Intelligent system for detecting errors and determining failure modes in noninvasive measurement of blood and tissue analytes

• US 6,788,965 B2
• Filed: 08/01/2002
• Issued: 09/07/2004
• Est. Priority Date: 08/03/2001
• Status: Active Grant

• Current Assignee: GLT Acquisition Corporation
• Original Assignee: Sensys Medical, Inc. (Sensys Medical Limited)
• Inventors: Makarewicz, Marcy, Briggs, Christopher C., Ruchti, Timothy L., Blank, Thomas B., Lorenz, Alexander D., Mattu, Mutua
• Primary Examiner(s): WINAKUR, ERIC FRANK
• Application Number:
  US10/211,478
• Publication Number:
  US 20030060692A1
• Time in Patent Office: 768 Days
• Field of Search: 600/310, 600/322, 600/316, 600/323, 600/326, 600/336, 600/473, 600/476, 356/39, 356/300, 128/920, 128/925, 702/19
• US Class Current: 600/316
• CPC Class Codes: A61B 5/0064 Body surface scanning
• CPC Class Codes: A61B 5/0075 by spectroscopy, i.e. measuring spectra, e.g. Raman spectroscopy, infrared absorption spectroscopy A61B5/0071 takes precedence
• CPC Class Codes: A61B 5/01 Measuring temperature of body parts ; Diagnostic temperature sensing, e.g. for malignant or inflamed tissue clinical contact thermometers G01K13/20
• CPC Class Codes: A61B 5/14532 for measuring glucose, e.g. by tissue impedance measurement
• CPC Class Codes: A61B 5/14546 for measuring analytes not otherwise provided for, e.g. ions, cytochromes
• CPC Class Codes: A61B 5/1459 invasive, e.g. introduced into the body by a catheter A61B5/1464 takes precedence
• CPC Class Codes: A61B 5/1495 Calibrating or testing of in-vivo probes
• CPC Class Codes: A61B 5/441 Skin evaluation, e.g. for skin disorder diagnosis
• 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: G06N 3/004 Artificial life, i.e. computing arrangements simulating life
• CPC Class Codes: G16H 50/20 for computer-aided diagnosis, e.g. based on medical expert systems

Abstract

An intelligent system for detecting errors and determining failure modes operates on an absorbance spectrum of in vivo skin tissue. Application of the system results in improved prediction accuracy through rejection of invalid and poor samples. System components include a noninvasive blood glucose meter, such as a near IR spectrometer, an error detection system (EDS); a system for diagnosing and mitigating errors; and a reporting method. In the EDS, a pattern classification engine and hierarchy of levels analyzes, detects and diagnoses instrument, interface and sample errors manifested in the spectrum to determine suitability of an absorbance spectrum for blood glucose measurement. The final component of the system evaluates the error condition, diagnoses the specific mode of failure (if necessary) and reports actions to be taken. Sub-components and levels of the EDS can operate independently of the other system elements to the benefit of a noninvasive glucose measurement system.

Claims

1. An intelligent system for detecting errors in spectroscopic determination of blood and/or tissue analytes comprising:

an apparatus for measuring a spectrum at a selected tissue site on a subject; and

an error detection system (EDS), said EDS comprising any of;

one or more individual modules, a module embodying at least one of a plurality of methods for detecting errors in spectral measurements; and

one or more subsystems, a subsystem comprising a plurality of said individual modules having a commonality;

wherein said one or more modules and said one or more subsystems are combined to form;

a hierarchic system comprising at least one level, a level comprising one of;

at least one subsystem, at least one module, and a combination of at least one subsystem and at least one module;

said EDS further comprising an error manager;

wherein said intelligent system detects conditions unsuitable for analyte determination.

2. The system of claim 1, wherein said modules output an acceptability measure, each module having a range of acceptability specified for its associated acceptability measure, wherein said ranges of acceptability are based on a set of exemplary samples, the set containing both erroneous measurements of known origin and valid measurements;

and

wherein an acceptability measure outside of its associated range of acceptability constitutes an error.

3. The system of claim 1, wherein said apparatus for measuring a spectrum comprises a spectrometer instrument.

4. The system of claim 1, wherein said analyte comprises glucose.

5. The system of claim 1, wherein measurement is any of noninvasive and in vivo.

6. The system of claim 1, further comprising a preprocessing and feature extraction system.

7. The system of claim 6, wherein said preprocessing and feature extraction system comprises:

means for low-level processing of raw intensity spectra, spectra including sample spectra and reference spectra;

means for averaging processed intensity spectra;

means for calculating an absorbance spectrum from said averaged, processed intensity spectra;

means for preprocessing said absorbance spectrum;

means for extracting features from a preprocessed absorbance spectrum;

a tissue template set, the tissue template comprising a set of measurements taken at onset of a measurement period; and

a calibration model, said calibration model applied to any of said preprocessed absorbance spectrum and extracted features to measure said analyte, said calibration model also applied to estimate relative precision of a measurement and/or determine certainty of a measurement in view of past measurements.

8. The system of claim 7, wherein low-level processing includes

subtraction of an electrical offset from said raw intensity spectra.

9. The system of claim 7, wherein averaging includes any of:

simple average calculation; and

robust estimation of mean intensity at each wavelength.

10. The system of claim 7, wherein said absorbance spectrum comprises a specific set of wavelengths in the near IR region that has been optimized for feature extraction and measurement.

11. The system of claim 7, wherein preprocessing includes any of:

scaling;

normalizing;

smoothing;

derivatizing;

filtering; and

transformations that attenuate noise and instrumental variation without unduly affecting signal of interest.

12. The system of claim 7, wherein said features include any of:

simple features, comprising values of a processed spectral measurement at which slope equals zero;

derived features, comprising additional features derived from simple features through mathematical transformation; and

abstract features developed through linear and/or nonlinear transformations of said preprocessed spectrum.

13. The system of claim 12, wherein simple features include any of:

critical points;

normalization points;

fat band points;

protein band points; and

water band points.

14. The system of claim 13, wherein derived features are mathematically derived from said critical, normalization, fat band, protein band, and water band points.

15. The system of claim 12, wherein abstract features do not have a specific interpretation related to a physical system, abstract features including scores from a principal components analysis.

16. The system of claim 7, wherein said tissue template set includes any of:

intensity spectra for sample and reference;

an absorbance spectrum;

a preprocessed absorbance spectrum;

and a set of extracted features.

17. The system of claim 1, said intelligent system further comprising a state classification and rule-based decision system.

18. The system of claim 17, wherein said state classification and rule-based decision system comprises:

means for collecting acceptability measures from said modules;

means for normalizing said acceptability measures;

a classifier, wherein said classifier determines at least one operating state for each of said levels; and

a rule-based decision engine, wherein a decision is made regarding acceptability of an analyte measurement based on said determined operating states.

19. The system of claim 18, wherein said classifier is developed from a data set of exemplary spectral measurements from a representative population sample.

20. The system of claim 19, wherein classes are defined by assigning said measurements from said data set to classes.

21. The system of claim 19, wherein class definition employs a supervised approach, wherein classes are defined through known differences in the data.

22. The system of claim 19, wherein class definition employs an unsupervised approach, wherein said acceptability measures are used to explore and develop clusters of the data in feature space, so that within cluster homogeneity and between cluster separation is optimized.

23. The system of claim 22, wherein a large set of samples from a multiplicity of subjects is used to create a database, the database including acceptability measures, measurements, and reference measurements.

24. The system of claim 23, wherein acceptability measures are sorted according to level of sophistication.

25. The system of claim 24, wherein an abstract factor analysis is performed to account for redundancy of information presented by said acceptability measures;

and

wherein a cluster analysis is performed to identify classes, said classes constituting states, that are associated with various levels of measurement error.

26. The system of claim 25, wherein states are combined to provide an error diagnosis.

27. The system of claim 22, wherein said data set is created through introduction of errors associated with defined categories, and wherein a statistical classifier is used to map said acceptability measures to a final decision.

28. The system of claim 22, wherein said classifier is designed by determining an optimal mapping from feature space to a class estimate that minimizes misclassifications.

29. The system of claim 22, wherein statistically based class definitions provide crisp class definitions.

30. The system of claim 19, wherein class assignment and decisions are based on fuzzy set theory.

31. The system of claim 30, wherein membership in fuzzy sets is defined by a continuum of grades and a set of membership functions that map feature space into an interval for each class, wherein assigned membership grade represents a degree of class membership, so that a sample can simultaneously be a member of more than one class.

32. The system of claim 1, wherein said modules include any of:

an online error check module;

an instrument error-detection module;

an instrument QC (quality control)-checking module;

a signal-processing module;

a sampling error-detection module;

a spectral anomaly-detection module;

surface contact error-detection module;

a hydration-checking module;

a sample variation-detection module;

a sample transient-detection module;

a patient skin temperature-measuring module;

a sample consistency-assessment module;

a sample stability-assessment module;

a tissue transient-detection module;

a skin temperature transient-detection module;

a data consistency module;

a sample structure variation-detection module;

an instrument drift-monitoring module;

an instrument stability-monitoring module;

an instrument performance-monitoring module;

a classification module;

a calibration set-comparison module;

an instrument operation-comparison module;

a measurement precision-estimation module;

a measurement range-assessment module; and

an expected value-prediction module.

33. The system of claim 32, wherein subsystems are defined according to sophistication of included modules, said subsystems including any of:

a low-level subsystem;

a mid-level subsystem; and

a high-level subsystem.

34. The system of claim 33, wherein said low-level subsystem includes modules for testing data immediately after collection of reference and sample spectra, said spectra comprising intensity spectra.

35. The system of claim 34, wherein testing is based on acceptability specifications for noninvasive glucose measurement, and wherein an action resulting from deviation from a specified level of acceptability includes any of:

rejection of a collected spectrum;

rejection of a tissue sample; and

generation of an instrument malfunction error.

36. The system of claim 33, wherein said low-level subsystem includes any of:

the online error check module;

the instrument error detection module;

the instrument QC-checking module;

the signal processing module; and

the sampling error detection module.

37. The system of claim 33, wherein a system manager inputs instrument performance specifications and target spectra for each type of material to be scanned to said low-level subsystem.

38. The system of claim 37, wherein said specifications include any of:

noise limits;

minimum operating temperature limits;

maximum signal levels;

wavelength accuracy limits; and

precision limits.

39. The system of claim 33, wherein said mid-level subsystem comprises a plurality of sublevels, said sublevels corresponding to type of information necessary for analysis.

40. The system of claim 33, wherein said mid-level subsystem uses a tissue template to determine if instrument performance and/or a sampled tissue volume have changed relative to an earlier time.

41. The system of claim 33, wherein said mid-level subsystem includes any of:

the spectral anomaly-detection module;

the surface contact error-detection module;

the hydration-checking module;

the sample variation-detection module;

the sample transient-detection module

the patient skin temperature-measuring module

the sample consistency-assessment module;

the sample stability-assessment module

the tissue transient-detection module;

the skin temperature transient-detection module;

the data consistency module;

the sample structure variation-detection module;

the instrument drift-monitoring module;

the instrument stability-monitoring module; and

the instrument performance-monitoring module.

42. The system of claim 33, wherein actions taken by said mid-level subsystem include any of:

instrument maintenance;

recollect tissue template;

sample rejection; and

instrument QC check.

43. The system of claim 33, wherein said high-level subsystem relies on:

a calibration model and parameters relating to the calibration model;

a patient database;

patient history;

a tissue template; and

measurement specifications common to all instruments and all patients.

44. The system of claim 33, wherein said high-level subsystem includes any of:

the classification module;

the calibration set-comparison module;

the instrument operation-comparison module;

the measurement precision-estimation module;

the measurement range-assessment module; and

the expected value-prediction module.

45. The system of claim 33, wherein actions taken by said high-level subsystem include any of:

change calibration model;

recalibrate patient;

instrument failure;

invalid glucose measurement;

instrument maintenance;

recollect tissue template;

sample rejection; and

instrument QC check.

46. The system of claim 33, wherein levels of said hierarchic system receive and inherit information from lower levels.

47. The system of claim 46, wherein errors generated at each level are inherited by succeeding levels for error diagnosis until a critical error is encountered.

48. The system of claim 46, wherein a composite of acceptability measures from each module is input to a state classification and decision system to diagnose specific source of said error.

49. The system of claim 48, further comprising a database of corrective instructions.

50. A method for detecting errors in spectroscopic measurement of blood and/or tissue analytes comprising the steps of:

measuring at least one spectrum at a selected tissue site on a subject;

applying one or more individual methods for detecting errors in spectral measurements to said measured spectrum, wherein said methods are implemented individually, in subsystems according to a commonality among selected methods, or in a hierarchy wherein a level of said hierarchy implements any of;

at least one method, at least one subsystem, and a combination of at least one method and at least one subsystem;

collecting output of said methods;

detecting conditions inconsistent with analyte determination based on said output;

recording generated errors; and

reporting a decision regarding acceptability of a measurement.

51. The method of claim 50, further comprising the steps of:

outputting an acceptability measure by each method; and

defining a range of acceptability for each acceptability measure wherein ranges of acceptability are based on a set of exemplary samples, the set including both erroneous measurements of known origin and valid measurements; and

generating an error when an acceptability measure is outside of its associated range of acceptability.

52. The method of claim 50, wherein said analyte comprises glucose.

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

processing raw spectra, spectra including sample spectra and reference spectra;

averaging processed intensity spectra;

calculating an absorbance spectrum from said processed intensity spectra;

preprocessing said absorbance spectrum;

extracting features from a preprocessed absorbance spectrum;

generating a tissue template set, the tissue template comprising a set of measurements taken at onset of a measurement period; and

providing a calibration model, said calibration model applied to any of said preprocessed absorbance spectrum and extracted features to measure said analyte, said calibration model also applied to estimate relative precision of a measurement and/or determine certainty of a measurement in view of past measurements.

54. The method of claim 53, wherein processing raw spectra comprises:

subtracting an electrical offset from said raw intensity spectra.

55. The method of claim 53, wherein averaging comprises the steps of:

calculating a simple average; and

producing a robust estimation of mean intensity at each wavelength.

56. The method of claim 53, wherein said absorbance spectrum comprises a specific set of wavelengths in the near IR region that has been optimized for feature extraction and measurement.

57. The method of claim 53, wherein preprocessing comprises:

scaling;

normalizing;

smoothing;

derivatizing;

filtering; and

transformations that attenuate noise and instrumental variation without unduly affecting signal of interest.

58. The method of claim 53, wherein said features include any of:

simple features, comprising values of a processed spectral measurement at which slope equals zero;

derived features, comprising additional features derived from simple features through mathematical transformation; and

abstract features developed through linear and/or nonlinear transformations of said preprocessed spectrum.

59. The method of claim 58, wherein simple features include any of:

critical points;

normalization points;

fat band points;

protein band points; and

water band points.

60. The method of claim 59, wherein derived features are mathematically derived from said critical, normalization, fat band, protein band, and water band points.

61. The method of claim 58, wherein abstract features do not have a specific interpretation related to a physical system, abstract features including scores from a principal components analysis.

62. The method of claim 53, wherein said tissue template set includes any of:

intensity spectra for sample and reference;

an absorbance spectrum;

a preprocessed absorbance spectrum;

and a set of extracted features.

63. The method of claim 50, wherein said outputs constitute acceptability measures, and wherein said step of collecting said outputs comprises:

inputting said acceptability measures to a state classification and rule-based decision system.

64. The method of claim 63, further comprising the step of:

normalizing said acceptability measures;

determining at least one operating state for each of said levels; and

making a decision regarding acceptability of an analyte measurement based on said determined operating states.

65. The method of claim 64, wherein the step of determining at least one operating state for each of said levels comprises the step of:

providing a classifier, wherein said classifier is developed from a data set of exemplar spectral measurements from a representative population sample.

66. The method of claim 65, the step of determining at least one operating state further comprising the step of:

defining classes by assigning said measurements from said data set to classes.

67. The method of claim 66, wherein the step of defining classes comprises:

defining classes through known differences in the data, based on a supervised approach.

68. The method of claim 66, wherein the step of defining classes comprises:

exploring and developing clusters of the data in feature space, wherein class definition employs an unsupervised approach, wherein said acceptability measures are used to, so that within cluster homogeneity and between cluster separation is optimized.

69. The method of claim 68, wherein a large set of samples from a multiplicity of subjects is used to create a database, the database including acceptability measures, measurements, and reference measurements.

70. The method of claim 68, wherein acceptability measures are sorted according to level of sophistication.

71. The method of claim 68, wherein the step of defining classes further comprises:

performing an abstract factor analysis to account for redundancy of information presented by said acceptability measures; and

performing a cluster analysis to identify classes, said classes constituting states, associated with various levels of measurement error.

72. The method of claim 71, further comprising the step of

combining states to provide an error diagnosis.

73. The method of claim 65, wherein said data set is created through introduction of errors associated with defined categories, and wherein a statistical classifier is used to map said acceptability measures to a final decision.

74. The method of claim 73, wherein said classifier is designed by determining an optimal mapping from feature space to a class estimate that minimizes misclassifications.

75. The method of claim 65, wherein statistically based class definitions provide crisp class definitions.

76. The method of claim 65, wherein class assignment and decisions are based on fuzzy set theory.

77. The method of claim 76, wherein membership in fuzzy sets is defined by a continuum of grades and a set of membership functions that map feature space into an interval for each class, wherein assigned membership grade represents a degree of class membership, so that a sample can simultaneously be a member of more than one class.

78. The method of claim 50, wherein said individual methods include any of:

an online error check method;

an instrument error-detection method;

an instrument QC (quality control)-checking method;

a signal-processing method;

a sampling error-detection method;

a spectral anomaly-detection method;

surface contact error-detection method;

a hydration-checking method;

a sample variation-detection method;

a sample transient-detection method;

a patient skin temperature-measuring method;

a sample consistency-assessment method;

a sample stability-assessment method;

a tissue transient-detection method;

a skin temperature transient-detection method;

a data consistency method;

a sample structure variation-detection method;

an instrument drift-monitoring method;

an instrument stability-monitoring method;

an instrument performance-monitoring method;

a classification method;

a calibration set-comparison method;

an instrument operation-comparison method;

a measurement precision-estimation method;

a measurement range-assessment method; and

an expected value-prediction method.

79. The method of claim 78, wherein said online error check method comprises the step of:

determining whether proper material has been scanned and whether material characteristics are similar to previously set standards.

80. The method of claim 78, wherein said instrument error-detection method comprises the step of:

detecting errors based on a series of tests that evaluate signal levels compared to a target range at particular wavelengths indicative of failure modes.

81. The method of claim 78, wherein failure modes include any of:

illumination system failure;

excessive instrument temperature;

damaged illumination/detection elements; and

excessive changes in light intensity.

82. The method of claim 78, wherein the instrument QC method comprises the step of:

determining if the instrument is operating according to instrument specifications and reporting an error if not.

83. The method of claim 78, wherein said specifications include specifications for any of:

instrument noise (at each wavelength and overall);

peak signal level; and

x-axis variation.

84. The method of claim 78, wherein said signal-processing method comprises any of the steps of:

applying baseline correction;

applying ensemble averaging;

applying wavelength standardization;

applying finite impulse response filtering (FIR);

differentiating;

applying multiplicative scatter correction;

applying standard normal transformation; and

calculating absorbance;

wherein all scanned materials are processed into a set of spectra that can be used to perform error detection and measure blood glucose.

85. The method of claim 78, wherein said sampling error-detection method comprises the step of:

detecting gross sampling errors, gross sampling errors including any of;

lifting or moving body part bearing measurement site during scanning;

moving a reference during scanning; and

improper application of a coupling medium.

86. The method of claim 78, wherein said spectral anomaly detection method comprises the step of:

detecting changes in relative absorbance of constituents at various depths corresponding to modification of a sampled tissue volume.

87. The method of claim 78, wherein said surface contact error detection method comprises the steps of:

extracting spectral features related to surface contact and comparing said extracted features to related features from a tissue template spectrum, wherein an error condition results if surface contact deviates significantly from said tissue template or from an a priori level.

88. The applying of claim 78, wherein said hydration-checking method comprises the steps of:

extracting spectral features related to sample hydration;

comparing said extracted features to features from a tissue template and previously calculated features; and

reporting an error condition if patient'"'"'s hydration has changed significantly from that of tissue template.

89. The method of claim 78, wherein said sample variation-detection method comprises the step of:

detecting mechanical distortion of an optically sampled tissue volume, wherein extracted features from a sample spectrum are compared to features from a tissue template by means of a distance measure, and

reporting an error if said sample spectrum exceeds a pre-set limit.

90. The method of claim 78, wherein said sample transient-detection method comprises the step of:

detecting rapid changes in coupling between a patient interface module and sample site, or rapid change in the sample by comparing extracted features with a previously established library representing a plurality of error conditions, error conditions including;

arm movement while scanning;

perspiration;

excessive pressure;

poor patient-instrument coupling;

tissue distortion;

tissue heating;

rapid changes in patient physiology;

fluid displacement in the tissue; and

poor sampling conditions.

91. The method of claim 78, wherein said patient skin temperature-measuring method comprises either of the steps of:

detecting patient skin temperature through a direct measurement with a skin temperature probe;

or

detecting patient skin temperature spectroscopically based on a predetermined skin temperature calibration model;

wherein the skin temperature is compared to skin temperature associated with a tissue template, and temperatures that vary from a target range produce an error condition.

92. The method of claim 78, wherein said sample consistency-assessment method comprises the step of:

comparing a sample spectrum with prior patient scans and calculating similarity;

wherein an error condition results if the sample spectrum differs substantially from the prior scans.

93. The method of claim 78, wherein said sample stability-assessment module comprises the step of:

comparing level of analyte marker bands with pre-set limits by comparing a processed sample spectrum with a tissue template spectrum over a selected wavelength range and calculating similarity between the two spectra;

wherein an error condition results if the sample spectrum differs substantially from the tissue template.

94. The method of claim 78, wherein said tissue transient-detection method comprises the step of:

examining time history of sample spectra features related to any of;

changes in physiological state;

changes in local tissue morphology; and

patient-instrument coupling errors;

wherein an error condition occurs if range and time-related correlation of a feature exceeds a predetermined limit.

95. The method of claim 78, wherein said skin temperature transient-detection method comprises the step of:

comparing range and time correlation of measured skin temperature over samples collected subsequent to collecting a tissue template, wherein a sample that is either out of range or that displays a unidirectional temperature change results in an error condition.

96. The method of claim 78, wherein said data consistency method comprises the step of:

comparing sample spectra with historical data;

wherein an error condition results if variability within the sample spectrum is inconsistent with said historical data.

97. The method of claim 78, wherein said sample structure variation-detection method comprises the step of:

comparing a tissue fingerprint of a newly collected sample spectrum with that of an original stored tissue template, said tissue fingerprint comprising a plurality of distinct features related to analyte absorption;

wherein an error condition results if the newly collected sample spectrum differs substantially from said tissue template, an error indicating a gross change in sampled tissue volume.

98. The method of claim 78, wherein said instrument drift-monitoring method comprises the step of:

comparing performance parameters with data collected with a tissue template;

wherein an error condition results if an instrument is outside of normal instrument operating procedures.

99. The method of claim 78, wherein said instrument stability-monitoring method comprises the step of:

comparing sample spectra with data from a history file containing instrument parameters, parameters including any of RMS (root mean squared) noise, wavelength shift and signal intensity;

wherein an error condition occurs if instrument performance has changed over a short time period.

100. The method of claim 78, wherein said instrument performance-monitoring method comprises the step of:

predicting potential instrument-related failure based on monitoring changes in any of instrument noise, temperatures, wavelength stability, and signal intensity over life of the instrument.

101. The method of claim 78, wherein said classification method comprises the step of:

determining a suitable patient class for calibration on the basis of a calibration database;

wherein sample spectral falling outside a classified set are classed as outliers so that an error conditions results.

102. The method of claim 78, wherein said calibration set-comparison method comprises the step of:

comparing a sample spectrum with a calibration set using any of feature extraction and cluster analysis to determine consistency of the sample spectrum with spectral data used to generate a calibration model, wherein spectra falling outside a range of calibration create an error condition.

103. The method of claim 78, wherein said instrument operation-comparison method comprises the step of:

comparing operational state of the instrument with similar instruments, including an instrument used to collect calibration data, wherein history information from said similar instruments is pooled to compare characteristics of the instrument with the similar instruments.

104. The method of claim 78, wherein said measurement precision-estimation method comprises the step of:

applying a calibration model to each of a set of tissue absorbance replicates to compare any of range, trend and standard deviation of glucose measurements based on said replicates with a maximum acceptable value;

wherein measurements exceeding said value are rejected, so that a confidence estimate of a final, averaged glucose measurement is produced.

105. The method of claim 78, wherein said measurement range-assessment method comprises the step of:

averaging glucose measurements associated with a set of replicate spectral scans to yield a final glucose measurement;

wherein an error condition results if said final measurement is outside a preset range.

106. The method of claim 78, wherein said expected value-prediction method comprises the steps of:

applying a prediction module to predict erroneous glucose measurements, wherein said prediction model uses a time series of past glucose measurements to extrapolate a future prediction; and

comparing said predication with a measurement based on a newly acquired sample spectrum;

wherein an error results if a large discrepancy exists between the measurement from the newly acquired spectrum and the prediction, where the prediction has a high degree of certainty.

107. The method of claim 78, wherein subsystems are defined according to sophistication of included methods, said subsystems including any of:

a low-level subsystem;

a mid-level subsystem; and

a high-level subsystem.

108. The method of claim 107, wherein said low-level subsystem includes methods for testing data immediately after collection of reference and sample spectra, said spectra comprising intensity spectra.

109. The method of claim 107, wherein testing is based on acceptability specifications for noninvasive glucose measurement, and wherein an action resulting from deviation from a specified level of acceptability includes any of:

rejection of a collected spectrum;

rejection of a tissue sample; and

generation of an instrument malfunction error.

110. The method of claim 107, wherein said low-level subsystem includes any of:

the online error check method;

the instrument error detection method;

the instrument QC-checking method;

the signal processing method; and

the sampling error detection method.

111. The method of claim 107, wherein a system manager inputs instrument performance specifications and target spectra for each type of material to be scanned to said low-level subsystem.

112. The method of claim 111, wherein said specifications include any of:

noise limits;

minimum operating temperature limits;

maximum signal levels;

wavelength accuracy limits; and

precision limits.

113. The method of claim 107, wherein actions taken by said low-level subsystem include any of:

sample rejection; and

instrument QC check.

114. The method of claim 107, wherein said mid-level subsystem comprises a plurality of sublevels, said sublevels corresponding to type of information necessary for analysis.

115. The method of claim 107, wherein said mid-level subsystem uses a tissue template to determine if instrument performance and/or a sampled tissue volume have changed relative to an earlier time.

116. The system of claim 107, wherein said mid-level subsystem includes any of:

the spectral anomaly-detection method;

the surface contact error-detection method;

the hydration-checking method;

the sample variation-detection method;

the sample transient-detection method;

the patient skin temperature-measuring method;

the sample consistency-assessment method;

the sample stability-assessment method;

the tissue transient-detection method;

the skin temperature transient-detection method;

the data consistency method;

the sample structure variation-detection method;

the instrument drift-monitoring method;

the instrument stability-monitoring method; and

the instrument performance-monitoring method.

117. The method of claim 107, wherein actions taken by said mid-level subsystem include any of:

instrument maintenance;

recollect tissue template;

sample rejection; and

instrument QC check.

118. The method of claim 107, wherein said high-level subsystem relies on:

a calibration model and parameters relating to the calibration model;

a patient database;

patient history;

a tissue template; and

measurement specifications common to all instruments and all patients.

119. The method of claim 107, wherein said high-level subsystem includes any of:

the classification method;

the calibration set-comparison method;

the instrument operation-comparison method;

the measurement precision-estimation method;

the measurement range-assessment method; and

the expected value-prediction method.

120. The method of claim 107, wherein actions taken by said high-level subsystem include any of:

change calibration model;

recalibrate patient;

instrument failure;

invalid glucose measurement;

instrument maintenance;

recollect tissue template;

sample rejection; and

instrument QC check.

121. The method of claim 107, wherein levels of said hierarchic system receive and inherit information from lower levels.

122. The method of claim 121, wherein errors generated at each level are inherited by succeeding levels for error diagnosis until a critical error is encountered.

123. The method of claim 121, wherein a composite of acceptability measures from each module is input to a state classification and decision system to diagnose specific source of said error.

124. The method of claim 123, further comprising the step of:

providing corrective instructions from a database of corrective instructions.