Dye data acquisition method, dye data acquisition device, and dye data acquisition program

Abstract

Provided is a data processing device (5) for processing spectrum data, which is a distribution of intensity values of fluorescence in each of C (C is an integer of 2 or more) wavelength bands with N (N is an integer of 2 or more) objects to be analyzed as objects, on the basis of the intensity values of the objects to be analyzed in each of the C wavelength bands, and for processing the spectrum data on the basis of the intensity values of the objects to be analyzed in each of the C wavelength bands. N analysis objects are clustered into L object groups (L is an integer of 2 to N-1 inclusive), L clustering matrices in which the intensity values of the C wavelength bands are arranged for each of the clustered object groups are generated, and statistical values of the intensity values of the object groups in the C wavelength bands are calculated for each of the L clustering matrices. The statistical values in the C wavelength bands of each of the L clustering matrices are used to perform de-mixing for the C wavelength bands, and K pieces of dye data representing the distribution of each of K fluorescent dyes (K is an integer of 2-C inclusive) are generated.

Description

Technical Field

[0001] One aspect of the embodiment relates to a pigment data acquisition method, a pigment data acquisition device, and a pigment data acquisition program. Background Art

[0002] Flow cytometry has long been a well-known technique for counting, screening, and analyzing the characteristics of samples such as cells using lasers. For example, Patent Document 1 below discloses clustering cells using fluorescence data of various colors of light emitted from cells by a flow cytometer. Furthermore, Non-Patent Document 2 below discloses a technique for clustering fluorescence spectra generated by unmixing the fluorescence data output by a flow cytometer to generate fluorescence spectra for each fluorochrome.

[0003] Prior art literature

[0004] Patent Literature

[0005] Patent Document 1: Japanese Patent Application Laid-Open No. 2020-87465

[0006] Patent Document 2: Japanese Patent Application Laid-Open No. 2021-36224 Summary of the Invention

[0007] Problems to be solved by the invention

[0008] Conventional methods such as those described above tend to struggle to obtain high-precision fluorescence spectra when the mixing matrix used for unmixing is unknown. For example, this can be difficult to obtain when observation conditions vary due to changes in the type of pigment being observed or the type of excitation light used.

[0009] Therefore, one aspect of the embodiment has been made in view of this problem, and the problem is to provide a pigment data acquisition method, a pigment data acquisition device, and a pigment data acquisition program that can obtain a high-precision fluorescence spectrum even when observation conditions change.

[0010] Technical solutions to problems

[0011] A pigment data acquisition method according to a first aspect of the embodiment comprises: a data acquisition step of acquiring spectral data for N (N is an integer greater than or equal to 2) objects, the spectral data being a distribution of fluorescence intensity values ​​at each of C (C is an integer greater than or equal to 2) detection wavelengths; a clustering step of clustering the N objects into L (L is an integer greater than or equal to 2 and less than or equal to N-1) object groups based on the intensity values ​​of each object at each of the C detection wavelengths, and generating L cluster matrices in which the intensity values ​​at each of the C detection wavelengths are arranged for each clustered object group; a calculation step of calculating, for each of the L cluster matrices, a statistical value of the intensity values ​​of the object group at the C detection wavelengths; and a data generation step of performing demixing for the C detection wavelengths using the statistical values ​​at the C detection wavelengths of each cluster matrix in the L cluster matrices to generate K pieces of pigment data representing the distribution of each of K (K is an integer greater than or equal to 2 and less than or equal to C) fluorescent pigments.

[0012] Alternatively, the pigment data acquisition device of the second aspect of the embodiment takes N (N is an integer greater than 2) objects as objects, processes spectral data as the distribution of fluorescence intensity values ​​at each of C (C is an integer greater than 2) detection wavelengths, clusters the N objects into L (L is an integer greater than 2 and less than N-1) object groups based on the intensity value of each object at each of the C detection wavelengths, generates L clustering matrices in which the intensity value of each of the C detection wavelengths is arranged according to each clustered object group, calculates statistical values ​​of the intensity values ​​of the object group at the C detection wavelengths for each of the L clustering matrices, uses the statistical values ​​at the C detection wavelengths of each clustering matrix in the L clustering matrices, performs demixing with the C detection wavelengths as the object, and generates K pigment data representing the distribution of each of K (K is an integer greater than 2 and less than C) fluorescent pigments.

[0013] Alternatively, a pigment data acquisition program according to a third aspect of the embodiment is used to generate pigment data representing the distribution of fluorescent pigments of N (N is an integer greater than or equal to 2) objects as objects based on the distribution of fluorescence intensity values ​​at each of C (C is an integer greater than or equal to 2) detection wavelengths, i.e., spectral data. The pigment data acquisition program causes a computer to execute the following steps: clustering the N objects into L (L is an integer greater than or equal to 2 and less than or equal to N-1) object groups based on the intensity values ​​of each object at each of the C detection wavelengths, and generating L clustering matrices in which the intensity values ​​of each of the C detection wavelengths are arranged for each clustered object group; calculating, for each of the L clustering matrices, a statistical value of the intensity values ​​of the object group at the C detection wavelengths; and performing demixing with the C detection wavelengths as objects using the statistical values ​​at the C detection wavelengths of each clustering matrix in the L clustering matrices to generate K pigment data representing the distribution of each of K (K is an integer greater than or equal to 2 and less than or equal to C) fluorescent pigments.

[0014] According to the first, second, or third aspects, spectral data of fluorescence distribution at C detection wavelengths is acquired for each of N objects. This spectral data is clustered into L object groups, and L cluster matrices are generated, each arranging the intensity values ​​of the object groups at each of the C detection wavelengths. Statistics of the intensity values ​​for each of the L cluster matrices are calculated. Furthermore, demixing is performed based on the statistics of the L cluster matrices, resulting in K pieces of pigment data representing the distribution of each of the K fluorescent pigments. This allows for demixing to be performed appropriately for the observation conditions, even if variations occur in the quantity of each pigment, resulting in highly accurate fluorescence spectral data.

[0015] The pigment data acquisition method of the fourth aspect of the embodiment comprises: a data acquisition step of acquiring spectral data with N (N is an integer greater than 2) objects as objects, the spectral data being the distribution of the intensity value of fluorescence at each of C (C is an integer greater than 2) detection wavelengths; a clustering step of clustering the C detection wavelengths into M (M is an integer greater than 2 and less than C-1) detection wavelength groups based on the intensity value for each of the N objects, and generating M clustering matrices in which the intensity value of each of the N objects is arranged according to each clustered detection wavelength group; a calculation step of calculating, for each of the M clustering matrices, a statistical value of the intensity value of the detection wavelength group of the N objects; and a data generation step of using the statistical value of the N objects in each clustering matrix in the M clustering matrices to perform demixing with the N objects as objects, and generating K pigment data representing the distribution of each of K (K is an integer greater than 2 and less than M) fluorescent pigments.

[0016] Alternatively, the pigment data acquisition device of the fifth aspect of the embodiment processes spectral data of the distribution of intensity values ​​of fluorescence at each of C detection wavelengths (C is an integer greater than 2) with N objects as objects. The pigment data acquisition device clusters the C detection wavelengths into M detection wavelength groups (M is an integer greater than 2 and less than C-1) based on the intensity value for each of the N objects, generates M clustering matrices in which the intensity value of each of the N objects is arranged according to each clustered detection wavelength group, calculates statistical values ​​of the intensity values ​​of the detection wavelength groups of the N objects for each of the M clustering matrices, uses the statistical values ​​of the N objects in each clustering matrix in the M clustering matrices, performs demixing with the N objects as objects, and generates K pigment data representing the distribution of each of K fluorescent pigments (K is an integer greater than 2 and less than M).

[0017] Alternatively, the pigment data acquisition program of the sixth aspect of the embodiment is used to generate pigment data representing the distribution of fluorescent pigments in an object based on the distribution of fluorescence intensity values ​​at each of C (C is an integer greater than 2) detection wavelengths with N (N is an integer greater than 2) objects as the object, that is, spectral data. The pigment data acquisition program causes the computer to execute: for each of the N objects, based on the intensity value, clustering the C detection wavelengths into M (M is an integer greater than 2 and less than C-1) detection wavelength groups, and generating M clustering matrices in which the intensity value of each of the N objects is arranged according to each clustered detection wavelength group; for each of the M clustering matrices, calculating the statistical value of the intensity value of the detection wavelength group of the N objects; and using the statistical value of the N objects in each clustering matrix in the M clustering matrices, performing demixing with the N objects as the object, and generating K pigment data representing the distribution of each of K (K is an integer greater than 2 and less than M) fluorescent pigments.

[0018] According to the fourth, fifth, or sixth aspects, spectral data of fluorescence distribution at C detection wavelengths is acquired for each of N objects. This spectral data is clustered into M detection wavelength groups, and M cluster matrices are generated, which arrange the intensity values ​​of the detection wavelength groups for each of the N objects. Statistics of the intensity values ​​of the detection wavelength groups are calculated for each of the M cluster matrices. Furthermore, demixing is performed based on the statistics of the M cluster matrices, resulting in K pieces of pigment data representing the distribution of K fluorescent pigments for each object. Thus, even if the type of pigment to be observed or the type of excitation light used for observation changes, demixing can be performed that is appropriate for the observation conditions, resulting in highly accurate fluorescence spectral data, i.e., pigment data.

[0019] Effects of the Invention

[0020] According to one aspect of the embodiment, a fluorescence spectrum can be obtained with high precision when observation conditions vary. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 1 is a schematic configuration diagram of a pigment data acquisition system 1 according to an embodiment.

[0022] Figure 2 yes Figure 1 Schematic diagram of the data acquisition device 3.

[0023] Figure 3 Yes Figure 1 A block diagram of an example of the hardware structure of the data processing device 5.

[0024] Figure 4 Yes Figure 1A block diagram of the functional structure of the data processing device 5.

[0025] Figure 5 It means by Figure 4 The matrix data Y' reconstructed by the statistical value calculation unit 203 and the matrix data Y' reconstructed by Figure 4 FIG. 1 is a diagram of an image of the pigment matrix data X′ derived by the matrix estimation unit 204 .

[0026] Figure 6 This is a flowchart showing the procedure of the pigment data acquisition method according to the embodiment.

[0027] Figure 7 This is a graph for explaining a general method of identifying a population of an analyte in flow cytometry.

[0028] Figure 8 This is a graph showing the distribution of colorant data related to two colorants acquired by omitting the clustering process in the colorant data acquisition system 1 .

[0029] Figure 9 This is a graph showing the distribution of colorant data related to two types of colorants acquired by executing clustering processing in the colorant data acquisition system 1 .

[0030] Figure 10 This is a graph showing the distribution of colorant data related to two colorants acquired by omitting the clustering process in the colorant data acquisition system 1 .

[0031] Figure 11 This is a graph showing the distribution of colorant data related to two types of colorants acquired by executing clustering processing in the colorant data acquisition system 1 .

[0032] Figure 12 This is a graph showing a histogram of the intensity of a certain fluorescence wavelength band and a histogram of the expression level of the pigment B acquired by the pigment data acquisition system 1 according to the embodiment.

[0033] Figure 13 It is a schematic configuration diagram of a data acquisition device 3A according to a modified example. DETAILED DESCRIPTION

[0034] Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. In the following description, the same elements or elements having the same functions will be denoted by the same reference numerals, and repeated description will be omitted.

[0035] Figure 1This is a schematic diagram of a pigment data acquisition system 1 as an embodiment of a pigment data acquisition device. The pigment data acquisition system 1 is a device for generating pigment data, which is used to determine the amount of pigment (fluorescent pigment) contained in a sample such as cells or microparticles as an analysis object. The pigment data generated by the pigment data acquisition system 1 can be used for purposes such as counting, screening, and characteristic analysis of cells, etc. through analysis of the data. Therefore, the pigment data acquisition system 1 is required to generate a large amount of pigment data of the analysis object with high productivity. The pigment data acquisition system 1 includes a data acquisition device 3 that performs flow cytometry-based analysis on the analysis object, and a data processing device 5 that processes the data acquired by the data acquisition device 3. The data acquisition device 3 and the data processing device 5 can be configured to transmit and receive data between them using wired or wireless communication, or can be configured to input and output data via a recording medium. In addition, the data acquisition device 3 and the data processing device 5 can also be configured as an integrated device.

[0036] Figure 2 yes Figure 1 The data acquisition device 3 is a schematic structural diagram of the data acquisition device 3. The data acquisition device 3 is a system for performing flow cytometry, generally called a flow cytometer, and is composed of a fluid system 52, an optical system 53, and an electronic system (signal processing device) 54.

[0037] The fluidic system 52 includes a flow cell 56, into which a sample fluid containing analytes such as cells or microparticles is injected, allowing the analytes contained in the sample fluid to be arranged and passed through a narrow channel 55. The flow cell 56 also has a function (not shown) for sorting (classifying and diverting) selected analytes through electric field control or the like.

[0038] The optical system 53 is a system for optically analyzing an analyte passing through a flow cell 56 using flow cytometry. This optical system 53 includes light sources 7a, 7b, 7c, and 7d, dichroic mirrors 57a, 57b, and 57c, a reflector 57d, a lens 8, filters 9a, 9b, 9c, and 9d, dichroic mirrors 10b and 10c, and photodetectors 11a, 11b, 11c, and 11d. The optical system guides various types of light generated from the analyte by flow cytometry to the photodetectors 11a, 11b, 11c, and 11d. Light sources 7a, 7b, 7c, and 7d each generate light (excitation light) having different central wavelengths (excitation wavelengths). Light sources 7a, 7b, 7c, and 7d are, for example, laser light sources, light-emitting diodes, or superluminescent diodes. The dichroic mirror 57a transmits light emitted from light source 7a to lens 8, while reflecting light emitted from other light sources 7b, 7c, and 7d to lens 8. The dichroic mirror 57b reflects the light emitted from the light source 7b toward the dichroic mirror 57a, and transmits the light emitted from the light sources 7c and 7d toward the dichroic mirror 57a. The dichroic mirror 57c reflects the light emitted from the light source 7c toward the dichroic mirror 57b, and transmits the light emitted from the light source 7d toward the dichroic mirror 57b. The reflector 57d reflects the light emitted from the light source 7d toward the dichroic mirror 57c. The lens 8 focuses the light emitted from the light sources 7a, 7b, 7c, and 7d onto the channel 55 within the flow cell 56. The optical filter 9a transmits forward scattered light (i.e., at least one of the light sources 7a, 7b, 7c, and 7d) generated from the sample fluid by the irradiation of light. The dichroic mirror 10b reflects the fluorescence (first fluorescence) in the first wavelength band (first detection wavelength) of the fluorescence generated from the sample fluid by the irradiation of light emitted from the light sources 7a, 7b, 7c, and 7d, respectively, and transmits the remaining fluorescence generated from the sample fluid. The dichroic mirror 10c reflects the fluorescence (second fluorescence) in the second wavelength band (second detection wavelength) of the fluorescence that has passed through the dichroic mirror 10b, and transmits the fluorescence (third fluorescence) in the remaining wavelength band (third detection wavelength) of the fluorescence that has passed through the dichroic mirror 10b. The optical filter 9b transmits the first fluorescence in the first wavelength band reflected by the dichroic mirror 10b, and the filter 9c transmits the second fluorescence in the second wavelength band reflected by the dichroic mirror 10c. The optical filter 9d transmits the third fluorescence in the third wavelength band of the fluorescence that has passed through the dichroic mirror 10c. The photodetectors 11a, 11b, 11c, and 11d are respectively positioned on the optical axes of the forward scattered light, the first fluorescence, the second fluorescence, and the third fluorescence, and measure the intensities of the forward scattered light, the first fluorescence, the second fluorescence, and the third fluorescence. The photodetectors 11 a , 11 b , 11 c , and 11 d are, for example, photomultiplier tubes, avalanche photodiodes, HPDs (Hybrid PhotoDetectors), or SiPMs (Silicon Photomultipliers).Alternatively, the photodetectors 11a, 11b, 11c, and 11d may be formed from a single photodetector having multiple detection ports. Examples of such single photodetectors include multi-anode photomultiplier tubes. The optical system 53 can measure the intensity of each fluorescence at C (C is an integer greater than or equal to 2) detection wavelengths, depending on the number of combinations of dichroic mirrors, filters, and photodetectors.

[0039] In addition, the optical system 53 can measure the intensity of the forward scattered light, the first fluorescence, the second fluorescence, and the third fluorescence respectively while switching to irradiate light of multiple wavelength bands from the light sources 7a, 7b, 7c, and 7d. Thus, the intensity of the fluorescence of multiple wavelength bands can be measured efficiently. In addition, the optical system 53 can also measure the intensity of the forward scattered light, the first fluorescence, the second fluorescence, and the third fluorescence respectively while continuously irradiating light of multiple wavelength bands from the light sources 7a, 7b, 7c, and 7d. In addition, the optical system 53 can also have a structure capable of observing side scattered light by using the same structure as described above.

[0040] The electronic system 54 is a device for collecting data on the light intensity measured by the optical system 53. Specifically, the electronic system 54 is electrically connected to the multiple photodetectors 11a, 11b, 11c, and 11d. It performs A / D conversion on the intensity signals representing the intensity detected by each channel of the multiple photodetectors 11a, 11b, 11c, and 11d, generates array data containing a one-dimensional array of the intensity values ​​for each analyte after A / D conversion, and transmits this data to the outside. The electronic system 54 can generate array data by directly A / D converting the intensity signals detected by the multiple photodetectors 11a, 11b, 11c, and 11d, or it can modify the intensity signals or the intensity values ​​after A / D conversion using various parameters to generate array data based on the modified intensity signals or modified intensity values.

[0041] Next, refer to Figure 3 and Figure 4 The structure of the data processing device 5 will be described. Figure 3 is a block diagram showing an example of the hardware configuration of the data processing device 5. Figure 4 It is a block diagram showing the functional structure of the data processing device 5.

[0042] like Figure 3As shown, the data processing device 5 is physically a computer or the like, including a CPU (Central Processing Unit) 101 as a processor, a RAM (Random Access Memory) 102 or a ROM (Read Only Memory) 103 as a recording medium, a communication module 104, and an input / output module 106, all of which are electrically connected. Furthermore, the data processing device 5 may include a display, keyboard, mouse, touch panel display, and the like as input / output devices, and may also include a data recording device such as a hard disk drive or semiconductor memory. Furthermore, the data processing device 5 may be composed of multiple computers.

[0043] like Figure 4 As shown, the data processing device 5 includes a data acquisition unit 201 , a clustering unit 202 , a statistic calculation unit 203 , a matrix estimation unit 204 , and a data generation unit 205 as functional components. Figure 4 Each functional unit of the data processing device 5 shown in FIG. 1 is realized by reading a program (pigment data acquisition program of the embodiment) into hardware such as the CPU 101 and the RAM 102, and operating the communication module 104 and the input / output module 106 under the control of the CPU 101, and reading and writing data in the RAM 102. The CPU 101 of the data processing device 5 executes the computer program. Figure 4 Each functional part of the data processing device 5 is used to perform a function and execute the processing corresponding to the pigment data acquisition method described later. In addition, CPU 101 can be independent hardware or can be installed in a programmable logic device such as FPGA as a soft processor. RAM and ROM can also be independent hardware or can be built into a programmable logic device such as FPGA. The various data required for executing the computer program and the various data generated by executing the computer program are all stored in storage media such as built-in memories such as ROM 103 and RAM 102 or hard disk drives. Below, the functions of the functional components of the data processing device 5 are described in detail.

[0044] The data acquisition unit 201 acquires, from the data acquisition device 3, array data related to fluorescence in C pre-specified wavelength bands (C is an integer greater than or equal to 2) for N (N is an integer greater than or equal to 2) analyte objects. These C array data are spectral data obtained by measuring the distribution of fluorescence intensity values ​​in each of the C wavelength bands from the N analyte objects while switching between irradiating the analyte with light in multiple wavelength bands in the data acquisition device 3. The number C of array data acquired (the number C of observed fluorescence wavelength bands) is pre-specified to be greater than the maximum number of pigments that may be contained in the analyte objects.

[0045] Clustering unit 202 reads the C array data acquired by data acquisition unit 201 and performs clustering on the N analytes constituting the C array data based on the intensity values ​​of the analytes in each of the C wavelength bands. Prior to clustering, clustering unit 202 generates matrix data Y, which is a one-dimensional arrangement of the intensity values ​​of the N analytes constituting the C array data.

[0046] Then, based on the distribution information of the intensity values ​​of each fluorescence wavelength band, the clustering unit 202 clusters the N analysis objects into L object groups (L is an integer greater than 2 and less than N-1). For example, the clustering unit 202 selects a wavelength band where intensity values ​​are likely to differ between pigments, creates a histogram of the intensity values ​​in this wavelength band, and sets a threshold value between two peaks based on this histogram. Furthermore, by comparing the set threshold value with the intensity values ​​of the arrangement data, the clustering unit 202 can cluster the analysis objects into two object groups. The clustering unit 202 can also classify the analysis objects into three or more object groups using the same function. The number of object groups L that the clustering unit 202 clusters is pre-set as a parameter stored in the data processing device 5, corresponding to the number of possible pigment types present in the analysis objects. The clustering unit 202 then reconstructs the matrix data Y, which is the matrix data Y containing the intensity values ​​of the analysis objects in the arrangement data arranged one-dimensionally in parallel, by dividing it into cluster matrices for each of the L object groups. The number L of object groups clustered by the clustering unit 202 may be set in advance based on the type of detection wavelength bands and the number of detection wavelength bands (C). The number L of object groups clustered by the clustering unit 202 may also be set independently of these.

[0047] The statistic calculation unit 203 and the matrix estimation unit 204 calculate a mixing matrix A for generating K pigment data representing the distribution of each of K pigments (K is an integer greater than or equal to 2 and less than or equal to C) from the C arrangement data, based on the L cluster matrices obtained for the analysis object. Generally, the relationship between matrix data Y, which is the observation matrix, and pigment matrix data X, which is the K pigment data arranged one-dimensionally for each analysis object, is expressed using the mixing matrix A as shown in the following equation, based on the non-negative matrix factorization (NMF) operation method.

[0048] Y=AX

[0049] Here, Y is the matrix data of C rows and N columns, A is the matrix data of C rows and K columns, and X is the matrix data of K rows and N columns. Conversely, if the value of the mixing matrix A is obtained, the color matrix data X uses the inverse matrix A of the mixing matrix A. -1 and matrix data Y are derived from the following equation (this process is called unmixing).

[0050] X=A-1 Y

[0051] Here, the statistical value calculation unit 203 compresses the matrix data Y generated by the clustering unit 202 in units of object groups clustered by the clustering unit 202, thereby reconstructing the matrix data Y'. Specifically, the statistical value calculation unit 203 calculates a statistical value for each object group in the clustered matrix based on the intensity values ​​of each row of the matrix data Y, and compresses the object group in each row into a single object having the calculated statistical value. Thus, the statistical value calculation unit 203 once again generates matrix data Y' having C rows and L columns. As the statistical value, the statistical value calculation unit 203 can calculate the average value based on the cumulative value of the intensity value, the mode (most frequent value) of the intensity value, or the median value of the intensity value.

[0052] The matrix estimation unit 204 utilizes the property that the following equation also holds true for the matrix data Y′ reconstructed by the statistical value calculation unit 203 and the color matrix data X′ compressed similarly from the color matrix data X, including the mixing matrix A.

[0053] Y'=AX'

[0054] The mixing matrix A is derived based on the matrix data Y'. Figure 5 An image of the matrix data Y′ reconstructed by the statistical value calculation unit 203 and the corresponding pigment matrix data X′ is shown. Figure 5 Each square shown in FIG represents one element of the matrix data. For example, it is divided into three object groups PGr 01 ~PGr 03 The pigment matrix data X and matrix data Y are set to each object group PGr 01 ~PGr 03 The statistical values ​​are representative values ​​and are compressed into three columns of pigment matrix data X' and matrix data Y'.

[0055] The matrix estimation unit 204 derives the mixing matrix A based on the matrix data Y' as follows. Specifically, the matrix estimation unit 204 sets an initial value for the mixing matrix A, calculates the loss function (loss value) Los described below while sequentially changing the value of the mixing matrix A, and derives the mixing matrix A that reduces the value of the loss function Los. Furthermore, a regularization term such as the L1 norm λ|A| (λ is a coefficient indicating the degree of emphasis on the regularization term) may be added to this loss function.

[0056] [Formula 1]

[0057]

[0058] In the above formula, j is a parameter indicating the position of the matrix data row (corresponding to the fluorescence wavelength band). The matrix subscript 1j represents the matrix data in the jth row of the first cluster matrix, the matrix subscript 2j represents the matrix data in the jth row of the second cluster matrix, and the matrix subscript 3j represents the matrix data in the jth row of the third cluster matrix. Furthermore, the parameters a, b, and c represent the average of the statistical values ​​in each column of the matrix data Y'.

[0059] As described above, the matrix estimation unit 204 calculates a loss function for each of the L cluster matrices divided by the clustering unit 202, referring to the statistical values ​​of the C matrix data Y'. The loss function Los is calculated based on the sum of the L loss functions, and the mixing matrix A is obtained based on this loss function Los. In this case, the matrix estimation unit 204 corrects the loss function calculated for each of the L cluster matrices by dividing it by the average values ​​a, b, and c of the statistical values ​​of the C matrix data Y'. The loss function Los is then calculated by summing the corrected loss functions. Alternatively, the matrix estimation unit 204 may correct the loss function for each of the L cluster matrices by dividing the row component of each wavelength band of the fluorescence of the difference value Y'-AX' by the C statistical values ​​corresponding to each wavelength band of the fluorescence.

[0060] Furthermore, the above formula can also be generalized as follows. Specifically, the matrix estimation unit 204 derives the mixing matrix A and the pigment matrix data X' based on the matrix data Y' as follows. Specifically, the matrix estimation unit 204 sets initial values ​​for the mixing matrix A and the pigment matrix data X', and while sequentially changing the values ​​of the mixing matrix A and the pigment matrix data X', calculates the loss function (loss value) Los using the following formula, thereby deriving the mixing matrix A and the pigment matrix data X' that reduce the value of the loss function Los. Furthermore, a regularization term such as the L1 norm λ|A| (λ is a coefficient indicating the degree of emphasis on the regularization term) may be added to this loss function. Furthermore, the calculation may be performed while imposing constraints such that the mixing matrix A and the pigment matrix data X' are non-negative values.

[0061] [Formula 2]

[0062]

[0063] In the above formula, j is a parameter indicating the position of the row of the matrix data (corresponding to the wavelength band of the fluorescence), and i is a parameter indicating the position of the column of the matrix data (corresponding to the i-th cluster). ij The weight of each element of the matrix data can also be calculated based on the value of each element or its standard deviation. ij All are set to the same value regardless of the weight of each element. In addition, the average value of the statistical value of each column of the matrix data Y' in the above formula is set to a, b, c, ..., and replaced by w1j =1 / a, w 2j =1 / b, w 3j =1 / c is the same as the formula of the loss function Los shown above.

[0064] As described above, the matrix estimation unit 204 refers to the statistical values ​​of the C matrix data Y' and calculates the loss function for each of the L cluster matrices divided by the clustering unit 202. i The loss function Los is calculated by summing the values ​​of the difference values ​​Y'-AX' and the mixing matrix A is obtained based on the loss function Los. In addition, the matrix estimation unit 204 may also correct the loss function Los of each cluster matrix of the L cluster matrices by dividing the row component of each wavelength band of the fluorescence of the difference value Y'-AX' by the C statistical values ​​corresponding to each wavelength band of the fluorescence. i , and thus perform calculations.

[0065] The data generation unit 205 uses the mixing matrix A derived by the matrix estimation unit 204 to unmix the C permutation data obtained for the analysis object, thereby obtaining K pigment data. Specifically, the data generation unit 205 unmixes the inverse matrix A of the mixing matrix A. -1 The data is applied to the matrix data Y generated by the clustering unit 202 based on the C permutation data to calculate the pigment matrix data X. The data generation unit 205 then reconstructs K pigment data from the pigment matrix data X and outputs the reconstructed K pigment data. The output destination can be an output device of the data processing device 5, such as a display or touch panel display, or an external device connected to the data processing device 5 for data communication. Furthermore, the pigment data generated by the data generation unit 205 can also be used for data analysis purposes such as counting, screening, and analyzing characteristics of analysis objects.

[0066] Next, the procedure of observation processing for an analysis object using the pigment data acquisition system 1 of the present embodiment, that is, the flow of the pigment data acquisition method of the present embodiment will be described. Figure 6 1 is a flowchart showing the procedure of observation processing by the pigment data acquisition system 1 .

[0067] First, the data acquisition device 3 generates and transmits the results of observation processing for multiple analytes (step S1). Next, the data acquisition unit 201 of the data processing device 5 acquires the array data for C fluorescence wavelength bands from the data acquisition device 3 (step S2; data acquisition step).

[0068] Furthermore, the clustering unit 202 of the data processing device 5 performs clustering on the C array data, clustering the N analysis objects in the array data into L object groups (step S3; clustering step). Next, the statistical value calculation unit 203 of the data processing device 5 calculates the statistical values ​​of the L object groups, thereby regenerating matrix data Y' based on the matrix data Y generated by the clustering unit 202 (step S4; calculation step).

[0069] Next, the matrix estimation unit 204 of the data processing device 5 derives a mixing matrix A based on the matrix data Y' (step S5; data generation step). Finally, the data generation unit 205 of the data processing device 5 uses the mixing matrix A to unmix the matrix data Y generated based on the C permutation data of the analysis object, thereby obtaining and outputting K pigment data (step S6; data generation step). Through the above processing, the observation process for multiple analysis objects is completed.

[0070] According to the pigment data acquisition system 1 described above, spectral data representing the distribution of fluorescence in C wavelength bands is acquired for each of N analysis objects. This spectral data is clustered into L object groups, and L cluster matrices are generated, which arrange the intensity values ​​of the object groups in each of the C fluorescence wavelength bands. Statistics of the intensity values ​​for each of the L cluster matrices are calculated. Furthermore, demixing is performed based on the statistics of the L cluster matrices, resulting in K pieces of pigment data representing the distribution of each of the K fluorescent pigments. This allows for demixing to be performed appropriately for the observation conditions, even if the amount of each pigment varies, resulting in highly accurate fluorescence spectral data.

[0071] Flow cytometry is a method that can identify or quantify cell populations by staining cells, cell surface proteins, and intracellular proteins. It can be used for applications such as cell sorting and immunophenotyping. According to this embodiment, based on the acquired staining data, for example, by analyzing the combination of antibodies expressed on blood cells, it is possible to accurately identify which cells are tumorous.

[0072] In this embodiment, by irradiating the object with light of multiple wavelength bands, the distribution of intensity values ​​in each of C fluorescence wavelength bands, i.e., spectral data, is obtained. Thus, by using light of multiple wavelength bands, fluorescence from multiple fluorescent pigments can be efficiently observed. As a result, highly accurate pigment data can be obtained when multiple fluorescent pigments are used as observation targets.

[0073] Furthermore, in this embodiment, N analyte objects are clustered based on the distribution information of intensity values ​​in each of C fluorescence wavelength bands. Using this distribution information in this manner enables clustering based on similarities in fluorescence wavelength distribution. As a result, unmixing tailored to observation conditions can be performed, enabling accurate pigment data to be obtained.

[0074] Furthermore, in this embodiment, the statistical value of the arrangement data is calculated based on the cumulative value, mode, or median of the intensity values ​​of the object groups. In this case, unmixing can be performed based on the overall trend of the intensity values ​​of the L object groups in the clustering matrix. Unmixing that is appropriate for the observation conditions can be performed, resulting in highly accurate pigment data.

[0075] Furthermore, in this embodiment, the statistical values ​​of the C fluorescence wavelength bands of each of the L cluster matrices are used to determine the mixing matrix A, and demixing is performed using the mixing matrix A. In this case, even if the number of analysis objects, the types of excitation light used for observation, or the number of observed fluorescence bands increases, the amount of calculation required for demixing can be reduced. Furthermore, by performing demixing using the statistical values ​​of the intensity values ​​of each clustered cluster matrix, the accuracy of pigment data separation can be improved. In the prior art, the mixing matrix A is derived by calculation based on reference information (fluorescence spectrum, absorption spectrum, etc.) of each pigment. In this embodiment, even when such reference information is unknown, the mixing matrix can be estimated based on the matrix data Y obtained from the arrangement data. As a result, the accuracy of the separated data can be improved while achieving increased productivity when acquiring pigment data.

[0076] Furthermore, in this embodiment, non-negative matrix factorization is used to calculate the loss function Los based on the statistics of C fluorescence wavelength bands for each of L cluster matrices, and the mixing matrix A is obtained based on the sum of the loss functions Los. In this case, the loss function Los is calculated based on the statistics of the L object groups in the clustered array data, and the mixing matrix A is obtained based on the sum of these loss functions Los. This mixing matrix A is then used to perform unmixing. This further improves the accuracy of the generated pigment data, even when the amounts of pigments contained in multiple analytes vary. Specifically, when the loss function Los is calculated based on array data without clustering, the separation accuracy of relatively large-volume pigments is prioritized, resulting in a decrease in the separation accuracy of relatively small-volume pigments. However, in this embodiment, the separation accuracy of multiple pigments can be improved simultaneously. Furthermore, the productivity of pigment data acquisition can be improved.

[0077] Furthermore, in this embodiment, the loss function Los of each of the L clustering matrices is corrected using coefficients a, b, and c based on statistical values, and the mixing matrix A is calculated based on the sum of the corrected loss functions Los. Thus, the loss function Los is calculated based on the statistical values ​​of the L object groups in the clustered array data, and each loss function Los is corrected based on the statistical values ​​when the sum of the loss functions Los is calculated. This allows for highly accurate generation of pigment data even when the fluorescence intensities of the individual pigments in the array data of the analysis objects vary. Specifically, when calculating the loss function Los without correcting the L loss functions Los, the separation accuracy of pigments with relatively strong intensities is prioritized, resulting in a decrease in the separation accuracy of pigments with relatively weak intensities. However, in this embodiment, the separation accuracy of multiple pigments can be uniformly improved.

[0078] Figure 7 This diagram illustrates a common method for identifying populations of analytes in flow cytometry. In typical flow cytometry, a scatter plot DP is generated based on the fluorescence intensities of several cells in the "Detection Wavelength 1" and "Detection Wavelength 2" wavelength bands. Based on this scatter plot DP, a histogram HG1 of the "Detection Wavelength 1" intensity and a histogram HG2 of the "Detection Wavelength 2" intensity are created. Then, gating is performed based on a threshold value determined from the distributions of HG1 and HG2, or based on the scatter plot DP, to identify the "Pigment A" and "Pigment B" populations.

[0079] However, fluorescence from a single pigment contains multiple fluorescence wavelength bands, and the distribution of the wavelength bands contained in the fluorescence varies complexly between different pigments. Furthermore, a single analyte may contain multiple pigments. In the aforementioned conventional population identification method, when the distributions of histograms HG1 and HG2 overlap between pigments, it is difficult to accurately separate the two pigments using a threshold value, and the pigment cannot be accurately identified. Furthermore, even when gating the scatter plot DP, it is difficult to accurately separate the two pigments, making it impossible to accurately identify the pigment. Furthermore, it is also difficult to identify multiple pigments contained in a single analyte. In contrast, according to this embodiment, by performing an unmixing process on the arrangement data of the C fluorescence wavelength bands, it is possible to quantitatively obtain the expression levels of each pigment with high precision.

[0080] Figures 8 to 11 The distribution of pigment data on two pigments acquired in the pigment data acquisition system 1 is shown. Figures 8 to 11 In the figure, part (a) is a graph showing a scatter plot of intensity values ​​of two fluorescence wavelength bands, and part (b) is a graph showing the expression levels (distribution) of two pigments, "pigment A" and "pigment B." Figure 8This shows the result of obtaining pigment data by omitting clustering processing in the data processing device 5 when there is a difference in the number of each pigment (when the number of cells of pigment A is 1 / 100 of the number of cells of pigment B). Figure 9 The results of pigment data obtained by clustering processing in the data processing device 5 when there is a difference in the number of each pigment (the number of cells of pigment A is 1 / 100 of the number of cells of pigment B) are shown. Figure 10 This shows the result of obtaining pigment data by omitting clustering processing in the data processing device 5 when there is a difference in the intensity of each pigment (when the intensity of the cells containing pigment A is 1 / 10 of the intensity of the cells containing pigment B). Figure 11 The diagram shows the result of obtaining pigment data by clustering processing performed by the data processing device 5 when there is a difference in the intensity of each pigment (when the intensity of cells containing pigment A is 1 / 10 of the intensity of cells containing pigment B).

[0081] like Figure 8 As shown in FIG, when clustering is omitted, there are single-color cells of pigment A within the range indicated by the dotted line, but the expression levels of the two pigments are not separately determined. Figure 9 As shown, when clustering processing is performed, the expression level of the single-color cells of pigment A is calculated within the range indicated by the dotted line, and the expression levels of the two pigments can be separated with high accuracy.

[0082] like Figure 10 As shown in , when clustering is omitted, the expression levels of the two pigments are not separated. Figure 11 As shown, when the clustering process is performed, the expression levels of the two pigments are separated and the expression levels of the two pigments are calculated with high accuracy.

[0083] exist Figure 12 In the figure, part (a) shows histograms of the intensity of a certain fluorescence wavelength band obtained by the pigment data acquisition system 1 for cells expressing pigment A and cells expressing pigment B, respectively. Part (b) shows a histogram of the expression level of pigment B obtained by the pigment data acquisition system 1. As such, the histograms of the intensity of the fluorescence wavelength bands overlap significantly between the two pigments, making it difficult to identify the pigments based on these histograms. In contrast, the pigment data obtained using this embodiment allows for the high-precision separation of the expression level of pigment B.

[0084] Although various embodiments of the present invention have been described above, the present invention is not limited to the above-described embodiments, and may be modified or applied to other embodiments within a scope that does not change the gist of the present invention.

[0085] As a clustering method of the object group in the data processing device 5 , a method using machine learning such as the K-means method, a method using deep learning, or the like may be adopted.

[0086] In addition, as a method for clustering the object group in the data processing device 5, in addition to the K-means method, a decision tree, a support vector machine, KNN (K nearest neighbor), a self-organizing map, a spectral clustering, a Gaussian mixture model, DBSCAN (density-based spatial clustering with noise), Affinity Propagation, MeanShift, Ward (hierarchical clustering), Agglomerative Clustering, OPTICS (point sorting to identify cluster structures), BIRCH (balanced iterative reduction clustering based on hierarchical structures), or a method using deep learning can also be used. In addition, the matrix data Y can also be preprocessed before clustering. For example, the C-dimensional data of each analysis object can also be reduced in dimensionality by using Phasor Analysis, principal component analysis, singular value decomposition, independent component analysis, linear discriminant analysis, t-SNE (t-distributed stochastic neighbor embedding), UMAP (uniform manifold approximation and projection), or other machine learning methods.

[0087] The clustering unit 202 of the data processing device 5 can also cluster the matrix data Y into M wavelength groups (M is an integer greater than 2 and less than C-1) (detection wavelength groups) based on the intensity values ​​of each wavelength band, thereby generating M cluster matrices that arrange the intensity values ​​of each detection object according to each wavelength group. The generation of these M cluster matrices can be performed before or after the generation of the L cluster matrices described above. In this case, the statistical value calculation unit 203 compresses the matrix data Y using the statistical values ​​of the clustered wavelength groups using the same method as described above, thereby reconstructing the matrix data Y'. In this way, by also clustering between the fluorescence wavelength bands, the S / N ratio of the pigment data can be improved. In addition, the amount of calculation required for demixing can be further reduced. In addition, by using the statistical values ​​of the intensity values ​​of each clustered wavelength group for demixing, the accuracy of the pigment data separation can be maintained.

[0088] Alternatively, the clustering unit 202 of the data processing device 5 may perform clustering only on the M wavelength groups, rather than clustering the L object groups in the above-described embodiment, and may generate only M clustering matrices. In this case, the data generation unit 205 of the data processing device 5 performs unmixing using the statistical values ​​of the N analysis objects in each of the M clustering matrices to generate K (K is an integer greater than or equal to 2 and less than or equal to M) pieces of pigment data.

[0089] The data acquisition device 3 in the above embodiment may also be changed to Figure 13 The structure shown. Figure 13 The data acquisition device 3A of the modified example shown is different from the data acquisition device 3A in that it is equipped with a spectrometer 58 and a detector 59 as the optical system 53. The spectrometer 58 splits the fluorescence generated from the sample fluid into multiple wavelength bands (for example, wavelength bands of 1024 channels). The detector 59 is a one-dimensional detector or a two-dimensional detector having multiple pixels (for example, 1024 pixels), which measures the intensity of the fluorescence of each wavelength band incident on each pixel from the spectrometer 58 and outputs it to the electronic system 54. According to such a data acquisition device 3A, it is possible to generate arrangement data of multiple fluorescence wavelength bands. For example, the arrangement data of 1024 channels generated by the data acquisition device 3A is divided into 512 cluster matrices by clustering two adjacent channels into one wavelength group by the data processing device 5.

[0090] In the first aspect, preferably, in the data acquisition step, spectral data representing the distribution of intensity values ​​at each of C detection wavelengths is acquired by irradiating the object with excitation light having multiple excitation wavelengths. Furthermore, in the second aspect, preferably, the apparatus further comprises data acquisition means for acquiring spectral data representing the distribution of intensity values ​​at each of C detection wavelengths by irradiating the object with excitation light having multiple excitation wavelengths. Thus, by using excitation light having multiple excitation wavelengths, fluorescence from multiple fluorescent pigments can be efficiently observed. As a result, highly accurate pigment data can be obtained when multiple fluorescent pigments are observed.

[0091] Furthermore, in the first aspect, preferably, the clustering step includes clustering the N objects based on distribution information of intensity values ​​at each of the C detection wavelengths. Using this distribution information allows clustering based on similarities in wavelength distribution. As a result, unmixing can be performed that is appropriate for observation conditions, and highly accurate pigment data can be obtained.

[0092] In the first aspect, preferably, the statistical value is calculated based on the cumulative value, mode, or median of the intensity values ​​of the object group in the calculation step. In this case, unmixing suitable for observation conditions can be performed to obtain high-quality pigment data.

[0093] Furthermore, in the first aspect, preferably, in the data generation step, a mixing matrix is ​​obtained using the statistical values ​​of the C detection wavelengths for each of the L cluster matrices, and demixing is performed using the mixing matrix. In this case, even if the number of objects, the types of excitation light used for observation, or the number of fluorescent bands to be observed increases, the computational complexity of demixing can be reduced. Furthermore, by performing demixing using the statistical values ​​of the intensity values ​​of each clustered cluster matrix, the accuracy of pigment data separation can be improved. As a result, the accuracy of the separated data can be improved while also achieving increased productivity when acquiring pigment data.

[0094] Furthermore, in the first aspect, it is preferred that, in the data generation step, non-negative matrix factorization is used to calculate loss values ​​based on the statistical values ​​of the C detection wavelengths for each of the L cluster matrices, and a mixing matrix is ​​obtained based on the sum of the loss values. In this case, the accuracy of the separated data can be improved, and the productivity of obtaining pigment data can be increased.

[0095] Furthermore, in the first aspect, preferably, in the data generation step, the loss value of each of the L cluster matrices is corrected based on the statistical value, and a mixing matrix is ​​obtained based on the sum of the corrected loss values. This improves the accuracy of pigment data separation even when there is variation in the intensity of each fluorescent pigment.

[0096] Furthermore, in the first aspect, preferably, in the clustering step, for each of the N objects, C detection wavelengths are clustered into M detection wavelength groups (M being an integer greater than 2 and less than C-1) based on the intensity value, and M clustering matrices are further generated by clustering the intensity values ​​of each of the N objects for each detection wavelength group. This further reduces the computational complexity involved in performing unmixing. Furthermore, by performing unmixing using the statistical values ​​of the intensity values ​​of each clustered detection wavelength group, the accuracy of the separation of the pigment data can be maintained.

[0097] The pigment data acquisition method of the embodiment is [1] "A pigment data acquisition method, comprising: a data acquisition step of acquiring spectral data of a distribution of fluorescence intensity values ​​at each of C detection wavelengths (C is an integer of 2 or more) with N objects as objects; a clustering step of clustering the N objects into L object groups (L is an integer of 2 or more and N-1 or less) based on the intensity values ​​of each object at each of the C detection wavelengths, and generating L cluster matrices in which the intensity values ​​of each of the C detection wavelengths are arranged according to each clustered object group; a calculation step of calculating statistical values ​​of the intensity values ​​of the object groups at the C detection wavelengths according to the L cluster matrices; and a data generation step of using the statistical values ​​of the C detection wavelengths of each cluster matrix of the L cluster matrices, performing demixing with the C detection wavelengths as objects, and generating the K pigment data representing the distribution of each of K fluorescent pigments (K is an integer of 2 or more and C or less).

[0098] The pigment data acquisition method of the embodiment may also be [2] "In the pigment data acquisition method described in [1] above, in the data acquisition step, the distribution of the intensity values ​​of each of the C detection wavelengths, i.e., the spectral data, is obtained by irradiating the object with each of the excitation light of multiple excitation wavelengths."

[0099] The pigment data acquisition method of the embodiment may also be [3] "In the pigment data acquisition method described in [2] above, in the clustering step, the N objects are clustered based on the distribution information of each of the C detection wavelengths of the intensity value."

[0100] The pigment data acquisition method of the embodiment may also be [4] "The pigment data acquisition method described in any one of [1] to [3] above, wherein in the calculation step, the statistical value is calculated based on the cumulative value, mode or median of the intensity values ​​of the object group."

[0101] The pigment data acquisition method of the embodiment may also be [5] "In the pigment data acquisition method described in any one of [1] to [4] above, in the data generation step, the statistical values ​​of the C detection wavelengths of each of the L clustering matrices are used to obtain a mixing matrix, and the mixing matrix is ​​used for demixing."

[0102] The pigment data acquisition method of the embodiment may also be [6] "In the pigment data acquisition method described in [5] above, in the data generation step, non-negative matrix decomposition is used to calculate the loss value based on the statistical value of the C detection wavelengths of each clustering matrix of L clustering matrices, and the mixing matrix is ​​obtained based on the sum of the loss values."

[0103] The pigment data acquisition method of the embodiment may also be [7] "the pigment data acquisition method as described in [6] above, in the data generation step, the loss value of each clustering matrix of L clustering matrices is corrected based on the statistical value, and the mixing matrix is ​​obtained based on the sum of the corrected loss values."

[0104] The pigment data acquisition method of the embodiment may also be [8] "In the pigment data acquisition method described in any one of [1] to [7] above, in the clustering step, for each of the N objects, based on the intensity value, the C detection wavelengths are clustered into M (M is an integer greater than 2 and less than C-1) detection wavelength groups, and further M clustering matrices are generated in which the intensity values ​​of each of the N objects are arranged according to the clustered detection wavelength groups."

[0105] Description of Reference Numerals

[0106] 1... pigment data acquisition system, 3, 3A... data acquisition device, 5... data processing device, 201... data acquisition unit, 202... clustering unit, 203... statistical value calculation unit, 204... matrix estimation unit, 205... data generation unit, A... mixing matrix.

Claims

1. A method for obtaining pigment data, wherein: have: a data acquisition step of acquiring spectral data for N objects, the spectral data being a distribution of fluorescence intensity values ​​at each of C detection wavelengths, wherein N is an integer greater than or equal to 2 and C is an integer greater than or equal to 2; a clustering step of clustering the N objects into L object groups based on the intensity value of each object at each of the C detection wavelengths, and generating L cluster matrices in which the intensity value at each of the C detection wavelengths is arranged for each of the clustered object groups, where L is an integer greater than or equal to 2 and less than or equal to N-1; a calculation step of calculating a statistical value of the intensity values ​​of the object group at C detection wavelengths for each of the L cluster matrices; and The data generation step uses the statistical values ​​at the C detection wavelengths of each of the L clustering matrices to perform unmixing with the C detection wavelengths as the object, and generates the K pigment data representing the distribution of each of the K fluorescent pigments, wherein K is an integer greater than 2 and less than C.

2. The pigment data acquisition method according to claim 1, wherein: In the data acquisition step, The spectral data, which is the distribution of the intensity values ​​at each of the C detection wavelengths, is acquired by irradiating the object with each of the excitation lights having a plurality of excitation wavelengths.

3. The pigment data acquisition method according to claim 2, wherein: In the clustering step, the N objects are clustered based on distribution information of the intensity values ​​at each of the C detection wavelengths.

4. The pigment data acquisition method according to any one of claims 1 to 3, wherein: In the calculation step, the statistical value is calculated based on a cumulative value, a mode, or a median value of the intensity values ​​of the object group.

5. The pigment data acquisition method according to any one of claims 1 to 4, wherein In the data generating step, a mixing matrix is ​​obtained using the statistical values ​​of the C detection wavelengths in each of the L clustering matrices, and demixing is performed using the mixing matrix.

6. The pigment data acquisition method according to claim 5, wherein: In the data generation step, non-negative matrix factorization is used to calculate loss values ​​based on the statistical values ​​of the C detection wavelengths in each of L clustering matrices, and the mixing matrix is ​​obtained based on the sum of the loss values.

7. The pigment data acquisition method according to claim 6, wherein: In the data generation step, the loss value is corrected based on the statistical value for each of the L clustering matrices, and the mixing matrix is ​​obtained based on the sum of the corrected loss values.

8. The pigment data acquisition method according to any one of claims 1 to 7, wherein: In the clustering step, According to each of the N objects, the C detection wavelengths are clustered into M detection wavelength groups based on the intensity value, and then M clustering matrices are generated in which the intensity value of each of the N objects is arranged according to each of the clustered detection wavelength groups, where M is an integer greater than 2 and less than C-1.

9. A pigment data acquisition device, wherein: Processing spectrum data as a distribution of fluorescence intensity values ​​at each of C detection wavelengths for N objects, where N is an integer greater than or equal to 2 and C is an integer greater than or equal to 2, Based on the intensity value of each object at each of the C detection wavelengths, the N objects are clustered into L object groups, and L cluster matrices are generated in which the intensity value at each of the C detection wavelengths is arranged for each of the clustered object groups, where L is an integer greater than or equal to 2 and less than or equal to N-1. Calculating a statistical value of the intensity values ​​of the object group at C detection wavelengths for each of the L cluster matrices, Using the statistical values ​​at the C detection wavelengths of each of the L clustering matrices, unmixing is performed with the C detection wavelengths as the object to generate the K pigment data representing the distribution of each of the K fluorescent pigments, where K is an integer greater than 2 and less than C.

10. The pigment data acquisition device according to claim 9, wherein: The apparatus further includes a data acquisition device for acquiring the spectrum data as the distribution of the intensity values ​​at each of the C detection wavelengths by irradiating the object with excitation light of each of a plurality of excitation wavelengths.

11. A pigment data acquisition program, wherein: for generating pigment data representing the distribution of fluorescent pigments of N objects based on spectral data representing the distribution of fluorescence intensity values ​​at each of C detection wavelengths for the objects, wherein N is an integer greater than or equal to 2 and C is an integer greater than or equal to 2, The pigment data acquisition program causes the computer to execute: a step of clustering the N objects into L object groups based on the intensity value of each object at each of the C detection wavelengths, and generating L clustering matrices in which the intensity value at each of the C detection wavelengths is arranged for each of the clustered object groups, wherein L is an integer greater than or equal to 2 and less than or equal to N-1; a step of calculating a statistical value of the intensity values ​​of the group of objects at C detection wavelengths for each of L cluster matrices; and Using the statistical values ​​at the C detection wavelengths of each of the L clustering matrices, unmixing is performed with the C detection wavelengths as the object to generate the K pigment data representing the distribution of each of the K fluorescent pigments, wherein K is an integer greater than 2 and less than C.

12. A method for obtaining pigment data, wherein: have: a data acquisition step of acquiring spectral data for N objects, the spectral data being a distribution of fluorescence intensity values ​​at each of C detection wavelengths, wherein N is an integer greater than or equal to 2 and C is an integer greater than or equal to 2; a clustering step of clustering the C detection wavelengths into M detection wavelength groups based on the intensity value for each of the N objects, and generating M cluster matrices in which the intensity value of each of the N objects is arranged according to each of the clustered detection wavelength groups, wherein M is an integer greater than or equal to 2 and less than or equal to C-1; a calculation step of calculating a statistical value of the intensity values ​​of the detection wavelength groups of the N objects for each of the M cluster matrices; and The data generation step uses the statistical values ​​of the N objects in each of the M clustering matrices to perform demixing with the N objects as objects to generate the K pigment data representing the distribution of each of the K fluorescent pigments, wherein K is an integer greater than 2 and less than M.

13. A pigment data acquisition device, wherein: Processing spectral data representing the distribution of fluorescence intensity values ​​at each of C detection wavelengths for N objects, where N is an integer greater than or equal to 2 and C is an integer greater than or equal to 2. For each of the N objects, the C detection wavelengths are clustered into M detection wavelength groups based on the intensity value, and M cluster matrices are generated in which the intensity value of each of the N objects is arranged according to each of the clustered detection wavelength groups, where M is an integer greater than or equal to 2 and less than or equal to C-1. calculating a statistical value of the intensity values ​​of the detection wavelength groups of the N objects for each of the M cluster matrices; Using the statistical values ​​of the N objects in each of the M clustering matrices, unmixing is performed with the N objects as objects to generate the K pigment data representing the distribution of each of the K fluorescent pigments, where K is an integer greater than 2 and less than M.

14. A pigment data acquisition program, wherein: for generating pigment data representing the distribution of fluorescent pigments of N objects based on spectral data representing the distribution of fluorescence intensity values ​​at each of C detection wavelengths for the objects, wherein N is an integer greater than or equal to 2 and C is an integer greater than or equal to 2, The pigment data acquisition program causes the computer to execute: a step of clustering the C detection wavelengths into M detection wavelength groups based on the intensity values ​​for each of the N objects, and generating M clustering matrices in which the intensity values ​​of each of the N objects are arranged for each of the clustered detection wavelength groups, wherein M is an integer greater than or equal to 2 and less than or equal to C-1; A step of calculating a statistical value of the intensity values ​​of the detection wavelength groups of the N objects for each of the M cluster matrices; and Using the statistical values ​​of the N objects in each of the M clustering matrices, demixing is performed on the N objects to generate the K pigment data representing the distribution of each of the K fluorescent pigments, wherein K is an integer greater than 2 and less than M.

Images