JP6281973B2 - Method for expressing properties of a mixture sample, method for evaluating properties of a mixture sample, method for identifying attributes of a mixture sample, and method for processing an electromagnetic wave signal derived from a mixture sample - Google Patents

Description

  The present invention relates to a technique for analyzing (analyzing) a mixture sample, and in particular, a method for expressing the characteristics of a mixture sample, a method for evaluating the characteristics of a mixture sample, a method for identifying attributes of a mixture sample, and a mixture sample. The present invention relates to a method for processing an electromagnetic wave signal.

Mixtures are broadly classified into “simple mixtures” and “complex / heterogeneous mixtures”.
“Simple mixture” refers to a combination of a relatively small number of identifiable substances.
“Complex / heterogeneous mixture” refers to (1) a large number of substances mixed in various concentrations, (2) containing unknown components, and (3) showing an interaction between substances.

  Here, the numerous substances include low-molecular to high-molecular compounds, simple substances (substances composed of a single element), and the like. Further, most of the interaction between substances is considered to be an interaction between molecules caused by an intermolecular force acting between molecules. This intermolecular interaction causes the compounds and simple substances in the “complex and heterogeneous mixture” to be supramolecularly assembled and complexed, thereby determining the structure and function of the entire “complex and heterogeneous mixture”. Thus, a “complex / heterogeneous mixture” exhibits unique properties as a whole.

  Although detailed description is omitted here, the intermolecular force that is the essence of the intermolecular interaction includes Coulomb force, orientation force, induced force, dispersion force, charge transfer force, exchange repulsive force, etc. Actually observed intermolecular interactions include electrostatic interactions, van der Waals interactions, and charge transfer interactions.

  Examples of the “complex / heterogeneous mixture” include fossil fuels (coal, petroleum, etc.), synthetic polymers, and biological samples. Of course, the present invention is not limited to these, and the “complex / heterogeneous mixture” includes many and many kinds of mixtures such as water pollutants and malodorous substances.

  Coal is a mixture with a very complex composition containing many hydrocarbon compounds (molecular weights of several hundred to several thousand) that are solid at room temperature, and is used as a power generation fuel and coke raw material. The chemical structure of coal as a whole (coal structure) still has many unknown points, which delays technological development that enables efficient and clean use of coal. In recent years, the elucidation of coal structure by intermolecular interaction and the design and development of innovative and efficient coal utilization processes based on the elucidation have been underway.

  Petroleum is a mixture with a very complicated structure containing many hydrocarbon compounds (molecular weight of several hundreds or less) that are liquid at room temperature, and is used as a fuel for transportation and as a chemical raw material. Until now, petroleum has been refined based on the average structural formula and general properties, but in recent years, crude oil and petroleum products have been regarded as a collection of molecules, based on their detailed composition and chemical structure. The development of petroleum technology that analyzes and predicts physical properties and reactivity is thriving.

  Synthetic polymers are generally complex mixtures of a large number of polymerized homologs and are used as leading materials in various fields including the electronics, biomedical, and aerospace fields. The required properties of synthetic polymer materials are becoming more sophisticated, and the chemical structure of synthetic polymers is becoming more complicated. The current molecular characterization technique (molecular characterization) is not enough to cope with the development speed of such synthetic polymers. In the characteristic analysis of synthetic polymers, it is important to understand not only the average values of structural parameters such as the average molecular weight of each synthesized polymer, but also the physical properties based on their distribution and chain measurements.

  Biological samples are blood, plasma, serum, saliva, pleural effusion, nasal fluid, intracellular fluid, intercellular fluid, lymph fluid, cerebrospinal fluid, bile acid, synovial fluid, pericardial fluid, ascites, stool, intraocular fluid, Sputum, urine, cells, tissues, organs, etc. In addition to low molecular weight metabolites such as amino acids, amines, nucleotides, sugars, and lipids, biological samples include biopolymers such as nucleic acids and proteins, cell membranes, and their degradation products and fragments. The physicochemical properties of individual substances constituting a biological sample vary in polarity, ionization degree (ion charge), boiling point, molecular weight, etc., and the interaction between substances is complicated.

  Here, genomics targeting nucleic acids, proteomics targeting proteins, and metabolomics targeting low molecular weight metabolites comprehensively identify and quantify nucleic acids, proteins, and low molecular weight metabolites contained in biological samples, respectively. These are analytical techniques intended to be used, and these are used in genetic and metabolic studies as omics techniques.

  When analyzing a specific substance in a mixture, usually pretreatment such as filtration, extraction, separation, etc. is performed to remove as much as possible the substance other than the target of analysis, and then instrumental analysis (mass spectrometry, electromagnetic waves) Specific substances are identified and quantified using techniques such as analysis and electroanalysis) and antigen-antibody reactions. When analyzing substances contained in a mixture by the omics technology, each substance is identified and identified according to the case where each substance is separated and purified as much as possible, and then a specific substance in the mixture is analyzed.・ In most cases, quantification is performed.

  Nowadays, electromagnetic wave analysis, mass spectrometry, and the like are widely used for the purpose of qualitatively quantifying an object or examining the physical properties of the object. Among them, electromagnetic wave analysis is a method for investigating the qualitative / quantitative or physical properties of an object by showing the intensity of physical observations as a function of frequency, energy, time, etc., and is the most widely used method of instrumental analysis. ing. Electromagnetic analysis was initially developed with the primary purpose of obtaining chemical information on single or single compounds, so pretreatment such as separation and extraction for each component is required when qualitative and quantitative analysis of mixtures is required. As

  The NMR spectrum method, which is one of electromagnetic wave analyses, is an analysis method using electromagnetic waves in the radio wave region. This NMR spectrum method was developed for the purpose of obtaining chemical information of a single compound, and is a technique for obtaining chemical information such as the number of atoms and molecular structure from the peaks on the spectrum. The NMR spectrum method is non-destructive and excellent in qualitative, quantitative, and reproducibility, so that the application range of analysis and analysis is extended to a mixture. When “simple mixtures” are targeted, certain methods such as quantitative nuclear magnetic resonance (Quantitative Nuclear Magnetic Resonance) and DOSY (Three Dimensional-Diffusion-Ordered NMR Spectroscopy) are used. Achievements have been obtained.

  In the quantitative NMR (qNMR) method, analysis is performed by comparing the integral value of the signal derived from the measurement target substance and the signal derived from the internal standard substance on the NMR spectrum. If a reference substance having a known molecular weight is prepared as an upper standard, it is possible to examine the purity of other compounds in the same solution measured simultaneously by using solution 1H NMR (proton nuclear magnetic resonance). . At present, quantitative NMR has become a highly reliable test method described in food additive official regulations and reference information of herbal pharmacopoeia.

  The DOSY method is a technique for separating a spectrum of a mixed sample composed of a plurality of compounds by utilizing the difference in self-diffusion coefficient of each molecular species, and does not perform pretreatment such as separation / extraction in the mixed sample. A spectrum of each compound is obtained.

  At present, the NMR spectrum method has been extended to biological samples, and in many cases, it is applied to analysis for the purpose of identification and quantification of substances constituting the biological sample. NMR spectroscopy has also been applied to metabolomics. For example, 1H NMR spectroscopy using the pulse sequence CPMG method (Carr-Purcell Meiboom-Gill sequence), which is a kind of spin echo method that can selectively detect only low-molecular substances with high mobility in a sample, is metabolomics. Has become one of the major analytical techniques. The spin echo method and the CPMG method will be briefly described as follows.

  Protons derived from polymers with small mobility in the sample are cross-relaxed (cross relaxation: energy released to magnetic nuclei in nearby molecules), and T2 relaxation time (longitudinal relaxation time) and T2 relaxation time (transverse relaxation time). ) Both become shorter. Since the peak width on the NMR spectrum is inversely proportional to the T2 relaxation time, the peak of the polymer compound (attributed to protons derived from the polymer) becomes wider. For this reason, the high molecular compound itself is not easy to analyze, and the peak is a peak of a low molecular weight compound having high mobility that can be easily identified and quantified (generally a T2 relaxation time is long and a sharp peak). ) Makes it difficult to analyze low molecular weight compounds. A measurement method that solves these problems is the spin echo method. In the spin echo method, a signal is acquired after a short time after the proton is excited, so protons derived from a polymer compound having a short T2 relaxation time such as protein are not detected, and a low molecular weight having a long T2 relaxation time is not detected. Only the peaks are clearly detected. However, in this measurement method, some signals may be inverted depending on conditions. In view of this, the CPMG method has been devised which excites using a special pulse.

JP-T-2002-543392 Special table 2012-500391 gazette

Psychogios N, Hau DD, Peng J, Guo AC, Mandal R, et al. (2011) The Human Serum Metabolome.PLoS ONE 6 (2): e16957.doi: 10.1371 / journal.pone.0016957

Here, when analyzing a mixture sample, especially the problems of the prior art in the case of performing electromagnetic wave analysis, mass spectrometry, etc. on a “complex / heterogeneous mixture” such as a biological sample will be examined.
Hereinafter, <pretreatment> / <identification / quantification of contained substances> generally performed in the analysis of a mixture sample will be described.

<Pretreatment>
(1) Pretreatment for analysis of the mixture includes filtration, extraction and separation.
(2) Pretreatment is a technique originally developed to selectively purify a specific substance or substance group. Therefore, when the pretreatment is performed on the “complex / heterogeneous mixture”, some substances contained in the “complex / heterogeneous mixture” are removed. Naturally, it becomes impossible to identify and quantify the substances removed in the pretreatment process, and the characteristics of the entire “complex and heterogeneous mixture” are lost. Therefore, it is not appropriate to perform the pretreatment for the analysis of “complex and heterogeneous mixture”.

<Identification and quantification of contained substances>
(1) The types of spectroscopic methods that can be identified and quantified without pretreatment such as filtration, extraction, and separation are limited to NMR spectral methods. When performing identification and quantification using NMR spectroscopy without pretreatment, basically analysis based on foreseeing knowledge of the NMR spectrum of a single known compound or the NMR spectrum of a known compound in a “simple mixture” is performed. Is called. In the NMR spectrum method, the molecular structure of a substance is determined based on a shift in the resonance frequency (chemical shift value) of the observation nucleus due to a change in the magnetic field environment in the measurement environment of the observation nucleus (for example, proton).
(2) In a “complex / heterogeneous mixture”, each compound is inhomogeneously assembled and compounded by intersubstance interactions (intermolecular interactions), and the magnetic field in the measurement environment is a “complex / heterogeneous mixture”. ”Indicates a unique distribution. For this reason, it is difficult to accurately identify and quantify each compound in the “complex / heterogeneous mixture” with only such foresight knowledge.
(3) The electromagnetic wave signal generated as a result of the intermolecular interaction in the “complex / heterogeneous mixture” can be NMR spectrum information indicating the unique characteristics of the “complex / heterogeneous mixture”. No analysis focusing on such NMR spectrum information has been performed.
As described above, the NMR spectrum method is also used for identifying and quantifying the components in the mixture without pretreatment, but has the following problems (a) to (c).
(A) In quantitative NMR (qNMR), a “complex / heterogeneous mixture” is not applicable.
(B) The application range of the DOSY method is limited to a mixture of the same kind and several kinds of compounds, and the quantitativeness is poor.
(C) The CPMG method is quantitative with respect to a peak derived from a highly mobile proton, and is widely applied to the analysis of low molecular metabolites in metabolomics. However, in a “complex / heterogeneous mixture”, the magnetic field environment changes inhomogeneously due to interactions between substances, etc., so the substance can be identified and identified based on the chemical shift value of the compound peak obtained from the compound alone. The reliability of the quantitative results is low.
(4) Information on unknown components in the “complex / heterogeneous mixture” cannot be obtained. For example, the technique described in Patent Document 1 relates to NMR spectrum analysis of serum that is a “complex / heterogeneous mixture” and is derived from glucose and a high molecular compound (lipoprotein) that exhibit a complex peak pattern. It can be said that it is suitable for peak analysis. However, this is an identification / quantitative analysis focusing only on a specific substance, information on unknown components cannot be obtained, and information of the entire “complex / heterogeneous mixture” is not used.

  In this way, when analyzing a “complex / heterogeneous mixture” such as a biological sample, it is not appropriate to perform pretreatment, and in the method for identifying and quantifying the contained substances, It is difficult to accurately grasp the characteristics of the “heterogeneous mixture” as a whole.

  On the other hand, a technique for analyzing a mixture without identifying contained substances is also known. However, at present, there is a limit to accurately grasping the characteristics of the whole mixture because an analysis method similar to the analysis method for identification and quantification is used. For example, the technique described in Patent Document 2 uses an NMR spectrum output as an xy trace. For this reason, when targeting biological samples with high concentrations of high molecular compounds such as serum, broad peaks due to protons with low motility derived from high molecular compounds are distributed over a wide band, and low molecular compounds or high molecular compounds Information on minute peaks derived from protons with high mobility cannot be obtained.

Most of the spectroscopic analyzes of “complex and heterogeneous mixtures” in the prior art are based on the assumption that “mixtures are aggregates of pure substances (simple substances and compounds)” and the electromagnetic properties of pure pure substances alone are considered. Use it to identify and quantify pure substances. For this reason, information on substances (known substances and unknown substances) included in the “complex / heterogeneous mixture” that are difficult to detect and identify has not been handled. In the case where the contained pure substance is not identified, changes in the electromagnetic wave signal (for example, time fluctuation) due to intermolecular interaction in the “complex / heterogeneous mixture” were not considered. Furthermore, the conventional analysis of the mixture requires chemical expertise, and has problems in terms of simplicity and speed.
Therefore, an object of the present invention is to solve at least one of the above-described problems in the conventional technique for analyzing a mixture.

According to one aspect of the present invention, there is provided a method for expressing characteristics of a complex / heterogeneous mixture sample including a biological sample collected from a living organism , wherein the complex / heterogeneous mixture sample is applied to a nuclear magnetic resonance (NMR) analyzer. A spectrogram of the FID signal is obtained by repeatedly obtaining a FID signal by applying a frequency analysis over a short period of time over the entire FID signal, and the characteristics of the complex / heterogeneous mixture sample by the calculated spectrogram Expressing.

According to another aspect of the present invention, there is provided a method for evaluating characteristics of a complex / heterogeneous mixture sample including a biological sample collected from an organism , wherein the complex / heterogeneous mixture sample is analyzed by a nuclear magnetic resonance (NMR) analyzer. A FID signal is obtained by applying to the FID signal, and a spectrogram of the FID signal is calculated by repeating frequency analysis over a short time over the entire FID signal, and the complex / heterogeneous mixture is calculated based on the calculated spectrogram Evaluating the properties of the sample.
Here, the attenuation characteristic value of each frequency component of the calculated spectrogram may be further calculated, and the characteristics of the mixture sample may be evaluated based on the calculated attenuation characteristic value of each frequency component.

According to another aspect of the present invention, there is provided a method for identifying an attribute of a complex / heterogeneous mixture sample including a biological sample collected from an organism , wherein the attribute of the complex / heterogeneous mixture sample is unknown. the mixture samples to obtain an FID signal by multiplying the nuclear magnetic resonance (NMR) analyzer, calculating the spectrogram of the FID signal by repeating a short time frequency analysis throughout the FID signal, the from the spectrogram Creating a spectrogram image for the test mixture sample, and creating based on the spectrogram image for the test mixture sample and at least one known mixture sample of the same type as the test mixture sample To identify the attribute of the test mixture sample by comparing with the attribute identification image It includes a.

According to another aspect of the present invention, there is provided a method for identifying an attribute of a complex / heterogeneous mixture sample including a biological sample collected from an organism , wherein the attribute of the complex / heterogeneous mixture sample is unknown. the mixture samples to obtain an FID signal by multiplying the nuclear magnetic resonance (NMR) analyzer, calculating the spectrogram of the FID signal by repeating a short time frequency analysis throughout the FID signal, wherein the numerical values of spectrogram Calculating feature quantities for the test mixture sample using column data; and feature quantities for the test mixture sample and a plurality of known mixtures whose attributes are determined in the same type as the test mixture sample Comparing an attribute identification index set based on the sample to identify the attribute of the test mixture sample.

According to another aspect of the present invention, there is provided a method for identifying an attribute of a complex / heterogeneous mixture sample including a biological sample collected from an organism , wherein the attribute of the complex / heterogeneous mixture sample is unknown. the mixture samples to obtain an FID signal by multiplying the nuclear magnetic resonance (NMR) analyzer, calculating the spectrogram of the FID signal by repeating a short time frequency analysis throughout the FID signals, each of the spectrogram Calculating an attenuation characteristic value of a frequency component, creating a frequency component / attenuation characteristic value graph for the test mixture sample from the attenuation characteristic value of each frequency component, and a frequency component for the test mixture sample / Attenuation characteristic value graph and at least one known mixture whose attributes are the same as those of the test mixture sample By comparing the attribute identifying graph created based on a sample, comprising, identifying attributes of the mixture tested sample.

According to another aspect of the present invention, there is provided a method for identifying an attribute of a complex / heterogeneous mixture sample including a biological sample collected from an organism , wherein the attribute of the complex / heterogeneous mixture sample is unknown. the mixture samples to obtain an FID signal by multiplying the nuclear magnetic resonance (NMR) analyzer, calculating the spectrogram of the FID signal by repeating a short time frequency analysis throughout the FID signals, each of the spectrogram Calculating the attenuation characteristic value of the frequency component, calculating the characteristic amount for the test mixture sample using the attenuation characteristic value of each frequency component, and the characteristic amount for the test mixture sample, Compared to the test mixture sample and the attribute identification index set based on multiple known mixture samples of the same type and attributes. Includes, identifying attributes of the mixture tested sample.

According to yet another aspect of the present invention, a method for processing a FID signal obtained by subjecting a complex and heterogeneous mixture sample comprising a biological sample taken from an organism to a nuclear magnetic resonance (NMR) analyzer is provided. calculating the spectrogram of the FID signal by repeating a short time frequency analysis throughout the FID signals, and the numerical string data of the spectrogram, and at least one of the spectrogram image created from the spectrogram, the complex- Outputting information indicating the characteristics of the heterogeneous mixture sample.

According to yet another aspect of the present invention, a method for processing a FID signal obtained by subjecting a complex and heterogeneous mixture sample comprising a biological sample taken from an organism to a nuclear magnetic resonance (NMR) analyzer is provided. calculating the spectrogram of the FID signal by repeating a short time frequency analysis throughout the FID signal, calculating the attenuation characteristic values of the respective frequency components of the spectrogram, and the attenuation characteristic values of the respective frequency components Outputting at least one of the numerical string data and the frequency component / attenuation characteristic value graph created from the attenuation characteristic value of each frequency component as information indicating the characteristics of the complex / heterogeneous mixture sample. .

Spectrograms (numerical string data and spectrogram images) calculated from FID signals obtained by applying a mixture sample to a nuclear magnetic resonance (NMR) analyzer, and applying the mixture sample to a nuclear magnetic resonance (NMR) analyzer The attenuation characteristic value (numerical string data, frequency component / attenuation characteristic value graph) of each frequency component of the spectrogram calculated from the FID signal obtained by the above is information indicating the characteristics of the mixture sample (that is, unique data). It has been confirmed that it can be. The information (data) hardly loses various kinds of information included in the FID signal, and information on the types and amounts of components (including unknown components) and amounts contained in the mixture sample. It also includes information on the effects of interactions. Further, analysis of the information (data) (including comparative), such as images and sound engineering for pan can be performed using, requires little chemical expertise (including priori knowledge), Moreover, it can be performed with high accuracy. Further, the FID signal can be acquired without destroying the mixed compound sample.

For this reason, according to the method of expressing the characteristics of the mixture sample, it is possible to analyze the mixture sample by utilizing information that has not been used or was not used conventionally.
In addition, according to the method for evaluating the characteristics of the mixture sample, the characteristics of the mixture sample (especially complex / heterogeneous mixture) can be measured with high accuracy, non-destructive, simple, and quick in a completely different manner. It becomes possible to evaluate.
Furthermore, according to the method for identifying the characteristics of the mixture sample, a highly accurate, non-destructive, convenient, and rapid method is used for a mixture sample whose attribute is unknown (particularly a complex or heterogeneous mixture) in a completely different manner. It is possible to identify the attribute.
Furthermore, according to the method of processing an electromagnetic wave signal derived from the mixture sample, the electromagnetic wave signal derived from the mixture sample can be output as data that is relatively easy to analyze and compare.

It is a figure which shows the NMR signal which is an example of the electromagnetic wave signal originating in a mixture sample. It is a block diagram which shows schematic structure of the signal processing apparatus in 1st Embodiment. It is a flowchart which shows an example of the process which the said signal processing apparatus in 1st Embodiment implements. It is a flowchart which shows an example of the process which identifies the attribute of a test mixture sample in 1st Embodiment. It is a block diagram which shows schematic structure of the signal processing apparatus in 2nd Embodiment. It is a flowchart which shows an example of the process which the said signal processing apparatus in 2nd Embodiment implements. It is a flowchart which shows an example of the process which identifies the attribute of a test mixture sample in 2nd Embodiment. It is a block diagram which shows schematic structure of the signal processing apparatus (apparatus which processes an electromagnetic wave signal) in 3rd Embodiment. It is a figure which shows an example of a spectrogram image (actually a color image). It is a flowchart which shows an example of the process which the said signal processing apparatus in 3rd Embodiment implements. It is a flowchart which shows an example of the process which identifies the attribute of a test mixture sample in 3rd Embodiment. It is a block diagram which shows an example of the attribute identification apparatus (apparatus which identifies the attribute of a test mixture) in 3rd Embodiment. It is a flowchart which shows the other example of the process which identifies the attribute of a test mixture sample in 3rd Embodiment. It is a block diagram which shows the other example of the attribute identification apparatus in 3rd Embodiment. It is a block diagram which shows schematic structure of the signal processing apparatus in 4th Embodiment. It is a flowchart which shows an example of the process which the said signal processing apparatus in 4th Embodiment implements. It is a flowchart which shows an example of the process which identifies the attribute of the test mixture sample in 4th Embodiment. It is a block diagram which shows an example of the attribute identification device in 4th Embodiment. It is a flowchart which shows the other example of the process which identifies the attribute of the test mixture sample in 4th Embodiment. It is a block diagram which shows the other example of the attribute identification apparatus in 4th Embodiment. It is a figure which shows an example of the spectrogram image of the NMR signal of rat plasma. It is a figure which shows an example of the frequency component / time constant graph of the NMR signal of rat plasma. It is a figure which shows an example of the principal component analysis result with respect to the spectrogram (numerical string data) of the NMR signal of rat plasma. It is a figure which shows an example of the PLS discriminant analysis result with respect to the spectrogram (numerical string data) of the NMR signal of rat plasma. It is a figure which shows an example of the principal component analysis result with respect to the time constant (numerical string data) of each frequency component of the NMR signal of rat plasma. It is a figure which shows an example of the PLS discriminant analysis result with respect to the time constant (numerical string data) of each frequency component of the NMR signal of rat plasma. It is a figure which shows an example of the spectrogram image of the NMR signal of a mouse cell extract. It is a figure which shows an example of the frequency component / time constant graph of the NMR signal of a mouse cell extract. It is a figure which shows an example of the principal component analysis result with respect to the spectrogram (numerical string data) of the NMR signal of a mouse cell extract. It is a figure which shows an example of the PLS discriminant analysis result with respect to the spectrogram (numerical string data) of the NMR signal of a mouse cell extract. It is a figure which shows an example of the principal component analysis result with respect to the time constant (numerical string data) of each frequency component of the NMR signal of a mouse cell extract. It is a figure which shows an example of the PLS discriminant analysis result with respect to the time constant (numerical string data) of each frequency component of the NMR signal of a mouse cell extract. It is a figure which shows an example of the color display of the NIR signal of rat plasma. It is a figure which shows the other example of the identification result using SVM with respect to the RGB value (numerical string data) calculated from the NIR signal of rat plasma.

First, the outline of the present invention will be described.
An important feature of the present invention is that the mixture is not regarded as an aggregate of individual substances but as a whole structural unit (integrated body). In the present invention, an electromagnetic wave signal derived from the mixture sample is acquired from the mixture sample, and the acquired electromagnetic wave signal is handled as a unique electromagnetic wave signal representing one structural unit (integrated body). That is, in the present invention, the electromagnetic wave signal derived from the mixture sample is not used as a composite signal of each component constituting the mixture sample (that is, an aggregate of the electromagnetic wave signals of each component), but one unique to the mixture sample. Treated as an electromagnetic wave signal (single data). By such handling, information on unknown components in the mixture sample and information on the influence of interstitial interaction can be included in the analysis data.

  The present invention provides a new method for expressing the characteristics of a mixture sample on the premise of the above important features. This expression method enables analysis of a mixture by utilizing information that has not been used or was not used before, and for example, those who have little chemical expertise (including foresight) It is possible to evaluate the characteristics of the mixture sample and to identify the attributes of the mixture sample.

  Here, the “electromagnetic wave signal derived from the mixture sample” includes an electromagnetic wave signal obtained by causing a predetermined energy or chemical reaction material including electromagnetic waves and heat to act on the mixture sample, an electromagnetic wave signal emitted from the mixture sample itself, and the like. For example, an electromagnetic wave signal (mainly a time domain signal) acquired (observed) by applying the mixture sample to various analyzers corresponds to an electromagnetic wave signal derived from the mixture sample. Examples of the various analyzers include a nuclear magnetic resonance (NMR) analyzer, a mass analyzer, a chromatograph device, an absorption analyzer, and an emission analyzer.

Hereinafter, embodiments and examples of the present invention will be described with reference to the accompanying drawings.
In the following description, the electromagnetic wave signal derived from the mixture sample is mainly an NMR signal (NMR FID signal), that is, an electromagnetic wave signal obtained by applying the mixture sample to an NMR analyzer. However, as described above, the present invention is not limited to this.

<First Embodiment>
As described above, conventional analysis of a mixture requires chemical expertise, and information on the structure and function of a “complex / heterogeneous mixture” contained in an electromagnetic wave signal derived from a mixture sample. Department is not used.

In the first embodiment, an electromagnetic wave signal derived from the mixture sample is acquired from the mixture sample, audible range data is generated from the acquired electromagnetic wave signal, and the generated audible range data is converted into sound and output. The characteristics of the mixture sample are expressed by the output sound.
The audible range data is basically generated by frequency-converting (for example, frequency shifting) the acquired electromagnetic wave signal into an audible range (human audible range). Therefore, here, a case will be described in which audible range data is generated by frequency-converting the acquired electromagnetic wave signal to an audible range. However, the present invention is not limited to this, and depending on the acquired electromagnetic wave signal, the electromagnetic wave signal itself can be used as the audible range data (in this case, audible range data can be generated from the electromagnetic wave signal derived from the mixture sample). include).

  An electromagnetic wave signal (eg, NMR signal) derived from a mixture sample is characterized by amplitude, frequency, and waveform as well as “sound”. Considering “the three elements of sound”, “sound volume” is determined by “amplitude”, and “pitch” is determined by “frequency”. “Tone” is characterized by “waveform”, that is, “time change of frequency”, and this “timbre” can distinguish sounds of various attributes (for example, piano sounds and violin sounds). Is possible. The human auditory sense has the ability to highly discriminate “tone colors”. Of course, the sound can be identified by the device.

  Therefore, by converting the electromagnetic wave signal derived from the mixture sample into sound and outputting it, it is possible to evaluate the characteristics of the mixture sample using human hearing. For this reason, even those who do not have chemical expertise should recognize intuitively and accurately that there are similarities and differences between mixture samples (especially complex and heterogeneous mixtures). Is possible. For example, if the “sound” of two mixture samples can be recognized as different “sounds”, both can be easily identified.

  In addition, since the electromagnetic wave signal derived from the mixture sample is only converted into sound and output, the possibility that a part of various information of the mixture sample contained in the electromagnetic wave signal is lost and not used is greatly suppressed. .

  Furthermore, since the electromagnetic wave signal derived from the mixture sample is handled as sound, general-purpose acoustic engineering technology can be used for the analysis of the electromagnetic wave signal. For example, when it is clear that the electromagnetic wave signal (or the audible range data) contains a signal derived from impurities or a solvent or external noise, these can be easily removed, and a mixture sample ( In particular, it can be expected to improve the analysis accuracy of complex and heterogeneous mixtures.

In the first embodiment, for example, the characteristics of the mixture sample are evaluated according to the following steps.
(1) First step (acquisition of electromagnetic wave signal derived from the mixture sample)
An electromagnetic wave signal (observation signal) derived from the mixture sample is acquired and stored as it is (that is, in RAW format). For example, an electromagnetic wave signal (for example, FID signal of NMR or terahertz wave) derived from the mixture sample is obtained by applying the mixture sample to a predetermined analyzer or the like. The acquired electromagnetic wave signal is stored as it is without any processing, and the information of the mixture sample included in the electromagnetic wave signal is not lost. FIG. 1 shows an example of an electromagnetic wave signal (NMR signal) derived from the mixture sample.

(2) Second step (processing of electromagnetic wave signal derived from the mixture sample)
The electromagnetic wave signal acquired (stored) in the first step is frequency-converted into an audible range (about 20 Hz to 20 kHz) to generate audible range data, and the generated audible range data is converted into sound and output. However, as described above, the electromagnetic wave signal acquired in the first step may be directly used as audible range data.

In the present embodiment, the electromagnetic wave signal processing (audible range data generation and output processing) is performed by the signal processing device 1A.
FIG. 2 is a block diagram showing a schematic configuration of the signal processing apparatus 1A.
As shown in FIG. 2, the signal processing device 1A includes a processing unit 2A and an output unit 3A.
The processing unit 2A receives an electromagnetic wave signal derived from the mixture sample, and converts the frequency of the input electromagnetic wave signal to an audible range to generate audible range data. The generated audible range data is stored (recorded) in an uncompressed audio format (for example, an uncompressed WAV format) and output to the output unit 2A.
The output unit 2A converts the audible range data output from the processing unit 2A into sound and outputs the sound.

FIG. 3 is a flowchart illustrating an example of processing performed by the signal processing device 1A.
In FIG. 3, in step S1, the input electromagnetic wave signal is frequency-converted into an audible range using, for example, a predetermined audible range conversion coefficient vector to generate audible range data.
In step S2, the audible range data generated in step S1 is stored (recorded) in an uncompressed audio format (for example, an uncompressed WAV format).
In step S3, the audible range data is D / A converted, for example, and output as sound.

  Here, the output unit 2A may be configured separately from the signal processing device 1A. In this case, the signal processing unit 1A performs the processes of S1 and S2, and outputs the audible range data stored (recorded) in the uncompressed audio format in step S2. Then, a sound reproduction device (corresponding to the output unit 2A) separate from the signal processing device 1A inputs the audible range data, performs the process of step S3, converts the audible range data into sound, and outputs it. To do. For example, when a computer is used as the audio reproducing device, information for DA conversion (WAV format header information including the number of bits per sample, the sampling frequency, and the number of channels) is given to the computer. The audible range data is output as sound from the sound output of the computer by a sound library function prepared by the OS.

(3) Third step (characteristic evaluation of the mixture sample)
The sound output from the signal processing apparatus 1 </ b> A hardly loses information on the mixture sample, and can be information indicating the characteristics of the mixture sample with high accuracy. Therefore, it is possible to evaluate the characteristics of the mixture sample by the output sound. The characterization of the mixture sample is basically performed using human hearing. Of course, the characteristics of the mixture sample may be evaluated by processing or analyzing the audible range data before being converted into sound by acoustic engineering technology.

  The sound output from the signal processing apparatus 1A and / or the analysis result thereof can be used as an index for evaluating the similarity or difference between the mixture samples. For example, for a predetermined type of mixture whose attribute classification is not set, the attribute classification can be set based on the sound of the corresponding mixture sample. In addition, a new attribute category can be set (reset) for a predetermined type of mixture whose attribute category is set according to a predetermined classification criterion. Furthermore, the sound output from the signal processing apparatus 1A can be used to identify the attribute of the mixture sample whose attribute is unknown. The attribute identification of such a mixture sample will be described below as a fourth step.

(4) Fourth step (attribute identification of a mixture sample with unknown attributes)
Identification of the attributes of a mixture sample with unknown attributes (hereinafter referred to as “test mixture sample”) is performed using a mixture sample (hereinafter referred to as “known mixture sample”) that has the same attributes as the test mixture sample. Do. Specifically, by comparing the sound based on the audible range data for the test mixture sample and the sound based on the attribute identification sound data generated based on the known mixture sample, the test mixture sample Attributes can be identified.

FIG. 4 is a flowchart showing a process for identifying the attribute of the test mixture.
In FIG. 4, step S11 obtains an electromagnetic wave signal derived from each of the test mixture sample and at least one of the known mixture samples. The acquisition of the electromagnetic wave signal is performed in the same manner as in the first step. Here, when one or a plurality of attributes have been determined for the mixture corresponding to the test mixture sample, an electromagnetic wave signal derived from the known mixture sample corresponding to (belonging to) each determined attribute is acquired. It is preferable to do this. That is, it is preferable to acquire the electromagnetic wave signal derived from the known mixture sample of the same number or more as the number of attributes determined.

  In step S12, each electromagnetic wave signal acquired in step S11 is frequency-converted into an audible range to generate audible range data, and each generated audible range data is recorded in an uncompressed audio format. The process of step S12 is performed in the same manner as the second process.

  In step S13, each audible range data generated in step S12 is converted into sound and output. Specifically, the sound based on the audible range data for the test mixture sample and the sound based on the audible range data for the at least one known mixture sample are output so that they can be compared. For example, a sound based on the audible range data for the test mixture sample and a sound based on the audible range data for the known mixture sample are alternately output.

  In step S14, the sounds output in step S13 are compared to identify the attributes of the test mixture sample. Specifically, determining whether a sound based on audible range data for the test mixture sample is similar to or most similar to a sound based on audible range data for at least one of the known mixture samples To identify the attributes of the test mixture sample. If the sound based on the audible range data for the test mixture sample is determined not to be similar to any of the sounds based on the audible range data for the at least one known mixture sample, the test mixture Leave sample attributes unknown.

  Here, the process of step S11, S12 about at least one said known mixture sample may be implemented separately from this flow. For example, after generation and recording of audible range data for at least one of the known mixture samples, the processing of steps S11 and S12 is performed on the test mixture sample, and the processing of step S13 is performed in step S14. You may make it identify the attribute of a test mixture sample.

  In addition, the identification of the attribute of the test mixture sample in step S14 is basically performed using human hearing. That is, each electromagnetic wave signal acquired in step S11 is input to the signal processing device 1A, and each sound output from the signal processing device 1A is compared by a predetermined person using its auditory sense. Identify However, the present invention is not limited to this, and an attribute identification device may be manufactured, and the attribute identification device may be configured to perform the processes of steps S12 to S14. In this case, the attribute identification device is configured to identify the attribute of the test mixture sample using, for example, an acoustic engineering technique.

Second Embodiment
In the second embodiment, an electromagnetic wave signal derived from the mixture sample is obtained from the mixture sample, and the obtained electromagnetic wave signal is frequency-converted into a visible region (human visible region) to obtain visible region data. The characteristics of the mixture sample are expressed by the output color.

  In this way, by converting the electromagnetic wave signal derived from the mixture sample into a color and outputting it, it becomes possible to evaluate the characteristics of the mixture sample using human vision (color vision). For this reason, as in the first embodiment, even those who do not have chemical expertise can intuitively and accurately know that there are similarities and differences between “complex and heterogeneous mixtures”. It becomes possible to recognize. In addition, since the electromagnetic wave signal is merely converted into a color and output, the possibility that a part of the information included in the electromagnetic wave signal is not used is greatly suppressed. Furthermore, since it is possible to use general-purpose optical engineering techniques for the analysis of electromagnetic wave signals derived from the mixture sample, an improvement in the analysis accuracy of the “complex / heterogeneous mixture” sample can be expected.

In 2nd Embodiment, the characteristic of a mixture sample is evaluated according to the following processes.
(1) First step (acquisition of electromagnetic wave signal derived from the mixture sample)
This is the same as the first step in the first embodiment.

(2) Second step (processing of electromagnetic wave signal derived from the mixture sample)
First obtained in step (saved) electromagnetic wave signal (including frequency shift) (RAW format) frequency conversion into the visible range (4.0 × ¹⁰ 14 Hz~7.5 × about ¹⁰ 14 Hz) and to visible As data, the visible range data is converted into sound and output.

In the present embodiment, the processing of the electromagnetic wave signal is performed by the signal processing device 1B.
FIG. 5 is a block diagram showing a schematic configuration of the signal processing apparatus 1B.
As illustrated in FIG. 5, the signal processing device 1B includes a processing unit 2B and an output unit 3B.
The processing unit 2B receives an electromagnetic wave signal (RAW format) derived from the mixture sample, and converts the frequency of the input electromagnetic wave signal to a visible range to generate visible range data. The generated visible range data is output to the output unit 2B.
The output unit 2B converts the visible range data output from the processing unit 2B into a color and outputs the color.

FIG. 6 is a flowchart illustrating an example of processing performed by the signal processing device 1B.
In FIG. 6, in step S21, the input electromagnetic wave signal (RAW format) is frequency-converted into a visible range using, for example, a predetermined visible range conversion coefficient vector, and visible range data is generated and recorded.

In step S22, the RGB value of the visible range data generated in step S21 is calculated using an RGB bandpass filter. The RGB band pass filter is a narrow band pass filter having a characteristic of transmitting red band, green band and blue band signals. In this embodiment, the center frequency of the red band is 4.3 × 10 ¹⁴ Hz, the center frequency of the green band is 5.5 × 10 ¹⁴ Hz, and the center frequency of the blue band is 6.9 × 10 ¹⁴ Hz. Hz.

The RGB values of the visible range data calculated in step S22 may be converted into other color system values. For example, using the L ^* a ^* b ^* conversion table or HSV conversion table, the RGB values of the visible range data calculated in step S22 are converted into L ^* a ^* b ^* values (L ^* a ^* b ^* conversion). Or may be converted into an HSV value (HSV conversion).
In step S23, the visible range data is output as a color.

  Here, the output unit 2B may be configured separately from the signal processing device 1B. In this case, the signal processing unit 1B performs the processing of steps S21 and S22 and stores (records) the visible range data. Then, a display device (corresponding to the output unit 2B) separate from the signal processing device 1B inputs the visible range data, performs the process of step S23, and outputs the visible range data as a color.

(3) Third step (characteristic evaluation of the mixture sample)
Similar to the sound in the first embodiment, the color output from the signal processing device 1B can be information that accurately indicates the characteristics of the mixture sample. Therefore, it is possible to evaluate the characteristics of the mixture sample according to the output color. That is, the color output from the signal processing device 1B can be used as an index for evaluating the similarity or difference between the mixture samples, for example, and the attribute classification of the mixture can be set (or reset). And can be used to identify attributes of a mixture sample whose attributes are unknown.
Of course, you may evaluate the characteristic of the said mixture sample by processing or analyzing the said visible range data before converting into a color.

(4) Fourth step (attribute identification of a mixture sample with unknown attributes)
Also in the present embodiment, as in the first embodiment, the attribute of the test mixture sample whose attribute is unknown is identified using a known mixture sample whose attribute is determined in the same type as the test mixture sample. Specifically, by comparing the color based on the visible range data for the test mixture sample with the color based on the color data for attribute identification generated based on the known mixture sample, the test mixture sample Attributes can be identified.

FIG. 7 is a flowchart showing a process for identifying attributes of the test mixture.
In FIG. 7, step S31 is the same as step S11 of FIG. That is, an electromagnetic wave signal (RAW format) derived from each of the test mixture sample and at least one known mixture sample is acquired.

  In step S32, each electromagnetic wave signal (RAW format) acquired in step S31 is frequency-converted into a visible range to generate visible range data, and an RGB value of each generated visible range data is calculated. The processing in step S32 is performed in the same manner as in the second step.

  In step S33, each visible range data is output as a color using the RGB values. Specifically, the color based on the visible range data for the test mixture sample and the color based on the visible range data for at least one known mixture sample are output in a comparable manner. For example, the color based on the visible range data for the test mixture sample and the color based on the visible range data for the known mixture sample are simultaneously displayed on one display screen.

  In step S34, the colors output in step S33 are compared to identify the attributes of the test mixture sample. Specifically, determining whether the color based on the visible range data for the test mixture sample is similar to or most similar to the color based on the visible range data for the at least one known mixture sample To identify the attributes of the test mixture sample. If it is determined that the color based on the visible range data for the test mixture sample is not similar to any of the colors based on the visible range data for the at least one known mixture sample, the test mixture Leave sample attributes unknown.

  Here, the processing of steps S32 and S33 for at least one of the known mixture samples may be performed separately from the present flow. For example, after making visible state data of at least one known mixture sample ready to be output as a color, the processing of steps S32 and S33 is performed on the test mixture sample, and in step S34, the test mixture sample is analyzed. You may make it identify an attribute.

  In addition, the identification of the attribute of the test mixture sample in step S34 is basically performed using human vision (color vision). That is, each electromagnetic wave signal acquired in step S31 is input to the signal processing device 1B, and each color output from the signal processing device 1B is compared by a predetermined person using its visual sense (color vision), and the test mixture sample Identify the attributes. However, the present invention is not limited to this, and an attribute identification device may be manufactured and the attribute identification device may perform steps S32 to S34 as in the first embodiment.

  In the second embodiment, it is also possible to create matrix data of the RGB values and perform analysis by pattern recognition using multivariate analysis or the like. In this case, a method similar to that of the third embodiment described below can be used.

<Third Embodiment>
For example, in the conventional NMR spectrum method, an NMR signal (FID signal) corresponding to an electromagnetic wave signal derived from a mixture sample is Fourier-transformed, and then an analysis focusing on the signal intensity distribution in each frequency component on the frequency spectrum is performed. ing. For this reason, the time resolution is sacrificed for the purpose of improving the frequency resolution (that is, time information is lost). Also, frequency components with low signal strength are hardly considered.

  In the third embodiment, paying attention to the fact that NMR signals derived from nuclei having different motility have different decay times, the frequency included in the NMR signal and the signal intensity information associated with the time are utilized in a balanced manner. Specifically, in the third embodiment, an electromagnetic wave signal derived from the mixture sample is acquired from the mixture sample, a spectrogram of the acquired electromagnetic wave signal is calculated, and the characteristics of the mixture sample are expressed by the calculated spectrogram. .

  In this way, by expressing the characteristics of the mixture sample by the spectrogram of the electromagnetic wave signal derived from the mixture sample, for example, comprehensive information on the structure and function of the “complex / heterogeneous mixture” contained in the NMR signal can be obtained. Analysis is possible, and significant improvement in accuracy can be expected in the characterization of “complex and heterogeneous mixtures”. This is not limited to the NMR signal, but can be said to be common to various electromagnetic wave signals derived from the mixture sample.

In the third embodiment, the characteristics of the mixture sample are evaluated according to the following steps.
(1) First step (acquisition of electromagnetic wave signal derived from the mixture sample)
This is the same as the first step in the first and second embodiments.

(2) Second step (processing of electromagnetic wave signal derived from the mixture sample)
The electromagnetic wave signal acquired (stored) in the first step is subjected to short-time frequency analysis to calculate a spectrogram (time frequency expression) of the electromagnetic wave signal.
The short-time frequency analysis is a method mainly used for the analysis of “sound” and “speech”, and the time change of the frequency component of the entire signal can be known.
The spectrogram refers to the result of calculating the frequency spectrum by short-time frequency analysis, that is, passing a composite signal such as a time-frequency signal through a window function, and can be output as a spectrogram image or numerical sequence data It is. The spectrogram image may be one in which the intensity (amplitude) of the signal component is displayed in brightness or color on a two-dimensional graph (time / frequency), for example. It can also be data.

In the present embodiment, the electromagnetic wave signal processing (spectrogram calculation processing) is performed by the signal processing device 1C.
FIG. 8 is a block diagram showing a schematic configuration of the signal processing apparatus 1C.
As illustrated in FIG. 8, the signal processing device 1C includes a processing unit 2C and an output unit 3C.
The processing unit 2C inputs an electromagnetic wave signal derived from the mixture sample, and calculates a spectrogram of the electromagnetic wave signal (that is, a spectrogram of the mixture sample) by repeating frequency analysis for a short time throughout the inputted electromagnetic wave signal. To do. As the short-time frequency analysis, for example, a short-time Fourier transform is applicable. However, the present invention is not limited to this, and wavelet transform or the like may be used.

  The output unit 3C outputs at least one of the numerical sequence data of the spectrogram calculated by the processing unit 2C and the spectrogram image created from the calculated spectrogram. FIG. 9 shows an example of the spectrogram image (actually a color image).

FIG. 10 is a flowchart illustrating an example of a process (the spectrogram calculation process) performed by the signal processing apparatus 1C.
As shown in FIG. 10, when the electromagnetic wave signal is input, the signal processing apparatus 1C performs processing for cutting out a target section using a window function and calculating a power spectrum from the Fourier transform of the waveform. The entire electromagnetic wave signal is processed by iterating while shifting along. Then, the signal processing apparatus 1C outputs at least one of the calculated spectrogram value string data and the spectrogram image.

Specifically, when the signal processing apparatus 1C uses, for example, a Hamming window (window length: N) as the window function, the N samples starting from the “τ _s ” sample (initial value is 0) of the electromagnetic wave signal The waveform obtained by multiplying the above range by the Hamming window is Fourier transformed. Here, N is set to a power of 2 when the fast Fourier transform is used. The signal processing apparatus CA calculates the power spectrum by squaring the norm of each sample of the complex Fourier spectrum obtained by the Fourier transform, and the calculated power spectrum is stored as “τ _s / T _d ” in a two-dimensional array for storage. Assign (store) the column. Next, while the signal processing device 1C updates the start point of the cutout by adding the T _{d (window} shift length) sample "tau _s", "tau _s" + N is less than the number of samples of said electromagnetic wave signal ( The same process is repeated until the end of the waveform is reached (steps S41 to S46).
As a result, a spectrogram of the electromagnetic wave signal (that is, a spectrogram for the mixture sample) is calculated, and a power spectrum at each “τ _s / T _d ” is stored in the two-dimensional array.

When the signal processing apparatus 1C outputs the numerical sequence data of the calculated spectrogram, the two-dimensional array storing the power spectrum at each “τ _s / T _d ” is converted into binary data or ASCII data ( For example, it is output as CSV format) (step S47). On the other hand, when outputting the calculated spectrogram as an image (spectrogram image), the signal processing unit 1C uses each element of the two-dimensional array storing the power spectrum at each “τ _s / T _d ” as the HSV color specification. System value conversion (hue conversion) is performed (step S48). For example, H (hue) is converted by applying a coefficient so that the power spectrum value is 0, blue (−120 °), and yellow (−60 °) at maximum, and S (saturation), the power spectrum. Regardless of the value, a constant (constant) is assumed, and V (lightness) is a value proportional to the value of the power spectrum.

(3) Third step (characteristic evaluation of the mixture sample)
The numerical string data of the spectrogram and the spectrogram image output from the signal processing device 1C have almost no information on the mixture sample, and can be information indicating the characteristics of the mixture sample with high accuracy. For example, on the frequency axis, information on the types of substances contained in a mixture sample and interactions between substances can be obtained, and on the time axis, information on interactions between substances can be obtained in addition to the physical properties of the entire mixture sample. It is done. In the spectrogram image, the difference in signal intensity at each point can be easily grasped by displaying the difference in signal intensity at each point in “hue”.

  Therefore, it is possible to evaluate the characteristics of the mixture sample by analyzing the numerical string data of the spectrogram and the spectrogram image. The analysis of the numerical sequence data of the spectrumgram or the spectrum image, in other words, the characteristic evaluation of the mixture sample uses, for example, a known statistical method (multivariate analysis or the like) or an image engineering method with a predetermined apparatus. Or by a predetermined person using the statistical technique or vision (color vision).

The spectrogram (numerical string data or spectrogram image) of the electromagnetic wave signal and / or the analysis result thereof can be used as an index for evaluating the similarity or difference between the mixture samples. For example, for a given type of mixture whose attribute category is not set, the attribute category can be set by analyzing spectrograms (numerical string data and spectrogram images) of the corresponding mixture samples. It becomes.
Similarly, a new attribute category can be set (reset) for a mixture of a predetermined type whose attribute category is set according to a predetermined classification criterion.
Furthermore, the spectrogram (numerical string data or spectrogram image) of the electromagnetic wave signal can be used to identify the attribute of the mixture sample whose attribute is unknown. The attribute identification of such a mixture sample will be described below as a fourth step.

(4) Fourth step (attribute identification of a mixture sample with unknown attributes)
Also in the present embodiment, as in the first and second embodiments, the identification of the attribute of the test mixture sample whose attribute is unknown uses a known mixture sample whose attribute is the same as that of the test mixture sample. Do it. Specifically, the attribute of the test mixture sample is identified by comparing the spectrogram for the test mixture sample with the spectrogram for the known mixture sample or an index based thereon. Hereinafter, the case where the spectrogram image is used and the case where the spectrogram numeric string data is used will be described separately.

(4-1) When Using the Spectrogram Image FIG. 11 is a flowchart showing a process for identifying the attribute of the test mixture sample using the spectrogram image.
In FIG. 11, step S51 is the same as step S11 of FIG. That is, an electromagnetic wave signal derived from each of the test mixture sample and at least one known mixture sample is acquired.

  In step S52, a spectrogram (time frequency expression) is calculated by repeating short-time frequency analysis over the entire electromagnetic wave signal acquired in step S51, and a spectrogram image of each electromagnetic wave signal is created. That is, a spectrogram image for the test mixture sample and a spectrogram image for at least one known mixture sample are created. The spectrogram image is created in the same manner as in the second step.

  In step S53, the spectrogram images created in step S52 are compared to identify attributes of the test mixture sample. Specifically, it is determined whether the spectrogram image for the test mixture sample is similar to or most similar to the spectrogram image for the at least one known mixture sample, thereby Identify attributes. If it is determined that the spectrogram image for the test mixture sample is not similar to any of the spectrogram images for the at least one known mixture sample, the attribute of the test mixture sample remains unknown. To do.

  Here, the process of step S51, S52 with respect to at least one said known mixture sample may be implemented separately from this flow. For example, after creating a spectrogram image of at least one of the known mixture samples, the processing of steps S51 and S52 is performed on the test mixture sample, and the attributes of the test mixture sample are identified in step S53. You may do it.

In addition, the identification of the attribute of the test mixture sample in step S53 can be performed by human vision (color vision) or a predetermined device. When performed by human vision (color vision), each electromagnetic wave signal acquired in step S51 is input to the signal processing device 1C, and each spectrogram image output from the signal processing device 1C is displayed by a predetermined person. The attributes of the test mixture sample are identified by comparison using (color vision).
On the other hand, when it is performed by a predetermined apparatus, an apparatus for identifying the attribute of the mixture sample (hereinafter referred to as “attribute identification apparatus”) having the following configuration is manufactured, and the attribute identification apparatus is the step S52 in FIG. , S53 is performed.

  In the flowchart shown in FIG. 11, the spectrogram image for the test mixture sample is compared with the spectrogram image for at least one known mixture sample, but the spectrogram for the at least one known mixture sample is compared. Instead of an image, an attribute identification image created for each attribute based on these may be used.

FIG. 12 is a block diagram showing a schematic configuration of an attribute identification device 10A for identifying attributes of the test mixture sample using the spectrogram image.
As illustrated in FIG. 12, the attribute identification device 10A includes a signal processing unit 11A, an attribute identification unit 12A, and an identification result output unit 13A.

  The signal processing unit 11A has a function similar to that of the above-described signal processing device 1C, and performs processing corresponding to step S52 in FIG. That is, the signal processing unit 11A inputs an electromagnetic wave signal derived from each of the test mixture sample and the at least one known mixture sample, and repeats a short-time frequency analysis over each of the input electromagnetic wave signals. The spectrogram of the electromagnetic wave signal is calculated. Then, the signal processing unit 11A creates a spectrogram image for the test mixture sample and a spectrogram image for at least one known mixture sample, and outputs them to the attribute identification unit 12A.

  The attribute identifying unit 12A performs a process corresponding to step S53 in FIG. 11 using, for example, an image engineering technique. That is, the attribute identification unit 12A compares the spectrogram image of the test mixture sample with the spectrogram image of at least one known mixture sample, identifies the attribute of the test mixture sample, and determines the identification result. It outputs to the identification result output part 13A. The identification result includes an unknown attribute.

  The identification result output unit 13A outputs the identification result obtained by the attribute identification unit 12A. For example, the identification result output unit 13A displays the identification result on a display unit (not shown). Thereby, the user of 10 A of attribute identification apparatuses can grasp | ascertain the attribute of the said test mixture sample.

(4-2) When using the spectrogram numeric string data FIG. 13 is a flowchart showing a process of identifying attributes of the test mixture sample using the spectrogram numeric string data.
In FIG. 13, in step S61, an electromagnetic wave signal derived from each of the test mixture sample and the plurality of known mixture samples is acquired. The acquisition of the electromagnetic wave signal is performed in the same manner as in the first step.

  In step S62, the spectrogram (time frequency expression) of each electromagnetic wave signal is calculated by repeating the short-time frequency analysis over the entire electromagnetic wave signal acquired in step S61. The calculation of the spectrogram is performed in the same manner as in the second step. Thus, spectrogram value string data for the test mixture sample and spectrogram value string data for the plurality of known mixture samples are obtained.

  In step S63, a multivariate analysis is performed on the spectrogram value sequence data for the plurality of known mixture samples, and the spectrogram value sequence data for the plurality of known mixture samples is separated into data groups for each attribute. Create a composite variable for. Examples of the multivariate analysis include principal component analysis, PLS discriminant analysis (PLS-DA), SIMCA method, and SVM (Support Vector Machine) method.

  In step S64, a characteristic amount for each of the plurality of known mixture samples, that is, a composite variable value (score) of the numerical sequence data of the spectrogram is calculated.

In step S65, an attribute identification index is set based on the feature quantities of the plurality of known mixture samples. Specifically, the feature amounts (synthetic variable values) for each of the plurality of known mixture samples calculated in step S64 are grouped for each attribute, and the feature amounts for the known mixture sample for each grouped attribute are grouped. The attribute identification index is set based on (combined variable value). The attribute identification index is set with a predetermined range for each attribute. Each of the attribute identification indicators includes a common feature of the known mixture sample having the same attribute, that is, a feature corresponding to each attribute and / or a feature representing each attribute.
Here, the attribute identification index may be a numerical value or a figure or table.

  In step S66, a characteristic amount of the test mixture sample, that is, a composite variable value of the numerical sequence data of the spectrogram is calculated.

  In step S67, the feature amount (synthetic variable value) for the sample mixture sample calculated in step S66 is compared with the attribute identification index set in step S65 to determine the attribute of the sample mixture sample. Identify. Specifically, it is determined whether the synthetic variable value for the test mixture sample corresponds to (includes) the attribute identification index set for each attribute, and the attribute of the test mixture sample Identify In addition, when the synthetic variable value about the said test mixture sample does not correspond to any of the said attribute identification parameter | index (it is not contained), the attribute of the said test mixture sample is made into unknown. Moreover, the identification of the attribute of the test mixture sample in step S63 can be performed by a predetermined person or a predetermined apparatus, similarly to step S53 of FIG.

  Here, the processes in steps S61 and S62 and the processes in steps S63 to S65 for the plurality of known mixture samples may be performed separately from the present flow. For example, after the attribute identification index is set, the processing of steps S61, S62, and S66 is performed on the test mixture sample, and the attributes of the test mixture sample are identified in step S67. Good.

FIG. 14 is a block diagram showing a schematic configuration of an attribute identification device 10B that identifies the attribute of the test mixture sample using the numerical string data of the spectrogram. The attribute identification device 10B performs the processes of steps S62 to S67 in FIG.
As shown in FIG. 14, the attribute identification device 10B includes a signal processing unit 11B, an index setting unit 20B, an attribute identification unit 12B, and an identification result output unit 13B.

  The signal processing unit 11B has a function similar to that of the above-described signal processing device 1C, and performs processing corresponding to step S62 in FIG. That is, the signal processing unit 11B inputs the electromagnetic wave signal derived from each of the test mixture sample and the plurality of known mixture samples, and repeats the short-time frequency analysis over the entire input electromagnetic wave signal to repeat each electromagnetic wave. Calculate the spectrogram of the signal. Then, the signal processing unit 11B outputs spectrogram value sequence data for the plurality of known mixture samples to the index setting unit 20B, and outputs spectrogram value sequence data for the test mixture sample to the attribute identification unit 12B. .

  The index setting unit 20B performs a process corresponding to steps S63 to S65 in FIG. That is, the index setting unit 20B performs multivariate analysis on the spectrogram value sequence data of the plurality of input mixture samples, and creates the composite variable, and each of the plurality of known mixture samples. A feature amount (composite variable value) is calculated. Then, the index setting unit 20B groups the calculated feature values (synthetic variable values) for each of the plurality of known mixture samples for each attribute, and the feature values for the known mixture sample for each grouped attribute. The attribute identification index is set based on (combined variable value). The index setting unit 20B outputs the created composite variable and the set attribute identification index to the attribute identification unit 12B.

  The attribute identification unit 12B performs processing corresponding to steps S66 and S67 of FIG. That is, the attribute identification unit 12B calculates a feature amount (synthetic variable value) of the numerical sequence data of the spectrogram for the test mixture sample, and compares the calculated feature amount (synthetic variable value) with the attribute identification index. Then, the attribute of the test mixture sample is identified. Then, the attribute identification unit 12B outputs the identification result to the identification result output unit 13B. The identification result includes unknown attributes.

  The identification result output unit 13B outputs the identification result obtained by the attribute identification unit 12B. For example, the output unit 13B displays the identification result on the display unit (not shown). Thereby, the user of the attribute identification device 10B can grasp the attribute of the test mixture sample.

  Here, the index setting unit 20B may be configured separately from the attribute identification device 10B. In this case, the attribute identification device 10B (attribute identification unit 12B) inputs the synthetic variable and the attribute identification index from the index setting unit 20B when identifying the attribute of the test mixture sample or in advance. Configured as follows.

<Fourth embodiment>
In the fourth embodiment, an electromagnetic wave signal derived from the mixture sample is acquired from the mixture sample, a spectrogram of the acquired electromagnetic wave signal is calculated, and a time constant indicating an attenuation characteristic of each frequency component is calculated from the calculated spectrogram. Further, the characteristics of the mixture sample are evaluated based on the calculated time constant of each frequency component.
As a result, in the fourth embodiment as well, as in the third embodiment, the information on the structure and function of the “complex / heterogeneous mixture” included in the electromagnetic wave signal derived from the mixture sample is comprehensively analyzed. be able to. In addition, a significant improvement in accuracy can be expected in the characteristic evaluation (characteristic analysis) of the “complex / heterogeneous mixture”.

In 4th Embodiment, the characteristic of a mixture sample is evaluated according to the following processes.
(1) First step (acquisition of electromagnetic wave signal derived from the mixture sample)
This is the same as the first step in the first to third embodiments.

(2) Second step (processing of electromagnetic wave signal derived from mixture material)
A spectrogram (time frequency expression) is calculated by repeating short-time frequency analysis over the entire electromagnetic wave signal (RAW signal) acquired (saved) in the first step. This point is the same as the second step of the third embodiment. However, in the fourth embodiment, a time constant τ indicating the attenuation characteristic of each frequency component is further calculated from the calculated spectrogram. The time constant τ represents the degree of decrease in signal intensity (amplitude) of each frequency component, and is output as a frequency component / time constant graph (for example, a histogram) or numerical string data (for example, a one-dimensional array). Can do.

In the present embodiment, processing of the electromagnetic wave signal (processing for calculating attenuation characteristics (time constant) of each frequency component) is performed by the signal processing device 1D.
FIG. 15 is a block diagram illustrating a schematic configuration of the signal processing device 1D.
As illustrated in FIG. 15, the signal processing device 1D includes a processing unit 2D and an output unit 3D.
The processing unit 2D inputs an electromagnetic wave signal derived from the mixture sample, and calculates a spectrogram of the electromagnetic wave signal (a spectrogram of the mixture sample) by repeating short-time frequency analysis over the entire input electromagnetic wave signal. Further, the processing unit 2D calculates a time constant indicating the attenuation characteristic of each frequency component from the calculated spectrogram.
The output unit 3D outputs at least one of numerical sequence data of the calculated time constant and a frequency component / time constant graph created from the calculated time constant.

FIG. 16 is a flowchart illustrating an example of processing (calculation processing of the attenuation characteristic value (time constant τ)) performed by the signal processing device 1D.
The signal processing unit 1D calculates the spectrogram of each input electromagnetic wave signal in the same manner as the processing in steps S61 to S66 of FIG. 13, and then performs the processing shown in FIG. 16 for each frequency component of each calculated spectrogram. To implement.

Signal attenuation characteristics are expressed by the following equation (1) (α: initial amplitude, τ: time constant).
f (t) = αe− ^{t / τ} (1)
Taking the logarithm of both sides, the following equation (2) is obtained.
logf (t) = − (1 / τ) t + logα (2)
Therefore, the signal processing unit 1D converts each frequency component of the spectrogram into a logarithm as shown on the right side of the equation (2) (step S71), and estimates a regression line that most closely approximates the time series sample by the least square method (step S71). S72) A time constant τ is calculated from the slope of the regression line and stored in a one-dimensional array for storage (step S73).

When the signal processing apparatus 1D outputs the numerical string data of the calculated time constant τ of each frequency component, the one-dimensional array storing the time constant τ is converted into binary data or ASCII data (for example, CSV format). ) (Step S74).
On the other hand, when outputting the calculated time constant τ of each frequency component as an image, the signal processing device 1D creates a frequency component / time constant graph (for example, a histogram) (step S75).

(3) Third step (characteristic evaluation of the mixture sample)
The numerical sequence data and frequency component / time constant graph of the time constant τ of each frequency component of the spectrogram output from the signal processing device 1D are the same as the numerical sequence data and spectrogram image of the spectrogram in the third embodiment. This can be information that accurately indicates the characteristics of the mixture sample. Therefore, by analyzing the numerical string data of each frequency component and the frequency component / time constant graph, it is possible to evaluate the characteristics of the mixture sample as in the third embodiment.

  That is, the numerical sequence data of the time constant τ of each frequency component and the frequency component / time constant graph can be used, for example, as an index for evaluating the similarity or difference between the mixture samples. Can be set (or reset) and can be used to identify attributes of mixture samples with unknown attributes.

(4) Fourth step (attribute identification of a mixture sample with unknown attributes)
Also in the present embodiment, as in the first to third embodiments, the attribute of the test mixture sample whose attribute is unknown is identified using a known mixture sample whose attribute is determined in the same type as the test mixture sample. can do. Specifically, by comparing the time constant of each frequency component calculated from the spectrogram for the test mixture sample with the time constant of each frequency component calculated from the spectrogram for the known mixture sample, The attributes of the test mixture sample can be identified. Hereinafter, the case where the frequency component / time constant graph is used and the case where numerical sequence data of the time constant of each frequency component are used will be described.

(4-1) Using Frequency Component / Time Constant Graph FIG. 17 is a flowchart showing processing for identifying attributes of the test mixture sample using the frequency component / time constant graph.
In FIG. 17, step S81 is the same as the process of step S11 of FIG.
In step S82, the spectrogram (time frequency expression) of each electromagnetic wave signal is calculated by repeating short-time frequency analysis over the entire electromagnetic wave signal acquired in step S81.

  In step S83, for each spectrogram calculated in step S82, a time constant τ (attenuation characteristic) of each frequency component is calculated to create a frequency component / time constant graph. That is, a frequency component / time constant graph for the test mixture sample and a frequency component / time constant graph for at least one known mixture sample are created. The creation of the frequency component / time constant flag is performed in the same manner as in the second step.

  In step S84, the frequency component / time constant graph created in step S83 is compared to identify the attribute of the test mixture sample. Specifically, it is determined whether the frequency component / time constant graph for the test mixture sample is similar to or most similar to the frequency component / time constant graph for at least one known mixture sample. Identify attributes of the test mixture sample. If it is determined that the frequency component / time constant graph for the test mixture sample is not similar to any of the frequency component / time constant graphs for at least one of the known mixture samples, Leave the attribute unknown.

  Here, the processing of steps S81 to S83 for at least one of the known mixture samples may be performed separately from the present flow. For example, after the frequency component / time constant graph for at least one of the known mixture samples is created, the processing of steps S81 to S83 is performed on the test mixture sample, and the attributes of the test mixture sample in step S84 May be identified.

In addition, the attribute of the test mixture sample in step S84 can be identified by human vision (color vision) or a predetermined device. It is performed by human vision (color vision) or a predetermined device. When performed by human vision (color vision), each electromagnetic wave signal acquired in step S81 is input to the signal processing device 1D, and each frequency component / time constant graph output from the signal processing device 1D is displayed as a predetermined graph. A human uses the vision (color vision) to compare and identify the attributes of the test mixture sample.
On the other hand, when it is performed by a predetermined device, an attribute identification device having the following configuration is manufactured, and the attribute identification device performs the processing of steps S82 to S84 in FIG.

  In the flowchart shown in FIG. 17, the frequency component / time constant graph for the test mixture sample is compared with the frequency component / time constant graph for at least one known mixture sample. Instead of the frequency component / time constant graph for the two known mixture samples, an attribute identification graph created for each attribute based on these may be used.

FIG. 18 is a block diagram showing a schematic configuration of an attribute identification device 10C for identifying attributes of the test mixture sample using the frequency component / time constant graph.
As illustrated in FIG. 18, the attribute identification device 10C includes a signal processing unit 11C, an attribute identification unit 12C, and an identification result output unit 13C.

  The signal processing unit 11C has a function similar to that of the above-described signal processing device 1D, and performs processing corresponding to steps S82 and S83 in FIG. That is, the signal processing unit 11C inputs the electromagnetic wave signal derived from each of the test mixture sample and the at least one known mixture sample, and repeats the frequency analysis for a short time over the entire input electromagnetic wave signal. A spectrogram of each electromagnetic wave signal is calculated, and a time constant (attenuation characteristic) of each frequency component of the calculated spectrogram is calculated. Then, the signal processing unit 11B creates a frequency component / time constant graph for the test mixture sample and a frequency component / time constant graph for at least one known mixture sample, and outputs them to the attribute identification unit 12C.

  The attribute identifying unit 12C performs a process corresponding to step S84 in FIG. 17 using, for example, an image engineering technique. That is, the attribute identification unit 12C compares the frequency component / time constant graph for the test mixture sample with the frequency component / time constant graph for at least one known mixture sample, and And the identification result is output to the identification result output unit 13C. The identification result includes unknown attributes.

  The identification result output unit 13C outputs the identification result obtained by the attribute identification unit 12C. For example, the output unit 13B displays the identification result on a display unit (not shown). Thereby, the user of the attribute identification device 10C can grasp the attribute of the test mixture sample.

(4-2) When using the numerical sequence data of the time constant of each frequency component FIG. 19 shows the process of identifying the attribute of the test mixture sample using the numerical sequence data of the time constant of each frequency component It is a flowchart.
In FIG. 19, steps S91 and S92 are the same as the processes of steps S61 and S62 of FIG. That is, an electromagnetic wave signal derived from each of the test mixture sample and the plurality of known mixture samples is acquired, and a spectrogram (time) of each electromagnetic wave signal is obtained by repeating short-time frequency analysis over the entire acquired electromagnetic wave signals. Frequency expression).

In step S93, for each spectrogram calculated in step S92, a time constant τ (attenuation characteristic) of each frequency component is calculated (see the second step).
As a result, it is possible to obtain the numerical sequence data of the time constant of each frequency component for the test mixture sample and the numerical sequence data of the time constant of each frequency component for each of the plurality of known mixture samples. .

  In step S94, multivariate analysis is performed on the numerical sequence data of the time constants of the frequency components for the plurality of known mixture samples, and the numerical sequences of the time constants of the frequency components for the plurality of known mixture samples. Create a composite variable to separate the data into data groups for each attribute.

  In step S95, a feature value for each of the plurality of known mixture samples, that is, a composite variable value (score) of the numerical value string data of the time constant of each frequency component is calculated.

  In step S96, an attribute identification index is set based on the feature amount (synthetic variable value) for the plurality of known mixture samples. Specifically, feature quantities (synthetic variable values) for each of the plurality of known mixture samples calculated in step S95 are grouped for each attribute, and feature quantities for the known mixture samples for each grouped attribute ( The attribute identification index is set based on the composite variable value). The attribute identification index is set with a predetermined range for each attribute. Note that the attribute identification index set in step S96 may be a numerical value or a figure or table.

  In step S97, the characteristic amount of the test mixture sample, that is, the composite variable value of the numerical sequence data of the time constant of each frequency component is calculated.

  In step S98, the feature amount (synthetic variable value) for the test mixture sample calculated in step S97 is compared with the attribute identification index set in step S96, and the attribute of the test mixture sample is compared. Identify. Specifically, it is determined whether the synthetic variable value for the test mixture sample corresponds to (includes) the attribute identification index set for each attribute, and the attribute of the test mixture sample Identify In addition, when the synthetic variable value about the said test mixture sample does not correspond to any of the said attribute identification parameter | index (it is not contained), the attribute of the said test mixture sample is made into unknown. Moreover, the identification of the attribute of the test mixture sample in this step S98 can be performed by a predetermined person or a predetermined apparatus as in step S84 of FIG.

  Here, the processes of steps S91 to S93 and the processes of steps S94 to S96 for each of the plurality of known mixture samples may be performed separately from the present flow. For example, after the attribute identification index is set, the processing of steps S91 to S93 and S97 may be performed on the test mixture sample, and the attributes of the test mixture sample may be identified in step S98. .

FIG. 20 is a block diagram showing a schematic configuration of an attribute identification device 10D for identifying attributes of the test mixture sample by using numerical sequence data of time constants of the frequency components. The attribute identification device 10D performs the processes of steps S92 to S98 in FIG.
As illustrated in FIG. 20, the attribute identification device 10D includes a signal processing unit 11D, an index setting unit 20D, an attribute identification unit 12D, and an identification result output unit 13D.

  The signal processing unit 11D has a function similar to that of the above-described signal processing device 1D, and performs processing corresponding to steps S92 and S93 in FIG. That is, the signal processing unit 11D receives the electromagnetic wave signal derived from each of the test mixture sample and the plurality of known mixture samples, and repeats the frequency analysis for a short time throughout the input electromagnetic wave signals. A spectrogram of the electromagnetic wave signal is calculated, and a time constant (attenuation characteristic) of each frequency component of the calculated spectrogram is calculated. Then, the signal processing unit 11D outputs, to the index setting unit 20D, the numerical sequence data of the time constant of each frequency component for a plurality of the known mixture samples, while the time of each frequency component for the test mixture sample. The constant numeric string data is output to the attribute identification unit 12D.

  The index setting unit 20D performs processing corresponding to steps S94 to S96 in FIG. That is, the index setting unit 20D performs the multivariate analysis on the numerical sequence data of the time constants of the frequency components for the plurality of input mixture samples, and generates the composite variable, The feature amount (synthetic variable value) of the numerical sequence data of the time constant of each frequency component for each of the mixture samples is calculated. Then, the index setting unit 20D groups the calculated feature values (synthetic variable values) for each attribute, and based on the feature values (synthetic variable values) for the known mixture sample for each grouped attribute. Set an index for attribute identification. The index setting unit 20D outputs the composite variable and the attribute identification index to the attribute identification unit 12D.

  The attribute identification unit 12D performs processing corresponding to steps S97 and S98 in FIG. That is, the attribute identification unit 12D calculates the feature quantity (synthetic variable value) of the numerical value string data of the time constant of each frequency component for the test mixture sample, and calculates the calculated feature quantity (synthetic variable value) and the attribute. The attribute of the test mixture sample is identified by comparing with an identification index. Then, the attribute identification unit 12D outputs the identification result to the identification result output unit 13D. The identification result also includes unknown attributes

  The identification result output unit 13D outputs the identification result obtained by the attribute identification unit 12D. For example, the output unit 13D displays the identification result on the display unit (not shown). Thereby, the user of the attribute identification device 10D can grasp the attributes of the test mixture sample.

  Here, the index setting unit 20D may be configured separately from the attribute identification device 10D. In this case, the attribute identification device 10D (attribute identification unit 12D) inputs the synthetic variable and the attribute identification index from the index setting unit 20D when identifying the attribute of the test mixture sample or in advance. Configured as follows.

  In the above description, the signal processing apparatus 1D calculates the time constant τ as a value indicating the attenuation characteristic of each frequency component. However, the initial amplitude α may be calculated instead of the time constant τ. This initial amplitude α also corresponds to an attenuation characteristic value. In this case, the signal processing device 1D outputs numerical sequence data and / or frequency component / initial amplitude graph of the initial amplitude α of each frequency component. In the characteristic evaluation of the mixture sample (third step) and the attribute identification of the mixture sample whose attribute is unknown (fourth step), the numerical sequence data of the time constant, the numerical sequence data of the initial amplitude, and the frequency component / The time constant graph may be read as the frequency component / initial amplitude graph. Further, the signal processing device 1D may calculate the time constant τ and the initial amplitude α as values indicating the attenuation characteristics of the respective frequency components.

<Example>
(A) Example Using NMR Signal The inventors created a program and software (hereinafter referred to as “NMR signal processing software”) that operates on a computer and processes an NMR signal derived from a mixture sample. This NMR signal processing software reads an NMR signal derived from a mixture sample, (a) outputs the read NMR signal as sound, and (b) performs a short-time frequency analysis on the read NMR signal to calculate a spectrogram. (C) calculating attenuation characteristic values (time constant τ and / or initial amplitude α) of each frequency component from the calculated spectrogram, and outputting frequency / attenuation characteristic values. It is possible to output as a graph (image) and numerical string data. That is, the NMR signal processing software can input and process NMR signals, and store and output them as sound, images, and numerical data. Data storage (output) methods include sound formats (RAW format, WAV format, etc.), image formats (JPEG format, TIFF format, PNG format, etc.), and numeric string formats (CSV format, etc.). The sound format data can be played back by general-purpose playback software and listened to as sound, and can be analyzed acoustically using general-purpose sound analysis software. Image format data allows visual (color vision) comparison of the results of short-time frequency analysis of NMR signals.
Hereinafter, as specific examples of the mixture sample, experimental methods and analysis results of “Example 1 rat plasma” and “Example 2 mouse cell extract” are described.

Example 1 Rat Plasma [Subjects and Methods]
・ Animal experiment (production of rat hemorrhagic shock model)
Male Sprague-Dawley rats weighing 350-400 g were used. For anesthesia, ketamine hydrochloride (65 mg / kg) and xylazine (0.7 mg / kg) (2- [2,6-dimethylphenylamino] -4H-5,6-dihydrothiazine hydrochloride) were intraperitoneally administered. Body temperature was measured with a rectal thermometer and maintained at 37.5-38.5 ° C. using a warming lamp. A polyethylene tube (outer diameter 0.96 mm, inner diameter 0.58 mm) filled with heparinized physiological saline (10 units / mL) was placed in the femoral artery and vein. The polyethylene tube inserted into the artery was connected to a transducer and an arterial pressure monitor, and blood pressure was continuously measured. Thereafter, the rat was fixed in a Bollman-type resting cage. When the mean arterial pressure of the rat was stabilized at 100 to 110 mmHg, blood was removed at a rate of 1 mL / min until the mean blood pressure reached 40 mmHg to create a hemorrhagic shock state. Thereafter, blood removal and return were repeated as necessary, and fine adjustment was performed so that the blood pressure level was maintained at 40 mmHg for 30 minutes.

Collection and storage of plasma Arterial blood collected from rats in a hemorrhagic shock state for 30 minutes was centrifuged at 4 ° C., and the supernatant was stored at −80 ° C. (“shock plasma”).
Normal rats were anesthetized as described above, arterial blood was collected, and the plasma obtained by centrifugation was used as control plasma (“control plasma”).

-Preparation of NMR measurement sample To 50-100 μL of prepared plasma, 500-550 μL of heavy water containing a known amount of sodium (3-trimethylsilyl) tetradeuteriopropionate-2,2,3,3-d4 (TMSP) is added, and 5 mm diameter NMR made of glass It moved to the sample tube and was set as the NMR measurement sample. Heavy water was added for the purpose of obtaining a deuterium lock signal for correction of magnetic field inhomogeneity, and TMSP was added as an indicator for signal confirmation.

・ Acquisition of proton NMR signal Acquisition of proton NMR signal is based on ECX NMR equipment (equipped with superconducting magnet with 300MHz proton resonance frequency, TH5 probe, 16 autosample changer, Delta ^™ NMR processing and control software (version 4.3.2)) I went there. For acquisition of proton NMR signal, a macro program for automatic measurement incorporated in the Delta ^™ system was used. The probe internal temperature was 23 ° C. The observation range of the NMR signal was 4500 Hz. As a measurement sequence, a single pulse mode for acquiring a one-dimensional NMR signal was used, and light water signals were suppressed by the homogate method and the DANTE method. The integration was performed 400 times so that a sufficient signal-to-noise ratio for data analysis was obtained. The obtained NMR signal data of each sample was stored as it was (here, JDF (ECX standard) format).

・ Data analysis (processing by NMR signal processing software and analysis of processing results)
(1) Output of sound of NMR signal The NMR signal data of each rat plasma stored in JDF format was input to the NMR signal processing software, and output (reproduced) as sound for each file of each JDF format. As a result, it was confirmed that the control plasma and the shock plasma are clearly output as different sounds, that is, the control plasma and the shock plasma can be distinguished by human hearing. Next, it was confirmed that these data were saved in the uncompressed WAV format, and the saved file could be played back by general-purpose playback software and analyzed using general-purpose sound analysis software.

(2) Calculation of spectrogram and creation of spectrogram image The NMR signal data of each rat plasma stored in JDF format is input to the NMR signal processing software, and the frequency analysis of each NMR signal data is repeated over the entire signal for a short time. The spectrum was calculated. The calculation results were displayed as spectrogram images on the computer display. Each spectrogram image was stored in PNG format.
FIG. 21 shows an example of a spectrogram image of control plasma and an example of a spectrum image of shock plasma. As shown in FIG. 21, the spectrogram image of control plasma and the spectrum image of shock plasma were clearly different, and it was confirmed that both could be identified by comparing the spectrogram images.

(3) Calculation of attenuation characteristic value (time constant τ) and creation of frequency component / time constant graph NMR signal data of each rat plasma stored in JDF format is input to the NMR signal processing software, and each NMR signal data The spectrum was calculated, and the attenuation characteristic value (time constant τ) of each frequency component was calculated from the calculated spectrogram. The calculation result was displayed as a frequency component / time constant graph on the computer display. Each frequency component / time constant graph was stored in PNG format.
FIG. 22 shows an example of a frequency component / time constant graph of control plasma and an example of a frequency component / time constant graph of shock plasma. As shown in FIG. 22, the frequency component / time constant graph of control plasma and the frequency component / time constant graph of shock plasma are clearly different, and both can be identified by comparing the frequency component / time constant graph. confirmed.

(4) Identification of spectrogram value string data by multivariate analysis NMR signal data of each rat plasma stored in JDF format was input to the NMR signal processing software, and a spectrum was calculated for each NMR signal data. The numerical sequence data as the calculation result was output and saved in CSV format. The stored numerical sequence data was input to general-purpose multivariate analysis software to perform principal component analysis (PCA). The result is shown in FIG. As shown in FIG. 23, the principal component score calculated from each data was distributed in a manner that the principal component 1 (Pc1) was clustered into a control group and a shock group in different regions on the score plot. Subsequent classification by the PLS-DA method confirmed that the control group and the shock group could be accurately identified as shown in FIG.

(5) Identification of numerical value data of attenuation characteristic value (time constant τ) by multivariate analysis NMR signal data of each rat plasma stored in JDF format is input to the NMR signal processing software, and the spectrum of each NMR signal data The attenuation characteristic value (time constant τ) of each frequency component was calculated from the calculated spectrogram. The numerical sequence data (τ value) as the calculation result was output and stored in CSV format. The stored numerical sequence data was input to general-purpose multivariate analysis software to perform principal component analysis (PCA). The result is shown in FIG. As shown in FIG. 25, it was confirmed that the principal component scores calculated from the respective data are clustered and distributed in the control group and the shock group in regions where the principal component 1 (Pc1) is different on the score plot. Subsequent classification by the PLS-DA method confirmed that the control group and the shock group could be accurately identified as shown in FIG.

(Example 2) Mouse cell extract [subject and method]
-Cell collection (1) Adipose tissue-derived stem cells (ASCs)
Five-week-old male mice (C57BL / 6) were used. After euthanizing with ether in an anesthesia bottle, the mouse was fixed in the supine position, the abdomen was transversely incised, and a fat mass from the groin to the back was collected. The fat mass was washed with PBS (Phosphate buffered saline) and adhering blood vessels were excised. After the fat mass was shredded, it was transferred to a 50 mL centrifuge tube together with PBS, and collagenase was further added to obtain a 0.15% collagenase solution. The centrifuge tube was placed in a hot tub and incubated at 37 ° C. for 30 minutes. After the fat mass in the centrifuge tube disappeared macroscopically, FBS (Fetal bovine serum) was added to stop the collagenase activity, followed by centrifugation. After the supernatant after centrifugation was removed by suction, 10 mL of control medium (culture solution) was placed in the same tube, and the cell pellet was sufficiently stirred, and 1000 μL was pipetted into each of 10 100 mL dishes. Each dish was shaken well, and after confirming that the cells spread evenly with the naked eye, incubation at 37 ° C. was started. The medium was changed 2-3 times a week.
After 100% confluence in about 1 day from about 4 days, all the control medium was aspirated and washed with PBS. Thereafter, 1 mL of Trypsin-EDTA was added to each 100 mL dish and incubated at 37 ° C. for 10 minutes. After the medium turned macroscopically yellow, cell detachment was confirmed with a microscope, and 2 mL of control medium was added for neutralization. Then, 0.5-1 mL of the same solution was dispensed into a new 100 mL dish containing 5 mL of control medium in advance, and incubation at 37 ° C. was started. Cells subcultured in this way were collected, rapidly frozen in liquid nitrogen, and stored at -80 ° C.

(2) Bone marrow mesenchymal stem cells (BSCs)
Five-week-old male mice (C57BL / 6) were used. After euthanizing with ether in an anesthesia bottle, the femur and tibia were collected, and then the bone marrow was washed out using a syringe with 20 mL PBS with a 23 gauge needle and collected in a 50 mL tube. A cell strainer was set in another 50 mL tube, and the bone marrow fluid collected earlier was transferred here and centrifuged. The supernatant was aspirated without sucking the cell pellet. To the bone marrow for 3 mice, 1 mL of control medium was added to the tube, and the cells were suspended after shaking. Then, 1 mL of the same solution was dispensed into a new 100 mL dish containing 5 mL of control medium in advance, and incubation at 37 ° C. was started. Cells subcultured in this way were collected, rapidly frozen in liquid nitrogen, and stored at -80 ° C.

-Preparation of NMR measurement sample (1) Cell disruption / separation extraction In order to prepare a mouse cell extract for NMR measurement, neutral extraction was performed according to the protocol of Yoshioka et al. This extraction method is basically the same as the Folch method. Folch designed this method to extract lipid components from animal tissues, but Yoshioka did not apply any special treatment to tissues or cells removed from living organisms, and chemically changed the contained substances. Focusing on the fact that a large number of chemical substances can be extracted without causing problems, improvements were made. In particular, it is also useful in that polar organic compounds can be separated from nonpolar organic compounds.
First, the cryopreserved cells were sufficiently stirred in 1.5 mL of a cooled chloroform / methanol (2: 1) mixture, and the cells were crushed and the contents were extracted by shaking, followed by centrifugation. Next, 0.5 mL of distilled water was added to the extract from which insoluble components were removed by filtration, and shake extraction was performed again under the same conditions as described above. The extract was sufficiently stirred and allowed to stand at 4 ° C. for 24 hours, followed by centrifugation to separate the two solution layers and the insoluble residue. In this experiment, the solvent was completely distilled off from the upper layer solution using a vacuum evaporator, and the residue obtained by distillation was used as an NMR measurement sample.

(2) Preparation of NMR measurement sample solution 180 μL of heavy water (D 2 O) was added to the above residue and redissolved, and placed in a 3 mm NMR sample tube. Into a separately prepared 5 mm NMR sample tube, 300 μL of heavy water containing 25 μM sodium (3-trimethylsilyl) tetradeuteriopropionate-2,2,3,3-d4 (TMSP) was placed, and the 3 mm NMR sample tube containing the measurement sample was placed in the 5 mm NMR tube. The sample was placed in a sample tube and fixed to obtain an NMR measurement sample. Heavy water was used for the purpose of obtaining a deuterium lock signal for correction of magnetic field inhomogeneity, and TMSP was used as an internal indicator for signal confirmation.

・ Acquisition of proton NMR signal Acquisition of proton NMR signal is based on ECX NMR equipment (equipped with superconducting magnet with 300MHz proton resonance frequency, TH5 probe, 16 autosample changer, Delta ^™ NMR processing and control software (version 4.3.2)) I went there. For acquisition of proton NMR signal, a macro program for automatic measurement incorporated in the Delta ^™ system was used. The probe internal temperature was 23 ° C. The observation range of the NMR signal was 4500 Hz. As a measurement sequence, a single pulse mode for acquiring a one-dimensional NMR signal was used, and light water signals were suppressed by the homogate method and the DANTE method. Accumulation was performed 8000 times so that a sufficient signal-to-noise ratio for data analysis was obtained. The acquired NMR signal data of each sample was stored as it is (here, JDF format (ECX standard format)).

・ Data analysis (processing by NMR signal processing software and analysis of processing results)
(1) Output of sound of NMR signal NMR signal data of each mouse cell extract stored in JDF format was input to the NMR signal processing software, and output (reproduced) as sound for each file of each JDF format. It was confirmed that the ASCs extract and the BSCs extract are output as clearly different sounds. Next, it was confirmed that these data were saved in the WAV format, and that the saved file could be played back by general-purpose playback software and analyzed using general-purpose sound analysis software.

(2) Calculation of spectrogram and creation of spectrogram image The NMR signal data of each mouse cell extract saved in JDF format is input to the NMR signal processing software, and for each NMR signal data, a short time frequency analysis is performed over the entire signal. The spectrum was calculated repeatedly. The calculation results were displayed as spectrogram images on the computer display. Each spectrogram image was stored in PNG format.
FIG. 27 shows an example of a spectrogram image of an ASCs extract and an example of a spectrum image of a BSCs extract (actually color). As shown in FIG. 27, the spectrogram image of the ASCs extract and the spectrum image of the BSCs extract are clearly different, and it was confirmed that the spectrogram images can be identified by comparing them.

(3) Calculation of attenuation characteristic value (time constant τ) and creation of frequency component / time constant graph NMR signal data of each mouse cell extract stored in JDF format is input to the NMR signal processing software, and each NMR signal A spectrum was calculated for the data, and an attenuation characteristic value (time constant τ) of each frequency component was calculated from the calculated spectrogram. The calculation result was displayed as a frequency component / time constant graph on the computer display. Each frequency component / time constant graph was stored in PNG format.
FIG. 28 shows an example of a frequency component / time constant graph of the ASCs extract and an example of a frequency component / time constant graph of the BSCs extract. As shown in FIG. 28, the frequency component / time constant graph of the ASCs extract is clearly different from the frequency component / time constant graph of the BSCs extract, and both are identified by comparing the frequency component / time constant graph. It was confirmed that it was possible.

(4) Identification of spectrogram value string data by multivariate analysis NMR signal data of each mouse cell extract stored in JDF format was input to the NMR signal processing software, and a spectrum was calculated for each NMR signal data. Numerical sequence data as a calculation result was output and saved in CSV format. The stored numerical sequence data was input to general-purpose multivariate analysis software to perform principal component analysis (PCA). The result is shown in FIG. As shown in FIG. 29, the principal component scores calculated from the respective data are distributed in a manner such that principal component 1 (Pc1) is clustered into ASCs extract group and BSCs extract group in different regions on the score plot. . Subsequent classification by the PLS-DA method confirmed that the ASCs extract group and the BSCs extract group can be accurately identified as shown in FIG.

(5) Identification of attenuation characteristic value (time constant τ) numerical data by multivariate analysis NMR signal data of each mouse cell extract stored in JDF format is input to the NMR signal processing software, and each NMR signal data The spectrum was calculated for, and the attenuation characteristic value (time constant τ) of each frequency component was calculated from the calculated spectrogram. The numerical sequence data (τ value) as the calculation result was output and stored in CSV format. The stored numerical sequence data (τ value) was input into general-purpose multivariate analysis software to perform principal component analysis (PCA). The result is shown in FIG. As shown in FIG. 31, the principal component scores calculated from each data were distributed in a manner that the principal component 1 (Pc1) was clustered into different areas in the ASCs extract group and the BSCs extract group on the score plot. . Subsequent classification by the PLS-DA method confirmed that the ASCs extract group and the BSCs extract group can be accurately identified as shown in FIG.

(B) Example Using Near Infrared (NIR) Signal The inventors have executed a program and software (hereinafter referred to as “electromagnetic wave signal”) for processing an electromagnetic wave signal (for example, NIR signal) derived from a mixture sample by running on a computer. Processing software "). This electromagnetic wave signal processing software can read an electromagnetic wave signal derived from a mixture sample and output (display) the read electromagnetic wave signal as a color (color image). That is, the electromagnetic wave signal processing software can process the electromagnetic wave signal and output it as a color image. The electromagnetic wave signal processing software can store and output the processed data in an image format (such as PNG format) and a numerical string format (such as CSV format).
Hereinafter, the experimental method and analysis result of “rat plasma” will be described as specific examples of the mixture sample.

Example 3 Rat Plasma [Subjects and Methods]
・ Animal experiments (production of sepsis model)
Male Sprague-Dawley rats weighing 280-330 g are intraperitoneally administered with 2 mg / kg or 10 mg / kg lipopolysaccharide (LPS, Escherichia coli 0111: B4) to induce sepsis. Six hours after administration, venous blood collected under 3-5% isoflurane anesthesia was centrifuged, and the supernatant was stored at −80 ° C. (LPS plasma 1 (2 mg / kg), LPS plasma 2 (10 mg / kg)). .
Plasma obtained by performing the same treatment from a rat administered with physiological saline instead of LPS was used as control plasma.

-Preparation of NIR measurement sample A NIR measurement sample was prepared by placing a plasma sample in a glass sample tube.
-Acquisition of NIR signal A near infrared sensor was used.
The measurement conditions are as follows.
Wavelength: 900-1700 nm, accumulation time mode: automatic, accumulation time upper limit: 1024 ms, gain: GAIN_LOW, number of scans: 10 times, measurement mode: number of times designated measurement mode, storage data: count value, absorbance, spectrum (standardized absorbance) .
The NIR measurement sample was placed in the measurement port of the near infrared sensor probe, and each sample was measured 10 times. Each measurement data was output and saved in CSV format.

Data processing The following processing and output were performed using the electromagnetic wave signal processing software.
(1) Color display of the NIR signal The dynamic range of the wavelength of the acquired NIR signal (eg, 900 to 1700 nm in the case of the near infrared region) is linearly mapped to the visible region (400 to 800 nm), and the signal in the visible region is displayed. It was read as (converted). Next, the intensities (average values) in each of the main bands of the three primary colors of RGB (for example, 690 to 710 nm for red, 536 to 556 nm for green, and 425 to 445 nm for blue) determined from the RGB sensitivity curve were calculated.
(2) Identification by visual (color vision) evaluation of NIR data of each color-displayed mixture Using the RGB values calculated for the NIR signal for each rat plasma sample, the color image is displayed on the computer display and a color image Saved as data (PNG format). FIG. 33 shows an example of each color (color image) of control plasma, LPS plasma 1, and LPS plasma 2 (actually color). As shown in FIG. 33, the color of control plasma, the color of LPS plasma 1 and the color of LPS plasma 2 are different from each other, and it was confirmed that the three could be identified by comparing the displayed colors.
(3) Identification of RGB data (numerical string data) by SVM (see FIG. 34)
RGB data of a plurality of control plasmas, a plurality of LPS plasmas 1 and a plurality of LPS plasmas 2 were output and stored in CSV format. This data is read into general-purpose multivariate analysis software, and control plasma (75 data), LPS plasma 1 (77 data), LPS plasma 2 (73 data) are training data, and the remaining control plasma (36 data), LPS plasma. 1 (36 data) and LPS plasma 2 (36 data) were used as test data for discrimination by SVM. SVM learning was performed using the training data, and the performance of the classifier was evaluated for the training data by cross validation with a segment size of 9. As a result, a high identification rate was obtained. Next, when classifying the test data with this classifier, an identification rate of 100% was obtained, and it was confirmed that the classifier had a high accuracy in identifying unknown data.

  As mentioned above, although embodiment and Example of this invention were described, this invention is not limited to each above-mentioned embodiment and each Example, A various deformation | transformation etc. are based on the technical idea of this invention. Is possible. In addition, the present invention is an epoch-making technique for evaluating the characteristics of a mixture. For example, an attribute of a mixture sample whose attribute is unknown is easily and quickly identified based on a known mixture sample whose attribute is clear. can do. Furthermore, the present invention is not limited to attribute identification, and can be applied as a mixture sample analysis technique in various fields.

  Usually, in the medical field, treatment starts after symptoms appear. However, in the future, the emergence of new medical care (preemptive medicine) that predicts disease before symptoms appear and performs therapeutic intervention to prevent or delay the onset of the disease is awaited. In the medical field, use of the present invention opens up the possibility of developing new clinical testing methods and medical devices, including ultra-early diagnosis in preemptive medicine, determination of treatment policy, determination of treatment effect, and prognosis prediction.

  In addition to the medical field, the present invention is applied as a new sensing technology to, for example, various industrial monitoring devices, analytical instruments, detectors (abnormality detection, process monitoring, etc. in chemical product manufacturing, food processing, water treatment, etc.) It is expected to be applied to food application, food inspection and development, plant breeding, polymer chemistry, etc.

  Furthermore, the structure and function of “complex / heterogeneous mixture”, which has been unknown in detail, will be elucidated by evaluating the characteristics (characteristic analysis) of “complex / heterogeneous mixture” using the present invention. New academic fields can be expected to be created.

  1A, 1B, 1C, 1D ... signal processing device, 2A, 2B, 2C, 2D ... processing unit, 3A, 3B, 3C, 3D ... output unit, 10A, 10B, 10C, 10D ... attribute identification device, 11A, 11B, 11C, 11D ... signal processing unit, 12A, 12B, 13C, 12D ... attribute identification unit, 13A, 13B, 13C, 13D ... unit, 20A, 20B ... index setting unit

Claims (23)

A method for expressing the characteristics of complex and heterogeneous mixture samples including biological samples taken from organisms ,
Obtaining a FID signal by subjecting the complex / heterogeneous mixture sample to a nuclear magnetic resonance (NMR) analyzer ; and
Calculating a spectrogram of the FID signal by repeating short-time frequency analysis over the entire FID signal, and expressing the characteristics of the complex / heterogeneous mixture sample by the calculated spectrogram;
Including a method. A method for characterizing a complex / heterogeneous mixture sample containing a biological sample taken from a living organism ,
Obtaining a FID signal by subjecting the complex / heterogeneous mixture sample to a nuclear magnetic resonance (NMR) analyzer ; and
Calculating a spectrogram of the FID signal by repeating short-time frequency analysis over the entire FID signal, and evaluating the characteristics of the complex / heterogeneous mixture sample based on the calculated spectrogram;
Including a method. The method according to claim 2 , wherein an attenuation characteristic value of each frequency component of the spectrogram is calculated, and characteristics of the complex / heterogeneous mixture sample are evaluated based on the calculated attenuation characteristic value of each frequency component. A method for identifying the attributes of complex and heterogeneous mixture samples including biological samples taken from organisms ,
Obtaining a FID signal by applying a test sample of a complex / heterogeneous mixture sample having unknown attributes to a nuclear magnetic resonance (NMR) analyzer ;
Calculating the spectrogram of the FID signal by repeating a short time frequency analysis throughout the FID signal,
Creating a spectrogram image of the test mixture sample from the spectrogram; and
A spectrogram image of the test mixture sample is compared with an image for attribute identification created based on at least one known mixture sample of the same kind as the test mixture sample and whose attributes are determined. Identifying the attributes of the mixture sample;
Including a method. The method according to claim 4 , wherein the attribute identification image includes a spectrogram image of the at least one known mixture sample created in the same manner as a spectrogram image of the test mixture sample. A method for identifying the attributes of complex and heterogeneous mixture samples including biological samples taken from organisms ,
Obtaining a FID signal by applying a test sample of a complex / heterogeneous mixture sample having unknown attributes to a nuclear magnetic resonance (NMR) analyzer ;
Calculating the spectrogram of the FID signal by repeating a short time frequency analysis throughout the FID signal,
Calculating a feature amount for the test mixture sample using the numerical sequence data of the spectrogram, and
Comparing the feature quantity of the test mixture sample with an attribute identification index set based on a plurality of known mixture samples of the same type as the test mixture sample and having attributes determined; Identifying the attributes of the sample,
Including, methods The attribute identification index is the feature amount of the known mixture sample grouped for each attribute among the feature amounts of the plurality of known mixture samples calculated in the same manner as the feature amount of the test mixture sample. The method according to claim 6 , wherein the method is set for each attribute based on the attribute. The feature amount, the a feature based on multivariate analysis with respect to the numerical sequence data spectrogram for said plurality of known mixtures sample calculated in the same manner as the spectrogram for the test mixture sample, according to claim 6 or 7 the method of. A method for identifying the attributes of complex and heterogeneous mixture samples including biological samples taken from organisms ,
Obtaining a FID signal by applying a test sample of a complex / heterogeneous mixture sample having unknown attributes to a nuclear magnetic resonance (NMR) analyzer ;
Calculating the spectrogram of the FID signal by repeating a short time frequency analysis throughout the FID signal,
Calculating an attenuation characteristic value of each frequency component of the spectrogram,
Creating a frequency component / attenuation characteristic value graph for the test mixture sample from the attenuation characteristic value of each frequency component; and
The frequency component / attenuation characteristic value graph for the test mixture sample is compared with the attribute identification graph created based on at least one known mixture sample of the same type as the test mixture sample and whose attributes are determined. Identifying the attributes of the test mixture sample,
Including a method. The attribute discrimination graph includes a frequency component / attenuation characteristic value graph for the at least one known mixture samples were prepared in the same manner as the frequency component / attenuation characteristic value graph of the mixture tested sample, claim 9 The method described in 1. A method for identifying the attributes of complex and heterogeneous mixture samples including biological samples taken from organisms ,
Obtaining a FID signal by applying a test sample of a complex / heterogeneous mixture sample having unknown attributes to a nuclear magnetic resonance (NMR) analyzer ;
Calculating the spectrogram of the FID signal by repeating a short time frequency analysis throughout the FID signal,
Calculating an attenuation characteristic value of each frequency component of the spectrogram,
Calculating a characteristic amount for the test mixture sample using the attenuation characteristic value of each frequency component; and
Comparing the feature quantity of the test mixture sample with an attribute identification index set based on a plurality of known mixture samples of the same type as the test mixture sample and having attributes determined; Identifying the attributes of the sample,
Including a method. The attribute identification index is the feature amount of the known mixture sample grouped for each attribute among the feature amounts of the plurality of known mixture samples calculated in the same manner as the feature amount of the test mixture sample. The method according to claim 11 , wherein the method is set for each attribute based on the attribute. The feature amount is used for multivariate analysis with respect to numerical sequence data of attenuation characteristic values of frequency components for the plurality of known mixture samples calculated in the same manner as attenuation characteristic values of frequency components for the test mixture sample. a feature based method according to claim 11 or 12. A method of processing a FID signal acquired by subjecting a complex / heterogeneous mixture sample containing a biological sample taken from a living organism to a nuclear magnetic resonance (NMR) analyzer , comprising:
Calculating the spectrogram of the FID signal by repeating a short time frequency analysis throughout the FID signals, and,
Outputting at least one of the numerical sequence data of the spectrogram and the spectrogram image created from the spectrogram as information indicating the characteristics of the complex / heterogeneous mixture sample,
Including a method. A method of processing a FID signal acquired by subjecting a complex / heterogeneous mixture sample containing a biological sample taken from a living organism to a nuclear magnetic resonance (NMR) analyzer , comprising:
Calculating the spectrogram of the FID signal by repeating a short time frequency analysis throughout the FID signal,
Calculating an attenuation characteristic value of each frequency component of the spectrogram; and
At least one of the numerical string data of the attenuation characteristic value of each frequency component and the frequency component / attenuation characteristic value graph created from the attenuation characteristic value of each frequency component is the characteristic of the complex / heterogeneous mixture sample. Output as information to show,
Including a method. A device for identifying the attributes of complex and heterogeneous mixture samples including biological samples taken from organisms ,
The FID signal acquired by applying the test sample of the complex / heterogeneous mixture sample with unknown attributes to a nuclear magnetic resonance (NMR) analyzer is input, and the entire input FID signal is short. A signal processing unit that calculates a spectrogram of the FID signal by repeating time-frequency analysis, and generates and outputs a spectrogram image of the test mixture sample from the calculated spectrogram;
A spectrogram image of the test mixture sample is compared with an image for attribute identification created based on at least one known mixture sample of the same kind as the test mixture sample and whose attributes are determined. An attribute identifier for identifying the attributes of the mixture sample;
Having a device. A device for identifying the attributes of complex and heterogeneous mixture samples including biological samples taken from organisms ,
The FID signal acquired by applying the test sample of the complex / heterogeneous mixture sample with unknown attributes to a nuclear magnetic resonance (NMR) analyzer is input, and the entire input FID signal is short. A signal processing unit that calculates a spectrogram of the FID signal by repeating time-frequency analysis, and calculates and outputs a feature amount of the sample mixture sample using numerical sequence data of the calculated spectrogram;
The feature amount of the test mixture sample is compared with an index for attribute identification set based on a plurality of known mixture samples whose attributes are the same as the test mixture sample, and the test mixture sample An attribute identifier for identifying the attributes of
Having a device. A device for identifying the attributes of complex and heterogeneous mixture samples including biological samples taken from organisms ,
The FID signal acquired by applying the test sample of the complex / heterogeneous mixture sample with unknown attributes to a nuclear magnetic resonance (NMR) analyzer is input, and the entire input FID signal is short. The spectrogram of the FID signal is calculated by repeating time-frequency analysis, the attenuation characteristic value of each frequency component of the calculated spectrogram is calculated, and the test mixture sample is calculated from the calculated attenuation characteristic value of each frequency component. A signal processing unit for creating and outputting a frequency component / attenuation characteristic value graph;
The frequency component / attenuation characteristic value graph for the test mixture sample is compared with the attribute identification graph created based on at least one known mixture sample of the same type as the test mixture sample and whose attributes are determined. An attribute identifying unit for identifying an attribute of the test mixture sample;
Having a device. A device for identifying the attributes of complex and heterogeneous mixture samples including biological samples taken from organisms ,
The FID signal acquired by applying the test sample of the complex / heterogeneous mixture sample with unknown attributes to a nuclear magnetic resonance (NMR) analyzer is input, and the entire input FID signal is short. A spectrogram of the FID signal is calculated by repeating time-frequency analysis, an attenuation characteristic value of each frequency component of the calculated spectrogram is calculated, and the test mixture sample is calculated using the calculated attenuation characteristic value of each frequency component A signal processing unit that calculates and outputs a feature amount for
The feature amount of the test mixture sample is compared with an index for attribute identification set based on a plurality of known mixture samples whose attributes are the same as the test mixture sample, and the test mixture sample An attribute identifier for identifying the attributes of
Having a device. An apparatus for processing a FID signal obtained by applying a complex / heterogeneous mixture sample containing a biological sample collected from an organism to a nuclear magnetic resonance (NMR) analyzer ,
A processing unit for calculating the spectrogram of the FID signal by repeating a short time frequency analysis throughout the inputted FID signal,
Output that outputs at least one of the numerical sequence data of the spectrogram calculated by the processing unit and the spectrogram image created from the spectrogram calculated by the processing unit as information indicating the characteristics of the complex / heterogeneous mixture sample And
Having a device. An apparatus for processing a FID signal obtained by applying a complex / heterogeneous mixture sample containing a biological sample collected from an organism to a nuclear magnetic resonance (NMR) analyzer ,
A processing unit for the calculating the spectrogram of the FID signal by repeating a short time frequency analysis throughout the inputted FID signal, and calculates an attenuation characteristic value of each frequency component of the calculated spectrogram,
At least one of the numerical string data of the attenuation characteristic value of each frequency component calculated by the processing unit and the frequency component / attenuation characteristic value graph created from the attenuation characteristic value of each frequency component calculated by the processing unit, An output unit that outputs information indicating characteristics of the complex / heterogeneous mixture sample;
Having a device. A computer program for causing a computer to execute the method according to claim 14 or 15 . A computer-readable recording medium storing the computer program according to claim 22 .