Method of and apparatus for determining the similarity of a biological analyte from a model constructed from known biological fluids

Claims

What is claimed:

1. A method of determining noninvasively and in vivo one or more unknown values of known characteristics, such as the concentration of an analyte, of a biological fluid in a mammal, said method comprising the steps of:

(a) irradiating in vivo and noninvasively said biological fluid having said unknown values of said known characteristics with infrared energy having at least several wavelengths so that there is differential absorption of at least some of said wavelengths by said biological fluid as a function of said wavelengths and said characteristics, said differential absorption causing intensity variations of said wavelengths incident from said biological fluid as a function of said wavelengths and said unknown values of said known characteristics;

(b) measuring said intensity variations from said biological fluid; and

(c) calculating said unknown values of said known characteristics in said biological fluid from said measured intensity variations utilizing an algorithm and a mathematical calibration model, said algorithm including all independent sources of intensity variations v. wavelengths information obtained from irradiating a set of samples in which said values of said known characteristics are known, said algorithm also including more wavelengths than samples in said set of samples, said model being constructed from said set of samples and being a function of said known values and characteristics and said intensity variations v. wavelengths information obtained from irradiating said set of samples.

2. The method as set forth in claim 1, wherein said set of samples is irradiated in vivo and noninvasively.

3. The method as set forth in claim 1, wherein said infrared energy is near infrared energy.

4. The method as set forth in claim 1, wherein said algorithm is selected from the group including partial least squares and principal component regression.

5. The method as set forth in claim 1, wherein said biological fluid is irradiated by directing said infrared energy to be incident on a portion of said mammal which includes said biological fluid, so that said biological fluid partially absorbs said incident infrared energy.

6. The method as set forth in claim 5, wherein said infrared energy incident on said portion of said mammal passes through said portion.

7. The method as set forth in claim 6, wherein said portion is a digit.

8. The method as set forth in claim 5, wherein said infrared energy incident on said portion of said mammal is partially absorbed by said portion and partially diffusely reflected from said portion.

9. The method as set forth in claim 8, wherein said portion is a head.

10. The method as set forth in claim 5, wherein said irradiating is accomplished by coupling said infrared energy to said portion of said mammal by a fiber optic device.

11. The method as set forth in claim 10, wherein said infrared energy from said portion of said mammal is transmitted from said portion by said fiber optic device.

12. The method as set forth in claim 10, wherein said infrared energy from said portion of said mammal is transmitted by a second fiber optic device.

13. The method as set forth in claim 11, further including the step of detecting outlier samples in said set of samples.

14. The method as set forth in claim 13, wherein said detection of outlier samples in said set includes the step of comparing said intensity v. wavelength responses of each sample in said set to said model, the comparing step yielding a measure of the magnitude of the difference between said intensity v. wavelength responses of each of said samples and said model.

15. The method as set forth in claim 14, further including performing a statistical test to indicate the probability of said magnitude being caused by random chance, and classifying as outliers those of said samples in said set having an excessively low probability.

16. The method as set forth in claim 15 wherein the statistical test is the F ratio test.

17. The method as set forth in claim 13, wherein said detection of outlier samples in said set includes the step of comparing said known values of said characteristics of each sample in said set to the calculated value of said characteristic, said calculated value being based on an estimate derived from said model of said value of said characteristic for each said sample to determine a measure of the magnitude of the difference between said known value of said characteristic and said calculated value of said characteristic.

18. The method as set forth in claim 17, further including performing a statistical test to indicate the probability of said magnitude being caused by random chance, and classifying as outliers those of said samples in said set having an excessively low probability.

19. The method as set forth in claim 18, wherein the statistical test is the F ratio test.

20. The method as set forth in claim 11, further including the step of determining whether said intensity variations v. wavelengths response of said biological fluid with unknown values of said known characteristics is an outlier.

21. The method as set forth in claim 20, wherein said determination of whether said intensity variations v. wavelengths response of said biological fluid is an outlier includes the step of comparing said intensity variations v. wavelengths from said biological fluid to said model, the comparing step yielding a measure of the magnitude of the difference between said intensity variations v. wavelengths of said biological fluid and said model.

22. The method as set forth in claim 21, further including performing a statistical test to indicate the probability of said magnitude being caused by random chance, and classifying as outliers those of said values having an excessively low probability.

23. The method as set forth in claim 22, wherein the statistical test is the F ratio test.

24. The method as set forth in claim 1, wherein said known characteristics are minor components of said biological fluid and said unknown values are less than 2.0 weight percent of said biological fluid.

25. A method of determining invasively and in vivo one or more unknown values of known characteristics, such as the concentration of an analyte, of a biological fluid in a mammal, said method comprising the steps of:

(a) coupling invasively and in vivo a source of infrared energy with an internal portion of said mammal;

(b) irradiating in vivo and invasively said biological fluid having said unknown values of said known characteristics with said infrared energy having at least several wavelengths so that there is differential absorption of at least some of said wavelengths by said biological fluid as a function of said wavelengths and said characteristics, said differential absorption causing intensity variations of said wavelengths incident from said biological fluid as a function of said wavelengths and said characteristics having unknown values;

(c) measuring said intensity variations from said biological fluid; and

(d) calculating said unknown values of said known characteristics in said biological fluid from said measured intensity variations from said biological fluid utilizing an algorithm and a mathematical calibration model, said algorithm including all independent sources of intensity variations v. wavelengths information obtained from irradiating a set of samples in which said values of said known characteristics are known, said algorithm also being capable of using more wavelengths than samples in said set of samples, said model being constructed from said set of samples and being a function of said known values and characteristics and said intensity variations v. wavelengths information obtained from irradiating said set of samples.

26. The method as set forth in claim 25, wherein said samples are irradiated invasively and in vivo.

27. The method as set forth in claim 25, wherein said algorithm is selected from the group including partial least squares and principal component regression.

28. The method as st forth in claim 25, wherein said source of infrared energy is coupled with said mammal by at least partially implanting a fiber optic device in said mammal.

29. The method as set forth in claim 28, wherein at least a portion of said fiber optic device is used as an attenuated total reflectance (ATR) device.

30. The method as set forth in claim 29, comprising passing said fiber optic device through a portion of said mammal.

31. The method as set forth in claim 25, comprising implanting an ATR crystal in said mammal, said ATR device being coupled to said source of infrared energy.

32. The method as set forth in claim 25, comprising implanting said source of infrared energy and apparatus for measuring said intensity variations in said mammal proximate to a supply of said biological fluid.

33. The method as set forth in claim 32, comprising positioning said source of infrared energy and said measuring apparatus on opposite sides of said biological fluid.

34. The method as set forth in claim 33, comprising positioning said source of infrared energy and said measuring apparatus on opposite sides of at least one blood vessel of said mammal.

35. The method as set forth in claim 25, wherein said infrared energy is either in the mid-infrared or near-infrared spectrum.

36. The method as set forth in claim 25, further including the step of detecting outlier samples in said set of samples.

37. The method as set forth in claim 36, wherein said detection of outlier samples in said set includes the step of comparing said intensity v. wavelength responses of each sample in said set to said model to determine a measure of the magnitude of the difference between said intensity v. wavelength responses of each of said samples and said model.

38. The method as set forth in claim 37, further including performing a statistical test to indicate the probability of said magnitude being caused by random chance, and classifying as outliers those of said samples in said set having an excessively low probability.

39. The method as set forth in claim 38, wherein the statistical test is the F ratio test.

40. The method as set forth in claim 25, wherein said detection of outlier samples in said set includes the step of comparing said known values of said characteristics of each sample in said set to the calculated value of said characteristic, said calculated value being based on an estimate derived from said model of said value of said characteristic for each said sample to determine a measure of the magnitude of the difference between said known value of said characteristic and said calculated value of said characteristic.

41. The method as set forth in claim 40, further including performing a statistical test to indicate the probability of said magnitude being caused by random chance, and classifying as outliers those of said samples in said set having an excessively low probability.

42. The method as set forth in claim 41, wherein the statistical test is the F ratio test.

43. The method as set forth in claim 25, further including the step of determining whether said intensity variations v. wavelengths of said biological fluid with unknown values of said known characteristics is an outlier.

44. The method as set forth in claim 43, wherein said determination of whether said intensity variations v. wavelengths of said biological fluid is an outlier includes the step of comparing said intensity variations v. wavelengths from said biological fluid to said model, said comparing step yielding a measure of the magnitude of the difference between said intensity variations v. wavelengths between said biological fluid and said model.

45. The method as set forth in claim 44, further including performing a statistical test to indicate the probability of said magnitude being caused by random chance, and classifying as outliers those of said values having an excessively low probability.

46. The method as set forth in claim 45, wherein the statistical test is the F ratio test.

47. The method as set forth in claim 25, wherein said known characteristics are minor components of said biological fluid and said unknown values are less than 2.0 weight percent of said biological fluid.

48. A method of determining in vitro one or more unknown values of known characteristics, such as the concentration of an analyte, of biological fluid, said method comprising the steps of:

(a) irradiating in vitro said biological fluid having said unknown values of said known characteristics with infrared energy having at least several wavelengths so that there is differential absorption of at least some of said wavelengths by said biological fluid as a function of said wavelengths and said characteristics, said differential absorption causing intensity variations of said wavelengths incident from said biological fluid as a function of said wavelengths and said characteristics having unknown values;

(b) measuring said intensity variations from said biological fluid; and

(c) calculating said unknown values of said known characteristics in said biological fluid from said measured intensity variations from said biological fluid, utilizing an algorithm and a mathematical calibration model, said algorithm including all independent sources of intensity variations v. wavelengths information obtained from irradiating a set of samples in which said values of said known characteristics are known, said algorithm also including more wavelengths than samples in said set of samples, said model being constructed from said set of samples and being a function of said known values and characteristics, said intensity variations v. wavelengths information being obtained by irradiating said set of samples.

49. The method as set forth in claim 48, wherein said samples are irradiated in vitro.

50. The method as set forth in claim 48, wherein said algorithm is selected from the group including, partial least squares and principal component regression.

51. The method as set forth in claim 48, further including the step of detecting outlier samples in said set of samples.

52. The method as set forth in claim 48, further including the step of determining whether said intensity variations v. wavelengths response of said biological fluid with unknown values of said characteristics is an outlier.

53. The method as set forth in claim 48, wherein said known characteristics are minor components of said biological fluid and said unknown values are less than 2.0 weight percent of said biological fluid.

54. Apparatus for determining at least one unknown value of a known characteristic, such as the concentration of an analyte, in a biological fluid, said apparatus comprising:

(A) a source of infrared energy having at least several wavelengths;

(B) means for coupling said at least several wavelengths of said infrared energy to said biological fluid to enable said biological fluid to differentially absorb at least some of said wavelengths, said differential absorption causing intensity variations of said infrared energy incident from said biological fluid as a function of said at least several wavelengths of said energy and said unknown value of said known characteristic;

(C) means for measuring said intensity variations; and

(D) computer means including:

i. a stored model constructed from a set of samples in which the values of said known characteristic are known, said model being a function of said known values from said set of samples and intensity v. wavelength information obtained from said set of samples, and

ii. an algorithm including (a) all independent sources of said intensity variations v. said wavelengths information from both said set of samples and said biological fluid and (b) more wavelengths than samples, said algorithm utilizing said model for calculating said unknown value of said known characteristic of said biological fluid from said measured intensity variations from said biological fluid.

55. The apparatus as set forth in claim 54, further including means for determining whether said intensity variations v. wavelength of said known characteristic in said biological fluid is an outlier.

56. The apparatus as set forth in claim 54, wherein said means for coupling includes an attenuated total reflectance (ATR) device.

57. The apparatus as set forth in claim 56, further including a flow cell, said ATR device being positioned in said cell for in vitro sampling.

58. The apparatus as set forth in claim 56, wherein said ATR device includes a biocompatible coating on at least a portion thereof for enabling said portion of said ATR device to contact said biological fluid in vivo.

59. The apparatus as set forth in claim 54, wherein said means for coupling includes a fiber optic device.

60. The apparatus as set forth in claim 59, wherein a portion of said fiber optic device forms an ATR device.

61. The apparatus as set forth in claim 59, wherein said fiber optic device includes a portion for transmitting said infrared energy to said biological fluid and a separate portion for transmitting said infrared energy from said biological fluid to said apparatus.

62. The apparatus as set forth in claim 54, wherein said apparatus includes a first portion adapted to be positioned on one side of an in vivo source of biological fluid and a second portion adapted to be positioned on another side of said in vivo source of biological fluid.

63. The apparatus as set forth in claim 54, wherein said means for coupling includes means for transmitting said infrared energy to an in vivo source of said biological fluid and means for measuring diffuse infrared reflection from said in vivo source.

64. The apparatus as set forth in claim 54, further including means for transmitting signals from said computer means to a means for changing said value of said known characteristic.

65. The apparatus as set forth in claim 64, wherein said means for changing is an insulin pump.

66. The apparatus of claim 54, wherein said means for measuring said intensity variations includes an array of infrared sensors, means for frequency dispersing said intensity variations of said at least several wavelengths onto said sensors, different ones of said sensors being provided for different ones of said at least several wavelengths, so that said different ones of said sensors derive different electric signals.

67. The apparatus of claim 66, wherein said array includes multiple individual filters, each having a bandpass for each one of said at least several wavelengths, said filters being positioned relative to said sensors so that said infrared energy passed through each of said filters is incident on a corresponding one of said sensors.

68. The apparatus of claim 66, wherein said array includes a sheet of infrared gradient wavelength responsive material having areas positioned relative to said sensors so that said infrared energy passed through different areas of said sheet is incident on a corresponding one of said sensors.

69. The apparatus of claim 66, wherein each of said sensors is constructed to be responsive to a different one of said at least several wavelengths.

70. A method of determining in vivo at least one unknown concentration of a known characteristic in a biological fluid in a mammal, said characteristic being a trace component in said biological fluid, said concentration being less than 2.0 weight percent of said biological fluid, said method comprising:

(a) irradiating in vivo said biological fluid having said unknown concentration of said known characteristic with infrared energy having said at least several wavelengths so that there is differential absorption of at least some of said wavelengths by said biological fluid as a function of said wavelengths and said unknown concentration, said differential absorption causing intensity variations of said wavelengths incident from said biological fluid as a function of said wavelengths and said unknown concentration;

(b) measuring said intensity variations from said biological fluid;

(c) calculating said unknown concentration in said biological fluid from said measured intensity variations from said biological fluid utilizing an algorithm and a mathematical calibration model, said algorithm including all independent sources of intensity variations v. wavelengths information obtained from irradiating a set of samples in which the concentrations of said known characteristic are known, said algorithm including more wavelengths than samples in said set of samples, said model being constructed from said set of samples and being a function of said known concentrations of said known characteristic and the intensity variations v. wavelengths information obtained from irradiating said set of samples; and

(d) determining whether said intensity variations v. wavelengths of said biological fluids is an outlier.

71. The method as set forth in claim 70, wherein said known characteristic is glucose.

FIELD OF THE INVENTION

The present invention relates generally to determining the nature, i.e., the similarity or concentration, of a biological analyte in comparison with a model constructed from plural known biological fluids, and, more particularly, to such a method and apparatus wherein a sample of biological fluid containing the analyte is irradiated with infrared energy having at least several wavelengths to cause differential absorption by the analyte as a function of the wavelengths and properties of the analyte.

BACKGROUND ART

For various care and treatment of mammal patients, it is necessary to determine concentrations of certain species in biological fluids. For instance, diabetics must be apprised of their blood glucose concentrations to enable insulin dosage to be adjusted. To determine blood glucose concentrations, blood is presently drawn several times per day by the diabetic, usually via a finger prick. If the blood glucose concentrations in such individuals are not properly maintained, the individuals become susceptible to numerous physiological problems, such as blindness, circulatory disorders, coronary artery disease, and renal failure. For these reasons, a substantial improvement in the quality of life of persons suffering from various maladies, such as diabetes mellitus, could be attained if the concentrations of species in body fluids are non-invasively and/or continuously determined. For example, for diabetic patients having external or implantable insulin pumps, a feedback loop for these pumps could be controlled by continuously monitoring glucose concentrations, to enable an artificial pancreas to be developed.

Exemplary systems have been previously proposed to monitor glucose in blood, as is necessary, for example, to control diabetic patients. This prior art is represented, for example, by Kaiser, U.S. Pat. No. 4,169,676, Muller, U.S. Pat. No. 4,427,889, and Dahne et al, European Patent Publication No. 0 160 768, and Bauer et al, Analytica Chimica Acta 197 (1987) pp. 295-301.

In Kaiser, glucose in blood is determined by irradiating a sample of the blood with a carbon dioxide laser source emitting a coherent beam, at a single frequency, in the mid-infrared region. An infrared beam derived from the laser source is coupled to the sample by way of an attenuated total reflectance crystal for the purpose of contacting the blood sample. The apparatus uses double beam instrumentation to examine the difference in absorption at the single frequency in the presence and absence of a sample. The reliability of the Kaiser device is materially impaired in certain situations because of the reliance on a single frequency beam for reasons explained below. Also, we have found from calculations based on available information that Kaiser's statement anent optical energy penetrating the skin to the depth of the blood capillaries is unlikely due to water absorption of the mid-infrared beam.

Muller discloses a system for quantifying glucose in blood by irradiating a sample of the blood with energy in a single beam from a laser operating at two frequencies in the mid-infrared region. The infrared radiation is either transmitted directly to the sample or by way of an attenuated total reflectance crystal for in vitro sampling. One frequency that irradiates the sample is in the 10.53-10.65 micrometer range, while the other irradiating frequency is in the 9.13-9.17 micrometer range. The radiation at the first frequency establishes a baseline absorption by the sample, while glucose absorption by the sample is determined from the intensity reduction caused by the sample at the second wavelength. The absorption ratio by the sample at the first and second frequencies quantifies the glucose of the sample. There is no glucose absorption at the first wavelength.

Dahne et al employs near-infrared spectroscopy for non-invasively transmitting optical energy in the nearinfrared spectrum through a finger or earlobe of a subject. Also discussed is the use of near-infrared energy diffusely reflected from deep within the tissue. Responses are derived at two different wavelengths to quantify glucose in the subject. One of the wavelengths is used to determine background absorption, while the other wavelength is used to determine glucose absorption. The ratio of the derived intensity at the two different wavelengths determines the quantity of glucose in the analyte biological fluid sample.

Bauer et al discloses monitoring glucose through the use of Fourier-transform infrared spectrometry wherein several absorbance versus wavelength curves are illustrated. A glucose concentration versus absorbance calibration curve, discussed in the last paragraph on p. 298, is constructed from several samples having known concentrations, in response to the intensity of the infrared energy absorbed by the samples at one wavelength, indicated as preferably 1035 cm.sup.-1.

All of the foregoing prior art techniques thus use only a single frequency analysis or ratio of two frequencies to determine a single proportionality constant describing a relationship between absorption of the infrared energy by the sample and concentration of a constituent of the biological fluid sample being analyzed, usually glucose. Hence, the prior art analysis is univariate since absorption by the constituent of interest at a single wavelength is determined.

However, univariate analysis has a tendency to be inaccurate in situations wherein there are concentration variations of any substance which absorbs at the analysis frequency. Biological systems are subject to numerous physiological perturbations over time and from person to person. The perturbations cause inaccuracies in univariate analysis, thereby decreasing the accuracy and precision of such analysis. The physiological perturbations involving any substance which absorbs at the analysis frequencies do not permit an operator of a system utilizing univariate analysis to recognize the resulting inaccuracy. In addition, nonlinearities may arise from spectroscopic instrumentation, refractive index dispersion, or interactions between molecules of the sample which cannot generally be modelled by univariate techniques. In addition, unknown biological materials in the sample have a tendency to interfere with the analysis process, particularly when these materials are present in varying amounts. Also the univariate techniques are usually not capable of identifying outlier samples, i.e., samples with data or constituents or spectra among the calibration or unknown data which differ from the remainder of the calibration set.

The described prior art systems utilizing midinfrared energy are not feasible for non-invasive in vivo determinations of glucose concentrations because of penetration depth limitations.

The most frequently employed prior art techniques for determining the concentration of molecular substances in biological fluids have used enzymatic, chemical and/or immunological methods. However, all of these techniques require invasive methods to draw a blood sample from a subject; typically, blood must be drawn several times a day by a finger prick, such as presently employed by a diabetic. For example, in the determination of glucose by diabetics, such invasive techniques must be performed using present technology. It would be highly desirable to provide a lessinvasive, continuous or semi-continuous system for automatically analyzing glucose concentrations in the control of diabetes mellitus.

It is, accordingly, an object of the present invention to provide a new and improved method of and apparatus for determining characteristics of a biological analyte sample.

Another object of the present invention is to provide a new and improved apparatus for and method of using infrared energy for analyzing biological fluids wherein the apparatus and method are particularly suitable for analyzing samples having concentrations of substances which variably or differentially absorb the infrared energy.

Another object of the invention is to provide a new and improved method of and apparatus for utilizing infrared energy to determine a characteristic, e.g., concentration, of a biological analyte by comparison of the absorption characteristics of said sample with a mathematical model constructed from several spectra of biological fluids having known absorption versus wavelength characteristics at known analyte concentrations.

A further object of the invention is to provide a new and improved apparatus for and method of analyzing biological fluids with infrared energy wherein interference with the infrared energy due to numerous physiological perturbations over time and between people does not have a particularly adverse effect on the results.

An additional object of the invention is to provide a new and improved apparatus for and method of using infrared energy to analyze biological fluids, wherein non-linearities due to various causes, for example, spectroscopic instrumentation, refractive index dispersion, and/or inter-molecular interactions, do not have an adverse effect on the analysis results.

An additional object of the present invention is to provide a new and improved apparatus for and method of using infrared energy to determine the nature of a biological sample wherein the presence of unknown biological materials in the sample does not interfere with the analysis of the sample, as long as these unknown biological materials are present in a calibration set which is used to derive a mathematical model which represents the response of known fluids to the infrared energy.

A further object of the invention is to provide a new and improved apparatus for and method of using infrared energy to determine characteristics of biological fluids wherein outlier samples subsisting in a calibration set used to derive a mathematical model are identified and either eliminated or accommodated so as not to have an adverse effect on the determination.

Another object of the invention is to provide a method of and apparatus for identifying the presence of outliers. The quality of the calibration results and the reliability of the unknown sample analyses can be critically dependent on the detection of outlier samples. In the calibration set, an outlier is a sample that does not exhibit the characteristic relationship between composition and the absorbance spectrum of the other calibration samples. During prediction, an outlier is a sample that is not representative of samples in the calibration set. Outliers in the calibration samples can impair the precision and accuracy of the calibration and limit the quality of the analyses of all future samples. The results of the analyses of outlier unknown samples by multivariate calibration cannot be considered reliable, and samples containing outliers should be analyzed by other methods. Thus, efficient detection of outlier samples is crucial for the successful application and wide acceptance of multivariate spectral analyses. For example, outliers occur as a result of changes in instrumental response, incorrect analyte determination by the reference method, unique type of sample, unexpected components, unusual baseline, incorrectly labeled or documented sample, etc.

Still an additional object of the invention is to provide a new and improved biological fluid analysis apparatus and method which is particularly adaptable, in certain embodiments, to non-invasive determinations.

THE INVENTION

In accordance with one aspect of the present invention, the concentration of a biological fluid containing an analyte is determined from a model constructed from plural known biological fluid samples. The model is a function of the concentration of materials in the known samples as a function of absorption at at least several wavelengths of infrared energy. The infrared energy is coupled to the analyte containing sample so there is differing absorption of the infrared energy as a function of the several wavelengths and characteristics of the analyte containing sample. The differing absorption causes intensity variations of the infrared energy passing through the analyte containing sample as a function of the several wavelengths. The thus-derived intensity variations of the infrared energy as a function of the several wavelengths are compared with the calibration model relating concentration to plural absorption versus wavelength patterns derived from the plural known biological fluid samples having various concentrations. The comparison enables the determination of the analyte concentration from the measured intensity variations for the biological fluid containing the analyte. In the preferred embodiment, the comparison is made in a computer by the partial least squares method, although other multivariate techniques can be employed.

In the computer implementation, the intensity variations as a function of wavelength are converted into plural first electric signals, such that different ones of the first electric signals are assigned to different ones of the wavelengths. The magnitude of each of the different first signals is determined by the intensity of the transmitted energy at the wavelength assigned to that particular first signal. The transmitted energy in the presence of the analyte containing sample is statistically compared with the transmitted energy in the absence of the sample to determine the absorption by the biological analyte containing fluid.

A multivariate statistical method, preferably using the partial least squares technique in a manner known in the statistical art, enables a model to be constructed of the infrared absorption versus wavelength characteristics and analyte concentrations. Following determination of the calibration model, the infrared absorption versus wavelength of the unknown fluid enables estimation of the analyte concentration. For examp