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

• US 20060167350A1
• Filed: 02/23/2005
• Published: 07/27/2006
• Est. Priority Date: 01/27/2005
• Status: Abandoned Application

• Current Assignee: Sensys Medical, Inc. (Sensys Medical Limited)
• Original Assignee: Sensys Medical, Inc. (Sensys Medical Limited)
• Inventors: Thennadil, Suresh, Ruchti, Timothy L., Monfre, Stephen L., Blank, Thomas B.
• Application Number:
  US11/065,223
• Publication Number:
  US 20060167350A1
• Time in Patent Office: Days
• US Class Current: 600/322
• CPC Class Codes: A61B 5/0059 using light, e.g. diagnosis by transillumination, diascopy, fluorescence A61B5/0093, A61B5/1455, A61B5/02416 and A61B5/0261 take 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 5/7264 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: G16H 50/20 for computer-aided diagnosis, e.g. based on medical expert systems

Abstract

A method of multi-tier classification and calibration in noninvasive blood analyte prediction is provided that 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 that have 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 uses 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 classification method for noninvasively determining a target analyte concentration, comprising the steps of:

providing a measured tissue spectrum of a subject;

extracting at least one feature from said spectrum; and

in a least one tier, using said extracted feature to classify said spectrum into at least one class of a set of classes.

2. The method of claim 1, wherein said target analyte concentration comprises glucose concentration.

3. The method of claim 1, wherein said spectrum comprises a near-infrared spectrum.

4. The method of claim 1, wherein said feature comprises a spectral feature.

5. The method of claim 1, wherein said extracting step comprises the step of:

representing structural properties and physiological state of said spectrum by applying at least one mathematical transformation to enhance a quality or aspect of said measured spectrum for interpretation.

6. The method of claim 5, wherein said representing step comprises the step of:

representing features in a vector, zε

^M that is determined from a preprocessed measurement through;

z=f(λ

,x)

where λ

is wavelength and where x is said measured tissue spectrum.

7. The method of claim 1, wherein said feature exhibits a structure indicative of a chemical constituent of said subject.

8. The method of claim 1, wherein said feature comprises any of:

a simple feature; and

an abstract feature.

9. The method of claim 1, wherein said classifying step comprises the step of classifying through at least two tiers.

10. The method of claim 9, wherein said classifying step comprises the step of classifying through at least three tiers.

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

using a decision rule to make class assignments.

12. The method of claim 1, said classes comprising a group of measurements wherein similarity between measurements within a group is greater than similarity between groups.

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

defining said classes on the basis of structural and state similarity;

wherein variation in tissue characteristics within a class is smaller than variation between classes.

14. The method of claim 1, wherein said set of classes comprises previously defined classes.

15. The method of claim 14, wherein said previously defined classes comprise classes based upon previously defined extracted spectral features.

16. The method of claim 1, wherein said classifying step comprises making any of:

a supervised class assignment; and

an unsupervised class assignment.

17. The method of claim 1, wherein said classifying step uses any of:

a crisp function; and

a fuzzy function.

18. The method of claim 1, wherein said classifying step comprises using any of:

a priori information;

a physical measurement of said subject; and

said measured tissue spectrum.

19. The method of claim 18, wherein said a priori information comprises any of:

age;

gender;

hematocrit level;

dermal thickness; and

temperature.

20. The method of claim 18, wherein said physical measurement comprises any of:

thickness of adipose tissue;

tissue hydration;

scattering properties of said tissue; and

skin thickness.

21. The method of claim 18, wherein said classifying step using said measured tissue spectrum comprises using 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.

22. The method of claim 1, wherein said classifying step comprises the step of

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

23. The method of claim 1, wherein said classes are mutually exclusive, wherein variation between classes is described statistically.

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

providing a model for said class; and

estimating said target analyte property using said model.

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

assigning degree of membership of said spectrum to at least two of said classes.

26. The method of claim 25, wherein said assigning step comprises using a fuzzy membership function.

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

assigning degree of class membership to said measured spectrum in at least two of said classes;

providing localized calibration models for said classes where said estimation spectrum has class membership;

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

combining said estimates to determine said analyte property.

28. The method of claim 1, wherein determining said analyte concentration represented by said measured spectrum comprises:

passing said measured spectrum and its class to a calibration wherein said analyte concentration for the measurement is given by;

ŷ

=g(c,x)

wherein g(·

) is the model, c is the class, x is said spectrum, and y is said analyte concentration.

29. The method of claim 1, wherein said target analyte concentration for said spectrum is given by:

ŷ

=g_k(x)

where g_k(·

) is a calibration model associated with the k^th class of said spectrum, x is said spectrum, and y is said target analyte property.

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

preprocessing said spectrum prior to said step of classifying.

31. A pattern classification method for estimating a target analyte property, comprising steps of:

providing a measured tissue spectrum from a subject; and

through at least one tier, classifying said measured spectrum, based upon at least one extracted tissue feature, into at least one class of a set of classes.

32. The method of claim 31, wherein said classifying step comprises the step of classifying based on any of:

a priori information; and

a physical measurement.

33. The method of claim 31, further comprising the step of:

preprocessing said tissue spectrum prior to said step of classifying.

34. The method of claim 31, further comprising the step of:

assigning degree of membership of said spectrum to at least two of said classes.

35. The method of claim 34, wherein said assigning step comprises using a fuzzy membership function.

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

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

providing localized calibration models for said classes where said estimation spectrum has class membership;

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

combining said interim analyte property estimates to determine said analyte property.

37. The method of claim 31, wherein said extracted tissue feature of said spectrum comprises representation with a portion of said spectrum.

38. The method of claim 31, further comprising the step of:

representing said extracted feature representing structural properties and physiological state of said subject by applying at least one mathematical transformation to enhance a quality or aspect of sample measurement for interpretation.

39. The method of claim 31, wherein said feature exhibits a structure indicative of a chemical constituent of said subject.

40. The method of claim 31, wherein said feature comprises any of:

a simple feature; and

an abstract feature.

41. The method of claim 36, wherein said interim analyte property estimates are combined according to said degree of class membership.

42. A pattern classification method for estimating a level of a target analyte comprising steps of:

providing a measured tissue spectrum from a subject;

in at least one tier, classifying said measured spectrum into previously defined classes.

43. The method of claim 42, wherein said previously defined classes comprise classes based upon previously defined extracted spectral features.

44. The method of claim 42, wherein said previously defined classes comprise any of:

age;

gender;

hematocrit level;

temperature;

thickness of adipose tissue;

tissue hydration;

scattering properties of said tissue;

skin thickness;

magnitude of protein absorbance;

magnitude of fat absorbance;

spectral characteristics;

pathlength estimates;

volume fraction of blood in tissue; and

a spectral feature, wherein said feature comprises a portion of said spectrum.

45. A pattern classification method for estimating a target analyte property, comprising the steps of:

providing a measured tissue spectrum representative of tissue from a subject;

in at least one tier, classifying said measured spectrum into a class, wherein said class is one of a plurality of classes;

providing a model for said class associated with said measured spectrum; and

estimating said target analyte property using said model and said class associated with said measured spectrum.

46. The method of claim 45, wherein said classes are mutually exclusive and wherein variation between classes is described statistically.