HK1019635B - Method and apparatus for multi-spectral analysis in noninvasive nir spectroscopy - Google Patents

Description

Multi-spectral analysis method and device for nondestructive near-infrared spectrometer

Technical Field

The present invention relates to a method and apparatus for determining the concentration of a target analyte in a sample using multi-spectral analysis. The invention can be applied to various chemical analyses, in particular to the nondestructive spectrophotometric analysis of the analyte in blood.

Background

Measuring the concentration of various components in blood finds application in various procedures for diagnosing and treating conditions and diseases of the human body. Measuring glucose in blood is an important application. In particular, while it is desirable for diabetic patients to monitor the concentration of glucose in the blood on a regular basis, it is often necessary or desirable for insulin-dependent or type I diabetic patients to monitor the glucose in the blood multiple times a day. In addition, the measurement of cholesterol concentration in blood provides important information for the treatment or prevention of coronary artery disease, and the measurement of other organic analytes in blood, such as bilirubin and ethanol, is also important in various diagnostics.

One of the most accurate and widely used methods of obtaining the concentration of an analyte in blood is to draw a blood sample from a patient and then analyze the blood sample in a laboratory using highly accurate and sensitive assay techniques or using less accurate self-test methods. In particular, conventional methods of monitoring the glucose level in blood require each test to take a blood sample from a diabetic patient (e.g., by a needle), and read the glucose level using a glucose meter (a spectrophotometer that reads the glucose concentration) or colorimetric calibration. Such invasive blood draws can be painful and tiring for the diabetic, who may have a potential for infection, especially if they require frequent testing. These considerations lead to the desire of the diabetic to cancel the monitoring process.

Accordingly, there is a need for a simple and accurate method and apparatus for non-invasive measurement of analyte concentrations in blood, particularly for monitoring glucose levels in the blood of diabetic patients. One approach to this problem has been to use conventional Near Infrared (NIR) analysis, where absorption measurements at one or more specific wavelengths are used to extract analyte-specific information from a given sample.

The near infrared absorption spectrum of a liquid sample contains a large amount of information about the various organic components of the sample. In particular, vibrational, rotational and stretching energy associated with organic molecular structures (e.g., carbon-carbon, carbon-hydrogen, carbon-nitrogen and nitrogen-hydrogen chemical bonds) can produce perturbations in the near infrared spectral region that can be detected, which are related to the concentration of such organic components present in the sample. However, in complex sample matrices, the near infrared spectrum also contains some amount of interference, due in part to structural similarities in the analytes, the relative levels of analyte concentrations, and the interference relationships between the analytes and the electronic and chemical "noise" amplitudes inherent in the particular system. This interference reduces the effectiveness and accuracy of measurements obtained using near infrared spectroscopy to determine the analyte concentration of the liquid sample. However, a number of non-invasive blood analyte determination results have been described for near infrared devices and methods.

U.S. patent 5,360,004 to Purdy et al describes a method and apparatus for determining the concentration of an analyte in blood wherein a portion of the body is irradiated with radiation containing two or more distinct bands of continuous wavelength incident radiation. Purdy et al emphasize the filtering technique and two particularly radiation-blocking peaks appear at approximately 1440 and 1935nm in the NIR absorption spectrum of water. This selective blocking is used to avoid heating effects due to absorption of radiation by water in the irradiated body.

In contrast, U.S. patent 5,267,152 to Yang et al describes an apparatus and technique for non-invasively measuring glucose concentration in blood using only one infrared spectral band that includes NIR water absorption peaks (e.g., "transmission window for water" that includes these wavelengths between 1300 and 1900 nm). Optically controlled light is directed onto a tissue source and light is collected by an integrating sphere. The collected light is analyzed and the glucose concentration in the blood is calculated using the stored reference calibration curve.

Apparatus for determining the concentration of an analyte in a complex sample is also described.

For example, U.S. patent 5,242,602 to Richardson et al describes a method for detecting various active or inactive water treatment components by analysis of aqueous systems. These methods involve determining the absorption or emission spectra of components over a range of 200 to 2500nm and applying a chemometric algorithm to extract a portion of the spectral data obtained, quantitatively representing various properties.

U.S. patent 5,252,829 to Nygaard et al describes a method and apparatus for measuring the concentration of urea in a milk sample using infrared attenuation measurement techniques. Multivariate techniques are performed to determine the spectral contributions of the known components using partial least squares algorithms, principal component regression, multiple linear regression, or artificial neural network learning. Scaling is performed by accounting for component contributions that block the analyte signal of interest. Thus, Nygaard et al describe a technique for measuring infrared attenuation of multiple analytes and compensating for background analyte effects to obtain higher accuracy measurements.

U.S. patent No. 4,975,581 to Robinson et al describes a method and apparatus for determining the concentration of an analyte in a biological sample based on a comparison of the known analyte concentration to the absorption of infrared energy between the samples (i.e., the difference between the absorption at several wavelengths). The comparison is performed using partial least squares analysis or other multivariate techniques.

U.S. patent No. 4,882,492 to Schlager describes a method and apparatus for non-invasively determining the concentration of an analyte in blood. Modulated infrared radiation is directed onto a tissue sample (e.g., an earlobe) and the infrared radiation either passes through the tissue or falls on the skin surface where the spectrum of the infrared radiation is altered by the target analyte (glucose). The spectrally modified radiation is then split into a beam, one beam being directed into the negative correlation cell and the other beam being directed into the reference cell. The intensity of the radiation passing through the two cells is compared to determine the concentration of the analyte in the sample.

Us patent 4,306,152 to Ross et al describes an optical fluid analyzer designed to minimize the effect of background absorption (i.e., the total or baseline optical absorption of a liquid sample) on the accuracy of measurements in turbid samples or liquid samples that are otherwise difficult to analyze. The device measures the optical signal at the characteristic optical absorption of the sample component of interest and another signal at a wavelength selected to be near background absorption, and then subtracts to reduce the background component of the analyte-dependent signal.

The accuracy of the information obtained using the above-described methods and devices is limited by the spectral interference caused by the background, i.e., the non-analyte and other sample components that also have absorption spectra in the near infrared range. A certain degree of background noise represents an inherent limitation of the system, especially when the analyte content is low. In view of this limitation, attempts have been made to improve the signal-to-noise ratio, for example by avoiding the water absorption peaks to enable higher radiation intensities to be employed, by reducing the amount of spectral information being analyzed, or by employing subtraction or compensation techniques based on approximate background absorption. Despite the improvements made by these techniques, there remains a need to provide a method and apparatus that can more accurately determine the concentration of an analyte in a liquid matrix, and in particular monitor the glucose level in blood.

Specification

Accordingly, it is a primary object of the present invention to address the needs described in the prior art by providing a method for determining the concentration of an analyte present in a sample containing a variable background matrix and possibly substantial constituent interference. The method accounts for structural similarity of the various components present in the sample, the relative magnitude of analyte concentrations, and spectral interference caused by differences in the various sample components and instruments.

The method of the invention generally comprises: (1) finding several different non-overlapping wavelength regions in the near infrared range, which have a high correlation with the concentration of the analyte; (2) illuminating the sample with incident radiation containing these wavelength regions, thereby obtaining spectrally attenuated radiation due to interaction with sample components; (3) detecting the spectrally attenuated radiation; (4) measuring the intensity of the spectrally attenuated radiation at a wavelength in the non-overlapping wavelength region; (5) the measurement results are corrected to obtain a value indicative of the analyte concentration.

It is a further object of the present invention to provide a spectrophotometer device for determining the concentration of an analyte present in a sample containing a variable background matrix and substantial component interference. The device is employed in multispectral analysis to obtain spectral information containing analyte-specific signals and signals related to instrument background noise and interference spectral information. Chemometric techniques are employed to construct filter elements that can enhance the correlation of analyte-specific information with analyte concentration and to derive system algorithms that can determine analyte concentration values.

In one aspect of the invention, a device is provided that includes a dedicated optical delivery unit that enhances the correlation of analyte-specific information with analyte concentration. The dedicated optical delivery unit comprises a positive correlation filter adapted to selectively highlight wavelengths having a high correlation with the selected analyte concentration. The wavelength to be highlighted is interfaced with a device that receives the information and converts the information into a signal representative of the intensity of the wavelength.

Brief Description of Drawings

FIG. 1 is a diagrammatic representation of an apparatus constructed in accordance with the present invention.

FIG. 2 is a diagrammatic representation of an associated spectrometer arrangement constructed in accordance with the present invention.

Figure 3 is a scan taken over a number of times during an in vivo glucose tolerance study.

FIG. 4 shows a graph of the results obtained by non-invasive measurement of blood glucose concentration using the method of the present invention.

Modes for carrying out the invention

Before the present invention is described in detail, it is to be understood that this invention is not limited to the particular components of the devices or methods described, as such may, of course, vary. It is also to be understood that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting. It must be noted that, as used in the specification and the claims, the singular form may include plural referents unless the context clearly dictates otherwise. Thus, for example, reference to "an analyte" also includes mixtures of analytes, reference to "an optical delivery unit" also includes two or more optical delivery units, reference to "a device that reflectively transmits radiation" also includes two or more devices, reference to "a wavelength" includes two or more wavelengths, reference to "a stoichiometric algorithm" includes two or more algorithms, and the like.

In this specification and the claims which follow, reference will be made to a number of terms which shall have the following meanings:

"chemometrics" refers to mathematical, statistical, and pattern recognition techniques employed in chemical analysis applications. See Brown et al (1990) anal. chem.62: 84-101. Chemometrics are used here to develop and utilize non-invasive diagnostic instruments employing advanced signal processing and calibration techniques. Signal processing is employed to improve the accessibility of important physical information in the analyte signal. Examples of signal processing techniques include fourier transforms, first and second derivatives, digital or adaptive filtering.

In chemometrics, "calibration" refers to the processing of data measuring chemical concentrations for purposes of quantification. In particular, statistical scaling using chemometric methods can be used to extract specific information from a complex set of data. Such scaling methods include linear regression, multiple linear regression, partial linear regression, and principal component analysis. In other applications, scaling may be performed using artificial neural networks, intelligent algorithms, and rotational principal component analysis.

Instruments that detect information about one or more components of a complex chemical matrix necessarily rely on analytical algorithms (e.g., those derived using chemometrics) to reveal information specific to one or more chemical components. Chemometrics can be employed to compare unknown to calibrated standards and databases, provide advanced forms of group analysis, and extract features from unknown samples that can be used as statistical and mathematical model information.

"principal component analysis (PAC)" is a data reduction method that can be performed in applying chemometric techniques to spectroscopic measurements of chemical analytes in complex matrices. PAC is used to reduce the dimensionality of a large number of interrelated variables while maintaining information that distinguishes one component from another. This reduction is achieved by eigenvector transformation of the original set of interrelated variables (e.g., absorption spectra) into a small number of uncorrelated Principal Component (PC) variables representing most of the information in the original set of variables. The new set of variables is ordered such that its small number of first variables retains the most change that occurs in all of its original variables. See Jolliffe L.T. et al, "principal Components analysis," spring-Verlag, New York (1986). More specifically, each PC is a linear combination of all raw measured variables. The first is the vector in the direction of maximum variance of the observed variable. The next individual PCs were chosen to represent the largest deviation of the measured data and to be orthogonal to the previously calculated PCs. Thus, the individual PCs are arranged in descending order of importance.

The term "weighting factor" includes the weighting coefficients of partial least squares regression and/or principal components regression, or any constant derived from any statistical scale that can be used to calculate an unknown sample value (e.g., the concentration of an analyte). The "wavelength weighting factor" is one embodiment of a weighting constant employed in constructing an optical filter device capable of highlighting specific wavelength information from spectral data. Wavelength specific information may be employed to determine a desired value (e.g., concentration of analyte) associated with a sample undergoing analysis. The wavelength weighting factors may be embodied as a particular filter density (e.g., neutral or wavelength specific), filter thickness, etc., which can all be determined using the statistical scaling techniques described above.

The term "optical transfer unit" includes any photosensitive element that partially absorbs incident radiation in the visible, ultraviolet, or infrared spectral regions, where partial absorption is wavelength selective. For the present invention, the optical transfer unit typically comprises an optical filter arrangement of absorption characteristics derived from partial least squares or principal component regression analysis. An optical filter device is used to selectively highlight wavelengths that are highly correlated with the concentration of the selected analyte. "highly correlated" or "closely correlated" refers to a quantitative correlation between absorption at a particular wavelength and a particular analyte concentration, where the correlation coefficient (r) for both variables is 0.9 or higher.

By "positive correlation filter" is meant an optical filter device whose absorption spectrum is sufficient to highlight specific wavelengths corresponding to target analytes but not other wavelengths of analytes that also absorb light. Thus, a positive correlation filter provides an optimal transfer function that is highly correlated with the concentration of the analyte in the sample being measured. An ideal positive correlation filter should be completely correlated with the target analyte (i.e., the correlation coefficient r should be +1.0) and not at all correlated with all other interfering absorbing analytes in a particular sample (r should be 0.0). Here, the synthesis of the positive optical filter is performed using chemometric techniques to determine the appropriate wavelength weighting factors.

"neutral density filter" refers to a standard optical filter device having a flat absorption spectrum. A neutral density filter may be employed in conjunction with an associated filter in the filter system to provide a weighting factor to attenuate the absorption of the analyte at the selected wavelength and to further improve the accuracy of the correlation provided by the system. The neutral density filter may have an absorption spectrum sufficient to attenuate radiation of all wavelengths in the range of interest equally.

As used herein, "aqueous medium" includes any substrate that consists of or contains water. Thus, an aqueous medium includes a medium having water as the major component, i.e., having a water content of at least about 50%, and water as the solvent but less than about 50%. The aqueous medium is specifically defined herein to include tissue cells of a mammal.

The term "blood analyte" refers to a blood component that is absorbable in the near infrared range, and measurement of blood components is useful in patient monitoring or health care.

As used herein, the term "near infrared" includes radiation in the spectral range of about 660nm to 3500nm, typically about 1050 to 2850nm, and more often about 1100 to 2500 nm.

The term "background absorption" refers to the entire or baseline optical absorption of an aqueous sample being analyzed, the absorption of a selected component at one or more characteristic wavelengths deviating from this background absorption to a degree indicative of the concentration of the selected component. When the background absorption benchmark is higher than the characteristic absorption of the selected component, such that a large number of interfering components are found in a complex aqueous medium, accurately measuring the small magnitude of the change in absorption at the characteristic wavelength of the component of interest requires the application of the chemometric techniques described herein. This is particularly true for applications where the total concentration of the component of interest is relatively low compared to aqueous media, such as in the measurement of analytes in blood.

General procedure

A spectrophotometric method for determining the concentration of an analyte in a liquid sample using near infrared radiation is provided. In order to obtain a set of measurements that can be used to determine analyte concentration with greater accuracy, the method of the present invention differs from previous techniques in that it utilizes all of the spectral information contained in the near infrared range.

The method comprises the following steps: (1) selecting several different non-overlapping near-infrared wavelength regions, wherein each wavelength region defines a spectral range; (2) illuminating the sample with near infrared light comprising a selected spectral range to obtain attenuated spectrally modified radiation; (3) collecting and measuring the intensity of spectrally attenuated radiation at one or more wavelengths contained within each selected spectral range; (4) these measurements are correlated to obtain a value indicative of the analyte concentration.

The spectral information obtained using this method may incorporate some mathematical transformations to obtain more accurate values of analyte concentration. For example, radiation absorption at a particular wavelength can be correlated with analyte structure and concentration using standard statistical techniques, such as Partial Least Squares (PLS) analysis, or Principal Component Regression (PCR) analysis. For example, Geladi et al (1986) Analica Chimica Acta 185: 1-17 describe PLS techniques. For a description of the PCR technique, reference may be made to Jolliffe L.T. Primary component analysis, Sprinter-Verlag, New York (1986).

Thus, in determining the concentration of an analyte in blood from a body tissue sample, one method involves selecting three non-overlapping wavelength regions from the near infrared range of 1100 to 3500nm, specifically a first wavelength region of 1100 to 1350nm, a second wavelength region of 1430 to 1450nm or 1930 to 1950nm, and a third wavelength region of 2000 to 2500nm, where each region defines a "spectral range". The first wavelength region contains wavelengths at which proteins and other cellular components exhibit predominant spectral activity, the second wavelength region is dominated by the absorption spectrum of water, and the third wavelength region contains wavelengths at which analyte organic molecules exhibit significant spectral activity. In regions where these components are not the predominant species, they also contribute to the absorption spectrum. The spectrally attenuated radiation obtained from each region thus contains a large amount of relevant information that must be reduced by statistical methods to obtain analyte-specific information.

The invention also relates to the use of signal processing to improve the accessibility of important physical information in an analysis signal. Therefore, the intensity value of the signal obtained at the specific wavelength can be processed, and the influence of instrument noise is reduced. The processed signal is then subjected to multivariate analysis using known statistical techniques.

The Principal Component Analysis (PCA) method of data reduction is the preferred method of reducing the dimensionality of a large number of related variables while preserving the information that distinguishes one component from another that is actually employed in the present invention. An original set of interrelated variables (e.g., absorption spectra) are eigenvector transformed into a small number of uncorrelated Principal Component (PC) variables representing most of the information in the original set of variables, and data reduction is possible using this eigenvector transform. The new set of variables ordered as its small number of first variables retains the most changes that occur in all of its original variables.

The principal component vector can be transformed by orthogonal rotation of the relative absorbance mean to obtain a relative value of the absorbance at a known wavelength and a wavelength that is attributable to the analyte. By doing this analysis of the information obtained for each of the three spectral regions, cross-correlating the principal component vectors through a linear algorithm, and removing the effect of interfering analytes by subtraction, the values obtained can be used in a system algorithm to determine the concentration of the analyte.

Multivariate techniques are employed to provide a model of the intensity of radiation at a particular wavelength in each spectral region as a function of analyte concentration in a particular sample matrix (e.g., body tissue). The model is constructed using two sets of exemplary measurements taken simultaneously, a first set of measurements, the "predicted set" comprising spectral data, such as intensity of radiation at a selected wavelength, and a second set of measurements, the "calibrated set" comprising analyte concentrations that have been determined using a destructive sampling technique with a higher degree of accuracy. This process is performed over a range of analyte concentrations, providing a set of calibration data and a set of prediction data.

The measurements obtained in the calibration and prediction groups are subjected to multivariate analysis, for example using a commercially available multivariate model development software program, to provide an initial model. The initial model is applied to the prediction data to derive an analyte concentration that can be compared to the value obtained by the invasive technique. By performing the above steps one after the other, a refined model is developed, which can be used to build a systematic algorithm for analyzing the data obtained by the method of the present invention.

The use of multivariate techniques as described above also enables the design of light sensitive elements, such as positive correlation filter systems, that enhance the correlation of the spectral information with analyte concentration. In particular, the solution obtained using multivariate analysis can be used to determine an optical parameter, such as an absorption characteristic, of the positive correlation filter system.

In the practice of the invention, non-analyte specific information for different non-overlapping spectral regions is also used, such as normalizing each spectral scan, subtracting background and baseline interference, or providing signal values for detecting inaccurate measurements.

In determining the analyte concentration of blood in a body tissue sample, measurements taken in the spectral range of about 1320-. By collecting and measuring the radiation intensity in this range, the obtained values can be used to estimate the actual intensity of the near infrared light used to illuminate the sample. This value can be used to normalize each scan and correct for fluctuations in light source intensity that affect the accuracy of the analyte concentration values obtained using the method of the present invention.

In addition, measurements taken over the spectral range of about 1430-1450nm or about 1930-1950nm provide a substantially non-reflective, highly attenuated signal due to the two main absorption peaks occurring at about 144 and 1935nm in the near infrared absorption spectrum of water. By collecting and measuring the radiation intensity in this one or both spectral ranges, the obtained value can be used as an estimate of the intensity of near infrared light that is not completely absorbed by the reference sample. This value can be used to subtract background or baseline information from analyte-specific signals obtained in other spectral regions and/or to provide an internal reference for detecting inaccurate measurements. To correct for the bedding effect caused by specular reflection (which varies with skin texture and age), this value can be subtracted from each spectral measurement obtained using the method of the present invention.

Measurements of the substantially unattenuated signal obtained from the first spectral region (e.g., spanning the spectral range of about 1320-. If the signals in the two spectral regions have relatively comparable values, then most of the radiation used to illuminate the tissue sample appears to be reflected by the skin surface and thus fails to penetrate the skin to interact with the analyte in the blood. Using this information, invalid measurements due to failure to obtain an appropriate instrument scan of the tissue can be identified.

The method of the present invention can be performed using any number of spectrophotometer configurations. Referring now to FIG. 1, a particular apparatus for determining the concentration of an analyte in a liquid sample is generally indicated by reference numeral 10. The apparatus includes a radiation source 12 that provides a plurality of distinct, non-overlapping wavelength regions in the range of about 600 to 3500 nm. Many suitable radiation sources are known to those skilled in the art, such as incandescent light sources directed at an interference filter, halogen light sources modulated by an associated chopper wheel, laser light sources, laser diode arrays, or high-speed Light Emitting Diode (LED) arrays. In one particular arrangement, the radiation source 12 provides radiation in three distinct wavelength regions, specifically a first wavelength region in the near infrared, typically in the range of about 1100 to 1350 nm; a second wavelength region, typically in the range of about 1930 to 1950 nm; and a third wavelength region, typically in the range of about 2000 to 3500 nm.

The apparatus 10 also includes sample interface optics 14 which direct incident radiation from a radiation source into a sample medium 16 containing an analyte and, after contact with the sample medium, collect spectrally modified radiation emerging as diffusely reflected light from the sample medium and pass it to a first lens system 18, thereby directing the light into first and second optical paths, indicated at 20 and 22, respectively. The first lens system 18 may comprise a partially reflective mirror arrangement as is well known to those skilled in the art.

In various configurations, sample interface optics 14 may be designed to enable device 10 to be brought into intimate contact with medium 16, such as by placing the device on the sample medium in direct contact therewith, thereby placing the radiation source in close proximity to the sample being analyzed for beam emission. After the beam is emitted, the reflected radiation is collected by a photosensitive device, such as a beam condensing device or a beam deflecting optical element. Alternatively, sample interface optics 14 may include a fiber optic waveguide coupled to the device, thereby enabling the device to be remotely located and operated. Other configurations are also provided in which a single bundle of optical fibers is used to both transmit and receive radiation to and from the medium. An optical pole disposed at one end of the single bundle of optical fibers emits near infrared radiation into the sample medium 16 and receives spectrally modified radiation that is returned to the apparatus 10 through the single bundle of optical fibers. Sapphire or high grade quartz may be used as the optical element in the fiber waveguide because these materials have good transmission characteristics in the near infrared spectral range.

The reflected light in the first optical path 20 is coupled to a first filter means 22 configured to pass light of a particular wavelength independent of the analyte concentration. In one configuration, the first filter means may comprise a narrow band pass filter having a near infrared absorption characteristic to selectively pass radiation having a wavelength substantially independent of the analyte concentration. The radiation emerging from the first filter 22 is then brought into contact with a first detection means 24. The coupling of the radiation to the first detection means may be via focusing means 26, such as a collimator lens or the like. Alternatively, the apparatus 10 may comprise a radiation detector capable of directly receiving radiation from the first filter device.

The first detection device detects the passing radiation and converts it into a signal representing the intensity of the radiation independent of the analyte. In one particular arrangement, the first detection device 24 comprises a lead sulfide photodetector capable of scanning a wavelength range of about 1100 to at least 3500nm at 1nm intervals.

The signal obtained from the first detection means can conveniently be converted to a digital signal, for example representative of the intensity of radiation at a wavelength independent of the analyte, using an analogue to digital converter. The digitized information may be conveniently input into a microprocessor or other electronic memory device as is well known to those skilled in the art.

Still referring to fig. 1, the reflected light in the second optical path 22 is sent to a tunable filter arrangement 28 that can adjust its absorption characteristics based on externally generated or already generated signals by the apparatus 10. Tunable filter arrangements typically include a screening filter, such as a neutral density filter, whose absorption characteristics can be adjusted to vary the attenuation of the radiation intensity represented by an external signal or system command. The degree of attenuation provided by the tunable filter 28 is dependent on a predetermined factor selected to ensure that the radiation exiting the tunable filter will be maintained at a constant value regardless of the intensity of the radiation prior to filtering. In one particular arrangement, the attenuation provided by the tunable filter arrangement is adjusted by a feedback signal generated by the first sensing arrangement 24.

The attenuated radiation from the tunable filter arrangement 28 is coupled to a primary analyte filter 30 having optical properties that selectively pass one or more wavelengths of the respective non-overlapping wavelength regions emitted by the radiation source 12. The wavelength passed through the primary analyte filter is selected to be related to the concentration of the analyte.

A second filter means 32 is provided in the device 10 in association with the primary analyte filter 30 such that wavelengths that selectively pass through the primary analyte filter interact with the second filter means, with the intensity of each passing wavelength being attenuated independently by the second filter means. For example, a set of independent weighting factors derived using chemometric techniques can determine the attenuation provided by the second filter arrangement.

In one particular configuration, the weighting factors are determined using partial least squares or principal component regression of the raw spectra obtained from the analyte-containing sample. The second filter arrangement 32 can be constructed using a suitable substrate layer capable of transmitting radiation in the range of at least 600 to 3500 nm. The base layer is typically plated with one or more layers of metals and/or oxides commonly used in the art to provide a variety of attenuation filter densities. Such coatings may be applied to the substrate using an emulsion or Chemical Vapor Deposition (CVD) techniques well known to those skilled in the art. In another arrangement, the second filter arrangement is a photographic mask having spectral lines of optical density on the mask that are proportional to weighting factors determined using rotational principal components or least squares analysis techniques.

After attenuation by the second filter means, the respective wavelengths are connected to a second detection means 34, such as a PbS detector or the like. As mentioned above, the wavelengths exiting the second filter means can be connected to the second detection means via a focusing means 36, such as a collimator lens. Alternatively, the apparatus 10 may comprise a radiation detector which directly receives the radiation from the second filter means.

The second detection means detects the attenuated wavelengths emerging from the second filter means and converts them into a signal that enables the analyte concentration to be determined using an analyte-specific algorithm. Specifically, the signal obtained from the second detection means is converted into a digital signal using an analog-to-digital converter. The digitized information is provided to a microprocessor where it is used to provide the analyte concentration that can be viewed on a display device and/or recorded on an output recorder.

Measurements of analyte concentrations in a variety of complex media, such as aqueous media with complex spectral backgrounds, can be obtained using the device 10. In one application, the device may be used to determine the concentration of an analyte in blood, particularly an organic analyte in blood, such as, but not limited to, glucose, urea (BUN), lipids, bilirubin, and ethanol. The analyte in the blood may be present in an in vitro sample medium (e.g., a blood sample) or the device may be used to measure blood analytes in tissue. However, the device 10 is particularly suited for use in field settings (e.g., measuring ethanol in blood) or home health monitoring (e.g., determining glucose levels in blood).

Referring now to FIG. 2, another apparatus for measuring the concentration of an analyte in a complex aqueous medium is indicated generally by the reference numeral 60. The apparatus includes a radiation source 62 that provides a plurality of distinct non-overlapping wavelength regions in the range of about 600 to 3500 nm. Radiation from the source 62 is directed to a photosensitive device 64, such as a collimating lens, selective filter device, or the like, which receives the radiation and directs the radiation into an optical path and/or passes selected wavelengths.

The near infrared radiation exiting the device 64 is coupled to a beam splitter 66, which splits the radiation into two beams, indicated at 68 and 70, respectively. A first beam 68 from the beam splitter 66 is incident on a sample medium 72 containing an analyte of unknown concentration. In fig. 2, sample media 72 includes a sample chamber formed from a suitable substrate capable of transmitting radiation in the near infrared range of interest. In one instance, the sample may comprise a serum sample, wherein the concentration of the analyte in the blood is to be determined. Alternatively, the first light beam 68 may be directed to a sample surface, such as a tissue surface, using a direct interface device or a simple interface device, such as the fiber optic waveguide devices described above. In this manner, the concentration of an analyte in a blood sample present in the tissue sample can be determined without damage using measurements of reflected near infrared absorption spectra after interaction of radiation with the tissue sample.

Radiation, including spectrally modified radiation that has interacted with a sample constituent (e.g., an analyte of interest), is then collected and directed into optical delivery element 74 disposed in the optical path. Optical transfer element 74 comprises a positive correlation filter system having an absorption spectrum sufficient to receive radiation and selectively highlight one or more wavelengths that are highly correlated with the concentration of the analyte of interest and substantially uncorrelated with interfering components present in the sample. Thus, positive correlation filter systems pass a large portion of the selected wavelength range, which provide analyte-specific information as well as information about the measurement background and information that can be used to correct for instrument variations or interference effects. Radiation exiting optical delivery element 74 is received by detection device 76, which converts the spectrally modified radiation into a signal representative of the intensity of the radiation. The detection means may comprise a broad spectrum photodetector such as a PbS photodetector.

Still referring to fig. 2, the second beam 70 from the beam splitter 66 is incident on a photosensor 78 disposed in the optical path. In one configuration, the photosensor 78 includes a neutral density filter arrangement having absorption characteristics sufficient to attenuate radiation equally over a selected near infrared wavelength range. In another configuration, the light sensitive element 78 is an optical transfer element that includes a positive correlation filter system having an absorption spectrum that is the same as the absorption spectrum of the optical transfer element 74. Radiation exiting the photosensor 78 is received by the detection device 80 and converted to a signal representative of its intensity.

The positive correlation filter system may be formed from a single substrate layer with a photosensitive coating having absorption characteristics capable of selectively emphasizing one or more wavelengths that are highly correlated with a particular analyte concentration. In certain system configurations, the positive correlation filter comprises a plurality of filter layers, each layer having a selected filter density and/or filter thickness suitable to provide the desired absorption characteristics. In one aspect, at least one layer of the system has a filter density and/or thickness that includes a wavelength weighting factor, where the weighting factor increases the positive correlation of the passed wavelength to the concentration of the analyte in the selected sample medium.

The signals generated by the detection means 76 and 80 are then connected to means 82 for converting these signals into digital signals representing the ratio of the intensity of the radiation emitted from the source 62 to the corresponding spectrally modified radiation emitted from the sample. In this way variations in the intensity of the radiation emitted from the source 62 can be corrected, thereby eliminating a potential source of error in the measurements obtained in the system. Alternatively, the ratio of the signals may be converted to a digital form and interpreted to determine the concentration of the analyte by an internal microprocessor 84 system or related systems using methods well known to those skilled in the art.

If desired, the microprocessor can be programmed to calculate the analyte concentration by applying a stoichiometric algorithm to the ratio signal. Suitable algorithms may be determined using the above-described chemometric techniques, such as least squares analysis or rotational principal component analysis of the original absorption spectrum of the analyte of interest.

It should be understood that while the invention has been described in conjunction with the specific preferred embodiments thereof, the foregoing description, as well as the examples set forth below, are intended to illustrate and not limit the scope of the invention. Other aspects, advantages and modifications within the scope of the invention will be apparent to those skilled in the art.

Examples of the present invention

By using the method of the invention, a non-invasive glucose measurement result can be obtained. In particular, reflectance optical measurements are made in the near infrared spectral region from about 1100nm to 3500 nm. Spectral scans were collected from the forearm of the volunteer using a tungsten-mercury (W-Hg) radiation source, a lead sulfide (PbS) detector, and an instrument with a scan rate of nm/0.4 second.

Certain spectral ranges are known to be useful in finding information about glucose concentration from forearm tissue scans. In vivo glucose tolerance studies are combined with non-invasively obtained determination of glucose concentration in blood in vivo to determine specific spectral regions. In particular, figure 3 shows the number-related scans obtained during the in vivo tolerance study (1-6). As can be seen in the figure, a significant change in the difference in reflection intensity over the range of about 2120 to 2180nm was recorded during the course of the times studied. This increase in change was directly related to the increase in glucose content in blood during the tolerance test, indicating that the range of 2120 to 2180nm contains specific spectral information of glucose.

Once a specific spectral range is found, non-invasive glucose measurements can be obtained using information from four different spectral ranges. The first spectral range includes radiation generated at about 1320 to 1340 nm. This range provides a strong reflected signal, in which there is no main absorption band for glucose. The individual scans may be normalized with the information obtained over the first spectral range in order to correct for fluctuations in the radiation source and for variations due to mechanical disturbances.

The second spectral range includes radiation generated over any of about 1440 to 1460nm or about 1940 to 1960 nm. These ranges provide a signal that is substantially free of reflections due to the strongly absorbing bands of water that attenuate diffusely reflected radiation. The information obtained over these ranges can be used as a background and baseline for subtraction from other measurements. These measurements allow for shimming of the undulations caused by the values of the specular reflected signals, which can be used to detect inappropriate measurements.

The third spectral range includes radiation generated at about 1670 to 1690 nm. This range provides analyte-specific information due to the presence of the glucose vibration harmonic band.

The fourth spectral range includes radiation generated at about 2120 to 2280 nm. This range provides analyte specific information due to glucose combined with the vibrational band.

The signals for the other spectral regions are normalized with the signals obtained for the first range. This process eliminates problems associated with source variations and serves to provide an internal reference when repeated for each spectral scan. Measurement variations caused by optical interface variations (e.g. patient movement) are thus greatly reduced.

Background information may be eliminated by subtracting the signal obtained in the second range from the signal obtained in the third and fourth analyte specific ranges. In this way, the bedding effect caused by specular reflection is corrected, which may vary with skin texture and age.

The normalized and baseline corrected third and fourth range signals may be applied in the chemometric analysis being analyzed. Fig. 4 shows the normalized difference between the signals in the second and third ranges.

As can be seen from the results shown in fig. 4, an increase in the glucose content in the blood leads to an increase in the signal difference in the two ranges.

Claims (21)

1. An apparatus for determining the concentration of an analyte in a sample, comprising:

(a) means for irradiating the sample with incident radiation comprising a plurality of different non-overlapping wavelength spectral regions in the 1100-3500nm spectrum;

(b) means for collecting reflected radiation from the sample and directing said reflected radiation into first and second optical paths, wherein said first optical path comprises radiation from the first wavelength spectral region 1100-1350 nm;

(c) a first filter means disposed in said first optical path, wherein said first filter means is capable of selectively passing radiation substantially independent of analyte concentration;

(d) first detecting means for receiving radiation selectively passed through said first filter means and converting said radiation selectively passed through said first filter means into a signal representative of the intensity of said radiation;

(e) a tunable filter arrangement disposed in said second optical path, wherein said tunable filter arrangement attenuates radiation intensity in said second optical path;

(f) a primary analyte filter means capable of receiving radiation attenuated by said tunable filter means and selectively passing therethrough one or more individual wavelengths, wherein said one or more individual wavelengths have a particular correlation with said analyte concentration;

(g) a second filter means capable of receiving one or more individual wavelengths emitted from the primary analyte filter means and attenuating the intensity of each individual wavelength;

(h) second detection means for receiving the individual wavelengths attenuated by said second filter means and converting the detected wavelengths into signals representative of the intensity of said wavelengths.

2. The apparatus of claim 1, wherein: the first filter device comprises a narrow-band-pass filter.

3. The apparatus of claim 2, wherein: the tunable filter arrangement includes a neutral density filter for use in cooperation with an associated filter in the filter system.

4. The apparatus of claim 3, wherein: the signal obtained from the first detection means is used to adjust the attenuation provided by the tunable filter arrangement.

5. The apparatus of claim 1, wherein: the second filter means comprises a neutral density filter for use in cooperation with an associated filter in the filter system.

6. The apparatus of claim 5, wherein: establishing the attenuation provided by the second filter means using a weighting factor.

7. An apparatus for determining the concentration of an analyte in a sample, comprising:

(a) a radiation source capable of emitting radiation comprising a plurality of different non-overlapping spectral regions in the 1100-3500nm spectrum;

(b) means for splitting into first and second beam paths radiation emitted by (a) a portion of the radiation source;

(c) means for illuminating said sample with radiation in a first optical path, thereby providing reflected radiation;

(d) means for collecting reflected radiation from said sample and directing said reflected radiation into a reflected optical path;

(e) a first optical transfer cell disposed in the reflected light path, the cell comprising a first positive correlation filter device having absorption characteristics suitable for absorbing the reflected radiation and projecting one or more wavelengths from the reflected radiation, wherein the one or more wavelengths have a high correlation with the concentration of an analyte in the sample;

(f) means for receiving one or more wavelengths highlighted by said first optical transfer element and converting said one or more wavelengths into a signal representative of the intensity of said highlighted wavelength;

(g) a second optical transfer cell disposed in said second optical path, said second cell comprising a neutral density filter arrangement having absorption characteristics sufficient to attenuate equally the intensity of radiation in said second optical path over a selected 1100-3500nm wavelength region;

(h) means for receiving attenuated radiation from said second optical transfer element and converting said radiation into a signal representative of its intensity;

(i) means for calculating the concentration of the analyte in said sample using the signals generated by means (f) and (h).

8. The apparatus of claim 7, wherein: the second optical transfer unit includes a second positive correlation filter device having an absorption characteristic identical to that of the first positive correlation filter device.

9. The apparatus of claim 7, wherein: said means for calculating the concentration of an analyte in the sample converts the signals generated by means (f) and (h) into digital signals representing the ratio of the intensity of radiation from the light source to the corresponding intensity of radiation from the sample.

10. The apparatus of claim 7, wherein: said means for calculating the concentration of an analyte in a sample comprises means for applying a stoichiometric algorithm to the signals generated by means (f) and (h).

11. The apparatus of claim 7, wherein: the first positive correlation filter device includes a plurality of films, each film having selective absorption characteristics that highlight a plurality of wavelengths of the filter device that are highly correlated with analyte concentration.

12. An apparatus for determining the concentration of an analyte in a sample, comprising:

(a) a radiation source capable of emitting radiation comprising a plurality of different non-overlapping wavelength spectral regions in the 1100-3500nm spectral region;

(b) means for splitting radiation emitted by said radiation source into first and second beam paths;

(c) means for illuminating the sample with radiation in the first beam path, thereby providing reflected radiation;

(d) means for collecting reflected radiation from the sample and directing said reflected radiation into a reflected light path;

(e) a first optical transfer element disposed in the path of said reflected light, said first element comprising first positive correlation filter means having absorption characteristics suitable for absorbing reflected radiation and projecting one or more wavelengths from said reflected radiation, wherein said one or more wavelengths have a high correlation with the concentration of an analyte in the sample;

(f) means for receiving one or more wavelengths highlighted by the first optical transfer element and converting them to a signal representative of the intensity of the highlighted wavelength;

(g) a second optical transfer unit disposed in a second beam path, the second unit including a second positive correlation filter device having an absorption characteristic identical to that of the first positive correlation filter device;

(h) means for receiving the radiation attenuated by the second optical transfer element and converting it into a signal representative of its intensity;

(i) means for calculating the concentration of the analyte in the sample using the signals generated by the means (f) and (h) sections.

13. The apparatus of claim 12, wherein: the first positive correlation filter device includes a plurality of films, each film having selective absorption characteristics that highlight a plurality of wavelengths of the filter device that are highly correlated with analyte concentration.

14. The apparatus of claim 13, wherein: the absorption characteristics of at least one of the layers in the first and second positive correlation filter arrangements are established using weighting factors.

15. The apparatus of claim 6 or 14, wherein: the weighting factors are derived using chemometric techniques.

16. The apparatus of claim 15, wherein: the weighting factors are derived using a rotated principal component analysis of the absorption spectrum of the analyte.

17. The apparatus of claim 1, wherein: the wavelength of the incident radiation is in the range of about 1100 to 3500 nm.

18. The apparatus of claim 7 or 12, wherein: the radiation source emits radiation at a wavelength in the range of about 1100 to 3500 nm.

19. The apparatus of claim 1, 7 or 12, wherein: the sample comprises body tissue and the analyte comprises an organic blood analyte.

20. The apparatus of claim 19, wherein: the blood analyte is selected from the group consisting of glucose, urea (BUN), lipids, bilirubin, and ethanol.

21. The apparatus of claim 20, wherein: the blood analyte is glucose.