Method and Apparatus for Noninvasively Estimating a Property of an Animal Body Analyte from Spectral Data
A method and apparatus for calibration development using clustering is disclosed. More particularly, the invention relates to subsequent calibration development using clusters that are individually interference compensated and to subsequent estimation. Estimation of analyte property values from data, such as noninvasive spectra, is improved by a calibration method that uses clusters that are individually interference-compensated.
1 . A method for generating a model for noninvasive estimation of an analyte property of blood or tissue from spectral data, comprising the steps of:
clustering the spectral data into a plurality of clusters;
removing interference from each of said clusters to yield respective interference-reduced clusters;
aggregating said interference-reduced clusters into a calibration data set; and
generating said model from said calibration data set.
2 . The method of claim 1 , wherein said step of clustering comprises the step of unsupervised clustering.
3 . The method of claim 2 , wherein the clustering step comprises the step of establishing the clusters with a cluster spectral variation within the clusters that is less than spectral variation between said clusters.
4 . The method of claim 3 , wherein the clustering step comprises the step of clustering the spectral data into a plurality of clusters in accordance with predetermined parameters, and further comprising the step of:
identifying as an outlier spectral data which fails said predetermined parameters.
5 . The method of claim 3 , further comprising the step of:
developing a basis set for at least one of said clusters, wherein said basis set represents a common interference within said one cluster.
6 . The method of claim 5 , wherein said step of interference removing further comprises the step of applying a basis set to at least one of said clusters.
7 . The method of claim 3 , further comprising the step of:
implementing said model to estimate said analyte property.
8 . The method of claim 7 , further comprising the step of:
collecting a noninvasive spectrum, wherein said step of implementing further comprises the step of processing said noninvasive spectrum to yield said analyte property.
9 . The method of claim 8 , wherein said analyte property comprises a glucose concentration.
10 . The method of claim 9 , further comprising the step of:
prior to said step of implementing, removing interference from said noninvasive spectrum according to a predefined interference basis.
11 . The method of claim 10 , where said interference basis comprises a compact representation of a source of spectral variance within said spectral data related to a chemical structure or a physical phenomenon of the blood/tissue.
12 . The method of claim 10 , wherein said step of removing further comprises the step of subtracting a rotation, said rotation representing an orthogonal measurement relative to said interference basis.
13 . The method of claim 8 , further comprising the step of:
prior to said step of implementing, assigning said noninvasive spectrum to one of said at least two clusters through a classification step, wherein said classification determines closest proximity of said noninvasive spectrum to each of said at least two clusters.
14 . The method of claim 13 , wherein said classification uses a distance metric.
15 . The method of claim 3 , wherein said step of removing interference comprises the step of using a feature extracted from said spectral data in forming said at least two clusters.
16 . The method of claim 15 , wherein said feature comprises a compact representation of a source of spectral variance within said spectral data related to a chemical structure or a physical phenomenon of the blood/tissue.
17 . The method of claim 16 , further comprising the step of:
determining an outlier, wherein said outlier comprises a spectrum wherein said feature is not readily fit or modeled.
18 . The method of claim 16 , further comprising the step of:
separating said at least two clusters based upon said feature, wherein said feature provides a distinguishing characteristic between said at least two clusters.
19 . The method of claim 16 , further comprising the step of:
identifying homogeneity within a cluster based upon said feature.
20 . The method of claim 1 , further comprising the step of:
forming a basis set of locally represented interference for each of said at least two clusters, wherein said step of removing interference comprises the step of removing said represented interference.
21 . The method of claim 2 , wherein said unsupervised method comprises a nonlinear model.
22 . The method of claim 21 , wherein said nonlinear model comprises a neural network.
23 . The method of claim 2 , wherein said unsupervised method represents natural structure of said spectral data to define said interference.
24 . The method of claim 2 , wherein said unsupervised method reduces a nonlinear span of said spectral data in a calibration space.
25 . The method of claim 2 , wherein said unsupervised method updates independently of an expert system.
26 . An apparatus for noninvasive analyte property estimation of a human subject having skin and a sample site, comprising:
a source of light energy;
a detector of light energy for acquiring spectral data;
a sample interface, the source and detector being disposed therein; and
an estimation component for estimating the analyte property from the spectral data, the estimation component comprising a calibration model component having an architecture of at least two clustered spectral data sets and means for removing separate cluster related interference from each of said clustered spectral data sets.
27 . The apparatus of claim 26 , further comprising:
a tissue stabilizer for reducing skin movement at the sample site.
28 . The apparatus of claim 27 , wherein said tissue stabilizer is adapted to contact the skin of the subject during use.
29 . The apparatus of claim 28 , wherein said tissue stabilizer is adapted to circumferentially surround and contact the sample site during use.
30 . The apparatus of claim 29 , wherein said tissue stabilizer is adapted to reduce stress and strain at the sample site.
31 . The apparatus of claim 27 , wherein said tissue stabilizer is adapted to contact the human subject no less than one-half inch from the sample site.
32 . The apparatus of claim 31 , wherein said tissue stabilizer is adapted to contact the human subject no less than one inch from the sample site.
33 . The apparatus of claim 28 , wherein said tissue stabilizer comprises a curved surface approximating curvature about the sample site.
34 . The apparatus of claim 33 , wherein said tissue stabilizer is adapted to contact the human subject no less than one-half inch from the sample site.
35 . The apparatus of claim 33 , wherein said tissue stabilizer is adapted to contact the human subject no less than one-half inch from the outer surface of a sample probe of said analyzer.
36 . The apparatus of claim 35 , wherein said analyzer is foldable into a carrying case.
37 . The apparatus of claim 36 , wherein at least one outer surface of said analyzer comprises at least one outer surface of said carrying case when said analyzer is in a folded state, wherein said carrying case further comprises a handle.
38 . An apparatus for noninvasive analyte property estimation of a subject having skin and a sample site, comprising:
a source of light energy;
a detector of light energy for acquiring spectral data;
a sample interface, the source and detector being disposed therein; and
an estimation component comprising a processor and memory, the memory being coupled to the processor and comprising a plurality of processor instructions for:
controlling the acquisition of the spectral data;
clustering the spectral data into a plurality of clusters;
removing interference from each of the clusters to yield respective interference-reduced clusters;
aggregating the interference-reduced clusters into a calibration data set; and
generating a model for the analyte property estimation from the calibration data set.