CA2299285C - Method and apparatus for generating basis sets for use in spectroscopic analysis - Google Patents
Method and apparatus for generating basis sets for use in spectroscopic analysis Download PDFInfo
- Publication number
- CA2299285C CA2299285C CA002299285A CA2299285A CA2299285C CA 2299285 C CA2299285 C CA 2299285C CA 002299285 A CA002299285 A CA 002299285A CA 2299285 A CA2299285 A CA 2299285A CA 2299285 C CA2299285 C CA 2299285C
- Authority
- CA
- Canada
- Prior art keywords
- sample
- analyte
- components
- basis set
- spectral
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
- 238000005284 basis set Methods 0.000 title claims abstract description 132
- 238000000034 method Methods 0.000 title claims description 72
- 238000004611 spectroscopical analysis Methods 0.000 title description 8
- 238000001228 spectrum Methods 0.000 claims abstract description 115
- 230000003595 spectral effect Effects 0.000 claims abstract description 91
- 239000012491 analyte Substances 0.000 claims abstract description 86
- 230000002452 interceptive effect Effects 0.000 claims abstract description 52
- 210000002966 serum Anatomy 0.000 claims abstract description 43
- 238000004458 analytical method Methods 0.000 claims abstract description 35
- 238000002835 absorbance Methods 0.000 claims description 121
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims description 109
- 108010088751 Albumins Proteins 0.000 claims description 58
- 102000009027 Albumins Human genes 0.000 claims description 58
- 102000006395 Globulins Human genes 0.000 claims description 41
- 108010044091 Globulins Proteins 0.000 claims description 41
- HVYWMOMLDIMFJA-DPAQBDIFSA-N cholesterol Chemical compound C1C=C2C[C@@H](O)CC[C@]2(C)[C@@H]2[C@@H]1[C@@H]1CC[C@H]([C@H](C)CCCC(C)C)[C@@]1(C)CC2 HVYWMOMLDIMFJA-DPAQBDIFSA-N 0.000 claims description 37
- 238000012937 correction Methods 0.000 claims description 36
- XSQUKJJJFZCRTK-UHFFFAOYSA-N Urea Chemical compound NC(N)=O XSQUKJJJFZCRTK-UHFFFAOYSA-N 0.000 claims description 29
- 230000000694 effects Effects 0.000 claims description 25
- 239000004202 carbamide Substances 0.000 claims description 24
- 239000000470 constituent Substances 0.000 claims description 20
- 235000012000 cholesterol Nutrition 0.000 claims description 18
- 150000003626 triacylglycerols Chemical class 0.000 claims description 14
- 102000004169 proteins and genes Human genes 0.000 claims description 13
- 108090000623 proteins and genes Proteins 0.000 claims description 13
- 229910052739 hydrogen Inorganic materials 0.000 claims description 12
- 239000001257 hydrogen Substances 0.000 claims description 12
- 239000007788 liquid Substances 0.000 claims description 8
- 230000036961 partial effect Effects 0.000 claims description 8
- 238000010183 spectrum analysis Methods 0.000 claims description 8
- 238000000491 multivariate analysis Methods 0.000 claims description 7
- 230000035515 penetration Effects 0.000 claims description 4
- 238000001514 detection method Methods 0.000 claims description 3
- UFHFLCQGNIYNRP-UHFFFAOYSA-N Hydrogen Chemical compound [H][H] UFHFLCQGNIYNRP-UHFFFAOYSA-N 0.000 claims 2
- WQZGKKKJIJFFOK-GASJEMHNSA-N Glucose Natural products OC[C@H]1OC(O)[C@H](O)[C@@H](O)[C@@H]1O WQZGKKKJIJFFOK-GASJEMHNSA-N 0.000 abstract description 119
- 239000008103 glucose Substances 0.000 abstract description 119
- 210000003743 erythrocyte Anatomy 0.000 abstract description 6
- 230000002500 effect on skin Effects 0.000 abstract description 3
- 239000000523 sample Substances 0.000 description 99
- URAYPUMNDPQOKB-UHFFFAOYSA-N triacetin Chemical compound CC(=O)OCC(OC(C)=O)COC(C)=O URAYPUMNDPQOKB-UHFFFAOYSA-N 0.000 description 48
- 239000000872 buffer Substances 0.000 description 38
- 239000001087 glyceryl triacetate Substances 0.000 description 24
- 235000013773 glyceryl triacetate Nutrition 0.000 description 24
- 229960002622 triacetin Drugs 0.000 description 24
- 238000010521 absorption reaction Methods 0.000 description 18
- 238000001320 near-infrared absorption spectroscopy Methods 0.000 description 17
- 238000013459 approach Methods 0.000 description 13
- 230000002277 temperature effect Effects 0.000 description 13
- 210000004369 blood Anatomy 0.000 description 12
- 239000008280 blood Substances 0.000 description 12
- 230000000593 degrading effect Effects 0.000 description 9
- 239000011159 matrix material Substances 0.000 description 9
- 238000005259 measurement Methods 0.000 description 9
- 239000008363 phosphate buffer Substances 0.000 description 9
- 230000005855 radiation Effects 0.000 description 9
- 239000000126 substance Substances 0.000 description 9
- 238000000862 absorption spectrum Methods 0.000 description 8
- WQZGKKKJIJFFOK-VFUOTHLCSA-N beta-D-glucose Chemical compound OC[C@H]1O[C@@H](O)[C@H](O)[C@@H](O)[C@@H]1O WQZGKKKJIJFFOK-VFUOTHLCSA-N 0.000 description 8
- 230000005540 biological transmission Effects 0.000 description 8
- 210000004027 cell Anatomy 0.000 description 8
- 230000006870 function Effects 0.000 description 8
- 238000012545 processing Methods 0.000 description 8
- 239000007787 solid Substances 0.000 description 8
- 238000001914 filtration Methods 0.000 description 7
- 230000003287 optical effect Effects 0.000 description 7
- 230000008859 change Effects 0.000 description 6
- 238000013480 data collection Methods 0.000 description 6
- 238000010586 diagram Methods 0.000 description 6
- 230000002829 reductive effect Effects 0.000 description 6
- 239000000243 solution Substances 0.000 description 6
- 239000003153 chemical reaction reagent Substances 0.000 description 5
- 238000007405 data analysis Methods 0.000 description 5
- 230000003247 decreasing effect Effects 0.000 description 5
- 238000005316 response function Methods 0.000 description 5
- 210000003491 skin Anatomy 0.000 description 5
- 238000012935 Averaging Methods 0.000 description 4
- JVTAAEKCZFNVCJ-UHFFFAOYSA-N lactic acid Chemical compound CC(O)C(O)=O JVTAAEKCZFNVCJ-UHFFFAOYSA-N 0.000 description 4
- 238000005457 optimization Methods 0.000 description 4
- 238000010238 partial least squares regression Methods 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 210000001519 tissue Anatomy 0.000 description 4
- 239000007864 aqueous solution Substances 0.000 description 3
- 235000013405 beer Nutrition 0.000 description 3
- 230000008901 benefit Effects 0.000 description 3
- 230000007423 decrease Effects 0.000 description 3
- 238000006073 displacement reaction Methods 0.000 description 3
- 239000003792 electrolyte Substances 0.000 description 3
- 238000002474 experimental method Methods 0.000 description 3
- 239000000203 mixture Substances 0.000 description 3
- 239000010453 quartz Substances 0.000 description 3
- VYPSYNLAJGMNEJ-UHFFFAOYSA-N silicon dioxide Inorganic materials O=[Si]=O VYPSYNLAJGMNEJ-UHFFFAOYSA-N 0.000 description 3
- OYPRJOBELJOOCE-UHFFFAOYSA-N Calcium Chemical compound [Ca] OYPRJOBELJOOCE-UHFFFAOYSA-N 0.000 description 2
- CURLTUGMZLYLDI-UHFFFAOYSA-N Carbon dioxide Chemical compound O=C=O CURLTUGMZLYLDI-UHFFFAOYSA-N 0.000 description 2
- LFQSCWFLJHTTHZ-UHFFFAOYSA-N Ethanol Chemical compound CCO LFQSCWFLJHTTHZ-UHFFFAOYSA-N 0.000 description 2
- 210000000601 blood cell Anatomy 0.000 description 2
- 239000011575 calcium Substances 0.000 description 2
- 229910052791 calcium Inorganic materials 0.000 description 2
- 239000002775 capsule Substances 0.000 description 2
- 235000013877 carbamide Nutrition 0.000 description 2
- 238000006243 chemical reaction Methods 0.000 description 2
- DDRJAANPRJIHGJ-UHFFFAOYSA-N creatinine Chemical compound CN1CC(=O)NC1=N DDRJAANPRJIHGJ-UHFFFAOYSA-N 0.000 description 2
- 239000003814 drug Substances 0.000 description 2
- 229940079593 drug Drugs 0.000 description 2
- 239000012530 fluid Substances 0.000 description 2
- 235000001727 glucose Nutrition 0.000 description 2
- 238000009499 grossing Methods 0.000 description 2
- 238000010438 heat treatment Methods 0.000 description 2
- 238000002329 infrared spectrum Methods 0.000 description 2
- 238000012804 iterative process Methods 0.000 description 2
- 239000004310 lactic acid Substances 0.000 description 2
- 235000014655 lactic acid Nutrition 0.000 description 2
- 210000003205 muscle Anatomy 0.000 description 2
- 239000008188 pellet Substances 0.000 description 2
- IOLCXVTUBQKXJR-UHFFFAOYSA-M potassium bromide Chemical compound [K+].[Br-] IOLCXVTUBQKXJR-UHFFFAOYSA-M 0.000 description 2
- 239000000843 powder Substances 0.000 description 2
- 238000007781 pre-processing Methods 0.000 description 2
- 238000012628 principal component regression Methods 0.000 description 2
- 241000894007 species Species 0.000 description 2
- 238000012549 training Methods 0.000 description 2
- 244000198134 Agave sisalana Species 0.000 description 1
- 241000283690 Bos taurus Species 0.000 description 1
- VEXZGXHMUGYJMC-UHFFFAOYSA-M Chloride anion Chemical compound [Cl-] VEXZGXHMUGYJMC-UHFFFAOYSA-M 0.000 description 1
- 238000004477 FT-NIR spectroscopy Methods 0.000 description 1
- 102000008946 Fibrinogen Human genes 0.000 description 1
- 108010049003 Fibrinogen Proteins 0.000 description 1
- 102000017011 Glycated Hemoglobin A Human genes 0.000 description 1
- 108010014663 Glycated Hemoglobin A Proteins 0.000 description 1
- 208000013016 Hypoglycemia Diseases 0.000 description 1
- 238000004566 IR spectroscopy Methods 0.000 description 1
- DGAQECJNVWCQMB-PUAWFVPOSA-M Ilexoside XXIX Chemical compound C[C@@H]1CC[C@@]2(CC[C@@]3(C(=CC[C@H]4[C@]3(CC[C@@H]5[C@@]4(CC[C@@H](C5(C)C)OS(=O)(=O)[O-])C)C)[C@@H]2[C@]1(C)O)C)C(=O)O[C@H]6[C@@H]([C@H]([C@@H]([C@H](O6)CO)O)O)O.[Na+] DGAQECJNVWCQMB-PUAWFVPOSA-M 0.000 description 1
- 102000011782 Keratins Human genes 0.000 description 1
- 108010076876 Keratins Proteins 0.000 description 1
- JVTAAEKCZFNVCJ-UHFFFAOYSA-M Lactate Chemical compound CC(O)C([O-])=O JVTAAEKCZFNVCJ-UHFFFAOYSA-M 0.000 description 1
- 241001465754 Metazoa Species 0.000 description 1
- 238000004497 NIR spectroscopy Methods 0.000 description 1
- OAICVXFJPJFONN-UHFFFAOYSA-N Phosphorus Chemical compound [P] OAICVXFJPJFONN-UHFFFAOYSA-N 0.000 description 1
- ZLMJMSJWJFRBEC-UHFFFAOYSA-N Potassium Chemical compound [K] ZLMJMSJWJFRBEC-UHFFFAOYSA-N 0.000 description 1
- 238000008050 Total Bilirubin Reagent Methods 0.000 description 1
- LEHOTFFKMJEONL-UHFFFAOYSA-N Uric Acid Chemical compound N1C(=O)NC(=O)C2=C1NC(=O)N2 LEHOTFFKMJEONL-UHFFFAOYSA-N 0.000 description 1
- TVWHNULVHGKJHS-UHFFFAOYSA-N Uric acid Natural products N1C(=O)NC(=O)C2NC(=O)NC21 TVWHNULVHGKJHS-UHFFFAOYSA-N 0.000 description 1
- PNNCWTXUWKENPE-UHFFFAOYSA-N [N].NC(N)=O Chemical compound [N].NC(N)=O PNNCWTXUWKENPE-UHFFFAOYSA-N 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 239000013060 biological fluid Substances 0.000 description 1
- 239000012472 biological sample Substances 0.000 description 1
- 230000000903 blocking effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000011088 calibration curve Methods 0.000 description 1
- 239000001569 carbon dioxide Substances 0.000 description 1
- 229910002092 carbon dioxide Inorganic materials 0.000 description 1
- 239000013626 chemical specie Substances 0.000 description 1
- 230000008878 coupling Effects 0.000 description 1
- 238000010168 coupling process Methods 0.000 description 1
- 238000005859 coupling reaction Methods 0.000 description 1
- 229940109239 creatinine Drugs 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 230000009977 dual effect Effects 0.000 description 1
- 210000000624 ear auricle Anatomy 0.000 description 1
- 238000000295 emission spectrum Methods 0.000 description 1
- 238000011067 equilibration Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 229940012952 fibrinogen Drugs 0.000 description 1
- 230000003345 hyperglycaemic effect Effects 0.000 description 1
- 201000001421 hyperglycemia Diseases 0.000 description 1
- 230000002218 hypoglycaemic effect Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000012417 linear regression Methods 0.000 description 1
- 238000002483 medication Methods 0.000 description 1
- 235000013336 milk Nutrition 0.000 description 1
- 239000008267 milk Substances 0.000 description 1
- 210000004080 milk Anatomy 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000001459 mortal effect Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 229910052698 phosphorus Inorganic materials 0.000 description 1
- 239000011574 phosphorus Substances 0.000 description 1
- 239000011591 potassium Substances 0.000 description 1
- 229910052700 potassium Inorganic materials 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 238000011002 quantification Methods 0.000 description 1
- 238000004445 quantitative analysis Methods 0.000 description 1
- 238000013139 quantization Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 230000010076 replication Effects 0.000 description 1
- 230000000717 retained effect Effects 0.000 description 1
- 150000003839 salts Chemical class 0.000 description 1
- 239000012723 sample buffer Substances 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 210000004927 skin cell Anatomy 0.000 description 1
- 239000011734 sodium Substances 0.000 description 1
- 229910052708 sodium Inorganic materials 0.000 description 1
- 239000002904 solvent Substances 0.000 description 1
- 230000009897 systematic effect Effects 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 230000001131 transforming effect Effects 0.000 description 1
- 229940116269 uric acid Drugs 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
- G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
- G01N21/27—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands using photo-electric detection ; circuits for computing concentration
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
- G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
- G01N21/27—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands using photo-electric detection ; circuits for computing concentration
- G01N21/274—Calibration, base line adjustment, drift correction
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
- G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
- G01N21/31—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry
- G01N21/35—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light
- G01N21/359—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light using near infrared light
Abstract
One or more basis sets are applied to a spectroscopic signal during analysis to produce an accurate spectral representation from which analyte concentration may be accurately determined. A basis set includes all interfering components found in a sample, such as serum. With regard to an analyte, such as glucose, it is necessary to define those components of a sample that have a larger interference than that of glucose. A basis set may be generated, for example, that produces a transform for the red blood cells that interfere or scatter the light; and also for skin effects. Once the spectra of all these components is known, it is then necessary to determine how each of these components interact, e.g. taking serum data, extracting each of the components, and then comparing the spectra for the individual components with that of the components in solution. The invention characterizes each component in a sample, as well as all other possible interferants and, after producing an accurate representation of each component at each frequency of interest, identifies and subtracts each interferant from the spectra produced at the frequency of interest. The basis sets may take the form of transforms that may be stored in a look-up table for use during analysis.
Description
METHOD AND APPARATUS FOR GENERATING
BASIS SETS FOR USE IN
SPECTROSCOPIC ANALYSIS
BACKGROUND OF THE INVENTION
TECHNICAL FIELD
1o The invention relates to determining the concentration of a target analyte in a sample.
More particularly, the invention relates to a method and apparatus for generating basis sets for use in determining the concentration of a target analyte in a sample, for example using mufti-spectral analysis.
DESCRIPTION OF THE PRIOR ART
Data analysis during spectroscopic analysis refers to the process of finding optimum wavelengths and generating accurate calibrations to relate a given set of spectroscopic data to reference laboratory values for the composition of a set of samples, such that it is possible to analyze, i. e. predict, the values of future samples of unknown composition. Calibration of spectroscopic instruments that are used to perform spectroscopic measurements is typically accomplished by application of multiple regression of the absorbance at some number of wavelengths against the reference laboratory values, i.e. mathematically determining the best possible fit of a straight line to a set of data (see, for example, H. Mark, Princ~les and Practice of Spectroscopic Calibration, John Wiley & Sons, Inc. (1991)).
An error free calibration, i.e. a sample for which Beer's law applies, is one in which the constituent of interest, and which is the only constituent in the sample, is dissolved 3o in a completely nonabsorbing solvent, and has only a single absorbance band. In this case, the concentration of the constituent is known exactly over a broad range for the set of calibration samples; and the spectrometer has no noise, nonlinearity, or other fault. In such an idealized case, the height of the absorbance peak is strictly proportional to the concentration of the constituent. Thus, it is possible to calibrate a system using only two samples because two points determine the line, and the slope of the line and intercept of data are readily determined using known mathematical formulae.
Unfortunately, the ideal case does not prevail in the real world. For example, spectroscopic measurements are subject to such phenomena as skew in the data, which is caused by physical changes in the instrument, sample, or experiment. For example, interfering and/or dominating constituents in the sample other than the constituent of interest can affect the data. Temperature, medium, pathlength, and scattering effects must also be considered.
Near-infrared (near-IR) absorbance spectra of liquid samples contain a large amount of information about the various organic constituents of the sample.
Specifically, the vibrational, rotational, and stretching energy associated with organic molecular structures (e.g. carbon-hydrogen, oxygen-hydrogen, and nitrogen-hydrogen chemical bonds) produce perturbations in the near-IR region which can be detected and related to the concentration of various organic constituents present in the sample.
However, in complex sample matrices, near-IR spectra also contain an appreciable amount of 2o interference, due in part to similarities of structure amongst analytes, relative levels of analyte concentration, interfering relationships between analytes, and the magnitude of electronic and chemical noise inherent in a particular system. Such interference reduces the efficiency and precision of measurements obtained using near-IR-spectrometry to determine the concentration of liquid sample analyses.
For example, temperature is a critical parameter for near-IR spectroscopic analysis of aqueous based samples. Major water absorption bands are centered at approximately 3800, 5200, and 6900 nm, but the exact positions of these bands are temperature sensitive. These bands shift to higher frequencies at higher temperatures.
Changes in 3o temperature also alter the extent of water hydrogen bonding to other chemical species, which causes significant shifts in band positions. The large water content of most clinical samples, e.g. when determining glucose concentration in an aqueous solution, necessitates precise control of the sample temperature.
BASIS SETS FOR USE IN
SPECTROSCOPIC ANALYSIS
BACKGROUND OF THE INVENTION
TECHNICAL FIELD
1o The invention relates to determining the concentration of a target analyte in a sample.
More particularly, the invention relates to a method and apparatus for generating basis sets for use in determining the concentration of a target analyte in a sample, for example using mufti-spectral analysis.
DESCRIPTION OF THE PRIOR ART
Data analysis during spectroscopic analysis refers to the process of finding optimum wavelengths and generating accurate calibrations to relate a given set of spectroscopic data to reference laboratory values for the composition of a set of samples, such that it is possible to analyze, i. e. predict, the values of future samples of unknown composition. Calibration of spectroscopic instruments that are used to perform spectroscopic measurements is typically accomplished by application of multiple regression of the absorbance at some number of wavelengths against the reference laboratory values, i.e. mathematically determining the best possible fit of a straight line to a set of data (see, for example, H. Mark, Princ~les and Practice of Spectroscopic Calibration, John Wiley & Sons, Inc. (1991)).
An error free calibration, i.e. a sample for which Beer's law applies, is one in which the constituent of interest, and which is the only constituent in the sample, is dissolved 3o in a completely nonabsorbing solvent, and has only a single absorbance band. In this case, the concentration of the constituent is known exactly over a broad range for the set of calibration samples; and the spectrometer has no noise, nonlinearity, or other fault. In such an idealized case, the height of the absorbance peak is strictly proportional to the concentration of the constituent. Thus, it is possible to calibrate a system using only two samples because two points determine the line, and the slope of the line and intercept of data are readily determined using known mathematical formulae.
Unfortunately, the ideal case does not prevail in the real world. For example, spectroscopic measurements are subject to such phenomena as skew in the data, which is caused by physical changes in the instrument, sample, or experiment. For example, interfering and/or dominating constituents in the sample other than the constituent of interest can affect the data. Temperature, medium, pathlength, and scattering effects must also be considered.
Near-infrared (near-IR) absorbance spectra of liquid samples contain a large amount of information about the various organic constituents of the sample.
Specifically, the vibrational, rotational, and stretching energy associated with organic molecular structures (e.g. carbon-hydrogen, oxygen-hydrogen, and nitrogen-hydrogen chemical bonds) produce perturbations in the near-IR region which can be detected and related to the concentration of various organic constituents present in the sample.
However, in complex sample matrices, near-IR spectra also contain an appreciable amount of 2o interference, due in part to similarities of structure amongst analytes, relative levels of analyte concentration, interfering relationships between analytes, and the magnitude of electronic and chemical noise inherent in a particular system. Such interference reduces the efficiency and precision of measurements obtained using near-IR-spectrometry to determine the concentration of liquid sample analyses.
For example, temperature is a critical parameter for near-IR spectroscopic analysis of aqueous based samples. Major water absorption bands are centered at approximately 3800, 5200, and 6900 nm, but the exact positions of these bands are temperature sensitive. These bands shift to higher frequencies at higher temperatures.
Changes in 3o temperature also alter the extent of water hydrogen bonding to other chemical species, which causes significant shifts in band positions. The large water content of most clinical samples, e.g. when determining glucose concentration in an aqueous solution, necessitates precise control of the sample temperature.
With regard to temperature, K. Hazen, M. Arnold, G. Small, Temperature-Insensitive Near-Infrared Spectroscopic Measurement of Glucose in Aqueous Solutions, Applied Spectroscopy, Vol. 48, No. 4, pp. 477-483 (1994) disclose the use of a digital Fourier filter that is combined with partial least squares (PLS) regression to generate a calibration model for glucose that is insensitive to sample temperature. The calibration model is initially created using spectra collected over the 5000 to 4000 nm spectral range with samples maintained at 37°C. The model is evaluated by judging the ability to determine glucose concentrations from a set of prediction spectra.
Absorption spectra in the prediction set are obtained by ratioing single-beam spectra collected from solutions at temperatures ranging from 32°C to 41 °C to reference spectra collected at 37°C. The temperature sensitivity of the underlying water absorption bands creates large baseline variations in the prediction spectra that are effectively eliminated by the Fourier filtering step.
See, also, G. Small, M. Arnold, L. Marquardt, Strategies for Coupling Digital Filtering with Partial Least-Squares Regression: Application to Determination of Glucose in Plasma by Fourier Transform Near-Infrared Spectroscopy, Analytical Chemistry, Vol. 65, No. 22, pp. 3279-3289 (1993) (Gaussian-shaped bandpass digital 2o filters are implemented by use of Fourier filtering techniques and employed to preprocess spectra to remove variations due to the background absorbance of the [bovine] plasma matrix. PLS regression is used with the filtered spectra to compute calibration models for glucose); M. Arnold, G. Small, Determination of Physiological Levels of Glucose in an Aqueous Matrix with Digitally Filtered Fourier Transform Near-Infrared Spectra, Analytical Chemistry, Vol. 62, No. 14, pp. 1457-1464 (1990) (and G. Small, M. Arnold, Method and Apparatus for Non-Invasive Detection of Physiological Chemicals, Particularly Glucose, U.S. Patent No. 5,459,317 (17 October 1995)) (...A digital Fourier filter... removes both high-frequency noise and low-frequency base-line variations from the spectra. Numerical optimization 3o procedures are used to identify the best location and width of a Gaussian-shaped frequency response function for this Fourier filter. A dynamic area calculation, coupled with a simple linear base-line correction, provides an integrated area from the processed spectra that is linearly related to glucose concentration...); and K. Hazen, Glucose Determination in Biological Matrices Using Near-Infrared Spectroscopy, Ph.D. Thesis, Univ. of Iowa (August 1995) (glucose determinations in water, serum, blood, and the body are performed using near-IR spectroscopy, multivariate analysis is used to correlate minor spectral variations with analyte concentrations.
A number of near-IR devices and methods have been described that may be used in connection with the foregoing techniques to provide noninvasive blood analyte determinations:
1o U.S. Patent No. 5,360,004 to Purdy et al. describes a method and apparatus for the determination of blood analyte concentrations, wherein a body portion is irradiated with radiation containing two or more distinct bands of continuous-wavelength incident radiation. Purdy et al. emphasize filtration techniques to specifically block radiation at the two peaks in the near-IR absorption spectrum for water, occurring at about 1440 and 1935 nm. Such selective blocking is carried out in order to avoid a heating effect that may be due to the absorption of radiation by water in the body part being irradiated.
By contrast, U.S. Patent No. 5,267,152 to Yang et al. describes noninvasive devices 2o and techniques for measuring blood glucose concentration using only the portion of the IR spectrum which contains the near-IR water absorption peaks (e.g. the water transmission window, which includes those wavelengths between 1300 and 1900 nm), where water absorbance reaches a minimum at 1600 nm. Optically controlled light is directed to a tissue source and then collected by an integrating sphere. The collected light is analyzed and blood glucose concentration calculated using a stored reference calibration curve.
U.S. Patent No. 5,606,164 to Price et al. describes a method and apparatus for measuring the concentration of an analyte present in a biological fluid. near-IR
3o radiation is applied to calibration samples to produce calibration data.
Unknown sample data is analyzed using data pretreatment followed by projection into the calibration model space with prediction of analyte concentration using the calibration model.
Absorption spectra in the prediction set are obtained by ratioing single-beam spectra collected from solutions at temperatures ranging from 32°C to 41 °C to reference spectra collected at 37°C. The temperature sensitivity of the underlying water absorption bands creates large baseline variations in the prediction spectra that are effectively eliminated by the Fourier filtering step.
See, also, G. Small, M. Arnold, L. Marquardt, Strategies for Coupling Digital Filtering with Partial Least-Squares Regression: Application to Determination of Glucose in Plasma by Fourier Transform Near-Infrared Spectroscopy, Analytical Chemistry, Vol. 65, No. 22, pp. 3279-3289 (1993) (Gaussian-shaped bandpass digital 2o filters are implemented by use of Fourier filtering techniques and employed to preprocess spectra to remove variations due to the background absorbance of the [bovine] plasma matrix. PLS regression is used with the filtered spectra to compute calibration models for glucose); M. Arnold, G. Small, Determination of Physiological Levels of Glucose in an Aqueous Matrix with Digitally Filtered Fourier Transform Near-Infrared Spectra, Analytical Chemistry, Vol. 62, No. 14, pp. 1457-1464 (1990) (and G. Small, M. Arnold, Method and Apparatus for Non-Invasive Detection of Physiological Chemicals, Particularly Glucose, U.S. Patent No. 5,459,317 (17 October 1995)) (...A digital Fourier filter... removes both high-frequency noise and low-frequency base-line variations from the spectra. Numerical optimization 3o procedures are used to identify the best location and width of a Gaussian-shaped frequency response function for this Fourier filter. A dynamic area calculation, coupled with a simple linear base-line correction, provides an integrated area from the processed spectra that is linearly related to glucose concentration...); and K. Hazen, Glucose Determination in Biological Matrices Using Near-Infrared Spectroscopy, Ph.D. Thesis, Univ. of Iowa (August 1995) (glucose determinations in water, serum, blood, and the body are performed using near-IR spectroscopy, multivariate analysis is used to correlate minor spectral variations with analyte concentrations.
A number of near-IR devices and methods have been described that may be used in connection with the foregoing techniques to provide noninvasive blood analyte determinations:
1o U.S. Patent No. 5,360,004 to Purdy et al. describes a method and apparatus for the determination of blood analyte concentrations, wherein a body portion is irradiated with radiation containing two or more distinct bands of continuous-wavelength incident radiation. Purdy et al. emphasize filtration techniques to specifically block radiation at the two peaks in the near-IR absorption spectrum for water, occurring at about 1440 and 1935 nm. Such selective blocking is carried out in order to avoid a heating effect that may be due to the absorption of radiation by water in the body part being irradiated.
By contrast, U.S. Patent No. 5,267,152 to Yang et al. describes noninvasive devices 2o and techniques for measuring blood glucose concentration using only the portion of the IR spectrum which contains the near-IR water absorption peaks (e.g. the water transmission window, which includes those wavelengths between 1300 and 1900 nm), where water absorbance reaches a minimum at 1600 nm. Optically controlled light is directed to a tissue source and then collected by an integrating sphere. The collected light is analyzed and blood glucose concentration calculated using a stored reference calibration curve.
U.S. Patent No. 5,606,164 to Price et al. describes a method and apparatus for measuring the concentration of an analyte present in a biological fluid. near-IR
3o radiation is applied to calibration samples to produce calibration data.
Unknown sample data is analyzed using data pretreatment followed by projection into the calibration model space with prediction of analyte concentration using the calibration model.
Devices have also been described for use in determination of analyte concentrations in complex samples, for example:
U.S. Patent No. 5,242,602 to Richardson et al. describes methods for analyzing aqueous systems to detect multiple components. The methods involve determination of the absorbance or emission spectrum of the components over the range of 200 to 2500 nm, and application of chemometrics algorithms to extract segments of the spectral data obtained to quantify multiple performance indicators.
U.S. Patent No. 5,252,829 to Nygaard et al. describes a method and apparatus for measuring the concentration of urea in a milk sample using an infrared attenuation measuring technique. Multivariate techniques are carned out to determine spectral contributions of known components using partial least squares algorithms, principal component regression, multiple linear regression or artificial neural network learning.
Calibration is carned out by accounting for the component contributions that block the analyte signal of interest. Thus, Nygaard et al. describe a technique of measuring multiple analyte infrared attenuations and compensating for the influence of background analyses to obtain a more accurate measurement.
U.S. Patent No. 4,975,581 to Robinson et al, describes a method and apparatus for determining analyte concentration in a biological sample based on a comparison of infrared energy absorption (i. e. differences in absorption at several wavelengths) between a known analyte concentration and a sample. The comparison is performed using partial least squares analysis or other multivariate techniques.
U.S. Patent No. 4,882,492 to Schlager describes a method and apparatus for noninvasive determination of blood analyte concentrations. Modulated IR
radiation is directed against a tissue sample (e.g. an ear lobe) and either passed through the tissue 3o or impinged on a skin surface where it is spectrally modified by a target analyte (glucose). The spectrally modified radiation is then split, wherein one portion is directed through a negative correlation cell and another through a reference cell.
U.S. Patent No. 5,242,602 to Richardson et al. describes methods for analyzing aqueous systems to detect multiple components. The methods involve determination of the absorbance or emission spectrum of the components over the range of 200 to 2500 nm, and application of chemometrics algorithms to extract segments of the spectral data obtained to quantify multiple performance indicators.
U.S. Patent No. 5,252,829 to Nygaard et al. describes a method and apparatus for measuring the concentration of urea in a milk sample using an infrared attenuation measuring technique. Multivariate techniques are carned out to determine spectral contributions of known components using partial least squares algorithms, principal component regression, multiple linear regression or artificial neural network learning.
Calibration is carned out by accounting for the component contributions that block the analyte signal of interest. Thus, Nygaard et al. describe a technique of measuring multiple analyte infrared attenuations and compensating for the influence of background analyses to obtain a more accurate measurement.
U.S. Patent No. 4,975,581 to Robinson et al, describes a method and apparatus for determining analyte concentration in a biological sample based on a comparison of infrared energy absorption (i. e. differences in absorption at several wavelengths) between a known analyte concentration and a sample. The comparison is performed using partial least squares analysis or other multivariate techniques.
U.S. Patent No. 4,882,492 to Schlager describes a method and apparatus for noninvasive determination of blood analyte concentrations. Modulated IR
radiation is directed against a tissue sample (e.g. an ear lobe) and either passed through the tissue 3o or impinged on a skin surface where it is spectrally modified by a target analyte (glucose). The spectrally modified radiation is then split, wherein one portion is directed through a negative correlation cell and another through a reference cell.
Intensity of the radiation passing through the cells are compared to determine analyte concentration in the sample.
U.S. Patent No. 4,306,152 to Ross et al. describes an optical fluid analyzer designed to minimize the effect of background absorption (i.e. the overall or base level optical absorbance of the fluid sample) on the accuracy of measurement in a turbid sample or in a liquid sample which is otherwise difficult to analyze. The apparatus measures an optical signal at the characteristic optical absorption of a sample component of interest and another signal at a wavelength selected to approximate background absorption, to and then subtracts to reduce the background component of the analyte dependent signal.
U.S. Patent No. 4,893,253 to Lodder describes a method for analyzing intact capsules and tablets by using near-infrared reflectance spectroscopy. The method detects adulterants in capsules by obtaining spectra for a training set of unadulterated samples, representing each spectrum as a point in a hyperspace, creating a number of training set replicates and a bootstrap replicate distribution, calculating the center of the bootstrap replicate distribution, obtaining a spectrum for an adulterated sample, transforming the spectrum into a point in hyperspace, and identifying the adulterated 2o sample as abnormal based on a relationship between the adulterated sample's hyperspatial point and the bootstrap replication distribution.
See, also, R. Rosenthal, L. Paynter, L. Mackie, Non-Invasive Measurement of Blood Glucose, U.S. Patent No. 5,028,787 (2 July 1991) (A near-infrared quantitative analysis instrument and method non-invasively measures blood glucose by analyzing near-infrared energy following interactance with venous or arterial blood, or transmission through a blood containing body part.).
The accuracy of information obtained using the above described methods and devices 3o is limited by the spectral interference caused by background, i.e. non-analyte, sample constituents that also have absorption spectra in the near-IR range.
Appreciable levels of background noise represent an inherent system limitation particularly when very little analyte is present. In light of this limitation, attempts have been made to improve signal-to-noise ratios, e.g. by avoiding water absorption peaks to enable the use of increased radiation intensity, by reducing the amount of spectral information to be analyzed, or by using subtraction or compensation techniques based on an approximation of background absorption. As discussed above, these techniques have focused primarily upon examining all constituents of a spectrum simultaneously.
Although such techniques have provided some improvement, there remains a need to provide a method and apparatus for performing a more precise determination of the concentration of analytes, for example in a liquid matrix, i.e. where an accurate representation of each and every sample component is obtained during analysis.
to SUMMARY OF THE INVENTION
The invention provides one or more basis sets that are applied to a spectroscopic signal during analysis to produce an accurate spectral representation from which analyte concentration may be accurately determined. The presently preferred embodiment of the invention is applicable for the determination of such analytes as glucose in serum, as determined using non-invasive techniques. For example, in the basis sets, near-IR absorbance features over the 1100 to 2500 nm spectral region are provided for water, albumin protein, globulin protein, triacetin, cholesterol, BIJN, and 2o glucose. In addition, sample temperature effects are also included, along with instrument noise levels.
A basis set includes all interfering components found in a sample, such as serum.
These components can include, for example, water, temperature/hydrogen bonding effects, albumin globulin protein, triglycerides, cholesterol, urea, and all organic components. The basis set also includes electrolytes, such as Na+, K+ and Clw The basis set does not include those components that do not interfere, such as anything that in terms of concentration is less than the background signal or noise level. With regard to an analyte, such as glucose, it is necessary to define those components of a sample that have a larger interference than that of glucose.
Instead of considering only the analytes that are mentioned above, which are all in blood or serum, a basis set may be generated, for example, that produces a transform for the red blood cells that interfere or scatter the light; and also for skin effects.
Once the spectra of each of these components is known, it is then necessary to determine how the components interact, e.g. taking serum data, extracting each of the components, and then comparing the spectra for the individual components with that of the components in solution.
Thus, once a basis set is generated for glucose in the presence of water, it is to determined that water interferes with glucose, and it is determined how to remove the water, then a basis set for a next component can be generated, such as for temperature effect. In the example of non-invasive glucose concentration determination, the invention sequentially adds basis sets for other components, e.g. globulin, protein, triglycerides, urea, or cholesterol, in the presence of water, to build up to a serum matrix. Once basis sets are generated for serum, it is then possible to generate basis sets for red blood cells, muscle layers, skin layers, fat layers, even the whole body.
It is significant to note that the basis set approach herein thus characterizes each component in a sample, as well as all other possible interferants and, after producing 2o an accurate representation of each component at each frequency of interest, subtracts each interferant from the spectra produced at the frequency of interest. In this way, all interferants may be identified within the context of all other relevant sample components, and thence removed from the spectra, leaving substantially only the signal produced by the analyte of interest.
The various basis sets may be also combined mathematically to generate a set of transforms that may be stored in a look-up table for use during analysis. In this way, a fast real time determination of analyte concentration may be made using relatively simple, low power computer hardware, e.g. a low power embedded controller.
U.S. Patent No. 4,306,152 to Ross et al. describes an optical fluid analyzer designed to minimize the effect of background absorption (i.e. the overall or base level optical absorbance of the fluid sample) on the accuracy of measurement in a turbid sample or in a liquid sample which is otherwise difficult to analyze. The apparatus measures an optical signal at the characteristic optical absorption of a sample component of interest and another signal at a wavelength selected to approximate background absorption, to and then subtracts to reduce the background component of the analyte dependent signal.
U.S. Patent No. 4,893,253 to Lodder describes a method for analyzing intact capsules and tablets by using near-infrared reflectance spectroscopy. The method detects adulterants in capsules by obtaining spectra for a training set of unadulterated samples, representing each spectrum as a point in a hyperspace, creating a number of training set replicates and a bootstrap replicate distribution, calculating the center of the bootstrap replicate distribution, obtaining a spectrum for an adulterated sample, transforming the spectrum into a point in hyperspace, and identifying the adulterated 2o sample as abnormal based on a relationship between the adulterated sample's hyperspatial point and the bootstrap replication distribution.
See, also, R. Rosenthal, L. Paynter, L. Mackie, Non-Invasive Measurement of Blood Glucose, U.S. Patent No. 5,028,787 (2 July 1991) (A near-infrared quantitative analysis instrument and method non-invasively measures blood glucose by analyzing near-infrared energy following interactance with venous or arterial blood, or transmission through a blood containing body part.).
The accuracy of information obtained using the above described methods and devices 3o is limited by the spectral interference caused by background, i.e. non-analyte, sample constituents that also have absorption spectra in the near-IR range.
Appreciable levels of background noise represent an inherent system limitation particularly when very little analyte is present. In light of this limitation, attempts have been made to improve signal-to-noise ratios, e.g. by avoiding water absorption peaks to enable the use of increased radiation intensity, by reducing the amount of spectral information to be analyzed, or by using subtraction or compensation techniques based on an approximation of background absorption. As discussed above, these techniques have focused primarily upon examining all constituents of a spectrum simultaneously.
Although such techniques have provided some improvement, there remains a need to provide a method and apparatus for performing a more precise determination of the concentration of analytes, for example in a liquid matrix, i.e. where an accurate representation of each and every sample component is obtained during analysis.
to SUMMARY OF THE INVENTION
The invention provides one or more basis sets that are applied to a spectroscopic signal during analysis to produce an accurate spectral representation from which analyte concentration may be accurately determined. The presently preferred embodiment of the invention is applicable for the determination of such analytes as glucose in serum, as determined using non-invasive techniques. For example, in the basis sets, near-IR absorbance features over the 1100 to 2500 nm spectral region are provided for water, albumin protein, globulin protein, triacetin, cholesterol, BIJN, and 2o glucose. In addition, sample temperature effects are also included, along with instrument noise levels.
A basis set includes all interfering components found in a sample, such as serum.
These components can include, for example, water, temperature/hydrogen bonding effects, albumin globulin protein, triglycerides, cholesterol, urea, and all organic components. The basis set also includes electrolytes, such as Na+, K+ and Clw The basis set does not include those components that do not interfere, such as anything that in terms of concentration is less than the background signal or noise level. With regard to an analyte, such as glucose, it is necessary to define those components of a sample that have a larger interference than that of glucose.
Instead of considering only the analytes that are mentioned above, which are all in blood or serum, a basis set may be generated, for example, that produces a transform for the red blood cells that interfere or scatter the light; and also for skin effects.
Once the spectra of each of these components is known, it is then necessary to determine how the components interact, e.g. taking serum data, extracting each of the components, and then comparing the spectra for the individual components with that of the components in solution.
Thus, once a basis set is generated for glucose in the presence of water, it is to determined that water interferes with glucose, and it is determined how to remove the water, then a basis set for a next component can be generated, such as for temperature effect. In the example of non-invasive glucose concentration determination, the invention sequentially adds basis sets for other components, e.g. globulin, protein, triglycerides, urea, or cholesterol, in the presence of water, to build up to a serum matrix. Once basis sets are generated for serum, it is then possible to generate basis sets for red blood cells, muscle layers, skin layers, fat layers, even the whole body.
It is significant to note that the basis set approach herein thus characterizes each component in a sample, as well as all other possible interferants and, after producing 2o an accurate representation of each component at each frequency of interest, subtracts each interferant from the spectra produced at the frequency of interest. In this way, all interferants may be identified within the context of all other relevant sample components, and thence removed from the spectra, leaving substantially only the signal produced by the analyte of interest.
The various basis sets may be also combined mathematically to generate a set of transforms that may be stored in a look-up table for use during analysis. In this way, a fast real time determination of analyte concentration may be made using relatively simple, low power computer hardware, e.g. a low power embedded controller.
BRIEF DESCRIPTION OF THE DRAWINGS
Fig. 1 is a flow diagram showing the generation of a basis set according to the invention;
s Fig. 2 is a block schematic diagram of an instrument that incorporates one or more basis sets according to the invention;
Fig. 3 is a block schematic diagram of an instrument that implements an algorithm to which incorporates one or more basis sets according to the invention;
Fig. 4 is a plot of water absorbance vs. wavelength;
Fig. 5 a is a plot of water absorbance for a varying temperature vs.
wavelength;
is Fig. 5b is a plot showing temperature effect;
Fig. 6a is a plot of water absorbance vs. wavelength showing absorbance when a nm long pass filter is used;
Fig. 6b is a plot of absorbance vs. wavelength showing protein~aq~ - water absorbance;
Fig. 7 is a plot of absorbance vs. wavelength showing albumir~aq~ - buffer (pathlength corrected);
Fig. 8 is a plot of absorbance vs. wavelength showing globulint$q~ - buffer;
Fig. 9 is a plot of absorbance vs. wavelength showing albumin~aq~ - buffer (pathlength corrected);
Fig. 10 is a plot showing pathlength corrections required for albumin, globulin, and triacetin;
Fig. 1 is a flow diagram showing the generation of a basis set according to the invention;
s Fig. 2 is a block schematic diagram of an instrument that incorporates one or more basis sets according to the invention;
Fig. 3 is a block schematic diagram of an instrument that implements an algorithm to which incorporates one or more basis sets according to the invention;
Fig. 4 is a plot of water absorbance vs. wavelength;
Fig. 5 a is a plot of water absorbance for a varying temperature vs.
wavelength;
is Fig. 5b is a plot showing temperature effect;
Fig. 6a is a plot of water absorbance vs. wavelength showing absorbance when a nm long pass filter is used;
Fig. 6b is a plot of absorbance vs. wavelength showing protein~aq~ - water absorbance;
Fig. 7 is a plot of absorbance vs. wavelength showing albumir~aq~ - buffer (pathlength corrected);
Fig. 8 is a plot of absorbance vs. wavelength showing globulint$q~ - buffer;
Fig. 9 is a plot of absorbance vs. wavelength showing albumin~aq~ - buffer (pathlength corrected);
Fig. 10 is a plot showing pathlength corrections required for albumin, globulin, and triacetin;
Fig. 11 is a plot of absorbance vs. wavelength showing triacetin~aq~ - buffer (pathlength corrected);
Fig. 12a is a plot of absorbance vs. wavelength showing urea- buffer;
Fig. 12 b is a plot of absorbance vs. wavelength showing urea - buffer (baseline corrected);
Fig. 13 is a plot of absorbance vs. wavelength showing glucose - buffer;
Fig. 14 is a plot of absorbance vs. wavelength for solid samples;
Fig. 15 is a plot of normalized absorbance vs. wavelength for solid samples;
Fig. 16 is a second plot of normalized absorbance vs. wavelength for solid samples;
Fig. 17 is a plot of standard error (mg/dL) vs. number of PLS factors for glucose~aq~;
Fig. 18a is another view of the fourth plot of standard error (mg/dL) vs.
resolution (nm) for glucose~a~;
Fig. 18b is another view of the fourth plot of standard error (mg/dL) vs.
resolution (nm) for glucose~aq~ showing an expanded y-axis;
Fig. 19 is another view of the eighth plot of standard error (mg/dL) vs.
resolution (nm) for glucose~aq~;
Fig. 20 is another view of the eighth plot of standard error (mg/dL) vs.
resolution (nm) for glucose~~~ showing averaged points of six PLS factors;
Fig. 21 is a plot of standard error (mg/dL) vs. resolution (nrn) for glucose in serum;
Fig. 22 is a second plot of standard error (mg/dL) vs. resolution (nm) for glucose in serum;
Fig. 23 is a third plot of standard error (mg/dL) vs. resolution (nm) for glucose in serum;
Fig. 24 is a fourth plot of standard error (mg/dL) vs. resolution (nm) for glucose in serum;
1o Fig. 25 is a fifth plot of standard error (mg/dL) vs. resolution (nm) for glucose in serum;
Fig. 26 is a sixth plot of standard error (mg/dL) vs. resolution (nm) for glucose in serum;
~5 Fig. 27 is a seventh plot of standard error (mg/dL) vs. resolution (nm) for glucose in serum;
Fig. 28 is an eighth plot of standard error (mg/dL) vs. resolution (nm) for glucose in 20 serum;
Fig. 29 is a ninth plot of standard error (mg/dL) vs. resolution (nm) for glucose in serum;
25 Fig. 30 is a tenth plot of standard error (mg/dL) vs. resolution (nm) for glucose in serum;
Fig. 31 is another view of the eighth plot of standard error (mg/dL) vs.
resolution (nm) for glucose in serum;
Fig. 32 is a plot of absorbance vs. wavelength showing raw absorbance for a sample containing water, albumin, globulin, and triacetin;
Fig. 33 is a plot of absorbance vs. wavelength for a sample containing albumin, globulin, and triacetin and from which water is subtracted and temperature and pathlength are corrected;
Fig. 34 is a plot of absorbance vs. wavelength showing linearity for albumin spectra where temperature and pathlength are corrected;
Fig. 35 is a plot of absorbance vs. wavelength for a sample containing globulin and triacetin and from which water and albumin are subtracted;
to Fig. 36 is a plot of absorbance vs. wavelength showing linearity for globulin spectra where temperature and pathlength are corrected; and Fig. 37 is a plot of absorbance vs. wavelength for a sample containing triacetin and from which water, albumin, and globulin are subtracted.
DETAILED DESCRIPTION OF THE INVENTION
The following discussion describes what the basis set is, how it is collected, the 2o instrument that is required, the data collection parameters, the data analysis as far as such factors as temperature and path length are concerned, and what are considered to be additional basis sets.
The simplest example of a basis set is a basis set that includes all interfering components in a sample, such as serum. These components can include, for example, water, temperature/hydrogen bonding effects, albumin, and globulin protein, triglycerides, cholesterol, urea, all organic components, and Na+, K+, and Cl-. To a lesser degree, the basis set may include additional electrolytes.
3o The basis set does not include those components that do not interfere, such as anything that in terms of concentration is less than the background signal or noise.
With regard to an analyte, such as glucose, it is necessary to define those components of a sample that have a larger interference than that of glucose. Instead of considering only the analytes that are mentioned above, which are all in blood or serum, a basis set may be generated, for example, that produces a transform for the red blood cells that interfere or scatter the light; and also for skin effects.
Once the spectra of all these components is known, it is then necessary to determine how each of these components interact, e.g. taking serum data, extracting each of the components, and then comparing the spectra for the individual components with that of the components in solution.
to Once a basis set is generated for glucose in the presence of water, it is determined that water interferes with glucose, and it is determined how to remove the water, then a basis set for a next component can be generated, such as for temperature effect. In the example of non-invasive glucose concentration determination, the invention sequentially adds basis sets for other components, e.g. globulin, protein, triglycerides, urea, or cholesterol, in the presence of water, to build up to a serum matrix.
Once basis sets are generated for serum, it is then possible to generate basis sets for red blood cells, muscle layers, skin layers, fat layers, even the whole body.
It is significant to note that the basis set approach herein thus characterizes each 2o component in a sample, as well as all other possible interferants, and subtracts each interferant from the spectra produced at the frequency of interest. In this way, all interferants are identified within the context of the sample and systematically from the spectra, leaving substantially only the signal produced by the analyte of interest.
The various basis sets may be combined mathematically to generate a set of transforms that may be stored in a look-up table for use during analysis. In this way, a fast, real time determination of analyte concentration may be made using relatively simple, low power computer hardware, e.g. a low power embedded controller.
3o Once it is determined which components are present in the sample, it is necessary to determine the best method of collecting spectra for these components. However, it is also necessary to define instrument specifications, such as signal to noise ratio, resolution, and wavelength reproducibility, before the spectra can be collected. These instrument considerations are discussed in detail below.
The procedure for generating basis sets is iterative. In some embodiments of the invention it is necessary to consider such factors as scatter correction, refractive index correction, depth of penetration into the tissue, total optical path length, and temperature. Once the basis sets are generated, they may be applied to a spectroscopic input signal. The signal thus processed by the basis sets is then preprocessed using standard chemometric techniques, such as smoothing and second derivative analysis.
to Another approach to processing after application of the basis sets is that of deconvolution. If deconvolution is used, then it is necessary to perform temperature correction after the data collection and scatter correction. This approach uses the basis sets to identify and isolate various components of the sample in an iterative fashion. 'Thereafter, multivariate analysis may be applied, which may include partial least squares analysis or principal components analysis. Such processing is a matter of choice and is well known in the art. For example, in a glucose concentration C for which there is a spectrum of interferograms n x m, where m is the number of interferograms and n is the number of interferogram points, a data reduction is performed, in which:
C=bo+blP~ +... b"P"; (1) where P; are PLS factor scores derived from the interferogram points and concentration values; and b; provides the regression coefficients.
Unique to the invention, various transforms such as deconvolution are performed with reference to the basis set. Pre-processing also relies on the basis set in the invention.
For example, in the case of a Fourier filter in which certain frequencies pass through 3o the filter, it is necessary to know what frequency the filter passed. In this way, it is possible to determine if the analyte is passed by the filter. The basis set is referenced to identify the analyte concentration at various frequencies, such that the Fourier filter only need be applied at those frequencies of interest, and not across a broad range of absorbance frequencies (as is practiced in the prior art). Thus, the basis set may loosely be thought of in this application as a filter for the filter.
Various molecular relations may be considered to be basis sets in themselves, such as carbon-hydrogen, oxygen-hydrogen, and nitrogen-hydrogen bonding. In such cases, there are more absorbance bands than can be accounted for by these fundamental components. This means that there are related effects for those portions of molecular structures to which these components are bonded. Thus, even though these molecules or pieces of molecules may be found in common among different constituents, it is 1o possible to assign them to a constituent and then discard them during deconvolution because of the signature across the spectra of a particular constituent.
Fig. 1 is a flow diagram showing the generation of a basis set according to the invention. The first step of the process involves identifying relevant interfering components of the sample at the same frequency as that of the analyte (100).
This step and subsequent steps may be performed using known spectroscopic and chemometric techniques, as is discussed in greater detail below. Once the interfering components are identified, the relevant interfering components are then all identified at other frequencies to quantify absorbance at these other frequencies (102).
The 2o interfering components, once quantified, are then removed at the frequency of the analyte (104). Each iteration of the foregoing steps may be described as a separate basis set. Thus, the invention produces a plurality of basis sets for an analyte.
Fig. 2 is a block schematic diagram of an instrument that incorporates one or more basis sets according to the invention. In operation, a device 10, collects spectra 20 using standard or modified (see below) spectroscopic devices. The spectra are provided to the input port/buffer 21 of a system that includes a processor 22.
The input port/buffer may include an analog-to-digital conversion function, such that spectral data collected by the spectroscopic device are converted to digital data. The processor operates upon such digital input data in accordance with various transforms stored in one or more look-up tables (LUTs). The LUTs contain transforms that incorporate the various basis sets. The transform process performed by the processor uses the basis sets to identify and remove substantially all interfering constituents from the spectral signal produced by the spectroscopic device. Once processing with regard to the basis sets is completed, the digital signal contains substantially only the analyte information. This information is then further processed in accordance with known spectroscopic calibration and chemometric techniques and provided to an output port/buffer 24. The output information may then be observed on a display 26, in any desired format, to provide an accurate indication of analyte concentration within the sample.
As also discussed both above and below, the basis sets generated in accordance with 1o the invention herein may be stored in a lookup table or they may be mixed in with the other transform information. In producing such look-up tables, the basis sets first exist as matrix raw data collected during the iterative process of generating the several basis sets. In view of the several basis sets generated in the preferred embodiment of the invention, there may be different matrices in the look-up tables, or there may be a single matrix that generates a transform which is representative of all of the basis sets and that is applied directly to the raw data. Thus, one embodiment of the invention takes each of the components and builds them into a complex matrix that comprises an algorithm for identifying and removing interferants. In this way, the invention provides a system that accurately represents how the components appear within the 2o spectra of interest when such components are all combined. It is this ability of the invention to identify each relevant component of a sample individually within the context of each other component that allows the ultimate determination of look-up table entries for an analyte of interest.
Although possible, the presently preferred embodiment of the invention does not provide spectra of glucose that has been corrected for all interferants at all concentration levels in a lookup table. Rather, there are a series of spectra of the analyte at certain different physiological concentrations of interest. For example, in the case of glucose, there are look-up basis set values for hypo- and hyperglycemia concentrations. Thus, the invention does not need to represent all of the information in all of the basis sets in the look-up tables. Rather, it is only necessary to represent information over the whole range of glucose that occurs in the body. The approach is taken for albumin protein, and other sample components. As discussed above, a single equation may be written for all of the spectral information in this matrix, or one or more look-up tables may be provided. In any event, this approach of storing only useful spectral information in the look-up tables reduces the memory and processing power requirements of the instrument.
As discussed more fully below, the basis sets are first generated and, thereafter, incorporated into an instrument for use during analysis. To determine those values that are to be put into the look-up tables it is necessary to go through any number of basis sets. As discussed above, it is necessary to identify the major interfering to components that affect the analyte in the sample and generate basis sets for each and every one of these components.
Fig. 3 is a block schematic diagram of an instrument that implements an algorithm which incorporates one or more basis sets according to the invention. Fig. 3 provides a detailed overview of the soflware/firmware component 30 of the device discussed in connection with Fig. 2. It should be appreciated that the invention herein is readily applied to any spectroscopic system. Thus, the system described in connection with Figs. 2 and 3 is provided only as an example of a presently preferred embodiment of the invention and not by way of limitation.
During processing within the instrument, the digitized spectral information is first applied to the basis sets 31. As discussed above (and in greater detail below), the basis sets are reduced to transforms that remove interfering constituents and/or components (chemical and/or physical) from the spectral information. The basis sets 2s may be applied before or in connection with a physical model 32 that corrects for such interfering physical factors as scattering, pathlength, and/or temperature.
After the spectral information is applied to the basis sets and (optionally) the physical model, the signal thus produced is deconvolved 33 to correct the signal to a reference.
The signal is next preprocessed 34 and digitally filtered 35. Preprocessing may employ such techniques as Kubelka-Munk transformation, mean centering, normalization, baseline correction, scatter correction, and interference correction, although it is presently preferred that the basis sets be used to resolve such issues as scatter correction, baseline correction, and interference correction.
Correction techniques that may be applied, for example to scattering, can include multiplicative scatter correction, standard normal variate correction, and extended multiplicative signal correction. The digital filtering function may be accomplished by such techniques as Gaussian filtering, low and high bandpass filters, and Lorentzian filtering.
Spectral wavelengths for the analyte are selected 36 and a multivariate analysis, such as higher order partial least squares (PLS) is performed 37. Such analysis techniques to may include principal component regression, partial least squares, rotated principal components, or correlation principal components analysis.
The preferred embodiment of the invention provides a plurality of basis sets that are used to quantify an analyte in a liquid sample. For purpose of illustration and example, the invention is now described in connection with glucose quantification in noninvasive spectra.
Data Collection. The first step of the process involves identifying major interfering chemical analytes and structures in the body. These factors include, inter alia, water 2o percentage present in the sample, temperature/hydrogen bonding effects on water, albumin protein, globulin protein, triglycerides, cholesterol, urea, glucose, lactate, ethanol, also Na+, K+, Cl-, and other electrolytes, glycosylated hemoglobin, skin, keratin, fibrinogen, and red blood cells. One advantage of the invention is the basis set may be generated in such way that it includes spectra for all interfering components.
Noninterfering components include, for example, components of lower molar absorptivity concentration, such as low dosage drugs and medications.
In the presently preferred embodiment of the invention, data collection instrumentation should take into account the following:
~ Signal is defined for each analyte by first determining the delta absorption from top of absorbance band to base, and then by defining the slope of change in absorbance versus concentration for samples spanning the physiological concentration at all frequencies.
~ Noise is defined as root mean square (rms) noise of analyte absorbance in the band of interest.
~ Sisal to Noise is defined as (slope X concentration) / noise. 'This value 1o must be greater than one for a minimum specified concentration to be analyzed.
~ Resolution in the presently preferred embodiment of the invention requires a minimum of seven points per peak.
Another factor to be considered is wavelength reproducibility. In the invention, a modified NIRS 5000 spectrometer is used to achieve the above criteria.
The data collection parameters for the basis set include the following:
Pathlengths due to absorbance of water, which is the primary interferant. It is necessary to select different pathlengths for each spectral window for an optimal basis set.
In the presently preferred embodiment of the invention, these pathlengths are:
~ 0 to 2 mm for combination band region;
~ 5 to 10 mm for first overtone region; and ~ 10 mm or greater for second overtone region.
While not necessary, it is possible to generate a basis set over the entire frequency range in a single data collection to compare information in different regions.
In the preferred embodiment of the invention, this dictates a 1 mm pathlength.
Optimal signal to noise levels are obtained separately. It is also necessary to provide continuous spectra to identify and model parameters, such as change in refractive index as a function of frequency.
For applications that involve the use of diffuse reflection, pathlength considerations are not taken into account because light penetration is proportional to the inverse of water absorbance, as defined by the system, based upon molecules interacting with specific concentrations at specific refractive indices.
In some embodiments of the invention it is desirable to use optical filters.
Because water behaves as natural short pass filter, it is advantageous to use long pass cutoff filters in conjunction with water bands to form a bandpass filter (although a system that provides sufficient resolution, i.e. a sufficient number of analog-to-digital (A/D) bit, may not require a filter). In the presently preferred embodiment of the invention, any filter in the midst of the H20 absorbance band may be used, e.g. the following filters may be used:
~ 1950 nm long pass filter for the combination band;
~ 1450 nm long pass filter for the first overtone; and ~ 1100 nm long pass filter for the second overtone.
The number of averaged scans for each spectra must be determined, where noise decreases with an increasing number of scans. In the presently preferred embodiment of the invention, noise vs. number of averaged scans is set to 64 averaged scans.
Replicate spectra. Experiments were conducted in which four replicates were collected due to the temperature coefficient of the spectrometer. The following results were obtained (which were due to the spectrometer used to make the measurements --these results are not indicative of a general phenomenon):
~ First replicate - outlier due to temperature;
~ Second replicate - small outlier characteristics; and ~ Fourth replicate - acceptable for further analysis.
to For purpose of experiments conducted in connection with the invention, the following additional parameters were defined:
~ Ionic strength is 0.1 M to match that of the body;
~ pH 7.35 phosphate buffer to approximate that of the body;
~ Temperature maintained at 38.0 ~ 0.2°C to match that of the body;
~ Components: ACS reagent grade chemicals used as standards.
Data Analysis. Data analysis must take into account temperature variations. In the presently preferred embodiment of the invention, temperature variations of 0.
I °C are observed to severely obscure the analyte absorbance bands (even concentrated albumin). Laboratory and instrument temperatures are impossible to control to 0.01 °C
for daily use. This effect is amplified in regions of high water absorbance and large changes of water absorbance due to temperature.
Data analysis must take into account pathlenQth. This consideration is similar to differential measurements taken in dual beam spectrometers, where one beam is 3o focused through the sample and a second beam is focused through a pathlength corresponding to the pathlength interference in the sample.
It is desirable to control pathlength to 0.0001 mm. For a 1 mm cell with buffer present there is a 1 mm pathlength of water. When an analyte is present, the pathlength of water is reduced due to displacement. The displacement is linearly proportional to the concentration of the analyte present. While various components of the sample may be rotated out if their concentration is unknown, such processing is unnecessary upon using the invention because such concentrations are known.
Temperature and pathlen~th correction algorithm. The following discussion provides an exemplary temperature and pathlength correction algorithm in accordance with the l0 invention.
1. Response function: residual (sample - buffer) about zero for regions where the analyte does not absorb.
2. Residual as function of spectral range is inversely weighted by the spectral noise of the sample.
3. Thousands of buffers collected at roughly 38°C are compared with the sample to match temperature. By using thousands of buffers, a good temperature match can be found.
4. For each buffer tested, incremental pathlengths of water are tested to match pathlength of buffer in the sample. For example, to get a pathlength of .99 mm, the buffer being tested as a possible background is multiplied by 0.9900. For example, albumin protein from .95 mm to 1 mm at 0.0005 mm steps is tested with each buffer.
The following is a Matlab~ temperature/pathlength correction program for selected parameters that must be optimized for each analyte, such as pathlength, and regions for the response function:
temppath.M
PROBLEM: Basis Set spectra require background subtraction of temperature and pathlength.
This program corrects the temperature by searching for a buffer collected at the same temperature as the sample and match the amount of buffer present in the sample.
clear to enter wavelength region wavelength = 1100:2:2498;
load sample spectra load a1b12_1 txt sample = albi2-1;
is [o p] = size (sample);
load buffer spectra usually use all buffers collected to date 20 load albbufftxt buff = albbuff;
[m n] = size (buff);
Code minimizes residual over user set regions 25 % These regions can not have absorbance from the analyte they are fine tuned iteratively.
in this case - three regions are used in the response function.
b-1 stmt = find(wavelength>= 1640 & wavelength < 1655);
3o b_2nd~t = find(wavelength>= 2077 & wavelength < 2085);
b-3rd_pt = find(wavelength>= 1640 & wavelength < 1655);
s-1 stmt = b_1 st-pt;
s 2nd~t = b 2nd~t;
s 3rd~t = b 3rd~t;
initialize to large residual pathlength = 0;
for aaa = 1 : p 40 best min(aaa) = 1000000;
end pathlength optimization determine pathlength matching water in sample 45 % (water absorbance * pathlength) for j = 0.95 : 0.0005 : 0.997 %.98 j=manual-pathlength pathlength = pathlength + 1;
temperature optimization for each pathlength, test every buffer for temperature match for temp = l :n avg b-1 stmt = mean(buff(b-1 st~t,temp))*j;
avg b 2nd~t = mean(buff(b 2nd~t,temp))*j;
avg b 3rd-pt = mean(buff(b 3rd~t,temp))*j, to repeat for every sample and replicate for sample num = 1 : p avg s-1 st-pt = mean(sample(s lst~t,sample_num));
avg s 2nd-pt = mean(sample(s 2nd~t,sample_num)) ;
avg s 3rd~t = mean(sample(s 3rd~t,sample num));
15 diff_1 stmt = abs( avg s-1 st-pt - avg b-1 stmt );
diff_2nd-pt = abs(avg s 2nd~t - avg b 2nd-pt );
diff 3rd~t = abs( avg s 3rd-pt - avg b 3rd~t );
store results of each loop results(sample num) = diff 1 stmt + diff 2nd-pt + diff 3rd~t;
usually add in weighting function as inverse of noise for each region here if response function for given sample is best - record parameters if results(sample_num) < best_min(sample num) best_min(sample num) = results(sample_num);
best-pathlength(sample_num) = j;
best temp(sample num) = temp;
end % end if end % end sample end % end temperature end % end pathlength % dump best parameters to screen for interpretation best_min best~athlength best temp plot temperature and pathlength corrected spectra hold off clg hold on = ysoo 2soo -o.ol o.o~~;
axis(v);
for sample num = 1:p best sample corr(:,sample num) = sample(:,sample num) - buff( :,best temp(sample num)) * best~athlength(sample_num);
plot (wavelength,best sample corr(:,sample_num));
intensity(sample num) = best sample_corr(481,sample_num);
end The resulting spectra are clean and baseline resolved. The spectra are selected to cover physiological concentrations for each analyte.
to The following example illustrates the generation of a first basis set "Basis Set L"
Example - Basis Set I
near-IR absorbance features over the 1500 to 2500 nm spectral region are provided for water, albumin protein, globulin protein, triacetin, cholesterol, urea, and glucose with a 1 mm pathlength. In addition, sample temperature effects are included along with instrument noise levels.
Experimental: Spectra of the major constituents of serum were collected over their 2o respective physiological ranges. Sample preparation consisted of dissolving dried, reagent grade solid samples in a 0.1 M phosphate buffer adjusted to pH 7.35.
All spectra were collected on a NIRS 5000 in transmission mode, with a 1 mm path length infrasil quartz cell, with a 120 second equilibration period, at 38.0°C, with 64 averaged scans, done in quadruplicate. A single instrument was used for all data acquisition.
Results and Discussion: Spectra are analyzed in order of decreasing absorbance changes in the two spectral windows from 2050 to 2350 nm and 1550 to 1850 nm.
The first replicate is discarded in all cases due to a consistent variation in temperature 3o caused by the instability of the NIRS 5000 spectrometer and photons heating the sample (data not included). The sample is in equilibrium by the second sample replicate.
Water Spectra: The near-IR is dominated by three large water absorbance bands centered at 2500, 1950, and 1450 nm as presented in Fig. 4. The high absorbance limits analysis done in aqueous solution in the near-IR to three spectral regions. The region from 2350 to 2050 nm is referred to herein as the combination band region; the region from 1850 to 1550 nm is referred to herein as the first overtone spectral region;
and the region from 1400 to 1100 nm is referred to as the second overtone spectral region.
The NIRS spectrometer sets the gain and hence the dynamic range of the detector based upon the spectral region with the most light intensity reaching the detector.
This is the second overtone spectral region for aqueous samples. However, the combination band region has the largest absorbance, followed by the first overtone region, and then the second overtone region. Due to the low absorbance of water in the 1300 nm region versus the 2200 nm region, a relatively small dynamic range is left for the 2200 nm region where glucose bands are the largest. Therefore, a 1500 nm long pass filter was employed which forces the NIRS system to set the gain based upon the first overtone spectral region. Hence, in the initial basis set no spectral information is provided for the second overtone region, optimum signal to noise levels are provided for the first overtone spectral region, and slightly degraded signal to noise levels are obtained for the combination bands. Among many modifications made to the NIRS system is an order sorter which allows a different gain setting for each of the three spectral regions during a single scan for the next basis data set.
Temperature Effects on Water Spectra: All three water absorbance bands in the near-IR shift to higher frequency with increasing temperature. Buffer spectra collected from 38.2 to 43.0°C are presented in Fig. 5A. The instrument should be modified to collect lower temperature spectra. A slight broadening of the lines can be observed on each of the water absorbance band shoulders. Subtracting a spectrum of water collected at 38.2°C from spectra of water collected at higher temperatures reveals the 3o magnitude and direction of the shift. Negative absorbance bands that increase with temperature are observed at 2000 and 1480 nm. As the water bands shift to higher temperature, there is less water absorbance in these regions, so that subtracting out a water absorbance band from a lower temperature results in too much background being subtracted. Positive absorbance bands that correlate with increasing temperature are observed at 2300, 1890, and 1400 nm. With increasing temperature, the water absorbance increasingly moves into these spectral regions.
Subtracting out the 38.2°C water spectrum does not subtract out enough in these regions.
The large water absorbance, coupled with the temperature shift, greatly hinders near-IR analysis. Fig. 5B reveals that in the subtraction, no useful information is obtained where the raw absorbance is greater than 3.0 ~ 0.1, indicating the limit of the dynamic range of the NIRS system. Therefore, the regions above 2460 nm and from 2010 to l0 1890 nm result in no analytically useful information and may be discarded for data collected with a 1 mm pathlength. Information in these spectral regions may be obtained by adjusting the pathlength. Due to the water absorption, the width of the regions that need to be discarded increases as the pathlength analyzed increases. In addition, the temperature effects are seen to span the entire combination band region and first overtone spectral region. As will be shown, these changes in baseline are roughly equal in magnitude to the highly absorbing protein and much greater in magnitude than all other spectral analytes examined.
Albumin Protein: After water and temperature effects, serum spectra are primarily composed of absorption from albumin protein which has a physiological range of 2.6 to 7.9 g/dL. Albumin protein absorbance bands are difficult to see in the presence of water, as shown in Fig. 6A. Subtracting out a buffer spectrum results in protein absorbance peaks at 2285, 2170, 1730, and 1690 nm, as shown in Fig. 6B. Large negative absorption bands also appear in the resulting spectra where water absorbs.
These bands are not primarily due to variation in temperature as a derivative of the water band would appear as seen in Fig. 5B. The negative bands are due to displacement of water by albumin and scattering.
A program, such as the Matlab~ program described above, is used to determine the 3o best buffer in terms of temperature and best calculated pathlength to be used as a background spectrum for subtraction. In Fig. 6B, the buffer and albumin in buffer spectra both had the same 1 mm fixed pathlength. Because albumin is present in the 1 to 12 g/dL range in this example and water is 100 g/dL, the albumin occupies a significant volume of the cell and less water is present per unit volume. A
program was written that multiplies the water spectrum by a percentage that can be sequentially varied over a wide range. The optimum calculated pathlength for each albumin in buffer spectrum was determined by minimizing the sum of the absolute value of the residuals in locations where albumin does not absorb and temperature effects are at a minimum (2085 to 2077 and 1655 to 1640 nm). The residual in the overtone region was weighted twice as much to compensate for the higher noise in the combination band region. To further minimize temperature effects, all buffer spectra collected were run through this optimization to find the best buffer in terms of temperature matching with the sample. Each albumin in buffer spectrum was run through this algorithm independently.
The results of subtracting the best buffer at the adjusted pathlength for each albumin spectrum are presented in Fig. 7. Additional albumin absorbance bands are now visible at 2060 nm and 2335 nrn. Expansion of the graph about the 2060 nm absorbance bands reveals increasing absorbance for each increase in albumin concentration. The albumin band centered at 2170 nm is more symmetrical than the one seen in Fig. 6B. The two peaks in the first overtone spectral region have a better baseline correction and now increase in absorption linearly with increasing concentration. However, negative absorbance bands are still evident where water absorbs at 2020 and 1870 nm. The region between 2000 and 1900 nm is an artifact of the mathematical correction over a region where the absorbance is greater than 3 and the system does not respond. In addition, there is a large difference between absorbance and the 1 g/dL albumin spectrum. This difference should be equal to the difference in absorbance from 1 to 2 g/dL. This offset can be reduced if the combination band region and the overtone region are treated individually. It should be pointed out that no baseline correction, smoothing, or scatter correction has been employed at this point.
3o Globulin Protein: Physiological concentrations of globulin (0.7 to 8.1 g/dL) are less absorbing in the near-IR than albumin. Straight subtraction of the phosphate buffer allows the same peaks to be observed that are seen in albumin protein, as shown in Fig. 8. The temperature and pathlength correction algorithm discussed above was run with exactly the same parameters as for albumin and the same additional extra peaks were found, as shown on Fig. 9. Overlaying the albumin and globulin spectra reveals that the globulin absorbance band centered at 2170 nm is slightly broader than that of albumin protein.
The calculated pathlengths required for the background subtraction from each of the spectra are presented in Fig. 10. For albumin, the correction is linear with increasing concentration, but has a y-axis intercept of 0.996 mm. 'This is consistent with the poor baseline observed in Fig. 7. The corrections for globulin are also linear, but greater to corrections are required per mg/dL analyte. This is consistent with the scattering tendencies of globulin. The y-axis intercept is 1.00, consistent with the excellent background subtraction.
Triglycerides: Triacetin is used to simulate triglycerides. The physiological range of triacetin is 50 to 450 mg/dL. The temperature and pathlength correction algorithm is again employed, but different regions are used in determining the minimum residual (2420 to 2440, 2080 to 2090, and 1575 to 1635, weighted 1:5:20). Six triacetin absorbance bands result centered at 2320, 2250, 2130, 1760, 1715, and 1675 nm, as shown on Fig. 11. The resulting pathlengths required for correction are linear with 2o concentration, but much smaller deviations from 1 mm result due to the lower concentration of triacetin versus protein in serum, as shown on Fig. 10. The signal levels of the smaller triacetin absorbance bands approach the noise level of the spectrometer.
Urea: Twelve urea in buffer spectra were collected. Due to the small physiological concentration of urea (6 to 123 mg/dL), the algorithm used to optimize the background by changing the effective pathlength of the buffer subtracted fails because no significant Figs. of water are displaced. No temperature matching algorithm is employed, but buffer spectra collected with each sample are used. A straight 3o background subtraction followed by a two point baseline correction (2094 to 2106 and 2320 to 2332 nm) was performed and is presented in Fig. 12. A single absorbance band is present centered at 2190 nm. No overtone peak is present. This is consistent with this absorption being related to N-H, whereas all of the other analyses have O-H
fundamental vibrations. Only four spectra are presented due to Iarge baseline drifts that obscure the linearity of the additional spectra. Higher concentration samples can be run to obtain a higher S/N and cleaner spectra, although the same resulting basis set is obtained.
Glucose: A complete glucose in buffer study was performed over the combination and first overtone spectral region of which a subset is presented here.
Glucose was examined from 30 to 600 mg/dL (also from 0 to 5000 mg/dL) to cover the physiological as well as hypoglycemic and hyperglycemic levels of glucose. A
to straight subtraction of buffer from glucose in buffer shows absorbance bands centered at 2326, 2272, 1800, 2150, 1730, and 1590 nm, as shown in Fig. 13.
Conclusions: Consistent with theory, for all analytes, the combination band spectral region yields larger absorbance than the first overtone spectral region.
However, longer pathlengths quickly degrade the signal to noise level in the combination band region due to the Iarge water absorbance, whereas the spectral quality in the first overtone spectral region should increase with small millimeter increases in pathlength.
The absorbance bands in the region of glucose absorbance in decreasing order of absorbance are water, temperature effects, albumin protein, globulin protein, 2o cholesterol, triglycerides, urea, and glucose. While every analyte analyzed absorbs more than glucose and over the same general spectral region, every analyte has a distinct absorbance signature. In principle, the serum spectra or the noninvasive spectra, can be deconvoluted.
***
The invention contemplates the generation of additional basis sets, such that substantially all interfering components are identified and factored into the spectroscopic analysis.
The following example illustrates the generation of a second basis set "Basis Set IL"
Example - Basis Set II
A study was rerun on dried, crushed, and pressed solid samples to give absorbance spectra with no water. A second basis set was collected based upon spectra of solid or neat components of human serum. The resulting absorbance spectra show the combination, first, and second overtone absorbance bands. In addition, for a given component the relative absorbance between regions may be compared. Combined, another method of wavelength selection is made available.
to Experimental: Pure component spectra of the liquid form of water (pH 7.35, 0.1 M
phosphate buffer 38.0 t 0.2°C), triacetin, and lactic acid were collected. Albumin, globulin, cholesterol, urea, and glucose exist as a solid in their pure state.
For these analytes, each was individually ground with a mortal and pestle to a fine powder in the absence of potassium bromide. The powder was then compressed into a transparent pellet in a specially designed press that fits into the NIRS 5000 transmission module.
Four replicates of each component were then obtained in the transmission mode.
'The pathlength of each analyte was not controlled.
Results and Discussion: The raw absorbance spectra for water, albumin, globulin, 2o cholesterol, triacetin, urea, glucose, and lactic acid are presented in Fig. 14. Because the pathlength of each pellet was not controlled, the relative absorbance between components can not be compared. The relative absorbance between frequencies for a given analyte can be compared. The large baseline offsets are due to the thickness of the sample and resulting total light throughput. This plot is included to show the total absorbance of each analyte relative to the dynamic range of the NIRS 5000.
For each component in Fig. 14, the minimum absorbance was subtracted out and the resulting spectra was normalized to 1 absorbance unit, as shown in Fig. 15.
The resulting full scale plots make it easier to compare absorbance as a function of 3o frequency and differences between components. For all three spectral regions, i.e.
combination (2050 to 2350 nm), first (1550 to 1850 nm), and second overtone (1100 to 1400 nm), the absorbance bands are observed to be distinct. In principle, each component can be deconvoluted. It should be noted that when interacting with water, these absorbance bands may shift and broaden. Comparing with the aqueous absorbance from Basis Set I (above) reveals the absorbance bands of the neat or solid water ( 140), albumin ( 141 ), globulin ( 142), and triacetin ( 143 ) to be in the same location with the same widths. Both urea (144) and glucose (145) reveal additional resolution of peaks that have broadened and merged in the presence of water.
Several key spectral signatures emerge from this Example. First, the combination band region contains absorbance for each of the individual analytes. These absorbances are in every case more intense than those in the first and second overtone 1o spectral regions. Cholesterol (146) absorbance drops off rapidly in this region as does triacetin. Neither interferes significantly with the glucose absorbance band centered at 2150 nm. The only interference is from water, albumin, and globulin which are shown in the Example - Linearity Study (below) to be removable by simple subtraction.
In the first overtone spectral region every component has an absorbance band except urea with its N-H bonds. Here the intensities of the absorbance bands range from 15% to 50% that of the corresponding combination band absorbance. It should be recognized that these values are for a fixed pathlength and can be adjusted based upon 2o total pathlength.
The second overtone spectral region has absorbance bands for every component examined, but the relative absorbances are the smallest, as shown on Fig. 16.
The glucose band ( 145) seen here is very difficult to see in the presence of water ( 140).
Conclusions: Each of the three regions contains information about every analyte with the exception of urea in the first overtone spectral region. The absorbance bands are highly overlapped and are generally less intense at higher frequencies. The absorbance bands are all distinct.
***
The following example illustrates the generation of a third basis set "Basis Set IIL"
Example - Basis Set III
The first basis set was repeated with no edge filter present to allow comparison of all spectral ranges. The first Example above used a 1500 nm long pass filter to force the NIRS spectrometer to gain range on the 1700 nm spectral region. This Example could be repeated with increased optical pathlengths to yield higher signal to noise levels in the first and second overtone spectral regions.
***
l0 The following example illustrates the generation of a fourth basis set "Basis Set IV."
Example - Basis Set IV
It is necessary to measure interactions of molecules in solution. In this Example, a serum data set is collected.
Data Sets: 'The first data set consists of spectra of glucose dissolved in a 0.1 M
phosphate buffer adjusted to pH 7.35. Reagent grade glucose was weighed and 2o diluted to a known volume with the 0.1 M phosphate buffer. Spectra were collected in the transmission mode with a 1 mm quartz cell using the NIRS 5000 spectrometer over the range of 1100 to 2500 nm with readings taken every 2 nm. A 1500 nm long pass filter was placed before the sample to force the NIRS spectrometer to set the gain on the peak signal at 1600 nm. Before and after every sample, 7 spectra of the 0.1 M
phosphate buffer were collected. A total of 64 glucose~ag~ samples were collected with 7 sequential replicates of each sample. The glucose samples covered a dynamic range of approximately 20 to 600 mg/dL. All samples were maintained at 38.0 ~
0.2°C.
The second data set consists of serum samples prepared by Western States Plasma.
3o Each serum sample was analyzed using a standard SMAC (serial mufti-channel automated chemistry) analysis yielding concentrations for calcium, ionized calcium (calculated), phosphorus, glucose, uric acid, urea nitrogen (BLTN), creatinine, creatinineBLJN ratio, total protein, albumin, globulin, A/G ratio, total bilirubin, ALT, ALP, LD (LDH), AST, GGT, sodium, potassium, chloride, carbon dioxide, triglycerides, and cholesterol. To extend the dynamic range and level the concentration distribution, reagent grade urea and glucose were quantitatively added to the serum samples. The NIRS 5000 spectrometer was used in the fashion described above with the same wavelength region, pathlength, temperature control, and long pass filter. A 0.1 M phosphate buffer adjusted to pH 7.35 was run before and after each serum sample. A total of 196 serum samples were collected with 4 sequential replicates of each sample. The glucose analyte covered a dynamic range of approximately 20 to 600 mg/dL.
to Experimental: Glucose is determined in each data set using PLS regression analysis.
The data sets are broken up into calibration and prediction keeping all replicate spectra together. A data point was originally collected every 2 nm from 1100 to 2500 nm. Additional data sets are formed from this data set by keeping every other point, every 3rd point, every 4th point, ..., to every 32°d point. PLS
calibration models and predictions are then determined using 1 to 10 PLS factors.
Results and Discussion: For each resolution, the resulting standard error of calibration (SEC) and standard error of prediction (SEP) is determined for 1 to 10 PLS
2o factors, as shown on Fig. 17. Here, selection of the optimum number of factors needs to be achieved. As different ranges need to be compared, differences in the number of PLS factors employed can lead to erroneous conclusions. Statistical approaches to determining the optimum number of factors have failed. Because the SEP does not increase as the system is over-modeled, and further because the SEC and SEP
yield similar results with 10 factors, it was decided for the purposes of this Example only to compare standard errors from range to range using the results obtained with ten PLS
factors.
Ten spectral ranges are analyzed in both the glucose in water and glucose in serum 3o data sets. These are summarized in Table 1 below. Ranges 1 to 3 and 5 to 7 correspond to the full width at zero height of the six glucose absorbance bands isolated in the near-IR. Ranges 4 and 8 splice together regions 1 to 3 and 5 to 7, respectively. Ranges 9 and 10 expand regions 4 and 8 into regions of increasing water absorbance, increasing noise, and no additional glucose information.
Table 1: Spectral Ranges Employed Ran a NumberS ectral Ran a nm 1 2078 to 2243 2 2243 to 2272 3 2297 to 2366 4 2078 to 2366 5 1587 to 1674 6 1674 to 1709 7 1709 to 1754 8 1587 to 1754 9 2000 to 2500 1520 to 1805 Clearly, the wider spectral region that incorporates more glucose information (and water and temperature) results in a lower standard error at any resolution than any of the three individual glucose absorbance bands.
The nominal resolution of the NIRS spectrometer is 10 nm for the standard 0.040"
exit slit used in this Example. Still, the standard error is observed to increase slightly as the resolution degrades from 2 to 10 nm. This is due to the manner in which the data sets were created from the original 2 nm resolution data set. For instance, in the ~5 6 nm resolution data set generated, every third spectral point is kept.
This means that two-thirds of the data are discarded. The discarded data has glucose, water, and temperature signal. In addition, by keeping these extra points, the effective noise is decreased by signal averaging. In as much as the true resolution of the NIRS
5000 is 10 nm, 100% of the slope observed on the SE vs. resolution graph is due to this 2o systematic error. In addition, the same slope is observed from 10 to 32 nm resolution.
The original data set with points every two nanometers was again broken down into data sets with resolution ranging from 2 to 32 nm at 2 nm intervals. This time, the data was averaged instead of just discarding extra points. For example, at 6 nm resolution points at 1100, 1102, and 1104 nm were averaged to a single point.
The next point averaged the data points at 1106, 1108, and 1110 nm. The PLS
analysis was then repeated and the standard errors with the tenth factor determined, as shown in Fig. 18. The increase in standard error observed with degrading resolution is observed to range from 5 to 10 mg/dL standard error as opposed to 5 to 25 mg/dL
standard error from 2 to 32 nm resolution. Clearly, the failure to average the data points results in an increase of the slope of standard error versus resolution. While the standard error roughly doubles from 2 to 32 nm resolution, the data indicates that for a glucose in water solution, the acceptable resolution may be 32 nm or more.
'This 1o makes chemical sense in as much as the narrowest absorbance band in this Example is 54 nm wide. In must be pointed out that there are no spectral interferences in this Example. Therefore, the actual acceptable resolution can only degrade from this resolution.
2°C for the first overtone region, for data sets generated at 2 to 32 nm resolution using averaged data, the increase in standard error with degrading resolution is greatly reduced, as shown on Fig. 19. In addition, for this spectral region, less than ten points are retained at resolutions greater than 16 nm. The PLS algorithm used only operates on as many factors as there are data points. If queried for standard errors with 2o additional factors, the standard error for the number of factors equal to the number of points available is generated. Because the standard errors continue to decrease with an increasing number of factors in this Example (see Fig. 17), the comparison of standard errors for various resolutions using ten PLS factors is not valid. A
direct comparison of standard errors at degrading resolution for the 1587 to 1754 nm spectral region with six PLS factors is presented in Fig. 20. The increased standard error observed with degrading resolution is now not observed with resolutions under 15 nm. 'This is a true comparison of standard errors for this spectral region.
The results in Fig. 18 for the 2078 to 2366 nm spectral region are still valid due to its large range which contains ten or more points up to 30 nm resolution.
Glucose in Serum: The SEC and SEP plots versus resolution for glucose in the serum study for the ten different spectral regions are provided in Figs. 21 to 30.
The results are generally the same as for glucose in water.
The combination band region is analyzed first. Range 1 with the largest glucose absorbance band yields the lowest standard errors for a region isolating a single glucose absorbance band, as shown on Fig. 21. Ranges 2 and 3 yield larger standard errors and have smaller glucose absorbance bands with a decreased signal to noise level, as shown on Figs. 22 and 23. Analysis of ranges 2 and 3 at degraded resolutions is limited by the number of data points present in each range. Range 4 which couples the first three regions demonstrates the lowest standard errors, as shown on Fig. 24.
Again, the averaging of points reduces the increase in standard error with degrading 1o resolution. The increase in standard error from 35 to 50 mg/dL observed as resolution degrades from 2 to 30 nm is entirely due to the loss of information in extracting rather than averaging data points. While the standard errors are higher than in the glucose in water Example, this Example demonstrates that even in the presence of all of the spectral interferences, except skin and blood cells, the resolution is essentially not an effect until after a resolution of 30 nm. This is the same result as for glucose in water.
The number of PLS factors incorporated is not an issue due to the fact that 10 points are present even at 30 nm resolution. Range 9 incorporates all of range 4 and extends past where glucose absorbs at both higher and lower frequencies, as shown in Fig. 29.
No resolution effect on standard error is observed from 2 to 32 nm.
The effects of resolution in the first overtone spectral region are more difficult to interpret due to decreased signal to noise and the narrower spectral ranges chosen.
Range 5 has the largest glucose absorbance band in the overtone spectral window and results in the lowest standard errors. Ranges 6 and 7 were shown to have very poor signal to noise levels for glucose in water (not presented). The standard errors are essentially mean centered prediction values, as shown on Figs. 26 and 27. The effect is worsened at degrading resolution due to the number of points in each spectral range.
Range 8 reveals real glucose predictions, as shown on Fig. 28. This range was reanalyzed with the averaged rather than the selected data, see Fig. 31. Using ten PLS
3o factors, the increasing standard error with degrading resolution observed is virtually identical to the nonaveraged data due to the number of points present in the data. This is shown by comparing the standard errors with only six PLS factors (6 points present at 30 nm resolution). No resolution effect is observed until a resolution of 20 nm.
Range 10 which expands to higher and lower frequencies from range 8 has 10 data points present at 30 nm resolution and shows no resolution effect until 20 nm, as shown in Fig. 30.
Conclusions: The glucose in water data set has sufficient signal to noise to determine glucose with the specifications required. The rise in standard error for the narrow glucose absorbance bands with degrading resolution is not real. It is partially the result of selecting the points rather than averaging the points to generate new data sets.
In addition, the new data sets did not contain enough data points to compare analysis to of 2 nm resolution data and 32 nm resolution data with ten PLS factors.
Resolution effects may be addressed by using fewer PLS factors in this comparison or by using larger spectral ranges. For both methods, the resolution effects are minimal to 30 nm in the combination band region and 15 nm in the first overtone spectral window.
Because it is preferred to get the highest signal to noise ratio possible from the instrument, it is acceptable to have 30 nm resolution. That is, by having less (but, nonetheless, acceptable) resolution, e.g. by having 30 nm resolution instead of 10 nm resolution, the instrument captures more signal relative to noise. Thus, even though the resolution is coarser, more information is contained in signal generated by the 2o instrument. As a result, the resolution selected in the preferred embodiment of the invention provides a more accurate picture of the spectra, even though the instrument has coarser resolution. This is because there is a higher signal to noise ratio at the resolution required. In contrast, if extra resolution were available in the instrument, but there was a lower signal to noise ratio, less information would be available for processing by the basis sets.
In the Example, the glucose in serum data sets resulted in roughly three times the standard error as in the glucose data set. Again, analysis is limited to either large spectral windows or to comparisons with fewer PLS factors for narrower ranges.
In 3o the combination band spectral region, the increase in standard error observed with degrading resolution is minimal to 30 nm resolution. In the first overtone spectral window, the slope to standard error versus resolution is minimal to 20 nm resolution.
These results are virtually identical to those generated in the glucose in water study.
The effects of the proteins, triglycerides, cholesterol, urea, salts, and minor organic constituents are observed not to effect the required resolution.
***
Example - Basis Set V
It is necessary to measure effect of scattering of whole blood cells. This basis set is generated as follows:
~ Collect blood data set in transmission and as diffuse reflectance.
~ Repeat component extraction.
~ Couple in scatter correction ~ Deconvolve (see deconvolution discussion below).
2o Example - Basis Set VI
It is necessary to measure the effect of skin. Animal studies are performed and all prior analysis techniques are repeated. Noninvasive studies can be viewed as extensions of the basis set.
***
Uses of Basis Sets.
Chemical and physical knowledge of a system are required for such factors as:
~ Intelligent wavelength selection, e.g. knowledge of the location and degree of interferences of each analyte.
~ Interpretation of noise levels as a function of region.
~ Interpretation of signal levels for each analytes as a function of wavelength.
~ Selection of optimal signal to noise regions for each analyte.
Resolution specifications for an instrument implementation of the invention are set forth above. The number of analog-to-digital (A/D) bits required to provide appropriate instrument resolution can be calculated from noninvasive spectra and glucose intensities (absorbance). For this determination, it is necessary to know the maximum intensity of the whole system and the intensity of glucose at the required standard error. If the maximum intensity of the sample is 10 to the negative absorbants unit, it is only necessary to calculate the intensity of the body scan, including all absorbants. To determine the intensity of glucose, the required standard error is 9 mg/dL. The intensity of glucose and water, and the intensity of water is used (as described above) to calculate the intensity of the glucose and water minus the intensity of water. This results in a value for the intensity of glucose. Once the intensity of glucose is determined, it is then necessary to determine the change in intensity of glucose, e.g. by drawing in a base line to the peak, and plotting the change in intensity of glucose versus glucose concentration. This provides a best fit of the data that can be fitted to a line to calculate the change in intensity at 9 mg/dL. Once this value is obtained, the ratio of this value to maximum intensity of the glucose is readily calculated. This ratio defines the number of bits that are required in the system for analog-to-digital conversion. For example, if the ratio is 50,000, then a 16 bit A/D
is required because sufficient quantization must be provided to avoid aliasing problems. Thus, the basis set is useful in defining instrument parameters.
Interpretation of multivariate results. Multivariate results are difficult to validate.
Standard errors must correlate with basis set information. If noisy regions are added, 3o the signal to noise ratio decreases. It is therefore necessary to correlate standard errors with the signal to noise ratio.
With regard to the removal of second, third, ... order light in a grating based spectrometer, a long pass filter is required. The basis set dictates the specifications of the filter.
With regard to the removal of scatter, such determinations are based upon refractive index change. In the preferred embodiment of the invention, the basis sets remove scatter and temperature effects. This step is repeated for additional analytes, and the reduced spectra are further processed using multivariate approaches.
Deconvolution of noninvasive spectra. The partial deconvolution reduces the rank of first temperature and water, then proteins, then organic constituents. The resulting spectra can then be fed into the multivariate approaches. However, the reduced dynamic range of signals forces PLS to lock in on smaller analytes, such as urea and glucose, instead of water and temperature.
There are a limited number of interferences for glucose in the near-IR. The major interferences have convenient breaks in concentration. The largest concentrations /
effects are temperature and water. Processing should remove the Refractive index, which is on the order of 100 g/dL.
Large concentration gaps exists between water and the proteins. Iterative deconvolution can be used to take advantage of this fact.
Albumin and globulin proteins are on the order of 1 to 7 g/dL. These interferants are easily identified and removed by spectral subtraction or rotation.
Example - Linearity Introduction: The basis set is used to determine the location and intensity of each of the major species interfering with glucose. It also demonstrates that for a given component, the absorbance increases linearly with increasing concentration. In this Example, it is shown that the absorbance of multiple components is the sum of the individual components, as assumed by Beer's law. This is critical to the herein described approach of using spectral subtraction of chemical information to enhance the signal to noise level of glucose.
Experimental: Spectra were collected in quadruplicate with a NIRS 5000 spectrometer configured in the transmission mode with a 1 mm pathlength quartz sample cell. All samples are reagent grade and were prepared in a 0.1 M
phosphate buffer at pH 7.35 and spectra were collected at 38.0 ~ 0.2°C.
Six single analyte solutions were prepared: 4000 & 8000 mg/dL albumin, 2000 &
4000 mg/dL globulin, and 200 & 400 mg/dL triacetin. Eight additional samples were prepared consisting of all possible permutations and combinations of the above six sample concentrations. For example, one sample consisted of 8000 mg/dL
albumin, 2000 mgldL globulin, and 200 mg/dL triacetin.
Results and Discussion: Three spectra of water, 8000 mg/dL albumin, 2000 mg/dL
globulin and 200 mg/dL triacetin appear primarily as water absorbance bands, as shown on Fig. 32. Subtraction of the water with the same algorithm used in the basis data set that attempts to match pathlength and temperature effects (discussed above) was employed to minimize the residual about zero absorbance over the spectral ranges 1640 to 1655 nm and 2077 to 2085 nm, as shown in Fig. 33. Results of incomplete temperature and pathlength subtraction dominate in the regions surrounding 1890 to 2010 nm where no signal results due to large water absorbance. The resulting spectra show the six dominant protein absorbance bands centered at 1690, 1730, 2060, 2170, 2285, and 2335 nm.
Spectra of the single analyte albumin samples are shown in Fig. 34. The 8000 mg/dL
albumin peaks are nearly exactly double the 4000 mg/dL albumin peaks indicating that Beer's law is holding. The average of the 8000 mg/dL albumin spectra was subtracted from the spectra in Fig. 33 to yield the spectra shown in Fig. 35.
Overlaid 3o with this are the 2000 mgldL globulin spectra. Clearly, the basic shape of the globulin spectra is discernible after subtraction of the 100,000 mg/dL (100 g/dL) water and the 8000 mg/dL albumin. The difference is the sum of the 200 mg/dL triacetin and baseline drift.
Spectra of the single analyte globulin samples are shown in Fig. 36. The 4000 mg/dL
globulin peaks (260, 261, 262) are nearly exactly double the 2000 mg/dL
globulin peaks (263, 264). Again, the average of the 2000 mg/dL globulin spectra is subtracted from the spectra shown in Fig. 33 to yield the spectra in Fig. 37. Overlaid with this are the standard 200 mg/dL triacetin spectra. Once again, the 200 mg/dL
triacetin peaks centered at 1675, 1715, 1760, 2130, 2250, and 2320 mg/dL can be seen after the subtraction of 100,000 mg/dL water, 8000 mgldL albumin, and the 2000 mg/dL
globulin. Unknown concentrations may be subtracted by rotation.
io Conclusions: For a relatively simple mixture, subtraction of the high concentration water, albumin, and globulin results in spectra of triacetin. Clearly, small errors in temperature and pathlength correction propagate into large errors of baseline for the lower concentration analytes. It is also possible that the error in subtraction may be due to scattering. To correct for this, a standard multiple scatter correction algorithm may be used. Clearly, straight subtraction can yield spectra that visually appear to yield higher signal to noise for the lower concentration analytes.
NOTE: the only two species in serum that have higher near-IR absorption than 2o glucose that were not included in this Example are cholesterol and urea.
***
In applying the invention, direct spectral subtraction is replaced with iterative subtraction, based upon regions of minimal or defined absorbance of remaining analytes. In another, equally preferred embodiment of the invention, another concentration gap may be taken advantage of for purposes of isolating the analyte vis-a-vis interferants. Two presently preferred approaches include:
3o ~ Analyze with multivariate techniques because the dynamic range of interferences and glucose is the same; and ~ Further removal of triglycerides, cholesterol, urea by deconvolution/
subtraction.
One approach to generating basis sets is iterative. For example, within a sample, after subtracting water, a determination of albumin and globulin is made. Once albumin and globulin are determined, and there is knowledge of water concentration, the albumin and globulin may be again removed, only this time more accurately.
This iterative process proceeds to some predetermined limit of precision, and then triglycerides and cholesterol are integrated into the analysis.
to Although the invention is described herein with reference to the preferred embodiment, one skilled in the art will readily appreciate that other applications may be substituted for those set forth herein without departing from the spirit and scope of the present invention. Accordingly, the invention should only be limited by the Claims included below.
Fig. 12a is a plot of absorbance vs. wavelength showing urea- buffer;
Fig. 12 b is a plot of absorbance vs. wavelength showing urea - buffer (baseline corrected);
Fig. 13 is a plot of absorbance vs. wavelength showing glucose - buffer;
Fig. 14 is a plot of absorbance vs. wavelength for solid samples;
Fig. 15 is a plot of normalized absorbance vs. wavelength for solid samples;
Fig. 16 is a second plot of normalized absorbance vs. wavelength for solid samples;
Fig. 17 is a plot of standard error (mg/dL) vs. number of PLS factors for glucose~aq~;
Fig. 18a is another view of the fourth plot of standard error (mg/dL) vs.
resolution (nm) for glucose~a~;
Fig. 18b is another view of the fourth plot of standard error (mg/dL) vs.
resolution (nm) for glucose~aq~ showing an expanded y-axis;
Fig. 19 is another view of the eighth plot of standard error (mg/dL) vs.
resolution (nm) for glucose~aq~;
Fig. 20 is another view of the eighth plot of standard error (mg/dL) vs.
resolution (nm) for glucose~~~ showing averaged points of six PLS factors;
Fig. 21 is a plot of standard error (mg/dL) vs. resolution (nrn) for glucose in serum;
Fig. 22 is a second plot of standard error (mg/dL) vs. resolution (nm) for glucose in serum;
Fig. 23 is a third plot of standard error (mg/dL) vs. resolution (nm) for glucose in serum;
Fig. 24 is a fourth plot of standard error (mg/dL) vs. resolution (nm) for glucose in serum;
1o Fig. 25 is a fifth plot of standard error (mg/dL) vs. resolution (nm) for glucose in serum;
Fig. 26 is a sixth plot of standard error (mg/dL) vs. resolution (nm) for glucose in serum;
~5 Fig. 27 is a seventh plot of standard error (mg/dL) vs. resolution (nm) for glucose in serum;
Fig. 28 is an eighth plot of standard error (mg/dL) vs. resolution (nm) for glucose in 20 serum;
Fig. 29 is a ninth plot of standard error (mg/dL) vs. resolution (nm) for glucose in serum;
25 Fig. 30 is a tenth plot of standard error (mg/dL) vs. resolution (nm) for glucose in serum;
Fig. 31 is another view of the eighth plot of standard error (mg/dL) vs.
resolution (nm) for glucose in serum;
Fig. 32 is a plot of absorbance vs. wavelength showing raw absorbance for a sample containing water, albumin, globulin, and triacetin;
Fig. 33 is a plot of absorbance vs. wavelength for a sample containing albumin, globulin, and triacetin and from which water is subtracted and temperature and pathlength are corrected;
Fig. 34 is a plot of absorbance vs. wavelength showing linearity for albumin spectra where temperature and pathlength are corrected;
Fig. 35 is a plot of absorbance vs. wavelength for a sample containing globulin and triacetin and from which water and albumin are subtracted;
to Fig. 36 is a plot of absorbance vs. wavelength showing linearity for globulin spectra where temperature and pathlength are corrected; and Fig. 37 is a plot of absorbance vs. wavelength for a sample containing triacetin and from which water, albumin, and globulin are subtracted.
DETAILED DESCRIPTION OF THE INVENTION
The following discussion describes what the basis set is, how it is collected, the 2o instrument that is required, the data collection parameters, the data analysis as far as such factors as temperature and path length are concerned, and what are considered to be additional basis sets.
The simplest example of a basis set is a basis set that includes all interfering components in a sample, such as serum. These components can include, for example, water, temperature/hydrogen bonding effects, albumin, and globulin protein, triglycerides, cholesterol, urea, all organic components, and Na+, K+, and Cl-. To a lesser degree, the basis set may include additional electrolytes.
3o The basis set does not include those components that do not interfere, such as anything that in terms of concentration is less than the background signal or noise.
With regard to an analyte, such as glucose, it is necessary to define those components of a sample that have a larger interference than that of glucose. Instead of considering only the analytes that are mentioned above, which are all in blood or serum, a basis set may be generated, for example, that produces a transform for the red blood cells that interfere or scatter the light; and also for skin effects.
Once the spectra of all these components is known, it is then necessary to determine how each of these components interact, e.g. taking serum data, extracting each of the components, and then comparing the spectra for the individual components with that of the components in solution.
to Once a basis set is generated for glucose in the presence of water, it is determined that water interferes with glucose, and it is determined how to remove the water, then a basis set for a next component can be generated, such as for temperature effect. In the example of non-invasive glucose concentration determination, the invention sequentially adds basis sets for other components, e.g. globulin, protein, triglycerides, urea, or cholesterol, in the presence of water, to build up to a serum matrix.
Once basis sets are generated for serum, it is then possible to generate basis sets for red blood cells, muscle layers, skin layers, fat layers, even the whole body.
It is significant to note that the basis set approach herein thus characterizes each 2o component in a sample, as well as all other possible interferants, and subtracts each interferant from the spectra produced at the frequency of interest. In this way, all interferants are identified within the context of the sample and systematically from the spectra, leaving substantially only the signal produced by the analyte of interest.
The various basis sets may be combined mathematically to generate a set of transforms that may be stored in a look-up table for use during analysis. In this way, a fast, real time determination of analyte concentration may be made using relatively simple, low power computer hardware, e.g. a low power embedded controller.
3o Once it is determined which components are present in the sample, it is necessary to determine the best method of collecting spectra for these components. However, it is also necessary to define instrument specifications, such as signal to noise ratio, resolution, and wavelength reproducibility, before the spectra can be collected. These instrument considerations are discussed in detail below.
The procedure for generating basis sets is iterative. In some embodiments of the invention it is necessary to consider such factors as scatter correction, refractive index correction, depth of penetration into the tissue, total optical path length, and temperature. Once the basis sets are generated, they may be applied to a spectroscopic input signal. The signal thus processed by the basis sets is then preprocessed using standard chemometric techniques, such as smoothing and second derivative analysis.
to Another approach to processing after application of the basis sets is that of deconvolution. If deconvolution is used, then it is necessary to perform temperature correction after the data collection and scatter correction. This approach uses the basis sets to identify and isolate various components of the sample in an iterative fashion. 'Thereafter, multivariate analysis may be applied, which may include partial least squares analysis or principal components analysis. Such processing is a matter of choice and is well known in the art. For example, in a glucose concentration C for which there is a spectrum of interferograms n x m, where m is the number of interferograms and n is the number of interferogram points, a data reduction is performed, in which:
C=bo+blP~ +... b"P"; (1) where P; are PLS factor scores derived from the interferogram points and concentration values; and b; provides the regression coefficients.
Unique to the invention, various transforms such as deconvolution are performed with reference to the basis set. Pre-processing also relies on the basis set in the invention.
For example, in the case of a Fourier filter in which certain frequencies pass through 3o the filter, it is necessary to know what frequency the filter passed. In this way, it is possible to determine if the analyte is passed by the filter. The basis set is referenced to identify the analyte concentration at various frequencies, such that the Fourier filter only need be applied at those frequencies of interest, and not across a broad range of absorbance frequencies (as is practiced in the prior art). Thus, the basis set may loosely be thought of in this application as a filter for the filter.
Various molecular relations may be considered to be basis sets in themselves, such as carbon-hydrogen, oxygen-hydrogen, and nitrogen-hydrogen bonding. In such cases, there are more absorbance bands than can be accounted for by these fundamental components. This means that there are related effects for those portions of molecular structures to which these components are bonded. Thus, even though these molecules or pieces of molecules may be found in common among different constituents, it is 1o possible to assign them to a constituent and then discard them during deconvolution because of the signature across the spectra of a particular constituent.
Fig. 1 is a flow diagram showing the generation of a basis set according to the invention. The first step of the process involves identifying relevant interfering components of the sample at the same frequency as that of the analyte (100).
This step and subsequent steps may be performed using known spectroscopic and chemometric techniques, as is discussed in greater detail below. Once the interfering components are identified, the relevant interfering components are then all identified at other frequencies to quantify absorbance at these other frequencies (102).
The 2o interfering components, once quantified, are then removed at the frequency of the analyte (104). Each iteration of the foregoing steps may be described as a separate basis set. Thus, the invention produces a plurality of basis sets for an analyte.
Fig. 2 is a block schematic diagram of an instrument that incorporates one or more basis sets according to the invention. In operation, a device 10, collects spectra 20 using standard or modified (see below) spectroscopic devices. The spectra are provided to the input port/buffer 21 of a system that includes a processor 22.
The input port/buffer may include an analog-to-digital conversion function, such that spectral data collected by the spectroscopic device are converted to digital data. The processor operates upon such digital input data in accordance with various transforms stored in one or more look-up tables (LUTs). The LUTs contain transforms that incorporate the various basis sets. The transform process performed by the processor uses the basis sets to identify and remove substantially all interfering constituents from the spectral signal produced by the spectroscopic device. Once processing with regard to the basis sets is completed, the digital signal contains substantially only the analyte information. This information is then further processed in accordance with known spectroscopic calibration and chemometric techniques and provided to an output port/buffer 24. The output information may then be observed on a display 26, in any desired format, to provide an accurate indication of analyte concentration within the sample.
As also discussed both above and below, the basis sets generated in accordance with 1o the invention herein may be stored in a lookup table or they may be mixed in with the other transform information. In producing such look-up tables, the basis sets first exist as matrix raw data collected during the iterative process of generating the several basis sets. In view of the several basis sets generated in the preferred embodiment of the invention, there may be different matrices in the look-up tables, or there may be a single matrix that generates a transform which is representative of all of the basis sets and that is applied directly to the raw data. Thus, one embodiment of the invention takes each of the components and builds them into a complex matrix that comprises an algorithm for identifying and removing interferants. In this way, the invention provides a system that accurately represents how the components appear within the 2o spectra of interest when such components are all combined. It is this ability of the invention to identify each relevant component of a sample individually within the context of each other component that allows the ultimate determination of look-up table entries for an analyte of interest.
Although possible, the presently preferred embodiment of the invention does not provide spectra of glucose that has been corrected for all interferants at all concentration levels in a lookup table. Rather, there are a series of spectra of the analyte at certain different physiological concentrations of interest. For example, in the case of glucose, there are look-up basis set values for hypo- and hyperglycemia concentrations. Thus, the invention does not need to represent all of the information in all of the basis sets in the look-up tables. Rather, it is only necessary to represent information over the whole range of glucose that occurs in the body. The approach is taken for albumin protein, and other sample components. As discussed above, a single equation may be written for all of the spectral information in this matrix, or one or more look-up tables may be provided. In any event, this approach of storing only useful spectral information in the look-up tables reduces the memory and processing power requirements of the instrument.
As discussed more fully below, the basis sets are first generated and, thereafter, incorporated into an instrument for use during analysis. To determine those values that are to be put into the look-up tables it is necessary to go through any number of basis sets. As discussed above, it is necessary to identify the major interfering to components that affect the analyte in the sample and generate basis sets for each and every one of these components.
Fig. 3 is a block schematic diagram of an instrument that implements an algorithm which incorporates one or more basis sets according to the invention. Fig. 3 provides a detailed overview of the soflware/firmware component 30 of the device discussed in connection with Fig. 2. It should be appreciated that the invention herein is readily applied to any spectroscopic system. Thus, the system described in connection with Figs. 2 and 3 is provided only as an example of a presently preferred embodiment of the invention and not by way of limitation.
During processing within the instrument, the digitized spectral information is first applied to the basis sets 31. As discussed above (and in greater detail below), the basis sets are reduced to transforms that remove interfering constituents and/or components (chemical and/or physical) from the spectral information. The basis sets 2s may be applied before or in connection with a physical model 32 that corrects for such interfering physical factors as scattering, pathlength, and/or temperature.
After the spectral information is applied to the basis sets and (optionally) the physical model, the signal thus produced is deconvolved 33 to correct the signal to a reference.
The signal is next preprocessed 34 and digitally filtered 35. Preprocessing may employ such techniques as Kubelka-Munk transformation, mean centering, normalization, baseline correction, scatter correction, and interference correction, although it is presently preferred that the basis sets be used to resolve such issues as scatter correction, baseline correction, and interference correction.
Correction techniques that may be applied, for example to scattering, can include multiplicative scatter correction, standard normal variate correction, and extended multiplicative signal correction. The digital filtering function may be accomplished by such techniques as Gaussian filtering, low and high bandpass filters, and Lorentzian filtering.
Spectral wavelengths for the analyte are selected 36 and a multivariate analysis, such as higher order partial least squares (PLS) is performed 37. Such analysis techniques to may include principal component regression, partial least squares, rotated principal components, or correlation principal components analysis.
The preferred embodiment of the invention provides a plurality of basis sets that are used to quantify an analyte in a liquid sample. For purpose of illustration and example, the invention is now described in connection with glucose quantification in noninvasive spectra.
Data Collection. The first step of the process involves identifying major interfering chemical analytes and structures in the body. These factors include, inter alia, water 2o percentage present in the sample, temperature/hydrogen bonding effects on water, albumin protein, globulin protein, triglycerides, cholesterol, urea, glucose, lactate, ethanol, also Na+, K+, Cl-, and other electrolytes, glycosylated hemoglobin, skin, keratin, fibrinogen, and red blood cells. One advantage of the invention is the basis set may be generated in such way that it includes spectra for all interfering components.
Noninterfering components include, for example, components of lower molar absorptivity concentration, such as low dosage drugs and medications.
In the presently preferred embodiment of the invention, data collection instrumentation should take into account the following:
~ Signal is defined for each analyte by first determining the delta absorption from top of absorbance band to base, and then by defining the slope of change in absorbance versus concentration for samples spanning the physiological concentration at all frequencies.
~ Noise is defined as root mean square (rms) noise of analyte absorbance in the band of interest.
~ Sisal to Noise is defined as (slope X concentration) / noise. 'This value 1o must be greater than one for a minimum specified concentration to be analyzed.
~ Resolution in the presently preferred embodiment of the invention requires a minimum of seven points per peak.
Another factor to be considered is wavelength reproducibility. In the invention, a modified NIRS 5000 spectrometer is used to achieve the above criteria.
The data collection parameters for the basis set include the following:
Pathlengths due to absorbance of water, which is the primary interferant. It is necessary to select different pathlengths for each spectral window for an optimal basis set.
In the presently preferred embodiment of the invention, these pathlengths are:
~ 0 to 2 mm for combination band region;
~ 5 to 10 mm for first overtone region; and ~ 10 mm or greater for second overtone region.
While not necessary, it is possible to generate a basis set over the entire frequency range in a single data collection to compare information in different regions.
In the preferred embodiment of the invention, this dictates a 1 mm pathlength.
Optimal signal to noise levels are obtained separately. It is also necessary to provide continuous spectra to identify and model parameters, such as change in refractive index as a function of frequency.
For applications that involve the use of diffuse reflection, pathlength considerations are not taken into account because light penetration is proportional to the inverse of water absorbance, as defined by the system, based upon molecules interacting with specific concentrations at specific refractive indices.
In some embodiments of the invention it is desirable to use optical filters.
Because water behaves as natural short pass filter, it is advantageous to use long pass cutoff filters in conjunction with water bands to form a bandpass filter (although a system that provides sufficient resolution, i.e. a sufficient number of analog-to-digital (A/D) bit, may not require a filter). In the presently preferred embodiment of the invention, any filter in the midst of the H20 absorbance band may be used, e.g. the following filters may be used:
~ 1950 nm long pass filter for the combination band;
~ 1450 nm long pass filter for the first overtone; and ~ 1100 nm long pass filter for the second overtone.
The number of averaged scans for each spectra must be determined, where noise decreases with an increasing number of scans. In the presently preferred embodiment of the invention, noise vs. number of averaged scans is set to 64 averaged scans.
Replicate spectra. Experiments were conducted in which four replicates were collected due to the temperature coefficient of the spectrometer. The following results were obtained (which were due to the spectrometer used to make the measurements --these results are not indicative of a general phenomenon):
~ First replicate - outlier due to temperature;
~ Second replicate - small outlier characteristics; and ~ Fourth replicate - acceptable for further analysis.
to For purpose of experiments conducted in connection with the invention, the following additional parameters were defined:
~ Ionic strength is 0.1 M to match that of the body;
~ pH 7.35 phosphate buffer to approximate that of the body;
~ Temperature maintained at 38.0 ~ 0.2°C to match that of the body;
~ Components: ACS reagent grade chemicals used as standards.
Data Analysis. Data analysis must take into account temperature variations. In the presently preferred embodiment of the invention, temperature variations of 0.
I °C are observed to severely obscure the analyte absorbance bands (even concentrated albumin). Laboratory and instrument temperatures are impossible to control to 0.01 °C
for daily use. This effect is amplified in regions of high water absorbance and large changes of water absorbance due to temperature.
Data analysis must take into account pathlenQth. This consideration is similar to differential measurements taken in dual beam spectrometers, where one beam is 3o focused through the sample and a second beam is focused through a pathlength corresponding to the pathlength interference in the sample.
It is desirable to control pathlength to 0.0001 mm. For a 1 mm cell with buffer present there is a 1 mm pathlength of water. When an analyte is present, the pathlength of water is reduced due to displacement. The displacement is linearly proportional to the concentration of the analyte present. While various components of the sample may be rotated out if their concentration is unknown, such processing is unnecessary upon using the invention because such concentrations are known.
Temperature and pathlen~th correction algorithm. The following discussion provides an exemplary temperature and pathlength correction algorithm in accordance with the l0 invention.
1. Response function: residual (sample - buffer) about zero for regions where the analyte does not absorb.
2. Residual as function of spectral range is inversely weighted by the spectral noise of the sample.
3. Thousands of buffers collected at roughly 38°C are compared with the sample to match temperature. By using thousands of buffers, a good temperature match can be found.
4. For each buffer tested, incremental pathlengths of water are tested to match pathlength of buffer in the sample. For example, to get a pathlength of .99 mm, the buffer being tested as a possible background is multiplied by 0.9900. For example, albumin protein from .95 mm to 1 mm at 0.0005 mm steps is tested with each buffer.
The following is a Matlab~ temperature/pathlength correction program for selected parameters that must be optimized for each analyte, such as pathlength, and regions for the response function:
temppath.M
PROBLEM: Basis Set spectra require background subtraction of temperature and pathlength.
This program corrects the temperature by searching for a buffer collected at the same temperature as the sample and match the amount of buffer present in the sample.
clear to enter wavelength region wavelength = 1100:2:2498;
load sample spectra load a1b12_1 txt sample = albi2-1;
is [o p] = size (sample);
load buffer spectra usually use all buffers collected to date 20 load albbufftxt buff = albbuff;
[m n] = size (buff);
Code minimizes residual over user set regions 25 % These regions can not have absorbance from the analyte they are fine tuned iteratively.
in this case - three regions are used in the response function.
b-1 stmt = find(wavelength>= 1640 & wavelength < 1655);
3o b_2nd~t = find(wavelength>= 2077 & wavelength < 2085);
b-3rd_pt = find(wavelength>= 1640 & wavelength < 1655);
s-1 stmt = b_1 st-pt;
s 2nd~t = b 2nd~t;
s 3rd~t = b 3rd~t;
initialize to large residual pathlength = 0;
for aaa = 1 : p 40 best min(aaa) = 1000000;
end pathlength optimization determine pathlength matching water in sample 45 % (water absorbance * pathlength) for j = 0.95 : 0.0005 : 0.997 %.98 j=manual-pathlength pathlength = pathlength + 1;
temperature optimization for each pathlength, test every buffer for temperature match for temp = l :n avg b-1 stmt = mean(buff(b-1 st~t,temp))*j;
avg b 2nd~t = mean(buff(b 2nd~t,temp))*j;
avg b 3rd-pt = mean(buff(b 3rd~t,temp))*j, to repeat for every sample and replicate for sample num = 1 : p avg s-1 st-pt = mean(sample(s lst~t,sample_num));
avg s 2nd-pt = mean(sample(s 2nd~t,sample_num)) ;
avg s 3rd~t = mean(sample(s 3rd~t,sample num));
15 diff_1 stmt = abs( avg s-1 st-pt - avg b-1 stmt );
diff_2nd-pt = abs(avg s 2nd~t - avg b 2nd-pt );
diff 3rd~t = abs( avg s 3rd-pt - avg b 3rd~t );
store results of each loop results(sample num) = diff 1 stmt + diff 2nd-pt + diff 3rd~t;
usually add in weighting function as inverse of noise for each region here if response function for given sample is best - record parameters if results(sample_num) < best_min(sample num) best_min(sample num) = results(sample_num);
best-pathlength(sample_num) = j;
best temp(sample num) = temp;
end % end if end % end sample end % end temperature end % end pathlength % dump best parameters to screen for interpretation best_min best~athlength best temp plot temperature and pathlength corrected spectra hold off clg hold on = ysoo 2soo -o.ol o.o~~;
axis(v);
for sample num = 1:p best sample corr(:,sample num) = sample(:,sample num) - buff( :,best temp(sample num)) * best~athlength(sample_num);
plot (wavelength,best sample corr(:,sample_num));
intensity(sample num) = best sample_corr(481,sample_num);
end The resulting spectra are clean and baseline resolved. The spectra are selected to cover physiological concentrations for each analyte.
to The following example illustrates the generation of a first basis set "Basis Set L"
Example - Basis Set I
near-IR absorbance features over the 1500 to 2500 nm spectral region are provided for water, albumin protein, globulin protein, triacetin, cholesterol, urea, and glucose with a 1 mm pathlength. In addition, sample temperature effects are included along with instrument noise levels.
Experimental: Spectra of the major constituents of serum were collected over their 2o respective physiological ranges. Sample preparation consisted of dissolving dried, reagent grade solid samples in a 0.1 M phosphate buffer adjusted to pH 7.35.
All spectra were collected on a NIRS 5000 in transmission mode, with a 1 mm path length infrasil quartz cell, with a 120 second equilibration period, at 38.0°C, with 64 averaged scans, done in quadruplicate. A single instrument was used for all data acquisition.
Results and Discussion: Spectra are analyzed in order of decreasing absorbance changes in the two spectral windows from 2050 to 2350 nm and 1550 to 1850 nm.
The first replicate is discarded in all cases due to a consistent variation in temperature 3o caused by the instability of the NIRS 5000 spectrometer and photons heating the sample (data not included). The sample is in equilibrium by the second sample replicate.
Water Spectra: The near-IR is dominated by three large water absorbance bands centered at 2500, 1950, and 1450 nm as presented in Fig. 4. The high absorbance limits analysis done in aqueous solution in the near-IR to three spectral regions. The region from 2350 to 2050 nm is referred to herein as the combination band region; the region from 1850 to 1550 nm is referred to herein as the first overtone spectral region;
and the region from 1400 to 1100 nm is referred to as the second overtone spectral region.
The NIRS spectrometer sets the gain and hence the dynamic range of the detector based upon the spectral region with the most light intensity reaching the detector.
This is the second overtone spectral region for aqueous samples. However, the combination band region has the largest absorbance, followed by the first overtone region, and then the second overtone region. Due to the low absorbance of water in the 1300 nm region versus the 2200 nm region, a relatively small dynamic range is left for the 2200 nm region where glucose bands are the largest. Therefore, a 1500 nm long pass filter was employed which forces the NIRS system to set the gain based upon the first overtone spectral region. Hence, in the initial basis set no spectral information is provided for the second overtone region, optimum signal to noise levels are provided for the first overtone spectral region, and slightly degraded signal to noise levels are obtained for the combination bands. Among many modifications made to the NIRS system is an order sorter which allows a different gain setting for each of the three spectral regions during a single scan for the next basis data set.
Temperature Effects on Water Spectra: All three water absorbance bands in the near-IR shift to higher frequency with increasing temperature. Buffer spectra collected from 38.2 to 43.0°C are presented in Fig. 5A. The instrument should be modified to collect lower temperature spectra. A slight broadening of the lines can be observed on each of the water absorbance band shoulders. Subtracting a spectrum of water collected at 38.2°C from spectra of water collected at higher temperatures reveals the 3o magnitude and direction of the shift. Negative absorbance bands that increase with temperature are observed at 2000 and 1480 nm. As the water bands shift to higher temperature, there is less water absorbance in these regions, so that subtracting out a water absorbance band from a lower temperature results in too much background being subtracted. Positive absorbance bands that correlate with increasing temperature are observed at 2300, 1890, and 1400 nm. With increasing temperature, the water absorbance increasingly moves into these spectral regions.
Subtracting out the 38.2°C water spectrum does not subtract out enough in these regions.
The large water absorbance, coupled with the temperature shift, greatly hinders near-IR analysis. Fig. 5B reveals that in the subtraction, no useful information is obtained where the raw absorbance is greater than 3.0 ~ 0.1, indicating the limit of the dynamic range of the NIRS system. Therefore, the regions above 2460 nm and from 2010 to l0 1890 nm result in no analytically useful information and may be discarded for data collected with a 1 mm pathlength. Information in these spectral regions may be obtained by adjusting the pathlength. Due to the water absorption, the width of the regions that need to be discarded increases as the pathlength analyzed increases. In addition, the temperature effects are seen to span the entire combination band region and first overtone spectral region. As will be shown, these changes in baseline are roughly equal in magnitude to the highly absorbing protein and much greater in magnitude than all other spectral analytes examined.
Albumin Protein: After water and temperature effects, serum spectra are primarily composed of absorption from albumin protein which has a physiological range of 2.6 to 7.9 g/dL. Albumin protein absorbance bands are difficult to see in the presence of water, as shown in Fig. 6A. Subtracting out a buffer spectrum results in protein absorbance peaks at 2285, 2170, 1730, and 1690 nm, as shown in Fig. 6B. Large negative absorption bands also appear in the resulting spectra where water absorbs.
These bands are not primarily due to variation in temperature as a derivative of the water band would appear as seen in Fig. 5B. The negative bands are due to displacement of water by albumin and scattering.
A program, such as the Matlab~ program described above, is used to determine the 3o best buffer in terms of temperature and best calculated pathlength to be used as a background spectrum for subtraction. In Fig. 6B, the buffer and albumin in buffer spectra both had the same 1 mm fixed pathlength. Because albumin is present in the 1 to 12 g/dL range in this example and water is 100 g/dL, the albumin occupies a significant volume of the cell and less water is present per unit volume. A
program was written that multiplies the water spectrum by a percentage that can be sequentially varied over a wide range. The optimum calculated pathlength for each albumin in buffer spectrum was determined by minimizing the sum of the absolute value of the residuals in locations where albumin does not absorb and temperature effects are at a minimum (2085 to 2077 and 1655 to 1640 nm). The residual in the overtone region was weighted twice as much to compensate for the higher noise in the combination band region. To further minimize temperature effects, all buffer spectra collected were run through this optimization to find the best buffer in terms of temperature matching with the sample. Each albumin in buffer spectrum was run through this algorithm independently.
The results of subtracting the best buffer at the adjusted pathlength for each albumin spectrum are presented in Fig. 7. Additional albumin absorbance bands are now visible at 2060 nm and 2335 nrn. Expansion of the graph about the 2060 nm absorbance bands reveals increasing absorbance for each increase in albumin concentration. The albumin band centered at 2170 nm is more symmetrical than the one seen in Fig. 6B. The two peaks in the first overtone spectral region have a better baseline correction and now increase in absorption linearly with increasing concentration. However, negative absorbance bands are still evident where water absorbs at 2020 and 1870 nm. The region between 2000 and 1900 nm is an artifact of the mathematical correction over a region where the absorbance is greater than 3 and the system does not respond. In addition, there is a large difference between absorbance and the 1 g/dL albumin spectrum. This difference should be equal to the difference in absorbance from 1 to 2 g/dL. This offset can be reduced if the combination band region and the overtone region are treated individually. It should be pointed out that no baseline correction, smoothing, or scatter correction has been employed at this point.
3o Globulin Protein: Physiological concentrations of globulin (0.7 to 8.1 g/dL) are less absorbing in the near-IR than albumin. Straight subtraction of the phosphate buffer allows the same peaks to be observed that are seen in albumin protein, as shown in Fig. 8. The temperature and pathlength correction algorithm discussed above was run with exactly the same parameters as for albumin and the same additional extra peaks were found, as shown on Fig. 9. Overlaying the albumin and globulin spectra reveals that the globulin absorbance band centered at 2170 nm is slightly broader than that of albumin protein.
The calculated pathlengths required for the background subtraction from each of the spectra are presented in Fig. 10. For albumin, the correction is linear with increasing concentration, but has a y-axis intercept of 0.996 mm. 'This is consistent with the poor baseline observed in Fig. 7. The corrections for globulin are also linear, but greater to corrections are required per mg/dL analyte. This is consistent with the scattering tendencies of globulin. The y-axis intercept is 1.00, consistent with the excellent background subtraction.
Triglycerides: Triacetin is used to simulate triglycerides. The physiological range of triacetin is 50 to 450 mg/dL. The temperature and pathlength correction algorithm is again employed, but different regions are used in determining the minimum residual (2420 to 2440, 2080 to 2090, and 1575 to 1635, weighted 1:5:20). Six triacetin absorbance bands result centered at 2320, 2250, 2130, 1760, 1715, and 1675 nm, as shown on Fig. 11. The resulting pathlengths required for correction are linear with 2o concentration, but much smaller deviations from 1 mm result due to the lower concentration of triacetin versus protein in serum, as shown on Fig. 10. The signal levels of the smaller triacetin absorbance bands approach the noise level of the spectrometer.
Urea: Twelve urea in buffer spectra were collected. Due to the small physiological concentration of urea (6 to 123 mg/dL), the algorithm used to optimize the background by changing the effective pathlength of the buffer subtracted fails because no significant Figs. of water are displaced. No temperature matching algorithm is employed, but buffer spectra collected with each sample are used. A straight 3o background subtraction followed by a two point baseline correction (2094 to 2106 and 2320 to 2332 nm) was performed and is presented in Fig. 12. A single absorbance band is present centered at 2190 nm. No overtone peak is present. This is consistent with this absorption being related to N-H, whereas all of the other analyses have O-H
fundamental vibrations. Only four spectra are presented due to Iarge baseline drifts that obscure the linearity of the additional spectra. Higher concentration samples can be run to obtain a higher S/N and cleaner spectra, although the same resulting basis set is obtained.
Glucose: A complete glucose in buffer study was performed over the combination and first overtone spectral region of which a subset is presented here.
Glucose was examined from 30 to 600 mg/dL (also from 0 to 5000 mg/dL) to cover the physiological as well as hypoglycemic and hyperglycemic levels of glucose. A
to straight subtraction of buffer from glucose in buffer shows absorbance bands centered at 2326, 2272, 1800, 2150, 1730, and 1590 nm, as shown in Fig. 13.
Conclusions: Consistent with theory, for all analytes, the combination band spectral region yields larger absorbance than the first overtone spectral region.
However, longer pathlengths quickly degrade the signal to noise level in the combination band region due to the Iarge water absorbance, whereas the spectral quality in the first overtone spectral region should increase with small millimeter increases in pathlength.
The absorbance bands in the region of glucose absorbance in decreasing order of absorbance are water, temperature effects, albumin protein, globulin protein, 2o cholesterol, triglycerides, urea, and glucose. While every analyte analyzed absorbs more than glucose and over the same general spectral region, every analyte has a distinct absorbance signature. In principle, the serum spectra or the noninvasive spectra, can be deconvoluted.
***
The invention contemplates the generation of additional basis sets, such that substantially all interfering components are identified and factored into the spectroscopic analysis.
The following example illustrates the generation of a second basis set "Basis Set IL"
Example - Basis Set II
A study was rerun on dried, crushed, and pressed solid samples to give absorbance spectra with no water. A second basis set was collected based upon spectra of solid or neat components of human serum. The resulting absorbance spectra show the combination, first, and second overtone absorbance bands. In addition, for a given component the relative absorbance between regions may be compared. Combined, another method of wavelength selection is made available.
to Experimental: Pure component spectra of the liquid form of water (pH 7.35, 0.1 M
phosphate buffer 38.0 t 0.2°C), triacetin, and lactic acid were collected. Albumin, globulin, cholesterol, urea, and glucose exist as a solid in their pure state.
For these analytes, each was individually ground with a mortal and pestle to a fine powder in the absence of potassium bromide. The powder was then compressed into a transparent pellet in a specially designed press that fits into the NIRS 5000 transmission module.
Four replicates of each component were then obtained in the transmission mode.
'The pathlength of each analyte was not controlled.
Results and Discussion: The raw absorbance spectra for water, albumin, globulin, 2o cholesterol, triacetin, urea, glucose, and lactic acid are presented in Fig. 14. Because the pathlength of each pellet was not controlled, the relative absorbance between components can not be compared. The relative absorbance between frequencies for a given analyte can be compared. The large baseline offsets are due to the thickness of the sample and resulting total light throughput. This plot is included to show the total absorbance of each analyte relative to the dynamic range of the NIRS 5000.
For each component in Fig. 14, the minimum absorbance was subtracted out and the resulting spectra was normalized to 1 absorbance unit, as shown in Fig. 15.
The resulting full scale plots make it easier to compare absorbance as a function of 3o frequency and differences between components. For all three spectral regions, i.e.
combination (2050 to 2350 nm), first (1550 to 1850 nm), and second overtone (1100 to 1400 nm), the absorbance bands are observed to be distinct. In principle, each component can be deconvoluted. It should be noted that when interacting with water, these absorbance bands may shift and broaden. Comparing with the aqueous absorbance from Basis Set I (above) reveals the absorbance bands of the neat or solid water ( 140), albumin ( 141 ), globulin ( 142), and triacetin ( 143 ) to be in the same location with the same widths. Both urea (144) and glucose (145) reveal additional resolution of peaks that have broadened and merged in the presence of water.
Several key spectral signatures emerge from this Example. First, the combination band region contains absorbance for each of the individual analytes. These absorbances are in every case more intense than those in the first and second overtone 1o spectral regions. Cholesterol (146) absorbance drops off rapidly in this region as does triacetin. Neither interferes significantly with the glucose absorbance band centered at 2150 nm. The only interference is from water, albumin, and globulin which are shown in the Example - Linearity Study (below) to be removable by simple subtraction.
In the first overtone spectral region every component has an absorbance band except urea with its N-H bonds. Here the intensities of the absorbance bands range from 15% to 50% that of the corresponding combination band absorbance. It should be recognized that these values are for a fixed pathlength and can be adjusted based upon 2o total pathlength.
The second overtone spectral region has absorbance bands for every component examined, but the relative absorbances are the smallest, as shown on Fig. 16.
The glucose band ( 145) seen here is very difficult to see in the presence of water ( 140).
Conclusions: Each of the three regions contains information about every analyte with the exception of urea in the first overtone spectral region. The absorbance bands are highly overlapped and are generally less intense at higher frequencies. The absorbance bands are all distinct.
***
The following example illustrates the generation of a third basis set "Basis Set IIL"
Example - Basis Set III
The first basis set was repeated with no edge filter present to allow comparison of all spectral ranges. The first Example above used a 1500 nm long pass filter to force the NIRS spectrometer to gain range on the 1700 nm spectral region. This Example could be repeated with increased optical pathlengths to yield higher signal to noise levels in the first and second overtone spectral regions.
***
l0 The following example illustrates the generation of a fourth basis set "Basis Set IV."
Example - Basis Set IV
It is necessary to measure interactions of molecules in solution. In this Example, a serum data set is collected.
Data Sets: 'The first data set consists of spectra of glucose dissolved in a 0.1 M
phosphate buffer adjusted to pH 7.35. Reagent grade glucose was weighed and 2o diluted to a known volume with the 0.1 M phosphate buffer. Spectra were collected in the transmission mode with a 1 mm quartz cell using the NIRS 5000 spectrometer over the range of 1100 to 2500 nm with readings taken every 2 nm. A 1500 nm long pass filter was placed before the sample to force the NIRS spectrometer to set the gain on the peak signal at 1600 nm. Before and after every sample, 7 spectra of the 0.1 M
phosphate buffer were collected. A total of 64 glucose~ag~ samples were collected with 7 sequential replicates of each sample. The glucose samples covered a dynamic range of approximately 20 to 600 mg/dL. All samples were maintained at 38.0 ~
0.2°C.
The second data set consists of serum samples prepared by Western States Plasma.
3o Each serum sample was analyzed using a standard SMAC (serial mufti-channel automated chemistry) analysis yielding concentrations for calcium, ionized calcium (calculated), phosphorus, glucose, uric acid, urea nitrogen (BLTN), creatinine, creatinineBLJN ratio, total protein, albumin, globulin, A/G ratio, total bilirubin, ALT, ALP, LD (LDH), AST, GGT, sodium, potassium, chloride, carbon dioxide, triglycerides, and cholesterol. To extend the dynamic range and level the concentration distribution, reagent grade urea and glucose were quantitatively added to the serum samples. The NIRS 5000 spectrometer was used in the fashion described above with the same wavelength region, pathlength, temperature control, and long pass filter. A 0.1 M phosphate buffer adjusted to pH 7.35 was run before and after each serum sample. A total of 196 serum samples were collected with 4 sequential replicates of each sample. The glucose analyte covered a dynamic range of approximately 20 to 600 mg/dL.
to Experimental: Glucose is determined in each data set using PLS regression analysis.
The data sets are broken up into calibration and prediction keeping all replicate spectra together. A data point was originally collected every 2 nm from 1100 to 2500 nm. Additional data sets are formed from this data set by keeping every other point, every 3rd point, every 4th point, ..., to every 32°d point. PLS
calibration models and predictions are then determined using 1 to 10 PLS factors.
Results and Discussion: For each resolution, the resulting standard error of calibration (SEC) and standard error of prediction (SEP) is determined for 1 to 10 PLS
2o factors, as shown on Fig. 17. Here, selection of the optimum number of factors needs to be achieved. As different ranges need to be compared, differences in the number of PLS factors employed can lead to erroneous conclusions. Statistical approaches to determining the optimum number of factors have failed. Because the SEP does not increase as the system is over-modeled, and further because the SEC and SEP
yield similar results with 10 factors, it was decided for the purposes of this Example only to compare standard errors from range to range using the results obtained with ten PLS
factors.
Ten spectral ranges are analyzed in both the glucose in water and glucose in serum 3o data sets. These are summarized in Table 1 below. Ranges 1 to 3 and 5 to 7 correspond to the full width at zero height of the six glucose absorbance bands isolated in the near-IR. Ranges 4 and 8 splice together regions 1 to 3 and 5 to 7, respectively. Ranges 9 and 10 expand regions 4 and 8 into regions of increasing water absorbance, increasing noise, and no additional glucose information.
Table 1: Spectral Ranges Employed Ran a NumberS ectral Ran a nm 1 2078 to 2243 2 2243 to 2272 3 2297 to 2366 4 2078 to 2366 5 1587 to 1674 6 1674 to 1709 7 1709 to 1754 8 1587 to 1754 9 2000 to 2500 1520 to 1805 Clearly, the wider spectral region that incorporates more glucose information (and water and temperature) results in a lower standard error at any resolution than any of the three individual glucose absorbance bands.
The nominal resolution of the NIRS spectrometer is 10 nm for the standard 0.040"
exit slit used in this Example. Still, the standard error is observed to increase slightly as the resolution degrades from 2 to 10 nm. This is due to the manner in which the data sets were created from the original 2 nm resolution data set. For instance, in the ~5 6 nm resolution data set generated, every third spectral point is kept.
This means that two-thirds of the data are discarded. The discarded data has glucose, water, and temperature signal. In addition, by keeping these extra points, the effective noise is decreased by signal averaging. In as much as the true resolution of the NIRS
5000 is 10 nm, 100% of the slope observed on the SE vs. resolution graph is due to this 2o systematic error. In addition, the same slope is observed from 10 to 32 nm resolution.
The original data set with points every two nanometers was again broken down into data sets with resolution ranging from 2 to 32 nm at 2 nm intervals. This time, the data was averaged instead of just discarding extra points. For example, at 6 nm resolution points at 1100, 1102, and 1104 nm were averaged to a single point.
The next point averaged the data points at 1106, 1108, and 1110 nm. The PLS
analysis was then repeated and the standard errors with the tenth factor determined, as shown in Fig. 18. The increase in standard error observed with degrading resolution is observed to range from 5 to 10 mg/dL standard error as opposed to 5 to 25 mg/dL
standard error from 2 to 32 nm resolution. Clearly, the failure to average the data points results in an increase of the slope of standard error versus resolution. While the standard error roughly doubles from 2 to 32 nm resolution, the data indicates that for a glucose in water solution, the acceptable resolution may be 32 nm or more.
'This 1o makes chemical sense in as much as the narrowest absorbance band in this Example is 54 nm wide. In must be pointed out that there are no spectral interferences in this Example. Therefore, the actual acceptable resolution can only degrade from this resolution.
2°C for the first overtone region, for data sets generated at 2 to 32 nm resolution using averaged data, the increase in standard error with degrading resolution is greatly reduced, as shown on Fig. 19. In addition, for this spectral region, less than ten points are retained at resolutions greater than 16 nm. The PLS algorithm used only operates on as many factors as there are data points. If queried for standard errors with 2o additional factors, the standard error for the number of factors equal to the number of points available is generated. Because the standard errors continue to decrease with an increasing number of factors in this Example (see Fig. 17), the comparison of standard errors for various resolutions using ten PLS factors is not valid. A
direct comparison of standard errors at degrading resolution for the 1587 to 1754 nm spectral region with six PLS factors is presented in Fig. 20. The increased standard error observed with degrading resolution is now not observed with resolutions under 15 nm. 'This is a true comparison of standard errors for this spectral region.
The results in Fig. 18 for the 2078 to 2366 nm spectral region are still valid due to its large range which contains ten or more points up to 30 nm resolution.
Glucose in Serum: The SEC and SEP plots versus resolution for glucose in the serum study for the ten different spectral regions are provided in Figs. 21 to 30.
The results are generally the same as for glucose in water.
The combination band region is analyzed first. Range 1 with the largest glucose absorbance band yields the lowest standard errors for a region isolating a single glucose absorbance band, as shown on Fig. 21. Ranges 2 and 3 yield larger standard errors and have smaller glucose absorbance bands with a decreased signal to noise level, as shown on Figs. 22 and 23. Analysis of ranges 2 and 3 at degraded resolutions is limited by the number of data points present in each range. Range 4 which couples the first three regions demonstrates the lowest standard errors, as shown on Fig. 24.
Again, the averaging of points reduces the increase in standard error with degrading 1o resolution. The increase in standard error from 35 to 50 mg/dL observed as resolution degrades from 2 to 30 nm is entirely due to the loss of information in extracting rather than averaging data points. While the standard errors are higher than in the glucose in water Example, this Example demonstrates that even in the presence of all of the spectral interferences, except skin and blood cells, the resolution is essentially not an effect until after a resolution of 30 nm. This is the same result as for glucose in water.
The number of PLS factors incorporated is not an issue due to the fact that 10 points are present even at 30 nm resolution. Range 9 incorporates all of range 4 and extends past where glucose absorbs at both higher and lower frequencies, as shown in Fig. 29.
No resolution effect on standard error is observed from 2 to 32 nm.
The effects of resolution in the first overtone spectral region are more difficult to interpret due to decreased signal to noise and the narrower spectral ranges chosen.
Range 5 has the largest glucose absorbance band in the overtone spectral window and results in the lowest standard errors. Ranges 6 and 7 were shown to have very poor signal to noise levels for glucose in water (not presented). The standard errors are essentially mean centered prediction values, as shown on Figs. 26 and 27. The effect is worsened at degrading resolution due to the number of points in each spectral range.
Range 8 reveals real glucose predictions, as shown on Fig. 28. This range was reanalyzed with the averaged rather than the selected data, see Fig. 31. Using ten PLS
3o factors, the increasing standard error with degrading resolution observed is virtually identical to the nonaveraged data due to the number of points present in the data. This is shown by comparing the standard errors with only six PLS factors (6 points present at 30 nm resolution). No resolution effect is observed until a resolution of 20 nm.
Range 10 which expands to higher and lower frequencies from range 8 has 10 data points present at 30 nm resolution and shows no resolution effect until 20 nm, as shown in Fig. 30.
Conclusions: The glucose in water data set has sufficient signal to noise to determine glucose with the specifications required. The rise in standard error for the narrow glucose absorbance bands with degrading resolution is not real. It is partially the result of selecting the points rather than averaging the points to generate new data sets.
In addition, the new data sets did not contain enough data points to compare analysis to of 2 nm resolution data and 32 nm resolution data with ten PLS factors.
Resolution effects may be addressed by using fewer PLS factors in this comparison or by using larger spectral ranges. For both methods, the resolution effects are minimal to 30 nm in the combination band region and 15 nm in the first overtone spectral window.
Because it is preferred to get the highest signal to noise ratio possible from the instrument, it is acceptable to have 30 nm resolution. That is, by having less (but, nonetheless, acceptable) resolution, e.g. by having 30 nm resolution instead of 10 nm resolution, the instrument captures more signal relative to noise. Thus, even though the resolution is coarser, more information is contained in signal generated by the 2o instrument. As a result, the resolution selected in the preferred embodiment of the invention provides a more accurate picture of the spectra, even though the instrument has coarser resolution. This is because there is a higher signal to noise ratio at the resolution required. In contrast, if extra resolution were available in the instrument, but there was a lower signal to noise ratio, less information would be available for processing by the basis sets.
In the Example, the glucose in serum data sets resulted in roughly three times the standard error as in the glucose data set. Again, analysis is limited to either large spectral windows or to comparisons with fewer PLS factors for narrower ranges.
In 3o the combination band spectral region, the increase in standard error observed with degrading resolution is minimal to 30 nm resolution. In the first overtone spectral window, the slope to standard error versus resolution is minimal to 20 nm resolution.
These results are virtually identical to those generated in the glucose in water study.
The effects of the proteins, triglycerides, cholesterol, urea, salts, and minor organic constituents are observed not to effect the required resolution.
***
Example - Basis Set V
It is necessary to measure effect of scattering of whole blood cells. This basis set is generated as follows:
~ Collect blood data set in transmission and as diffuse reflectance.
~ Repeat component extraction.
~ Couple in scatter correction ~ Deconvolve (see deconvolution discussion below).
2o Example - Basis Set VI
It is necessary to measure the effect of skin. Animal studies are performed and all prior analysis techniques are repeated. Noninvasive studies can be viewed as extensions of the basis set.
***
Uses of Basis Sets.
Chemical and physical knowledge of a system are required for such factors as:
~ Intelligent wavelength selection, e.g. knowledge of the location and degree of interferences of each analyte.
~ Interpretation of noise levels as a function of region.
~ Interpretation of signal levels for each analytes as a function of wavelength.
~ Selection of optimal signal to noise regions for each analyte.
Resolution specifications for an instrument implementation of the invention are set forth above. The number of analog-to-digital (A/D) bits required to provide appropriate instrument resolution can be calculated from noninvasive spectra and glucose intensities (absorbance). For this determination, it is necessary to know the maximum intensity of the whole system and the intensity of glucose at the required standard error. If the maximum intensity of the sample is 10 to the negative absorbants unit, it is only necessary to calculate the intensity of the body scan, including all absorbants. To determine the intensity of glucose, the required standard error is 9 mg/dL. The intensity of glucose and water, and the intensity of water is used (as described above) to calculate the intensity of the glucose and water minus the intensity of water. This results in a value for the intensity of glucose. Once the intensity of glucose is determined, it is then necessary to determine the change in intensity of glucose, e.g. by drawing in a base line to the peak, and plotting the change in intensity of glucose versus glucose concentration. This provides a best fit of the data that can be fitted to a line to calculate the change in intensity at 9 mg/dL. Once this value is obtained, the ratio of this value to maximum intensity of the glucose is readily calculated. This ratio defines the number of bits that are required in the system for analog-to-digital conversion. For example, if the ratio is 50,000, then a 16 bit A/D
is required because sufficient quantization must be provided to avoid aliasing problems. Thus, the basis set is useful in defining instrument parameters.
Interpretation of multivariate results. Multivariate results are difficult to validate.
Standard errors must correlate with basis set information. If noisy regions are added, 3o the signal to noise ratio decreases. It is therefore necessary to correlate standard errors with the signal to noise ratio.
With regard to the removal of second, third, ... order light in a grating based spectrometer, a long pass filter is required. The basis set dictates the specifications of the filter.
With regard to the removal of scatter, such determinations are based upon refractive index change. In the preferred embodiment of the invention, the basis sets remove scatter and temperature effects. This step is repeated for additional analytes, and the reduced spectra are further processed using multivariate approaches.
Deconvolution of noninvasive spectra. The partial deconvolution reduces the rank of first temperature and water, then proteins, then organic constituents. The resulting spectra can then be fed into the multivariate approaches. However, the reduced dynamic range of signals forces PLS to lock in on smaller analytes, such as urea and glucose, instead of water and temperature.
There are a limited number of interferences for glucose in the near-IR. The major interferences have convenient breaks in concentration. The largest concentrations /
effects are temperature and water. Processing should remove the Refractive index, which is on the order of 100 g/dL.
Large concentration gaps exists between water and the proteins. Iterative deconvolution can be used to take advantage of this fact.
Albumin and globulin proteins are on the order of 1 to 7 g/dL. These interferants are easily identified and removed by spectral subtraction or rotation.
Example - Linearity Introduction: The basis set is used to determine the location and intensity of each of the major species interfering with glucose. It also demonstrates that for a given component, the absorbance increases linearly with increasing concentration. In this Example, it is shown that the absorbance of multiple components is the sum of the individual components, as assumed by Beer's law. This is critical to the herein described approach of using spectral subtraction of chemical information to enhance the signal to noise level of glucose.
Experimental: Spectra were collected in quadruplicate with a NIRS 5000 spectrometer configured in the transmission mode with a 1 mm pathlength quartz sample cell. All samples are reagent grade and were prepared in a 0.1 M
phosphate buffer at pH 7.35 and spectra were collected at 38.0 ~ 0.2°C.
Six single analyte solutions were prepared: 4000 & 8000 mg/dL albumin, 2000 &
4000 mg/dL globulin, and 200 & 400 mg/dL triacetin. Eight additional samples were prepared consisting of all possible permutations and combinations of the above six sample concentrations. For example, one sample consisted of 8000 mg/dL
albumin, 2000 mgldL globulin, and 200 mg/dL triacetin.
Results and Discussion: Three spectra of water, 8000 mg/dL albumin, 2000 mg/dL
globulin and 200 mg/dL triacetin appear primarily as water absorbance bands, as shown on Fig. 32. Subtraction of the water with the same algorithm used in the basis data set that attempts to match pathlength and temperature effects (discussed above) was employed to minimize the residual about zero absorbance over the spectral ranges 1640 to 1655 nm and 2077 to 2085 nm, as shown in Fig. 33. Results of incomplete temperature and pathlength subtraction dominate in the regions surrounding 1890 to 2010 nm where no signal results due to large water absorbance. The resulting spectra show the six dominant protein absorbance bands centered at 1690, 1730, 2060, 2170, 2285, and 2335 nm.
Spectra of the single analyte albumin samples are shown in Fig. 34. The 8000 mg/dL
albumin peaks are nearly exactly double the 4000 mg/dL albumin peaks indicating that Beer's law is holding. The average of the 8000 mg/dL albumin spectra was subtracted from the spectra in Fig. 33 to yield the spectra shown in Fig. 35.
Overlaid 3o with this are the 2000 mgldL globulin spectra. Clearly, the basic shape of the globulin spectra is discernible after subtraction of the 100,000 mg/dL (100 g/dL) water and the 8000 mg/dL albumin. The difference is the sum of the 200 mg/dL triacetin and baseline drift.
Spectra of the single analyte globulin samples are shown in Fig. 36. The 4000 mg/dL
globulin peaks (260, 261, 262) are nearly exactly double the 2000 mg/dL
globulin peaks (263, 264). Again, the average of the 2000 mg/dL globulin spectra is subtracted from the spectra shown in Fig. 33 to yield the spectra in Fig. 37. Overlaid with this are the standard 200 mg/dL triacetin spectra. Once again, the 200 mg/dL
triacetin peaks centered at 1675, 1715, 1760, 2130, 2250, and 2320 mg/dL can be seen after the subtraction of 100,000 mg/dL water, 8000 mgldL albumin, and the 2000 mg/dL
globulin. Unknown concentrations may be subtracted by rotation.
io Conclusions: For a relatively simple mixture, subtraction of the high concentration water, albumin, and globulin results in spectra of triacetin. Clearly, small errors in temperature and pathlength correction propagate into large errors of baseline for the lower concentration analytes. It is also possible that the error in subtraction may be due to scattering. To correct for this, a standard multiple scatter correction algorithm may be used. Clearly, straight subtraction can yield spectra that visually appear to yield higher signal to noise for the lower concentration analytes.
NOTE: the only two species in serum that have higher near-IR absorption than 2o glucose that were not included in this Example are cholesterol and urea.
***
In applying the invention, direct spectral subtraction is replaced with iterative subtraction, based upon regions of minimal or defined absorbance of remaining analytes. In another, equally preferred embodiment of the invention, another concentration gap may be taken advantage of for purposes of isolating the analyte vis-a-vis interferants. Two presently preferred approaches include:
3o ~ Analyze with multivariate techniques because the dynamic range of interferences and glucose is the same; and ~ Further removal of triglycerides, cholesterol, urea by deconvolution/
subtraction.
One approach to generating basis sets is iterative. For example, within a sample, after subtracting water, a determination of albumin and globulin is made. Once albumin and globulin are determined, and there is knowledge of water concentration, the albumin and globulin may be again removed, only this time more accurately.
This iterative process proceeds to some predetermined limit of precision, and then triglycerides and cholesterol are integrated into the analysis.
to Although the invention is described herein with reference to the preferred embodiment, one skilled in the art will readily appreciate that other applications may be substituted for those set forth herein without departing from the spirit and scope of the present invention. Accordingly, the invention should only be limited by the Claims included below.
Claims (46)
1. ~A method for determining the concentration of a target analyte in a sample using multi-spectral analysis, comprising the steps of:
generating at least one basis set that includes at least one interfering component in said sample; and applying at least one transform incorporating said at least one basis set to a spectroscopic signal representative of said sample;
wherein a component of said sample corresponding to said analyte is identified by application of said at least one transform.
generating at least one basis set that includes at least one interfering component in said sample; and applying at least one transform incorporating said at least one basis set to a spectroscopic signal representative of said sample;
wherein a component of said sample corresponding to said analyte is identified by application of said at least one transform.
2. ~The method of Claim 1, wherein said sample is serum; and wherein said basis set comprises any of a set of interfering spectral components that include at least water, temperature and/or hydrogen effects, bonding effects, albumin, globulin, protein, triglycerides, cholesterol, urea, scatter correction, refractive index correction, depth of penetration, and organic, body, and physical components.
3. ~The method of Claim 1, wherein said basis set excludes those components that do not interfere with detection of said analyte.
4. ~The method of Claim 1, further comprising the step of:
identifying all relevant interfering components.
identifying all relevant interfering components.
5. ~The method of Claim 4, further comprising the step of:
determining how each of said interfering components interact.
determining how each of said interfering components interact.
6. ~The method of Claim 5, further comprising the step of:
extracting each of said interfering components.
extracting each of said interfering components.
7. ~The method of Claim 6, further comprising the step of:
comparing spectra for each of said interfering components with that of each of said interfering components in solution.
comparing spectra for each of said interfering components with that of each of said interfering components in solution.
8. ~The method of Claim 1, wherein a basis set is sequentially and iteratively generated for each of said interfering components.
9. ~The method of Claim 1, further comprising the steps of:
characterizing each of said interfering components in said sample; and subtracting each of said interfering components from spectra produced at a frequency of interest.
characterizing each of said interfering components in said sample; and subtracting each of said interfering components from spectra produced at a frequency of interest.
10. ~The method of Claim 8, further comprising the step of:
combining said basis sets mathematically to generate a set of transforms that may be stored in a look-up table for use during analysis.
combining said basis sets mathematically to generate a set of transforms that may be stored in a look-up table for use during analysis.
11. ~The method of Claim 1, further comprising the steps of:
applying said at least one transform incorporating said at least one basis set to a signal produced during said multi-spectral analysis to identify a signal representative of said analyte;
applying multivariate analysis to said signal.
applying said at least one transform incorporating said at least one basis set to a signal produced during said multi-spectral analysis to identify a signal representative of said analyte;
applying multivariate analysis to said signal.
12. ~The method of Claim 11, wherein said multivariate analysis comprises a partial least squares analysis, followed by a principal components analysis.
13. ~A method of generating at least one basis set for application in determining the concentration of a target analyte in a sample using multi-spectral analysis, said method comprising the steps of:
identifying at least one relevant interfering component of said sample at a same frequency as that of said analyte;
identifying said at least one relevant interfering component at other frequencies to quantify absorbance of said interfering components at said other frequencies; and removing said at least one interfering component at said analyte frequency.
identifying at least one relevant interfering component of said sample at a same frequency as that of said analyte;
identifying said at least one relevant interfering component at other frequencies to quantify absorbance of said interfering components at said other frequencies; and removing said at least one interfering component at said analyte frequency.
14. ~The method of Claim 13, wherein each step of said method is repeated for each of said at least one interfering component to produce a plurality of basis sets for an analyte.
15. ~The method of Claim 13, further comprising the step of determining the concentration of said target analyte in said sample with said basis set by:
collecting spectra data with a spectroscopic device;
converting said spectral data collected by said spectroscopic device to digital data;
operating upon such digital input data in accordance with various transforms stored in one or more look-up tables (LUTs), wherein said LUTs contain transforms that incorporate said basis set, and wherein said transforms use said basis set to identify and remove interfering constituents from the spectral signal produced by said spectroscopic device.
collecting spectra data with a spectroscopic device;
converting said spectral data collected by said spectroscopic device to digital data;
operating upon such digital input data in accordance with various transforms stored in one or more look-up tables (LUTs), wherein said LUTs contain transforms that incorporate said basis set, and wherein said transforms use said basis set to identify and remove interfering constituents from the spectral signal produced by said spectroscopic device.
16. ~The method of Claim 13, further comprising the step of storing said basis set in a look-up table.
17. ~The method of Claim 13, said basis set comprising:
a series of spectra of said analyte at different physiological concentrations of interest.
a series of spectra of said analyte at different physiological concentrations of interest.
18. ~The method of Claim 15, wherein at least one transform incorporating said basis set is applied before or in connection with a physical model that corrects for interfering physical factors that include any of scattering, pathlength, and temperature.
19. ~The method of Claim 13, further comprising the step of providing a plurality of basis sets that are used to quantify an analyte in a liquid sample.
20. ~The method of Claim 13, further comprising the step of selecting different pathlengths for each spectral window.
21. ~The method of Claim 20, wherein said pathlengths comprise:
1 mm for a combination band region;
2 to 8 mm for a first overtone region; and mm or greater for a second overtone region.
1 mm for a combination band region;
2 to 8 mm for a first overtone region; and mm or greater for a second overtone region.
22. ~The method of Claim 13, wherein at least one transform incorporating one or more basis sets are applied to a spectroscopic signal during analysis to produce an accurate spectral representation from which analyte concentration may be accurately determined.
23. ~The method of Claim 13, wherein said basis set includes all interfering components found in said sample.
24. ~An apparatus for determining the concentration of a target analyte in a sample using multi-spectral analysis, comprising:
means for generating a spectroscopic signal representative of said sample; and means for applying at least one transform incorporating at least one basis set that includes at least one interfering component in said sample to said spectroscopic signal;
wherein a component of said sample corresponding to said analyte is identified by application of said at least one transform incorporating said at least one basis set.
means for generating a spectroscopic signal representative of said sample; and means for applying at least one transform incorporating at least one basis set that includes at least one interfering component in said sample to said spectroscopic signal;
wherein a component of said sample corresponding to said analyte is identified by application of said at least one transform incorporating said at least one basis set.
25. ~The apparatus of Claim 24, wherein said sample is serum; and wherein said basis set comprises any of a set of interfering spectral components that includes at least water, temperature and/or hydrogen effects, bonding effects, albumin, globulin, protein, triglycerides, cholesterol, urea, scatter correction, refractive index correction, depth of penetration, and organic, body, and physical components.
26. ~The apparatus of Claim 24, wherein said basis set excludes those components that do not interfere with detection of said analyte.
27. ~The apparatus of Claim 24, said basis set further comprising:
all relevant interfering components.
all relevant interfering components.
28. ~The apparatus of Claim 27, wherein a basis set is generated by determining how each of said interfering components interact.
29. ~The apparatus of Claim 28, wherein a basis set is further generated by extracting each of said interfering components.
30. ~The apparatus of Claim 29, wherein a basis set is further generated by comparing spectra for each of said interfering components with that of each of said interfering components in solution.
31. ~The apparatus of Claim 24, further comprising:
at least one transform incorporating a basis set for each of said interfering components.
at least one transform incorporating a basis set for each of said interfering components.
32. ~The apparatus of Claim 24, wherein a basis set is generated by characterizing each of said interfering components in said sample; and subtracting each of said interfering components from spectra produced at a frequency of interest.
33. ~The apparatus of Claim 31, wherein said at least one transform comprises a mathematically generated set of transforms that may be stored in a look-up table for use during analysis, said set of transforms incorporating said basis set.
34. ~The apparatus of Claim 24, wherein said at least one transform is applied to a signal produced during said multi-spectral analysis to identify a signal representative of said analyte; and wherein multivariate analysis is applied to said signal.
35. ~The apparatus of Claim 34, wherein said multivariate analysis comprises a partial least squares analysis, followed by a principal components analysis.
36. ~An apparatus for determining the concentration of a target analyte in a sample using multi-spectral analysis, comprising:
a spectroscopic device for collecting spectral data;
means for converting said spectral data collected by said spectroscopic device to digital data;
a processor for operating upon such digital input data in accordance with various transform stored in one or more look-up tables (LUTs), wherein said LUTs contain transforms that incorporate a basis set;~
wherein a basis set comprises spectral information representative of relevant interfering components of said sample at a same frequency as that of said analyte; and spectral information representative of substantially all of said relevant interfering components at other frequencies to quantify absorbance of said interfering components at said other frequencies in which spectral information of said interfering components is removed from a sample spectra at said analyte frequency;
a basis set being stored in a memory for use by said processor during multi-spectral analysis; and wherein said transforms use at least one basis set to identify and remove substantially all interfering constituents from the spectral signal produced by said spectroscopic device.
a spectroscopic device for collecting spectral data;
means for converting said spectral data collected by said spectroscopic device to digital data;
a processor for operating upon such digital input data in accordance with various transform stored in one or more look-up tables (LUTs), wherein said LUTs contain transforms that incorporate a basis set;~
wherein a basis set comprises spectral information representative of relevant interfering components of said sample at a same frequency as that of said analyte; and spectral information representative of substantially all of said relevant interfering components at other frequencies to quantify absorbance of said interfering components at said other frequencies in which spectral information of said interfering components is removed from a sample spectra at said analyte frequency;
a basis set being stored in a memory for use by said processor during multi-spectral analysis; and wherein said transforms use at least one basis set to identify and remove substantially all interfering constituents from the spectral signal produced by said spectroscopic device.
37. ~The apparatus of Claim 36, wherein said transforms use a plurality of basis sets for an analyte.
38. ~The apparatus of Claim 36, a basis set comprising:
a series of spectra of said analyte at different physiological concentrations of interest.
a series of spectra of said analyte at different physiological concentrations of interest.
39. ~The apparatus of Claim 36, wherein one or more transforms are applied before or in connection with a physical model that corrects for interfering physical factors that include any of scattering, pathlength, and temperature.
40. ~The apparatus of Claim 36, said transforms using a plurality of basis sets that are used to quantify an analyte in a liquid sample.
41. ~The apparatus of Claim 36, wherein different pathlengths are selected for each spectral window.
42. ~The apparatus of Claim 41, wherein said pathlengths comprise:
1 mm for a combination band region;
to 10 mm for a first overtone region; and mm or greater for a second overtone region.
1 mm for a combination band region;
to 10 mm for a first overtone region; and mm or greater for a second overtone region.
43. ~The apparatus of Claim 36, wherein one or more transforms are applied to a spectroscopic signal during analysis to produce an accurate spectral representation from which analyte concentration may be accurately determined.
44. ~The apparatus of Claim 36, wherein a basis set includes all interfering components found in said sample.
45. ~The apparatus of Claim 36, wherein said spectral information is non-invasively collected.
46. ~The apparatus of Claim 36, wherein said means for converting spectral data to digital data comprises an analog-to-digital converter.
Applications Claiming Priority (3)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US08/911,588 US6115673A (en) | 1997-08-14 | 1997-08-14 | Method and apparatus for generating basis sets for use in spectroscopic analysis |
US08/911,588 | 1997-08-14 | ||
PCT/US1998/015673 WO1999009395A1 (en) | 1997-08-14 | 1998-07-27 | Method and apparatus for generating basis sets for use in spectroscopic analysis |
Publications (2)
Publication Number | Publication Date |
---|---|
CA2299285A1 CA2299285A1 (en) | 1999-02-25 |
CA2299285C true CA2299285C (en) | 2006-01-24 |
Family
ID=25430514
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CA002299285A Expired - Fee Related CA2299285C (en) | 1997-08-14 | 1998-07-27 | Method and apparatus for generating basis sets for use in spectroscopic analysis |
Country Status (16)
Country | Link |
---|---|
US (1) | US6115673A (en) |
EP (2) | EP1496350A3 (en) |
JP (1) | JP2001516019A (en) |
KR (1) | KR100389590B1 (en) |
CN (1) | CN1107861C (en) |
AR (1) | AR013417A1 (en) |
AU (1) | AU738441B2 (en) |
BR (1) | BR9811938A (en) |
CA (1) | CA2299285C (en) |
IL (1) | IL134456A (en) |
MY (1) | MY120544A (en) |
NO (1) | NO20000692L (en) |
NZ (1) | NZ502821A (en) |
TR (1) | TR200000402T2 (en) |
TW (1) | TWI226930B (en) |
WO (1) | WO1999009395A1 (en) |
Families Citing this family (301)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6240306B1 (en) | 1995-08-09 | 2001-05-29 | Rio Grande Medical Technologies, Inc. | Method and apparatus for non-invasive blood analyte measurement with fluid compartment equilibration |
US7890158B2 (en) | 2001-06-05 | 2011-02-15 | Lumidigm, Inc. | Apparatus and method of biometric determination using specialized optical spectroscopy systems |
US6871169B1 (en) * | 1997-08-14 | 2005-03-22 | Sensys Medical, Inc. | Combinative multivariate calibration that enhances prediction ability through removal of over-modeled regions |
US7383069B2 (en) * | 1997-08-14 | 2008-06-03 | Sensys Medical, Inc. | Method of sample control and calibration adjustment for use with a noninvasive analyzer |
US7010336B2 (en) * | 1997-08-14 | 2006-03-07 | Sensys Medical, Inc. | Measurement site dependent data preprocessing method for robust calibration and prediction |
US7039446B2 (en) * | 2001-01-26 | 2006-05-02 | Sensys Medical, Inc. | Indirect measurement of tissue analytes through tissue properties |
WO2002065090A2 (en) * | 2001-01-26 | 2002-08-22 | Sensys Medical | Noninvasive measurement of glucose through the optical properties of tissue |
JP2000069502A (en) * | 1998-08-24 | 2000-03-03 | Sony Corp | Video signal processing circuit and image pickup device using it |
US6475800B1 (en) * | 1999-07-22 | 2002-11-05 | Instrumentation Metrics, Inc. | Intra-serum and intra-gel for modeling human skin tissue |
WO2001023868A1 (en) * | 1999-09-28 | 2001-04-05 | Bran + Luebbe Gmbh | Pcb measurement in pork fat with nir |
US6816605B2 (en) | 1999-10-08 | 2004-11-09 | Lumidigm, Inc. | Methods and systems for biometric identification of individuals using linear optical spectroscopy |
AU2001237067A1 (en) * | 2000-02-18 | 2001-08-27 | Argose, Inc. | Reduction of inter-subject variation via transfer standardization |
US6587199B1 (en) * | 2000-02-25 | 2003-07-01 | Sensys Medical, Inc. | Embedded data acquisition and control system for non-invasive glucose prediction instrument |
ES2278749T3 (en) * | 2000-04-17 | 2007-08-16 | Becton Dickinson And Company | METHOD FOR ANALYZING SUBSTANCE MIXTURES. |
US7606608B2 (en) * | 2000-05-02 | 2009-10-20 | Sensys Medical, Inc. | Optical sampling interface system for in-vivo measurement of tissue |
US7519406B2 (en) | 2004-04-28 | 2009-04-14 | Sensys Medical, Inc. | Noninvasive analyzer sample probe interface method and apparatus |
US20060211931A1 (en) * | 2000-05-02 | 2006-09-21 | Blank Thomas B | Noninvasive analyzer sample probe interface method and apparatus |
US20070179367A1 (en) * | 2000-05-02 | 2007-08-02 | Ruchti Timothy L | Method and Apparatus for Noninvasively Estimating a Property of an Animal Body Analyte from Spectral Data |
AU2001256611A1 (en) * | 2000-05-16 | 2001-11-26 | Jeacle Limited | Photometric analysis of natural waters |
KR100597016B1 (en) * | 2000-08-07 | 2006-07-06 | 미쓰이 가가쿠 가부시키가이샤 | Production control method |
US6549861B1 (en) * | 2000-08-10 | 2003-04-15 | Euro-Celtique, S.A. | Automated system and method for spectroscopic analysis |
AU2001286439A1 (en) * | 2000-08-10 | 2002-02-25 | Euroceltique S.A. | Automated system and method for spectroscopic analysis |
US6675030B2 (en) | 2000-08-21 | 2004-01-06 | Euro-Celtique, S.A. | Near infrared blood glucose monitoring system |
JP4054853B2 (en) * | 2000-10-17 | 2008-03-05 | 独立行政法人農業・食品産業技術総合研究機構 | Blood analysis using near infrared spectroscopy |
US7599351B2 (en) * | 2001-03-20 | 2009-10-06 | Verizon Business Global Llc | Recursive query for communications network data |
US6574490B2 (en) | 2001-04-11 | 2003-06-03 | Rio Grande Medical Technologies, Inc. | System for non-invasive measurement of glucose in humans |
US7043288B2 (en) | 2002-04-04 | 2006-05-09 | Inlight Solutions, Inc. | Apparatus and method for spectroscopic analysis of tissue to detect diabetes in an individual |
IL143904A0 (en) * | 2001-06-21 | 2002-04-21 | Glucon Inc | Method and apparatus for measuring temperature |
US6697658B2 (en) | 2001-07-02 | 2004-02-24 | Masimo Corporation | Low power pulse oximeter |
US20040064299A1 (en) * | 2001-08-10 | 2004-04-01 | Howard Mark | Automated system and method for spectroscopic analysis |
US7077565B2 (en) * | 2001-11-15 | 2006-07-18 | Glucon, Inc. | Method for measuring temperature of substances from measurement of absorption coefficients |
US7009180B2 (en) * | 2001-12-14 | 2006-03-07 | Optiscan Biomedical Corp. | Pathlength-independent methods for optically determining material composition |
EP1850114A1 (en) * | 2001-12-14 | 2007-10-31 | Optiscan Biomedical Corporation | Spectroscopic method of determining an analyte concentration in a sample |
AU2002359713B2 (en) * | 2001-12-14 | 2008-10-30 | Optiscan Biomedical Corporation | Spectroscopic method of determining an analyte concentration in a sample |
US6862534B2 (en) * | 2001-12-14 | 2005-03-01 | Optiscan Biomedical Corporation | Method of determining an analyte concentration in a sample from an absorption spectrum |
US7355512B1 (en) | 2002-01-24 | 2008-04-08 | Masimo Corporation | Parallel alarm processor |
US20070149868A1 (en) * | 2002-03-08 | 2007-06-28 | Blank Thomas B | Method and Apparatus for Photostimulation Enhanced Analyte Property Estimation |
US8504128B2 (en) * | 2002-03-08 | 2013-08-06 | Glt Acquisition Corp. | Method and apparatus for coupling a channeled sample probe to tissue |
US7440786B2 (en) * | 2002-03-08 | 2008-10-21 | Sensys Medical, Inc. | Method and apparatus for presentation of noninvasive glucose concentration information |
US20050187439A1 (en) * | 2003-03-07 | 2005-08-25 | Blank Thomas B. | Sampling interface system for in-vivo estimation of tissue analyte concentration |
EP1499231A4 (en) * | 2002-03-08 | 2007-09-26 | Sensys Medical Inc | Compact apparatus for noninvasive measurement of glucose through near-infrared spectroscopy |
US20050054908A1 (en) * | 2003-03-07 | 2005-03-10 | Blank Thomas B. | Photostimulation method and apparatus in combination with glucose determination |
US7697966B2 (en) * | 2002-03-08 | 2010-04-13 | Sensys Medical, Inc. | Noninvasive targeting system method and apparatus |
US8718738B2 (en) * | 2002-03-08 | 2014-05-06 | Glt Acquisition Corp. | Method and apparatus for coupling a sample probe with a sample site |
US6850788B2 (en) | 2002-03-25 | 2005-02-01 | Masimo Corporation | Physiological measurement communications adapter |
US6654125B2 (en) | 2002-04-04 | 2003-11-25 | Inlight Solutions, Inc | Method and apparatus for optical spectroscopy incorporating a vertical cavity surface emitting laser (VCSEL) as an interferometer reference |
US7689268B2 (en) * | 2002-08-05 | 2010-03-30 | Infraredx, Inc. | Spectroscopic unwanted signal filters for discrimination of vulnerable plaque and method therefor |
US7259906B1 (en) | 2002-09-03 | 2007-08-21 | Cheetah Omni, Llc | System and method for voice control of medical devices |
JP2004156912A (en) * | 2002-11-01 | 2004-06-03 | Jasco Corp | Method and apparatus for measuring bod and method and apparatus for treating waste water |
US6920345B2 (en) | 2003-01-24 | 2005-07-19 | Masimo Corporation | Optical sensor including disposable and reusable elements |
US20050033127A1 (en) * | 2003-01-30 | 2005-02-10 | Euro-Celtique, S.A. | Wireless blood glucose monitoring system |
JP4565233B2 (en) * | 2003-02-20 | 2010-10-20 | 株式会社キャンパスクリエイト | Flow rate measuring method and measuring apparatus used therefor |
US7620674B2 (en) * | 2003-03-07 | 2009-11-17 | Sensys Medical, Inc. | Method and apparatus for enhanced estimation of an analyte property through multiple region transformation |
US7640140B2 (en) * | 2003-03-07 | 2009-12-29 | Sensys Medical, Inc. | Method of processing noninvasive spectra |
US7751594B2 (en) | 2003-04-04 | 2010-07-06 | Lumidigm, Inc. | White-light spectral biometric sensors |
WO2004090786A2 (en) | 2003-04-04 | 2004-10-21 | Lumidigm, Inc. | Multispectral biometric sensor |
US7460696B2 (en) | 2004-06-01 | 2008-12-02 | Lumidigm, Inc. | Multispectral imaging biometrics |
US7668350B2 (en) | 2003-04-04 | 2010-02-23 | Lumidigm, Inc. | Comparative texture analysis of tissue for biometric spoof detection |
US7271912B2 (en) * | 2003-04-15 | 2007-09-18 | Optiscan Biomedical Corporation | Method of determining analyte concentration in a sample using infrared transmission data |
US7500950B2 (en) | 2003-07-25 | 2009-03-10 | Masimo Corporation | Multipurpose sensor port |
US20070234300A1 (en) * | 2003-09-18 | 2007-10-04 | Leake David W | Method and Apparatus for Performing State-Table Driven Regression Testing |
US7483729B2 (en) | 2003-11-05 | 2009-01-27 | Masimo Corporation | Pulse oximeter access apparatus and method |
WO2005091970A2 (en) * | 2004-03-06 | 2005-10-06 | Michael Trainer | Methods and apparatus for determining the size and shape of particles |
US9297737B2 (en) | 2004-03-06 | 2016-03-29 | Michael Trainer | Methods and apparatus for determining characteristics of particles |
EP1722676B1 (en) | 2004-03-08 | 2012-12-19 | Masimo Corporation | Physiological parameter system |
US20080033275A1 (en) * | 2004-04-28 | 2008-02-07 | Blank Thomas B | Method and Apparatus for Sample Probe Movement Control |
US8868147B2 (en) | 2004-04-28 | 2014-10-21 | Glt Acquisition Corp. | Method and apparatus for controlling positioning of a noninvasive analyzer sample probe |
US8229185B2 (en) | 2004-06-01 | 2012-07-24 | Lumidigm, Inc. | Hygienic biometric sensors |
US7822452B2 (en) | 2004-08-11 | 2010-10-26 | Glt Acquisition Corp. | Method for data reduction and calibration of an OCT-based blood glucose monitor |
US8787630B2 (en) | 2004-08-11 | 2014-07-22 | Lumidigm, Inc. | Multispectral barcode imaging |
CA2584162C (en) * | 2004-10-21 | 2017-11-28 | Optiscan Biomedical Corporation | Method and apparatus for determining an analyte concentration in a sample having interferents |
US7722537B2 (en) * | 2005-02-14 | 2010-05-25 | Optiscan Biomedical Corp. | Method and apparatus for detection of multiple analytes |
US8190223B2 (en) | 2005-03-01 | 2012-05-29 | Masimo Laboratories, Inc. | Noninvasive multi-parameter patient monitor |
US20060217602A1 (en) * | 2005-03-04 | 2006-09-28 | Alan Abul-Haj | Method and apparatus for noninvasive targeting |
US8180579B2 (en) * | 2005-03-25 | 2012-05-15 | Lawrence Livermore National Security, Llc | Real time gamma-ray signature identifier |
US7801338B2 (en) | 2005-04-27 | 2010-09-21 | Lumidigm, Inc. | Multispectral biometric sensors |
US7962188B2 (en) | 2005-10-14 | 2011-06-14 | Masimo Corporation | Robust alarm system |
US7519253B2 (en) | 2005-11-18 | 2009-04-14 | Omni Sciences, Inc. | Broadband or mid-infrared fiber light sources |
US8182443B1 (en) | 2006-01-17 | 2012-05-22 | Masimo Corporation | Drug administration controller |
US8219172B2 (en) | 2006-03-17 | 2012-07-10 | Glt Acquisition Corp. | System and method for creating a stable optical interface |
US10188348B2 (en) | 2006-06-05 | 2019-01-29 | Masimo Corporation | Parameter upgrade system |
US8355545B2 (en) | 2007-04-10 | 2013-01-15 | Lumidigm, Inc. | Biometric detection using spatial, temporal, and/or spectral techniques |
US8175346B2 (en) | 2006-07-19 | 2012-05-08 | Lumidigm, Inc. | Whole-hand multispectral biometric imaging |
US7995808B2 (en) | 2006-07-19 | 2011-08-09 | Lumidigm, Inc. | Contactless multispectral biometric capture |
JP2009544108A (en) | 2006-07-19 | 2009-12-10 | ルミダイム インコーポレイテッド | Multispectral image for multiple biometric authentication |
US7801339B2 (en) | 2006-07-31 | 2010-09-21 | Lumidigm, Inc. | Biometrics with spatiospectral spoof detection |
US7804984B2 (en) | 2006-07-31 | 2010-09-28 | Lumidigm, Inc. | Spatial-spectral fingerprint spoof detection |
US8457707B2 (en) | 2006-09-20 | 2013-06-04 | Masimo Corporation | Congenital heart disease monitor |
US8840549B2 (en) | 2006-09-22 | 2014-09-23 | Masimo Corporation | Modular patient monitor |
US9861305B1 (en) | 2006-10-12 | 2018-01-09 | Masimo Corporation | Method and apparatus for calibration to reduce coupling between signals in a measurement system |
US7880626B2 (en) | 2006-10-12 | 2011-02-01 | Masimo Corporation | System and method for monitoring the life of a physiological sensor |
US8255026B1 (en) | 2006-10-12 | 2012-08-28 | Masimo Corporation, Inc. | Patient monitor capable of monitoring the quality of attached probes and accessories |
US8280473B2 (en) | 2006-10-12 | 2012-10-02 | Masino Corporation, Inc. | Perfusion index smoother |
WO2008073855A2 (en) | 2006-12-09 | 2008-06-19 | Masimo Corporation | Plethysmograph variability processor |
US8652060B2 (en) | 2007-01-20 | 2014-02-18 | Masimo Corporation | Perfusion trend indicator |
WO2008134135A2 (en) | 2007-03-21 | 2008-11-06 | Lumidigm, Inc. | Biometrics based on locally consistent features |
US8374665B2 (en) | 2007-04-21 | 2013-02-12 | Cercacor Laboratories, Inc. | Tissue profile wellness monitor |
US20090036759A1 (en) * | 2007-08-01 | 2009-02-05 | Ault Timothy E | Collapsible noninvasive analyzer method and apparatus |
CA3105353A1 (en) | 2007-10-10 | 2009-04-16 | Optiscan Biomedical Corporation | Fluid component analysis system and method for glucose monitoring and control |
WO2009111542A2 (en) | 2008-03-04 | 2009-09-11 | Glucolight Corporation | Methods and systems for analyte level estimation in optical coherence tomography |
JP5575752B2 (en) | 2008-05-02 | 2014-08-20 | マシモ コーポレイション | Monitor configuration system |
JP2011519684A (en) | 2008-05-05 | 2011-07-14 | マシモ コーポレイション | Pulse oximeter system with electrical disconnect circuit |
US8577431B2 (en) | 2008-07-03 | 2013-11-05 | Cercacor Laboratories, Inc. | Noise shielding for a noninvasive device |
US8630691B2 (en) | 2008-08-04 | 2014-01-14 | Cercacor Laboratories, Inc. | Multi-stream sensor front ends for noninvasive measurement of blood constituents |
US7959598B2 (en) | 2008-08-20 | 2011-06-14 | Asante Solutions, Inc. | Infusion pump systems and methods |
SE532941C2 (en) | 2008-09-15 | 2010-05-18 | Phasein Ab | Gas sampling line for breathing gases |
US8771204B2 (en) | 2008-12-30 | 2014-07-08 | Masimo Corporation | Acoustic sensor assembly |
US8588880B2 (en) | 2009-02-16 | 2013-11-19 | Masimo Corporation | Ear sensor |
US9323894B2 (en) | 2011-08-19 | 2016-04-26 | Masimo Corporation | Health care sanitation monitoring system |
WO2010102069A2 (en) | 2009-03-04 | 2010-09-10 | Masimo Corporation | Medical monitoring system |
US10007758B2 (en) | 2009-03-04 | 2018-06-26 | Masimo Corporation | Medical monitoring system |
US10032002B2 (en) | 2009-03-04 | 2018-07-24 | Masimo Corporation | Medical monitoring system |
US8388353B2 (en) | 2009-03-11 | 2013-03-05 | Cercacor Laboratories, Inc. | Magnetic connector |
US8571619B2 (en) | 2009-05-20 | 2013-10-29 | Masimo Corporation | Hemoglobin display and patient treatment |
US10475529B2 (en) | 2011-07-19 | 2019-11-12 | Optiscan Biomedical Corporation | Method and apparatus for analyte measurements using calibration sets |
US8473020B2 (en) | 2009-07-29 | 2013-06-25 | Cercacor Laboratories, Inc. | Non-invasive physiological sensor cover |
US8731250B2 (en) | 2009-08-26 | 2014-05-20 | Lumidigm, Inc. | Multiplexed biometric imaging |
US20110137297A1 (en) | 2009-09-17 | 2011-06-09 | Kiani Massi Joe E | Pharmacological management system |
JP5394501B2 (en) | 2009-10-02 | 2014-01-22 | シャープ株式会社 | Blood vessel condition monitoring method |
US20110082711A1 (en) | 2009-10-06 | 2011-04-07 | Masimo Laboratories, Inc. | Personal digital assistant or organizer for monitoring glucose levels |
KR101239439B1 (en) * | 2009-10-09 | 2013-03-06 | 숭실대학교산학협력단 | Detection Method of Cancer Cells Using Reflection-Absorption Infrared Spectroscopy |
US9839381B1 (en) | 2009-11-24 | 2017-12-12 | Cercacor Laboratories, Inc. | Physiological measurement system with automatic wavelength adjustment |
DE112010004682T5 (en) | 2009-12-04 | 2013-03-28 | Masimo Corporation | Calibration for multi-level physiological monitors |
US9153112B1 (en) | 2009-12-21 | 2015-10-06 | Masimo Corporation | Modular patient monitor |
US11289199B2 (en) | 2010-01-19 | 2022-03-29 | Masimo Corporation | Wellness analysis system |
JP2013521054A (en) | 2010-03-01 | 2013-06-10 | マシモ コーポレイション | Adaptive alarm system |
JP4918732B2 (en) * | 2010-03-05 | 2012-04-18 | 日本電気株式会社 | Light measuring apparatus and method |
US8584345B2 (en) | 2010-03-08 | 2013-11-19 | Masimo Corporation | Reprocessing of a physiological sensor |
US8570149B2 (en) | 2010-03-16 | 2013-10-29 | Lumidigm, Inc. | Biometric imaging using an optical adaptive interface |
WO2011114578A1 (en) | 2010-03-19 | 2011-09-22 | シャープ株式会社 | Measurement device, measurement method, measurement result processing device, measurement system, measurement result processing method, control program, and recording medium |
US9307928B1 (en) | 2010-03-30 | 2016-04-12 | Masimo Corporation | Plethysmographic respiration processor |
US8666468B1 (en) | 2010-05-06 | 2014-03-04 | Masimo Corporation | Patient monitor for determining microcirculation state |
JP5601098B2 (en) * | 2010-09-03 | 2014-10-08 | ソニー株式会社 | Fluorescence intensity correction method and fluorescence intensity calculation apparatus |
JP5710767B2 (en) | 2010-09-28 | 2015-04-30 | マシモ コーポレイション | Depth of consciousness monitor including oximeter |
US9211095B1 (en) | 2010-10-13 | 2015-12-15 | Masimo Corporation | Physiological measurement logic engine |
WO2012055509A1 (en) | 2010-10-29 | 2012-05-03 | Lonza Ltd | Diketopiperazine forming dipeptidyl linker |
US20120226117A1 (en) | 2010-12-01 | 2012-09-06 | Lamego Marcelo M | Handheld processing device including medical applications for minimally and non invasive glucose measurements |
US10332630B2 (en) | 2011-02-13 | 2019-06-25 | Masimo Corporation | Medical characterization system |
US9066666B2 (en) | 2011-02-25 | 2015-06-30 | Cercacor Laboratories, Inc. | Patient monitor for monitoring microcirculation |
US9986919B2 (en) | 2011-06-21 | 2018-06-05 | Masimo Corporation | Patient monitoring system |
US9532722B2 (en) | 2011-06-21 | 2017-01-03 | Masimo Corporation | Patient monitoring system |
US11439329B2 (en) | 2011-07-13 | 2022-09-13 | Masimo Corporation | Multiple measurement mode in a physiological sensor |
US9782077B2 (en) | 2011-08-17 | 2017-10-10 | Masimo Corporation | Modulated physiological sensor |
EP3584799B1 (en) | 2011-10-13 | 2022-11-09 | Masimo Corporation | Medical monitoring hub |
US9808188B1 (en) | 2011-10-13 | 2017-11-07 | Masimo Corporation | Robust fractional saturation determination |
US9943269B2 (en) | 2011-10-13 | 2018-04-17 | Masimo Corporation | System for displaying medical monitoring data |
US9778079B1 (en) | 2011-10-27 | 2017-10-03 | Masimo Corporation | Physiological monitor gauge panel |
CN102564965B (en) * | 2011-12-31 | 2014-10-08 | 聚光科技(杭州)股份有限公司 | Detecting method based on spectroscopy detection technology |
US9392945B2 (en) | 2012-01-04 | 2016-07-19 | Masimo Corporation | Automated CCHD screening and detection |
US11172890B2 (en) | 2012-01-04 | 2021-11-16 | Masimo Corporation | Automated condition screening and detection |
US10149616B2 (en) | 2012-02-09 | 2018-12-11 | Masimo Corporation | Wireless patient monitoring device |
EP2845086B1 (en) | 2012-03-25 | 2021-12-22 | Masimo Corporation | Physiological monitor touchscreen interface |
JP6490577B2 (en) | 2012-04-17 | 2019-03-27 | マシモ・コーポレイション | How to operate a pulse oximeter device |
EP2657681A1 (en) | 2012-04-26 | 2013-10-30 | Roche Diagnostics GmbH | Improvement of the sensitivity and the dynamic range of photometric assays by generating multiple calibration curves |
US9351672B2 (en) | 2012-07-16 | 2016-05-31 | Timothy Ruchti | Multiplexed pathlength resolved noninvasive analyzer apparatus with stacked filters and method of use thereof |
US9766126B2 (en) | 2013-07-12 | 2017-09-19 | Zyomed Corp. | Dynamic radially controlled light input to a noninvasive analyzer apparatus and method of use thereof |
US9585604B2 (en) | 2012-07-16 | 2017-03-07 | Zyomed Corp. | Multiplexed pathlength resolved noninvasive analyzer apparatus with dynamic optical paths and method of use thereof |
US9351671B2 (en) | 2012-07-16 | 2016-05-31 | Timothy Ruchti | Multiplexed pathlength resolved noninvasive analyzer apparatus and method of use thereof |
US9697928B2 (en) | 2012-08-01 | 2017-07-04 | Masimo Corporation | Automated assembly sensor cable |
US9877650B2 (en) | 2012-09-20 | 2018-01-30 | Masimo Corporation | Physiological monitor with mobile computing device connectivity |
US9749232B2 (en) | 2012-09-20 | 2017-08-29 | Masimo Corporation | Intelligent medical network edge router |
US9955937B2 (en) | 2012-09-20 | 2018-05-01 | Masimo Corporation | Acoustic patient sensor coupler |
US9560996B2 (en) | 2012-10-30 | 2017-02-07 | Masimo Corporation | Universal medical system |
US9787568B2 (en) | 2012-11-05 | 2017-10-10 | Cercacor Laboratories, Inc. | Physiological test credit method |
EP2936118B1 (en) * | 2012-12-20 | 2023-06-21 | Radiometer Medical ApS | An apparatus for detecting the concentration of a component in a sample |
US10660526B2 (en) | 2012-12-31 | 2020-05-26 | Omni Medsci, Inc. | Near-infrared time-of-flight imaging using laser diodes with Bragg reflectors |
US9993159B2 (en) | 2012-12-31 | 2018-06-12 | Omni Medsci, Inc. | Near-infrared super-continuum lasers for early detection of breast and other cancers |
WO2014143276A2 (en) | 2012-12-31 | 2014-09-18 | Omni Medsci, Inc. | Short-wave infrared super-continuum lasers for natural gas leak detection, exploration, and other active remote sensing applications |
EP3181048A1 (en) | 2012-12-31 | 2017-06-21 | Omni MedSci, Inc. | Near-infrared lasers for non-invasive monitoring of glucose, ketones, hba1c, and other blood constituents |
CA2895982A1 (en) | 2012-12-31 | 2014-07-03 | Omni Medsci, Inc. | Short-wave infrared super-continuum lasers for early detection of dental caries |
US9164032B2 (en) | 2012-12-31 | 2015-10-20 | Omni Medsci, Inc. | Short-wave infrared super-continuum lasers for detecting counterfeit or illicit drugs and pharmaceutical process control |
US9724025B1 (en) | 2013-01-16 | 2017-08-08 | Masimo Corporation | Active-pulse blood analysis system |
US9965946B2 (en) | 2013-03-13 | 2018-05-08 | Masimo Corporation | Systems and methods for monitoring a patient health network |
US9936917B2 (en) | 2013-03-14 | 2018-04-10 | Masimo Laboratories, Inc. | Patient monitor placement indicator |
US9891079B2 (en) | 2013-07-17 | 2018-02-13 | Masimo Corporation | Pulser with double-bearing position encoder for non-invasive physiological monitoring |
US10555678B2 (en) | 2013-08-05 | 2020-02-11 | Masimo Corporation | Blood pressure monitor with valve-chamber assembly |
WO2015038683A2 (en) | 2013-09-12 | 2015-03-19 | Cercacor Laboratories, Inc. | Medical device management system |
EP3054849B1 (en) | 2013-10-07 | 2022-03-16 | Masimo Corporation | Regional oximetry sensor |
US11147518B1 (en) | 2013-10-07 | 2021-10-19 | Masimo Corporation | Regional oximetry signal processor |
US10832818B2 (en) | 2013-10-11 | 2020-11-10 | Masimo Corporation | Alarm notification system |
US10279247B2 (en) | 2013-12-13 | 2019-05-07 | Masimo Corporation | Avatar-incentive healthcare therapy |
US11259745B2 (en) | 2014-01-28 | 2022-03-01 | Masimo Corporation | Autonomous drug delivery system |
GB2523989B (en) | 2014-01-30 | 2020-07-29 | Insulet Netherlands B V | Therapeutic product delivery system and method of pairing |
US10123729B2 (en) | 2014-06-13 | 2018-11-13 | Nanthealth, Inc. | Alarm fatigue management systems and methods |
US10231670B2 (en) | 2014-06-19 | 2019-03-19 | Masimo Corporation | Proximity sensor in pulse oximeter |
WO2016024791A1 (en) * | 2014-08-12 | 2016-02-18 | Samsung Electronics Co., Ltd. | Sample test method, microfluidic device, and test device |
KR102310652B1 (en) * | 2014-08-12 | 2021-10-12 | 삼성전자주식회사 | Test method for sample, microfluidic device and test apparatus |
US10111591B2 (en) | 2014-08-26 | 2018-10-30 | Nanthealth, Inc. | Real-time monitoring systems and methods in a healthcare environment |
US10231657B2 (en) | 2014-09-04 | 2019-03-19 | Masimo Corporation | Total hemoglobin screening sensor |
US10383520B2 (en) | 2014-09-18 | 2019-08-20 | Masimo Semiconductor, Inc. | Enhanced visible near-infrared photodiode and non-invasive physiological sensor |
US9459201B2 (en) | 2014-09-29 | 2016-10-04 | Zyomed Corp. | Systems and methods for noninvasive blood glucose and other analyte detection and measurement using collision computing |
WO2016057553A1 (en) | 2014-10-07 | 2016-04-14 | Masimo Corporation | Modular physiological sensors |
US10568553B2 (en) | 2015-02-06 | 2020-02-25 | Masimo Corporation | Soft boot pulse oximetry sensor |
EP3253289B1 (en) | 2015-02-06 | 2020-08-05 | Masimo Corporation | Fold flex circuit for optical probes |
CN107431301B (en) | 2015-02-06 | 2021-03-30 | 迈心诺公司 | Connector assembly with retractable needle for use with medical sensors |
US10737024B2 (en) | 2015-02-18 | 2020-08-11 | Insulet Corporation | Fluid delivery and infusion devices, and methods of use thereof |
US10524738B2 (en) | 2015-05-04 | 2020-01-07 | Cercacor Laboratories, Inc. | Noninvasive sensor system with visual infographic display |
US11653862B2 (en) | 2015-05-22 | 2023-05-23 | Cercacor Laboratories, Inc. | Non-invasive optical physiological differential pathlength sensor |
US10991135B2 (en) | 2015-08-11 | 2021-04-27 | Masimo Corporation | Medical monitoring analysis and replay including indicia responsive to light attenuated by body tissue |
CN108348162B (en) | 2015-08-31 | 2021-07-23 | 梅西莫股份有限公司 | Wireless patient monitoring system and method |
US11504066B1 (en) | 2015-09-04 | 2022-11-22 | Cercacor Laboratories, Inc. | Low-noise sensor system |
CA2999410C (en) * | 2015-09-25 | 2019-08-27 | Sanmina Corporation | System and method for health monitoring using a non-invasive, multi-band biosensor |
KR102461185B1 (en) | 2015-11-30 | 2022-11-01 | 삼성전자주식회사 | Apparatus for estimating blood level of an ingredient using specrum analysis |
US11679579B2 (en) | 2015-12-17 | 2023-06-20 | Masimo Corporation | Varnish-coated release liner |
US10557792B2 (en) | 2015-12-31 | 2020-02-11 | Abb, Inc. | Spectral modeling for complex absorption spectrum interpretation |
WO2017123525A1 (en) | 2016-01-13 | 2017-07-20 | Bigfoot Biomedical, Inc. | User interface for diabetes management system |
EP3453414A1 (en) | 2016-01-14 | 2019-03-13 | Bigfoot Biomedical, Inc. | Adjusting insulin delivery rates |
US10993662B2 (en) | 2016-03-04 | 2021-05-04 | Masimo Corporation | Nose sensor |
US10537285B2 (en) | 2016-03-04 | 2020-01-21 | Masimo Corporation | Nose sensor |
US9554738B1 (en) | 2016-03-30 | 2017-01-31 | Zyomed Corp. | Spectroscopic tomography systems and methods for noninvasive detection and measurement of analytes using collision computing |
US11191484B2 (en) | 2016-04-29 | 2021-12-07 | Masimo Corporation | Optical sensor tape |
JP6835877B2 (en) * | 2016-06-02 | 2021-02-24 | シージーン アイエヌシー | In-sample target analytical substance detection method using a signal change amount data set |
US10608817B2 (en) | 2016-07-06 | 2020-03-31 | Masimo Corporation | Secure and zero knowledge data sharing for cloud applications |
US10617302B2 (en) | 2016-07-07 | 2020-04-14 | Masimo Corporation | Wearable pulse oximeter and respiration monitor |
KR102539142B1 (en) | 2016-09-05 | 2023-06-01 | 삼성전자주식회사 | Device and method for analysis of spectra, and device for measurement the blood glucose |
EP3515535A1 (en) | 2016-09-23 | 2019-07-31 | Insulet Corporation | Fluid delivery device with sensor |
US9829378B1 (en) | 2016-10-13 | 2017-11-28 | Bentley Instruments, Inc. | Determining a size of cell of a transmission spectroscopy device |
WO2018071715A1 (en) | 2016-10-13 | 2018-04-19 | Masimo Corporation | Systems and methods for patient fall detection |
US10473586B2 (en) | 2016-11-22 | 2019-11-12 | Airware, Inc. | Enhanced optical data capture using NDIR for liquids |
US10041881B2 (en) * | 2016-11-22 | 2018-08-07 | Airware, Inc. | NDIR glucose detection in liquids |
US11504058B1 (en) | 2016-12-02 | 2022-11-22 | Masimo Corporation | Multi-site noninvasive measurement of a physiological parameter |
US10750984B2 (en) | 2016-12-22 | 2020-08-25 | Cercacor Laboratories, Inc. | Methods and devices for detecting intensity of light with translucent detector |
US10721785B2 (en) | 2017-01-18 | 2020-07-21 | Masimo Corporation | Patient-worn wireless physiological sensor with pairing functionality |
WO2018156648A1 (en) | 2017-02-24 | 2018-08-30 | Masimo Corporation | Managing dynamic licenses for physiological parameters in a patient monitoring environment |
US11086609B2 (en) | 2017-02-24 | 2021-08-10 | Masimo Corporation | Medical monitoring hub |
WO2018156809A1 (en) | 2017-02-24 | 2018-08-30 | Masimo Corporation | Augmented reality system for displaying patient data |
EP3585254B1 (en) | 2017-02-24 | 2024-03-20 | Masimo Corporation | Medical device cable and method of sharing data between connected medical devices |
US10327713B2 (en) | 2017-02-24 | 2019-06-25 | Masimo Corporation | Modular multi-parameter patient monitoring device |
US10388120B2 (en) | 2017-02-24 | 2019-08-20 | Masimo Corporation | Localized projection of audible noises in medical settings |
EP3592231A1 (en) | 2017-03-10 | 2020-01-15 | Masimo Corporation | Pneumonia screener |
WO2018194992A1 (en) | 2017-04-18 | 2018-10-25 | Masimo Corporation | Nose sensor |
US10918281B2 (en) | 2017-04-26 | 2021-02-16 | Masimo Corporation | Medical monitoring device having multiple configurations |
US10856750B2 (en) | 2017-04-28 | 2020-12-08 | Masimo Corporation | Spot check measurement system |
JP7159208B2 (en) | 2017-05-08 | 2022-10-24 | マシモ・コーポレイション | A system for pairing a medical system with a network controller by using a dongle |
WO2019014629A1 (en) | 2017-07-13 | 2019-01-17 | Cercacor Laboratories, Inc. | Medical monitoring device for harmonizing physiological measurements |
US10637181B2 (en) | 2017-08-15 | 2020-04-28 | Masimo Corporation | Water resistant connector for noninvasive patient monitor |
US11298021B2 (en) | 2017-10-19 | 2022-04-12 | Masimo Corporation | Medical monitoring system |
JP7282085B2 (en) | 2017-10-31 | 2023-05-26 | マシモ・コーポレイション | System for displaying oxygen status indicators |
USD925597S1 (en) | 2017-10-31 | 2021-07-20 | Masimo Corporation | Display screen or portion thereof with graphical user interface |
US11766198B2 (en) | 2018-02-02 | 2023-09-26 | Cercacor Laboratories, Inc. | Limb-worn patient monitoring device |
USD928199S1 (en) | 2018-04-02 | 2021-08-17 | Bigfoot Biomedical, Inc. | Medication delivery device with icons |
WO2019204368A1 (en) | 2018-04-19 | 2019-10-24 | Masimo Corporation | Mobile patient alarm display |
US11883129B2 (en) | 2018-04-24 | 2024-01-30 | Cercacor Laboratories, Inc. | Easy insert finger sensor for transmission based spectroscopy sensor |
CN112236826A (en) | 2018-05-04 | 2021-01-15 | 英赛罗公司 | Safety constraints for drug delivery systems based on control algorithms |
US10932729B2 (en) | 2018-06-06 | 2021-03-02 | Masimo Corporation | Opioid overdose monitoring |
US10779098B2 (en) | 2018-07-10 | 2020-09-15 | Masimo Corporation | Patient monitor alarm speaker analyzer |
US11627895B2 (en) * | 2018-08-10 | 2023-04-18 | Samsung Electronics Co., Ltd. | Apparatus and method for estimating analyte concentration, and apparatus and method for generating analyte concentration estimation model |
US11872156B2 (en) | 2018-08-22 | 2024-01-16 | Masimo Corporation | Core body temperature measurement |
CA3112209C (en) | 2018-09-28 | 2023-08-29 | Insulet Corporation | Activity mode for artificial pancreas system |
USD917564S1 (en) | 2018-10-11 | 2021-04-27 | Masimo Corporation | Display screen or portion thereof with graphical user interface |
USD917550S1 (en) | 2018-10-11 | 2021-04-27 | Masimo Corporation | Display screen or portion thereof with a graphical user interface |
US11565039B2 (en) | 2018-10-11 | 2023-01-31 | Insulet Corporation | Event detection for drug delivery system |
USD916135S1 (en) | 2018-10-11 | 2021-04-13 | Masimo Corporation | Display screen or portion thereof with a graphical user interface |
USD999246S1 (en) | 2018-10-11 | 2023-09-19 | Masimo Corporation | Display screen or portion thereof with a graphical user interface |
US11406286B2 (en) | 2018-10-11 | 2022-08-09 | Masimo Corporation | Patient monitoring device with improved user interface |
USD998631S1 (en) | 2018-10-11 | 2023-09-12 | Masimo Corporation | Display screen or portion thereof with a graphical user interface |
USD998630S1 (en) | 2018-10-11 | 2023-09-12 | Masimo Corporation | Display screen or portion thereof with a graphical user interface |
US11389093B2 (en) | 2018-10-11 | 2022-07-19 | Masimo Corporation | Low noise oximetry cable |
CA3115776A1 (en) | 2018-10-11 | 2020-04-16 | Masimo Corporation | Patient connector assembly with vertical detents |
EP3864869A1 (en) | 2018-10-12 | 2021-08-18 | Masimo Corporation | System for transmission of sensor data using dual communication protocol |
USD897098S1 (en) | 2018-10-12 | 2020-09-29 | Masimo Corporation | Card holder set |
US11464410B2 (en) | 2018-10-12 | 2022-10-11 | Masimo Corporation | Medical systems and methods |
EP3887800B1 (en) * | 2018-11-30 | 2024-04-17 | Institut National De La Sante Et De La Recherche Medicale - Inserm | Method for determining one content in protein and associated devices and methods |
US11684296B2 (en) | 2018-12-21 | 2023-06-27 | Cercacor Laboratories, Inc. | Noninvasive physiological sensor |
JP2022529948A (en) | 2019-04-17 | 2022-06-27 | マシモ・コーポレイション | Patient monitoring systems, equipment, and methods |
CN112304918B (en) * | 2019-07-30 | 2022-04-01 | 同方威视技术股份有限公司 | Method and device for identifying mixture based on Raman spectrum and Raman spectrum detection equipment |
USD917704S1 (en) | 2019-08-16 | 2021-04-27 | Masimo Corporation | Patient monitor |
USD919100S1 (en) | 2019-08-16 | 2021-05-11 | Masimo Corporation | Holder for a patient monitor |
USD921202S1 (en) | 2019-08-16 | 2021-06-01 | Masimo Corporation | Holder for a blood pressure device |
USD985498S1 (en) | 2019-08-16 | 2023-05-09 | Masimo Corporation | Connector |
USD919094S1 (en) | 2019-08-16 | 2021-05-11 | Masimo Corporation | Blood pressure device |
US11832940B2 (en) | 2019-08-27 | 2023-12-05 | Cercacor Laboratories, Inc. | Non-invasive medical monitoring device for blood analyte measurements |
US11801344B2 (en) | 2019-09-13 | 2023-10-31 | Insulet Corporation | Blood glucose rate of change modulation of meal and correction insulin bolus quantity |
US11935637B2 (en) | 2019-09-27 | 2024-03-19 | Insulet Corporation | Onboarding and total daily insulin adaptivity |
USD927699S1 (en) | 2019-10-18 | 2021-08-10 | Masimo Corporation | Electrode pad |
KR20220083771A (en) | 2019-10-18 | 2022-06-20 | 마시모 코오퍼레이션 | Display layouts and interactive objects for patient monitoring |
CN115176155A (en) | 2019-10-25 | 2022-10-11 | 塞卡科实验室有限公司 | Indicator compounds, devices including indicator compounds, and methods of making and using the same |
CN111077103A (en) * | 2019-11-30 | 2020-04-28 | 贵州中烟工业有限责任公司 | Method for measuring content of glyceryl triacetate |
US11833329B2 (en) | 2019-12-20 | 2023-12-05 | Insulet Corporation | Techniques for improved automatic drug delivery performance using delivery tendencies from past delivery history and use patterns |
US11551802B2 (en) | 2020-02-11 | 2023-01-10 | Insulet Corporation | Early meal detection and calorie intake detection |
US11547800B2 (en) | 2020-02-12 | 2023-01-10 | Insulet Corporation | User parameter dependent cost function for personalized reduction of hypoglycemia and/or hyperglycemia in a closed loop artificial pancreas system |
EP4104037A1 (en) | 2020-02-13 | 2022-12-21 | Masimo Corporation | System and method for monitoring clinical activities |
US11879960B2 (en) | 2020-02-13 | 2024-01-23 | Masimo Corporation | System and method for monitoring clinical activities |
US11324889B2 (en) | 2020-02-14 | 2022-05-10 | Insulet Corporation | Compensation for missing readings from a glucose monitor in an automated insulin delivery system |
EP4120901A1 (en) | 2020-03-20 | 2023-01-25 | Masimo Corporation | Wearable device for noninvasive body temperature measurement |
US11607493B2 (en) | 2020-04-06 | 2023-03-21 | Insulet Corporation | Initial total daily insulin setting for user onboarding |
USD933232S1 (en) | 2020-05-11 | 2021-10-12 | Masimo Corporation | Blood pressure monitor |
USD979516S1 (en) | 2020-05-11 | 2023-02-28 | Masimo Corporation | Connector |
USD974193S1 (en) | 2020-07-27 | 2023-01-03 | Masimo Corporation | Wearable temperature measurement device |
USD980091S1 (en) | 2020-07-27 | 2023-03-07 | Masimo Corporation | Wearable temperature measurement device |
US11684716B2 (en) | 2020-07-31 | 2023-06-27 | Insulet Corporation | Techniques to reduce risk of occlusions in drug delivery systems |
USD946596S1 (en) | 2020-09-30 | 2022-03-22 | Masimo Corporation | Display screen or portion thereof with graphical user interface |
USD946597S1 (en) | 2020-09-30 | 2022-03-22 | Masimo Corporation | Display screen or portion thereof with graphical user interface |
USD946598S1 (en) | 2020-09-30 | 2022-03-22 | Masimo Corporation | Display screen or portion thereof with graphical user interface |
CN112535531B (en) * | 2020-11-27 | 2022-08-19 | 广东省医疗器械研究所 | Biological tissue welding effect detection device |
US11904140B2 (en) | 2021-03-10 | 2024-02-20 | Insulet Corporation | Adaptable asymmetric medicament cost component in a control system for medicament delivery |
USD997365S1 (en) | 2021-06-24 | 2023-08-29 | Masimo Corporation | Physiological nose sensor |
USD1000975S1 (en) | 2021-09-22 | 2023-10-10 | Masimo Corporation | Wearable temperature measurement device |
WO2023049900A1 (en) | 2021-09-27 | 2023-03-30 | Insulet Corporation | Techniques enabling adaptation of parameters in aid systems by user input |
US11439754B1 (en) | 2021-12-01 | 2022-09-13 | Insulet Corporation | Optimizing embedded formulations for drug delivery |
US20230417656A1 (en) * | 2022-06-24 | 2023-12-28 | Abb Schweiz Ag | Enhancements to laser spectroscopy modeling by measurement of hydrocarbon fuel gas compositions |
Family Cites Families (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US4893253A (en) * | 1988-03-10 | 1990-01-09 | Indiana University Foundation | Method for analyzing intact capsules and tablets by near-infrared reflectance spectrometry |
JP3213307B2 (en) * | 1989-09-18 | 2001-10-02 | ミネソタ マイニング アンド マニユフアクチユアリング カンパニー | A method for predicting the properties of biological materials by near-infrared spectral analysis |
EP0535700B1 (en) * | 1991-10-04 | 1997-03-26 | The Perkin-Elmer Corporation | Method and apparatus for comparing spectra |
US5348002A (en) * | 1992-04-23 | 1994-09-20 | Sirraya, Inc. | Method and apparatus for material analysis |
CA2174719C (en) * | 1993-08-24 | 2005-07-26 | Mark R. Robinson | A robust accurate non-invasive analyte monitor |
US5462879A (en) * | 1993-10-14 | 1995-10-31 | Minnesota Mining And Manufacturing Company | Method of sensing with emission quenching sensors |
GB9415869D0 (en) * | 1994-08-05 | 1994-09-28 | Univ Mcgill | Substrate measurement by infrared spectroscopy |
WO1996032631A1 (en) * | 1995-04-13 | 1996-10-17 | Pfizer Inc. | Calibration tranfer standards and methods |
US5743262A (en) * | 1995-06-07 | 1998-04-28 | Masimo Corporation | Blood glucose monitoring system |
US5606164A (en) * | 1996-01-16 | 1997-02-25 | Boehringer Mannheim Corporation | Method and apparatus for biological fluid analyte concentration measurement using generalized distance outlier detection |
AU711324B2 (en) * | 1995-08-07 | 1999-10-14 | Roche Diagnostics Operations Inc. | Biological fluid analysis using distance outlier detection |
US5655530A (en) * | 1995-08-09 | 1997-08-12 | Rio Grande Medical Technologies, Inc. | Method for non-invasive blood analyte measurement with improved optical interface |
US5747806A (en) * | 1996-02-02 | 1998-05-05 | Instrumentation Metrics, Inc | Method and apparatus for multi-spectral analysis in noninvasive nir spectroscopy |
-
1997
- 1997-08-14 US US08/911,588 patent/US6115673A/en not_active Expired - Lifetime
-
1998
- 1998-07-27 WO PCT/US1998/015673 patent/WO1999009395A1/en active IP Right Grant
- 1998-07-27 JP JP2000510011A patent/JP2001516019A/en active Pending
- 1998-07-27 EP EP04024619A patent/EP1496350A3/en not_active Withdrawn
- 1998-07-27 TR TR2000/00402T patent/TR200000402T2/en unknown
- 1998-07-27 CA CA002299285A patent/CA2299285C/en not_active Expired - Fee Related
- 1998-07-27 NZ NZ502821A patent/NZ502821A/en unknown
- 1998-07-27 IL IL13445698A patent/IL134456A/en not_active IP Right Cessation
- 1998-07-27 KR KR10-2000-7001500A patent/KR100389590B1/en not_active IP Right Cessation
- 1998-07-27 BR BR9811938-9A patent/BR9811938A/en not_active Application Discontinuation
- 1998-07-27 EP EP98937229A patent/EP1004016A1/en not_active Withdrawn
- 1998-07-27 CN CN98810041A patent/CN1107861C/en not_active Expired - Fee Related
- 1998-07-27 AU AU85992/98A patent/AU738441B2/en not_active Ceased
- 1998-08-12 MY MYPI98003665A patent/MY120544A/en unknown
- 1998-08-13 AR ARP980104018A patent/AR013417A1/en active IP Right Grant
- 1998-09-10 TW TW087113352A patent/TWI226930B/en not_active IP Right Cessation
-
2000
- 2000-02-11 NO NO20000692A patent/NO20000692L/en not_active Application Discontinuation
Also Published As
Publication number | Publication date |
---|---|
EP1004016A1 (en) | 2000-05-31 |
AU738441B2 (en) | 2001-09-20 |
NO20000692D0 (en) | 2000-02-11 |
IL134456A0 (en) | 2001-04-30 |
CN1275200A (en) | 2000-11-29 |
EP1496350A3 (en) | 2005-06-22 |
MY120544A (en) | 2005-11-30 |
NZ502821A (en) | 2002-08-28 |
EP1496350A2 (en) | 2005-01-12 |
KR20010022896A (en) | 2001-03-26 |
WO1999009395A1 (en) | 1999-02-25 |
JP2001516019A (en) | 2001-09-25 |
CA2299285A1 (en) | 1999-02-25 |
CN1107861C (en) | 2003-05-07 |
BR9811938A (en) | 2000-09-05 |
US6115673A (en) | 2000-09-05 |
NO20000692L (en) | 2000-04-06 |
TWI226930B (en) | 2005-01-21 |
KR100389590B1 (en) | 2003-06-27 |
AU8599298A (en) | 1999-03-08 |
AR013417A1 (en) | 2000-12-27 |
TR200000402T2 (en) | 2000-08-21 |
IL134456A (en) | 2003-11-23 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CA2299285C (en) | Method and apparatus for generating basis sets for use in spectroscopic analysis | |
Hall et al. | Near-infrared spectrophotometry: a new dimension in clinical chemistry | |
US7620674B2 (en) | Method and apparatus for enhanced estimation of an analyte property through multiple region transformation | |
US7640140B2 (en) | Method of processing noninvasive spectra | |
Cuadrado et al. | Comparison and joint use of near infrared spectroscopy and Fourier transform mid infrared spectroscopy for the determination of wine parameters | |
EP0869348B1 (en) | Method of determining a glucose concentration in a target by using near-infrared spectroscopy | |
US5750994A (en) | Positive correlation filter systems and methods of use thereof | |
US6441388B1 (en) | Methods and apparatus for spectroscopic calibration model transfer | |
US7098037B2 (en) | Accommodating subject and instrument variations in spectroscopic determinations | |
US20060017923A1 (en) | Analyte filter method and apparatus | |
US6064897A (en) | Sensor utilizing Raman spectroscopy for non-invasive monitoring of analytes in biological fluid and method of use | |
EP1442699A1 (en) | A method of characterizing spectrometer instruments and providing calibration models to compensate for instrument variation | |
NZ331158A (en) | Method and apparatus for multi-spectral analysis in noninvasive nir spectroscopy | |
US6615151B1 (en) | Method for creating spectral instrument variation tolerance in calibration algorithms | |
McMurdy et al. | Raman spectroscopy-based creatinine measurement in urine samples from a multipatient population | |
JP4329360B2 (en) | Glucose concentration determination device | |
Boysworth et al. | Aspects of multivariate calibration applied to near-infrared spectroscopy | |
Alrezj et al. | Digital bandstop filtering in the quantitative analysis of glucose from near‐infrared and midinfrared spectra | |
Venkatesan et al. | A comparative study of principal component regression and partial least squares regression with application to FTIR diabetes data | |
Du et al. | Improvement of the partial least squares model performance for oral glucose intake experiments by inside mean centering and inside multiplicative signal correction | |
Li et al. | Comparison of performance of partial least squares regression, secured principal component regression, and modified secured principal component regression for determination of human serum albumin, γ-globulin and glucose in buffer solutions and in vivo blood glucose quantification by near-infrared spectroscopy | |
Yao et al. | Wavelength selection method based on absorbance value optimization to near-infrared spectroscopic analysis | |
Ham et al. | Multivariate determination of glucose using NIR spectra of human blood serum | |
MXPA00001560A (en) | Method and apparatus for generating basis sets for use in spectroscopic analysis | |
Shih | Quantitative biological Raman spectroscopy for non-invasive blood analysis |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
EEER | Examination request | ||
MKLA | Lapsed |