Noninvasive measurement of glucose through the optical properties of tissue

• US 6,990,364 B2
• Filed: 01/25/2002
• Issued: 01/24/2006
• Est. Priority Date: 01/26/2001
• Status: Expired due to Term

• Current Assignee: GLT Acquisition Corporation
• Original Assignee: Sensys Medical, Inc. (Sensys Medical Limited)
• Inventors: Ruchti, Timothy L., Monfre, Stephen L., Blank, Thomas B., Lorenz, Alexander, Thennadil, Suresh N.
• Primary Examiner(s): WINAKUR, ERIC FRANK
• Application Number:
  US10/297,736
• Publication Number:
  US 20040068163A1
• Time in Patent Office: 1,460 Days
• Field of Search: 600/310, 600/322, 600/316, 600/473
• US Class Current: 600/310
• CPC Class Codes: A61B 2560/0223 of calibration, e.g. protocols for calibrating sensors
• CPC Class Codes: A61B 5/14532 for measuring glucose, e.g. by tissue impedance measurement
• CPC Class Codes: A61B 5/1455 using optical sensors, e.g. spectral photometrical oximeters
• CPC Class Codes: A61B 5/1495 Calibrating or testing of in-vivo probes
• CPC Class Codes: G01N 2021/3595 using FTIR
• CPC Class Codes: G01N 21/31 Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry G01N21/72 takes precedence
• CPC Class Codes: G01N 21/33 using ultraviolet light G01N21/39 takes precedence
• CPC Class Codes: G01N 21/359 using near infrared light

Abstract

Methods and apparatus for noninvasive determination of blood analytes, such as glucose, through NIR spectroscopy utilize optical properties of tissue as reflected in key spectroscopic features to improve measurement accuracy and precision. Physiological conditions such as changes in water distribution among tissue compartments lead to complex alterations in the measured absorbance spectrum of skin and reflect a modification in the effective pathlength of light, leading to a biased noninvasive glucose measurement. Changes in the optical properties of tissue are detected by identifying key features responsive to physiological variations. Conditions not conducive to noninvasive measurement of glucose are detected. Noninvasive glucose measurements that are biased by physiological changes in tissue are compensated. In an alternate embodiment, glucose is measured indirectly based on natural physiological response of tissue to glucose concentration. A spectroscopic device capable of such measurements is provided.

Claims

1. A method for noninvasive measurement of a target analyte in a tissue sample, comprising the steps of:

measuring a spectrum of a tissue sample;

detecting at least one optical property of said tissue sample related to fluid shift between extravascular and intravascular tissue compartments triggered by changes in concentration of said analyte, as manifested by at least one spectral feature reflecting said fluid shift; and

either correcting a direct analyte measurement on the basis of said at least one optical property;

or

measuring said analyte indirectly on the basis of said at least one optical property.

2. The method of claim 1, wherein said step of measuring a spectrum comprises measuring a spectrum, m, of said tissue, said spectrum denoted by a vector m∈

<sup>1×

N</sup> of absorbance values pertaining to a set of N wavelengths, λ

∈

^N spanning near IR region of approximately 700 to 2500 nm.

3. The method of claim 2, wherein m comprises a set of wavelengths in the near IR that have been optimized for extraction of features.

4. The method of claim 1, further comprising the step of preprocessing said spectrum, preprocessing comprising any of:

filtering;

calculating a first or second derivative of said spectrum; and

scatter correction.

5. The method of claim 1, further comprising the step of extracting said at least one features, feature extraction comprising any mathematical transformation that enhances a quality or aspect of a sample measurement for interpretation so that structural properties and physiological state of said tissue sample are concisely represented.

6. The method of claim 5, wherein said feature extraction step comprises the steps of:

identifying distinct absorption bands that change in differing manners with respect to changes in pathlength, scattering and absorption properties of the sample;

developing a set of features that represents or reflects the optical properties of the tissue; and

applying said features to identify;

conditions suitable for analyze measurement; and

conditions unsuitable for analyte measurement.

7. The method of claim 5, wherein features include any of:

at least one simple features, comprising values of a spectral measurement or a processed spectral measurement at critical points, wherein a critical point is a point having a slope of zero;

a derived features, comprising at least one features derived from simple features through a mathematical transformation; and

an abstract features;

wherein simple and derived features generally have a physical interpretation, and wherein abstract features do not necessarily have a specific interpretation related to a physical system.

8. The method of claim 7, wherein an abstract features comprises scores from a principal component analysis, wherein leading principal components represent variation related to the structural properties and physiological state of the tissue sample, so that the optical properties of the tissue are represented.

9. The method of claim 7, wherein a feature is determined from a second derivative of said spectrum, wherein value of said second derivative spectrum at each critical point constitutes a feature that represents a key property of the sample.

10. The method of claim 9, wherein a feature includes any of:

a normalization point;

a fat band point;

a protein band point; and

a water band point;

wherein a normalization point is used to determine a derived feature, and fat, protein, and water points are respectively located in vicinity of an absorption band due to fat, protein, or water.

11. The method of claim 1, further comprising the step of:

determining difference between a tissue template and a preprocessed spectrum according to;

z=x−

(cx_t+d);

wherein x comprises a pre-processed spectrum or a selected set of features, x_t comprises a tissue template associated with a measurement period, and c and d are slope and intercept adjustments to the tissue template.

12. The method of claim 11, wherein said tissue template is determined through one or more spectral measurements combined according to a predetermined data selection criterion during each measurement period.

13. The method of claim 12, wherein a measurement period comprises a time period during which the state of the tissue sample is uniform and measurement site is constant.

14. The method of claim 12, further comprising the step of:

providing an associated set of reference analyte values, said values combined according to said predetermined data selection criterion to form a measurement bias adjustment.

15. The method of claim 11, wherein said tissue template comprises any set of features from a given subject or calibration set that future spectral measurements will be compared with, wherein c and d are determined are determined through least-squares fit of the tissue template over a particular wavelength range to the measured spectrum.

16. The method of claim 1, further comprising any of the steps of:

detecting conditions not conducive to analyte measurement; and

detecting outliers.

17. The method of claim 16, said step of detecting conditions not conducive to analyte measurement comprising the steps of:

monitoring at least one selected features; and

ensuring that variation in said at least one selected feature over a given measurement does not exceed that of a calibration set or another previously established limit.

18. The method of claim 1, further comprising the step of providing a calibration model, said model determined from a calibration set of exemplary paired data points each consisting of a preprocessed spectral measurement, x and an associated reference analyte value, y, said calibration set including analyte concentrations that span the expected range of variation and spectral variation representative of future spectral measurements, said model comprising an equation, a set of parameters and corresponding computer code implemented to measure a subject'"'"'s analyte level on the basis of a processed spectral measurement.

19. The method of claim 18, wherein said y values are determined from samples of blood, serum, plasma or interstitial fluid taken from a fingertip, a site near the measurement site or an alternate site.

20. The method of claim 19, wherein said alternate site comprises a sample site that has been designed or determined to reflect the sample site.

21. The method of claim 18, wherein said parameters are calculated using any of:

multivariate regression;

weighted multivariate regression;

principal component regression;

partial least squares regression; and

artificial neural networks.

22. The method of claim 18, wherein said step of correcting a direct analyte measurement on the basis of said detected changes comprises supplementing said model with selected features according to:

ŷ

=f(x_p,z)+b,

where ŷ

is an estimated glucose concentration, x_p∈

^N is a processed spectral measurement, z∈

^M is a set of features representative of the physiological state or optical properties of the tissue, f;

^N,M→

¹ is a model used to measure glucose on the basis of a preprocessed spectrum and at least one extracted features, and b is a baseline adjustment for glucose measurement associated with both a tissue template and said calibration model.

23. The method of claim 18, wherein said step of correcting a direct analyte measurement on the basis of said detected changes comprises supplementing said model with selected features according to:

ŷ

=f(x_p)−

(m_sg(z)+m_i)+b,

where ŷ

is an estimated glucose concentration, x_p∈

^N is a processed spectral measurement, z∈

^M is a set of features representative of the physiological state or optical properties of the tissue, wherein x_p and z are independent, where f;

^N→

¹ is a model used to measure glucose in the absence of physiological or other tissue variation, g;

^M→

¹ is a model used to map the features to a variable correlated to error in glucose measurement caused by a change in the optical properties of the tissue, m_s and m_i are slope and intercepts used to convert g(z) to correct units, and b is a baseline adjustment for glucose measurement associated with both a tissue template and said calibration model.

24. The method of claim 23, wherein f(·

) and g(·

) are separately determined experimentally, wherein f(·

) is determined by manipulating glucose while tissue optical properties remain constant, and wherein the optical properties of tissue are allowed to fluctuate and g(·

), m_s and m_i are determined on the basis of the error in glucose measurement where target value for g(·

) is given by;

r=y−

f(x_p)−

b

where y is a reference glucose concentration.

25. The method of claim 24, wherein said step of correcting a direct analyte measurement on the basis of said detected changes comprises supplementing said model with selected features according to:

ŷ

=x_pF−

(m_szG+m_i)+b,

wherein f(·

) and g(·

) are determined to be linear over range of measurement and where;

F∈

<sup>N×

1</sup> and G∈

<sup>M×

1</sup>.

26. The method of claim 18, wherein said step of measuring said analyte indirectly on the basis of said at least one spectral features comprises using extracted features to measure glucose indirectly according to:

ŷ

=(m_sg(z)+m_i)+b

where g;

^M→

¹ comprises said model, said model used to map set of features z to a variable correlated to a reference glucose level concentration and m_s and m_i are slope and intercepts used to convert g(z) to the correct units and b is a baseline adjustment for glucose measurement.

27. The method of claim 26, wherein said at least one feature is selected based an their combined correlation to the reference glucose concentration.

28. The method of claim 27, wherein said at least one features is selected based on any of:

a priori knowledge;

trial-and-error;

stepwise regression;

random search techniques;

genetic algorithms; and

evolutionary programming.

29. The method of claim 2, wherein g(·

) is determined according to;

ŷ

=(m_szG+m_i)+b,

where G∈

<sup>M×

1</sup>.

30. The method of claim 26, wherein z includes either a normalized second derivative fat b and or protein band, and wherein m_s, m_i and G are determined through any of multivariate regression, weighted multivariate regression and locally weighted regression.

31. The method of claim 26, wherein said step of measuring said analyte indirectly on the basis of said at least one spectral features comprises using an extracted feature to measure glucose indirectly according to:

ŷ

=g₁z₁+g₂z₂+g₃z₃+g₄+b,

where z₁, z₂ and z₃ comprise normalized second derivative values of a fat band and two water bands, respectively.

32. The method of claim 1, wherein said step of measuring said analyte indirectly comprises:

providing a noninvasive glucose calibration model wherein abstract features that reflect said changes in optical properties of said tissue are used as independent variables for said calibration;

preprocessing said measured spectrum; and

decomposing said preprocessed spectrum according to;

z=xP

where x∈

<sup>1×

N</sup> is the preprocessed spectrum, N is number of wavelengths selected for calibration, P∈

<sup>1×

M</sup> is a set of M eigenvectors or loadings obtained from a principal components analysis of a calibration set and z∈

<sup>1×

M</sup> is the set of abstract features used to measure glucose through application of said calibration model, wherein said model is either linear or nonlinear.

33. The method of claim 32, wherein said abstract feature comprises scores from a principal component analysis.

34. The method of claim 32, wherein providing said calibration model comprises the steps of:

preprocessing spectral measurements from said calibration set through a first derivative with a wide smoothing window;

scatter correcting said preprocessed measurements from said calibration set through multiplicative scatter correction or standard normal variate transformation;

selecting wavelengths from a range of approximately 1450-1850 nm and optionally 1180-1450 nm;

correcting said preprocessed, scatter-corrected data to a tissue template; and

applying partial least squares regression.

35. The method of claim 34, further comprising the steps of:

applying identical processing to said measured spectrum as applied in developing said calibration model.

36. The method of claim 35, further comprising the step of:

determining a glucose measurement according to;

ŷ

=mG+b

where G∈

<sup>M×

1</sup> is the calibration model derived from partial least-squares regression and b is a baseline correction.

37. The method of claim 1, wherein said analyte comprises glucose.

38. An apparatus for noninvasive measurement of a target analyte in a tissue sample comprising:

means for measuring a spectrum of a tissue sample;

means for detecting at least one optical property of said tissue sample related to fluid shift between extravascular and intravascular tissue compartments triggered by change in concentration of said analyte, as manifested by at least one spectral features reflecting said fluid shifts; and

either correcting a direct analyte measurement on the basis of said at least one optical property;

or

measuring said analyte indirectly on the basis of said at least one optical property.

39. The apparatus of claim 38, wherein said means for measuring a spectrum of a tissue sample comprises a spectrometer system, said spectrometer system comprising:

a source of near infrared (NIR) radiation;

a wavelength selection element;

a means for interfacing with the measurement site on the skin of a subject, wherein radiation is directed toward the measurement site from the source, and a light signal returned from the site is collected;

means for detecting said returned radiation; and

means for digitizing said returned signal.

40. The apparatus of claim 39, wherein said NIR source radiates energy in a range of approximately 700-2500 nm.

41. The apparatus of claim 39, wherein said NIR source comprises any of:

an LED array; and

a halogen lamp.

42. The apparatus of claim 39, further comprising a band pass filter to minimize effect of wavelengths radiated from said source that are outside a spectral range of interest.

43. The apparatus of claim 39, wherein said wavelength selection element comprises any of:

a dispersive element;

an interferometer; and

successive illumination of elements of an LED array.

44. The apparatus of claim 43, further comprising a reference wavelength standard, wherein changes in said wavelength selection element caused by environmental changes are compensated by scanning said standard proximate to or simultaneous with interrogation of said tissue.

45. The apparatus of claim 39, wherein said means for interfacing with the measurement site on the skin of a subject comprises:

at least one optical element that directs radiation to and/or from the tissue, said optical element including any of;

a light pipe;

a fiber optic;

a lens system; and

a light directing mirror system.

46. The apparatus of claim 45, said means for interfacing with the measurement site on the skin of a subject further comprising:

a guide to assist in interfacing said optical element and said measurement site.

47. The apparatus of claim 46, said means for interfacing with the measurement site further comprising a subject interface module, said subject interface module including at least an elbow rest and a wrist rest.

48. The apparatus of claim 45, said means for interfacing with the measurement site on the skin of a subject further comprising:

an optical coupling fluid, a portion of said fluid being interposed between said measurement site and said source to minimize specular reflectance from surface of the skin.

49. The apparatus of claim 39, wherein an area irradiated and an area from which returning radiation is collected from are separated by a preselected distance, said distance selected to target a tissue volume conducive to measurement of a property of interest.

50. The apparatus of claim 39, wherein said means for detecting said returned radiation comprises:

a single diodes or a diode array responsive to targeted wavelengths of interest.

51. The apparatus of claim 50, wherein said diodes include InGaAs detectors.

52. The apparatus of claim 50, wherein said diodes are selected such that material junctions in said diodes do not coincide with targeted wavelengths.

53. The apparatus of claim 50, wherein said diodes convert said signal to a voltage.

54. The apparatus of claim 39, said means for digitizing said returned signal comprises an ADC (analog-to-digital converter), wherein said signal, having been converted to a voltage, is digitally sampled for analysis on a microprocessor-based system.

55. The apparatus of claim 54, further comprising a display, wherein a result of said analysis is displayed.

56. The apparatus of claim 38, wherein said means for detecting at least one optical property of said tissue sample related to fluid shift in said tissue, as manifested by said at least one spectral feature reflecting said changes;

and

either correcting a direct analyte measurement on the basis of said at least one spectral feature;

or

measuring said analyte indirectly on the basis of said at least one spectral feature comprises an analyzer, said analyzer including;

means for measuring a spectrum, m of said tissue, said spectrum denoted by a vector m∈

<sup>1×

N</sup> of absorbance values pertaining to a set of N wavelengths, λ

∈

^N spanning near-infrared region of approximately 700 to 2500 nm.

57. The apparatus of claim 56, said analyzer further comprising:

means for preprocessing said spectrum.

58. The apparatus of claim 57, said analyzer further comprising:

means for extracting said at least one feature, comprising any mathematical transformation that enhances a quality or aspect of a sample measurement for interpretation so that structural properties and physiological state of said tissue sample are concisely represented.

59. The apparatus of claim 58, said analyzer further comprising:

a tissue template determined through one or more spectral measurements combined according to a predetermined data selection criterion during each a measurement period, said tissue template subtracted from said spectrum.

60. The apparatus of claim 59, said analyzer further comprising means for calculating a measurement bias adjustment.

61. The apparatus of claim 60, said analyzer further comprising

means for detecting outliers and conditions detrimental to spectroscopic glucose measurement though said at least one feature.

62. The apparatus of claim 61, said analyzer further comprising:

at least one calibration model, said model determined from a calibration set of exemplary paired data points each consisting of a spectral measurement, x and an associated reference analyte value, y, said calibration set including analyte concentrations that span the expected range of variation and spectral variation representative of future spectral measurements, said model comprising an equation, a set of parameters and corresponding computer code implemented to measure a subject'"'"'s analyte level on the basis of a processed spectral measurement.

63. The apparatus of claim 38, wherein said analyte comprises glucose.

64. A method for noninvasive measurement of a target analyte property in a tissue sample, comprising the steps of:

noninvasively measuring a spectrum of a tissue sample in vivo;

detecting at least one optical property of said tissue sample related to physiologic state of said tissue resulting from change in concentration of said analyte, as manifested by at least one spectral feature reflecting said physiologic state; and

either correcting a direct analyte measurement on the basis of said at least one optical property;

or

measuring said analyte indirectly on the basis of said at least one optical property.

65. A method for noninvasive measurement of glucose concentration in a tissue sample, comprising the steps of:

measuring a spectrum of a tissue sample;

detecting at least one optical property of said tissue sample related to state of said tissue resulting from change in concentration of glucose, as manifested by at lest one spectral feature reflecting said state; and

either correcting a direct glucose measurement on the basis of said at least one optical property;

or

measuring said glucose concentration indirectly on the basis of said at least one optical property.

66. A method for noninvasive measurement of a target analyte in a tissue sample, comprising the steps of:

measuring a spectrum of a tissue sample;

detecting conditions not conducive to analyte measurement;

if conditions are conducive to analyte measurement, detecting at least one optical property of said tissue sample related to physiology of said tissue resulting from change in concentration of said analyte, as manifested by at least one spectral feature reflecting said physiology; and

either correcting a direct analyte measurement on the basis of said at least one spectral feature;

or

measuring said analyte indirectly on the basis of said at least one optical property.