[go: up one dir, main page]

WO2016041842A1 - Procédé et système de détection d'un analyte cible - Google Patents

Procédé et système de détection d'un analyte cible Download PDF

Info

Publication number
WO2016041842A1
WO2016041842A1 PCT/EP2015/070723 EP2015070723W WO2016041842A1 WO 2016041842 A1 WO2016041842 A1 WO 2016041842A1 EP 2015070723 W EP2015070723 W EP 2015070723W WO 2016041842 A1 WO2016041842 A1 WO 2016041842A1
Authority
WO
WIPO (PCT)
Prior art keywords
target
sers
raman
factorization
target analyte
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.)
Ceased
Application number
PCT/EP2015/070723
Other languages
English (en)
Inventor
Tommy Sonne ALSTRØM
Mikkel Nørgaard SCHMIDT
Jan Larsen
Kasper Bayer FRØHLING
Anja Boisen
Michael Stenbaek Schmidt
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Danmarks Tekniske Universitet
Original Assignee
Danmarks Tekniske Universitet
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Danmarks Tekniske Universitet filed Critical Danmarks Tekniske Universitet
Publication of WO2016041842A1 publication Critical patent/WO2016041842A1/fr
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/62—Systems in which the material investigated is excited whereby it emits light or causes a change in wavelength of the incident light
    • G01N21/63—Systems in which the material investigated is excited whereby it emits light or causes a change in wavelength of the incident light optically excited
    • G01N21/65—Raman scattering
    • G01N21/658—Raman scattering enhancement Raman, e.g. surface plasmons
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N2201/00—Features of devices classified in G01N21/00
    • G01N2201/12—Circuits of general importance; Signal processing
    • G01N2201/129—Using chemometrical methods
    • G01N2201/1293—Using chemometrical methods resolving multicomponent spectra

Definitions

  • the present invention relates in one aspect to a method of detecting the presence of a target analyte in a sample by Surface Enhanced Raman Spectroscopy (SERS).
  • SERS Surface Enhanced Raman Spectroscopy
  • the invention relates to a system for detecting the presence of a target analyte in a sample by Surface Enhanced Raman Spectroscopy (SERS).
  • the sample may be in a solid, liquid or gaseous phase.
  • the present invention is concerned with the problem of detecting the presence of one or more specific target analytes in a sample, such as testing for the presence of specific biological and chemical species.
  • the target analytes are specified beforehand.
  • the samples are in a liquid or gaseous phase, but some samples may also be presented in a solid phase, such as in the form of a fine powder with particles in the nanoscale.
  • the result of a detection test is "positive" for a particular target analyte if the presence of that target analyte in the sample can be established.
  • the result is "negative” if the presence of that target analyte in the sample cannot be established.
  • Lowering the detection limit for detection test is therefore important for improving the reliability of such tests.
  • a large variety of techniques have been developed to analyze samples, amongst those optical spectroscopy techniques. When using optical spectroscopy, the energy dependence of the optical response of the sample to an optical stimulus is recorded as a spectrum, plotting e.g. intensity against energy of the optical response, and characteristic features in the spectrum are analyzed to identify constituents of the sample.
  • Raman spectroscopy is a technique analyzing the inelastic scattering of stimulating photons with molecules where the spectral response exhibits pronounced peaks that are shifted in energy with respect to the stimulus.
  • the energy shifted peaks reflect events of inelastic interaction between the probing stimulus and characteristic molecular states of molecules in the sample and therefore yield specific information about the sample on a molecular level.
  • a particularly sensitive Raman spectroscopy technique is Surface Enhanced Raman Spectroscopy (SERS), wherein molecules that are adsorbed to certain SERS-active substrate surfaces exhibit an enhanced Raman scattering signal with an enhancement factor of up to 10E12 or even more, thereby allowing, in principle, for the detection of single molecules. While such enhancement factors may be achieved in certain experimental set-ups, they are in practical applications of detecting a target analyte in a real world sample typically obscured by the unavoidable presence of other substances in the sample.
  • SERS Surface Enhanced Raman Spectroscopy
  • spectral contributions from different constituents may be mixed together in the same spectrum and may therefore be difficult to distinguish from each other.
  • Some attempts to de-mix such spectral information involve the acquisition of large numbers of spectra obtained under more or less systematically varied conditions to vary the contributions from the different constituents.
  • WO 2012/1 121 19 proposes a method for analyzing a liquid sample containing multiple unknown analyte components by Ra- man spectroscopy, in particular SERS.
  • the method includes forming a layer of ana- lytes on a SERS surface, obtaining Raman spectra of multiple respective pixels, applying a band target entropy minimization (BTEM) algorithm to obtain pure component spectra, and subsequently identifying analytes from the pure component spectra.
  • BTEM band target entropy minimization
  • the proposed method is for extraction of estimated pure component Ra- man spectra from the experimentally obtained set of mixed Raman spectra. Unknown analytes are identified by comparison of the estimated spectra to known spectra stored in spectral libraries. However, the successful de-mixing and detection relies on the presence of different spectral contributions at various locations of the SERS-active surface. The method is therefore not well-suited for detecting a specific target analyte at very low concentrations.
  • the Raman/SERS signal is enhanced through strong localized enhancement of electromagnetic fields observed in so-called hot-spots, such as formed in nano gaps between metal surfaces.
  • SERS the statistical behavior of the electromagnetic hot spot distribution. It is not only required to find a hot-spot, but there also have to be molecules captured in the hot-spot, which is increasingly unlikely as the concentration of the target molecules decreases.
  • the area of the SERS-active surface that is exposed to the sample may be increased and spectra from the enlarged area may be recorded in a Raman/SERS map. This increases the likelihood that spectra from hot spots containing molecules are recorded.
  • the challenge that arises from this strategy is the compilation of a large amount of irrelevant data diluting the signal of interest from the rare target analyte.
  • a first aspect of the invention relates to a method of detecting the presence of one or more specific target analytes in a sample, the method comprising the steps of
  • the data analysis unit comprises a factorization model with parameters that are estimat- ed via statistical inference techniques, such as using a Bayesian frame work, and a classifier that determines a detection result from the model based on the estimated parameters.
  • the sample containing analytes may be presented in any suitable form including solid, liquid or gas phase.
  • the sample may comprise sub-samples that may have different constituents, different form, and different analyte content.
  • the analytes may be captured on the SERS-active surface in any known manner.
  • a SERS- active surface is exposed to the sample so as to capture analytes from the sample by adherence to the SERS-active surface.
  • SERS-active surface is to be understood broadly to also include surface areas located on different SERS- substrates, or even on different SERS-devices, each SERS-device comprising one or more substrates with a surface adapted for SERS-type measurements.
  • Raman spectroscopy measurements are then performed on the SERS-active surface to obtain Raman/SERS-spectra.
  • an ensemble X of Raman/SERS- spectra are measured at different locations on the SERS-active surface.
  • at least a subset of the Raman/SERS spectra is measured at contiguous locations in the same portion of the SERS-active surface so as to provide a Raman/SERS spectral map from that portion.
  • the ensemble X of Raman/SERS- spectra may also include measurements taken at the same location on the SERS- active surface, e.g. by repetitive scanning.
  • the SERS-active surface is adapted to enhance the Raman signal in so-called hot-spots.
  • the spectrometer device may be configured for localized acquisition of Raman/SERS-spectral data may be acquired from the SERS-active surface by any known technique, including advanced techniques such as resonant excitation.
  • a specific analyte to be tested for in the detection is referred to as target analyte.
  • the detection procedure may be directed to testing for the presence of a single tar- get analyte or of multiple target analytes simultaneously.
  • the different target analytes may be present in the same sub-sample or may be present in different sub- samples measured in different measurement campaigns/series and on the SERS- active surface of different SERS-substrates/devices.
  • the method according to the present invention allows for simultaneously analyzing large data sets of Ra- man/SERS-spectra in order to simultaneously test for the presence of a target analyte.
  • the method even allows for testing for multiple target analytes at the same time and in the presence of unidentified analytes (contaminants), irrespective of whether the target analytes and other analytes are present in the same sub-sample or in different sub-samples, if the spectra are obtained on the same SERS-active sub- strate or on completely different SERS-devices, as long as the measured data can be assembled to an ensemble X of Raman/SERS-spectra that can be represented in a common matrix.
  • the ensemble of spectral data is then modelled, wherein modelling includes factor- izing the ensemble X of Raman/SERS-spectra as a loadings matrix A times a spectral component matrix S, wherein matrix elements of A and S are all real and non- negative.
  • Real numbers are in the present application denoted by the double-struck letter R.
  • the term "real and non-negative" covers all real numbers including zero and is in the present application denoted by the double-struck letter R with subscript "+”.
  • the factorisation model deviates from the actual measurement data by at least measurement noise.
  • This deviation may be dealt with, e.g. by adding noise contributions as a residual noise matrix E to the product of the loadings matrix A times the spectral components S, wherein matrix elements of E are all real. Since E describes measurement noise, the elements may also assume negative values.
  • the ensemble X of Ra- man/SERS-spectra may thus be written as:
  • the presence of the target analyte may be determined on the basis of all or some of the estimated parameters of the factorization-model.
  • the factorization-model with non-negative elements allows for an intuitive physical interpretation of the model parameters.
  • the basis vectors of S may be interpreted as spectral components of the mixed spectrum, wherein one or more of these spectral components may be identified as target components, each representing the spectral signature of a particular target analyte.
  • Information on the presence of the target analyte in the sample under investigation may therefore be derived directly from the estimated loadings for the selected target components of S, whereas information on the reliability of detection, the detection limit or the signal-to-noise ratio of the detection may be derived e.g. from the loadings of the target component and the estimated noise matrix E in combination.
  • the presence of the target analyte may be determined by a classifier using the factorization model with the estimated parameters as input.
  • the method may thus include feeding estimated factorization model parameters to a classifier, which on the basis of these parameters determines the presence of the one or more target analytes.
  • estimation of at least one of the one or more target components is subject to a population rule determining the population of the target component elements with values according to a predetermined function.
  • Each of the target components of S may be subject to a population rule.
  • the population rule governs the population of a given target component during estimation/inference of the elements of S that make up the target component vector.
  • spectral reference may be derived in any suitable way, e.g. from a library of spectral data, a theoretical model, or from a reference measurement for the target analyte. Different target analytes may be subjected to different population rules.
  • a population rule for the corresponding target com- ponent in S may be omitted.
  • Spectral components for unknown analytes are not subject to any population rule.
  • the predetermined function is one of a fixed value target spectrum that is kept fixed during estimation/inference, a parametric description of the target spectrum with target component parameters that are estimated/inferred by the joint estimation; and a probabilistic parametric description of the target spectrum with target component parameters that are estimated / inferred by the joint estimation.
  • the values of the target component for a given target analyte are set to the target spectrum selected in step (c) as representative of that target analyte, and the target component is kept fixed during the joint estimation of the parameters in the factorization-model.
  • This strict population rule is particularly useful when a highly reliable reference spectrum is available as target spectrum, e.g. from SERS-measurements on a reference sample containing a high concentration of the target analyte.
  • the values of the target component for a given target analyte populated according to a parametric description of the target spectrum selected in step (c) as representative of that target analyte, and the parameters of the parametric target component description are estimated along with the other model parameters during the joint estimation of the parameters in the factorization-model.
  • This parametric population rule is particularly useful e.g. when the target spectrum is available in the form of parametric peak functions describing characteristic spectral features occurring in the target spectrum of the target analyte.
  • a probabilistic parametric function may be useful as a less strict population rule to allow for further flexibility in the description of the target spectrum during estimation, e.g.
  • the parameters of a probabilistic parametric target component description are also estimated along with the remaining parameters of the factorization model.
  • the target components of one or more of the target analytes are not subject to any specific population rule and the respective target components are estimated on equal footing with the other spectral components of the spectral component matrix S.
  • step (c) the respective target spectra determined in step (c)
  • the respective target spectra determined in step (c) may be, for example, a reference spectrum from a spectral data library, a theoretical model, or a measurement on a reference sample of the target analyte.
  • at least a first target component and a second target component are subject to respective population rules, wherein the predetermined functions determining the population of the first and second target components are different from each other.
  • the population for some or all target components may be the same. However, depending on the available infor- mation on the spectral signatures for different target analytes, different population rules may formulated for each of the target components.
  • Target components populated according to different functions may be combined in the same model.
  • Choice of the population rule may e.g. depend on the quality/availability of spectral information about the various target analytes.
  • the target spectrum of at least one target analyte in step (c) is determined from reference Raman/SERS- measurements of said at least one target analyte captured on a SERS-active surface.
  • a target spectrum coming close to the expected experimental conditions of the detection situation is obtained.
  • the SERS-active surface for measuring the target spectrum is of the same type as the SERS-active substrate used for the detection measurement, thereby closely matching the experimental conditions of the detection situation.
  • the target analyte is captured on the SERS-active surface by exposing the SERS-active surface to a reference sample comprising the target analyte.
  • the reference sample comprises the target analyte with a sufficiently high concentration so as to obtain a target spectrum with well pronounced spectral features and with a high signal to noise ratio.
  • the concentration of the target analyte should be such as to allow for a clear distinction of the Raman/SERS-signal form the target analyte in the spectra measured on the reference sample.
  • a concentration of target analyte corresponding to a monolayer or more distributed across the surface of a SERS- active substrate on which the reference Raman/SERS measurements are performed is considered a high concentration that is suitable for obtaining a reliable target spectrum.
  • the reference Raman/SERS measurements include an ensemble R of reference Raman/SERS-spectra, and the target spectrum is extracted from a factorization-model of the ensemble R.
  • the target spectrum in step (c) may thus extracted from the ensemble R of reference Ra- man/SERS-spectra by performing steps analogue to steps (a)-(f) above on the reference sample, wherein the reference sample is prepared with a sufficiently high concentration of the target analyte as discussed above.
  • the target spectrum is replaced by a pre-determined reference spectrum for the target analyte - e.g.
  • the ensemble R may then be modelled as a product of a loadings matrix times a spectral component matrix and the spectral component that comes closest to the pre-determined reference spectrum is selected as the target spectrum for the detection measurement.
  • the factorization-model further includes an additive term with a residual noise matrix E, wherein matrix elements of E are all real.
  • the noise term E may be comprise values that are fixed, that are determined by parametric functions, and/or determined by probabilistic parametric functions.
  • the factorization-model further includes an additive term with a background matrix B, wherein matrix elements of B are all real and non-negative.
  • the background term B may be comprise values that are fixed, that are determined by parametric functions, and/or determined by probabilistic parametric functions.
  • both the noise contributions and the background contributions may be modelled as further spectral components in the spectral component matrix S, thereby contributing additive with appropriate loadings to the total Raman/SERS-signals.
  • the joint estimation is performed in a Bayesian framework.
  • the Bayesian framework uses sparse priors, such as exponential priors in a formulation analogue to the model described in Schmidt et al., "Bayesian Non-negative Matrix Factorization” (2009), Lecture notes in Computer Science, pp.540-547.
  • the sparse priors are applied to the loadings matrix A and the spectral components of S that are not governed by a population rule ("S free ").
  • S free population rule
  • the spectrally broad, typically smooth nature of the background signal is reflected by a smoothness constraint to B, and the measurement noise term E may be denoted as "weakly informative".
  • a total number K of spectral components forming the spectral component matrix S is treated as a parameter of the factorization model, which is estimated by the joint estimation, wherein advantageously, the total number is estimated using non-paramtric Bayesian techniques.
  • Another way of determining the total number of K of spectral components to be included in the factorization model is cross-validation where the method is performed using different numbers of K, and the choice of K is cross-validated against an eval- uation of achieved result, e.g. by analysing for the different values of K the deviation of the product of A and S with respect to the ensemble of measured data X.
  • a detection result for the presence of a given target analyte is determined as positive, if the loading of the target component exceeds a pre-determined threshold. Further according to one embodiment of the method, a positive detection result for the presence of a given target analyte in one location on the SERS-active surface is validated by regarding loadings of the corresponding target component in neighbouring locations on the SERS-active surface.
  • the parameters of the factorization are estimated in a Bayesian framework using a Gibbs-sampler. A suitable Gibbs sampler is found in the above-mentioned document by Schmidt et al.
  • a second aspect of the invention relates to a system for detecting the presence of one or more specific target analytes in a sample, the system comprising a SERS- active surface, a spectrometer device configured for the acquisition of Ra- man/SERS-spectral data from the SERS-active surface, and a computer including devices for data storage and data processing, wherein the computer is configured for performing the steps of
  • Analytes are captured on a SERS-active surface and an ensemble X of Raman/SERS-spectra is acquired from that SERS-active surface.
  • the ensemble X of Raman/SERS-spectra is stored in a data storage device, and then processed by a data processing unit that determines, for each target analyte, if an instance of positive detection can be established.
  • the data processing unit comprises a factorization model with parameters that are estimated via statistical inference techniques, such as using a Bayesian frame work, and a classifier that determines a detection result from the model based on the estimated parameters.
  • the spectrometer device includes an excitation light source adapted to excite Raman scattering events on the SERS-active surface, a detection unit for the spatially and spectrally resolved detection of light scattered from the SERS-active surface and for the conversion of the detected light to computer readable signals.
  • the spectrometer device further includes means for focusing the excitation light source to a spot and means for positioning the excitation spot on the SERS-active substrate and/or means for spectrally resolved detection of the light scattered from the SERS-active surface.
  • the SERS-active surface is adapted to enhance the Raman signal in so-called hot-spots.
  • the spectrometer device may be configured for localized acquisition of Raman/SERS-spectral data from a surface by any known technique, including advanced techniques such as resonant excitation Raman spectroscopy.
  • At least one of the one or more target components is populated during estimation of the parameters of the factorization model by
  • the system may be configured for detecting multiple target components, and more than one of these target components may be subject to respective population rules.
  • the population rules for populating different target components may be of the same type and/or may differ from each other.
  • a system according to any of the above-mentioned embodiments is configured to perform the method according to any of the above-mentioned embodiments.
  • FIG. 1 a schematic diagram of a system for detecting a target analyte according to one embodiment
  • FIG. 2 a SEM micrograph of a SERS-active substrate
  • FIG. 3 (a) a Raman/SERS-map and (b) a Raman/SERS-spectrum at a given location on the map,
  • FIG. 4 (a) a Raman/SERS-map and (b) three spectra at three locations (I, II, III) on the map,
  • FIG. 6 (a) a second loadings-map and (b) the corresponding second spectral component as estimated in the factorization model of Fig. 5, and in FIG. 7 (A) a Raman/SERS-map, (B) a loadings map, and (C) the corresponding target component as estimated in a factorization model of an ensemble of Raman/SERS-spectra for low concentrations of target analyte.
  • FIG.1 shows a schematic diagram of a system 100 for detecting a target analyte in a sample according to one embodiment of the invention.
  • the system comprises a measurement portion 101 and a data analysis portion 102.
  • the measurement por- tion 101 comprises a SERS-device 1 10 with a SERS-active surface 1 1 1 on which analytes 99 from a sample are captured.
  • the measurement portion 101 further comprises a spectrometer device 120 with an excitation light source 121 adapted to excite Raman scattering events on the SERS-active surface 1 1 1 , and a detection unit 122 for the spatially and spectrally resolved detection of light scattered from the SERS-active surface 1 1 1 and for conversion of the detected light to computer readable signals.
  • the spectrometer device 120 further includes means for focusing the excitation light from source 121 to a spot and means for positioning the excitation spot on the SERS-active surface 1 1 1.
  • Means for positioning the excitation spot for a performing a measurement may include optical means for moving the ex- citation light with respect to the SERS-surface, means for mechanically positioning the SERS-device 1 1 and/or mechanically positioning the spectrometer device 120.
  • the spectrometer device 120 is thus configured for the acquisition of an ensemble X of localized/spatially resolved Raman/SERS-spectra from the SERS-active surface 1 1 1 .
  • the spectrometer device 120 may be configured for acquisition of Ra- man/SERS-spectral data from a surface by any known Raman-spectroscopy technique, including advanced techniques such as resonant excitation Raman spectroscopy.
  • the measurement portion 101 has a SERS-device 1 10 with an arrangement of silver coated silicon nano-pillars as the SERS-active surface 1 1 1 on a silicon substrate. Preparation of such SERS-devices 1 10 is disclosed e.g. in WO 201 1/047690 A1 . An SEM-micrograph of silver coated silicon nano-pillars on a SERS-active surface is depicted in Fig.2.
  • the spectrometer device 120 used for acquiring the Raman/SERS-data of the examples is a Thermo Scientific DXR Raman Microscope using a 780 nm excitation laser.
  • a laser powered at 0.5 mW power was used in conjunction with a 50x optical objective which yields an approximately 1 ⁇ diameter laser spot size.
  • the total acquisition time in each spot was 1 s, wherein each spot was measured twice within that total acquisition time, before a 6 th order polynomial baseline correction was performed.
  • Each ensemble of Raman/SERS-spectra consisted of spectra gathered in larger squares (up to 10000 points) with a 1x1 ⁇ grid.
  • the target analyte was 17p-Estradiol (E2) in a fluorescent labelled version called Estradiol Glow (EG), which is available from Jena Bioscience GmbH, Germany.
  • High concentration solutions comprise 1 ⁇ EG and low concentrations comprise 10 nM EG.
  • the ensemble X of Raman/SERS-spectra is passed to the data analysis portion 102 and stored in a data storage 130 in the form of a matrix [ X ] of N spectra with D values each.
  • the ensemble X of Raman/SERS-spectra is then passed to a data modelling device 140 for modelling by a factorization model with a loadings matrix A times a spectral component matrix S and an additive term E representing the measure- ment noise.
  • the spectral component matrix S has K spectral component vectors with D values each
  • the loadings matrix A has N loading vectors with K elements for weighting of the K spectral components in a linear combination modelling the total Raman/SERS-signal in each of the N locations/instances.
  • the number K of spectral components is a model parameter that may be chosen beforehand. In the examples given below, K was chosen to be ten.
  • the parameters of the factorization model are estimated in a Bayesian framework using a Gibbs sampler, and the result of the estimation is passed from the data modelling device 140 to a classifier 150 for determining the presence of the target analyte from the parameters of the model.
  • the spectral component representing the Raman/SERS-spectral signature of the target analyte is referred to as the target component.
  • the target component is selected amongst the spectral components of the spectral component matrix S.
  • the presence of the target analyte may advantageously be established from the loadings of the spectral component representing the Raman/SERS-spectral signature of the target analyte as discussed above.
  • the target component for a given analyte may be selected amongst the group of spectral components ⁇ S fre e ⁇ with values that may be freely estimated.
  • the target component is identified by comparison with a known spectral response of the target analyte, such as derived from a spectral data library, from a theoretical model, or experimentally from a measurement on a reference sample. The comparison with the known spectral response is performed after estimation of the model parameters.
  • the known spectral response of the target analyte may be integrated directly in the model as forehand information, thereby significantly improving the performance of the data analysis portion in terms of lower detection limit and enhanced robustness of the detection as compared to a model neglecting such forehand information.
  • the target analyte to be tested for is known beforehand.
  • the target component of a given target analyte may be selected in a group of spectral components that are subject to a population rule for populating the values of the spectral components during the estimation routine.
  • the population rule for the target component of the given analyte may be that all values are fixed to a specific spectrum throughout the estimation, thus only allowing the loadings of this specific spectrum to be estimated in the model.
  • the fixed spectrum may be a known spectral response as discussed above.
  • due to the strict nature of the fixed spectrum population rule it is recommended to use a forehand knowledge of the target component that is close to the spectral response of the target analyte under the actual experimental conditions of the measurement portion 101 .
  • the population rule may describe the target component in a parametric form.
  • each of the target components may be subject to a different population rule.
  • Multiple spectral components subject to such population rules may be grouped as a group of fixed spectral components ⁇ S fix ⁇ and a group of parametric spectral components ⁇ Sp a ra ⁇ , respectively.
  • the parametric functions describing the parametric spectra may be probabilistic parametric functions.
  • the parameters of the parametric spectral components ⁇ S par a ⁇ are estimated along with the remaining parameters of the factorization model.
  • FIG.3 shows (a) a Raman/SERS-map and (b) a Raman/SERS-spectrum at a given location on the map as indicated by the arrow.
  • the Raman/SERS-map visualizes the recorded intensities at a specific Raman shift chosen to be where the spectral response of the target molecule is expected to have a peak. Areas with high intensities are identified and these areas are denoted as potential hot-spots containing target molecules.
  • the Raman/SERS data shown in Fig.3 illustrates an uncontami- nated Raman/SERS measurement.
  • Fig.3(a) maps the Raman/SERS-signal intensi- ties at 1 166 cm “1 and Fig.3(b) gives the Raman/SERS-spectrum for Estradiol Glow (EG) at the location where the Raman/SERS-map for 1 166 cm “1 has a maximum. Hot-spots containing molecules are readily identified on the Raman/SERS-map as the bright areas. In the Raman/SERS-spectrum for EG, the peaks at 1 166 cm “1 and 1580 cm “1 are considered the major discriminative. The spectra are collected over a 30 ⁇ 30 ⁇ area with a resolution of 1 ⁇ .
  • Raman/SERS maps Traditional analysis of Raman/SERS maps rely on choosing one or two Raman shifts where the dominant peaks for the molecule of interest are present. In the case where the target molecule is present (and at a hot-spot) the intensities are considered to be heavy tail distributed whereas Raman/SERS maps for blank substrates are considered to be normally distribut- ed. Detection is then performed by testing if a Raman/SERS-map has heavy tailed intensities. This approach is not specific and will be sensitive to outliers or contaminants that give high intensities at the selected Raman shifts. This becomes clear when considering FIG.4. Example 2
  • FIG.4 gives an example of a Raman/SERS-map that has been locally contaminated.
  • Fig.4(a) maps the Raman/SERS-signal intensities at 1 166 cm "1 .
  • the Raman/SERS- map at 1 166 cm "1 exhibits a number of maxima.
  • Fig.4(b) gives three spectra labelled (I, II, III) corresponding to the three locations of maximum intensity indicated by (I, II, III) on the map. Only the very weak spectrum labelled (III) is actually the Raman/SERS-spectrum for the target analyte EG.
  • the other spectra (I, II) are to be attributed to contaminants, most probably salt residuals stemming from sample preparation steps.
  • the contaminant signals obscure the target Raman/SERS signal, mislead the traditional data analysis methods and thereby prevent an appropriate detection of the target analyte. As illustrated in Fig.5 and Fig.6, this issue is addressed by using the factorization model according to the present for de-mixing the different spectral components.
  • A, S and E are jointly estimated in a Bayesian framework using exponential priors with a rate equal to one for both A and S in order to account for the sparse nature of these matrices in analogy to the model described in Schmidt et al., "Bayesian Non-negative Matrix Factorization” (2009), Lecture notes in Computer Science, pp.540-547.
  • an effective Gibbs sampler was derived that is also used here for parameter estimation.
  • a burn in period of 20 Gibbs-sweeps was used and then 50 sweeps were used to generate 50 samples from each component. The mean values from these samples were used as the parameter estimates.
  • a prior for the noise variance in E we chose a inverse gamma density with shape and scale parameters equal to zero (a flat improper prior), in order
  • Fig.5(a) and Fig.6(a) map first and second loadings for first and second spectral components, respectively. The loadings are extracted from the loadings matrix A.
  • Fig.5(b) and Fig.6(b) show the corresponding first and second spectral component for these first and second loadings as extracted from the spectral component matrix S.
  • the most dominant area for 1 166cm "1 area I in Fig. 4(a)
  • the mixed area (spectrum II on Fig.4(b))
  • the EG-spectrum is selected as the target component in the spectral component matrix S.
  • the loadings for the target component can now be used to identify where EG is likely to be present in the map.
  • the method correctly identifies spots where EG is present.
  • the proposed method successfully identifies a Raman/SERS-spectrum as one of the spectral components in S.
  • the method is applied on the data shown in Fig.3, and one of the spectral component basis vectors readily corresponds to a spectrum for EG.
  • the Raman/SERS-spectrum for EG as derived from the freely estimated spectral component matrix S for high concentration measurements may now be used as a fixed target component S fix in a new test for detecting the presence of EG as the target analyte at very low concentrations.
  • the sensitivity and selectivity of the method/system for detecting the target analyte, here EG is significantly improved as compared to traditional data analysis.
  • the Raman/SERS-spectrum of interest will only make up a minor portion in the data and the aforementioned approach is often not able to identify the target spectrum.
  • the factorization model is not able to automatically identify the EG spectrum as one of the spectral components in S. This is most probably, because the target molecule is giving a weak signal as well as being a rare occurrence.
  • Fig.7(A) maps the Raman/SERS signal intensity at 1 166 cm "1 ;
  • Fig.7(B) maps the loadings of the EG target component;
  • Fig.7(C) provides the Raman/SERS-spectrum at a location that is identified by the data-analysis as having EG present.
  • spectral component identified as EG in the high concentration measurements shown on Fig.5 as a vector in S when analysing the ensemble of Raman/SERS-spectra from the low concentration sample, and keeping that column fixed during parameter estimation (probing mode), areas with EG on the SERS- active substrate are successfully identified.

Landscapes

  • Health & Medical Sciences (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Chemical & Material Sciences (AREA)
  • Analytical Chemistry (AREA)
  • Biochemistry (AREA)
  • General Health & Medical Sciences (AREA)
  • General Physics & Mathematics (AREA)
  • Immunology (AREA)
  • Pathology (AREA)
  • Investigating, Analyzing Materials By Fluorescence Or Luminescence (AREA)
  • Spectrometry And Color Measurement (AREA)

Abstract

L'invention concerne un procédé et un système de détection de la présence d'un ou de plusieurs analytes cibles spécifiques dans un échantillon. Le procédé comprend les étapes consistant à exposer une surface active du point de vue de la diffusion Raman exaltée de surface à l'échantillon de façon à capturer des analytes provenant de l'échantillon par adhérence à la surface active du point de vue de la diffusion Raman exaltée de surface ; acquérir un ensemble X de spectres Raman/diffusion Raman exaltée de surface, mesurés à différents emplacements sur la surface active du point de vue de la diffusion Raman exaltée de surface ; pour chaque analyte cible, déterminer un spectre cible représentatif de cet analyte cible ; modéliser l'ensemble X de spectres Raman/diffusion Raman exaltée de surface mesurés à l'aide d'un modèle de factorisation, le modèle de factorisation comprenant une matrice de charges A par une matrice de composantes spectrales S, les éléments de matrice de A et S étant tous réels et non négatifs ; estimer conjointement des paramètres dans le modèle de factorisation, les paramètres estimés comprenant au moins A et S ; pour chaque analyte cible, sélectionner, parmi les composantes spectrales de S, une composante cible correspondant au spectre cible de cet analyte cible ; et déterminer la présence du ou des analytes cibles au moins sur la base des charges estimées pour les composantes cibles sélectionnées. Pendant l'estimation, le composant cible ou au moins l'un des composants cibles est défini selon une fonction prédéterminée.
PCT/EP2015/070723 2014-09-17 2015-09-10 Procédé et système de détection d'un analyte cible Ceased WO2016041842A1 (fr)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
EP14185146.9 2014-09-17
EP14185146 2014-09-17

Publications (1)

Publication Number Publication Date
WO2016041842A1 true WO2016041842A1 (fr) 2016-03-24

Family

ID=51542249

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/EP2015/070723 Ceased WO2016041842A1 (fr) 2014-09-17 2015-09-10 Procédé et système de détection d'un analyte cible

Country Status (1)

Country Link
WO (1) WO2016041842A1 (fr)

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2011047690A1 (fr) 2009-10-23 2011-04-28 Danmarks Tekniske Universitet Substrats de diffusion raman a exaltation de surface, consommables pour spectroscopie raman
WO2012112119A1 (fr) 2011-02-16 2012-08-23 Agency For Science, Technology And Research (A*Star) Procédé et appareil pour l'analyse d'un échantillon liquide par spectroscopie raman

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2011047690A1 (fr) 2009-10-23 2011-04-28 Danmarks Tekniske Universitet Substrats de diffusion raman a exaltation de surface, consommables pour spectroscopie raman
WO2012112119A1 (fr) 2011-02-16 2012-08-23 Agency For Science, Technology And Research (A*Star) Procédé et appareil pour l'analyse d'un échantillon liquide par spectroscopie raman

Non-Patent Citations (5)

* Cited by examiner, † Cited by third party
Title
LI ET AL., JOURNAL OF VLSI SIGNAL PROCESSING, 2007, pages 83
MÓNICA B. MAMIÁN-LÓPEZ ET AL: "Quantification of moxifloxacin in urine using surface-enhanced Raman spectroscopy (SERS) and multivariate curve resolution on a nanostructured gold surface", ANALYTICAL AND BIOANALYTICAL CHEMISTRY, vol. 405, no. 24, 1 September 2013 (2013-09-01), DE, pages 7671 - 7677, XP055225853, ISSN: 1618-2642, DOI: 10.1007/s00216-013-7200-y *
SCHMIDT ET AL.: "Bayesian Non-negative Matrix Factorization", LECTURE NOTES IN COMPUTER SCIENCE, 2009, pages 540 - 547
TAMÁS FIRKALA ET AL: "Investigation of drug distribution in tablets using surface enhanced Raman chemical imaging", JOURNAL OF PHARMACEUTICAL AND BIOMEDICAL ANALYSIS, vol. 76, 1 March 2013 (2013-03-01), US, pages 145 - 151, XP055225931, ISSN: 0731-7085, DOI: 10.1016/j.jpba.2012.12.017 *
ZHOU KENNETH J ET AL: "Nonnegative matrix factorization: a blind sources separation method to extract content of fluorophores mixture media", PROGRESS IN BIOMEDICAL OPTICS AND IMAGING, SPIE - INTERNATIONAL SOCIETY FOR OPTICAL ENGINEERING, BELLINGHAM, WA, US, vol. 8947, 4 March 2014 (2014-03-04), pages 89471V - 89471V, XP060033874, ISSN: 1605-7422, ISBN: 978-0-8194-9850-2, DOI: 10.1117/12.2035536 *

Similar Documents

Publication Publication Date Title
EP2035793B1 (fr) Procédés d'analyse spectroscopique
Liland et al. Optimal choice of baseline correction for multivariate calibration of spectra
CN110261405B (zh) 基于显微高光谱技术的绝缘子污秽成分识别方法
JP6091493B2 (ja) 試料に存在する成分を決定するための分光装置と分光法
CA2795585C (fr) Procede d'analyse spectroscopique
EP2853885B1 (fr) Procédé de mesure de spectres Raman pour l'inspection de médicaments
US8658973B2 (en) Auger elemental identification algorithm
JP2007502988A (ja) 蛍光寿命画像顕微法および分光法
EP3368885B1 (fr) Procédés et systèmes d'analyse de la qualité de fixation d'un tissu par spectroscopie en moyen infrarouge
CN107532989A (zh) 用于分析颗粒的方法
Schie et al. Estimation of spectra sample size for characterizing single cells using micro‐Raman spectroscopy
CN106404743A (zh) 一种结合拉曼光谱和近红外光谱的探测方法及探测装置
KR102866692B1 (ko) 이미지를 활용한 화학적 성분의 분석 방법, 장치 및 프로그램
Pavillon et al. Feature‐based recognition of Surface‐enhanced Raman spectra for biological targets
Angeyo et al. Principal component analysis‐assisted energy dispersive X‐ray fluorescence spectroscopy for non‐invasive quality assurance characterization of complex matrix materials
KR20170062812A (ko) 에너지 전이에 따른 형광 수명을 측정하는 방법
Heller et al. Hidden information in principal component analysis of ToF‐SIMS data: on the use of correlation loadings for the identification of significant signals and structure elucidation
Lau et al. A Bayesian Whittaker–Henderson smoother for general‐purpose and sample‐based spectral baseline estimation and peak extraction
JP6280910B2 (ja) 分光システムの性能を測定するための方法
WO2016041842A1 (fr) Procédé et système de détection d'un analyte cible
Aalders et al. Investigating the age of blood traces: How close are we to finding the holy grail of forensic science?
Globus et al. Sub-terahertz resonance spectroscopy of biological macromolecules and cells
Sattlecker et al. Assessment of robustness and transferability of classification models built for cancer diagnostics using Raman spectroscopy
Sun et al. Spectroscopic determination of soil organic carbon and total nitrogen content in pasture soils
JP2004219250A (ja) 粒子中の化学成分測定方法

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 15760466

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205 DATED 24.05.2017)

122 Ep: pct application non-entry in european phase

Ref document number: 15760466

Country of ref document: EP

Kind code of ref document: A1