Intelligent system for noninvasive blood analyte prediction

• US 6,280,381 B1
• Filed: 07/22/1999
• Issued: 08/28/2001
• Est. Priority Date: 07/22/1999
• Status: Expired due to Term

• Current Assignee: GLT Acquisition Corporation
• Original Assignee: Instrumentation Metrics Inc.
• Inventors: Ruchti, Timothy L., Malin, Stephen F.
• Primary Examiner(s): WINAKUR, ERIC FRANK
• Application Number:
  US09/359,191
• Time in Patent Office: 768 Days
• Field of Search: 600/310, 600/322, 600/323, 600/330, 600/331, 600/336, 600/473, 128/920, 128/924, 128/925, 356/402, 250/340, 250/341.1, 250/339.01
• 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
• CPC Class Codes: G16H 50/20 for computer-aided diagnosis, e.g. based on medical expert systems
• CPC Class Codes: Y10S 128/92 Computer assisted medical diagnostics

Abstract

An intelligent system for measuring blood analytes noninvasively operates on a near infrared absorbance spectrum of in vivo skin tissue. An hierarchical architecture employs a pattern classification engine to adapt the calibration to the structural properties and physiological state of the subject as manifested in the absorbance spectrum. A priori information about the primary sources of sample variability are used to establish general categories of subjects. By applying calibration schemes specific to the various categories, the spectral interference is reduced resulting in improved prediction accuracy and parsimonious calibrations. Two classification rules are disclosed. The first rule assumes the classes are mutually exclusive and applies specific calibration models to the various subject categories. The second rule uses fuzzy set theory to develop calibration models and blood analyte predictions. Therefore, each calibration sample has the opportunity to influence more than one calibration model according to its class membership. Similarly, the predictions from more than one calibration are combined through defuzzification to produce the final blood analyte prediction.

Claims

1. A method for compensating for covariation of spectrally interfering species, sample heterogeneity, state variations, and structural variations, comprising the steps of:

providing an intelligent pattern recognition system that is capable of determining calibration models that are most appropriate for a subject at the time of measurement;

developing said calibration models from the spectral absorbance of a representative population of subjects that have been segregated into classes;

defining said classes on the basis of structural and state similarity, wherein variation within a class is small compared to variation between classes;

classifying said subject, wherein classification occurs through extracted features of a tissue absorbance spectrum related to current subject state and structure; and

applying a combination of one or more of said calibration models.

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

defining subpopulations or classes of subjects whose structure and state produce similarly featured NIR absorbance spectra;

wherein said classes have improved homogeneity leading to a reduction in variation related to optical properties and composition of a sample.

3. An intelligent system for measuring blood analytes noninvasively by operating on a near infrared (NIR) absorbance spectrum of in vivo skin tissue, said system comprising:

a pattern classification engine for adapting a calibration model to the structural properties and physiological state of a subject as manifested in said NIR absorbance spectrum; and

means for reducing spectral interference by applying calibration schemes specific to general categories of subjects that have been segregated into classes;

wherein a priori information about primary sources of sample variability is used to establish said general categories of subjects.

4. The system of claim 3, wherein said pattern classification engine applies a classification rule that assumes that said classes are mutually exclusive;

and wherein said pattern classification engine applies specific calibration models to said general subject categories.

5. The system of claim 3, wherein said pattern classification engine uses fuzzy set theory to develop calibration models and blood analyte predictions;

wherein each calibration sample has an opportunity to influence more than one calibration model according to its class membership; and

wherein predictions from more than one calibration are combined through defuzzification to produce a final blood analyte prediction.

6. The system of claim 3, further comprising:

means for measuring blood analytes noninvasively over a diverse population of subjects at various physiological states;

said pattern classification engine classifying subjects according to their state and structure; and

said means for reducing spectral interference by applying a combination of one or more existing calibration models to predict the blood analytes.

7. An intelligent system for measuring blood analytes noninvasively by operating on a near infrared (NIR) absorbance spectrum of in vivo skin tissue, said system comprising:

an execution layer that receives tissue absorbance spectra from an instrument and that performs rudimentary preprocessing;

a coordination layer that performs feature extraction;

a classification system that is used to classify a subject according to extracted features that represent the state and structure of a sample;

wherein predictions from one or more existing calibration models are used to form an analyte estimate based on said classification.

8. The system of claim 7, further comprising:

a management level for receiving said classification and blood analyte prediction, said management level taking action based on the certainty of said estimate, said management level coordinating all algorithmic events, monitoring performance based on class, adapting rules as necessary, and maintaining information regarding system state.

9. The system of claim 7, wherein said classification system uses classes that are mutually exclusive.

10. The system of claim 7, wherein said classification system applies fuzzy set theory to form a classifier and prediction rules which allow membership in more than one class.

11. The system of claim 7, wherein said instrument performs absorbance measurement through any of transmissive, diffuse reflectance, or alternate methods.

12. The system of claim 7, wherein said tissue absorbance spectrum is a vector m=^N of absorbance values pertaining to a set of N wavelengths λ

ε

^N that span the near infrared, and wherein number of necessary wavelengths in said spectrum is a function of cross correlation between a target analyte and interfering species.

13. The system of claim 7, further comprising:

preprocessing means for scaling, normalization, smoothing, calculating derivatives, filtering, and other transformations that attenuate noise and instrumental variation without affecting the signal of interest.

14. The system of claim 13, wherein a preprocessed measurement, x=^N is determined according to:

15. The system of claim 7, wherein an entire spectrum is used for noninvasive applications with significant variation within and between individuals.

16. A pattern recognition method for estimating a concentration of a target blood analyte, comprising the step of:

classifying new spectral measurements into previously defined classes through structural and state similarities as observed in a tissue absorbance spectrum, according to a pattern classification method;

wherein class membership is an indication of which calibration model is most likely to accurately estimate the concentration of the target blood analyte;

said pattern classification method comprising the steps of;

extracting features; and

classifying said features according to a classification model and decision rule.

17. The method of claim 16, wherein said feature extraction step is any mathematical transformation that enhances a particular aspect or quality of data that is useful for interpretation.

18. The method of claim 16, wherein said classification model comprises means for determining a set of similarity measures with predefined classes.

19. The method of claim 18, further comprising the step of:

providing a classification system that assumes that said classes are mutually exclusive and that forces each measurement to be assigned to a single class.

20. The method of claim 18, further comprising the step of:

providing a fuzzy classifier that is not mutually exclusive, wherein said fuzzy classifier allows a sample to have membership in more than one class simultaneously and provides a number between zero and one indicating a degree of membership in each class.

21. The method of claim 20, wherein a calibration model is passed a vector of memberships, where a vector, c, is used to determine an adaptation of said calibration model suitable for blood analyte prediction or an optimal combination of several blood analyte predictions.

22. The method of claim 21, wherein a membership vector and preprocessed absorbance spectrum are both used by a single calibration model for blood analyte prediction, where the calculation is given by:

23. The method of claim 21, wherein separate calibrations are used for each class;

and wherein each calibration is generated using all measurements in a calibration set by exploiting a membership vector assigned to each measurement.

24. The method of claim 23, wherein said membership vector is used to determine an optimal combination of p blood analyte predictions from all classes through defuzzification.

25. The method of claim 24, wherein each of the p calibration models is developed using an entire calibration.

26. The method of claim 25, wherein calibration measurements are weighted by their respective membership in a kth class when a kth calibration model is calculated;

wherein weighted least squares is applied to calculate regression coefficients in a linear case, and wherein a covariance matrix is used in a factor based methods case.

27. The method of claim 24, wherein said defuzzification is a mapping from a vector of blood analyte predictions and a vector of class memberships to a single analyte prediction, wherein said defuzzifier can be denoted as transformation such that:

28. The method of claim 16, wherein said decision rule comprises means for assigning class membership on the basis of a set of measures calculated by a decision engine.

29. The method of claim 16, 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.

30. The method of claim 29, wherein said features are represented in a vector, zε

^M that is determined from a preprocessed measurement through;

31. The method of claim 30, wherein individual features are divided into two 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.

32. The method of claim 31, wherein features that can be calculated from NIR spectral absorbance measurements include 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 related to sex;

pathlength estimates;

volume fraction of blood in tissue; and

spectral characteristics related to environmental influences.

33. The method of claim 16, further comprising the step of:

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

34. The method of claim 16, further comprising the step of:

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

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

35. The method of claim 16, wherein said extraction step determines at least one calibration model that is most appropriate for measurement;

wherein a subject is assigned to one of many predefined classes for which a calibration model has been developed and tested.

36. The method of claim 16, wherein said pattern classification step further comprises the steps of:

measuring the similarity of a features to predefined classes; and

assigning class membership.

37. The method of claim 36, wherein said measuring step uses mutually exclusive classes and assigns each measurement to one class.

38. The method of claim 36, wherein said assigning step uses a fuzzy classification system that allows class membership in more than one class simultaneously.

39. The method of claim 16, further comprising the step of:

assigning measurements in an exploratory data set to classes.

40. The method of claim 39, further comprising the step of:

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

41. The method of claim 40, 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 combinations of feature divisions;

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 which minimizes the number of misclassifications; and

creating a model based on class definitions which transforms a measured set of features to an estimated classification.

42. The method of claim 41, wherein a mapping from feature space to a vector of class memberships is given by:

43. The method of claim 42, wherein blood analyte prediction occurs by application of said combination of one or more calibration models for blood analyte prediction.

44. The method of claim 43, wherein said calibration model comprises either of nonlinear partial least squares or artificial neural networks.

45. The method of claim 43, wherein a blood analyte prediction for a preprocessed measurement x with classification specified by c is given by:

 

ŷ

=g(c,x)

where g(•

) is a nonlinear calibration model which maps x and c to an estimate of the blood analyte concentration, {overscore (y)}.

46. The method of claim 43, wherein a different calibration is realized for each class.

47. The method of claim 46, wherein an estimated class is used to select one of p calibration models most appropriate for blood analyte prediction using a current measurement, wherein given that k is the class estimate for said measurement, blood analyte prediction is:

48. The method of claim 46, therein said calibrations are developed from a set of exemplar absorbance spectra with reference blood analyte values and pre-assigned classification definitions.

49. The method of claim 16, further comprising the step of:

providing an algorithm manager for reporting results to an operator, coordinating all algorithmic events, monitoring performance based on class, and adapting rules as necessary.

50. The method of claim 49, wherein both class estimates and blood analyte predictions are reported to said algorithm manager.

51. The method of claim 49, further comprising the step of:

notifying said operator by said algorithm manager that a prediction is invalid or a measurement does not fit into one of the existing classes;

wherein further spectral measurements are taken to determine if said error is due to an instrument, a measurement technique, or a sample; and

wherein said error detection and correction algorithm determines if more classes are necessary or if said instrument requires maintenance.