Data processing device and sample evaluation method

Abstract

In a data processing device, a multivariate analysis unit (18) performs multivariate analysis on a data set comprising a plurality of spectrograms obtained from a plurality of samples, and specifies the principal component of the data set as a multivariate analysis result. Each spectrogram has a first coordinate system. A distribution generation unit (20) generates a load distribution corresponding to the principal component as a multivariate analysis result. A drawing generation unit (22) generates a load drawing having a second coordinate system on the basis of the load distribution. The second coordinate system is the same as the first coordinate system.

Description

Technical Field

[0001] The present disclosure relates to a technology for processing data generated by NMR measurement performed by an NMR (nuclear magnetic resonance) device. Background Art

[0002] Methods for evaluating the properties of a mixture are known. The mixture is, for example, a biological sample such as serum, a mixed solution sample containing a polymer compound, etc.

[0003] Documents 1 and 2 describe methods for evaluating the characteristics of biological samples. Specifically, a method is described, comprising: a step of obtaining a FID (Free Induction Decay) signal from a biological sample using an NMR device; and a step of calculating a spectrogram by repeatedly performing time-frequency analysis on the FID signal as a whole. In addition, a method is described of generating a score plot by performing multivariate analysis on the spectrogram, and identifying the properties of the biological sample based on the score plot.

[0004] Documents 3 and 4 describe devices that perform bucket integration using bucket sets on target spectrum data to reduce target spectrum data to histogram data.

[0005] Prior art literature

[0006] Patent Literature

[0007] Document 1: Japanese Patent Application Publication No. 2015-114157

[0008] Document 2: Japanese Patent Application Publication No. 2019-158868

[0009] Document 3: Japanese Patent No. 5415476

[0010] Document 4: Japanese Patent No. 5020491 Summary of the invention

[0011] Problem that the invention aims to solve

[0012] As an example of time-frequency analysis, short-time Fourier transform (STFT) is known. As a parameter that needs to be set in short-time Fourier transform, there is a frame length. In short-time Fourier transform, time resolution and frequency resolution are in a trade-off relationship. In order to obtain the desired time resolution or the desired frequency resolution, it is necessary to set an appropriate frame length. It is desirable to provide information for setting an appropriate frame length to the user.

[0013] In the techniques described in Documents 1 and 2, although the properties of the sample can be identified from the score plot, the evaluation results of the score plot cannot be fed back to the parameters for time-frequency analysis. In addition, in the techniques described in Documents 1 and 2, it is impossible to examine the factors that cause the differences in properties between multiple samples (for example, the components and structures that affect the differences in properties). The same can be said for the techniques described in Documents 3 and 4.

[0014] An object of the present disclosure is to provide a new method for examining the characteristics of a sample from a spectrum generated by NMR measurement.

[0015] Solutions for solving problems

[0016] (1) The data processing device of the embodiment includes an acquisition unit, an analysis unit and a generation unit. The acquisition unit acquires multiple spectra generated by performing NMR measurement on multiple samples. The multiple spectra respectively have a first coordinate system, and the first coordinate system has a time axis and a frequency axis. The analysis unit determines the principal component of the data set and generates a load distribution corresponding to the principal component by performing multivariate analysis on the data set composed of the multiple spectra. The load distribution is composed of multiple loads corresponding to multiple variables in the multivariate analysis. The generation unit generates a load plot having a second coordinate system based on the load distribution, and the second coordinate system has a time axis and a frequency axis. The second coordinate system is the same as the first coordinate system.

[0017] According to the above configuration, by referring to or analyzing the load plot, in the time-frequency coordinate system, it is possible to easily identify the sample components that contribute more to the main component, the sample components that do not contribute much to the main component, etc. Based on the load plot, the parameter set for time-frequency analysis can be corrected, and the objects of multivariate analysis can be screened. The processor described later functions as the above-mentioned acquisition unit, analysis unit, and generation unit. In addition, the processor functions as some other units described below. The multivariate analysis unit and distribution generation unit described later are equivalent to the analysis unit.

[0018] In an embodiment, a specific spectrum is selected from a plurality of spectra. The data processing device of the embodiment further includes a display control unit, which displays the specific spectrum and the load plot in a row on a display. According to this configuration, the comparison between the specific spectrum and the load plot becomes easy.

[0019] The data processing device of the embodiment further includes a setting unit, which sets an analysis area defined by at least one of a time range and a frequency range for the load plot. The analysis unit performs multivariate analysis again on the limited data set corresponding to the analysis area. The data set usually includes an important part and an unimportant part in terms of clarifying the differences between multiple samples. According to the above configuration, multivariate analysis can be performed on an important part of the data set while excluding the unimportant part of the data set.

[0020] In an embodiment, on the load map, the time range and the frequency range are specified by a user, and the setting unit sets the analysis region according to the time range and the frequency range specified by the user.

[0021] In an embodiment, the setting unit sets a region that belongs to a time range and a frequency range and satisfies a specific load condition as an analysis region. For example, when the load at the coordinate of interest belongs to a specific load range, the specific load condition is satisfied. In an embodiment, the generating unit generates a load map by performing coordinate transformation on the load distribution.

[0022] The data processing device of the embodiment further includes: a unit for performing time-frequency analysis on multiple FID signals generated by NMR measurement of multiple samples according to a parameter set, thereby generating multiple spectra; a unit for correcting the parameter set based on the analysis area; and a unit for performing time-frequency analysis on multiple FID signals again according to the corrected parameter set, thereby generating multiple spectra constituting a limited data set. According to this structure, for example, a limited data set can be obtained at a desired time resolution or a desired frequency resolution. The parameter set can include a frame length, a frame interval, etc. The parameter set can also be corrected by the user.

[0023] The data processing device of the embodiment further includes a unit for cutting out a plurality of parts constituting a limited data set from the plurality of spectra based on the analysis region. According to this configuration, the plurality of spectra can be reused.

[0024] The program of the implementation mode is a program for executing a data processing method in a computer. The data processing method includes: an acquisition step of acquiring a plurality of spectra generated by performing NMR measurements on a plurality of samples, wherein the plurality of spectra respectively have a first coordinate system, and the first coordinate system has a time axis and a frequency axis; an analysis step of determining the principal component of the data set and generating a load distribution corresponding to the principal component by performing multivariate analysis on a data set consisting of a plurality of spectra, wherein the load distribution is composed of a plurality of loads corresponding to a plurality of variables in the multivariate analysis; and a generation step of generating a load plot having a second coordinate system based on the load distribution, and the second coordinate system has a time axis and a frequency axis. The second coordinate system is the same as the first coordinate system. The above-mentioned data processing method is equivalent to a sample evaluation method.

[0025] (2) One scheme of the present disclosure is a data processing device, characterized in that it includes: an acquisition unit, which acquires a plurality of spectra generated by performing NMR measurement on each of a plurality of samples and specified by time and frequency; a multivariate analysis unit, which generates a principal component for separating the plurality of samples according to the attributes of each sample by performing multivariate analysis on the plurality of spectra; a distribution generation unit, which generates a load distribution of a specific principal component corresponding to an index column determined by time information and frequency information contained in the plurality of spectra based on the result of the multivariate analysis performed by the multivariate analysis unit; and a drawing generation unit, which generates a load drawing by representing the load distribution of the specific principal component on a coordinate system specified by time and frequency.

[0026] The data processing device may further include a display control unit configured to display the spectrum of the designated sample and the load map in parallel on a display.

[0027] Alternatively, the data processing device further includes a determination unit, which determines an analysis area on the load drawing through a time range and a frequency range, and the multivariate analysis unit, the distribution generation unit, and the drawing generation unit perform respective processing on the analysis area as an object.

[0028] The determining unit may determine, when the time range and the frequency range are designated by a user on the load map, a region defined by the time range and the frequency range designated by the user as the analysis region.

[0029] The determination unit may determine, as the analysis region, a region which is defined by a time range and a frequency range and in which a load is included in a specific range.

[0030] The map generating unit may generate the load map by performing coordinate transformation on the load distribution.

[0031] One scheme of the present disclosure is a program, characterized in that a computer is made to function as the following units: an acquisition unit, which acquires multiple spectra generated by performing NMR measurement on each of a plurality of samples and specified by time and frequency; a multivariate analysis unit, which generates a principal component for separating the plurality of samples according to each sample attribute by performing multivariate analysis on the plurality of spectra; a distribution generation unit, which generates a load distribution of a specific principal component corresponding to an index column determined by time information and frequency information contained in the plurality of spectra based on the result of the multivariate analysis performed by the multivariate analysis unit; and a drawing generation unit, which generates a load drawing by representing the load distribution of the specific principal component on a coordinate system specified by time and frequency.

[0032] One scheme disclosed herein is a characteristic evaluation method, comprising: a first step of obtaining a plurality of spectra generated by performing NMR measurement on each of a plurality of samples and defined by time and frequency; a second step of generating a principal component for separating the plurality of samples according to the attributes of each sample by performing multivariate analysis on the plurality of spectra; a third step of generating a load distribution of a specific principal component corresponding to an index column determined by time information and frequency information contained in the plurality of spectra based on the result of the multivariate analysis performed in the second step; and a fourth step of generating a load plot by expressing the load distribution of the specific principal component in a coordinate system defined by time and frequency. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 It is a block diagram showing the configuration of a data processing system according to an embodiment.

[0034] Figure 2 This is a block diagram showing the hardware configuration of the data processing device according to the embodiment.

[0035] Figure 3 is a diagram showing an FID signal.

[0036] Figure 4 It is a figure which shows the spectrum of HSA.

[0037] Figure 5 It is a figure which shows the spectrum of HDL.

[0038] Figure 6 It is a figure which shows the spectrum of LDL.

[0039] Figure 7 is a diagram showing a score plot.

[0040] Figure 8 is a diagram showing load distribution.

[0041] Fig. 9 It is a figure which shows the loading plot corresponding to the 1st principal component (PC-1).

[0042] Fig.10 This is a diagram for explaining coordinate conversion between data in a two-dimensional format and data in a one-dimensional format.

[0043] Fig.11 is a diagram showing the spectrum and loading plot of HSA.

[0044] Fig.12 It is a diagram showing the analysis area set on the load map.

[0045] Fig.13 is a diagram showing a score plot.

[0046] Fig.14 is a diagram showing a load plot.

[0047] Fig.15 It is a figure which shows the spectrum generated from the serum of BKS18.

[0048] Fig.16 is a diagram showing a spectrum generated from the serum of Jcl.

[0049] Fig.17 is a diagram showing a score plot.

[0050] Fig.18 is a diagram showing a load plot.

[0051] Fig.19 is a diagram showing a score plot.

[0052] Fig. 20 is a diagram showing a load plot.

[0053] Fig.21 is a diagram showing spectra generated from serum of PD patients.

[0054] Fig. 22 is a diagram showing spectra generated from serum of non-PD patients.

[0055] Fig.23 is a diagram showing a score plot.

[0056] Fig.24 is a diagram showing a load plot.

[0057] Fig.25 is a diagram showing a load plot.

[0058] Fig.26 It is a diagram showing a data processing method according to an embodiment. DETAILED DESCRIPTION

[0059] Reference Figure 1 A data processing system according to an embodiment will be described. Figure 1 1 is a block diagram showing the configuration of a data processing system according to an embodiment. The data processing system according to the embodiment includes an NMR apparatus 10 and a data processing apparatus 12 .

[0060] The NMR device 10 is a device that irradiates a high-frequency signal to a sample placed in a static magnetic field and detects the high-frequency signal emitted from the sample. In this embodiment, the NMR device 10 performs NMR measurement on each of a plurality of different samples, thereby detecting the FID signal of each of the plurality of samples. The FID signal has a waveform indicating the time variation of the amplitude of the observed high-frequency signal.

[0061] Each sample is a mixture, for example, a biological sample such as serum, a mixed solution sample containing a polymer compound, etc. Of course, mixtures other than these can also be used as samples.

[0062] The data processing device 12 is a device that receives the FID signal detected by the NMR device 10 and processes the FID signal. The data processing device 12 processes the FID signal of each of the multiple samples. The data processing device 12 may be included in the NMR device 10, or may be a device physically different from the NMR device 10, or may be composed of multiple physically separated devices.

[0063] For example, the data processing device 12 includes a receiving unit 14 , a frequency analyzing unit 16 , a multivariate analyzing unit 18 , a distribution generating unit 20 , a plot generating unit 22 , a storage unit 24 , a display controlling unit 26 , a display unit 28 , an operating unit 30 , and a determining unit 31 .

[0064] The receiving unit 14 receives the FID signal of each sample detected by the NMR device 10. For example, the receiving unit 14 can receive the FID signal of each sample from the NMR device 10 by using wired communication or wireless communication. As another example, when the FID signal of each sample is stored in a storage device such as a memory or a hard disk and input to the data processing device 12 via the storage device, the receiving unit 14 can also receive the input FID signal. The receiving unit 14 is equivalent to an acquisition unit or an acquisition means.

[0065] The frequency analysis unit 16 generates a spectrogram for each of the multiple samples by repeatedly performing time-frequency analysis on the FID signal of each of the multiple samples. For example, short-time Fourier transform (STFT) is used as time-frequency analysis. A spectrogram is an image that represents time, frequency, and intensity. The frequency analysis unit 16 calculates the spectrum by performing frequency analysis while applying a window function to the FID signal obtained from each sample at each moment on the time axis. Each sample is a mixed sample, and each FID signal is a composite signal composed of multiple FID signal components corresponding to multiple sample components. The frequency analysis unit 16 generates a spectrogram based on multiple frequency spectra on the time axis. For example, a spectrogram is generated by representing the intensity (amplitude) of each frequency component using color or brightness on a two-dimensional coordinate system (first coordinate system) specified by the time axis and the frequency axis. A spectrogram is data in a two-dimensional format specified by time and frequency.

[0066] In addition, the spectrum of each sample may be generated by the NMR device 10. In this case, the receiving unit 14 receives the spectrum of each sample.

[0067] The multivariate analysis unit 18 performs multivariate analysis on a data set consisting of a plurality of spectra corresponding to a plurality of samples. For example, the multivariate analysis is principal component analysis (PCA), PLS discriminant analysis (PLS-DA), or SIMCA (Soft Independent Modeling of Class Analogy: partial space method) method. For example, two synthetic variables are generated. Of course, the number of synthetic variables generated is not particularly limited.

[0068] Each spectrum is equivalent to a two-dimensional intensity matrix. Specifically, the intensity matrix is ​​composed of N columns arranged in the direction of the time axis (horizontal axis), and each column is composed of M intensities arranged in the direction of the frequency axis (vertical axis). The M×N intensities that constitute each spectrum are called sub-datasets. When the number of samples that are the subject of analysis is S, the data set is composed of S sub-datasets. The data set is composed of S×M×N intensities. Each intensity is determined by time, frequency and sample number. Each intensity is actually treated as data in vector format.

[0069] For example, as a data set, in the case of imagining a two-dimensional data set, the data set has an index axis and a sample axis. The index axis is an axis representing M×N indices indicating M×N variables (specifically, intensities). The index is equivalent to a combination of time and frequency. The sample axis is an axis representing S samples.

[0070] In an embodiment, the multivariate analysis unit 18 performs principal component analysis on the data set constructed as above. Principal component analysis is a method of aggregating data in a multidimensional space into a low-dimensional space. In an embodiment, in the principal component analysis, the principal component that maximizes the variance of multiple spectra is searched, in other words, the principal component that significantly expresses the differences between multiple samples is searched. In fact, the search is repeated. Thus, in an embodiment, the first principal component and the second principal component are determined. More principal components can also be determined. In an embodiment, as a result of the principal component analysis, a score for the first principal component and a score for the second principal component are calculated for each spectrum, that is, for each sample. The score for the first principal component corresponds to the position on the axis of the first principal component. Similarly, the score for the second principal component corresponds to the position on the axis of the second principal component. Each principal component is defined by a linear equation containing multiple variables. In the linear equation, the coefficient acting on each variable is called a load (or load value). The load (loading) of each variable for a certain principal component represents the contribution (that is, weight) of each variable to the principal component. The multivariate analysis unit 18 generates a score map by, for example, plotting a plurality of scores corresponding to a plurality of samples on a two-dimensional coordinate system having a first principal component axis and a second principal component axis.

[0071] The distribution generation unit 20 generates a load distribution corresponding to each principal component based on the result of the multivariate analysis or as a result of the multivariate analysis. In an embodiment, the load distribution is a one-dimensional numerical sequence consisting of M×N loads corresponding to M×N variables. The M×N variables are determined by M×N indexes, so it can also be understood that the load distribution is composed of M×N loads corresponding to M×N indexes. The distribution generation unit 20 can also be integrated into the multivariate analysis unit 18.

[0072] The drawing generation unit 22 generates a loading plot as a two-dimensional graph from a load distribution (data in a one-dimensional format) corresponding to a specific principal component. The specific principal component is typically the first principal component. The specific principal component can also be selected by the user. In more detail, the load plot has a coordinate system (second coordinate system) having a time axis and a frequency axis. The drawing generation unit 22 plots each load constituting the load distribution on the coordinate system. At this time, the coordinates for drawing each load are determined from the index corresponding to each load. The size of the load is represented by color or brightness. In an embodiment, the second coordinate system of the load plot is the same as the first coordinate system of the spectrum. Conversely, the second coordinate system is matched with the first coordinate system so that the content of the spectrum can be specifically evaluated while referring to the load plot. According to the embodiment, the spectrum and the load plot can be easily compared. The load plot of the embodiment is different from a general load plot having a coordinate system having two principal component axes, and is a special load plot (TF-load plot) having a time (Time)-frequency (Frequency) coordinate system. In the load plot, the time variation of the load is shown for each frequency or frequency band. Such time variation cannot be seen in the conventional general load plot.

[0073] The storage unit 24 is implemented by a storage device and stores, for example, FID signals, spectrum data, results of multivariate analysis (eg, multiple scores, multiple loads, etc.), load distribution data, load maps, and the like.

[0074] The display control unit 26 controls the display of various information. For example, the display control unit 26 causes the display unit 28 to display FID signals, spectra, results of multivariate analysis, load distribution, load plots, and the like.

[0075] In this embodiment, the display control unit 26 displays the spectrum and the load plot side by side on the display unit 28. For example, when the user specifies a spectrum to be displayed, the display control unit 26 displays the specified spectrum and the load plot side by side on the display unit 28.

[0076] The display unit 28 is a display such as a liquid crystal display, an EL display, etc. The operation unit 30 is an input device such as a keyboard, a mouse, an input key, an operation panel, etc.

[0077] The determination unit 31 functions as a setting unit. Specifically, the determination unit 31 sets the area to be analyzed (hereinafter referred to as the "analysis area"). For example, the determination unit 31 sets the analysis area defined by the time range and the frequency range on the load drawing. For example, when the user specifies the time range and the frequency range on the load drawing, the determination unit 31 sets the two-dimensional area defined by the time range and the frequency range specified by the user as the analysis area. As another example, the determination unit 31 may also set an area that belongs to the time range and the frequency range and satisfies a specific load condition as the analysis area. For example, when the load at the coordinates of interest belongs to a specific load range, the specific load condition is satisfied. The specific load range is, for example, a range above a threshold value, a range above a lower limit value and less than an upper limit value, etc. The specific load range may be specified by the user or predetermined. For example, the determination unit 31 determines an area with a load above a threshold value in the load drawing as the analysis area. In addition, the determination unit 31 may not be included in the data processing device 12.

[0078] Below, refer to Figure 2 The hardware configuration of the data processing device 12 will be described. Figure 2 It is a block diagram showing the hardware configuration of the data processing device 12 .

[0079] For example, the data processing device 12 includes a communication device 32 , a UI (User Interface) 34 , a storage device 36 , and a processor 38 .

[0080] The communication device 32 includes one or more communication interfaces including a communication chip, a communication circuit, etc., and has the function of sending information to other devices and the function of receiving information from other devices. The communication device 32 may have a wireless communication function or a wired communication function.

[0081] UI34 is a user interface, including a display and an input device. The display is a liquid crystal display or an EL display, etc. The input device is a keyboard, a mouse, an input key or an operation panel, etc. The display unit 28 and the operation unit 30 are implemented by UI34. UI34 can also be a UI such as a touch panel that has both a display and an input device.

[0082] The storage device 36 is a device that constitutes one or more storage areas for storing data. The storage device 36 is, for example, a hard disk drive (HDD), a solid state drive (SSD), various memories (such as RAM, DRAM, NVRAM, ROM, etc.), other storage devices (such as optical disks, etc.), or a combination thereof. The storage unit 24 is implemented by the storage device 36.

[0083] Processor 38 controls the operation of each unit of data processing device 12. Frequency analysis unit 16, multivariate analysis unit 18, distribution generation unit 20, plot generation unit 22, display control unit 26, and determination unit 31 are implemented by processor 38. Storage device 36 may also be used for this implementation.

[0084] For example, the processor 38 is composed of a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), a DSP (Digital Signal Processor), other programmable logic devices, or electronic circuits.

[0085] Each function of the data processing device 12 can also be realized by the cooperation of hardware resources and software resources. For example, each function is realized by the CPU constituting the processor 38 reading and executing the program stored in the storage device 36. The program is stored in the storage device 36 via a recording medium such as a CD or DVD or via a communication path such as a network. As another example, each function of the data processing device 12 can also be realized by hardware resources such as electronic circuits.

[0086] The processing performed by the data processing device 12 will be described in detail below.

[0087] Figure 3 An example of an FID signal is shown. The FID signal 40 is an FID signal of a certain sample. The FID signal 40 is a signal detected by the NMR device 10, and has a waveform representing the time variation of the amplitude of the observed high-frequency signal. When the receiving unit 14 receives the FID signal 40, the frequency analysis unit 16 generates a spectrogram by performing a short-time Fourier transform on the FID signal 40.

[0088] Here, as an example, an aqueous solution of a serum standard substance is used as a sample. Specifically, three samples are used: HSA (human serum albumin solution), HDL (Lipoproteins, High Density, Human Plasma), and LDL (Lipoproteins, Low Density, Human Plasma).

[0089] Figure 4 The spectrum 42 of HSA is shown. Figure 5 A spectrum 44 of HDL is shown. Figure 6 A spectrum 46 of LDL is shown.

[0090] HSA, HDL and LDL are measured by the NMR device 10, thereby detecting the FID signals of HSA, HDL and LDL. The receiving unit 14 receives the FID signals of HSA, HDL and LDL. The frequency analysis unit 16 generates spectra of HSA, HDL and LDL by performing short-time Fourier transform on the FID signals of HSA, HDL and LDL.

[0091] In spectrograms 42, 44, and 46, the horizontal axis shows time and the vertical axis shows frequency. The intensity of each frequency component is represented by color (or brightness). The characteristics of the sample can be represented by the spectrogram. A color bar is shown on the right side of each spectrogram 42, 44, and 46. Since each sample is composed of multiple sample components, a striped pattern is seen in each spectrogram 42, 44, and 46. Specifically, each spectrogram 42, 44, and 46 has a plurality of stripes arranged in the frequency axis direction. Each stripe is parallel to the horizontal axis direction, and it shows the attenuation characteristics of the FID signal component.

[0092] The multivariate analysis unit 18 generates a plurality of synthetic variables by performing multivariate analysis on the data set composed of the spectra 42, 44, and 46. Here, as an example, the multivariate analysis unit 18 determines the first principal component (PC-1) and the second principal component (PC-2) by performing principal component analysis, and calculates two load distributions corresponding to the two principal components and the scores of each sample for each principal component.

[0093] Specifically, when the number of samples is S, the multivariate analysis unit 18 expresses each data (each intensity) constituting the S spectra in a vector format, and performs principal component analysis on the S×M×N data (that is, data set) expressed in a vector format. Through the principal component analysis, a load distribution is generated for each principal component. The load distribution is composed of M×N loads expressed in a vector format.

[0094] The multivariate analysis unit 18 determines the coordinates corresponding to the score of each sample (specifically, the first principal component score and the second principal component score) on the two-dimensional coordinate system defined by the first principal component (PC-1) axis and the second principal component (PC-2) axis, draws display elements at the coordinates, and thus generates a score plot. Figure 7 An example of the score plot is shown.

[0095] exist Figure 7In the score plot 48 shown, the horizontal axis shows the first principal component (PC-1) and the vertical axis shows the second principal component (PC-2). As an example, the contribution rate of the first principal component (PC-1) is 21%, and the contribution rate of the second principal component (PC-2) is 11%. The contribution rate is an indicator that represents the contribution of the principal component to the data as a whole. By referring to the contribution rate, the user can identify the importance of the principal component based on experience. The black circle mark shows the score of HSA. The black triangle mark shows the score of HDL. The white triangle mark shows the score of LDL.

[0096] The user can recognize the degree of separation of the plurality of samples by referring to the score plot 48. When the distance between the plurality of plots on the score plot 48 is short, it can be evaluated that the characteristics of the plurality of samples are similar to each other. On the other hand, when the distance between the plurality of plots is long, it can be evaluated that the characteristics of the plurality of samples are different from each other. Figure 7 In the example shown, the marker representing HSA is distributed separately from the marker representing HDL and the marker representing LDL on the score plot 48. Therefore, it can be evaluated that HSA is well separated from HDL and LDL in the score plot 48. When the degree of variance of the score in a certain principal component is large, it can be evaluated that the principal component has a greater contribution to the separation of a plurality of samples.

[0097] The distribution generating unit 20 generates a load distribution for a specific principal component based on the result of the principal component analysis performed by the multivariate analyzing unit 18. Here, as an example, the distribution generating unit 20 generates a load distribution for the first principal component (PC-1). Figure 8 The load distribution 50 is shown. The horizontal axis shows the index, and the vertical axis shows the load for the first principal component (PC-1). The index corresponds to a combination of time and frequency.

[0098] The map generation unit 22 generates a load map by expressing each load by color or brightness on a two-dimensional coordinate system defined by a time axis and a frequency axis based on the load distribution 50 .

[0099] Fig. 9 The load map 52 is shown. Specifically, the map generation unit 22 generates the load map 52 as a load distribution having a two-dimensional data format by performing coordinate conversion on the load distribution 50 having a one-dimensional data format.

[0100] In the load plot 52, the horizontal axis shows time and the vertical axis shows frequency. The magnitude of the load for the first principal component (PC-1) is expressed by color or brightness. The time axis in the load plot 52 is equivalent to the time axis in the spectrogram, and the frequency axis in the load plot 52 is equivalent to the frequency axis in the spectrogram.

[0101] Reference Fig.10The coordinate conversion from two-dimensional data to one-dimensional data and from one-dimensional data to two-dimensional data will be described. As described above, the spectrogram is two-dimensional data. On the other hand, the data input to the multivariate analysis unit 18 and the data output from the multivariate analysis unit 18 are one-dimensional data. Therefore, before performing the multivariate analysis, the two-dimensional spectrogram needs to be converted into one-dimensional data. In addition, in order to generate a two-dimensional load plot, the load distribution as one-dimensional data needs to be converted into two-dimensional data.

[0102] Fig.10 A specific example of coordinate conversion is shown. Reference numeral 54 indicates two-dimensional data, and reference numeral 56 indicates one-dimensional data. In the two-dimensional data 54, the longitudinal position is represented by m, and the lateral position is represented by n. The position in the one-dimensional data 56 is represented by k.

[0103] In the two-dimensional data 54, (n, m) indicates a position in the two-dimensional space or a value at the position. For example, (1, 1) indicates a position (1, 1) in the two-dimensional space or a value at the position (1, 1). Fig.10 In the example shown, m is any value from 1 to 3, and n is 1 or 2.

[0104] In the one-dimensional data 56, k indicates a position on the one-dimensional space or a value at the position. For example, (1) indicates a position (1) on the one-dimensional space or a value at the position (1).

[0105] The relationship among m, n and k is expressed by the following formula (1).

[0106] k=m+(n-1)×M···(1)

[0107] Here, M is the number of vertical positions in the two-dimensional data. Fig.10 In the example shown, M = 3. For example, when m = 1 and n = 1, k = 1.

[0108] According to the relationship defined by the above equation (1), two-dimensional data is converted into one-dimensional data, and one-dimensional data is converted into two-dimensional data.

[0109] For example, when principal component analysis is performed on a two-dimensional spectrum, the multivariate analysis unit 18 converts the spectrum as two-dimensional data into one-dimensional data (sub-data set) for each sample according to the above relationship. The multivariate analysis unit 18 performs principal component analysis on a data set that is a collection of multiple one-dimensional data corresponding to multiple samples. The distribution generation unit 20 generates a load distribution for each principal component as a result of the principal component analysis. In an embodiment, among multiple load distributions corresponding to multiple principal components, typically, the load distribution corresponding to the first principal component is used. That is, the drawing generation unit 22 converts the load distribution (one-dimensional data) corresponding to the first principal component into a load drawing (two-dimensional data) according to the above relationship.

[0110] The display control unit 26 displays the spectrum and the load plot in parallel on the display unit 28. For example, when the user specifies the spectrum 42 of HSA as a comparison target by using the operation unit 30 and gives an instruction for comparison display, as shown in FIG. Fig.11 As shown, the display control unit 26 displays the spectrum 42 and the load plot 52 side by side on the display unit 28 .

[0111] As described above, the load distribution 50 is converted into a load plot 52. The load plot 52 is a two-dimensional image having a time axis as the horizontal axis and a frequency axis as the vertical axis. In the load plot 52, each load is represented by a color or brightness. The horizontal axis of the load plot 52 is the same time axis as the horizontal axis of the spectrum 42, and the vertical axis of the load plot 52 is the same frequency axis as the vertical axis of the spectrum 42. On the spectrum 42, the intensity of the frequency component is represented by color or brightness, and on the load plot 52, the load is represented by color or brightness. The second coordinate system of the load plot 52 is the same coordinate system as the first coordinate system of the spectrum 42, so it is easy for the user to compare the spectrum 42 with the load plot 52.

[0112] exist Fig.11 In the example shown, the spectrum 42 and the load plot 52 are displayed side by side, but this display is only an example. The display control unit 26 may display the spectrum 44 or the spectrum 46 and the load plot 52 side by side on the display unit 28 instead of the spectrum 42. For example, when the user specifies the spectrum 44 or the spectrum 46 as a comparison object, the display control unit 26 displays the specified spectrum and the load plot 52 side by side on the display unit 28.

[0113] The display control unit 26 may display a plurality of spectra and the load plot 52 side by side on the display unit 28. For example, when the user specifies the spectra 44 and 46, the display control unit 26 displays the spectra 44 and 46 and the load plot 52 side by side on the display unit 28. This makes it easy for the user to compare a plurality of spectra and the load plot.

[0114] The user can conduct various investigations based on the two-dimensional load distribution represented as the load plot 52. For example, the user can conduct various investigations from the perspectives of load changes in the frequency axis direction or the time axis direction (e.g., changes in color or brightness), positions on the frequency axis and / or positions on the time axis where characteristic parts exist, and the relationship between multiple positions where multiple characteristic parts exist.

[0115] In addition, the user can conduct various investigations by comparing the spectrogram and the load plot. For example, the user can determine the position of the high-intensity part on the frequency axis on the spectrogram, and observe what kind of local state is produced at the determined position on the frequency axis on the load plot. In contrast, the user can determine the position with a large load or the position with a small load (such as the position on the frequency axis) on the load plot, and observe what kind of local state is produced at the determined position (such as the position on the frequency axis) on the spectrogram. The same can be said about the position on the time axis. In addition, the user can compare the intensity change and the load change by observing within the same frequency range in the spectrogram and the load plot, or compare the intensity change and the load change by observing within the same time range. The spectrogram and the load plot have the same coordinate system, so their comparison is easy.

[0116] In the spectrum, the position on the frequency axis where the intensity becomes high varies depending on the components contained in the sample and the structure of the components. Generally, in a spectrum generated from a known sample, it is known what kind of intensity distribution appears in which frequency region. By referring to the intensity distribution in the spectrum, the components contained in the sample can be determined. In addition, the decay rate (time change) of the NMR signal varies depending on the properties of the components, so the position of the intensity distribution on the time axis varies depending on the properties of the components contained in the sample.

[0117] By referring to the distribution form on the load plot (for example, load changes in the frequency axis direction, characteristic parts on the frequency axis, etc.) and comparing it with the distribution form on the spectrum (for example, intensity changes in the frequency axis direction, characteristic parts on the frequency axis, etc.), the user can examine the differences in components and structures that affect the differences in properties in multiple samples.

[0118] For example, by comparing the attenuation characteristics of each component in the spectrum with the load distribution in the load plot, the user can evaluate the characteristics of each component that makes up the sample, or analyze and examine the factors (for example, the composition and structure of the sample) that cause changes or differences in position on the frequency axis in the load plot.

[0119] For example, the user determines the position on the frequency axis where the intensity is high in the spectrum. Generally, it is known what kind of intensity distribution appears in which frequency region. Therefore, by determining the position on the frequency where the signal intensity is high, the components contained in the sample can be determined. In the load plot, when the load is large at a certain position on the frequency axis, it is inferred that the component corresponding to the determined position contributes more to the separation of multiple samples. In other words, it is inferred that the component has a high contribution rate to the separation of multiple samples. In this way, by comparing the spectrum and the load plot, it is possible to infer the components that can contribute to the separation of multiple samples.

[0120] The analysis result of the load plot can also be fed back to the time-frequency analysis performed by the frequency analysis unit 16. For example, in the load plot, when a characteristic load distribution is generated at a position on a certain frequency axis (for example, when a large load distribution or a small load distribution is generated locally), sometimes it is desirable to improve the frequency resolution to examine the characteristic load distribution in more detail. In this case, a short-time Fourier transform is performed on each FID signal on the basis of setting a value that improves the frequency resolution as the frame length of the short-time Fourier transform. Thus, the frequency resolution of the spectra corresponding to each sample is improved. On this basis, a score plot and a load plot are generated by principal component analysis of multiple spectra corresponding to multiple samples. By examining the load plot, it is possible to determine whether the frequency resolution or the time resolution should be emphasized in the short-time Fourier transform. Other parameters in the time-frequency analysis (for example, frame interval, window function form) can also be changed based on the load plot. When the parameter set for time-frequency analysis is changed, the time-frequency analysis is performed again on the multiple FID signals that have been obtained according to the changed parameter set, thereby generating multiple spectra. The principal component analysis is performed again on the data set consisting of multiple spectra. The score plot and loading plot are generated again as the result of the principal component analysis.

[0121] Below, refer to Fig.12 The processing of the determination unit 31 will be described. Fig.12 A load plot 52 is shown.

[0122] For example, the load map 52 is displayed on the display unit 28. The user operates the operation unit 30 to specify a time range and a frequency range on the load map 52. The determination unit 31 sets a region 58 defined by the time range and frequency range specified by the user as an analysis region.

[0123] When the analysis area is set, the frequency analysis unit 16 takes the part or component corresponding to the analysis area in each FID signal as the object, and performs short-time Fourier transform again, thereby generating a limited spectrum. The multivariate analysis unit 18 performs multivariate analysis on multiple limited spectra (limited data sets) corresponding to multiple samples. The distribution generation unit 20 generates a load distribution based on the results of the multivariate analysis, and the drawing generation unit 22 generates a load drawing from the load distribution. Thus, a load drawing corresponding to the analysis area is generated. The load drawing can be said to be a limited load drawing. Preferably, before performing the short-time Fourier transform again, the parameter set for the short-time Fourier transform is changed.

[0124] For example, the user specifies the region estimated to contribute to the separation of the plurality of samples as the analysis region by referring to the load plot 52. That is, the user specifies the analysis region excluding the region estimated not to contribute to the separation of the plurality of samples. In this way, the result of the principal component analysis can be improved.

[0125] exist Fig.12 In the example shown, the analysis area is determined by the rectangular area 58, but the analysis area may be determined by an area of ​​a shape other than a rectangle (e.g., a circle, an ellipse, or an area of ​​any shape). For example, the user may specify the analysis area in a manner that includes one or more characteristic parts.

[0126] The determination unit 31 may also determine an area that satisfies the load condition as the analysis area. For example, when the load is included in a specific range (for example, a range above a threshold value, a range below a threshold value, or a range above a lower limit value and below an upper limit value), the load condition is satisfied. Fig.12 In the example shown, the identification unit 31 may identify the analysis region so as to surround a portion having a load equal to or greater than a threshold value.

[0127] When the parameter set for time-frequency analysis does not need to be changed after setting the analysis region, multiple parts corresponding to the analysis region can be cut out from the multiple spectra that have been generated without re-executing the time-frequency analysis. The multivariate analysis unit 18 performs principal component analysis on the limited data set consisting of the multiple cut-out parts. A new score plot and a new load plot are generated as the result of the principal component analysis. The cutting out of multiple parts can be performed by the determination unit 31 or the multivariate analysis unit 18.

[0128] Several embodiments are described below.

[0129] Example 1

[0130] The samples of Example 1 are HSA, HDL, and LDL. The purpose of Example 1 is to perform NMR measurement on albumin and lipoprotein in human serum and to visualize the differences between the time-frequency characteristics of a plurality of NMR signals obtained from these samples.

[0131] The details of HSA, HDL, and LDL are shown below.

[0132] ·HSA (Human Serum Albumin Solution)

[0133] Manufacturer: NMIJ

[0134] Lot: 144

[0135] Material number: NMIJ CRM 6202-a

[0136] HDL (Lipoproteins High Density, Human Plasma)

[0137] Manufacturer: Calbiochem

[0138] Batch number: 3816722

[0139] Material number: 437641-10MG

[0140] LDL (Lipoproteins Low Density, Human Plasma)

[0141] Manufacturer: Calbiochem

[0142] Batch number: 3883470&3912892

[0143] Material number: 437644-10MG

[0144] (1) Device composition

[0145] JNM-ECZ400R manufactured by JEOL Ltd. was used as the NMR apparatus 10. DELTA software (version 5.3.2 (JEOL Ltd.)) was used as the software for data processing and control in the NMR apparatus 10. An application program (hereinafter referred to as "STFT tool") developed for MATLAB (TM) (The MathWorks, Inc.: MathWorks) and Unscrambler X version 11 (manufactured by Camo Software) were installed in the data processing apparatus 12.

[0146] (2) Sample creation and measurement

[0147] For each of the above-mentioned samples, the sample (100 μL) was mixed with 500 μL of deuterated water (heavy water, manufactured by Sigma-Aldrich), and the sample solution thus generated was injected into a sample tube for NMR with an outer diameter of 5 mm. The temperature in the NMR device 10 was set to 30°C. By performing 1H-NMR measurement on the above-mentioned three samples as objects, the FID signals of the three samples were detected. The resonance frequency of the signal from the light water present in each sample solution was set to the frequency of the observation center. By using the DANTE (Delays Alternating with Nutation for Tailored Excitation) pulse method, which is an example of a pulse program, as a measurement method, the signal from light water can be suppressed. As a result, the subsequent analysis process becomes easy and good quality results can be obtained. Depending on the purpose of the measurement, multiple sample tubes containing the same sample can be prepared, and multiple sample tubes containing multiple sample solutions with different dilution concentrations can also be prepared.

[0148] (3) Analysis

[0149] (3-1) Time-frequency analysis

[0150] The STFT tool is started, and the FID signals of the plurality of samples obtained by the above measurement are taken into the tool. The sampling frequency and the number of data points are input to the tool, and on this basis, time-frequency analysis is performed on each FID signal. Figure 4 The spectrum 42 shown is an example of a spectrum of HSA. Figure 5 The spectrum 44 shown is an example of a spectrum of HDL. Figure 6 The spectrum 46 shown is an example of a spectrum of LDL. In order to shorten the time required for analysis, the frequency range and time range to be analyzed may be narrowed down.

[0151] (3-2) Multivariate Analysis 1

[0152] Start Unscrambler X and import all the data (dataset) obtained by the above-mentioned time-frequency analysis into the software. In addition, input the data file name, the label for subsequent drawing display and the label for grouping processing into the software. Then, perform principal component analysis (PCA) on the data set. Referring to the generated score drawing, confirm whether each sample is well separated. Specifically, confirm the score drawing with the axis with the largest variance (the first principal component axis (PC-1)) and the axis with the second largest variance (the second principal component axis (PC-2)). When sufficient separation cannot be confirmed in the score drawing, it is necessary to confirm other drawings such as PC-3 and PC-4. Confirm the load distribution corresponding to the obtained score drawing and save it. Figure 7 The score map 48 shown is an example of the score map of Embodiment 1.

[0153] (3-3) Load plot Figure 1

[0154] Start the STFT tool, open a spectrum generated in the time-frequency analysis, and read the load distributions saved in the multivariate analysis 1 (the load distribution of PC-1, the load distribution of PC-2). You can also read the load distributions of the principal components after PC-3 while observing the separation quality and contribution rate. Using a specific load distribution among the multiple load distributions read in, a load plot is generated and displayed. The time axis and frequency axis in the load plot correspond to the time axis and frequency axis in the spectrum. Confirm in which area the characteristics appear in the load plot. Fig. 9 The load plot 52 shown is an example of a load plot of Example 1. From the frequency information at which the characteristic portion appears, it is possible to examine which component in the sample is related to the characteristic. In addition, from the time information at which the characteristic portion appears, it is possible to examine what kind of property the characteristic is derived from.

[0155] (3-4) Multivariate Analysis 2

[0156] In multivariate analysis 2, partial least squares discriminant analysis (PLS-DA) was performed instead of principal component analysis (PCA). Unscrambler X was started and partial least squares discriminant analysis (PLS-DA) was performed on the data set in the same manner as in multivariate analysis 1 above. Fig.13 A score plot 60 generated by partial least squares discriminant analysis (PLS-DA) is shown.

[0157] (3-5) Load plot Figure 2

[0158] Start the STFT tool and press the same as above to draw the load Figure 1In the same way as described in , open a spectrum generated in the time-frequency analysis and read the load distribution (the load distribution of PC-1 and the load distribution of PC-2) saved in the multivariate analysis 2. Generate and display a load plot using the read load distribution. Confirm in which area the feature appears in the load plot. Fig.14 The load plot 62 is shown. From the frequency information at which the characteristic portion appears, it is possible to examine which component in the sample is related to the characteristic. From the time information at which the characteristic portion appears, it is possible to examine what kind of property the characteristic comes from.

[0159] In Example 1, multiple samples can be clearly separated on the score plot. In addition, differences can be read from multiple frequency positions from multiple components included in HSA, multiple frequency positions from multiple components included in HDL, and multiple frequency positions from multiple components included in LDL. That is, it is shown that the load plot is effective as information for analyzing which component in each sample the separation on the score plot comes from.

[0160] Example 2

[0161] The purpose of Example 2 is to perform serum pattern analysis on each of diabetic model mice (BKS.Cg db / db) and healthy mice (Jcl:ICR) and to discriminate the onset of arteriosclerosis based on the analysis results.

[0162] (1) Background and Purpose

[0163] Diabetes can cause arteriosclerosis and various subsequent cardiovascular events caused by persistent high blood sugar. In clinical practice, detailed examination of arteriosclerotic lesions requires other examinations in addition to blood tests. If arteriosclerotic lesions can be evaluated through blood samples, it is considered to be clinically useful. However, diabetes and subsequent arteriosclerotic lesions are related to complex molecular biological processes, and it is not easy to detect all related substances. At present, there is no established test method that can detect and evaluate arteriosclerotic lesions from blood samples of diabetic patients. The NMR-based analysis method is an analysis method that can evaluate the physicochemical properties of serum. The inventors of this application proposed a hypothesis that the blood state related to the progression of arteriosclerotic lesions with diabetes as the background can be distinguished from the healthy state by using this analysis method, and analyzed the difference between the serum of diabetic model mice (BKS.Cg db / db) and the serum of healthy mice (Jcl:ICR). Diabetic model mice (BKS.Cg db / db) are mice that show symptoms of hyperglycemia at an early stage. Serum and carotid artery tissue were collected from diabetic model mice (BKS.Cg db / db) and healthy mice (Jcl:ICR). Each serum was analyzed using an NMR device, and each carotid artery tissue was pathologically examined. Based on the two NMR analysis results and the two pathological examination results, it was studied whether the NMR analysis results of serum can be used to detect symptoms related to arteriosclerosis and evaluate their progression.

[0164] (2) Animal husbandry

[0165] Diabetic model mice (BKS.Cg db / db) and healthy mice (Jcl: ICR) (both purchased from CLEA, Japan) were routinely raised in the clean room of the Experimental Animal Management Room of Nippon Medical University. The food used was a general solid sample (MF manufactured by Oriental Yeast Co., Ltd.), and the mice were allowed to eat freely and drink water freely.

[0166] (3) Methods of sample collection and pathological examination

[0167] After mice were subjected to maintenance anesthesia using isoflurane to obtain sufficient analgesia, the mice were opened, heart blood (about 1 mL) was collected, and then the mice were euthanized. After collecting heart blood, carotid artery tissue was collected in order to perform histological examination by HE staining. After collecting carotid artery tissue, the carotid artery tissue was immediately infiltrated with formalin solution, and the processing and HE staining after embedding were performed, and then, the observation of carotid artery tissue was performed using a microscope.

[0168] The blood was centrifuged to separate the serum, and the separated serum, i.e., the sample for NMR measurement, was stored at -80°C until the NMR measurement.

[0169] The same operation and treatment as in Example 1 were performed using serum samples obtained from mice with arteriosclerosis symptoms, thereby generating a spectrum, a score map, and a loading map for each serum sample.

[0170] Fig.15 The spectrum 64 of BKS is shown. Fig.16 The spectrum 66 of Jcl is shown. Fig.17 Shown is a score plot 68. The score plot 68 is a score plot generated by performing a principal component analysis (PCA). Fig.18 A loading plot 70 is shown. The loading plot 70 is a loading plot generated based on the results of a principal component analysis (PCA).

[0171] Fig.19 Another score plot 72 is shown. Score plot 72 is a score plot generated by partial least squares discriminant analysis (PLS-DA). Fig. 20 A loading plot 74 is shown. The loading plot 74 is a loading plot generated based on the results of a partial least squares discriminant analysis (PLS-DA).

[0172] Significant feature quantities (separation) were confirmed in the score plot of the sample with symptoms. When the factors that have a significant impact on the difference in feature quantities are confirmed in the spectrum, large feature quantities can be confirmed in the signals from HDL, LDL, and glucose. This result is consistent with the medical viewpoint that it is the cause of arteriosclerosis.

[0173] Pathological examination results

[0174] Pathological images of carotid artery tissue were observed. In healthy mice (Jcl: ICR), mild diffuse intimal hypertrophy was observed in all individuals, but no atherosclerotic plaques were confirmed. In diabetic model mice (BKS.Cgdb / db), the degree of intimal hypertrophy was stronger than that of healthy mice (Jcl: ICR) at 18 weeks of age, and the presence of atherosclerotic plaques became obvious in more than half of the individuals, which clearly showed that arteriosclerosis progressed in a short period of time at this age.

[0175] In Example 2, each sample can be clearly separated on the score plot. In addition, the difference in frequency position of each component such as glucose can be read from the load plot. This is consistent with the medical viewpoint that it is the cause of arteriosclerosis. That is, Example 2 also shows that the load plot is effective as information for analyzing which component in the sample the separation on the score plot comes from.

[0176] Example 3

[0177] The purpose of Example 3 is to identify Parkinson's disease (PD) patients by using serum samples.

[0178] (1) Purpose

[0179] Differentiation between Parkinson's disease (PD) and various diseases presenting with Parkinson's symptoms, such as multiple system atrophy and progressive supranuclear palsy, is not necessarily easy in the early stages of the disease, and no specific blood biomarkers have been reported for PD. As an example of a marker that can distinguish PD from non-PD presenting with Parkinson's disease with the highest accuracy at the current time point, DAT-SPECT and MIBG myocardial scintigraphy can be cited. By performing DAT-SPECT and MIBG myocardial scintigraphy on PD and non-PD, using the NMR analysis method developed by the inventors of the present application, it was studied whether PD and non-PD can be distinguished by serum samples.

[0180] (2) Target

[0181] With the approval of the Ethics Committee, serum was collected from 10 patients who presented with any one of tremor, bradykinesia, muscle rigidity or postural reflex disorder at the Chiba Kita General Hospital of Nippon Medical University from October 2020 to February 2022. A database was created based on blood biochemistry data, head MRI data, MIGB myocardial scintigraphy data, and DAT-SPECT data. For the 10 cases that completed the diagnosis and examination, NMR analysis was performed on the serum collected from each patient in the same steps as in Example 1. Thus, spectra, score plots, and load plots were generated for each serum.

[0182] Fig.21 A spectrum 76 obtained from the serum of a PD patient is shown. Fig. 22 Spectrum 78 obtained from serum of a non-PD patient is shown. Fig.23 A score plot 80 is shown. The score plot 80 is a score plot generated by partial least squares discriminant analysis (PLS-DA). Fig.24 A load plot 82 is shown. Fig.25 A load plot 84 is shown. The load plots 82 and 84 are load plots generated based on the results of partial least squares discriminant analysis (PLS-DA). Contour lines are shown on the load plot 84.

[0183] (3) Diagnosis

[0184] PD was diagnosed when the International Parkinson and Movement Disorder Society (MDS) diagnostic criteria (2015) of "clinically reliable PD" and "clinically approximately reliable PD" were met. MIBG myocardial scintigraphy was considered abnormal when the heart-to-mediastinum ratio was 2.2 or less in either the early or late phase. DAT-SPECT was confirmed to be abnormal by visual evaluation and quantitative evaluation based on SBR.

[0185] In Example 3, the PLS-DA method was used to analyze the serum of PD patients (abnormal MIBG and abnormal DAT-SPECT) and non-PD patients (normal MIBG and normal DAT-SPECT). As a result, each group formed a cluster on the score plot, and the two groups were distributed in significantly different areas. This shows the possibility of distinguishing PD patients and non-PD patients by NMR analysis of serum.

[0186] use Fig.26 The data processing method (sample evaluation method) of the embodiment is summarized. Fig.26 The flow of processing executed by the processor is shown.

[0187] In S10, NMR measurement is performed on a plurality of samples 88, thereby obtaining a plurality of FID signals 90. In S12, a short-time Fourier transform (STFT) is applied to the plurality of FID signals 90. In S14, a spectrogram 92 is generated based on a series of frequency spectra generated by the short-time Fourier transform for each FID signal 90. That is, a plurality of spectrograms 92 corresponding to the plurality of samples 88 are generated. Each spectrogram 92 has a first coordinate system. In the first coordinate system, the horizontal axis is the time axis and the vertical axis is the frequency axis.

[0188] In S16, a data set 94 is constructed based on the multiple spectra 92. Specifically, the data set 94 is composed of multiple sub-data sets 94a corresponding to the multiple samples. Each sub-data set 94a is composed of m×n intensities constituting each spectrum. i indicates the index axis, and j indicates the sample axis.

[0189] In S18, a principal component analysis (PCA) is performed on the data set 94. In S22, a score plot 98 is generated as a result of the principal component analysis. In the illustrated example, the score plot has a first principal component axis (PC-1) and a second principal component axis (PC-2). In S22, a general load plot may also be generated. For example, a general load plot is generated based on a load distribution corresponding to the first principal component and a load distribution corresponding to the second principal component, and has a first principal component axis (PC-1) and a second principal component axis (PC-2).

[0190] In S20, as a result of the principal component analysis, a load plot (TF-load plot) 96 of the embodiment is generated. For example, based on the load distribution 95 corresponding to the first principal component, a load plot 96 corresponding to the first principal component is generated. The load plot 96 has a second coordinate system. In the second coordinate system, the horizontal axis is the time axis and the vertical axis is the frequency axis. The second coordinate system is the same as the first coordinate system.

[0191] In S23, based on the user's designation, a specific spectrum 102 is selected from the generated plurality of spectra 92. In S24, the selected spectrum 102 and the load plot 96 are displayed on the display 100.

[0192] As needed, in S25, the user who has referred to the load plot 96 changes the parameter set for the short-time Fourier transform. For example, the frame length is changed. In this case, the short-time Fourier transform is performed again on the plurality of FID signals 90 according to the changed parameter set. The principal component analysis is applied to the new data set consisting of the plurality of new spectra generated thereby.

[0193] In addition, as needed, an analysis region 106 is set on the load plot 96. In this case, in S26, short-time Fourier transform is performed again on the plurality of FID signals 90 under the conversion conditions according to the analysis region 106. A plurality of new spectra generated by the re-execution of the short-time Fourier transform constitute a restricted data set. A principal component analysis is applied to the restricted data set. In addition, the above parameter set may also be changed before executing S26.

[0194] In the case where the analysis region 106 is set on the load plot 96, it is also possible to cut out a plurality of parts 94 corresponding to the analysis region 106 from the plurality of spectra 92 that have been generated without re-execution of the short-time Fourier transform. Specifically, in S27, the plurality of spectra 92 that have been generated are obtained, and in S29, the parts 94 corresponding to the analysis region 106 are cut out from each of the obtained spectra 92. In S30, the restricted data set 108 is formed by the cut-out plurality of parts 94. In S32, the principal component analysis is applied to the restricted data set 108. Thus, for example, in S36, a score plot and / or a general load plot is generated as a result of the principal component analysis. In S34, a load plot of the embodiment is generated based on the load distribution generated by the principal component analysis. The load plot is generated, for example, from the load distribution corresponding to the first principal component. Then, in S38, the load plot is displayed as needed. The analysis region may also be set for the load plot. In this case, the series of processes described above are performed again.

[0195] As described above, according to the present embodiment, the content of the spectrum and the content of the load drawing can be examined while comparing the spectrum and the load drawing. Thus, the factors that cause differences in characteristics and attributes in multiple samples can be analyzed. As a result, for example, in the medical field, preventive medicine (that is, medical treatment that predicts a disease before a case occurs and performs therapeutic intervention to prevent or delay the onset of the disease) can be expected. For example, ultra-early diagnosis, determination of treatment guidelines, determination of treatment effects, and prognosis prediction can be expected. In addition, according to the present embodiment, even without using known attribute recognition images, it is possible to search for cases.

Claims

1. A data processing device, characterized in that: Include: An acquisition unit (14) that acquires a plurality of spectra generated by performing NMR measurement on a plurality of samples, wherein each of the plurality of spectra has a first coordinate system, and the first coordinate system has a time axis and a frequency axis; An analysis unit (18, 20) that determines a principal component of the data set and generates a load distribution corresponding to the principal component by performing a multivariate analysis on the data set composed of the plurality of spectra, wherein the load distribution is composed of a plurality of loads corresponding to a plurality of variables in the multivariate analysis; as well as A generating unit (22) generates a load plot having a second coordinate system based on the load distribution, the second coordinate system having a time axis and a frequency axis, The second coordinate system is the same as the first coordinate system.

2. The data processing device according to claim 1, characterized in that: selecting a specific spectrum from among the plurality of spectra, The data processing device further comprises a display control unit (26), wherein the display control unit (26) arranges and displays the specific spectrum and the load plot on a display.

3. The data processing device according to claim 1, characterized in that: The method further comprises a setting unit (31) for setting an analysis region defined by at least one of a time range and a frequency range for the load map. The analysis unit (18, 20) performs the multivariate analysis again on the limited data set corresponding to the analysis area.

4. The data processing device according to claim 3, characterized in that: On the load plot, the time range and the frequency range are specified by the user, The setting unit (31) sets the analysis area according to the time range and the frequency range specified by the user.

5. The data processing device according to claim 3, characterized in that: The setting unit (31) sets a region that belongs to the time range and the frequency range and satisfies a specific load condition as the analysis region.

6. The data processing device according to claim 1, characterized in that: The generating unit (22) generates the load map by performing coordinate transformation on the load distribution.

7. The data processing device according to claim 3, characterized in that: The method further includes a unit (S29) for cutting out the limited data set from the data set based on the parsing region.

8. The data processing device according to claim 3, characterized in that: Also includes: A unit (16, S12) for generating the plurality of spectra by performing time-frequency analysis on a plurality of FID signals generated by performing the NMR measurement on the plurality of samples according to a parameter set; A unit (S26) for modifying the parameter set based on the analysis region; as well as A unit (16, S12) is provided for re-performing the time-frequency analysis on the plurality of FID signals according to the modified parameter set, thereby generating a plurality of spectrograms constituting the defined data set.

9. A program for executing a data processing method in a computer, characterized in that The data processing method comprises: An acquisition step (14) of acquiring a plurality of spectra generated by performing NMR measurement on a plurality of samples, wherein each of the plurality of spectra has a first coordinate system, and the first coordinate system has a time axis and a frequency axis; an analysis step (18, 20, S18), determining a principal component of the data set and generating a load distribution corresponding to the principal component by performing a multivariate analysis on the data set composed of the plurality of spectra, wherein the load distribution is composed of a plurality of loads corresponding to a plurality of variables in the multivariate analysis; and A generation step (22, S20) of generating a load plot having a second coordinate system based on the load distribution, wherein the second coordinate system has a time axis and a frequency axis, The second coordinate system is the same as the first coordinate system.

10. A sample evaluation method, characterized in that: Include: An acquisition step (14) of acquiring a plurality of spectra generated by performing NMR measurement on a plurality of samples, wherein each of the plurality of spectra has a first coordinate system, and the first coordinate system has a time axis and a frequency axis; an analysis step (18, 20, S18), determining a principal component of the data set and generating a load distribution corresponding to the principal component by performing a multivariate analysis on the data set composed of the plurality of spectra, wherein the load distribution is composed of a plurality of loads corresponding to a plurality of variables in the multivariate analysis; and A generation step (22, S20) of generating a load plot having a second coordinate system based on the load distribution, wherein the second coordinate system has a time axis and a frequency axis, The second coordinate system is the same as the first coordinate system.

Images