US20210338115A9 - System and method for obtaining blood glucose concentration using temporal independent component analysis (ica) - Google Patents
System and method for obtaining blood glucose concentration using temporal independent component analysis (ica) Download PDFInfo
- Publication number
- US20210338115A9 US20210338115A9 US16/183,325 US201816183325A US2021338115A9 US 20210338115 A9 US20210338115 A9 US 20210338115A9 US 201816183325 A US201816183325 A US 201816183325A US 2021338115 A9 US2021338115 A9 US 2021338115A9
- Authority
- US
- United States
- Prior art keywords
- spectra
- spectrum
- nir
- orthogonal pure
- nir spectra
- 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.)
- Granted
Links
Images
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/145—Measuring characteristics of blood in vivo, e.g. gas concentration, pH value; Measuring characteristics of body fluids or tissues, e.g. interstitial fluid, cerebral tissue
- A61B5/14532—Measuring characteristics of blood in vivo, e.g. gas concentration, pH value; Measuring characteristics of body fluids or tissues, e.g. interstitial fluid, cerebral tissue for measuring glucose, e.g. by tissue impedance measurement
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N33/00—Investigating or analysing materials by specific methods not covered by groups G01N1/00 - G01N31/00
- G01N33/48—Biological material, e.g. blood, urine; Haemocytometers
- G01N33/483—Physical analysis of biological material
- G01N33/487—Physical analysis of biological material of liquid biological material
- G01N33/49—Blood
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/145—Measuring characteristics of blood in vivo, e.g. gas concentration, pH value; Measuring characteristics of body fluids or tissues, e.g. interstitial fluid, cerebral tissue
- A61B5/1455—Measuring characteristics of blood in vivo, e.g. gas concentration, pH value; Measuring characteristics of body fluids or tissues, e.g. interstitial fluid, cerebral tissue using optical sensors, e.g. spectral photometrical oximeters
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7235—Details of waveform analysis
- A61B5/725—Details of waveform analysis using specific filters therefor, e.g. Kalman or adaptive filters
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7235—Details of waveform analysis
- A61B5/7253—Details of waveform analysis characterised by using transforms
- A61B5/7257—Details of waveform analysis characterised by using transforms using Fourier transforms
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7235—Details of waveform analysis
- A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
- A61B5/7267—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems involving training the classification device
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G06N99/005—
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H10/00—ICT specially adapted for the handling or processing of patient-related medical or healthcare data
- G16H10/40—ICT specially adapted for the handling or processing of patient-related medical or healthcare data for data related to laboratory analysis, e.g. patient specimen analysis
Definitions
- Systems, devices, and methods consistent with exemplary embodiments relate to glucose monitoring, and more particularly relates to obtaining blood glucose concentration using temporal independent component analysis (ICA).
- ICA temporal independent component analysis
- Glucose monitoring is used for testing level of glucose concentration in blood, and can be performed either invasively or non-invasively.
- invasive method skin of a person is pierced to obtain blood sample for testing, and in the non-invasive method, collection of blood sample is not required for obtaining the glucose concentration.
- non-invasive glucose monitoring include Mid Infrared (Mid IR), Near Infrared (NIR), and Raman spectroscopy.
- the NIR method is commonly used for continuous glucose monitoring, in which the IR waves are made to pass through the skin and absorption of the IR waves by the subcutaneous portion of skin is used in determining the glucose level.
- the absorption of the wave by the sample is defined by BEER Lambert law:
- E absorption coefficient
- C concentration of component in sample
- d penetration depth
- the NIR spectrum of the skin is composed of absorption of the IR waves by several components such as water, fat (or cholesterol), protein (e.g., collagen and keratin), amino acids, elastin and glucose. Therefore, the NIR spectrum of the skin can be obtained as the following equation:
- a NIR A Water +A Cholesterol +A Collagen +A Keratin +A Elastin +A Acid +A Glucose
- Related art methods for monitoring glucose levels using an NIR spectrum includes a non-contact analysis method of solid samples in NIR Diffusion reflectance measurement, wherein Independent Component Analysis is performed on the mixed spectra to separate the mixed spectra into the pure analyte spectra and their concentration profiles.
- the related art method also uses a scatter correction to remove the non-linear effects from the measured spectra.
- the related art method does not address the issue of instrumental/environmental residual drift reducing SNR in an actual scenario.
- the related art method uses independent component analysis (ICA) algorithm that assumes statistically independent source signals and non-zero kurtosis.
- ICA independent component analysis
- Another related art method for monitoring glucose levels is based on an approach for analysis of Near Infrared (NIR) data using Independent Component Analysis (ICA), wherein a Blind Source Separation is performed on a non-analyte mixture to identify the concentration of individual mixture.
- NIR Near Infrared
- ICA Independent Component Analysis
- the method uses a mixture made from starch, water and protein for experimentation.
- the ICA algorithm used in the related art method assumes interdependence of source signals and that the concentration of each components of the mixture is not time varying.
- the concentration of components may change with time.
- the related art method does not address the challenge of concentration of one component (e.g., glucose) being very low compared to that of other components (e.g., water and protein), which significantly degrades the accuracy of human skin NIR analysis.
- Another related art method for monitoring glucose levels includes a method for measuring blood glucose using only the portion of the IR spectrum which contains the NIR water absorption peaks, wherein the related art method uses electromagnetic (EM) radiation of a wavelength transmitted through the skin to the measurement region, for example, a blood vessel. The collected light is analyzed and compared against a stored reference calibration curve to calculate blood glucose concentration.
- EM electromagnetic
- the related art method assumes that the background interference is common for an entire range of the near infrared region.
- the related art method uses a reference calibration curve which varies from person to person and hence universality is not guaranteed. It is assumed in the related art method that all constituents of the human skin is known and well understood in the NIR spectrum. This is not true for the case of human skin. This will affect the accuracy in determination of the glucose level.
- One or more exemplary embodiments provide a method for obtaining blood glucose concentration using temporal independent component analysis (ICA).
- ICA temporal independent component analysis
- a method for obtaining blood glucose concentration using near infrared spectroscopy (NIR) data including: obtaining, by an independent component analysis (ICA) temporal module, orthogonal pure spectra from human NIR spectra; performing, by a processing module, one or more preproces sings and drift removal on the human NIR spectra and the orthogonal pure spectra to obtain preprocessed spectra; and obtaining, by a regression block, the blood glucose concentration from the preprocessed spectra.
- ICA independent component analysis
- the obtaining the orthogonal pure spectra may include: receiving, by a pre-data whitening unit, the human NIR spectra; obtaining, by the pre-data whitening unit, data whitened NIR spectra based on performing transformation on the human NIR spectra; calculating, by an iterative processing unit, an orthogonal pure spectrum from the data whitened NIR spectra; calculating new deflated NIR spectra to be transmitted to the iterative processing unit, to compute a new orthogonal pure spectrum based on removal of an effect of the previously calculated orthogonal pure spectrum; and combining one or more computed orthogonal pure spectrums to obtain the orthogonal pure spectra.
- the obtaining the data whitened NIR spectra may include: calculating Eigen vectors of the human NIR spectra using a singular value decomposition; and applying a whitening transformation using the Eigen vectors on the human NIR spectra to obtain the data whitened NIR spectra.
- the calculating the orthogonal pure spectrum may include: computing, by a single processing unit, an estimate spectrum, based on the data whitened NIR spectra and a residual error; and reiterating the computing of the estimate spectrum until convergence of learning parameters is achieved to obtain the orthogonal pure spectrum.
- the computing the estimate spectrum may include: randomly initializing the learning parameters, the learning parameters including a weight vector and a bias vector; obtaining the estimate spectrum based on the weight vector and the bias vector; and computing source statistics for the estimate spectrum, the source statistics including Cross correlation and Covariance matrix.
- the reiterating may include: calculating updated values of the weight vector and the bias vector based on the source statistics of the estimate spectrum; calculating an updated estimate spectrum based on the updated values of the weight vector and the bias vector; determining the updated estimate spectrum as the orthogonal pure spectrum in response to the convergence being achieved for the weight vector; and reiterating the computing of the estimate spectrum in response to the convergence not being achieved for the weight vector.
- the calculating the new deflated NIR spectra may include: deflating the data whitened NIR spectra based on the orthogonal pure spectrum obtained from the iterative processing unit; determining whether a certain number of orthogonal pure spectrums are obtained; and transmitting the deflated NIR spectra to the iterative processing unit to obtain a new orthogonal pure spectrum in response to the certain number of the orthogonal pure spectrums are not obtained.
- the obtaining the preprocessed spectra may include: performing an extended multiplicative scatter correction (EMSC) method on the human NIR spectra and the orthogonal pure spectra; performing, by using a Fast Fourier Transform (FFT) block, a filtering method to obtain filtered spectra; and performing, on the filtered spectra, drift removal to obtain the preprocessed spectra.
- ESC extended multiplicative scatter correction
- FFT Fast Fourier Transform
- the obtaining the filtered spectra may include: subsequent to performing the EMSC method, performing a Fourier domain filtering on the human NIR spectra to reduce noise on the human NIR spectra by using a Hanning Window; and removing drift by differentiating, with respect to a wavelength, Fourier domain filtered spectra, to obtain the filtered spectra.
- the obtaining the blood glucose concentration may include: extracting, by a feature extraction block, one or more features from the preprocessed spectra; obtaining a training data set and a validation data set from the one or more features; and obtaining the blood glucose concentration by performing regression on the training data set and the validation data set.
- a system for obtaining blood glucose concentration using near infrared spectroscopy (NIR) data including: an independent component analysis (ICA) temporal module configured to obtain orthogonal pure spectra from human NIR spectra; a processing module configured to perform one or more preproces sings and drift removal on the human NIR spectra and the orthogonal pure spectra to obtain preprocessed spectra; and a regression block configured to obtain the blood glucose concentration from the preprocessed spectra.
- ICA independent component analysis
- the ICA temporal module may include: a pre-data whitening unit configured to: receive the human NIR spectra; and obtain data whitened NIR spectra based on performing transformation on the human NIR spectra; an iterative processing unit configured to calculate an orthogonal pure spectrum from the data whitened NIR spectra; a deflation module configured to calculate new deflated NIR spectra to be transmitted to the iterative processing unit, to compute a new orthogonal pure spectrum based on removal of an effect of the previously calculated orthogonal pure spectrum; and a learning algorithm unit configured to combine one or more computed orthogonal pure spectrums to obtain the orthogonal pure spectra.
- a pre-data whitening unit configured to: receive the human NIR spectra; and obtain data whitened NIR spectra based on performing transformation on the human NIR spectra
- an iterative processing unit configured to calculate an orthogonal pure spectrum from the data whitened NIR spectra
- the pre-data whitening unit may obtain Eigen vectors of the human NIR spectra using a singular value decomposition, and apply a whitening transformation using the Eigen vectors on the human NIR spectra to obtain the data whitened NIR spectra.
- the learning algorithm unit may compute, by a single processing unit included in the learning algorithm unit, an estimate spectrum, based on the data whitened NIR spectra and a residual error.
- the learning algorithm unit may compute the estimate spectrum by performing: randomly initializing learning parameters, the learning parameters including a weight vector and a bias vector; obtaining the estimate spectrum based on the weight vector and the bias vector; and computing a source statistics for the estimate spectrum, the source statistics including Cross correlation and Covariance matrix; and reiterating computing of the estimate spectrum until convergence of the learning parameters is achieved to obtain the orthogonal pure spectrum.
- the reiterating may include: calculating updated values of the weight vector and the bias vector based on the source statistics of the estimate spectrum; calculating an updated estimate spectrum based on the updated values of the weight vector and the bias vector; determining the updated estimate spectrum as the orthogonal pure spectrum in response to the convergence being achieved for the weight vector; and reiterating the updated values of the weight vector and the bias vector to the single processing unit in response to the convergence not being achieved for the weight vector.
- the deflation module may deflate the data whitened NIR spectra based on the orthogonal pure spectrum obtained from the iterative processing unit; determine whether a certain number of orthogonal pure spectrums are obtained, and transmit the deflated NIR spectra to the iterative processing unit to obtain a new orthogonal pure spectrum in response to the certain number of the orthogonal pure spectrums not being obtained.
- the processing module may include: an extended multiplicative scatter correction (EMSC) module configured to perform an extended multiplicative scatter correction (EMSC) method on the human NIR spectra and the orthogonal pure spectra; a Fast Fourier Transform (FFT) block configured to perform a filtering method to obtain filtered spectra; and a drift removal module configured to perform drift removal on the filtered spectra to obtain the preprocessed spectra.
- ESC extended multiplicative scatter correction
- FFT Fast Fourier Transform
- the regression block may include: a feature extraction block configured to extract one or more features from the preprocessed spectra; a separation block configured to obtain a training data set and a validation data set from the one or more features; and a regression model identifier block configured to obtain the blood glucose concentration based on performing regression on the training data set and the validation data set.
- FIG. 1 illustrates a schematic flow diagram illustrating a method for obtaining blood glucose concentration using temporal independent component analysis (ICA), according to an exemplary embodiment
- FIG. 2 is a schematic block diagram illustrating units for obtaining blood glucose concentration using temporal independent component analysis (ICA), according to an exemplary embodiment
- FIG. 3 is a schematic diagram illustrating single processing unit for extracting spectra, according to an exemplary embodiment.
- the disclosure describes a method for obtaining blood glucose concentration using temporal independent component analysis (ICA).
- the method comprises an Independent Component Analysis (ICA) temporal module to obtain an orthogonal pure spectrum from human Near Infrared Spectroscopy (NIR) spectra.
- ICA Independent Component Analysis
- NIR Near Infrared Spectroscopy
- obtaining the orthogonal pure spectrum comprises receiving, by a pre-data whitening unit, human NIR spectra.
- the pre-data whitening unit provides data whitened NIR spectra after applying a transformation on the human NIR spectra.
- obtaining the data whitened NIR spectra comprises calculating Eigen vectors of the input NIR spectra using singular value decomposition method. Further, a whitening transformation is applied using the Eigen vectors on the input NIR spectra to obtain the data whitened NIR spectra.
- obtaining the orthogonal pure spectra comprises calculating, by an iterative processing unit, an orthogonal pure spectrum from the data whitened NIR spectra.
- the method for obtaining blood glucose concentration comprises calculating the orthogonal pure spectra, which comprises computing an estimate spectrum by a single processing unit, based on the data whitened NIR spectra and a residual error, while computing the estimate spectrum.
- the computing an estimate spectrum by the single processing unit comprises randomly initializing the learning parameters, which comprises a weight vector and a bias vector, obtaining an estimate spectrum based on the weight vector and the bias vector, and computing source statistics for the estimate spectrum, the source statistics comprising a Cross correlation and Covariance matrix.
- the method for obtaining blood glucose concentration comprises combining all computed orthogonal pure spectrum to obtain an orthogonal pure spectra.
- calculating individual orthogonal pure spectra comprises reiterating over the estimate spectrum until convergence of learning parameters is achieved to obtain an orthogonal pure spectrum
- the reiterating over the estimate spectrum to obtain the orthogonal pure spectrum comprises calculating updated values of the weight vector and the bias vector based on the source statistics of the estimate spectrum, calculating the updated estimate spectrum based on the updated values of the weight vector and the bias vector, assigning the updated estimate spectrum values as the orthogonal pure spectrum in response to convergence being achieved for the weight vector, and reiterating the above operations in response to convergence not being reached.
- the method of obtaining a pure spectrum comprises calculating new deflated NIR spectra to be transmitted back to the iterative processing unit, to compute a new orthogonal pure spectrum, after removing the effect of the orthogonal pure spectrum previously calculated.
- calculating the new deflated NIR spectra to be transmitted back to the iterative processing unit comprises deflating the data whitened NIR spectra based on the orthogonal pure spectrum obtained from the iterative processing unit.
- the method for obtaining blood glucose concentration comprises checking if a certain number of the orthogonal pure spectrums are obtained.
- the method for obtaining blood glucose concentration comprises sending back the deflated NIR spectra to the iterative processing unit to obtain a new orthogonal pure spectrum if the certain number of the orthogonal pure spectrums are not obtained.
- the method for obtaining blood glucose concentration is based on a processing module applying one or more preproces sings and drift removal techniques on the human NIR spectra and the orthogonal pure spectra to obtain a preprocessed spectra.
- the ICA temporal module filters the human NIR spectra and obtains the orthogonal pure spectra.
- the orthogonal pure spectra along with the human NIR spectra are transmitted to the processing module that receives the orthogonal pure spectra and human NIR spectra, and applies one or more processing and drift removal techniques to the human NIR spectra and the orthogonal pure spectra to obtain preprocessed spectra.
- the processing module Upon applying one or more preprocessings and drift removal techniques to the human NIR spectra and the orthogonal pure spectra components, the processing module obtains the preprocessed spectra.
- obtaining the preprocessed spectra comprises applying an extended multiplicative scatter correction (EMSC) method on the human NIR spectra and the orthogonal pure spectra.
- the method for obtaining blood glucose concentration comprises applying, by means of a Fast Fourier Transform (FFT) block, filtering methods to obtain filtered spectra, wherein obtaining the filtered spectra comprises applying a Fourier domain filtering on the human NIR spectra after performing the EMSC method to reduce the impact of noise on the human NIR spectra by using a Hanning Window, and differentiating, with respect to a wavelength, the Fourier domain filtered spectra to remove the impact of drift, which is constant with respect to the wavelength to obtain the filtered spectra.
- the method for obtaining blood glucose concentration comprises applying, on the filtered spectra, drift removal techniques to obtain the preprocessed spectra.
- the method for obtaining blood glucose concentration comprises a regression block obtaining a glucose concentration from the preprocessed spectra.
- the processing module transmits the preprocessed spectra to the regression block.
- the regression block receives the preprocessed spectra and provides the data to the feature extraction block, wherein the feature extraction block extracts one or more features from the obtained preprocessed spectra.
- the method for obtaining blood glucose concentration comprises obtaining a training data set and a validation data set from the feature. Further, the method for obtaining blood glucose concentration comprises obtaining the glucose concentration upon performing regression on the training data set and the validation data set.
- FIG. 1 illustrates a schematic flow diagram 100 illustrating a method for obtaining blood glucose concentration using temporal independent component analysis (ICA), according to an exemplary embodiment.
- ICA temporal independent component analysis
- an Independent Component Analysis (ICA) temporal module obtains orthogonal pure spectra from human NIR spectra.
- a processing module applies one or more preprocessings and drift removal techniques on the human NIR spectra and the orthogonal pure spectra to obtain preprocessed spectra.
- a regression block obtains a glucose concentration from the preprocessed spectra.
- FIG. 2 is a schematic block diagram of a user device (or an apparatus for obtaining blood glucose concentration) 200 illustrating components for obtaining blood glucose concentration using temporal independent component analysis (ICA), according to an exemplary embodiment.
- ICA temporal independent component analysis
- the user device 200 includes at least one processor 201 including an independent component analysis (ICA) temporal block 202 , a processing block 204 , and a regression block 206 .
- the ICA temporal block 202 comprises a pre-data whitening unit 208 , a learning algorithm unit 210 , and a deflation module 212 .
- the processing block 204 comprises an extended multiplicative scatter correction (EMSC) module 214 , a Fast Fourier Transform (FFT) block 216 , and a drift removal module 218 .
- EMC extended multiplicative scatter correction
- FFT Fast Fourier Transform
- the regression block 206 comprises a feature extraction block 220 , a separation block 222 , a training block 224 , and a regression model identifier block 226 .
- ICA Independent Component Analysis
- the ICA defines a generative model for the observed multivariate data, which is typically given as a large database of samples.
- the ICA is a case of Blind Source Separation, wherein the ICA is a statistical and computational technique for revealing hidden factors that underlie sets of random variables, measurements, or signals.
- the ICA is used to transform observed data x, using linear transformation W into maximally independent components y as
- ICA techniques are based on a principle of assuming non Gaussianity and statistical independence of source signals.
- the requirement of assuming non Gaussianity and statistical independence of source signals does not permit its use to many real life scenarios where the source signals (y) are commonly dependent on each other.
- CGM non-invasive continuous glucose monitoring
- the pure spectra of components in skin such as, but not limited to, glucose, water, fat, collagen, keratin, acid, and the like have a high correlation.
- a mean normalized spectrum extracted for one or more components is not capable of appropriately capturing the peaks of the pure spectra of components.
- the peaks of the pure spectra of components cannot be extracted from NIR spectra through conventional ICA techniques.
- the ICA temporal block 202 overcomes the above discussed problem, wherein the ICA temporal block 202 works on a batch learning method for sequential blind source extraction. Further, the ICA temporal block 202 works on signals obtained from non-additive white (i.i.d.) temporally correlated sources.
- the ICA temporal block 202 comprises the pre-data whitening unit 208 , the learning algorithm unit 210 , and the deflation module 212 .
- the pre-data whitening unit 208 receives human NIR spectra.
- the pre-data whitening unit 208 further provides data whitened NIR spectra after removing error from the human NIR spectra. Further, the pre-data whitening unit 208 calculates Eigen vectors of the human NIR spectra using singular value decomposition method. Further, a whitening transformation is applied by the pre-data whitening unit 208 using the Eigen vectors of the human NIR spectra to find the data whitened NIR spectra.
- the pre-data whitening unit 208 provides the data whitened NIR spectra to the learning algorithm unit 210 that calculates, using an iterative processing unit, an orthogonal pure spectrum from the data whitened NIR spectra.
- the learning algorithm unit 210 comprises a single processing unit, wherein the learning algorithm unit 210 computes an estimate spectrum based on the data whitened NIR spectra and a residual error while computing the estimate spectrum, wherein computing the estimate spectra from the single processing unit comprises randomly initializing the learning parameters, the learning parameters comprising a weight vector and a bias vector. Further, the learning algorithm unit 210 obtains an estimate spectrum based on the weight vector and the bias vector, and computes source statistics for the estimate spectrum comprising a Cross correlation and Covariance matrix.
- the learning algorithm unit 210 of the ICA temporal block 202 reiterates, over the estimate spectrum, operations of the single processing unit, to obtain orthogonal pure spectra, wherein the reiterating comprises calculating updated values of the weight vector and the bias vector based on the source statistics of the estimate spectrum, calculating the updated estimate spectrum based on the updated values of the weight vector and the bias vector, assigning the updated estimate spectrum as the orthogonal pure spectrum in response to convergence being achieved for the weight vector, and reiterating the operations of the single processing unit in response to convergence not being achieved.
- the learning algorithm unit 210 provides the orthogonal pure spectrum value to the deflation module 212 that calculates new deflated NIR spectra to be sent back to the iterative processing unit after removing the effect of the orthogonal pure spectra.
- the deflation module 212 deflates the data whitened NIR spectra based on the orthogonal pure spectrum obtained from the iterative processing unit. Further, the deflation module 212 checks whether a certain number of orthogonal pure spectrum are obtained. Further, the deflation module 212 transmits back the deflated NIR spectra to the iterative processing unit to obtain a new orthogonal pure spectrum in response to the required number of orthogonal pure spectra not being obtained.
- the ICA temporal block 202 does not assume statistical independence or non-zero kurtosis for the source signals (e.g., pure spectra), but only assumes different temporal structures for the pure spectra, which is true as they have different auto correlation.
- the ICA temporal block 202 uses a method based on second order statistics to compute orthogonal pure spectra, thus is computationally efficient than related art methods.
- A is an m ⁇ n unknown mixing matrix or concentrations
- s(k) is a vector of unknown pure spectra
- n(k) is an additive white (i.i.d.) noise vector.
- the human NIR spectra X is mean and standard deviation normalized to give X , wherein X is transmitted to a pre-data whitening block to calculate eigen vectors of the human NIR spectra and apply the whitening transformation, to obtain data whitened NIR spectra.
- the eigen vectors E and D can be calculated using:
- Equation (6) The whitening transformation to obtain the data whitened NIR spectra ⁇ tilde over (X) ⁇ is expressed in Equation (6).
- data whitening renders the covariance matrix of data whitened NIR spectra R xx to be equal to I n :
- each spectrum is calculated through Blind extraction technique. Assume the Data whitened NIR spectra to be x(k) and the first spectra y 1 (k) need to be extracted. Then, a single processing unit is described as:
- the outputs of the single processing unit y 1 (k) and ⁇ 1 (k) represent the extracted spectra, and the error after passing y 1 (k) by an FIR filter b 1 respectively.
- FIG. 3 is a schematic diagram illustrating single processing unit 300 for extracting spectra, according to an exemplary embodiment.
- the single processing unit 300 extracts one orthogonal pure spectrum from a plurality of orthogonal pure spectrums in the received data whitened NIR spectra.
- the single processing unit 300 estimates the optimal values of vectors w 1 and b 1 so as to extract spectra.
- a cost function J(w 1 , b 1 ) is defined as:
- the learning algorithm unit of ICA temporal block 202 minimizes cost function J(w 1 , b 1 ) to estimate each orthogonal pure spectrum.
- the deflation module 212 deflates the input to remove the effect of orthogonal pure spectrum derived in a previous operation from data whitened NIR spectra using Equation (15):
- Equation (16) can be simplified to:
- the processing block 204 of the user device 200 comprises the EMSC module 214 , the Fast Fourier Transform (FFT) block 216 , and the drift removal module 218 .
- the processing block 204 applies one or more preprocessings and drift removal techniques on the orthogonal pure spectra components to obtain preprocessed spectra.
- the processing block 204 receives the human NIR spectra and orthogonal pure spectra from the ICA temporal block 202 , and provides the data to the extended multiplicative scatter correction (EMSC) module 214 .
- the EMSC module 214 applies an extended multiplicative scatter correction (EMSC) method on the human NIR spectra, uses the orthogonal pure spectra and regress for their compositions in the NIR spectra.
- EMC extended multiplicative scatter correction
- Y any human NIR spectra comprising various pure spectra X 1 , X 2 , . . . , Xk for different blood components. Then, Y can be obtained using simple Linear Regression at any given wavelength as follows:
- a k are the strengths of blood component and a 0 is a DC component.
- the glucose spectra can be obtained by subtracting other components in the given spectra:
- the data can be further provided to the FFT block 216 , wherein FFT filtering methods are applied to reduce impact of noise on the spectra and obtain filtered spectra.
- FFT filtering methods a Fourier domain filtering is performed on the human NIR spectra after EMSC is applied to reduce the impact of noise on the spectra by using a Hanning Window of size N win , which is expressed in the following equations:
- the FFT block 216 comprises a differential block that acts as a supplement to the drift removal block to remove the impact of constant drift with respect to a wavelength.
- the differential block is denoted as
- the differential block differentiates the Fourier domain filtered spectra to remove the impact of drift, which is constant with respect to a wavelength, to obtain filtered spectra. Based on the filtered spectra obtained from the FFT block and the differential block, a correlation value for each feature index with respect to glucose without and with the FFT filtering and the differentiation by the differential block can be obtained. From the comparison, it can be observed that there is an increase in the correlation for the most of the indices after the FFT filtering and the differentiation.
- the data from the FFT block 216 can be provided to the drift removal module 218 that applies a drift removal method to obtain the preprocessed spectra by removing the effect of experimental/instrumental drift in the NIR spectra.
- drift removal method it is assumed that NIR spectra contain only linear drift.
- the user device 200 comprises the regression block 206 that performs regression on the processed data and calculates glucose output from the processed data.
- the regression block 206 comprises the feature extraction block 220 , the separation block 222 , the training block 224 , and the regression model identifier block 226 .
- the feature extraction block 220 receives the spectra obtained after data processing block as input and extracts one or more features associated with the processed data signal.
- the features are the wavelength spectrums which show consistent high correlations with the glucose concentration.
- the obtained features are provided to the separation block 222 that separates training data set and validation data set.
- the training data set is further provided to the training block 224 that receives the training data set and trains the regression model.
- the training block 224 uses a principal components regression (PCR) method for training the data set.
- PCR principal components regression
- the regression model identifier block 226 receives the trained data set from the training block 224 and validation data set from the separation block 222 , and obtains the glucose concentration upon performing regression on the training data set and the validation data set.
- At least one of the components, elements or units represented by a block in the drawings may be embodied as various numbers of hardware, software and/or firmware structures that execute respective functions described above, according to an exemplary embodiment.
- at least one of these components, elements or units may use a direct circuit structure, such as a memory, processing, logic, a look-up table, etc. that may execute the respective functions through controls of one or more microprocessors or other control apparatuses.
- at least one of these components, elements or units may be specifically embodied by a module, a program, or a part of code, which contains one or more executable instructions for performing specified logic functions.
- At least one of these components, elements or units may further include a processor such as a central processing unit (CPU) that performs the respective functions, a microprocessor, or the like.
- a processor such as a central processing unit (CPU) that performs the respective functions, a microprocessor, or the like.
- a bus is not illustrated in the above block diagrams, communication between the components, elements or units may be performed through the bus.
- Functional aspects of the above exemplary embodiments may be implemented in algorithms that execute on one or more processors.
- the components, elements or units represented by a block or processing steps may employ any number of related art techniques for electronics configuration, signal processing and/or control, data processing and the like..
- the “unit” or “module” used herein may be a hardware component, such as a processor or a circuit, and/or a software component that is executed by a hardware component such as a processor.
Landscapes
- Health & Medical Sciences (AREA)
- Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Physics & Mathematics (AREA)
- Biomedical Technology (AREA)
- Medical Informatics (AREA)
- General Health & Medical Sciences (AREA)
- Public Health (AREA)
- Pathology (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- Veterinary Medicine (AREA)
- Animal Behavior & Ethology (AREA)
- Surgery (AREA)
- Heart & Thoracic Surgery (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Mathematical Physics (AREA)
- Signal Processing (AREA)
- Hematology (AREA)
- Psychiatry (AREA)
- Chemical & Material Sciences (AREA)
- Physiology (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Software Systems (AREA)
- Theoretical Computer Science (AREA)
- Evolutionary Computation (AREA)
- Optics & Photonics (AREA)
- Immunology (AREA)
- Primary Health Care (AREA)
- Analytical Chemistry (AREA)
- Medicinal Chemistry (AREA)
- Ecology (AREA)
- Biochemistry (AREA)
- Urology & Nephrology (AREA)
- Food Science & Technology (AREA)
- Epidemiology (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
Abstract
Description
- This application claims priority from Indian Patent Application No. 201741042881, filed on Nov. 29, 2017 in the Indian Intellectual Property Office, and Korean Patent Application No. 10-2018-0075331, filed on Jun. 29, 2018 in the Korean Intellectual Property Office, the disclosures of which are hereby incorporated in their entireties by reference.
- Systems, devices, and methods consistent with exemplary embodiments relate to glucose monitoring, and more particularly relates to obtaining blood glucose concentration using temporal independent component analysis (ICA).
- Glucose monitoring is used for testing level of glucose concentration in blood, and can be performed either invasively or non-invasively. In the invasive method, skin of a person is pierced to obtain blood sample for testing, and in the non-invasive method, collection of blood sample is not required for obtaining the glucose concentration. Some of the typical methods used for non-invasive glucose monitoring include Mid Infrared (Mid IR), Near Infrared (NIR), and Raman spectroscopy. In recent years, the NIR method is commonly used for continuous glucose monitoring, in which the IR waves are made to pass through the skin and absorption of the IR waves by the subcutaneous portion of skin is used in determining the glucose level. The absorption of the wave by the sample is defined by BEER Lambert law:
-
- Where E is absorption coefficient, C is concentration of component in sample and d is penetration depth.
- If the sample is composed of different constituents having different coefficients (ϵ1, ϵ2, . . . ϵn) and concentrations (Cl, C2, . . . Cn), then overall absorption can be given as the following equation:
-
A=ϵ1 C 1 d+ϵ 2 C 2 d+ . . . +ϵ n C n d (2) - The NIR spectrum of the skin is composed of absorption of the IR waves by several components such as water, fat (or cholesterol), protein (e.g., collagen and keratin), amino acids, elastin and glucose. Therefore, the NIR spectrum of the skin can be obtained as the following equation:
-
A NIR =A Water +A Cholesterol +A Collagen +A Keratin +A Elastin +A Acid +A Glucose - Monitoring of glucose concentration non-invasively is very challenging as the concentration of glucose in blood is several orders lesser than that of other constituents and many times, the glucose information is buried under the noise and drift components of the NIR spectra. The orders of concentration of different constituents are shown in the below table:
-
Constituent Water Fat Protein Elastin/Acid Glucose Order of concentration(~) 10{circumflex over ( )}0 10{circumflex over ( )}−1 10{circumflex over ( )}−3 10{circumflex over ( )}−3 10{circumflex over ( )}−4 - Related art methods for monitoring glucose levels using an NIR spectrum includes a non-contact analysis method of solid samples in NIR Diffusion reflectance measurement, wherein Independent Component Analysis is performed on the mixed spectra to separate the mixed spectra into the pure analyte spectra and their concentration profiles. The related art method also uses a scatter correction to remove the non-linear effects from the measured spectra. However, the related art method does not address the issue of instrumental/environmental residual drift reducing SNR in an actual scenario. Further, the related art method uses independent component analysis (ICA) algorithm that assumes statistically independent source signals and non-zero kurtosis. However, in many of the cases, bio medical source signals are dependent on each other and have very low kurtosis value, which significantly degrades the accuracy in monitoring glucose levels.
- Another related art method for monitoring glucose levels is based on an approach for analysis of Near Infrared (NIR) data using Independent Component Analysis (ICA), wherein a Blind Source Separation is performed on a non-analyte mixture to identify the concentration of individual mixture. The method uses a mixture made from starch, water and protein for experimentation. However, the ICA algorithm used in the related art method assumes interdependence of source signals and that the concentration of each components of the mixture is not time varying. However, for a human body, the concentration of components may change with time. Further, the related art method does not address the challenge of concentration of one component (e.g., glucose) being very low compared to that of other components (e.g., water and protein), which significantly degrades the accuracy of human skin NIR analysis.
- Another related art method for monitoring glucose levels includes a method for measuring blood glucose using only the portion of the IR spectrum which contains the NIR water absorption peaks, wherein the related art method uses electromagnetic (EM) radiation of a wavelength transmitted through the skin to the measurement region, for example, a blood vessel. The collected light is analyzed and compared against a stored reference calibration curve to calculate blood glucose concentration. However, the related art method assumes that the background interference is common for an entire range of the near infrared region. Further, the related art method uses a reference calibration curve which varies from person to person and hence universality is not guaranteed. It is assumed in the related art method that all constituents of the human skin is known and well understood in the NIR spectrum. This is not true for the case of human skin. This will affect the accuracy in determination of the glucose level.
- Therefore, there is a need for a method for identifying the pure spectra of various skin components directly from the NIR spectra. Further, there is a need for a method for modifying the original ICA algorithm to obtain representations of orthogonal pure spectra even if the actual pure spectra are dependent on each other. Further, there is a need for a method, in which the obtained spectra is used by the temporal ICA algorithm to obtain the glucose concentration in the NIR spectra without the need for in vitro pure spectra. Further, there is need for a method for obtaining blood glucose concentration using temporal independent component analysis (ICA).
- One or more exemplary embodiments provide a method for obtaining blood glucose concentration using temporal independent component analysis (ICA).
- According to an aspect of an exemplary embodiment, there is provided a method for obtaining blood glucose concentration using near infrared spectroscopy (NIR) data, the method including: obtaining, by an independent component analysis (ICA) temporal module, orthogonal pure spectra from human NIR spectra; performing, by a processing module, one or more preproces sings and drift removal on the human NIR spectra and the orthogonal pure spectra to obtain preprocessed spectra; and obtaining, by a regression block, the blood glucose concentration from the preprocessed spectra.
- The obtaining the orthogonal pure spectra may include: receiving, by a pre-data whitening unit, the human NIR spectra; obtaining, by the pre-data whitening unit, data whitened NIR spectra based on performing transformation on the human NIR spectra; calculating, by an iterative processing unit, an orthogonal pure spectrum from the data whitened NIR spectra; calculating new deflated NIR spectra to be transmitted to the iterative processing unit, to compute a new orthogonal pure spectrum based on removal of an effect of the previously calculated orthogonal pure spectrum; and combining one or more computed orthogonal pure spectrums to obtain the orthogonal pure spectra.
- The obtaining the data whitened NIR spectra may include: calculating Eigen vectors of the human NIR spectra using a singular value decomposition; and applying a whitening transformation using the Eigen vectors on the human NIR spectra to obtain the data whitened NIR spectra.
- The calculating the orthogonal pure spectrum may include: computing, by a single processing unit, an estimate spectrum, based on the data whitened NIR spectra and a residual error; and reiterating the computing of the estimate spectrum until convergence of learning parameters is achieved to obtain the orthogonal pure spectrum.
- The computing the estimate spectrum may include: randomly initializing the learning parameters, the learning parameters including a weight vector and a bias vector; obtaining the estimate spectrum based on the weight vector and the bias vector; and computing source statistics for the estimate spectrum, the source statistics including Cross correlation and Covariance matrix.
- The reiterating may include: calculating updated values of the weight vector and the bias vector based on the source statistics of the estimate spectrum; calculating an updated estimate spectrum based on the updated values of the weight vector and the bias vector; determining the updated estimate spectrum as the orthogonal pure spectrum in response to the convergence being achieved for the weight vector; and reiterating the computing of the estimate spectrum in response to the convergence not being achieved for the weight vector.
- The calculating the new deflated NIR spectra may include: deflating the data whitened NIR spectra based on the orthogonal pure spectrum obtained from the iterative processing unit; determining whether a certain number of orthogonal pure spectrums are obtained; and transmitting the deflated NIR spectra to the iterative processing unit to obtain a new orthogonal pure spectrum in response to the certain number of the orthogonal pure spectrums are not obtained.
- The obtaining the preprocessed spectra may include: performing an extended multiplicative scatter correction (EMSC) method on the human NIR spectra and the orthogonal pure spectra; performing, by using a Fast Fourier Transform (FFT) block, a filtering method to obtain filtered spectra; and performing, on the filtered spectra, drift removal to obtain the preprocessed spectra.
- The obtaining the filtered spectra may include: subsequent to performing the EMSC method, performing a Fourier domain filtering on the human NIR spectra to reduce noise on the human NIR spectra by using a Hanning Window; and removing drift by differentiating, with respect to a wavelength, Fourier domain filtered spectra, to obtain the filtered spectra.
- The obtaining the blood glucose concentration may include: extracting, by a feature extraction block, one or more features from the preprocessed spectra; obtaining a training data set and a validation data set from the one or more features; and obtaining the blood glucose concentration by performing regression on the training data set and the validation data set.
- According to an aspect of another exemplary embodiment, there is provided a system for obtaining blood glucose concentration using near infrared spectroscopy (NIR) data, the system including: an independent component analysis (ICA) temporal module configured to obtain orthogonal pure spectra from human NIR spectra; a processing module configured to perform one or more preproces sings and drift removal on the human NIR spectra and the orthogonal pure spectra to obtain preprocessed spectra; and a regression block configured to obtain the blood glucose concentration from the preprocessed spectra.
- The ICA temporal module may include: a pre-data whitening unit configured to: receive the human NIR spectra; and obtain data whitened NIR spectra based on performing transformation on the human NIR spectra; an iterative processing unit configured to calculate an orthogonal pure spectrum from the data whitened NIR spectra; a deflation module configured to calculate new deflated NIR spectra to be transmitted to the iterative processing unit, to compute a new orthogonal pure spectrum based on removal of an effect of the previously calculated orthogonal pure spectrum; and a learning algorithm unit configured to combine one or more computed orthogonal pure spectrums to obtain the orthogonal pure spectra.
- The pre-data whitening unit may obtain Eigen vectors of the human NIR spectra using a singular value decomposition, and apply a whitening transformation using the Eigen vectors on the human NIR spectra to obtain the data whitened NIR spectra.
- The learning algorithm unit may compute, by a single processing unit included in the learning algorithm unit, an estimate spectrum, based on the data whitened NIR spectra and a residual error.
- The learning algorithm unit may compute the estimate spectrum by performing: randomly initializing learning parameters, the learning parameters including a weight vector and a bias vector; obtaining the estimate spectrum based on the weight vector and the bias vector; and computing a source statistics for the estimate spectrum, the source statistics including Cross correlation and Covariance matrix; and reiterating computing of the estimate spectrum until convergence of the learning parameters is achieved to obtain the orthogonal pure spectrum.
- The reiterating may include: calculating updated values of the weight vector and the bias vector based on the source statistics of the estimate spectrum; calculating an updated estimate spectrum based on the updated values of the weight vector and the bias vector; determining the updated estimate spectrum as the orthogonal pure spectrum in response to the convergence being achieved for the weight vector; and reiterating the updated values of the weight vector and the bias vector to the single processing unit in response to the convergence not being achieved for the weight vector.
- The deflation module may deflate the data whitened NIR spectra based on the orthogonal pure spectrum obtained from the iterative processing unit; determine whether a certain number of orthogonal pure spectrums are obtained, and transmit the deflated NIR spectra to the iterative processing unit to obtain a new orthogonal pure spectrum in response to the certain number of the orthogonal pure spectrums not being obtained.
- The processing module may include: an extended multiplicative scatter correction (EMSC) module configured to perform an extended multiplicative scatter correction (EMSC) method on the human NIR spectra and the orthogonal pure spectra; a Fast Fourier Transform (FFT) block configured to perform a filtering method to obtain filtered spectra; and a drift removal module configured to perform drift removal on the filtered spectra to obtain the preprocessed spectra.
- The regression block may include: a feature extraction block configured to extract one or more features from the preprocessed spectra; a separation block configured to obtain a training data set and a validation data set from the one or more features; and a regression model identifier block configured to obtain the blood glucose concentration based on performing regression on the training data set and the validation data set.
- The above and/or other aspects will become apparent and more readily appreciated by describing certain exemplary embodiments with reference to the accompanying drawings in which:
-
FIG. 1 illustrates a schematic flow diagram illustrating a method for obtaining blood glucose concentration using temporal independent component analysis (ICA), according to an exemplary embodiment; -
FIG. 2 is a schematic block diagram illustrating units for obtaining blood glucose concentration using temporal independent component analysis (ICA), according to an exemplary embodiment; and -
FIG. 3 is a schematic diagram illustrating single processing unit for extracting spectra, according to an exemplary embodiment. - In the following detailed description, exemplary embodiments will be described with reference to the accompanying drawings. Descriptions of well-known components and processing techniques are omitted so as to not unnecessarily obscure the embodiments herein. The examples used herein are intended merely to facilitate an understanding of ways in which the embodiments herein can be practiced and to further enable those of skill in the art to practice the embodiments herein. Accordingly, the examples should not be construed as limiting the scope of the embodiments herein. It should be understood that other embodiments may be utilized and that changes may be made without departing from the scope of the disclosure. The following detailed description is, therefore, not to be taken in a limiting sense, and the scope of the disclosure is defined only by the appended claims.
- The specification may refer to “an”, “one” or “some” embodiment(s) in several locations. This does not necessarily imply that each such reference is to the same embodiment(s), or that the feature only applies to a single embodiment. Single features of different embodiments may also be combined to provide other embodiments.
- As used herein, the singular forms “a”, “an” and “the” are intended to include the plural forms as well, unless expressly stated otherwise. It will be further understood that the terms “includes”, “comprises”, “including” and/or “comprising” when used in this specification, specify the presence of stated features, integers, s, operations, elements and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof. As used herein, the term “and/or” includes any and all combinations and arrangements of one or more of the associated listed items.
- Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure pertains. It will be further understood that terms, such as those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the relevant art and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein.
- The disclosure describes a method for obtaining blood glucose concentration using temporal independent component analysis (ICA). According to an exemplary embodiment, the method comprises an Independent Component Analysis (ICA) temporal module to obtain an orthogonal pure spectrum from human Near Infrared Spectroscopy (NIR) spectra. The human NIR spectra are received and provided to the ICA temporal module to be used to obtain the orthogonal pure spectra with respect to the human NIR spectra.
- In an exemplary embodiment, obtaining the orthogonal pure spectrum comprises receiving, by a pre-data whitening unit, human NIR spectra. The pre-data whitening unit provides data whitened NIR spectra after applying a transformation on the human NIR spectra. In another exemplary embodiment, obtaining the data whitened NIR spectra comprises calculating Eigen vectors of the input NIR spectra using singular value decomposition method. Further, a whitening transformation is applied using the Eigen vectors on the input NIR spectra to obtain the data whitened NIR spectra.
- Further, obtaining the orthogonal pure spectra comprises calculating, by an iterative processing unit, an orthogonal pure spectrum from the data whitened NIR spectra. In another exemplary embodiment, the method for obtaining blood glucose concentration comprises calculating the orthogonal pure spectra, which comprises computing an estimate spectrum by a single processing unit, based on the data whitened NIR spectra and a residual error, while computing the estimate spectrum. The computing an estimate spectrum by the single processing unit comprises randomly initializing the learning parameters, which comprises a weight vector and a bias vector, obtaining an estimate spectrum based on the weight vector and the bias vector, and computing source statistics for the estimate spectrum, the source statistics comprising a Cross correlation and Covariance matrix. Further, the method for obtaining blood glucose concentration comprises combining all computed orthogonal pure spectrum to obtain an orthogonal pure spectra.
- Further, calculating individual orthogonal pure spectra comprises reiterating over the estimate spectrum until convergence of learning parameters is achieved to obtain an orthogonal pure spectrum, wherein the reiterating over the estimate spectrum to obtain the orthogonal pure spectrum comprises calculating updated values of the weight vector and the bias vector based on the source statistics of the estimate spectrum, calculating the updated estimate spectrum based on the updated values of the weight vector and the bias vector, assigning the updated estimate spectrum values as the orthogonal pure spectrum in response to convergence being achieved for the weight vector, and reiterating the above operations in response to convergence not being reached.
- Further, the method of obtaining a pure spectrum comprises calculating new deflated NIR spectra to be transmitted back to the iterative processing unit, to compute a new orthogonal pure spectrum, after removing the effect of the orthogonal pure spectrum previously calculated. In an exemplary embodiment, calculating the new deflated NIR spectra to be transmitted back to the iterative processing unit comprises deflating the data whitened NIR spectra based on the orthogonal pure spectrum obtained from the iterative processing unit. Further, the method for obtaining blood glucose concentration comprises checking if a certain number of the orthogonal pure spectrums are obtained. Further, the method for obtaining blood glucose concentration comprises sending back the deflated NIR spectra to the iterative processing unit to obtain a new orthogonal pure spectrum if the certain number of the orthogonal pure spectrums are not obtained.
- Further, the method for obtaining blood glucose concentration is based on a processing module applying one or more preproces sings and drift removal techniques on the human NIR spectra and the orthogonal pure spectra to obtain a preprocessed spectra. The ICA temporal module filters the human NIR spectra and obtains the orthogonal pure spectra. The orthogonal pure spectra along with the human NIR spectra are transmitted to the processing module that receives the orthogonal pure spectra and human NIR spectra, and applies one or more processing and drift removal techniques to the human NIR spectra and the orthogonal pure spectra to obtain preprocessed spectra. Upon applying one or more preprocessings and drift removal techniques to the human NIR spectra and the orthogonal pure spectra components, the processing module obtains the preprocessed spectra.
- In an exemplary embodiment, obtaining the preprocessed spectra comprises applying an extended multiplicative scatter correction (EMSC) method on the human NIR spectra and the orthogonal pure spectra. Further, the method for obtaining blood glucose concentration comprises applying, by means of a Fast Fourier Transform (FFT) block, filtering methods to obtain filtered spectra, wherein obtaining the filtered spectra comprises applying a Fourier domain filtering on the human NIR spectra after performing the EMSC method to reduce the impact of noise on the human NIR spectra by using a Hanning Window, and differentiating, with respect to a wavelength, the Fourier domain filtered spectra to remove the impact of drift, which is constant with respect to the wavelength to obtain the filtered spectra. Further, the method for obtaining blood glucose concentration comprises applying, on the filtered spectra, drift removal techniques to obtain the preprocessed spectra.
- Further, the method for obtaining blood glucose concentration comprises a regression block obtaining a glucose concentration from the preprocessed spectra. The processing module transmits the preprocessed spectra to the regression block. The regression block receives the preprocessed spectra and provides the data to the feature extraction block, wherein the feature extraction block extracts one or more features from the obtained preprocessed spectra. Further, the method for obtaining blood glucose concentration comprises obtaining a training data set and a validation data set from the feature. Further, the method for obtaining blood glucose concentration comprises obtaining the glucose concentration upon performing regression on the training data set and the validation data set.
-
FIG. 1 illustrates a schematic flow diagram 100 illustrating a method for obtaining blood glucose concentration using temporal independent component analysis (ICA), according to an exemplary embodiment. - According to the flow diagram 100, at
operation 102, an Independent Component Analysis (ICA) temporal module obtains orthogonal pure spectra from human NIR spectra. Further, atoperation 104, a processing module applies one or more preprocessings and drift removal techniques on the human NIR spectra and the orthogonal pure spectra to obtain preprocessed spectra. Further, atoperation 106, a regression block obtains a glucose concentration from the preprocessed spectra. -
FIG. 2 is a schematic block diagram of a user device (or an apparatus for obtaining blood glucose concentration) 200 illustrating components for obtaining blood glucose concentration using temporal independent component analysis (ICA), according to an exemplary embodiment. - According to an exemplary embodiment shown in
FIG. 2 , theuser device 200 includes at least oneprocessor 201 including an independent component analysis (ICA)temporal block 202, aprocessing block 204, and aregression block 206. Further, the ICAtemporal block 202 comprises apre-data whitening unit 208, alearning algorithm unit 210, and adeflation module 212. Further, theprocessing block 204 comprises an extended multiplicative scatter correction (EMSC)module 214, a Fast Fourier Transform (FFT) block 216, and adrift removal module 218. Further, theregression block 206 comprises afeature extraction block 220, aseparation block 222, atraining block 224, and a regressionmodel identifier block 226. According to an exemplary embodiment, Independent Component Analysis (ICA) is a method used to separate multivariate signal into additive components. The ICA defines a generative model for the observed multivariate data, which is typically given as a large database of samples. The ICA is a case of Blind Source Separation, wherein the ICA is a statistical and computational technique for revealing hidden factors that underlie sets of random variables, measurements, or signals. - Consider an example, in which a random data vector x=(x1, x2, . . . xm)T is given as a weighted sum of independent components sp, p=1, . . . n, such that
-
x=Σ p a p *s p (3) - where ap are mixing weights.
- The ICA is used to transform observed data x, using linear transformation W into maximally independent components y as
-
y=W*x (4) - Conventional ICA techniques are based on a principle of assuming non Gaussianity and statistical independence of source signals. The requirement of assuming non Gaussianity and statistical independence of source signals does not permit its use to many real life scenarios where the source signals (y) are commonly dependent on each other. In non-invasive continuous glucose monitoring (CGM), the pure spectra of components in skin such as, but not limited to, glucose, water, fat, collagen, keratin, acid, and the like have a high correlation. Further, a mean normalized spectrum extracted for one or more components is not capable of appropriately capturing the peaks of the pure spectra of components. Thus, the peaks of the pure spectra of components cannot be extracted from NIR spectra through conventional ICA techniques.
- The ICA
temporal block 202 according to an exemplary embodiment overcomes the above discussed problem, wherein the ICAtemporal block 202 works on a batch learning method for sequential blind source extraction. Further, the ICAtemporal block 202 works on signals obtained from non-additive white (i.i.d.) temporally correlated sources. The ICAtemporal block 202 comprises thepre-data whitening unit 208, thelearning algorithm unit 210, and thedeflation module 212. - The
pre-data whitening unit 208 receives human NIR spectra. Thepre-data whitening unit 208 further provides data whitened NIR spectra after removing error from the human NIR spectra. Further, thepre-data whitening unit 208 calculates Eigen vectors of the human NIR spectra using singular value decomposition method. Further, a whitening transformation is applied by thepre-data whitening unit 208 using the Eigen vectors of the human NIR spectra to find the data whitened NIR spectra. - Further, the
pre-data whitening unit 208 provides the data whitened NIR spectra to thelearning algorithm unit 210 that calculates, using an iterative processing unit, an orthogonal pure spectrum from the data whitened NIR spectra. Thelearning algorithm unit 210 comprises a single processing unit, wherein thelearning algorithm unit 210 computes an estimate spectrum based on the data whitened NIR spectra and a residual error while computing the estimate spectrum, wherein computing the estimate spectra from the single processing unit comprises randomly initializing the learning parameters, the learning parameters comprising a weight vector and a bias vector. Further, thelearning algorithm unit 210 obtains an estimate spectrum based on the weight vector and the bias vector, and computes source statistics for the estimate spectrum comprising a Cross correlation and Covariance matrix. - Further, the
learning algorithm unit 210 of the ICAtemporal block 202 reiterates, over the estimate spectrum, operations of the single processing unit, to obtain orthogonal pure spectra, wherein the reiterating comprises calculating updated values of the weight vector and the bias vector based on the source statistics of the estimate spectrum, calculating the updated estimate spectrum based on the updated values of the weight vector and the bias vector, assigning the updated estimate spectrum as the orthogonal pure spectrum in response to convergence being achieved for the weight vector, and reiterating the operations of the single processing unit in response to convergence not being achieved. - Further, the
learning algorithm unit 210 provides the orthogonal pure spectrum value to thedeflation module 212 that calculates new deflated NIR spectra to be sent back to the iterative processing unit after removing the effect of the orthogonal pure spectra. Thedeflation module 212 deflates the data whitened NIR spectra based on the orthogonal pure spectrum obtained from the iterative processing unit. Further, thedeflation module 212 checks whether a certain number of orthogonal pure spectrum are obtained. Further, thedeflation module 212 transmits back the deflated NIR spectra to the iterative processing unit to obtain a new orthogonal pure spectrum in response to the required number of orthogonal pure spectra not being obtained. - According to an exemplary embodiment, the ICA
temporal block 202 does not assume statistical independence or non-zero kurtosis for the source signals (e.g., pure spectra), but only assumes different temporal structures for the pure spectra, which is true as they have different auto correlation. The ICAtemporal block 202 uses a method based on second order statistics to compute orthogonal pure spectra, thus is computationally efficient than related art methods. - For instance, consider that the ICA
temporal block 202 estimates each orthogonal pure spectrum from the human NIR spectra one at a time. Assume the human NIR spectra to be x(k)=[x1(k), x2(k), . . . xm(k)]T for each time instant k, wherein x can be represented as: -
x(k)=As(k)+n(k) (5) - where A is an m×n unknown mixing matrix or concentrations,
- s(k) is a vector of unknown pure spectra, and
- n(k) is an additive white (i.i.d.) noise vector.
- In the equation (5), it is required to determine maximally independent components/orthogonal pure spectrum, y(k), which can optimally represent s(k).
- Using the ICA temporal block method according to an exemplary embodiment, the human NIR spectra X is mean and standard deviation normalized to give
X , whereinX is transmitted to a pre-data whitening block to calculate eigen vectors of the human NIR spectra and apply the whitening transformation, to obtain data whitened NIR spectra. The eigen vectors E and D can be calculated using: -
[E, D]=eig(X ′*X ) - The whitening transformation to obtain the data whitened NIR spectra {tilde over (X)} is expressed in Equation (6).
-
{tilde over (X)}=ED½E·X (6) - Further, data whitening renders the covariance matrix of data whitened NIR spectra Rxx to be equal to In:
- This helps to ensure that data whitened NIR spectra {tilde over (X)} is orthogonal and their projections over each other are zero.
- In the learning block, each spectrum is calculated through Blind extraction technique. Assume the Data whitened NIR spectra to be x(k) and the first spectra y1(k) need to be extracted. Then, a single processing unit is described as:
-
y 1(k)=w 1 T *x(k)=Σj=1 m w 1j x j(k) (8) -
where w1=[w11, w12, . . . , w1m]T -
b1=[b11, b12, b1L]T - The outputs of the single processing unit y1(k) and ε1(k) represent the extracted spectra, and the error after passing y1(k) by an FIR filter b1 respectively.
-
FIG. 3 is a schematic diagram illustratingsingle processing unit 300 for extracting spectra, according to an exemplary embodiment. - According to an exemplary embodiment, the
single processing unit 300 extracts one orthogonal pure spectrum from a plurality of orthogonal pure spectrums in the received data whitened NIR spectra. Thesingle processing unit 300 estimates the optimal values of vectors w1 and b1 so as to extract spectra. Hence, a cost function J(w1, b1) is defined as: -
J(w 1 , b 1)=E{ε 2} (10) - From Equations 9 and 10, the result obtained can be defined as:
- where Covariance Matrix Rxx=E{xxT} and
- Cross correlation Matrices
-
- Further, the learning algorithm unit of ICA
temporal block 202 minimizes cost function J(w1, b1) to estimate each orthogonal pure spectrum. By differentiating J(w1, b1) with respect to w1 and b1 separately and equating them to 0, the result obtained would be: - Equations (12) and (13) together represent an iterative method, similar to an expectation-maximization (EM) method where the parameters of the previous iteration are used to learn new statistics. Also, from Equation (7), Rxx=In. Therefore, Equation (12) becomes:
- Therefore, the operations of a single processing unit for ICA temporal can be summarized as:
-
- 1. Randomly initializing w1 and b1.
- 2. Obtaining y1 using the current value of w1 (Equation (8)).
- 3. Computing the statistics of source (Rxx, and R) keeping w1 and b1 constant.
- 4. Updating the learning parameters w1 and b1 using Equations (12) and (13).
- 5. In response to w1 convergence being achieved, extracting current y1 spectra.
- Otherwise, above operations are repeated.
- Further, the
deflation module 212 deflates the input to remove the effect of orthogonal pure spectrum derived in a previous operation from data whitened NIR spectra using Equation (15): -
x i+1(k)=x i(k)−{tilde over (w)} 1 *y i(k) (15) - where xi+1(k) is the deflated NIR spectra.
- Further, {tilde over (w)}i is calculated by minimizing mean square cost function Ji({tilde over (w)}i)=E{xi+1 Txi+1} with respect to {tilde over (w)}i, which will give:
-
- where wi is the learning parameter obtained from the last iterative operation. Since, by the
pre-data whitening unit 208, the human NIR spectra were whitened to obtain data whitened spectra, Equation (16) can be simplified to: -
{tilde over (w)}i=wi (17) - The value of {tilde over (w)}i is fed back to the learning algorithm in an iterative fashion to derive the next orthogonal pure spectrum. The above operations are repeated till a certain number of orthogonal pure spectrums are obtained, wherein the orthogonal pure spectrum extracted is appropriately capable of capturing the peaks in all range of wavenumber, adequately representing the pure spectra.
- Further, the
processing block 204 of theuser device 200 comprises theEMSC module 214, the Fast Fourier Transform (FFT) block 216, and thedrift removal module 218. Theprocessing block 204 applies one or more preprocessings and drift removal techniques on the orthogonal pure spectra components to obtain preprocessed spectra. Theprocessing block 204 receives the human NIR spectra and orthogonal pure spectra from the ICAtemporal block 202, and provides the data to the extended multiplicative scatter correction (EMSC)module 214. TheEMSC module 214 applies an extended multiplicative scatter correction (EMSC) method on the human NIR spectra, uses the orthogonal pure spectra and regress for their compositions in the NIR spectra. For instance, let Y be any human NIR spectra comprising various pure spectra X1, X2, . . . , Xk for different blood components. Then, Y can be obtained using simple Linear Regression at any given wavelength as follows: -
Y(n)=a 0+Σk=1 M a k X k (18) - where ak are the strengths of blood component and a0 is a DC component.
- Taking X1 as the glucose spectra, the glucose spectra can be obtained by subtracting other components in the given spectra:
-
X 1 =Y−a 0−Σk=2 M a k X k (19) - The data can be further provided to the
FFT block 216, wherein FFT filtering methods are applied to reduce impact of noise on the spectra and obtain filtered spectra. In the FFT filtering method, a Fourier domain filtering is performed on the human NIR spectra after EMSC is applied to reduce the impact of noise on the spectra by using a Hanning Window of size Nwin, which is expressed in the following equations: -
X(k)=FFT(x(n)) -
x fd(t)=IFFT(X(k)*Hanning(N win) - Further, the
FFT block 216 comprises a differential block that acts as a supplement to the drift removal block to remove the impact of constant drift with respect to a wavelength. Mathematically, the differential block is denoted as -
- With respect to a wavelength, the differential block differentiates the Fourier domain filtered spectra to remove the impact of drift, which is constant with respect to a wavelength, to obtain filtered spectra. Based on the filtered spectra obtained from the FFT block and the differential block, a correlation value for each feature index with respect to glucose without and with the FFT filtering and the differentiation by the differential block can be obtained. From the comparison, it can be observed that there is an increase in the correlation for the most of the indices after the FFT filtering and the differentiation.
- Further, the data from the FFT block 216 can be provided to the
drift removal module 218 that applies a drift removal method to obtain the preprocessed spectra by removing the effect of experimental/instrumental drift in the NIR spectra. During drift removal method, it is assumed that NIR spectra contain only linear drift. - Further, the
user device 200 comprises theregression block 206 that performs regression on the processed data and calculates glucose output from the processed data. Theregression block 206 comprises thefeature extraction block 220, theseparation block 222, thetraining block 224, and the regressionmodel identifier block 226. Thefeature extraction block 220 receives the spectra obtained after data processing block as input and extracts one or more features associated with the processed data signal. The features are the wavelength spectrums which show consistent high correlations with the glucose concentration. The obtained features are provided to theseparation block 222 that separates training data set and validation data set. - The training data set is further provided to the
training block 224 that receives the training data set and trains the regression model. In an exemplary embodiment, thetraining block 224 uses a principal components regression (PCR) method for training the data set. This is merely an example and the disclosure is not limited thereto. The person having ordinarily skill in the art can use any of other similar known methods of regression for training the regression model, without departing from the scope of the invention. Further, the regressionmodel identifier block 226 receives the trained data set from thetraining block 224 and validation data set from theseparation block 222, and obtains the glucose concentration upon performing regression on the training data set and the validation data set. - At least one of the components, elements or units represented by a block in the drawings may be embodied as various numbers of hardware, software and/or firmware structures that execute respective functions described above, according to an exemplary embodiment. For example, at least one of these components, elements or units may use a direct circuit structure, such as a memory, processing, logic, a look-up table, etc. that may execute the respective functions through controls of one or more microprocessors or other control apparatuses. Also, at least one of these components, elements or units may be specifically embodied by a module, a program, or a part of code, which contains one or more executable instructions for performing specified logic functions. Also, at least one of these components, elements or units may further include a processor such as a central processing unit (CPU) that performs the respective functions, a microprocessor, or the like. Further, although a bus is not illustrated in the above block diagrams, communication between the components, elements or units may be performed through the bus. Functional aspects of the above exemplary embodiments may be implemented in algorithms that execute on one or more processors. Furthermore, the components, elements or units represented by a block or processing steps may employ any number of related art techniques for electronics configuration, signal processing and/or control, data processing and the like..
- The “unit” or “module” used herein may be a hardware component, such as a processor or a circuit, and/or a software component that is executed by a hardware component such as a processor.
- Although a few embodiments have been shown and described, it would be appreciated by those skilled in the art that changes may be made in the exemplary embodiments without departing from the principles and spirit of the disclosure, the scope of which is defined in the claims and their equivalents.
Claims (19)
Applications Claiming Priority (5)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
IN201741042881 | 2017-11-29 | ||
IN201741042881 | 2017-11-29 | ||
KR1020180075311A KR102574085B1 (en) | 2017-11-29 | 2018-06-29 | Apparatus and method for predicting blood glucose concentration |
KR10-2018-0075331 | 2018-06-29 | ||
KR10-2018-0075311 | 2018-06-29 |
Publications (3)
Publication Number | Publication Date |
---|---|
US20190159703A1 US20190159703A1 (en) | 2019-05-30 |
US11141083B2 US11141083B2 (en) | 2021-10-12 |
US20210338115A9 true US20210338115A9 (en) | 2021-11-04 |
Family
ID=66633901
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US16/183,325 Active 2039-10-25 US11141083B2 (en) | 2017-11-29 | 2018-11-07 | System and method for obtaining blood glucose concentration using temporal independent component analysis (ICA) |
Country Status (1)
Country | Link |
---|---|
US (1) | US11141083B2 (en) |
Families Citing this family (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US11478158B2 (en) | 2013-05-23 | 2022-10-25 | Medibotics Llc | Wearable ring of optical biometric sensors |
CN110346772A (en) * | 2019-08-22 | 2019-10-18 | 上海无线电设备研究所 | A kind of higher-frequency radar significantly ionosphere phase diameter Disturbance Rejection method |
US20230195079A1 (en) * | 2021-12-21 | 2023-06-22 | International Business Machines Corporation | Characterizing liquids based on features extracted from time-dependent, differential signal measurements |
Family Cites Families (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7436511B2 (en) | 1999-01-22 | 2008-10-14 | Sensys Medical, Inc. | Analyte filter method and apparatus |
KR20020028406A (en) | 2000-10-09 | 2002-04-17 | 정인수 | Iris image acquisition device and iris recognition method with high recognition rates |
AU2002355272A1 (en) | 2001-07-25 | 2003-02-17 | Argose, Inc. | Adjunct quantitative system and method for non-invasive measurement of in vivo analytes |
US6701170B2 (en) | 2001-11-02 | 2004-03-02 | Nellcor Puritan Bennett Incorporated | Blind source separation of pulse oximetry signals |
US7613488B1 (en) | 2002-12-20 | 2009-11-03 | Niresults Inc. | Apparatus and methods for compensation of blood volume effects on NIR spectroscopic measurements of blood analytes |
JP4474145B2 (en) | 2003-11-12 | 2010-06-02 | 株式会社日立メディコ | Optical measuring device |
JP4379155B2 (en) | 2004-03-02 | 2009-12-09 | 株式会社島津製作所 | Biological light measurement method and apparatus using near infrared light |
JP4631510B2 (en) | 2005-03-31 | 2011-02-16 | 株式会社島津製作所 | Brain function information monitoring device |
US20100324398A1 (en) * | 2007-05-11 | 2010-12-23 | Jung Tzyy-Ping | Non-invasive characterization of a physiological parameter |
WO2010039864A2 (en) | 2008-09-30 | 2010-04-08 | Drexel University | Functional near-infrared spectroscopy as a monitor for depth of anesthesia |
US20110200237A1 (en) | 2008-10-15 | 2011-08-18 | Nec Corporation | Pattern matching device and pattern matching method |
WO2012032536A2 (en) | 2010-09-06 | 2012-03-15 | Abhishek Sen | System and method for non-invasive determination of hemoglobin concentration in blood |
US20140372081A1 (en) | 2011-03-29 | 2014-12-18 | Drexel University | Real time artifact removal |
JP6326732B2 (en) | 2013-07-18 | 2018-05-23 | セイコーエプソン株式会社 | Calibration curve creation method and apparatus, and blood component calibration apparatus |
US10362983B2 (en) | 2014-05-12 | 2019-07-30 | Robert R. Alfano | Near infrared photonic prostatoscopy analyzer |
-
2018
- 2018-11-07 US US16/183,325 patent/US11141083B2/en active Active
Also Published As
Publication number | Publication date |
---|---|
US20190159703A1 (en) | 2019-05-30 |
US11141083B2 (en) | 2021-10-12 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US6701170B2 (en) | Blind source separation of pulse oximetry signals | |
US11141083B2 (en) | System and method for obtaining blood glucose concentration using temporal independent component analysis (ICA) | |
US9921201B2 (en) | Calibration curve creating method and calibration curve creation apparatus | |
US20160097676A1 (en) | Target component calibration device, electronic device, and target component calibration method | |
CN105147274B (en) | A kind of method that heart rate is extracted in the face video signal from visible spectrum | |
US10980484B2 (en) | Method of enabling feature extraction for glucose monitoring using near-infrared (NIR) spectroscopy | |
US20160103063A1 (en) | Calibration curve generation device, target component calibration device, electronic device, and glucose concentration calibration device | |
CN102631198B (en) | Dynamic spectrum data processing method based on difference value extraction | |
US20150025340A1 (en) | Calibration curve creating method and apparatus for the same, and blood component calibration apparatus | |
US7706478B2 (en) | Method and apparatus of source separation | |
US20140297197A1 (en) | Calibration curve creation method, calibration curve creation apparatus, and target component gauging apparatus | |
EP3007075A2 (en) | Calibration curve creation method and device, target component calibration method and device, and electronic apparatus | |
JP4379155B2 (en) | Biological light measurement method and apparatus using near infrared light | |
US10568584B2 (en) | Apparatus for estimating blood component level using spectrum analysis | |
US20180088035A1 (en) | Calibration apparatus and calibration curve creation method | |
KR102574085B1 (en) | Apparatus and method for predicting blood glucose concentration | |
US11089981B2 (en) | Methods and systems for performing universal calibration to non-invasively determine blood glucose concentration | |
US20180088034A1 (en) | Calibration apparatus, calibration curve creation method, and independent component analysis method | |
US11204314B2 (en) | Calibration apparatus and calibration curve creation method | |
US20180317817A1 (en) | Method of predicting blood glucose level using near-infrared (nir) spectroscopy data | |
Alrezj et al. | Pre-processing to enhance the quantitative analysis of glucose from NIR and MIR spectra | |
KR102515832B1 (en) | Glucose feature extraction method, Glucose monitoring apparatus and method | |
KR20200010991A (en) | Apparatus and method for determining blood glucose concentration | |
Haritopoulos et al. | Foetal heartbeat frequency estimation from real-world ECG signals using cyclostationary methods | |
Eguizabal Aguado et al. | Direct identification of breast cancer pathologies using blind separation of label-free localized reflectance measurements |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
AS | Assignment |
Owner name: SAMSUNG ELECTRONICS CO., LTD., KOREA, REPUBLIC OF Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:AGGARWAL, GORISH;BYNAM, KIRAN;JOS, SUJIT;AND OTHERS;SIGNING DATES FROM 20180921 TO 20181030;REEL/FRAME:047440/0217 |
|
FEPP | Fee payment procedure |
Free format text: ENTITY STATUS SET TO UNDISCOUNTED (ORIGINAL EVENT CODE: BIG.); ENTITY STATUS OF PATENT OWNER: LARGE ENTITY |
|
STPP | Information on status: patent application and granting procedure in general |
Free format text: DOCKETED NEW CASE - READY FOR EXAMINATION |
|
STPP | Information on status: patent application and granting procedure in general |
Free format text: RESPONSE TO NON-FINAL OFFICE ACTION ENTERED AND FORWARDED TO EXAMINER |
|
STPP | Information on status: patent application and granting procedure in general |
Free format text: NOTICE OF ALLOWANCE MAILED -- APPLICATION RECEIVED IN OFFICE OF PUBLICATIONS |
|
FEPP | Fee payment procedure |
Free format text: PETITION RELATED TO MAINTENANCE FEES GRANTED (ORIGINAL EVENT CODE: PTGR); ENTITY STATUS OF PATENT OWNER: LARGE ENTITY |
|
STPP | Information on status: patent application and granting procedure in general |
Free format text: PUBLICATIONS -- ISSUE FEE PAYMENT VERIFIED |
|
STCF | Information on status: patent grant |
Free format text: PATENTED CASE |