Multi-tier method of developing localized calibration models for non-invasive blood analyte prediction

• US RE41,333 E1
• Filed: 01/27/2005
• Issued: 05/11/2010
• Est. Priority Date: 07/22/1999
• Status: Expired due to Term

• Current Assignee: GLT Acquisition Corporation
• Original Assignee: Sensys Medical, Inc. (Sensys Medical Limited)
• Inventors: Thennadill, Suresh N., Ruchti, Timothy L., Monfre, Stephen L., Blank, Thomas B.
• Primary Examiner(s): WINAKUR, ERIC FRANK
• Application Number:
  US11/046,673
• Time in Patent Office: 1,930 Days
• Field of Search: 600/310, 600/322, 600/323, 600/330, 600/336, 600/473, 128/920, 128/923, 128/924, 128/925, 356/402, 250/340, 250/341.1, 250/339.01, 702/19
• US Class Current: 600/322
• CPC Class Codes: A61B 2560/0223 of calibration, e.g. protocols for calibrating sensors
• CPC Class Codes: A61B 2560/0233 Optical standards
• CPC Class Codes: A61B 5/1075 for measuring dimensions by non-invasive methods, e.g. for determining thickness of tissue layer A61B8/0858 takes precedence
• 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: A61B 5/7264 Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems neural networks per se G06N3/00; expert systems per se G06N5/00
• CPC Class Codes: G01N 21/274 Calibration, base line adjustment, drift correction
• CPC Class Codes: G01N 21/359 using near infrared light
• CPC Class Codes: G01N 21/4785 Standardising light scatter apparatus; Standards therefor
• CPC Class Codes: G01N 21/49 within a body or fluid

Abstract

A method of multi-tier classification and calibration in noninvasive blood analyte prediction minimizes prediction error by limiting co-varying spectral interferents. Tissue samples are categorized based on subject demographic and instrumental skin measurements, including in vivo near-IR spectral measurements. A multi-tier intelligent pattern classification sequence organizes spectral data into clusters having a high degree of internal consistency in tissue properties. In each tier, categories are successively refined using subject demographics, spectral measurement information and other device measurements suitable for developing tissue classifications.

The multi-tier classification approach to calibration utilizes multivariate statistical arguments and multi-tiered classification using spectral features. Variables used in the multi-tiered classification can be skin surface hydration, skin surface temperature, tissue volume hydration, and an assessment of relative optical thickness of the dermis by the near-IR fat band. All tissue parameters are evaluated using the NIR spectrum signal along key wavelength segments.

Claims

1. A method of developing a multi-tiered calibration model for estimating concentration of a target blood analyte from measured tissue spectra, comprising the steps of:

providing a calibration set, wherein said calibration set comprises a data set of exemplar spectral measurements from a representative sampling of a subject population;

initially, classifying said exemplar measurements into previously defined classes based on a priori a priori information pertaining to a corresponding subject;

further classifying said exemplar measurements into previously defined classes based on at least one instrumental measurement at a tissue measurement site;

extracting at least one feature from said exemplar measurements for still further classification, wherein a decision rule makes class assignments; and

calculating at least one localized calibration model based on said classified measurements and an associated set of reference values.

2. The method of claim 1, wherein said initial classification step comprises the steps of:

in a first tier, classifying said measured spectrum exemplar measurements into previously defined classes based on subject'"'"'s age; and

in a second tier, further classifying said measured spectrum exemplar measurements into previously defined classes based on subject'"'"'s sex.

3. The method of claim 1, wherein said further classification step further comprises the steps of:

in a third tier further classsifying said exemplar measurements into previously defined classes based on an estimation of stratum corneum hydration at said tissue measurement site; and

in a fourth tier, further classifying said exemplar measurements into previously defined classes based on skin temperature at said tissue measurement site.

4. The method of claim 3, wherein said stratum corneum hydration estimate is based on a measurement of ambient humidity at said tissue measurement site.

5. The method of claim 1, wherein said feature extraction step comprises any mathematical transformation that enhances a quality or aspect of sample measurement for interpretation to represent concisely structural properties and physiological state of a tissue measurement site, wherein a resulting set of features is used to classify a subject and determine a calibration model that is most useful for blood analyte prediction.

6. The method of claim 5, wherein said features are represented in a vector, zΣ

^M that is determined from a preprocessed measurement through;

z=f(λ

,x)

where f(•

);

^N→

^M is a mapping from a measurement space to a feature space, wherein decomposing f(•

) yields specific transformations, f_i(•

);

^N→

^M_i for determining a specific feature, wherein the dimension M_i indicating whether an i^th feature is a scalar or a vector and an aggregation of all features is the vector z, and wherein a feature exhibits a certain structure indicative of an underlying physical phenomenon when said feature is represented as a vector or a pattern.

7. The method of claim 6, wherein individual features are divided into categories, said categories comprising:

abstract features that do not necessarily have a specific interpretation related to a physical system; and

simple features that are derived from an a priori understanding of a sample and that can be related directly to a physical phenomenon.

8. The method of claim 7, wherein said simple features can be calculated from NIR spectral absorbance measurements, said simple features including any of:

thickness of adipose tissue;

hematocrit level;

tissue hydration;

magnitude of protein absorbance;

scattering properties of said tissue;

skin thickness;

temperature related effects;

age related effects;

spectral characteristics;

pathlength estimates;

volume fraction of blood in tissue; and

spectral characteristics related to environmental influences.

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

employing spectral decomposition to determine features related to a known spectral absorbance pattern.

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

employing factor-based methods to build a model capable of representing variation in a measured absorbance spectrum related to a demographic variable;

wherein projection of a measured absorption onto said model constitutes a feature that represents spectral variation related to said demographic variable.

11. The method of claim 1, wherein said feature extraction step assigns a measurement to one of many predefined classes.

12. The method of claim 1, further comprising the steps of;

measuring the similarity of a feature to predefined classes; and

assigning class membership.

13. The method of claim 1, further comprising the step of;

using measurements and class assignments to determine a mapping from features to class assignments.

14. The method of claim 13, further comprising the steps of:

defining classes from said features in a supervised manner, wherein each set of features is divided into two or more regions, and wherein classes are defined by combination of feature divisions;

performing a cluster analysis on the spectral data to determine groups of said defined classes that can be combined, wherein the final number of class definitions is significantly reduced;

designing a classifier subsequent to class definition through supervised pattern recognition by determining an optimal mapping or transformation from the feature space to a class estimate that minimizes the number of misclassifications; and

creating a model based on class definitions that transforms a measured set of features to an estimated classification, wherein said class definitions are optimized to satisfy specifications of a measurement system used to take said measurements.

15. The method of claim 14, wherein said optimized classes comprise groups of measurements wherein similarity between measurements within a group is greater than similarity between groups.

16. The method of claim 15, said step of calculating at least one localized calibration model comprising:

calculating weights, w, for said exemplar measurements according to;

W=(X^TX)<sup>−

1</sup>X^Ty,

 

where X represents a matrix of spectral measurements, and y represents a reference value of said target analyte concentration for each measurement.

17. The method of claim 16, wherein a vector of weights of spectral measurements within one of said groups comprises a regression vector for said group;

wherein said regression vector comprises a calibration model for said group.

18. A method of developing a multi-tiered calibration model for estimating concentration of a target blood analyte from measured tissue spectra, comprising the steps of:

providing a calibration set, wherein said calibration set comprises a data set of exemplar spectral measurements from a representative sampling of a subject population;

in at least one tier, classifying said exemplar measurements into previously defined classes; and

extracting at least one feature from said exemplar measurements for still further classification; and

calculating at least one localized calibration model based on said classified exemplar measurements and a set of associated reference values.

19. The method of claim 18, wherein said classifying step is based on any of:

abstract and simple features.

20. The method of claim 18, further comprising the step of mapping said exemplar measurements to estimates of said analyte based on either a linear or a nonlinear model.

21. The method of claim 18, wherein said classifying step is based on any of:

a prioria priori information; and

at least one instrumental measurement at a tissue measurement site at which optical samples were taken for said spectral measurements.

22. The method of claim 18, wherein said classifying step comprises multiple tiers.

23. The pattern classification method of claim 22, wherein said classifying step comprises any of the steps of:

classifying said exemplar measurements into previously defined classes based on subject'"'"'s age;

classifying said exemplar measurements into previously defined classes based on subject'"'"'s sex;

classifying said exemplar measurements into previously defined classes based on an estimation of stratum corneum hydration of said tissue measurement site; and

classifying said exemplar measurements into previously defined classes based on skin temperature at said tissue measurement site.

24. A method for developing a calibration model for estimating a target analyte property from measured tissue spectra, comprising the steps of:

providing a data set of exemplar spectral measurements from a sampling of a subject population;

classifying a majority of said exemplar measurements into classes using at least one feature of said exemplar measurements;

wherein said feature comprises a spectral feature,

wherein said classes comprise groups of measurements wherein similarity between measurements within a group is greater than similarity between groups, and

calculating at least one localized calibration model using said classified measurements and an associated set of reference values.

25. The method of claim 24, wherein said classifying step comprises classifying based on any of:

a priori information;

a physical measurement; and

an optical measurement at a tissue measurement site.

26. The method of claim 25, wherein said a priori information comprises any of:

age;

gender;

hematocrit level; and

temperature.

27. The method of claim 25, wherein said physical measurement comprises any of:

thickness of adipose tissue;

tissue hydration;

scattering properties of said tissue; and

skin thickness.

28. The method of claim 25, wherein said optical measurement comprises any of:

magnitude of protein absorbance;

magnitude of fat absorbance;

a spectral characteristic;

a pathlength estimate;

volume fraction of blood in tissue; and

a spectral feature.

29. The method of claim 25, wherein said classes at least partially share exemplar measurements.

30. The method of claim 25, further comprising the step of:

assigning degree of membership to at least some of said exemplar measurements according to a fuzzy membership function.

31. The method of claim 30, wherein at least one of said localized calibration models comprises coefficients calculated with exemplar measurements and said degree of membership.

32. The method of claim 31, further comprising the steps of:

providing an estimation spectrum;

assigning degree of class membership to said estimation spectrum in at least one of said classes;

estimating at least one interim analyte property with said localized calibration models; and

combining said estimates to determine said analyte property.

33. The method of claim 32, wherein said step of assigning comprises use of a fuzzy membership function.

34. The method of claim 32, wherein said step of combining uses said degree of class membership.

35. The method of claim 24, wherein said classifying step comprises:

classifying said exemplar measurements into previously defined classes based on at least one instrument measurement at a tissue measurement site.

36. The method of claim 24, wherein said feature extraction comprises the steps of:

representing structural properties and physiological state of a tissue measurement site through application of at least one mathematical transformation that enhances a quality or aspect of sample measurement for interpretation, and

using a resulting set of features i to classify a subject and determine a calibration model that is most useful for blood analyte prediction.

37. The method of claim 36, wherein said step of representing structural properties and physiological state comprises the step of:

representing features in a vector, zε

^M that is determined from a preprocessed measurement through;

z=f(λ

,x)

 

where f;

^N→

^M is a mapping space to a feature space, wherein decomposing f(•

) yields specific transformations, f_i(•

);

^N→

^M_i for determining a specific feature, wherein the dimension M_i indicates whether an i^th feature is a scalar or a vector and an aggregation of all features is the vector z.

38. The method of claim 24, wherein said feature exhibits a structure indicative of an underlying physical phenomenon when said feature is represented as a vector or a pattern.

39. The method of claim 24, wherein said feature comprises any of:

a simple feature; and

an abstract feature.

40. The method of claim 24, wherein a decision rule makes class assignments.

41. The method of claim 24, wherein said features comprise sets of features and wherein the step of defining classes in a supervised manner comprises the steps of:

dividing each set of features into two or more regions, wherein classes are defined by combinations of feature divisions, wherein classes are defined through known differences in data;

performing a cluster analysis on the exemplar measurements to determine groups of said defined classes that can be combined to reduce the final number of class definitions;

designing a classifier subsequent to class definition through supervised pattern recognition by determining an optimal mapping or transformation from the feature space to a class estimate that minimizes the number of misclassifications; and

creating a model based on class definitions that transforms a measured set of features to an estimated classification, wherein said class definitions are optimized to satisfy specifications of a measurement system used to take said measurements.

42. The method of claim 41, further comprising:

calculating weights, W, for said measurements, according to;

W=(X^TX)<sup>−

1</sup>X^TY,

 

where X represents a matrix of measurements, and Y represents a reference value of a target analyte concentration for each measurement.

43. The method of claim 42, wherein a vector of weights of spectral measurements within one of said groups comprises a regression vector for said group;

and

wherein said regression vector comprises a calibration model for said group.

44. The method of claim 24, wherein the steps of defining said classes in an unsupervised manner comprises:

developing clusters of data in feature space based on the measurements, wherein within-cluster homogeneity and between-cluster separation is maximized.

45. The method of claim 44, wherein clusters formed from features having physical meaning are interpreted based on the known underlying phenomenon causing variation in the feature space.

46. The method of claim 24, wherein said classes are defined on the basis of structural and state similarity, wherein variation in tissue characteristics within a class is smaller than the variation between classes.

47. The method of claim 24, wherein said classifying step is based on any of:

a simple feature; and

an abstract feature.

48. The method of claim 24, further comprising the step of:

preprocessing prior to said step of classifying.

49. A method for developing a calibration model for estimating a target analyte property from measured tissue spectra, comprising the steps of:

providing a data set of exemplar spectral measurements from a sampling of a subject population;

classifying a majority of said exemplar measurements into classes using at least one feature of said exemplar measurements; and

calculating at least one localized calibration model using said classified measurements and an associated set of reference values,

wherein the step of classifying comprises classifying through at least two tiers.

50. A method for developing a calibration model for estimating a target blood analyte property from measured tissue spectra, comprising the steps of:

providing a calibration set, wherein said calibration set comprises a data set of exemplar spectral measurements from a representative sampling of a subject population;

extracting at least one feature from at least one of said exemplar measurements;

classifying at least a portion of said exemplar measurements into classes using said feature; and

calculating at least one localized calibration model for at least one of said classes based on said classified measurements and an associated set of reference values,

wherein said step of extracting at least one feature comprises;

representing structural properties and physiological state of a tissue measurement site through application of at least one mathematical transformation that enhances a quality or aspect of sample measurement for interpretation, wherein a resulting set of features is used to classify a subject and determine a calibration model.

51. The method of claim 50, wherein said feature comprises a spectral feature.

52. The method of claim 50, wherein the step of classifying comprises classifying based on any of:

a priori information;

a physical measurement; and

an optical measurement of a tissue measurement site.

53. The method of claim 50, wherein the step of classifying measurements comprises:

classifying said exemplar measurements into previously defined classes based on at least one instrument measurement at a tissue measurement site.

54. The method of claim 50, wherein said feature comprises any of:

a simple feature; and

an abstract feature.

55. The method of claim 50, wherein the step of classifying comprises classifying said exemplar measurements, wherein said classes are defined in any of supervised and unsupervised manners.

56. The method of claim 50, wherein the step of extracting comprises a mathematical transformation resulting in any of:

a simple feature; and

an abstract feature.

57. The method of claim 50, wherein said classes at least partially share exemplar measurements.

58. The method of claim 50, wherein the step of classifying comprises classifying through at least two tiers.

59. The method of claim 50, wherein said classes are previously defined.

60. The method of claim 50, further comprising the step of:

preprocessing prior to said step of extracting.

61. The method of claim 50, wherein the step of classifying uses any of:

a crisp function; and

a fuzzy function.

62. A method for developing a calibration algorithm for calculating concentration of a target blood analyte from measured tissue spectra, comprising the steps of:

providing a data set of exemplar spectral measurements from a representative sampling of a subject population;

classifying at least one of said exemplar measurements into previously defined classes; and

calculating at least one localized calibration model using said classified measurements and an associated set of reference values,

wherein said classes comprise groups of measurements, wherein similarity between measurements within a group is greater than similarity between groups.

63. The method of claim 62, wherein said classes are defined by any of:

a priori information;

a physical measurement; and

an optical measurement at a tissue measurement site.

64. The method of claim 63, wherein said a priori information comprises any of:

age;

gender;

hematocrit level; and

temperature.

65. The method of claim 63, wherein said physical measurement comprises any of:

thickness of adipose tissue;

tissue hydration;

scattering properties of said tissue; and

skin thickness.

66. The method of claim 63, wherein said optical measurement comprises any of:

magnitude of protein absorbance;

magnitude of fat absorbance;

a spectral characteristic;

a pathlength estimate;

volume fraction of blood in tissue; and

a spectral feature.

67. The method of claim 62, wherein a decision rule makes class assignments.

68. A method for developing a multi-tier calibration model for determining concentration of a target blood analyte from measured tissue spectra, comprising the steps of:

providing a calibration set, wherein said calibration set comprises a data set of exemplar spectral measurements from a representative sampling of a subject population;

through at least two tiers, classifying said exemplar measurements into classes; and

calculating at least one localized calibration model using said classified measurements and an associated set of reference values.