the interfering substance is a topical substance applied to the outside of the user's skin.

6. The method according to any one of claims 1 to 5, wherein the interfering substance is interstitial material found in the user's interstitial fluid.

7. The method according to claim 5, wherein the interfering substance is a compound derived from outside the user's body.

8. The method according to any one of claims 1 to 7, comprising updating the model in accordance with the generated output value of the analyte being measured.

9. A method for generating a multivariate analysis model for a device for non-invasive measurement of an analyte using spectroscopic techniques, wherein the method is: During testing of the aforementioned model, the input spectrum to the model is spectrally spiked, Determining the output from the model in response to the spectrally spiked input and Methods that include...

10. The method according to any one of claims 1 to 9, wherein the spectra of the analyte to be measured at known concentrations and the spectra of the interfering substance are provided as a single superimposed spectrum on the model to be tested.

11. A device for non-invasive measurement of analyte concentration using spectroscopic techniques, wherein the device is Light source and A spectrometer for receiving the generated spectrum for measuring the concentration of the analyte It has, The device receives the spectrum, inputs the spectrum into a measurement model, and the model is tested using the method according to any one of claims 1 to 10.

12. A device for testing a model for non-invasive measurement of analyte concentration using spectroscopic techniques, wherein the device is a) Obtain the spectrum of the analyte to be measured at a known concentration and the spectrum of interfering substances, b) Process the received spectra of the analyte to be measured and the interfering substance, and generate an output value of the analyte to be measured using the model. A device configured in such a way.

13. A method for testing a multivariate model for generating analyte concentration values ​​using spectroscopic techniques, wherein the method is: (a) To provide a multivariate model for generating output values ​​of analyte concentrations based on spectral measurements, (b) Inputting the spectrum of a known concentration of the analyte to be measured into the model, (c) Inputting the spectrum of the interfering substance, (d) Processing the received spectra of the analyte to be measured and the interfering substance, and generating an output value of the analyte to be measured using the model. Methods that include...

Images