Method for discriminating type of hemoglobin, method for generating learned model, learned model, hemoglobin analysis system, and program

Abstract

The invention relates to a method for discriminating the type of hemoglobin, a method for generating a learned model, a learned model, a hemoglobin analysis system, and a program. In the method for discriminating the type of hemoglobin, a learning completion model is prepared, which is generated by using a data set including learning frequency analysis data and a known type of hemoglobin as a learning data set for machine learning. The frequency analysis data for learning is generated by frequency analysis of separation data generated by separation processing of a blood sample for learning having a known type of hemoglobin; a frequency analysis step for preparing frequency analysis data for examination by performing frequency analysis on separation data generated by a separation process for a blood sample for examination for which the type of hemoglobin is unknown; the type of hemoglobin contained in the test blood sample is discriminated by inputting the test frequency analysis data into the learned model.

Description

Technical Field

[0001] The present disclosure relates to a method for distinguishing a type of hemoglobin, a method for generating a learned model, a learned model, a hemoglobin analysis system, and a program. Background Art

[0002] Conventionally, methods have been proposed for measuring glycated hemoglobin (e.g., HbA1a, HbA1b, HbA1c, etc.) and variant hemoglobin (e.g., HbF, HbS, etc.) in a measurement sample such as blood by liquid chromatography (e.g., see Japanese Patent No. 6856434, Japanese Patent No. 5948727, Japanese Patent No. 6492856, and Japanese Patent Laid-Open No. 2017-203677).

[0003] In the separation and classification methods described in these prior art documents, the separation time is prolonged to identify the type of hemoglobin in the blood sample, and the determination is made based on the peak pattern at a specific time, so the measurement requires time.

[0004] The present disclosure aims to provide a method for distinguishing the type of hemoglobin, a method for generating a learned model, a learned model, a hemoglobin analysis system, and a program that can identify the type of hemoglobin in a blood sample with higher accuracy in distinguishing a blood sample. Summary of the invention

[0005] One aspect of the method for distinguishing the category of hemoglobin disclosed in the present invention is: preparing a learned model, the learned model being generated by using a data set containing learning frequency analysis data and known categories of hemoglobin as a learning data set for machine learning, the learning frequency analysis data being generated by performing frequency analysis on separated data generated by separation processing of a learning blood sample whose category of hemoglobin is known; preparing inspection frequency analysis data by performing frequency analysis on separated data generated by separation processing of an inspection blood sample whose category of hemoglobin is unknown; and distinguishing the category of hemoglobin contained in the inspection blood sample by inputting the inspection frequency analysis data into the learned model.

[0006] Thus, in one embodiment, it is possible to provide a method for distinguishing the type of hemoglobin, which can identify the type of hemoglobin in a blood sample even when using separation data in which the separation time for the separation process of the blood sample is short.

[0007] Effects of the Invention

[0008] According to the present disclosure, there is provided a method for distinguishing the type of hemoglobin, which can determine the type of hemoglobin with higher distinction accuracy for a blood sample. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Exemplary embodiments of the present disclosure are described in detail based on the following drawings, in which:

[0010] Figure 1 This is a schematic diagram of the structure of an HPLC device.

[0011] Figure 2 This is a block diagram of a control system of an HPLC device serving as a learned model generation device.

[0012] Figure 3 This is a flowchart for explaining a method for generating a learned model according to the present embodiment of the present disclosure.

[0013] Figure 4 This is a diagram showing data of machine learning for distinguishing the types of hemoglobin according to the present disclosure, and is a diagram of a chromatogram obtained by measuring a blood sample using an HPLC device.

[0014] Figure 5 Then Figure 4 The diagram showing the data of the machine learning for distinguishing the type of hemoglobin involved in the present disclosure is Figure 4 A diagram of a chromatogram showing the result of performing a standardization process on the chromatogram of FIG.

[0015] Figure 6 Then Figure 5 The diagram showing the data of the machine learning for distinguishing the type of hemoglobin involved in the present disclosure is for implementing Figure 5 The chromatogram obtained by the standardization process is subjected to wavelet transformation.

[0016] Figure 7 This is a block diagram of a control system of an HPLC device serving as a hemoglobin identification device.

[0017] Figure 8 This is a flowchart for explaining a method for determining the type of hemoglobin contained in an unknown sample using a learned model according to the present embodiment of the present disclosure.

[0018] Fig. 9A This is a diagram for explaining the shape of a peak when a blood sample is measured at different separation times using an HPLC apparatus, and is a chromatogram when hemoglobin E is measured.

[0019] Fig. 9B This is a diagram for explaining the shape of a peak when a blood sample is measured using an HPLC apparatus at different separation times, and is a chromatogram when abnormal hemoglobin A is measured.

[0020] Fig. 10A9 is a diagram for explaining the shape of a peak when a blood sample is measured with a different separation time using an HPLC apparatus different from the HPLC apparatus used for the measurement in FIG. 9 , and is a chromatogram when hemoglobin E is measured.

[0021] Fig. 10B 9 is a diagram for explaining the shape of a peak when a blood sample is measured with a different separation time using an HPLC apparatus different from the HPLC apparatus used for the measurement in FIG. 9 , and is a chromatogram when abnormal hemoglobin A is measured.

[0022] Fig.11A This is a diagram for explaining the shape of a peak when the same blood sample is measured using different HPLC apparatuses, and is a chromatogram when the measurement time is long.

[0023] Fig. 11B This is a diagram for explaining the shape of a peak when the same blood sample is measured using different HPLC apparatuses, and is a chromatogram when the measurement time is short.

[0024] Fig. 12A is a diagram showing data of machine learning for distinguishing the type of hemoglobin according to another embodiment of the present disclosure. Figure 4 A graph of the amplitude spectrum obtained by Fourier transforming the chromatogram.

[0025] Fig. 12B is a diagram showing data of machine learning for determining the type of hemoglobin according to another embodiment of the present invention. Figure 4 A diagram of the power spectrum obtained by Fourier transforming the chromatogram. DETAILED DESCRIPTION

[0026] Hereinafter, one embodiment of the present disclosure will be described.

[0027] (Structure of Learning Model Generating Device)

[0028] exist Figure 1 , a schematic configuration diagram of an HPLC device 10 using high performance liquid chromatography (HPLC) is shown as an example of a learned model generation device in the present embodiment. The learned model generation device is an example of a hemoglobin analysis system.

[0029] The HPLC apparatus 10 is configured to be provided with a blood collection tube 18 , automatically measure the concentrations of glycated hemoglobin (e.g., HbA1a, HbA1b, HbA1c, etc.) and variant Hb (e.g., HbF, HbS, etc.) in whole blood, and has a sample preparation unit 12 , an analysis unit 58 , and a photometry unit 16 .

[0030] The HPLC apparatus 10 includes a plurality of eluent bottles 22A, 22B, 22C, 22D, and 22E ( Figure 1 Each of the eluent bottles 22A to 22E holds eluents A to E to be supplied to an analytical column 60 described later. Each eluent differs in composition, component ratio, pH, osmotic pressure, etc., depending on the application.

[0031] The blood collection tube 18 is stored in, for example, a rack (not shown), and is configured to be movable to a position where blood can be collected by a nozzle 12N in the sample preparation unit 12 described later.

[0032] The sample preparation unit 12 is for preparing a sample to be introduced into the analytical column 60 from the blood collected from the blood collection tube 18. The sample preparation unit 12 includes a nozzle 12N and a dilution reservoir 24.

[0033] The nozzle 12N is used to collect various liquids including the blood sample 20 in the blood collection tube 18, and can suck and discharge the liquid and can move in the vertical direction and the horizontal direction. The operation of the nozzle 12N is controlled by the control unit 100 described later.

[0034] The analysis unit 58 is used to control the adsorption and desorption of components contained in the blood sample such as hemoglobin, which is a biological component, with respect to the filler of the analysis column 60, and to supply various biological components to the photometric unit 16. The set temperature in the analysis unit 58 is, for example, about 40° C. The analysis column 60 holds a filler for selectively adsorbing the components contained in the sample. As the filler, for example, methacrylic acid-methacrylic acid ester copolymer is used.

[0035] The analysis unit 58 includes a manifold 61 , a liquid delivery pump 62 , and an injection valve 63 in addition to the analysis column 60 .

[0036] The manifold 61 is used to selectively supply eluent from a specific eluent bottle among the plurality of eluent bottles 22A to 22E to the analytical column 60. The manifold 61 is connected to the eluent bottles 22A, 22B, 22C, 22D, and 22E via pipes 80A to 80E, respectively, and is connected to the injection valve 63 via pipe 84.

[0037] The liquid delivery pump 62 is used to apply power to move the eluent toward the injection valve 63 , and is provided in the middle of the pipe 84 .

[0038] The injection valve 63 collects a certain amount of the sample for introduction, and can introduce the sample for introduction into the analytical column 60, and has a plurality of introduction ports and discharge ports (not shown). The injection valve 63 is connected to an injection loop 64. The injection loop can hold a certain amount (e.g., several μL) of liquid. By appropriately switching the injection valve 63, it is possible to select a state in which the injection loop 64 is connected to the dilution tank 24 and the sample for introduction is supplied from the dilution tank 24 to the injection loop 64, and a state in which the injection loop 64 is connected to the analytical column 60 via the pre-filter PF and the piping 85 and the sample for introduction is introduced from the injection loop 64 to the analytical column 60. As such an injection valve 63, for example, a six-way valve can be used. In addition, the pre-filter PF is a filter for filtering the sample and the eluent.

[0039] The photometric unit 16 is used to optically detect hemoglobin contained in the sample supplied from the analytical column 60 , and is connected to a waste liquid tank 88 for discharging the sample supplied from the analytical column 60 via a pipe 87 .

[0040] Here, when the sample is introduced into the analytical column 60 , the time for which the analytical column 60 retains the components contained in the sample varies for each component, so the amount of the components contained in the sample supplied to the photometric unit 16 changes with time. In other words, the components contained in the sample are separated by the analytical column 60 .

[0041] Then, the photometric unit 16 is controlled by a control unit 100 described later to continuously measure the absorbance of a component included in the sample supplied from the analytical column 60 , and transmits the absorbance measurement result to the control unit 100 .

[0042] Therefore, separation data of the sample can be obtained by recording the absorbance measurement results sent from the photometric unit 16. That is, the analysis unit 58 is a member that performs separation processing in this embodiment, and the analysis unit 58 and the photometric unit 16 are examples of separation units in this embodiment.

[0043] Figure 2 1 shows a block diagram of a control system of the HPLC device 10. Figure 2 As shown, the HPLC device 10 includes a control unit 100. The control unit 100 is a structure in which a CPU (Central Processing Unit) 100A, a ROM (Read Only Memory) 100B, a RAM (Random Access Memory) 100C, a nonvolatile memory 100D, and an input / output interface (I / F) 100E are connected via a bus 100F. In addition, the HPLC device 10 includes an operation unit (not shown) for receiving input from an operator and a display unit 26.

[0044] The sample preparation unit 12 , the analysis unit 58 , the photometry unit 16 , and the display unit 26 are connected to the input / output interface 100E.

[0045] In the present embodiment, the display unit 26 is, for example, a liquid crystal display, and is controlled by the CPU 100A to display information from the control unit 100 to an operator.

[0046] In the present embodiment, as an example, the CPU 100A pre-records a basic program (not shown) for controlling the operation of the HPLC device 10 in the non-volatile memory 100D. In the present embodiment, as an example, the non-volatile memory 100D also stores a first program 120A for generating a learned model 120C. The CPU 100A reads and executes various programs stored in the non-volatile memory 100D. Alternatively, the first program 120A may be recorded on a recording medium such as a CD-ROM, and the CPU 100A may read and execute the first program 120A using a CD-ROM drive or the like.

[0047] In addition, the CPU 100A controls the manifold 61 and the injection valve 63 by executing the basic program, thereby changing the type of eluent introduced into the analytical column 60 or the flow rate of the eluent per unit time. Thus, the CPU 100A determines the flow rate of the components in the sample flowing out of the analytical column 60 per unit time, and sets the measurement time of the components contained in the sample (the time for measuring the absorbance by the photometric unit 16).

[0048] Next, refer to Figures 3 to 6 The following describes a process of generating a learned model used in the method for determining the type of hemoglobin according to the present embodiment.

[0049] (Process of Generating a Learning Model Based on the Method of the Present Embodiment)

[0050] In the learned model generation device of the present embodiment, the control unit 100 generates a learned model for distinguishing the type of hemoglobin contained in a sample by executing various programs according to the present embodiment. Specifically, when the CPU 100A reads the first program 120A, it executes the following processing.

[0051] (Processing by the CPU 100A to be executed by the first program 120A)

[0052] In this embodiment, the first program 120A causes the CPU 100A to execute Figure 3 Processing shown.

[0053] In step S12, the CPU 100A stores the absorbance corresponding to the component of the sample of known hemoglobin type (hereinafter referred to as known sample) measured by the photometric unit 16 in the RAM 100C, and creates a Figure 4 Then, the CPU 100A stores the created chromatogram in the RAM 100C. Then, the CPU 100A transfers to step S14.

[0054] In step S14, the CPU 100A performs normalization processing on the chromatogram stored in the RAM 100C to create a Figure 5 Then, the CPU 100A stores the created standardized data in the RAM 100C. Then, the CPU 100A transfers to step S16.

[0055] In addition, the specific transformation method of the normalization process is shown in the following process and formula 1.

[0056] 1. A normalization process is performed on each data point of the chromatogram (a scaling process in which a value obtained by subtracting the average value from the measured value is further divided by the standard deviation for each measured data point).

[0057] 2. The value obtained by adding 0.001 to the absolute value of the minimum value in the normalized data is added to each normalized data point.

[0058] 3. For each data point, take the logarithm to the base e (Napier's constant).

[0059] [Formula 1]

[0060] In addition, in the above formula 1, x' i is the value of each data point i after standardization, x i is the value at each data point i in the chromatogram. In addition, μ is the average value of each data point in the chromatogram, and σ is the standard deviation of the value of each data point in the chromatogram. In addition, Min is a function representing the minimum value among a plurality of values. In addition, ln is a logarithmic function with Napier's constant as the base.

[0061] In addition, the above-mentioned conversion method and parameters are examples of the normalization process involved in this embodiment, and the normalization process involved in this embodiment is not limited to the above-mentioned conversion method and parameters.

[0062] In step S16, CPU 100A performs wavelet analysis on the normalized data stored in RAM 100C to create Figure 6 Then, the CPU 100A records the created wavelet analysis data in the ROM 100B. Then, the CPU 100A transfers to step S18.

[0063] In step S18, the CPU 100A creates a learning data set that associates the type of hemoglobin contained in the sample, which is the basis for creating the wavelet analysis data, with the wavelet analysis data. Then, the CPU 100A records the created learning data set in the ROM 100B. Then, the CPU 100A moves to step S20.

[0064] In step S20, CPU 100A checks the number of recorded learning data sets. Then, if the number of learning data sets is less than the predetermined number, a positive determination is made in step S20, and the process moves to step S12. On the other hand, if the number of learning data sets reaches the predetermined number, a negative determination is made in step S20, and the process moves to step S22.

[0065] In step S22, the CPU 100A constructs a convolutional neural network and inputs it as a plurality of learning data sets recorded in the ROM 100B. Then, the CPU 100A learns the type and amount of hemoglobin contained in the sample and the characteristics of the wavelet analysis data based on the plurality of learning data sets input to the convolutional neural network. Then, the CPU 100A transfers to step S24.

[0066] In step S24, the CPU 100A records the learned parameters in the nonvolatile memory 100D. Then, the CPU 100A ends the processing by the first program 120A.

[0067] The learned parameters are part of a learned model 120C described later in combination with a subroutine (not shown). In other words, the first program 120A causes the CPU 100A to execute a process for generating the learned model 120C.

[0068] In addition, as described above, the first program 120A performs wavelet analysis on the normalized data. Thus, in this embodiment, the CPU 100A that executes the first program 120A can be said to be an example of a frequency analysis unit.

[0069] Next, refer to Figure 1 , Figure 7 and Figure 8 The process of determining the type of hemoglobin contained in a sample whose type of hemoglobin is unknown (hereinafter referred to as an unknown sample) using the learned model involved in the present embodiment is described. In addition, in the HPLC device 11 as the hemoglobin identification device involved in the present embodiment, the other structures of the HPLC device 10 except the control unit 100 are the same as those of the HPLC device 10.

[0070] (Structure of hemoglobin identification device)

[0071] Figure 7 1 is a block diagram of a control unit 101 of an HPLC device 11 as a hemoglobin identification device according to the present embodiment. The hemoglobin identification device is an example of a hemoglobin analysis system.

[0072] The HPLC device 11 as the hemoglobin discrimination device is different from the HPLC device 10 as the learned model generation device described above in that the structure of the program stored in the nonvolatile memory 100D is different. Specifically, the hemoglobin discrimination device stores in the nonvolatile memory a learned model 120C and a second program 120B for discriminating hemoglobin in a blood sample using the learned model 120C.

[0073] The CPU 100A in the hemoglobin discrimination device reads and executes various programs stored in the nonvolatile memory 100D. In addition, the learned model 120C is configured to include the learned parameters generated by the learned model generation device. The learned model 120C can be transmitted from the nonvolatile memory 100D of the learned model generation device in any manner, and as an example, it can be recorded in the nonvolatile memory 100D of the hemoglobin discrimination device via a network. In addition, the second program 120B can also be recorded on a recording medium such as a CD-ROM, and the CPU 100A can read it and execute it by using a CD-ROM drive or the like.

[0074] The other structures in the control unit 101 are the same as those of the control unit 100 of the HPLC device 10 of the above-mentioned learned model generation device. In the following description, when the HPLC device 10 of the learned model generation device and the HPLC device 11 as the hemoglobin discrimination device are not distinguished, they are referred to as HPCL devices 10 and 11. Similarly, in the following description, when the control unit 100 of the HPLC device 10 and the control unit 101 of the HPLC device 11 are not distinguished, they are referred to as control units 100 and 101.

[0075] (Process of Discriminating the Type of Hemoglobin by the Method of the Present Embodiment)

[0076] The hemoglobin identification device of the present embodiment identifies the type of hemoglobin contained in a sample by executing various programs according to the present embodiment by the control unit 101. Specifically, when the CPU 100A reads the second program 120B, it executes the following processing.

[0077] (Processing Executed by the Second Program 120B on the CPU 100A)

[0078] In this embodiment, the second program 120B causes the CPU 100A to execute Figure 8 Processing shown.

[0079] In step S42, the CPU 100A reads the learned model 120C (learned parameters and subroutines) recorded in the ROM 100B. Then, the CPU 100A proceeds to step S44.

[0080] In step S44, the CPU 100A stores the absorbance corresponding to the component of the unknown sample measured by the photometric unit 16 in the RAM 100C, and creates a Figure 4 Then, the CPU 100A stores the created chromatogram in the RAM 100C. Then, the CPU 100A transfers to step S46.

[0081] In step S46, the CPU 100A performs normalization processing on the chromatogram stored in the RAM 100C to create a Figure 5 Then, the CPU 100A stores the created standardized data in the RAM 100C. Then, the CPU 100A transfers to step S48.

[0082] In step S48, CPU 100A performs wavelet analysis on the normalized data stored in RAM 100C to create Figure 6 Then, the CPU 100A records the created wavelet analysis data in the ROM 100B.

[0083] In step S50, the CPU 100A inputs the wavelet analysis data recorded in the ROM 100B to the learned model 120C, and outputs the type of hemoglobin contained in the unknown sample. Then, the CPU 100A ends the processing by the second program 120B.

[0084] In addition, the hemoglobin identification device according to the present embodiment may further execute the following processing.

[0085] For example, in step S50 , when the type of hemoglobin contained in the unknown sample has not been identified, the CPU 100A may cause the display unit 26 to display that the type of hemoglobin has not been identified.

[0086] On the other hand, in step S50, when the type of hemoglobin contained in the unknown sample is determined, other processing may be further performed. For example, the CPU 100A may calculate the amount of hemoglobin contained in the unknown sample for each hemoglobin type based on the value of the peak corresponding to the hemoglobin type of the target of the chromatogram of the unknown sample, and display the calculated value on the display unit 26.

[0087] In addition, as described above, the second program 120B performs wavelet analysis on each of the normalized data. Thus, in this embodiment, the CPU 100A that executes the second program 120B can be said to be an example of a frequency analysis unit.

[0088] Next, the results of discriminating hemoglobin contained in an unknown sample using the method for generating the learned model 120C and the method for discriminating the type of hemoglobin involved in this embodiment will be described as Examples 1 and 2. In each example, the so-called "unknown sample" is a sample used in the discrimination test in the learned model 120C, and actually refers to a sample that has been discriminated for the type of hemoglobin contained in the sample and is not used for generating the learned model 120C.

[0089] In each example, a diabetes test device HA-8180V or HA-8190V manufactured by ARKRAY Co., Ltd. was used. In each test device, the test time was set to 90 seconds for HA-8180V and 24 seconds for HA-8190V. The sampling frequency was set to 2048 Hz (sampling interval was 488.28125 microseconds).

[0090] In each example, a blood sample containing hemoglobin S, hemoglobin C, hemoglobin D, or hemoglobin E and a blood sample in which hemoglobin S, hemoglobin C, hemoglobin D, and hemoglobin E were not observed were prepared. In addition, a sample in which hemoglobin S, hemoglobin C, hemoglobin D, and hemoglobin E were not observed will be referred to as a "healthy subject specimen" hereinafter.

[0091] (Example 1)

[0092] In Example 1, HA-8180V is used as a measuring device to create a learned model 120C and to discriminate the type of hemoglobin contained in an unknown sample using the learned model 120C.

[0093] In this embodiment, during the learning phase, use Figure 3 More specifically, in this embodiment, a chromatogram of a known sample is created, and after the created chromatogram is standardized, wavelet analysis data is created. Then, a learning data set including the wavelet analysis data is created, and a learning model 120C is created by learning using a convolutional neural network.

[0094] In addition, in this embodiment, in the estimation stage of unknown samples, the Figure 8More specifically, in this embodiment, a chromatogram of an unknown sample is created, and after the created chromatogram is standardized, wavelet analysis data is created. Then, the type of hemoglobin contained in the unknown sample is determined based on the learned model 120C for the wavelet analysis data.

[0095] Table 1 shows the amount of learning data (the number of learning data sets, the number of known samples) and the amount of test data (the number of samples used in the discrimination test, the number of unknown samples) used in the learning involved in this embodiment, the number of correct answers and the correct answer rate.

[0096]

Table 1

[0097] (Example 2)

[0098] In Example 2, HA-8190V is used as a measuring device to create a learned model 120C and to discriminate the type of hemoglobin contained in an unknown sample using the learned model 120C.

[0099] In this embodiment, during the learning phase, use Figure 3 More specifically, in this embodiment, a chromatogram of a known sample is created, and after the created chromatogram is standardized, wavelet analysis data is created. Then, a learning data set containing the wavelet analysis data is created, and a learning model is created by learning using a convolutional neural network.

[0100] In addition, in this embodiment, in the estimation stage of unknown samples, the Figure 8 More specifically, in this embodiment, a chromatogram of an unknown sample is created, and after the created chromatogram is standardized, wavelet analysis data is created. Then, the type of hemoglobin contained in the unknown sample is determined based on the learned model for the wavelet analysis data.

[0101] Table 2 shows the number of learning data and the number of test data used in the learning involved in this embodiment, the number of correct answers, and the correct answer rate.

[0102]

Table 2

[0103] (Summarize)

[0104] As described above, the following is shown by the HPLC apparatuses 10 and 11 in the present embodiment.

[0105] As shown in Tables 1 and 2, it is shown that the type of hemoglobin can be distinguished using the method for creating a learned model and the method for distinguishing the type of hemoglobin in Examples 1 and 2. In addition, based on the results, it is considered that the accuracy of determining the type of hemoglobin in a blood sample based on the learned model created using wavelet analysis data is improved. That is, it is shown that the method for creating a learned model and the method for distinguishing the type of hemoglobin in this embodiment can determine the type of hemoglobin in a blood sample even when the separation time is short.

[0106] (Method for distinguishing the type of hemoglobin according to reference example)

[0107] Another method for discriminating the type of hemoglobin contained in an unknown sample is to discriminate based on the shape of peaks appearing in a chromatogram (the positional relationship between peaks appearing in a chromatogram).

[0108] Fig. 9A This is a chromatogram obtained for a blood sample containing hemoglobin E using the HA-8180V as the measuring device and setting the flow rate to a measuring time of 200 seconds. Fig. 9A As shown in FIG. 1 , in this sample, two peaks were confirmed before and after the maximum peak, which is the peak of hemoglobin A0.

[0109] Likewise, Fig. 9B This is a chromatogram obtained for a blood sample containing abnormal hemoglobin A, with the measurement time set to 200 seconds. Fig. 9B As shown in the chromatogram, peak A appears at around 30 seconds as the divided area, but Fig. 9A Different from the peak B, no peak B appeared. Therefore, it is considered that according to this measurement condition, by observing the presence or absence of peak B in a sample having peak A, hemoglobin E and abnormal hemoglobin A can be distinguished. In this way, by setting the measurement time to 200 seconds, hemoglobin can be distinguished, but this measurement time is longer than the time set in the above-mentioned Example 1.

[0110] Here, Fig. 10A Is used with Fig. 9A The chromatogram was obtained using the same sample and the same apparatus as in Example 1, with the measurement time set to 90 seconds. Fig. 10A As shown, according to the chromatogram, as the divided area, peak A appears at around 25 seconds, but peak B does not appear.

[0111] In addition, similarly, Fig. 10B is with Fig. 9B The same sample is a chromatogram obtained using a different measuring device, HA-8190V, with the measuring time set to 90 seconds. Fig. 10BAs shown, according to the chromatogram, as the divided area, peak A appears at around 25 seconds, but peak B does not appear.

[0112] so, Fig. 9A The peak B generated in Fig. 10A Since the hemoglobin is not classified, it is easy to misjudge the type of hemoglobin contained in the sample when the measurement time is shortened.

[0113] In addition, for example, Fig.11A This is a chromatogram obtained for a blood sample containing hemoglobin S when the measurement time is set to 90 seconds. Fig.11A As shown, in the sample containing hemoglobin S, peak C appears about 28 seconds after the largest peak (hemoglobin A0 peak) appears about 20 seconds later.

[0114] on the other hand, Fig. 11B Is to use Fig.11A The chromatogram was obtained using the same sample and the same apparatus as in Example 2, with the measurement time set to 24 seconds. Fig. 11B As shown, according to the chromatogram, peak C does not appear after the largest peak that appears around 18 seconds.

[0115] As described above, even when the same HPLC apparatuses 10 and 11 are used, when the measurement time is shortened, the type of hemoglobin contained in the sample is easily erroneously determined.

[0116] On the other hand, in HA-8180V as an example of the HPLC apparatus 10, 11 in the present embodiment, it was shown that the type of hemoglobin in the blood sample can be determined in a measurement time of 90 seconds. In addition, in HA-8190V as an example of the HPLC apparatus 10, 11 in the present embodiment, it was shown that the type of hemoglobin in the blood sample can be determined in a measurement time of 24 seconds. Therefore, it is shown that by creating the learned model 120C using wavelet analysis data, it is possible to determine the type of hemoglobin in the blood sample even when using separation data with a short separation time for separation processing of the blood sample.

[0117] In addition, even when the separation time of the separation process of the blood sample is long, the peak of the chromatogram is not clear, and the discrimination accuracy of the hemoglobin type based on the chromatogram deteriorates. On the other hand, in the method for generating a learned model and the method for distinguishing the type of hemoglobin involved in the present embodiment, frequency analysis is performed on the components contained in the blood sample, so even if the peaks are repeated or the peaks appearing in the chromatogram are slightly unclear, the repeated peaks are separated as the frequency analysis results. Therefore, according to the method for generating a learned model and the method for distinguishing the type of hemoglobin involved in the present embodiment, even in a chromatogram with unclear peaks, the effect of improving the discrimination accuracy of the hemoglobin type can be expected.

[0118] In addition, for hemoglobin F, hemoglobin A2 (HbA2), and other variant hemoglobin types not implemented in the above-mentioned Examples 1 and 2, a learned model can be generated and a method for distinguishing the type of hemoglobin can be applied in the same manner as in the present embodiment.

[0119] (Other embodiments)

[0120] In the above description, the control units 100 and 101 are built into the HPLC devices 10 and 11, but the present invention is not limited thereto. For example, the main structure of the control units 100 and 101 may be arranged on a network, and the HPLC devices 10 and 11 may communicate with the structure arranged on the network, thereby operating as the control units 100 and 101. In addition, for example, the first program 120A and the second program 120B may be obtained by appropriately downloading from a server arranged on the network by the control unit 100 and 101 of the HPLC device 10. In addition, for example, the server arranged on the network may communicate with the HPLC devices 10 and 11, and after the HPLC devices 10 and 11 send the chromatogram to the server, the server sends the frequency analysis data to the HPLC devices 10 and 11, thereby operating as the frequency analysis unit.

[0121] In addition, the first program 120A, the second program 120B, and the learned model 120C are executed by the control units 100 and 101, but are not limited thereto, and may be executed by a server configured on the network. For example, the HPLC apparatuses 10 and 11 may also be configured as follows: the chromatogram of the unknown sample is sent to the server configured on the network, and the amount of hemoglobin contained in the unknown sample is received from the server. These configurations are all included in the hemoglobin analysis system in this embodiment.

[0122] In addition, in the method for distinguishing the type of hemoglobin involved in the present disclosure, as long as the learned model 120C can be prepared, the method for preparing the learned model 120C is not limited. In other words, as long as the learned model 120C can be prepared, the device for creating the chromatogram, the device for performing frequency analysis, the device for generating the learned model 120C, and the device for distinguishing are not limited.

[0123] More specifically, the device for creating the chromatogram may be a device that is separate from the HPCL devices 10 and 11 and has only the function of creating the chromatogram based on the blood sample. In addition, the device for performing frequency analysis may be a device that is separate from the HPCL devices 10 and 11 and has only the function of performing wavelet analysis based on the chromatogram. In addition, in the above description, the learning model generation device generates the learned model 120C, and the hemoglobin discrimination device discriminates the type of hemoglobin based on the learned model 120C, but these devices may also be the same device. In addition, these devices may all be built into the same HPLC devices 10 and 11, or some of the devices may be built into a common device, or they may be different devices. Therefore, according to the technology of this embodiment, if the learned model 120C created according to each model of the HPLC device 11 is prepared, the creation of the learned model 120C may not be performed in each individual HPLC device 10.

[0124] In the above description, the control units 100 and 101 execute various programs, and thus the processing involved in the embodiment is implemented by a software structure, but the present invention is not limited thereto. For example, the processing involved in the above embodiment may also be implemented by a hardware structure or a combination of a hardware structure and a software structure.

[0125] In the above description, wavelet analysis data is used as frequency analysis data, but the technology involved in the present disclosure is not limited to this.

[0126] For example, in frequency analysis, instead of wavelet analysis, a fast Fourier transform (Fast Fourier Transform) may be performed to use amplitude data as shown in FIG. 12(A) or power spectrum data as shown in FIG. 12(B). Alternatively, instead of wavelet analysis, data obtained by performing a short-time Fourier transform (Short-Time Fourier Transform) may be used.

[0127] In the above description, the learning data set uses the hemoglobin concentration and wavelet analysis data of the known sample, but it is not limited to this, and the chromatogram can also be included in the learning data set. In this case, in the discrimination process, the chromatogram of the unknown sample is included in the input data input to the learned model 120C.

[0128] In addition, in the above description, as an example of an analysis system for hemoglobin, HPLC devices 10 and 11 are used, but the technology involved in the present disclosure is not limited to this. That is, as long as the chromatograms of known samples and unknown samples can be obtained, it is not limited to HPLC devices 10 and 11, and an electrophoresis analysis device can also be used. In the electrophoresis analysis device, based on the created spectrum, it is possible to arbitrarily perform standardization processing, perform wavelet analysis, and create wavelet analysis data. According to the technology in the present disclosure, even in this case, it is possible to generate a learning model 120C and distinguish the type of hemoglobin.

[0129] In addition, in the above description, in the method for generating the learned model 120C, the learned model 120C is generated on the basis of constructing a convolutional neural network, but the technology involved in the present disclosure is not limited to this. That is, as long as the learned model 120C can be obtained, for example, a recurrent neural network (Recurrent Neural Network) can be constructed to perform learning. Even in this case, the learned model 120C can be generated.

[0130] In addition, preferred embodiments of the present disclosure are further described below.

[0131] (Note 1)

[0132] A method for distinguishing the category of hemoglobin includes preparing a learned model, wherein the learned model is generated by using a data set containing learning frequency analysis data and known categories of hemoglobin as a learning data set for machine learning, wherein the learning frequency analysis data is generated by performing frequency analysis on separated data generated by separating a learning blood sample whose category of hemoglobin is known; preparing inspection frequency analysis data by performing frequency analysis on separated data generated by separating a inspection blood sample whose category of hemoglobin is unknown; and distinguishing the category of hemoglobin contained in the inspection blood sample by inputting the inspection frequency analysis data into the learned model.

[0133] (Note 2)

[0134] The method for distinguishing the type of hemoglobin according to Supplementary Note 1, wherein the separated data for the learning frequency analysis data and the separated data for the inspection frequency analysis data are chromatograms.

[0135] (Note 3)

[0136] A method for distinguishing the type of hemoglobin according to Note 1 or Note 2, wherein the learning frequency analysis data and the inspection frequency analysis data are generated by performing frequency analysis on the separated data for the learning frequency analysis data and the separated data for the inspection frequency analysis data, which have been subjected to standardization processing, respectively.

[0137] (Note 4)

[0138] The method for distinguishing the type of hemoglobin according to any one of Supplementary Notes 1 to 3, wherein the learning frequency analysis data and the inspection frequency analysis data are generated by performing wavelet analysis.

[0139] (Note 5)

[0140] A method for discriminating the type of hemoglobin according to any one of Notes 1 to 4, wherein the machine learning includes a convolutional neural network.

[0141] (Note 6)

[0142] The method for distinguishing the type of hemoglobin according to any one of Supplementary Notes 1 to 5, wherein the hemoglobin is hemoglobin S, hemoglobin C, hemoglobin D, or hemoglobin E.

[0143] (Note 7)

[0144] A method for generating a learned model generates learned parameters by using a data set containing learning frequency analysis data and known hemoglobin categories as a learning data set for machine learning, wherein the learning frequency analysis data is generated by performing frequency analysis on separated data generated by separation processing of a learning blood sample with known hemoglobin categories.

[0145] (Note 8)

[0146] The method for generating a learned model according to Supplementary Note 7, wherein the separation data is a chromatogram.

[0147] (Note 9)

[0148] The method for generating a learned model according to Supplementary Note 7 or Supplementary Note 8, wherein the frequency analysis for generating the learning frequency analysis data is performed on the separated data after the normalization process.

[0149] (Note 10)

[0150] The method for generating a learned model according to any one of Supplementary Notes 7 to 9, wherein the frequency analysis for generating the learning frequency analysis data is wavelet analysis.

[0151] (Note 11)

[0152] A method for generating a learned model according to any one of Notes 7 to 10, wherein the machine learning includes a convolutional neural network.

[0153] (Note 12)

[0154] The method for generating a learned model according to any one of Supplementary Notes 7 to 11, wherein the hemoglobin is hemoglobin S, hemoglobin C, hemoglobin D, or hemoglobin E.

[0155] (Note 13)

[0156] A learned model that, when frequency analysis data generated by performing frequency analysis on separation data generated by separation processing of a blood sample is input, outputs the category of hemoglobin contained in the blood sample based on the frequency analysis data, wherein the learned model is generated by using learning data containing learning frequency analysis data and known categories of hemoglobin as learning data sets for machine learning, and the learning frequency analysis data is generated by performing frequency analysis on separation data generated by separation processing of a learning blood sample whose category of hemoglobin is known.

[0157] (Note 14)

[0158] A hemoglobin analysis system comprises: a separation unit, which separates a test blood sample of unknown hemoglobin type to generate test separation data; a frequency analysis unit, which performs frequency analysis on the test separation data to generate test frequency analysis data; a discrimination unit, which discriminates the type of hemoglobin contained in the test blood sample measured based on the test frequency analysis data using the learned model described in Note 13; and an output unit, which outputs the amount of each hemoglobin contained in the test blood sample discriminated by the discrimination unit.

[0159] (Note 15)

[0160] A program causes a computer to perform the following processing: inputting a learned model generated by using learning data including learning frequency analysis data and known hemoglobin categories as learning data sets for machine learning, wherein the learning frequency analysis data is generated by performing frequency analysis on separated data generated by performing separation processing on a learning blood sample whose category of hemoglobin is known; inputting inspection frequency analysis data generated by performing frequency analysis on separated data generated by performing separation processing on an inspection blood sample whose category of hemoglobin is unknown; and outputting the category of hemoglobin contained in the inspection blood sample based on the learned model.

Claims

1. A method for distinguishing the type of hemoglobin, preparing a learned model generated by using a data set including learning frequency analysis data and known hemoglobin categories as a learning data set for machine learning, wherein the learning frequency analysis data is generated by performing frequency analysis on separation data generated by separation processing of a learning blood sample whose hemoglobin category is known; performing frequency analysis on separation data generated by separation processing of a test blood sample of unknown hemoglobin type to prepare test frequency analysis data; The type of hemoglobin contained in the test blood sample is determined by inputting the test frequency analysis data into the learned model.

2. The method for distinguishing the type of hemoglobin according to claim 1, wherein: The separated data for the learning frequency analysis data and the separated data for the inspection frequency analysis data are chromatograms.

3. The method for distinguishing the type of hemoglobin according to claim 1, wherein: The learning frequency analysis data and the inspection frequency analysis data are generated by performing frequency analysis on the separated data for the learning frequency analysis data and the separated data for the inspection frequency analysis data, which have been subjected to standardization processing, respectively.

4. The method for distinguishing the type of hemoglobin according to claim 1, wherein: The learning frequency analysis data and the inspection frequency analysis data are generated by performing wavelet analysis.

5. The method for distinguishing the type of hemoglobin according to claim 1, wherein: The machine learning includes convolutional neural networks.

6. The method for distinguishing the type of hemoglobin according to claim 1, wherein: The hemoglobin is hemoglobin S, hemoglobin C, hemoglobin D or hemoglobin E.

7. A method for generating a learned model, Learning completed parameters are generated by using a data set containing learning frequency analysis data and known hemoglobin categories as a learning data set for machine learning, wherein the learning frequency analysis data is generated by frequency analysis of separation data generated by separation processing of a learning blood sample with known hemoglobin categories.

8. The method for generating a learning model according to claim 7, wherein: The separation data is a chromatogram.

9. The method for generating a learning model according to claim 7, wherein: The frequency analysis for generating the learning frequency analysis data is performed on the separated data after the normalization process.

10. The method for generating a learned model according to claim 7, wherein: The frequency analysis for generating the learning frequency analysis data is wavelet analysis.

11. The method for generating a learned model according to claim 7, wherein: The machine learning includes convolutional neural networks.

12. The method for generating a learned model according to claim 7, wherein: The hemoglobin is hemoglobin S, hemoglobin C, hemoglobin D or hemoglobin E.

13. A learned model, which, when receiving frequency analysis data generated by frequency analysis of separation data generated by separation processing of a blood sample, outputs the type of hemoglobin contained in the blood sample based on the frequency analysis data, wherein: The learned model is generated by using learning data including learning frequency analysis data and known hemoglobin categories as a learning data set for machine learning, wherein the learning frequency analysis data is generated by performing frequency analysis on separated data generated by separation processing of a learning blood sample with a known hemoglobin category.

14. A hemoglobin analysis system comprising: a separation unit that performs separation processing on a blood sample for inspection whose type of hemoglobin is unknown, and generates separation data for inspection; a frequency analysis unit that performs frequency analysis on the inspection separation data to generate inspection frequency analysis data; a discriminating unit that discriminates the type of hemoglobin contained in the test blood sample measured based on the test frequency analysis data using the learned model according to claim 13; and The output unit outputs the type of hemoglobin determined by the determination unit.

15. A computer program product comprising a program, the program causing a computer to execute the following processing: inputting a learned model generated by using learning data including learning frequency analysis data generated by performing frequency analysis on separation data generated by separation processing of a learning blood sample whose hemoglobin category is known and a known hemoglobin category as a learning data set for machine learning; inputting test frequency analysis data generated by frequency analysis of separation data generated by separation processing of a test blood sample of unknown hemoglobin type; and The type of hemoglobin contained in the test blood sample is output based on the learned model.

Images