Sample analysis device, sample analysis method, and program

Abstract

To provide a specimen analyzer, a specimen analysis method and a program with which it is possible to reduce the load of a computer that analyzes the data obtained by specimen measurement, using an artificial intelligence algorithm.SOLUTION: A specimen analyzer 4000 for analyzing an analyte in a specimen comprises a measurement unit 4000 that includes an optical detection unit for acquiring an optical signal from the specimen, and an analysis unit 300 that analyzes first and second data corresponding to the optical signal. The analysis unit 300 executes a first analysis operation on the first data by an artificial intelligence algorithm, and executes a second analysis operation to process the representative value of the second data that corresponds to the feature of the analyte.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present invention relates to a specimen analysis apparatus, a specimen analysis method, and a program for analyzing a specimen.

Background Art

[0002] Patent Document 1 describes a method of analyzing signals obtained by measuring cells with a flow cytometer using an artificial intelligence algorithm and classifying the cells according to the type.

Prior Art Document

Patent Document

[0007] According to the specimen analysis device of the present invention, the specimen obtained from the specimen measurement Data analysis processing, Non-AI The first analysis operation by the algorithm, AI algorithm By dividing the work between the second analysis operation, measurement The data is uniform, AI Compared to analyzing data using only algorithms, this method reduces the load on the analysis unit, which is the computer that processes the data.

[0008] The present invention relates to a method for analyzing specimens, specifically for analyzing analytes in specimens. The process involves mixing the sample, staining reagent, and hemolytic reagent to prepare a sample for measurement (S301), and then optically measuring the sample. The process (S302), Measurement data obtained by optical measurement This includes an analysis process (S201, S202) that analyzes the following. In the analysis process (S201, S202), A first analysis operation using a non-AI algorithm and a second analysis operation using an AI algorithm (60) are performed on the measurement data, (i) the first analysis operation includes classifying the analytes into groups of at least neutrophils, lymphocytes, monocytes and eosinophils, and obtaining count results for each classified group, where, based on the measurement data, at least two representative values ​​corresponding to each characteristic of the analytes contained in the measurement sample are calculated, a two-dimensional plot is generated using at least two representative values, the analytes are classified into groups of at least neutrophils, lymphocytes, monocytes and eosinophils based on the two-dimensional plot, and count results for each classified group are obtained, (ii) based on the classification of the analytes by the second analysis operation, information suggesting the presence of abnormal cells is obtained, wherein the abnormal cells include at least one selected from the group consisting of nucleated erythrocytes, immature granulocytes, blast cells, abnormal lymphocytes, atypical lymphocytes, reactive lymphocytes, plasma cells, megakaryocytes and tumor cells. .

[0009] According to the specimen analysis method of the present invention, the specimen obtained from the specimen measurement Data analysis processing, Non-AI The first analysis operation by the algorithm, AI algorithm By dividing the work between the second analysis operation, measurement The data is uniform, AI Compared to analyzing data using only algorithms, this method reduces the load on the computer processing the data.

[0010] The present invention relates to a program that causes a computer (300, 600, 3001, 3002, 6001, 6002) to perform the process of analyzing analytes in a sample. The process involves mixing the sample, staining reagent, and hemolytic reagent to prepare the measurement sample, optically measuring the measurement sample, and obtaining the measurement results from the optical measurement. Process to analyze data (82a, 82b, 82c) andincluding Analyze The process A first analysis operation using a non-AI algorithm and a second analysis operation using an AI algorithm are performed on the measurement data, (i) the first analysis operation includes classifying the analytes into groups of at least neutrophils, lymphocytes, monocytes and eosinophils, and obtaining count results for each classified group, wherein at least two representative values ​​corresponding to each characteristic of the analytes contained in the measurement sample are calculated based on the measurement data, a two-dimensional plot is generated using at least two representative values, the analytes are classified into groups of at least neutrophils, lymphocytes, monocytes and eosinophils based on the two-dimensional plot, and count results for each classified group are obtained, (ii) based on the classification of the analytes by the second analysis operation, information suggesting the presence of abnormal cells is obtained, wherein the abnormal cells include at least one selected from the group consisting of nucleated erythrocytes, immature granulocytes, blast cells, abnormal lymphocytes, atypical lymphocytes, reactive lymphocytes, plasma cells, megakaryocytes and tumor cells.

[0011] According to the program of the present invention, the measurement analysis process of the data obtained from the sample Non-AI is shared between the first analysis operation based on an algorithm and AI algorithm the second analysis operation, thereby measurement reducing the load on the computer that processes the data compared to the case where the data is uniformly AI analyzed using only an algorithm.

Advantages of the Invention

[0018] According to the present invention, the load on a computer that analyzes data obtained from measurement of a sample by an artificial intelligence algorithm can be reduced.

Brief Description of the Drawings

[0019] [Figure 1] FIG. 1 is a diagram schematically showing a configuration example of a sample analysis apparatus according to Embodiment 1. [[ID=3३]] [Figure 2] FIG. 2 is a diagram showing an outline of analysis when an optical detection unit according to Embodiment 1 is a detection unit based on flow cytometry. ] [Figure 3] FIG. 3 is a diagram schematically showing waveform data and representative values according to Embodiment 1. ] [Figure 4] FIG. 4 is a diagram showing an outline of analysis when an optical detection unit according to Embodiment 1 is a detection unit that detects transmitted light or scattered light from a measurement sample. [Figure 5] FIG. 4 is a flowchart showing an example of a sample analysis method according to Embodiment 1. [Figure 6] FIG. 6 is a flowchart showing an example of setting an analysis operation based on rules set in an analysis unit according to Embodiment 2. [Figure 7] FIG,. 7 is a flowchart showing an example in which analysis is executed according to measurement items according to Embodiment 3. [Figure 8]Figure 8 is a schematic example diagram showing a screen for setting up AI analysis or computational processing analysis for each measurement item according to Embodiment 3. [Figure 9] Figure 9 is a flowchart showing an example of how analysis is performed according to a measurement order, according to Embodiment 3. [Figure 10] Figure 10 is a schematic example diagram showing a screen for setting the analysis mode for a measurement order according to Embodiment 3. [Figure 11] Figure 11 is a flowchart showing an example of how analysis is performed according to the analysis mode of the device, according to Embodiment 3. [Figure 12] Figure 12 is a schematic example diagram showing a screen for setting the analysis mode of the analysis unit according to Embodiment 3. [Figure 13] Figure 13 is a flowchart showing an example of how analysis is performed according to the type of measurement order in Embodiment 3. [Figure 14] Figure 14 is a schematic example diagram showing a screen for setting the analysis mode for each type of measurement order according to Embodiment 3. [Figure 15] Figure 15 is a flowchart showing an example of how analysis is performed according to the type of measurement item and measurement order in Embodiment 3. [Figure 16] Figure 16 is a schematic example diagram showing a screen for setting up AI analysis or computational processing analysis for each measurement item and measurement order type according to Embodiment 3. [Figure 17] Figure 17 is a flowchart showing an example of how the necessity of AI analysis is determined based on a flag obtained through computational processing analysis, according to Embodiment 3. [Figure 18] Figure 18 is a schematic example diagram showing a screen for setting up AI analysis for each flag of the analysis results according to Embodiment 3. [Figure 19] Figure 19 is a flowchart showing an example of performing AI analysis on a specific analyte classified by computational processing analysis according to Embodiment 3. [Figure 20]Figure 20 is a schematic diagram illustrating a screen for setting whether or not to perform AI analysis for each type of analyte according to Embodiment 3. [Figure 21] Figure 21 is a diagram illustrating the classification method by computational processing analysis and AI analysis performed in the process shown in Figure 19, according to Embodiment 3. [Figure 22] Figure 22 is a flowchart showing an example of executing AI analysis when a specific classification is performed in computational processing analysis according to Embodiment 3. [Figure 23] Figure 23 is a schematic example diagram showing a screen for setting whether or not to perform AI analysis on a specific type of analyte according to Embodiment 3. [Figure 24] Figure 24 is a block diagram showing the configuration of the measurement unit according to Embodiment 4. [Figure 25] Figure 25 is a schematic diagram showing the configuration of the optical system of the FCM detection unit according to Embodiment 4. [Figure 26] Figure 26 is a block diagram showing the configuration of the analysis unit according to Embodiment 4. [Figure 27] Figure 27 is a block diagram showing the configuration of the measurement unit when the sample analyzer performs counting and classification of blood cells in a blood sample, according to Embodiment 4. [Figure 28] Figure 28 is a block diagram showing the configuration of the sample aspiration unit and sample preparation unit in the measurement unit of Figure 27 according to Embodiment 4. [Figure 29] Figure 29 is a block diagram showing another configuration of the sample preparation unit shown in Figure 28, according to Embodiment 4. [Figure 30] Figure 30 is a flowchart showing an example of how analysis is performed according to the measurement channel in Embodiment 4. [Figure 31] Figure 31 is a schematic example diagram showing a screen for setting up AI analysis or computational processing analysis for each measurement channel according to Embodiment 4. [Figure 32] Figure 32 is a schematic diagram illustrating the waveform data used in the analysis method according to Embodiment 4. [Figure 33]Figure 33 is a schematic diagram showing an example of a method for generating training data used to train an AI algorithm for determining the type of analyte in a sample, according to Embodiment 4. [Figure 34] Figure 34 shows the label values ​​corresponding to cell types according to Embodiment 4. [Figure 35] Figure 35 is a schematic diagram illustrating a method for analyzing waveform data of analytes in a sample using an AI algorithm, according to Embodiment 4. [Figure 36] Figure 36 is a flowchart showing an example of performing AI analysis on waveform data acquired by a WDF channel according to Embodiment 4. [Figure 37] Figure 37 is a flowchart illustrating an example of Embodiment 4 in which nucleated red blood cells and basophils are classified by AI analysis and other cells are classified by computational processing analysis based on waveform data acquired by the WDF channel. [Figure 38] Figure 38 is a flowchart showing an example of performing AI analysis on neutrophils / basophils identified by the analysis of computational processing in the WDF channel, according to Embodiment 4. [Figure 39] Figure 39 is a schematic block diagram showing the configuration of the measurement unit according to Embodiment 5. [Figure 40] Figure 40 is a schematic side view showing the measurement by the detection block according to Embodiment 5. [Figure 41] Figure 41 is a flowchart showing an example of analysis according to Embodiment 5. [Figure 42] Figure 42 is a block diagram showing the configuration of the sample analyzer according to Embodiment 6. [Figure 43] Figure 43 is a block diagram showing the configuration of the analysis unit according to Embodiment 6. [Figure 44] Figure 44 is a block diagram showing another configuration of the sample analyzer according to Embodiment 6. [Figure 45] Figure 45 shows an example configuration of a parallel processing processor according to Embodiment 6. [Figure 46]Figure 46 is a schematic diagram showing an example of a parallel processing processor installation according to Embodiment 6. [Figure 47] Figure 47 is a schematic diagram showing an example of a parallel processing processor installation according to Embodiment 6. [Figure 48] Figure 48 is a schematic diagram showing an example of a parallel processing processor installation according to Embodiment 6. [Figure 49] Figure 49 shows another example of the parallel processing processor according to Embodiment 6. [Figure 50] Figure 50 is a diagram showing an example configuration of a parallel processing processor that performs arithmetic processing according to Embodiment 6. [Figure 51] Figure 51 is a diagram illustrating the overview of matrix operations performed by the parallel processing processor according to Embodiment 6. [Figure 52] Figure 52 is a conceptual diagram illustrating the case of Embodiment 6, in which multiple arithmetic processes are executed in parallel on a parallel processing processor. [Figure 53] Figure 53 is a schematic diagram showing an overview of the computational processing related to the convolutional layer according to Embodiment 6. [Figure 54] Figure 54 is a flowchart showing the analysis operation of the analysis unit and measurement unit according to Embodiment 6. [Figure 55] Figure 55 is a flowchart showing the details of the AI ​​analysis in step S201 of Figure 54, according to Embodiment 6. [Figure 56] Figure 56 is a flowchart detailing step S2011 of Figure 55 according to Embodiment 6. [Figure 57] Figure 57 is a block diagram showing another configuration of the measurement unit according to Embodiment 6. [Figure 58] Figure 58 is a block diagram showing another configuration of the analysis unit according to Embodiment 6. [Figure 59] Figure 59 is a block diagram showing another configuration of the measurement unit according to Embodiment 6. [Figure 60] Figure 60 is a block diagram showing another configuration of the analysis unit according to Embodiment 6. [Figure 61] Figure 61 is a block diagram showing another configuration of the sample analyzer according to Embodiment 6. [Figure 62] Figure 62 is a block diagram showing other configurations of the measurement unit according to Embodiment 6. [Figure 63] Figure 63 is a block diagram showing another configuration of the analysis unit according to Embodiment 6. [Figure 64] Figure 64 is a diagram showing an example configuration of a parallel processing processor that performs arithmetic processing according to Embodiment 6. [Figure 65] Figure 65 is a block diagram showing the configuration of a computer according to Embodiment 6. [Figure 66] Figure 66 is a block diagram showing another configuration of the measurement unit according to Embodiment 6. [Figure 67] Figure 67 is a block diagram showing another configuration of the analysis unit according to Embodiment 6. [Figure 68] Figure 68 is a block diagram showing another configuration of the measurement unit according to Embodiment 6. [Figure 69] Figure 69 is a block diagram showing another configuration of the analysis unit according to Embodiment 6. [Figure 70] Figure 70 is a block diagram showing another configuration of the measurement unit according to Embodiment 6. [Figure 71] Figure 71 is a block diagram showing another configuration of the analysis unit according to Embodiment 6. [Figure 72] Figure 72 is a block diagram showing another configuration of the measurement unit according to Embodiment 6. [Figure 73] Figure 73 is a block diagram showing another configuration of the analysis unit according to Embodiment 6. [Figure 74] Figure 74 is a schematic diagram showing the configuration of a waveform data analysis system according to Embodiment 7. [Figure 75] Figure 75 is a block diagram showing the configuration of a deep learning device according to Embodiment 7. [Figure 76] Figure 76 is a functional block diagram of a deep learning device according to Embodiment 7. [Figure 77] Figure 77 is a flowchart showing the processing performed by the deep learning device according to Embodiment 7. [Figure 78] Figure 78 is a schematic diagram illustrating the structure of a neural network according to Embodiment 7, a schematic diagram showing operations at each node, and a schematic diagram showing operations between nodes. [Modes for carrying out the invention]

[0020] The outline and embodiments of the present invention will be described in detail below with reference to the accompanying drawings. In the following description and drawings, the same reference numerals indicate the same or similar components, and for convenience, descriptions of the same or similar components will be omitted.

[0021] [Embodiment 1] This embodiment discloses a sample analyzer, a sample analysis method, and a program capable of performing both analysis using an artificial intelligence (AI) algorithm and analysis without using an AI algorithm on data obtained by measuring a sample.

[0022] In AI algorithm-based analysis, data is analyzed, for example, through extensive matrix operations. Hereafter, for convenience, AI algorithm-based analysis will be referred to as "AI analysis." AI analysis involves, for example, convolution operations performed by AI algorithms.

[0023] In analyses that do not use AI algorithms, data is analyzed by, for example, calculating representative values ​​corresponding to the characteristics of the analyte. Hereinafter, for convenience, the analysis method that analyzes data by calculating representative values ​​corresponding to the characteristics of the analyte without using AI algorithms will be referred to as "computational analysis" or "non-AI analysis." The amount of representative values ​​processed in computational analysis is smaller than the amount of data input to the AI ​​algorithm in AI analysis. In computational analysis, the amount of data to be processed and the amount of computation are less compared to AI analysis, so the load on the computer performing the analysis is less compared to AI analysis. This makes it possible to shorten the TAT (Turn Around Time) of the analysis of measurement results.

[0024] According to the specimen analyzer, specimen analysis method, and program of Embodiment 1, the analysis of data obtained by measuring a specimen can be divided between AI analysis and computational processing analysis, thereby reducing the load on the computer performing the analysis.

[0025] Figure 1 is a schematic diagram showing an example configuration of the sample analyzer 4000 of Embodiment 1. In Figure 1, the upper diagram shows an example configuration of the sample analyzer 4000 of the embodiment, and the lower diagram shows an example of a modification of the configuration of Embodiment 1.

[0026] As shown in the upper part of Figure 1, the specimen analyzer 4000 of Embodiment 1 comprises, for example, a measurement unit 400 and an analysis unit 300. Alternatively, as shown in the lower part of Figure 1, the specimen analyzer 4000 may be composed of an integrally configured measurement unit 400 and analysis unit 300. Examples of specimen analyzers 4000 include blood cell analyzers, urine analyzers, blood coagulation analyzers, immunoassay analyzers, biochemical analyzers, and gene analyzers. The analytes to be analyzed by the specimen analyzer 4000 are, for example, cells, formed elements, proteins, genes, etc.

[0027] The measurement unit 400 measures the sample and acquires data about the sample. The analysis unit 300 analyzes the data acquired by the measurement unit 400. The analysis unit 300 may also have functions for setting the measurement conditions of the sample to be measured in the measurement unit 400 and controlling the execution of the measurement. The analysis unit 300 is configured as a separate device (e.g., a computer) from the measurement unit 400 and is connected to the measurement unit 400. The analysis unit 300 and the measurement unit 400 are connected by wire or wireless.

[0028] The measurement unit 400 includes an optical detection unit for measuring a measurement sample prepared from a specimen.

[0029] The optical detection unit is, for example, a detection unit based on flow cytometry and is used for measuring blood and urine samples. The optical detection unit acquires an optical signal by irradiating the sample being measured, which is flowing through a flow cell, with light. For example, the optical detection unit irradiates the sample being measured, which contains analytes (e.g., cells and formed elements), flowing through a flow cell with light, causing forward scattered light, side scattered light, and fluorescence to be generated from the analytes. A photodetector provided in the optical detection unit receives the generated light and outputs an optical signal corresponding to the intensity of the received light. The optical signal is a waveform analog signal corresponding to the time change of forward scattered light, side scattered light, and fluorescence. An A / D conversion unit provided in the optical detection unit converts the optical signal digitally and acquires waveform digital data (hereinafter referred to as "waveform data") corresponding to each of the analytes. In this case, the waveform data is used, for example, for classifying leukocyte types in blood samples, classifying the number of red blood cells and white blood cells in blood samples, or classifying formed elements in urine samples.

[0030] The optical detection unit may be configured to irradiate the sample contained in a container with light and detect the light transmitted from the sample or the scattered light scattered from the sample with a photodetector. In this case, the optical detection unit irradiates the sample, which contains the analyte and is left standing in the container, with light. The photodetector provided in the optical detection unit receives the transmitted light that has passed through the sample or the scattered light generated by the sample for a predetermined period of time and outputs an optical signal corresponding to the intensity of the received light. In this case, the optical signal is a waveform analog signal corresponding to the change in transmitted or scattered light over time due to the coagulation of the sample. The A / D conversion unit provided in the optical detection unit converts the optical signal to digital and acquires waveform digital data (hereinafter referred to as "coagulation waveform data") corresponding to the change in transmitted or scattered light over time. In this case, the coagulation waveform data is used, for example, for the analysis of the coagulation ability of a blood sample.

[0031] Next, we will explain an example of the analysis performed by the analysis unit 300 using the data acquired by the measurement unit 400.

[0032] Figure 2 shows an overview of the analysis when the optical detection unit is a flow cytometry-based detection unit.

[0033] In Figure 2, the left-hand diagram shows an overview of the computational processing analysis, and the right-hand diagram shows an overview of the AI ​​analysis. FSC, SSC, and FL in Figure 2 represent the optical signals corresponding to the forward scattered light intensity, side scattered light intensity, and fluorescence acquired by the optical detection unit of the measurement unit 400, respectively.

[0034] The measurement unit 400 identifies regions in the digital data obtained by digitally converting the optical signal that have values ​​greater than a predetermined threshold, as shown in the upper graph of Figure 3, as regions corresponding to analytes in the sample. Regions in the digital data obtained by digitally converting the optical signal that have values ​​greater than a predetermined threshold correspond to each of the analytes in the sample. Each graph in Figure 3 schematically shows the region corresponding to one analyte in the sample identified in the digital data (for example, the "waveform data" region in the upper graph of Figure 3). Note that the identification of regions with values ​​greater than a predetermined threshold may also be performed on the optical signal.

[0035] The measurement unit 400 acquires waveform data from the digital data obtained by digitally converting the optical signal, representing the region corresponding to each analyte in the sample. Waveform data is acquired for multiple analytes in the sample. The analysis unit 300 calculates representative values ​​from the waveform data that correspond to the characteristics of the analytes through computational processing. As shown in the graphs in Figure 3, the analysis unit 300 calculates quantities such as the peak value, width, and area of ​​the waveform data as representative values. The peak value is the maximum value of the waveform data, the width is the width of the waveform data in the time axis direction, and the area is the area enclosed by the waveform data.

[0036] In computational analysis, representative values ​​corresponding to the characteristics of the analyte are predetermined. For example, when classifying and counting blood cells, which are the analyte, the representative value predetermined in the computational analysis algorithm is the peak value. The analysis unit 300 obtains the predetermined representative value from the waveform data through a predetermined calculation and processes the obtained representative value for analyzing the analyte. The analysis unit 300 obtains the predetermined representative value for each of the multiple waveform data obtained by the measurement unit 400. In other words, the same type of representative value (e.g., peak value) is obtained from each of the multiple waveform data through a predetermined calculation by the analysis unit 300. The predetermined representative value may be obtained by the measurement unit 400 and the obtained representative value and waveform data may be transmitted to the analysis unit 300.

[0037] On the other hand, in AI analysis, representative values ​​are not predetermined because the AI ​​algorithm extracts features from the waveform data. The features of the waveform data extracted by the AI ​​algorithm (i.e., the features corresponding to the analyte) can change depending on what the AI ​​algorithm has learned, so there is no need to predetermine representative values ​​in AI analysis. Because the AI ​​algorithm can extract diverse features of the waveform data depending on what it has learned, not only representative values ​​but the waveform data itself is input to the AI ​​algorithm. Because the waveform data itself is input to the AI ​​algorithm, AI analysis places a higher computer load on the data processing and a longer TAT (Turn Around Time) required for the calculations compared to computational analysis.

[0038] As shown in the left-hand diagram of Figure 2, the analysis unit 300 obtains representative values ​​from waveform data acquired in relation to the analyte during computational analysis, and generates a scattergram SC based on these representative values. In the scattergram SC illustrated in Figure 2, the horizontal axis SSCP represents the peak value of the waveform data based on lateral scattered light, and the vertical axis FLP represents the peak value of the waveform data based on fluorescence. Multiple analytes are plotted on the scattergram SC. Based on the scattergram SC, the analysis unit 300 performs classification and analysis of the analytes in the sample.

[0039] As shown in the right-hand diagram of Figure 2, in AI analysis, the analysis unit 300 inputs waveform data corresponding to the analyte into the AI ​​algorithm 60 and performs classification and analysis of the analyte in the sample. The AI ​​algorithm 60 is a pre-trained AI algorithm, which is generated by inputting the above-mentioned waveform data into a pre-training AI algorithm and allowing it to learn. The representative values ​​obtained in computational analysis have a smaller data volume than the waveform data input into the AI ​​algorithm in AI analysis.

[0040] The types of analytes classified by computational processing analysis and AI analysis include, for example, the types of blood cells in a blood sample and the types of formed elements in a urine sample. For example, the analysis unit 300 performs AI analysis on the measurement item that classifies the types of white blood cells in a blood sample, and performs computational processing analysis on the other measurement items.

[0041] Figure 4 shows an overview of the analysis when the optical detection unit is a detection unit that detects transmitted or scattered light from the sample being measured.

[0042] The measurement unit 400 acquires digital data obtained by digitally converting optical signals as coagulation waveform data. In one measurement, one coagulation waveform data is acquired from one sample.

[0043] The graph in Figure 4 shows an example of coagulation waveform data based on transmitted light detected after irradiating a sample with light. The horizontal axis represents elapsed time, and the vertical axis represents absorbance. Absorbance is a value that indicates how much of the light irradiated onto the sample is absorbed by the sample. An absorbance of 0% indicates that almost all of the light irradiated onto the sample reaches the photodetector, while an absorbance of 100% indicates that almost no of the light irradiated onto the sample reaches the photodetector.

[0044] Note that transmitted light intensity may be used instead of absorbance. In this case, if the ratio of the vertical axis (transmitted light intensity) is set to increase as you move upwards, the coagulation waveform data will have a shape that decreases over time, similar to Figure 4.

[0045] The coagulation waveform data includes at least data corresponding to optical signals acquired from timing T2, which indicates the start of coagulation of the sample, to timing T3, which indicates the end of coagulation of the sample. The coagulation waveform data may also include data corresponding to optical signals acquired from the start timing T1 of photometry by the measurement unit 400 to the end timing T4 of photometry.

[0046] In computational analysis, the analysis unit 300 calculates representative values ​​from the coagulation waveform data that correspond to the characteristics of the analyte, and performs the analysis based on the calculated representative values. In computational analysis, the analysis unit 300 identifies the coagulation waveform data when the detected light intensity satisfies predetermined conditions as representative values. For example, the analysis unit 300 obtains the time (T-T2) required for the absorbance of the coagulation waveform data to decrease to a predetermined value (e.g., 50%) as a representative value, and provides the obtained representative value as a result indicating the time it takes for the blood sample to coagulate.

[0047] In AI analysis, the analysis unit 300 analyzes the coagulation waveform data based on the AI ​​algorithm 60 (see Figure 2). For example, the analysis unit 300 obtains whether or not there are any abnormalities in the measurement based on the features extracted from the coagulation waveform data by the AI ​​algorithm 60. Based on whether or not there are any abnormalities in the measurement, the analysis unit 300 determines whether or not there is a suspicion of a nonspecific reaction occurring.

[0048] For example, the analysis unit 300 analyzes whether or not there are abnormalities caused by interfering substances in the blood sample. Specifically, the analysis unit 300 uses coagulation waveform data related to PT (prothrombin time), which is an item for measuring coagulation ability related to prothrombin, a coagulation factor, to analyze whether or not there are abnormalities.

[0049] Furthermore, in AI analysis, the analysis unit 300 may input coagulation waveform data into the AI ​​algorithm 60 to obtain the time it takes for the blood sample to coagulate. Alternatively, in AI analysis, the analysis unit 300 may input coagulation waveform data into the AI ​​algorithm 60 to obtain the cause of any prolongation of the coagulation time.

[0050] Figure 5 is a flowchart showing an example of the sample analysis method according to Embodiment 1.

[0051] In step S1, the measurement unit 400 acquires an optical signal using an optical detection unit and obtains waveform data from the acquired optical signal.

[0052] In step S2, the analysis unit 300 performs AI analysis on the waveform data (first data) that is the target of AI analysis from the waveform data acquired by the measurement unit 400. For example, the analysis unit 300 identifies the waveform data corresponding to the measurement item that is the target of AI analysis as the first data, and performs AI analysis on the identified first data.

[0053] In step S3, the analysis unit 300 performs computational analysis on the waveform data (second data) that is the target of computational analysis from the waveform data acquired by the measurement unit 400. For example, the analysis unit 300 identifies the waveform data corresponding to the measurement item that is the target of computational analysis as the second data, and performs computational analysis on the identified second data.

[0054] In steps S2 and S3 described above, the case where the measurement classifying the types of white blood cells in a blood sample is the target of AI analysis will be explained as an example. The measurement unit 400 prepares a blood sample with reagents corresponding to the measurement of white blood cell classification, and measures the prepared measurement sample with an optical detection unit based on flow cytometry. The analysis unit 300 identifies the waveform data based on the measurement sample for white blood cell classification as the first data, since the measurement related to white blood cell classification is the target of AI analysis. The analysis unit 300 analyzes the first data using the AI ​​algorithm 60 and classifies the white blood cells. On the other hand, the analysis unit 300 identifies the waveform data based on the measurement sample other than white blood cell classification as the second data. The analysis unit 300 identifies representative values ​​corresponding to the characteristics of the analyte from the second data, performs computational analysis to process the identified representative values, and classifies blood cells other than white blood cells.

[0055] In step S4, the analysis unit 300 provides the analysis results obtained in steps S2 and S3. In step S4, for example, the analysis unit 300 displays the analysis results on the display unit or transmits the analysis results to another computer.

[0056] In step S1, the measurement unit 400 may acquire an optical signal from a single measurement sample using an optical detection unit, and then acquire waveform data from the acquired optical signal. In this case, the first data and the second data may each consist of multiple data points, and some of the data may be the same as each other.

[0057] Alternatively, in step S1, an optical detection unit may acquire optical signals from each of multiple measurement samples, including a sample taken from the same subject, and waveform data may be acquired from each of the acquired optical signals. In this case, in step S2, the analysis unit 300 performs AI analysis on the waveform data (first data) acquired from one measurement sample, and in step S3, performs computational analysis on the waveform data (second data) acquired from the other measurement samples. Multiple measurement samples, including a sample taken from the same subject, may be prepared using the same type of reagents, or they may be prepared using different types of reagents.

[0058] Alternatively, in step S1, an optical detection unit may acquire optical signals from each of the multiple measurement samples, including samples collected from different subjects, and waveform data may be acquired from each of the acquired optical signals. In this case, in step S2, the analysis unit 300 performs AI analysis on the waveform data (first data) acquired from one measurement sample, and in step S3, performs computational analysis on the waveform data (second data) acquired from the other measurement samples. The multiple measurement samples, including samples collected from different subjects, may be prepared using the same type of reagents or different types of reagents.

[0059] In the above embodiment 1, as part of the calculation and analysis process, in step S3, the analysis unit 300 identifies representative values ​​corresponding to the characteristics of the analyte from the second data and processes the identified representative values, but the embodiment is not limited to this. For example, in step S1, the measurement unit 400 may acquire representative values ​​from waveform data and output the waveform data and representative values ​​to the analysis unit 300, and in step S3, as part of the calculation and analysis process, the analysis unit 300 may process the representative values ​​acquired from the measurement unit 400.

[0060] [Embodiment 2] In Embodiment 2, AI analysis and computational processing analysis are selected based on rules set in the analysis unit 300.

[0061] The rules for selecting analysis operations are set by the user, for example, via the analysis unit 300. The user can set rules in the analysis unit 300 according to the laboratory's operational policies. This allows for appropriate adjustments to the division of labor between AI analysis and computational analysis, depending on the laboratory's operational policies.

[0062] The ability to set rules for analysis operations allows for flexible adjustments to the division of labor between AI analysis and computational analysis while reducing the load on the analysis unit 300. For example, if the accuracy of AI analysis improves by providing additional training to the AI ​​algorithm 60, rules can be set to increase the amount of data targeted for AI analysis. Also, for example, if prioritizing a shorter TAT (Turn Around Time) for the analysis of measurement results, rules can be set to increase the amount of data targeted for computational analysis.

[0063] Figure 6 is a flowchart showing an example of setting up analysis operations based on rules configured in the analysis unit 300.

[0064] In step S11, the measurement unit 400 acquires an optical signal using an optical detection unit and obtains waveform data from the acquired optical signal. In step S12, the analysis unit 300 refers to a rule for selecting an analysis operation and, based on the referred rule, identifies the waveform data to be subjected to AI analysis and computational processing analysis from the waveform data acquired in step S11.

[0065] In step S13, the analysis unit 300 determines whether the waveform data identified in step S12 contains waveform data subject to AI analysis. If the waveform data subject to AI analysis is included (S12: YES), in step S14, the analysis unit 300 performs AI analysis on the waveform data subject to AI analysis identified in step S12.

[0066] Next, in step S15, the analysis unit 300 determines whether there is any waveform data subject to computational processing analysis in addition to the waveform data analyzed by AI. If there is waveform data subject to computational processing analysis (S15: YES), in step S16, the analysis unit 300 performs computational processing analysis on the waveform data subject to computational processing analysis identified in step S12.

[0067] In step S12, the measurement unit 400 may acquire both waveform data subject to AI analysis and waveform data subject to computational analysis. For example, if the measurement unit 400 performs measurements related to leukocyte classification and reticulocytes according to the measurement order, it acquires waveform data for leukocyte classification and waveform data for reticulocyte measurement. If leukocyte classification is subject to AI analysis and reticulocyte measurement is subject to computational analysis, the analysis unit 300 determines that the waveform data for leukocyte classification, which is subject to AI analysis, is included (S13: YES), and performs AI analysis on the waveform data. Furthermore, the analysis unit 300 determines that the waveform data for reticulocyte classification, which is subject to computational analysis, is also included (S15: YES), and performs computational analysis on the waveform data.

[0068] On the other hand, if the waveform data acquired by the measurement unit 400 does not include waveform data subject to AI analysis (S13:NO), in step S16, the analysis unit 300 performs computational analysis on the waveform data subject to computational analysis identified in step S12. Also, if AI analysis is performed and no waveform data subject to computational analysis is found (S15:NO), computational analysis is not performed, and the process proceeds to step S17.

[0069] In step S17, the analysis unit 300 provides the analysis results.

[0070] The analysis unit 300 may determine in step S13 whether or not waveform data subject to computational analysis is included, and in step S15 whether or not waveform data subject to AI analysis is included. In this case, if the analysis unit 300 determines in step S13 that waveform data subject to computational analysis is included, it performs computational analysis in step S14. Furthermore, if the analysis unit 300 determines in step S15 that waveform data subject to AI analysis is included, it performs AI analysis in step S16.

[0071] [Embodiment 3] Embodiment 3 describes various examples in which AI analysis and computational processing analysis are divided amongst themselves.

[0072] For example, the division of labor between AI analysis and computational analysis is determined by the software program that the analysis unit 300 uses to perform waveform data analysis. The software program of the analysis unit 300 identifies the waveform data to be analyzed by AI and the waveform data to be analyzed by computation, respectively, and performs the analysis. The software program is designed according to requirements related to the inspection (e.g., improving turnaround time, increasing analytical accuracy).

[0073] Figure 7 is a flowchart showing an example of how analysis is performed according to the measurement items.

[0074] In Figure 7, steps S21, S22, and S23 have been added, replacing steps S12, S13, and S15, respectively, compared to Figure 6. The changes from Figure 6 will be explained below.

[0075] In step S21, the analysis unit 300 refers to a rule that includes whether to perform AI analysis or computational processing analysis based on the measurement items, and based on the referred rule, identifies the waveform data of the measurement items to be subjected to AI analysis and the waveform data of the measurement items to be subjected to computational processing analysis, respectively, from the waveform data acquired in step S11.

[0076] Figure 8 is a schematic example showing a screen for setting up AI analysis or computational processing analysis for each measurement item. The measurement items illustrated in Figure 8 relate to a blood cell analyzer.

[0077] The screen shown in Figure 8 is displayed, for example, on the display unit of the analysis unit 300. The screen in Figure 8 has checkboxes for setting up AI analysis and checkboxes for setting up computational analysis for each measurement item. For each measurement item, only one of the checkboxes for AI analysis or computational analysis can be selected. The user operates the checkboxes to select whether to perform AI analysis or computational analysis for each measurement item, and then operates the setting button. As a result, the rule is stored in the memory unit of the analysis unit 300.

[0078] Furthermore, while the user sets either AI analysis or computational processing analysis for each measurement item via the screen shown in Figure 8, the screen may be configured to allow both AI analysis and computational processing analysis to be set. This allows for comparison of the results of AI analysis and computational processing analysis. In addition, the selection of analysis for each measurement item may be set in advance at the time of shipment of the device, or only the administrator may be able to change the settings.

[0079] Returning to Figure 7, in step S22, the analysis unit 300 determines whether the waveform data identified in step S21 includes waveform data of the measurement item to be analyzed using AI. If the waveform data of the measurement item to be analyzed using AI is included (S22: YES), in step S14, the analysis unit 300 performs AI analysis on the waveform data of the measurement item to be analyzed using AI, which was identified in step S21.

[0080] For example, the measurement unit 400 prepares a leukocyte measurement sample by mixing reagents and samples corresponding to a given classification, according to a measurement order that classifies leukocytes (e.g., five classifications: neutrophils, lymphocytes, monocytes, eosinophils, and basophils). The measurement unit 400 acquires an optical signal corresponding to the leukocyte measurement sample using an optical detection unit. The measurement unit 400 acquires waveform data corresponding to the acquired optical signal. If the measurement items related to the classification of leukocytes (e.g., the count and percentage of each of neutrophils, lymphocytes, monocytes, eosinophils, and basophils) are subject to AI analysis, the analysis unit 300 performs AI analysis on the waveform data acquired by the measurement unit 400 through the measurement of the leukocyte measurement sample.

[0081] Next, in step S23, the analysis unit 300 determines whether the waveform data identified in step S12 contains waveform data for the measurement item to be subjected to computational analysis. If there is waveform data to be subjected to computational analysis (S23: YES), in step S16, the analysis unit 300 performs computational analysis on the waveform data identified in step S21 that pertains to the measurement item.

[0082] For example, the measurement unit 400 prepares a reticulocyte measurement sample by mixing reagents and a sample corresponding to a classification, according to a measurement order for classifying reticulocytes. The measurement unit 400 acquires an optical signal corresponding to the reticulocyte measurement sample using an optical detection unit. The measurement unit 400 acquires waveform data corresponding to the acquired optical signal. If the measurement items related to the classification of reticulocytes (e.g., the count and percentage of reticulocytes) are subject to computational processing and analysis, the analysis unit 300 performs computational processing and analysis on the waveform data acquired by the measurement unit 400 through the measurement of the reticulocyte measurement sample.

[0083] Note that the sample analyzer 4000 is not limited to a blood cell analyzer; it may also be a urine analyzer or a blood coagulation analyzer. For example, if the sample analyzer 4000 is a urine analyzer, the analysis unit 300 performs AI analysis on some of the measurement items and computational analysis on the remaining measurement items. If the sample analyzer 4000 is a blood coagulation analyzer, the analysis unit 300 performs computational analysis on all measurement items, and for some measurement items, performs AI analysis in addition to computational analysis to determine if there is a suspicion of a nonspecific reaction occurring.

[0084] Figure 9 is a flowchart showing an example of how analysis is performed according to the measurement order.

[0085] In Figure 9, compared to Figure 6, steps S31 and S32 have been added, replacing steps S12 and S13, respectively, and step S15 has been deleted. The changes from Figure 6 will be explained below.

[0086] In step S31, the analysis unit 300 identifies, based on the measurement order, whether the waveform data acquired in step S11 is waveform data subject to AI analysis or waveform data subject to computational processing analysis. The analysis mode for the measurement order is either the AI ​​analysis mode or the computational processing analysis mode, and is stored in the memory unit of the analysis unit 300 in association with the measurement order.

[0087] Figure 10 is a schematic example showing the screen for setting the analysis mode for a measurement order.

[0088] The screen shown in Figure 10 is displayed, for example, on the display unit of the analysis unit 300. In the screen shown in Figure 10, each row corresponds to a measurement order identified by a sample number. The screen in Figure 10 has checkboxes for setting the AI ​​analysis mode and checkboxes for setting the computational analysis mode for each measurement order. The user operates the checkboxes to select whether to have the analysis unit 300 perform AI analysis or computational analysis for each measurement order, and then operates the setting button. As a result, the analysis mode is stored in the memory unit of the analysis unit 300, corresponding to the measurement order.

[0089] Furthermore, the analysis mode associated with each measurement order is not limited to being set by the user via the analysis unit 300, but may also be set in advance on the host computer or the like when setting the measurement order.

[0090] Returning to Figure 9, in step S32, the analysis unit 300 determines whether the waveform data identified in step S31 is the waveform data of the measurement order to be analyzed by AI. If the identified waveform data is the waveform data of the measurement order to be analyzed by AI (S32: YES), in step S14, the analysis unit 300 performs AI analysis on the waveform data of that measurement order. On the other hand, if the identified waveform data is the waveform data of the measurement order to be analyzed by computational processing (S32: NO), in step S16, the analysis unit 300 performs computational processing analysis on the waveform data of that measurement order.

[0091] Figure 11 is a flowchart showing an example of how analysis is performed depending on the analysis mode of the device.

[0092] In Figure 11, compared to Figure 6, steps S41 and S42 have been added, replacing steps S12 and S13, respectively, and step S15 has been deleted. The changes from Figure 6 will be explained below.

[0093] In step S41, the analysis unit 300 refers to a rule that includes the analysis mode of the analysis unit 300, and based on the referred rule, identifies whether the waveform data acquired in step S11 is waveform data subject to AI analysis or waveform data subject to computational analysis. If the AI ​​analysis mode is set in the above rule, all waveform data will be subject to AI analysis, and if the computational analysis mode is set in the above rule, all data will be subject to computational analysis.

[0094] Figure 12 is a schematic example showing the screen for setting the analysis mode of the analysis unit 300.

[0095] The screen shown in Figure 12 is displayed, for example, on the display unit of the analysis unit 300. The screen in Figure 12 includes checkboxes for setting the AI ​​analysis mode and checkboxes for setting the computational processing analysis mode for the analysis unit 300. The user operates the checkboxes to select whether to have the analysis unit 300 perform AI analysis or computational processing analysis, and then operates the setting button. As a result, the rule is stored in the memory unit of the analysis unit 300.

[0096] Returning to Figure 11, in step S42, the analysis unit 300 determines whether the waveform data identified in step S41 is the waveform data to be analyzed using AI. If the identified waveform data is the data to be analyzed using AI (S42: YES), that is, if the analysis mode of the analysis unit 300 is AI analysis mode, in step S14, the analysis unit 300 performs AI analysis on the waveform data. On the other hand, if the identified waveform data is the waveform data to be analyzed using computational processing (S42: NO), that is, if the analysis mode of the analysis unit 300 is computational processing analysis mode, in step S16, the analysis unit 300 performs computational processing analysis on the waveform data.

[0097] Figure 13 is a flowchart showing an example of how analysis is performed depending on the type of measurement order.

[0098] In Figure 13, compared to Figure 6, steps S51 and S52 have been added, replacing steps S12 and S13, respectively, and step S15 has been deleted. The changes from Figure 6 will be explained below.

[0099] In step S51, the analysis unit 300 refers to a rule that includes an analysis mode corresponding to the type of measurement order, and based on the type of measurement order and the referred rule, identifies whether the waveform data acquired in step S11 is waveform data subject to AI analysis or waveform data subject to computational processing analysis. The types of measurement orders include "Normal," which corresponds to normal measurements such as initial inspections; "Rerun," which corresponds to re-inspections with the same measurement items as the initial inspection; and "Reflex," which corresponds to re-inspections with changed measurement items from the initial inspection. The above rule has either an AI analysis mode or a computational processing analysis mode set for each type of measurement order.

[0100] Figure 14 is a schematic example showing the screen for setting the analysis mode for each type of measurement order.

[0101] The screen shown in Figure 14 is displayed, for example, on the display unit of the analysis unit 300. The screen in Figure 14 includes checkboxes for setting the AI ​​analysis mode and the computational processing analysis mode for each type of measurement order (Normal, Rerun, Reflex). The user operates the checkboxes to select whether to perform AI analysis or computational processing analysis for each type of measurement order, and then operates the setting button. As a result, the rule is stored in the memory unit of the analysis unit 300.

[0102] Furthermore, the analysis mode associated with each type of measurement order is not limited to being set by the user via the analysis unit 300, but may also be pre-set by a host computer or the like according to the type of measurement order.

[0103] Returning to Figure 13, in step S52, the analysis unit 300 determines whether the waveform data identified in step S51 is the waveform data to be analyzed using AI. If the identified waveform data is the waveform data to be analyzed using AI (S52: YES), that is, if the analysis mode corresponding to the type of measurement order is the AI ​​analysis mode, in step S14, the analysis unit 300 performs AI analysis on the waveform data. On the other hand, if the identified waveform data is the waveform data to be analyzed using computational processing (S52: NO), that is, if the analysis mode corresponding to the type of measurement order is the computational processing analysis mode, in step S16, the analysis unit 300 performs computational processing analysis on the waveform data.

[0104] Figure 15 is a flowchart showing an example of how analysis is performed depending on the type of measurement item and measurement order.

[0105] In Figure 15, step S61 has been added in place of step S12, compared to Figure 6. The changes from Figure 6 will be explained below.

[0106] In step S61, the analysis unit 300 refers to the rules for selecting the analysis operation and, based on the measurement item and the type of measurement order, identifies the waveform data to be subjected to AI analysis and the waveform data to be subjected to computational processing analysis from the waveform data acquired in step S11.

[0107] Figure 16 is a schematic example diagram showing the screen for setting up AI analysis or computational processing analysis for each measurement item and measurement order type.

[0108] The screen in Figure 16 is displayed, for example, on the display unit of the analysis unit 300. Similar to Figure 8, the screen in Figure 16 has checkboxes for setting either AI analysis or computational processing analysis for each measurement item, and checkboxes for setting only computational processing analysis, similar to Figure 14, and checkboxes for setting either AI analysis or computational processing analysis for each measurement order type (Normal, Rerun, Reflex). The user operates the checkboxes in the upper list to select whether to perform AI analysis or computational processing analysis for each measurement item, and operates the checkboxes in the lower list to select whether to perform AI analysis or computational processing analysis for each measurement order type, and then operates the setting button. As a result, the rules are stored in the memory unit of the analysis unit 300.

[0109] As shown in Figure 16, once the settings are configured, if the measurement order type is "Normal," the analysis unit 300 identifies the waveform data acquired by the measurement unit 400 as the target of computational processing analysis. For example, if the measurement order type is "Normal," computational processing analysis is performed for all measurement items based on that measurement order, regardless of the analysis settings for each measurement item. Also, if the measurement order type is "Rerun" or "Reflex," the analysis unit 300 sets the waveform data acquired by the measurement unit 400 as the target of AI analysis or computational processing analysis according to the analysis settings configured for each measurement item. For example, if the measurement order type is "Rerun" or "Reflex," the measurement items related to nucleated red blood cells (NRBCs) and basophils (BASOs) will be subject to AI analysis, and the other measurement items will be subject to computational processing analysis.

[0110] Returning to Figure 15, in step S13, the analysis unit 300 determines whether the waveform data identified in step S61 contains data subject to AI analysis. If there is waveform data subject to AI analysis (S13: YES), in step S14, the analysis unit 300 performs AI analysis on the waveform data subject to AI analysis identified in step S61.

[0111] Next, in step S15, the analysis unit 300 determines whether the waveform data identified in step S61 includes waveform data subject to computational analysis. If there is waveform data subject to computational analysis (S15: YES), in step S16, the analysis unit 300 performs computational analysis on the waveform data identified in step S61.

[0112] Figure 17 is a flowchart showing an example where the necessity of AI analysis is determined based on flags generated by computational processing analysis.

[0113] In Figure 17, compared to Figure 6, steps S71 to S74 have been added after step S11, and steps S12 to S16 have been deleted. The changes from Figure 6 will be explained below.

[0114] In step S71, the analysis unit 300 performs computational analysis on the waveform data acquired in step S11 and sets flags that indicate abnormalities in the analytes in the sample based on the results of the computational analysis. These flags may include, for example, a flag indicating the detection of a predetermined abnormal cell or a flag indicating that the count value of a predetermined blood cell is abnormal. In step S72, the analysis unit 300 refers to a rule that includes whether or not to perform AI analysis on the analysis results of the flags.

[0115] Figure 18 is a schematic example illustrating the screen for setting up AI analysis for each flag in the analysis results.

[0116] The screen shown in Figure 18 is displayed, for example, on the display unit of the analysis unit 300. The screen in Figure 18 includes checkboxes for setting AI analysis for each flag assigned to the analysis results from the computational processing analysis. If the sample analyzer 4000 is a blood cell analyzer, the flags assigned to the analysis results from the computational processing analysis include a decrease or increase in blood cells, and the appearance of abnormal cells. The user selects whether to perform AI analysis for each flag by operating the checkboxes and then operates the setting button. As a result, the rules are stored in the memory unit of the analysis unit 300.

[0117] When a checkbox for a flag is checked, AI analysis is performed on the waveform data corresponding to that analysis result. In the example shown in Figure 18, the checkboxes for the analysis results of blast cells / abnormal lymphocytes, blast cells, abnormal lymphocytes, and atypical lymphocytes are checked, and if the computational analysis sets a flag indicating the presence of these blood cells, AI analysis is performed on these blood cells.

[0118] Returning to Figure 17, in step S73, the analysis unit 300 determines whether the sample is subject to AI analysis based on the flag assigned to the analysis result obtained in step S71 and the rule referenced in step S72. If the sample is subject to AI analysis (S73: YES), in step S74, the analysis unit 300 performs AI analysis on each waveform data. For example, if a flag indicating the detection of blast cells is generated in the computational analysis, AI analysis is performed on each waveform data obtained in step S11, according to the rule illustrated in Figure 18.

[0119] On the other hand, if the sample is not subject to AI analysis (S73:NO), the analysis unit 300 skips step S74.

[0120] Figure 19 is a flowchart showing an example of performing AI analysis on a specific analyte classified by computational processing analysis.

[0121] In Figure 19, compared to Figure 6, steps S81 to S84 have been added after step S11, and steps S12 to S16 have been deleted. The changes from Figure 6 will be explained below.

[0122] In step S81, the analysis unit 300 performs computational processing analysis on the waveform data acquired in step S11 and classifies the analytes. In step S82, the analysis unit 300 refers to rules including whether or not to perform AI analysis for each type of analyte.

[0123] Figure 20 is a schematic example diagram showing a screen for setting whether or not to perform AI analysis for each type of analyte.

[0124] The screen shown in Figure 20 is displayed, for example, on the display unit of the analysis unit 300. The screen in Figure 20 includes checkboxes for setting AI analysis for each type of analyte. If the sample analyzer 4000 is a blood cell analyzer, the types classified by computational analysis include eosinophils, neutrophils, lymphocytes, and monocytes. The user selects whether or not to perform AI analysis for each type of analyte by operating the checkboxes and then operates the setting button. This stores the rules in the memory unit of the analysis unit 300.

[0125] When a checkbox for a type of analyte is checked, AI analysis is performed on the waveform data classified under that type. In the example shown in Figure 20, AI analysis is performed on monocytes and lymphocytes.

[0126] Returning to Figure 19, in step S83, the analysis unit 300 identifies waveform data corresponding to analytes classified into a specific category (for example, monocytes and lymphocytes in the case of the rules shown in Figure 20) based on the type of analyte classified in step S81 and the rules referenced in step S82. In step S84, the analysis unit 300 performs AI analysis on the identified waveform data.

[0127] Figure 21 illustrates the classification method using computational processing analysis and AI analysis, which is performed in the process shown in Figure 19.

[0128] In the computational analysis, a scattergram is used, with two types of representative values ​​calculated from the waveform data as axes. For example, as shown in Figure 20, if the system is set to perform AI analysis on monocytes and lymphocytes, in step S81, the computational analysis identifies the plotted areas enclosed by dashed lines on the scattergram that correspond to monocytes and lymphocytes. Then, in step S83, the waveform data corresponding to the plotted areas is identified, and in step S84, AI analysis is performed on the identified waveform data.

[0129] Returning to Figure 19, in step S17, the analysis unit 300 provides the analysis results from computational processing analysis and AI analysis. At this time, the analysis unit 300 replaces the analysis results obtained by computational processing analysis in step S81, specifically those of the type that were the target of AI analysis, with the analysis results obtained by AI analysis. Alternatively, the analysis results from computational processing analysis and the analysis results from AI analysis may be provided together.

[0130] Figure 22 is a flowchart showing an example of performing AI analysis when a specific classification is made in computational processing analysis.

[0131] In Figure 22, compared to Figure 6, steps S91 to S95 are added after step S11, and steps S12 to S16 are deleted. The changes from Figure 6 will be explained below.

[0132] In step S91, the analysis unit 300 performs computational analysis on the waveform data acquired in step S11 and classifies the analytes. In step S92, the analysis unit 300 refers to rules including whether or not to perform AI analysis on certain types of analytes, such as cells that are not present in the peripheral blood of healthy individuals.

[0133] Figure 23 is a schematic example showing a screen for setting whether or not to perform AI analysis on a specific type of analyte.

[0134] The screen shown in Figure 23 is displayed, for example, on the display unit of the analysis unit 300. The screen in Figure 23 includes checkboxes for setting AI analysis for each type of specific analyte. If the sample analyzer 4000 is a hematology analyzer, the types classified by the computational analysis may include blast cells, abnormal lymphocytes, atypical lymphocytes, and immature granulocytes. The user selects whether or not to perform AI analysis for each type of specific analyte by operating the checkboxes and then operates the setting button. This stores the rules in the memory unit of the analysis unit 300.

[0135] When a checkbox for a specific type of analyte is selected, AI analysis is performed on the waveform data classified under that type. In the example shown in Figure 23, AI analysis is performed on blast cells, abnormal lymphocytes, and atypical lymphocytes.

[0136] Returning to Figure 22, in step S93, the analysis unit 300 determines whether an analyte classified into a specific category (for example, blast cells, abnormal lymphocytes, and atypical lymphocytes in the case of the rules shown in Figure 23) has been detected, based on the type of analyte classified in step S91 and the rule referenced in step S92. If an analyte classified into a specific category has been detected (S93: YES), in step S94, the analysis unit 300 identifies waveform data corresponding to the analyte classified into the specific category. In step S95, the analysis unit 300 performs AI analysis on the identified waveform data.

[0137] In step S17, the analysis unit 300 provides the analysis results obtained from computational processing analysis and AI analysis. At this time, the analysis unit 300 replaces the analysis results obtained from computational processing analysis in step S91 that were of the type targeted for AI analysis with the analysis results obtained from AI analysis and provides them. Alternatively, the analysis results obtained from computational processing analysis and the analysis results obtained from AI analysis may be provided together.

[0138] [Embodiment 4] Embodiment 4 shows a detailed configuration example in which computational processing analysis and AI analysis are performed separately in a sample analyzer 4000 that analyzes a sample based on flow cytometry.

[0139] Examples of samples measured by the sample analyzer 4000 of Embodiment 4 include biological samples collected from a subject. These samples may include, for example, peripheral blood such as venous blood and arterial blood, urine, and other bodily fluids. Other bodily fluids may include, for example, bone marrow fluid, ascites, pleural fluid, and cerebrospinal fluid. Hereinafter, other bodily fluids may simply be referred to as "bodily fluids." Blood samples are not limited as long as they allow for cell counting and determination of cell types. Preferably, the blood is peripheral blood. For example, the blood may be peripheral blood collected using an anticoagulant such as ethylenediaminetetraacetate sodium salt or potassium salt, or heparin sodium. Peripheral blood may be collected from an artery or a vein.

[0140] The cell types to be determined in this embodiment are based on morphological classification and vary depending on the type of sample. When the sample is blood and the blood is collected from a healthy person, the cell types to be determined in this embodiment include, for example, nucleated cells such as nucleated red blood cells and white blood cells, as well as red blood cells, platelets, etc. Nucleated cells include, for example, neutrophils, lymphocytes, monocytes, eosinophils, and basophils. Neutrophils include, for example, segmented neutrophils and band neutrophils. When the blood is collected from a non-healthy person, nucleated cells may include, for example, at least one selected from the group consisting of immature granulocytes and abnormal cells. Such cells are also included in the cell types to be determined in this embodiment. Immature granulocytes may include, for example, metamyelocytes, myelocytes, promyelocytes, myeloblasts, and other cells.

[0141] Furthermore, nucleated cells may include not only normal cells but also abnormal cells not found in the peripheral blood of healthy individuals. Examples of abnormal cells are cells that appear when a person suffers from a specified disease, such as tumor cells. In the case of the hematopoietic system, the specified disease may be selected from the group consisting of, for example, myelodysplastic syndrome, acute myeloblastic leukemia, acute promyelocytic leukemia, acute myelomonocytic leukemia, acute monocytic leukemia, erythroleukemia, acute megakaryoblastic leukemia, acute myeloid leukemia, acute lymphoblastic leukemia, lymphoblastic leukemia, chronic myeloid leukemia, or leukemia such as chronic lymphoblastic leukemia, malignant lymphomas such as Hodgkin lymphoma and non-Hodgkin lymphoma, and multiple myeloma.

[0142] Furthermore, abnormal cells may include, for example, erythroblasts, which are nucleated red blood cells such as lymphoblasts, plasma cells, atypical lymphocytes, reactive lymphocytes, proerythroblasts, basophilic erythroblasts, polychromatic erythroblasts, orthochromatic erythroblasts, promegaloblasts, basophilic megaloblasts, polychromatic megaloblasts, and orthochromatic megaloblasts, as well as megakaryocytes, which include micromegacaryocytes, and other cells not normally found in the peripheral blood of healthy individuals.

[0143] Furthermore, if the sample is urine, the cell types to be determined in this embodiment may include, for example, epithelial cells such as transitional epithelium and squamous epithelium, red blood cells, white blood cells, etc. Abnormal cells may include, for example, bacteria, filamentous fungi, fungi such as yeast, tumor cells, etc.

[0144] When the sample is a body fluid that does not normally contain blood components, such as ascites, pleural fluid, or cerebrospinal fluid, the cell types may include, for example, red blood cells, white blood cells, and large cells. The large cells mentioned here refer to cells that are larger than white blood cells and have been detached from the endometrium of the body cavity or the peritoneum of internal organs, such as mesothelial cells, histiocytes, and tumor cells.

[0145] When the sample is bone marrow fluid, the cell types to be determined in this embodiment may include mature hematopoietic cells and immature hematopoietic cells as normal cells. Mature hematopoietic cells include, for example, nucleated cells such as nucleated red blood cells and white blood cells, as well as red blood cells and platelets. Nucleated cells such as white blood cells include, for example, neutrophils, lymphocytes, plasma cells, monocytes, eosinophils, and basophils. Neutrophils include, for example, segmented neutrophils and band neutrophils. Immature hematopoietic cells include, for example, hematopoietic stem cells, immature granulocytes, immature lymphocytes, immature monocytes, immature red blood cells, megakaryocytes, and mesenchymal cells. Immature granulocytes may include, for example, metamyelocytes, myelocytes, promyelocytes, and myeloblasts. Immature lymphocytes include, for example, lymphoblasts. Immature monocytes include monoblasts, etc. Immature erythrocytes include nucleated erythrocytes such as proerythroblasts, basophilic erythroblasts, polychromatic erythroblasts, orthochromatic erythroblasts, promegaloblasts, basophilic megaloblasts, polychromatic megaloblasts, and orthochromatic megaloblasts. Megakaryocytic cells include megakaryoblasts, etc.

[0146] Abnormal cells that may be found in the bone marrow include, for example, hematopoietic tumor cells selected from the group consisting of leukemias such as myelodysplastic syndrome, acute myeloblastic leukemia, acute promyelocytic leukemia, acute myelomonocytic leukemia, acute monocytic leukemia, erythroleukemia, acute megakaryoblastic leukemia, acute myeloid leukemia, acute lymphoblastic leukemia, lymphoblastic leukemia, chronic myeloid leukemia, or chronic lymphoblastic leukemia, malignant lymphomas such as Hodgkin lymphoma and non-Hodgkin lymphoma, and multiple myeloma, as well as metastatic tumor cells of malignant tumors that have occurred in organs other than the bone marrow.

[0147] The signals obtained from analytes (e.g., cells and formed elements) in the sample exemplified above include analog optical signals such as forward scatter, side scatter, and fluorescence signals obtained by irradiating cells flowing through a flow cell with light. However, there are no particular limitations as long as the signals represent the characteristics of the analytes and allow for the classification of analytes by type.

[0148] The optical signal is preferably a light signal obtained as an optical response by irradiating a cell with light. The optical signal may include at least one selected from signals based on light scattering, signals based on light absorption, signals based on transmitted light, and signals based on fluorescence.

[0149] Signals based on light scattering may include scattered light signals and light loss signals resulting from light irradiation. Scattered light signals represent the characteristics of the analyte in the sample, depending on the angle of reception of the scattered light relative to the direction of propagation of the irradiated light. Forward scattered light signals are used to calculate representative values ​​representing the size of the analyte. Side scattered light signals, when the analyte in the sample is a cell, are used to calculate representative values ​​representing the complexity of the cell nucleus.

[0150] In forward-scattered light, "forward" refers to the direction of propagation of light emitted from the light source. "Forward" may include forward low angles, where the receiving angle is approximately 0° to 5°, and / or forward high angles, where the receiving angle is approximately 5° to 20°, when the angle of the irradiated light is set to 0°. "Sideways" is not limited as long as it does not overlap with "forward". "Sideways" may include receiving angles approximately 25° to 155°, preferably 45° to 135°, and more preferably 90°, when the angle of the irradiated light is set to 0°. Fluorescence in this embodiment is detected at the same receiving angles as side-scattered light.

[0151] Signals based on light scattering may include polarization or depolarization as signal components. For example, by irradiating an analyte in a sample with light and receiving the scattered light through a polarizing plate, only scattered light polarized at a specific angle can be received. Alternatively, by irradiating an analyte in a sample with light through a polarizing plate and receiving the resulting scattered light through a polarizing plate that transmits only polarization at a different angle than the irradiating polarizing plate, only depolarized scattered light can be received.

[0152] The optical loss signal represents the amount of light loss, based on the decrease in the amount of light received at the light-receiving unit due to scattering of light when it is irradiated onto the analyte. Preferably, the optical loss signal is obtained as optical loss in the optical axis direction of the irradiated light (axial optical loss). The optical loss signal can be expressed as the ratio of the amount of light received when the sample is flowing through the flow cell to the amount of light received when the sample is flowing through the flow cell, with the amount of light received at the light-receiving unit being 100% when the sample is not flowing through the flow cell. Axial optical loss, like the forward scattered light signal, is used to calculate a representative value representing the size of the analyte, but the signal obtained will differ depending on whether the cells are translucent or not.

[0153] The fluorescence-based signal may be fluorescence excited by irradiating an analyte labeled with a fluorescent substance with light, or it may be autofluorescence generated from an unstained analyte. If the analyte in the sample is a cell, the fluorescent substance may be a fluorescent dye that binds to nucleic acids or membrane proteins, or it may be a labeled antibody that has been modified with a fluorescent dye to bind to a specific protein in the cell.

[0154] The optical signal may be acquired in the form of image data obtained by irradiating the analyte in the sample with light and imaging the irradiated analyte. Image data can be obtained by imaging individual analytes flowing through the flow cell channel with an image sensor such as a TDI camera or a CCD camera. Alternatively, image data of cells may be obtained by coating, scattering, or dotting a sample containing cells or a measurement sample onto a glass slide and imaging the glass slide with an image sensor.

[0155] The signal obtained from the analyte in the sample is not limited to an optical signal, but may also be an electrical signal obtained from cells. For example, the electrical signal may be obtained by applying a DC current to a flow cell and using the change in impedance caused by the analyte flowing through the flow cell as the electrical signal. The electrical signal obtained in this way is used to calculate a representative value that reflects the volume of the analyte. Alternatively, the electrical signal may be the change in impedance when a radio frequency is applied to the analyte flowing through the flow cell. The electrical signal obtained in this way is used to calculate a representative value that reflects the conductivity of the analyte.

[0156] The signals obtained from the analytes in the sample may be a combination of at least two of the signals described above. By combining multiple signals, the characteristics of the analytes can be analyzed from multiple perspectives, enabling more accurate cell classification. The combination may be at least two of multiple optical signals, such as forward scattered light signals, side scattered light signals, and fluorescence signals, or scattered light signals at different angles, such as low-angle scattered light signals and high-angle scattered light signals. Alternatively, optical signals and electrical signals may be combined, and there are no particular limitations on the type and number of signals to be combined.

[0157] In this embodiment, the AI ​​algorithm 60 used for AI analysis is, for example, a deep learning algorithm. A deep learning algorithm is one type of artificial intelligence algorithm and consists of a neural network including multiple hidden layers. Data input to the neural network is processed by a large amount of matrix operations. Waveform data corresponding to each analyte is obtained from the digital data acquired by A / D conversion of the analog optical signal exemplified in Figure 2 above, and the acquired waveform data is input to the AI ​​algorithm 60 for analysis. For example, the AI ​​algorithm 60 classifies the type of analyte corresponding to the input waveform data.

[0158] In this embodiment, the type of analyte in the sample is not limited to being classified by the AI ​​algorithm 60. For each analyte passing through a predetermined position in the flow path, the signal intensity may be acquired at multiple time points while the analyte is passing through the predetermined position, and the type of each analyte may be determined based on the result of recognizing the signal intensity at multiple time points for each acquired analyte as a pattern. The pattern may be recognized as a numerical pattern of signal intensity at multiple time points, or as a shape pattern when the signal intensity at multiple time points is plotted as a graph. When recognized as a numerical pattern, the type of analyte can be determined by comparing the numerical pattern of the analyte with a numerical pattern of a known type. For example, Spearman's rank correlation, z-score, etc., can be used to compare the numerical pattern of the analyte with a control numerical pattern. The type of analyte can be determined by comparing the graph shape pattern of the analyte with a graph shape pattern of a known type. The comparison between the graph shape pattern of the analyte and graph shape patterns of known types can be performed, for example, using geometric shape pattern matching or using feature descriptors such as SIFT Descriptor.

[0159] (Example configuration) This section describes an example configuration of a sample analyzer 4000 when the measurement unit 400 is equipped with an FCM detection unit (a detection unit based on flow cytometry) for measuring a sample (for example, a blood sample, a urine sample, body fluid, or bone marrow fluid).

[0160] Figure 24 is a block diagram showing the configuration of the measurement unit 400.

[0161] As shown in Figure 24, the measurement unit 400 includes an FCM detection unit 410 for detecting analytes in the sample, an analog processing unit 420 for processing the analog optical signal output from the FCM detection unit 410, a device mechanism unit 430, a sample preparation unit 440, a sample aspiration unit 450, and a measurement unit control unit 460.

[0162] The sample aspiration unit 450, for example, aspirates a sample from a sample container and discharges the aspirated sample into a reaction vessel (e.g., a reaction chamber, a reaction cuvette). The sample preparation unit 440, for example, aspirates reagents for preparing a measurement sample and discharges the reagents into a reaction vessel containing the sample. The measurement sample is prepared by mixing the sample and reagents in the reaction vessel. The device mechanism unit 430 includes the mechanism within the measurement unit 400.

[0163] Figure 25 is a schematic diagram showing the configuration of the optical system of the FCM detection unit 410.

[0164] Light emitted from the light source 4111 is irradiated onto the analyte in the sample being measured as it passes through the flow cell (sheath flow cell) 4113 via the irradiation lens system 4112. As a result, scattered light and fluorescence are emitted from the analyte flowing through the flow cell 4113.

[0165] The wavelength of light emitted from the light source 4111 is not particularly limited, and a wavelength suitable for exciting the fluorescent dye is selected. Examples of light sources used for the light source 4111 include semiconductor laser light sources, argon laser light sources, gas laser light sources such as helium-neon lasers, and mercury arc lamps. Semiconductor laser light sources are particularly preferred because they are significantly less expensive than gas laser light sources.

[0166] The forward scattered light generated from the analyte in the flow cell 4113 is received by the photodetector 4116 via the focusing lens 4114 and the pinhole section 4115. The photodetector 4116 is, for example, a photodiode. The side scattered light generated from the analyte in the flow cell 4113 is received by the photodetector 4121 via the focusing lens 4117, the dichroic mirror 4118, the bandpass filter 4119, and the pinhole section 4120. The photodetector 4121 is, for example, a photodiode. The fluorescence generated from the analyte in the flow cell 4113 is received by the photodetector 4122 via the focusing lens 4117 and the dichroic mirror 4118. The photodetector 4122 is, for example, an avalanche photodiode. Photomultiplier tubes may be used as photodetectors 4116, 4121, and 4122.

[0167] The analog light-receiving signals (optical signals) output from each of the light-receiving elements 4116, 4121, and 4122 are input to the analog processing unit 420 via amplifiers 4151, 4152, and 4153, respectively.

[0168] The analog processing unit 420 performs processing such as noise reduction and smoothing on the optical signal input from the FCM detection unit 410, and outputs the processed optical signal to the A / D conversion unit 461.

[0169] Returning to Figure 24, the measurement unit control unit 460 comprises an A / D conversion unit 461, an IF (interface) unit 462, a bus 463, and IF units 464 and 465.

[0170] The A / D conversion unit 461 converts the analog optical signals output from the analog processing unit 420, from the start to the end of measurement of the sample, into digital data. When multiple types of optical signals (for example, optical signals corresponding to forward scattered light intensity, side scattered light intensity, and fluorescence intensity, respectively) are generated from a single sample, the A / D conversion unit 461 converts each optical signal from the start to the end of measurement into digital data. For example, as shown in Figure 25, three types of optical signals (forward scattered light signal, side scattered light signal, and fluorescence signal) are input to the A / D conversion unit 461 via multiple corresponding signal transmission paths 420a. The A / D conversion unit 461 converts each of the optical signals input from the multiple signal transmission paths 420a into digital data. Each signal transmission path 420a is configured, for example, to transmit analog optical signals as differential signals.

[0171] The A / D conversion unit 461 compares the signal level of the optical signal with a predetermined threshold and samples the optical signal having a signal level greater than the threshold. The A / D conversion unit 461 samples the analog optical signal at a predetermined sampling rate (for example, sampling 1024 points at 10 nanosecond intervals, sampling 128 points at 80 nanosecond intervals, or sampling 64 points at 160 nanosecond intervals). The A / D conversion unit 461 generates digital data (waveform data) of the forward scattered light signal, digital data (waveform data) of the side scattered light signal, and digital data (waveform data) of the fluorescence signal for each analyte by performing sampling processing on three types of optical signals corresponding to each analyte. Each digital data (waveform data) corresponds to each analyte in the sample.

[0172] The A / D conversion unit 461 assigns an index to each of the generated waveform data. The generated waveform data is, for example, digital data corresponding to each of the N analytes contained in a single sample. As a result, three types of waveform data are generated for each analyte, corresponding to three types of optical signals (forward scattered light signal, side scattered light signal, and fluorescence signal).

[0173] The waveform data generated by the A / D conversion unit 461 is transmitted to the analysis unit 300 via the IF units 462 and 465 and the bus 463. The device mechanism unit 430, the sample preparation unit 440, and the sample aspiration unit 450 are controlled by the analysis unit 300 via the IF units 464 and 465 and the bus 463.

[0174] Figure 26 is a block diagram showing the configuration of the analysis unit 300.

[0175] The analysis unit 300 comprises a processor 3001, RAM 3017, bus 3003, storage unit 3004, IF unit 3006, display unit 3011, and operation unit 3012. The analysis unit 300 is configured, for example, by a personal computer. The analysis unit 300 is connected to the measurement unit 400 via the IF unit 3006.

[0176] The processor 3001 is composed of, for example, a CPU. The processor 3001 executes programs loaded from the storage unit 3004 into the RAM 3017. The RAM 3017 is the so-called main memory. The processor 3001 analyzes waveform data acquired by the measurement unit 400 by executing an analysis program. The processor 3001 controls the analysis unit 300 and the measurement unit 400 by executing a control program.

[0177] The storage unit 3004 is composed of, for example, a hard disk drive (HDD) or a solid-state drive (SSD). The storage unit 3004 stores waveform data received from the measurement unit 400, a program for controlling the analysis unit 300 and the measurement unit 400, and a program for analyzing the waveform data. The program for analyzing the waveform data is configured to analyze the waveform data based on the computational analysis and AI analysis described above. The storage unit 3004 also stores rules for identifying the waveform data that will be the target of the AI ​​analysis and computational analysis, respectively, and rules for selecting the analysis operation.

[0178] The display unit 3011 is configured, for example, as a liquid crystal display. The display unit 3011 is connected to the processor 3001 via the bus 3003 and the IF unit 3006. The display unit 3011 displays, for example, the analysis results acquired by the measurement unit 400.

[0179] The control unit 3012 consists of, for example, a keyboard, mouse, and a pointing device including a touch panel. Users such as physicians and laboratory technicians can input measurement orders and measurement instructions based on those orders into the sample analyzer 4000 by operating the control unit 3012. Users can also input instructions to display analysis results by operating the control unit 3012. Analysis results include, for example, numerical results based on the analysis, graphs, charts, and flag information assigned to the sample.

[0180] Figure 27 is a block diagram showing the configuration of the measurement unit 400 when the sample analyzer 4000 performs counting and classification of blood cells in a blood sample.

[0181] The measurement unit 400 in Figure 27, in addition to the configuration in Figure 24, further includes an RBC / PLT detection unit 4101, an HGB detection unit 4102, analog processing units 4201 and 4202, and A / D conversion units 4611 and 4612.

[0182] The RBC / PLT detection unit 4101 is an electrical resistance type detection unit that measures blood cells based on the RBC / PLT measurement sample using the sheath flow DC detection method. The HGB detection unit 4102 measures hemoglobin based on the hemoglobin measurement sample using the SLS-hemoglobin method. The data obtained by A / D conversion of the analog signals acquired from the RBC / PLT detection unit 4101 and the HGB detection unit 4102 are subject to computational processing and analysis. Based on the data from the RBC / PLT detection unit 4101, the red blood cells and platelets in the blood sample are counted. Based on the data from the HGB detection unit 4102, the amount of hemoglobin in the blood sample is obtained.

[0183] Furthermore, the data obtained by A / D conversion of the analog signals acquired from the RBC / PLT detection unit 4101 and the HGB detection unit 4102 may also be subject to AI analysis. In addition, AI analysis and computational processing analysis may be used interchangeably for the data based on the RBC / PLT detection unit 4101 and the HGB detection unit 4102. This reduces the load on the analysis unit 300 that processes the data.

[0184] Figure 28 is a block diagram showing the configuration of the sample aspiration unit 450 and the sample preparation unit 440 in the measurement unit 400 shown in Figure 27.

[0185] The sample aspiration unit 450 includes a nozzle 451 for aspirating a blood sample (e.g., whole blood) from a blood collection tube TB, and a pump 452 for applying negative and positive pressure to the nozzle. The nozzle 451 is inserted into the blood collection tube TB by being moved up and down by the device mechanism unit 430 (see Figure 27). When the pump 452 applies negative pressure with the nozzle 451 inserted into the blood collection tube TB, the blood sample is aspirated through the nozzle 451. The device mechanism unit 430 may also include a hand member for inverting and agitating the blood collection tube TB before aspirating blood from it.

[0186] The sample preparation unit 440 comprises a WDF sample preparation unit 440a, a RET sample preparation unit 440b, a WPC sample preparation unit 440c, a PLT-F sample preparation unit 440d, and a WNR sample preparation unit 440e. Each of the sample preparation units 440a to 440e is equipped with a reaction chamber for mixing the sample with reagents (e.g., hemolytic agents and staining solutions). The sample preparation units 440a to 440e are used in the WDF channel, RET channel, WPC channel, PLT-F channel, and WNR channel, respectively.

[0187] Here, the sample analyzer 4000 is equipped with multiple measurement channels corresponding to each of the multiple types of measurement samples to be prepared. The sample analyzer 4000 is equipped with, for example, a WDF channel, a RET channel, a WPC channel, a PLT-F channel, and a WNR channel. The WDF channel is for detecting neutrophils, lymphocytes, monocytes, and eosinophils. The RET channel is for detecting reticulocytes. The WPC channel is for detecting blast cells and abnormal lymphoid cells. The PLT-F channel is for detecting platelets. The WNR channel is for detecting leukocytes other than basophils, basophils, and nucleated erythrocytes.

[0188] In the sample preparation units 440a to 440e, a hemolytic agent container containing a hemolytic agent, which is a reagent corresponding to the measurement channel, and a staining solution container containing a staining solution are connected via a flow path. For example, in the WDF sample preparation unit 440a, a hemolytic agent container containing a WDF hemolytic agent (e.g., LyzaCell WDF II; manufactured by Sysmex Corporation), which is a reagent for WDF measurement, and a staining solution container containing a WDF staining solution (e.g., FluoroCell WDF; manufactured by Sysmex Corporation) are connected via a flow path. Here, a configuration in which one sample preparation unit is connected to both the hemolytic agent container and the staining solution container is illustrated, but one sample preparation unit does not necessarily have to be connected to both the hemolytic agent container and the staining solution container, and one reagent container may be shared by multiple sample preparation units. Furthermore, the sample preparation unit and the reagent container do not need to be connected by a flow path; the reagent may be aspirated from the reagent container by a nozzle, the nozzle moves, and the aspirated reagent is discharged from the nozzle into the reaction chamber of the sample preparation unit.

[0189] The nozzle 451, which has aspirated the blood sample, is positioned above the reaction chamber of the sample preparation unit 440a to 440e corresponding to the measurement order by horizontal and vertical movement by the device mechanism unit 430. In this state, when the pump 452 applies positive pressure, the blood sample is discharged from the nozzle 451 into the corresponding reaction chamber. The sample preparation unit 440 supplies the hemolytic agent and staining solution corresponding to the reaction chamber from which the blood sample was discharged, and prepares the measurement sample by mixing the blood sample, hemolytic agent, and staining solution in the reaction chamber.

[0190] In the WDF sample preparation unit 440a, WDF measurement samples are prepared; in the RET sample preparation unit 440b, RET measurement samples are prepared; in the WPC sample preparation unit 440c, WPC measurement samples are prepared; in the PLT-F sample preparation unit 440d, PLT-F measurement samples are prepared; and in the WNR sample preparation unit 440e, WNR measurement samples are prepared. The prepared measurement samples are supplied from the reaction chamber to the FCM detection unit 410 via a flow channel, where cell measurement is performed using the flow cytometry method.

[0191] The measurement channels mentioned above (WDF, RET, WPC, PLT-F, WNR) correspond to the measurement items included in the measurement order. For example, the WDF channel corresponds to measurement items related to leukocyte classification, the RET channel corresponds to measurement items related to reticulocytes, the PLT-F channel corresponds to measurement items related to platelets, and the WNR channel corresponds to measurement items related to leukocyte count and nucleated erythrocytes. The measurement samples prepared in the above measurement channels are measured by the FCM detection unit 410.

[0192] The measurement results from the RBC / PLT detection unit 4101 correspond to the measurement items related to red blood cell count. The measurement results from the HGB detection unit 4102 correspond to the measurement items related to hemoglobin level.

[0193] Figure 29 is a block diagram showing other configurations of the sample preparation unit 440 shown in Figure 28.

[0194] In the example shown in Figure 29, the configuration of the measurement channels in the sample preparation unit 440 is changed according to the division of labor between AI analysis and computational processing analysis. Specifically, compared to the sample preparation unit 440 in Figure 28, the sample preparation unit 440 in Figure 29 has a WDF sample preparation unit 440a for the WDF channel and reagents (WDF hemolytic agent and WNR staining solution) connected to the WNR sample preparation unit 440e for the WNR channel.

[0195] As shown in Figure 29, when the sample preparation unit 440 is configured, the classification of basophils and nucleated red blood cells performed in the WNR channel is performed in the WDF channel. The analysis unit 300 classifies neutrophils, lymphocytes, monocytes, eosinophils, basophils, and nucleated red blood cells by performing AI analysis on the waveform data obtained from the measurement sample prepared in the WDF channel. In this case, for example, the waveform data corresponding to neutrophils, lymphocytes, monocytes, eosinophils, basophils, and nucleated red blood cells from the waveform data obtained from the measurement in the WDF channel is pre-trained as training data for the AI ​​algorithm 60. This generates an AI algorithm 60 capable of classifying neutrophils, lymphocytes, monocytes, eosinophils, basophils, and nucleated red blood cells from the waveform data of the WDF channel.

[0196] According to the configuration shown in Figure 29, for example, measurement samples for different specimens can be prepared in parallel in each reaction chamber of multiple WDF sample preparation units 404a. This allows for parallel measurement of WDF channels for different specimens.

[0197] Furthermore, in the configuration shown in Figure 29, the reaction chamber and reagents corresponding to the original measurement channel (WNR channel) are replaced with the reaction chamber and reagents corresponding to the later measurement channel (WDF channel). In this case, the analysis of the original measurement channel must be performed by analyzing the later measurement channel.

[0198] In the configuration shown in Figure 29, the analysis of the original measurement channel (WNR channel) is performed by AI analysis of waveform data from the later measurement channel (WDF channel). This makes it possible to replace the original measurement channel (WNR channel) with the later measurement channel (WDF channel). Therefore, the number of additional measurement channels can be increased without increasing the total number of measurement channels provided in the sample analyzer 4000. Increasing the number of WDF channels allows for the parallel measurement of different samples using multiple WDF channels, improving the throughput of measurements by the WDF channels. Furthermore, by dividing the AI ​​analysis and computational processing analysis, a significant effect is obtained in that the computational load required for AI analysis is reduced, and the throughput of sample processing is also improved.

[0199] In the process described with reference to Figure 7, an example was shown in which the measurement item is analyzed using either AI analysis or computational processing analysis. However, the choice of which method to use may be determined based on the measurement channel.

[0200] Figure 30 is a flowchart showing an example of how analysis is performed depending on the measurement channel.

[0201] In Figure 30, step S101 has been added in place of step S12, compared to Figure 6. The changes from Figure 6 will be explained below.

[0202] In step S101, the analysis unit 300 refers to a rule that includes whether to perform AI analysis or computational analysis based on the measurement channel, and based on the referred rule, identifies the waveform data to be subjected to AI analysis and the waveform data to be subjected to computational analysis, respectively, from the waveform data acquired in step S11.

[0203] Figure 31 is a schematic example diagram showing a screen for setting up AI analysis or computational processing analysis for each measurement channel. The measurement channels illustrated in Figure 31 relate to a blood cell analyzer.

[0204] The screen shown in Figure 31 is displayed, for example, on the display unit of the analysis unit 300. The screen in Figure 31 has checkboxes for setting up AI analysis and checkboxes for setting up computational analysis for each measurement channel. The user operates the checkboxes to select whether to perform AI analysis or computational analysis for each measurement channel and operates the setting button. As a result, the rule is stored in the memory unit of the analysis unit 300.

[0205] Returning to Figure 30, if the waveform data identified in step S101 includes waveform data subject to AI analysis (S13: YES), in step S14, the analysis unit 300 performs AI analysis on the waveform data subject to AI analysis identified in step S101. If the waveform data identified in step S101 includes waveform data subject to computational processing analysis (S15: YES), in step S16, the analysis unit 300 performs computational processing analysis on the waveform data subject to computational processing analysis identified in step S101.

[0206] In Figure 31, since the WDF channel is set as the target of AI analysis, the analysis of the measurement items associated with the WDF channel (e.g., leukocyte type) is performed by AI analysis of the waveform data obtained from the measurement sample prepared in the WDF channel. On the other hand, since the other channels are set as the target of computational processing analysis, the analysis of the measurement items associated with the other channels is performed by computational processing analysis of the waveform data obtained from the measurement samples prepared in those other channels.

[0207] Furthermore, if a WDF channel is set as the target of AI analysis, AI analysis may be performed on all measurement items corresponding to the WDF channel, or AI analysis may be performed on some of the measurement items corresponding to the WDF channel, and computational analysis may be performed on the remaining measurement items.

[0208] <Example of analytical method for analytes in a sample> Next, we will explain the method for generating training data 75 and the method for analyzing waveform data using the examples shown in Figures 32 to 35.

[0209] <Waveform data> Figure 32 is a schematic diagram illustrating the waveform data used in this analysis method.

[0210] As shown in the upper diagram of Figure 32, a measurement sample prepared from a specimen containing analyte A flows into the flow cell 4113. When light is shone on the analyte A flowing through the flow cell 4113, forward scattered light is generated in front of the direction of light propagation. Similarly, side scattered light and fluorescence are generated to the sides of the direction of light propagation. The forward scattered light, side scattered light, and fluorescence are received by the photodetectors 4116, 4121, and 4122, respectively, and signals corresponding to the received light intensity are output. As a result, analog optical signals representing the change in the signal over time are output from the photodetectors 4116, 4121, and 4122, respectively. The optical signal corresponding to the forward scattered light is called the "forward scattered light signal," the optical signal corresponding to the side scattered light is called the "side scattered light signal," and the optical signal corresponding to the fluorescence is called the "fluorescence signal." The optical signals are input to the A / D conversion unit 461 and converted into digital data.

[0211] The middle diagram in Figure 32 schematically illustrates the conversion to digital data by the A / D converter 461. Here, the analog optical signal is directly input to the A / D converter 461. The level of the optical signal may be converted directly to digital data, but processing such as noise reduction, baseline correction, and normalization may be performed as appropriate.

[0212] As shown in the middle diagram of Figure 32, the A / D converter 461 samples the forward scattered light signal, side scattered light signal, and fluorescence signal from the analog optical signals input from the photodetectors 4116, 4121, and 4122, starting at the point when the level of the forward scattered light signal exceeds a predetermined threshold, and ending at the point when the level of the forward scattered light signal falls below a predetermined threshold. Digital waveform data corresponding to one analyte is acquired from the waveform between the start and end points. The A / D converter 461 samples each optical signal at a predetermined sampling rate (for example, sampling 1024 points at 10 nanosecond intervals, sampling 128 points at 80 nanosecond intervals, or sampling 64 points at 160 nanosecond intervals).

[0213] For convenience, in this example, start and end points were set for the analog optical signal and waveform data was acquired. However, as mentioned above, the optical signal may be converted entirely into digital data, and then start and end points may be set for the digital data to acquire the waveform data.

[0214] The lower part of Figure 32 schematically shows the waveform data obtained by sampling. Through sampling, matrix data (one-dimensional array data) is obtained as waveform data corresponding to one analyte, with elements being digital values ​​representing the analog signal levels at multiple time points. The A / D conversion unit 461 generates waveform data for forward scattered light, waveform data for side scattered light, and fluorescence for each analyte. The A / D conversion unit 461 repeats the generation of waveform data until the number of acquired analytes reaches a predetermined number, or until a predetermined time has elapsed since the sample was flowed into the flow cell 4113. This results in digital data consisting of waveform data for N analytes contained in one sample. The set of sampling data for each analyte (for example, a set of 1024 digital values ​​every 10 nanoseconds from t=0ns to t=10240ns) corresponds to the waveform data.

[0215] Each waveform data generated by the A / D conversion unit 461 may be assigned an index to identify each analyte. For example, the index may be assigned an integer from 1 to N in the order of the generated waveform data, and the same index may be assigned to the waveform data of forward scattered light, side scattered light, and fluorescence obtained from the same analyte. By assigning the same index to the waveform data corresponding to the same analyte, the AI ​​algorithm 60, described later, can analyze the waveform data of forward scattered light, side scattered light, and fluorescence corresponding to each analyte as a set and classify the type of analyte.

[0216] <Generating training data> Figure 33 is a schematic diagram showing an example of a method for generating training data used to train an AI algorithm 50 for determining the type of analyte in a sample.

[0217] By measuring the analyte based on flow cytometry, an optical signal 70a corresponding to forward scattered light, an optical signal 70b corresponding to side scattered light, and an optical signal 70c corresponding to fluorescence are obtained from the analyte. Waveform data 72a, 72b, and 72c corresponding to the analyte are obtained based on the optical signals 70a, 70b, and 70c, respectively. For training data 75, for example, waveform data 72a, 72b, and 72c of analytes that have been determined to be highly likely to be of a specific type as a result of computational analysis of the analytes in the sample measured based on flow cytometry can be used.

[0218] The following describes an example using the 4000 specimen analyzer as a blood cell counter for analyzing blood samples.

[0219] The operator measures the blood sample with the FCM detection unit 410 and stores waveform data of forward scattered light, side scattered light, and fluorescence of individual analytes contained in the sample. Subsequently, the operator classifies the analytes (cells) in the sample into groups of neutrophils, lymphocytes, monocytes, eosinophils, basophils, immature granulocytes, and abnormal cells, for example, based on the peak values ​​of the waveform data based on side scattered light and the peak values ​​of the waveform data based on fluorescence. The operator obtains training data 75 by assigning label values ​​77 corresponding to the classified cell type to the waveform data of that cell. Since training data 75 is generated for each cell type, the label values ​​77 differ depending on the cell type, as shown in Figure 34.

[0220] At this time, the operator determines the mode, mean, or median of the peak values ​​of the waveform data based on the lateral scattered light and fluorescence of cells included in the neutrophil population, identifies representative cells based on these values, and assigns the label value "1" corresponding to neutrophils to the waveform data of the identified cells.

[0221] The method for generating training data 75 is not limited to this. For example, the operator may collect only specific cells using a cell sorter, measure those cells based on flow cytometry, and assign cell label values ​​to the resulting waveform data to obtain training data 75.

[0222] The waveform data 72a, 72b, and 72c are combined with label values ​​77 that represent the type of cell from which the data originated. The training data 75 includes three waveform data (waveform data based on optical signals 70a, 70b, and 70c) corresponding to each cell, in an associated manner. The training data 75 is then input into the AI ​​algorithm 50.

[0223] <Overview of Deep Learning> Figure 33 will be used as an example to explain the overview of neural network training.

[0224] The AI ​​algorithm 50 is composed of a neural network including multiple hidden layers. In this case, the neural network is, for example, a convolutional neural network having convolutional layers. The number of nodes in the input layer 50a of the neural network corresponds to the number of elements in the arrays contained in the waveform data 72a, 72b, and 72c of the input training data 75. The number of elements in the arrays is equal to the sum of the number of elements in the waveform data 72a, 72b, and 72c for forward scattered light, side scattered light, and fluorescence corresponding to one analyte.

[0225] In the example in Figure 33, each of the waveform data 72a, 72b, and 72c contains 1024 elements, so the number of nodes in the input layer 50a is 1024 × 3 = 3072. The waveform data 72a, 72b, and 72c are input to the input layer 50a of the neural network. The label values ​​77 of each waveform data from the training data 75 are input to the output layer 50b of the neural network to train it. An intermediate layer 50c is positioned between the input layer 50a and the output layer 50b.

[0226] <Method for analyzing waveform data> Figure 35 schematically illustrates the method of analyzing waveform data of analytes in a sample using the AI ​​algorithm 60.

[0227] In the waveform data analysis method illustrated in Figure 35, the analyte is measured based on flow cytometry, thereby acquiring an optical signal 80a corresponding to forward scattered light, an optical signal 80b corresponding to side scattered light, and an optical signal 80c corresponding to fluorescence from the analyte. Waveform data 82a, 82b, and 82c corresponding to the analyte are acquired based on the optical signals 80a, 80b, and 80c, respectively. Then, analysis data 85 consisting of the waveform data 82a, 82b, and 82c is generated.

[0228] It is preferable that the analysis data 85 and the training data 75 have at least the same acquisition conditions. The acquisition conditions include the conditions for measuring the analytes in the sample based on flow cytometry, such as the preparation conditions of the sample to be measured, the flow rate when the sample to be measured is flowed through the flow cell, the intensity of the light irradiated onto the flow cell, and the amplification factor of the light-receiving unit that receives scattered light and fluorescence. The acquisition conditions also include the sampling rate when converting the analog optical signal to digital.

[0229] The analysis data 85 includes three waveform data (waveform data based on optical signals 80a, 80b, and 80c) corresponding to each analyte, in an associated manner. The analysis data 85 is then input into the trained AI algorithm 60. The AI ​​algorithm 60 consists of a neural network with multiple hidden layers.

[0230] When the analysis data 85 is input to the input layer 60a of the neural network constituting the AI ​​algorithm 60, classification information 82 regarding the type of analyte corresponding to the analysis data 85 is output from the output layer 60b. An intermediate layer 60c is positioned between the input layer 60a and the output layer 60b. The classification information 82 includes the probability that the analyte belongs to each of several types. Furthermore, the type with the highest probability is determined to be the type to which the analyte belongs, and a label value 83, which is an identifier representing that type, and an analysis result 84, which is a string representing that type, are output.

[0231] In the example of FIG. 35, since the probability that the type of analyte corresponding to the analysis data 85 is neutrophils is the highest, "1" is output as the label value 83, and character data "neutrophils" is output as the analysis result 84. The output of the label value 83 and the analysis result 84 may be performed by the AI algorithm 60, or other computer programs may output the most preferable label value 83 and analysis result 84 based on the probability calculated by the AI algorithm 60.

[0232] The method for analyzing the waveform data in the examples shown in FIGS. 19 to 21 described above will be described based on FIGS. 32 and 35 described above.

[0233] In the case of the examples shown in FIGS. 19 to 21, first, the analysis unit 300 analyzes the acquired waveform data by calculation processing. Then, the analysis unit 300 performs AI analysis on the waveform data corresponding to a predetermined type of cells (in the examples of FIGS. 19 to 21, monocytes and lymphocytes) classified by the calculation processing analysis.

[0234] When classified into a predetermined cell by the calculation processing analysis, the cell is identified, for example, by the index of the waveform data in FIG. 32. Thereby, the waveform data classified into monocytes and lymphocytes by the calculation processing analysis is specified by the index in the AI analysis. The analysis unit 300 performs AI analysis on the waveform data specified based on the index according to the example of FIG. 35. The analysis unit 300 inputs, for example, the waveform data specified by the index to the AI algorithm 60 that has been learned to be able to classify monocytes and lymphocytes in more detail.

[0235] The method for analyzing the waveform data described with reference to FIG. 29 described above will be described based on FIGS. 32 and 35 described above.

[0236] In the analysis method described with reference to Figure 29, the analysis unit 300 performs classification and counting of nucleated red blood cells (NRBCs) and basophils (BASOs), as well as classification and counting of eosinophils, neutrophils, lymphocytes, and monocytes, for example, by AI analysis of waveform data obtained from measurements using a WDF channel. In this example, the AI ​​algorithm 60 is trained to be able to classify nucleated red blood cells, basophils, eosinophils, neutrophils, lymphocytes, and monocytes using waveform data. By using such an AI algorithm 60, it is possible to replace the WNR channel with a WDF channel.

[0237] Figure 36 is a flowchart showing an example of performing AI analysis on waveform data acquired in a WDF channel.

[0238] In step S111, the measurement unit 400 acquires an optical signal from the measurement sample prepared in the WDF channel and obtains waveform data from the acquired optical signal. In step S112, the analysis unit 300 performs AI analysis on the waveform data acquired in step S111. In step S113, the analysis unit 300 provides the analysis results for the waveform data of the WDF channel and the analysis results for the waveform data of the other channels together. How the analysis of the waveform data of the other channels is divided between AI analysis and computational analysis is determined, for example, based on one of the rules exemplified in the embodiments described above.

[0239] In other analytical methods based on the configuration shown in Figure 29, the analysis unit 300 performs classification and counting of nucleated red blood cells and basophils by, for example, AI analysis of waveform data obtained from the WDF channel. The analysis unit 300 performs computational analysis on the waveform data corresponding to cells that were not classified as either nucleated red blood cells or basophils, and performs classification and counting of eosinophils, neutrophils, lymphocytes, and monocytes. In this example, the AI ​​algorithm 60 is trained to classify analytes from waveform data into nucleated red blood cells, basophils, and other analytes.

[0240] The analysis unit 300 performs computational analysis on waveform data corresponding to cells that were not classified as either nucleated red blood cells or basophils. For example, peak values ​​are extracted from the waveform data corresponding to cells that were not classified as either nucleated red blood cells or basophils, and the cell type is classified based on a two-dimensional graph (scattergram) generated from the peak values ​​corresponding to lateral scattered light and the peak values ​​corresponding to fluorescence. For example, based on this two-dimensional graph, the cells are classified as either eosinophils, neutrophils, lymphocytes, monocytes, or others. Cells classified as other than eosinophils, neutrophils, lymphocytes, and monocytes in the analysis based on the two-dimensional graph are classified as, for example, debris.

[0241] Figure 37 is a flowchart showing an example of classifying nucleated red blood cells and basophils using AI analysis and classifying others using computational processing analysis, based on waveform data acquired from the WDF channel.

[0242] In step S121, the measurement unit 400 acquires an optical signal from the measurement sample prepared in the WDF channel and obtains waveform data from the acquired optical signal. In step S122, the analysis unit 300 performs AI analysis on the waveform data acquired in step S121. This classifies nucleated red blood cells and basophils. In step S123, the analysis unit 300 identifies waveform data corresponding to cells that are not classified as either nucleated red blood cells or basophils.

[0243] In step S124, the analysis unit 300 performs computational analysis on the waveform data identified in step S123. This classifies lymphocytes, monocytes, eosinophils, and neutrophils. In step S125, the analysis unit 300 provides the analysis results for the waveform data of the WDF channel and the waveform data of the other channels together.

[0244] In other analytical methods based on the configuration shown in Figure 29, the analysis unit 300 performs classification and counting of lymphocytes, monocytes, eosinophils, and neutrophils or basophils by computational processing and analysis of waveform data obtained in the WDF channel, for example. In the classification and counting of neutrophils or basophils, for example, cells classified as either neutrophils or basophils are counted. Subsequently, the analysis unit 300 performs AI analysis on the waveform data corresponding to lymphocytes, monocytes, eosinophils, and cells not classified as either neutrophils or basophils, as well as cells classified as either neutrophils or basophils. As a result, the analytes are classified into nucleated red blood cells, basophils, and other cells.

[0245] For example, the count results of cells classified as either neutrophils or basophils by computational analysis are subtracted from the count results of cells classified as basophils by AI analysis to calculate the respective count results for neutrophils and basophils. Cells that are not classified as either nucleated red blood cells or basophils by AI analysis are classified as, for example, debris.

[0246] Figure 38 is a flowchart showing an example of performing AI analysis on neutrophils / basophils identified by the analysis of computational processing in the WDF channel.

[0247] In step S131, the measurement unit 400 acquires an optical signal from the measurement sample prepared in the WDF channel and obtains waveform data from the acquired optical signal. In step S132, the analysis unit 300 performs computational analysis on the waveform data acquired in step S131. This classifies the cells into groups consisting of lymphocytes, monocytes, eosinophils, and neutrophils and basophils. In step S133, the analysis unit 300 identifies the waveform data corresponding to (1) cells that were not classified as lymphocytes, monocytes, eosinophils, or neutrophils or basophils, and (2) cells that were classified as neutrophils or basophils.

[0248] In step S134, the analysis unit 300 performs AI analysis on the waveform data identified in step S133. This classifies neutrophils and basophils. In step S135, the analysis unit 300 provides the analysis results of the waveform data for the WDF channel and the analysis results of the waveform data for the other channels.

[0249] [Embodiment 5] Embodiment 5 shows a detailed configuration example in which computational processing analysis and AI analysis are performed separately in a sample analyzer 4000 that analyzes the coagulation ability of blood samples.

[0250] Examples of samples measured by the sample analyzer 4000 of Embodiment 5 include biological samples collected from a subject. These samples may include, for example, whole blood or plasma. The sample analyzer 4000 of Embodiment 5 analyzes for abnormalities caused by interfering substances in the sample based on methods such as coagulation, synthetic substrate, immunoturbidimetry, agglutination, and chemiluminescent enzyme immunoassay (CLEIA). The sample analyzer 4000 of Embodiment 5 comprises, for example, a measurement unit 400 and an analysis unit 300, similar to the configuration example of Embodiment 1 shown in Figure 1.

[0251] (Example configuration) Figure 39 is a schematic block diagram showing the configuration of the measurement unit 400 according to Embodiment 5.

[0252] The measurement unit 400 in FIG. 39 includes a detection unit 470 instead of the FCM detection unit 410 and further includes a control unit 466 as compared with the measurement unit 400 shown in FIG. 24.

[0253] The detection unit 470 includes a light source unit 471 and a detection block 476. The light source unit 471 includes, for example, a halogen lamp. The light source unit 471 is configured to be able to emit light with a wavelength of 660 nm for blood coagulation time measurement, light with a wavelength of 405 nm for synthetic substrate measurement, and light with a wavelength of 800 nm for immunoturbidimetry measurement. The sample preparation unit 440 mixes a blood coagulation reagent with the specimen to prepare a measurement sample. The detection unit 470 irradiates the measurement sample composed of the blood coagulation reagent and the specimen with light from the light source unit 471 and detects the light transmitted through the specimen. Note that the detection unit 470 may irradiate the measurement sample with light from the light source unit 471 and detect the light scattered by the specimen.

[0254] The control unit 466 is composed of, for example, an FPGA. The control unit 466 is connected to the analysis unit 300 via a bus 463 and an IF unit 465. The control unit 466 controls each part of the measurement unit 400 based on an instruction from the analysis unit 300. [[ID=ll]] <00009'65> FIG. 40 is a side view schematically showing the measurement by the detection block 476.

[0256] The detection block 476 includes a holding unit 472, an optical fiber 473, a condenser lens 474, and a light receiving unit 475.

[0257] The holding unit 472 holds the reaction vessel C1, which contains the measurement sample prepared from the sample and the sample corresponding to the measurement item. This allows the measurement sample to stand still. Light emitted from the light source unit 471 (see Figure 39) is guided by the optical fiber 473 to the focusing lens 474. The focusing lens 474 focuses the light from the optical fiber 473 onto the reaction vessel C1. The transmitted light that has been focused onto the reaction vessel C1 and passed through the measurement sample inside the reaction vessel C1 is received by the light receiving unit 475. The light receiving unit 475 is, for example, a photodiode. The light receiving unit 475 outputs an optical signal based on the intensity of the received transmitted light.

[0258] Returning to Figure 39, the analog processing unit 420 processes the analog optical signal output from the light receiving unit 475 (see Figure 40) and outputs it to the A / D conversion unit 461. The A / D conversion unit 461 converts the analog optical signal to digital. As described above, the digital data obtained by digitally converting the optical signal is the coagulation waveform data shown in Figure 4. The control unit 466 transmits the acquired coagulation waveform data to the analysis unit 300.

[0259] Figure 41 is a flowchart showing an example of analysis according to Embodiment 5. In Embodiment 5, the processor 3001 (see Figure 26) of the analysis unit 300 performs computational processing analysis and AI analysis on the coagulation waveform data.

[0260] In step S141, the measurement unit 400 acquires an optical signal in the detection unit 470 and obtains coagulation waveform data from the acquired optical signal. In step S142, the analysis unit 300 performs computational processing and analysis on the coagulation waveform data acquired in step S141. For example, as explained with reference to Figure 4, the analysis unit 300 obtains the time (T-T2) required for the absorbance of the coagulation waveform data to decrease to 50% as a result indicating the time it takes for the blood sample to coagulate.

[0261] In step S143, the analysis unit 300 performs AI analysis on the coagulation waveform data acquired in step S141. Based on the features extracted by the AI ​​algorithm 60 from the coagulation waveform data, the analysis unit 300 determines whether or not there are any abnormalities in the measurement. Based on the presence or absence of abnormalities in the measurement, the analysis unit 300 determines whether or not there is a suspicion of a nonspecific reaction occurring.

[0262] In step S144, the analysis unit 300 provides results indicating the time it takes for the blood sample obtained in step S142 to coagulate, and results indicating whether or not there are any abnormalities in the measurement obtained in step S143.

[0263] In Figure 41, AI analysis is always performed in step S143, but the analysis unit 300 may perform the process in step S143 based on a pre-set rule indicating whether or not AI analysis is necessary.

[0264] The sample analyzer 4000 in Embodiment 5 was a blood coagulation analyzer that optically measures the change in turbidity of the sample due to the coagulation of a blood sample, but it is not limited to this. For example, it may be a blood coagulation analyzer that measures the change in the amplitude motion of a steel ball in the sample due to the change in viscosity of the sample due to the coagulation of a blood sample, by receiving a high frequency emitted from a high-frequency transmitting coil. Furthermore, although the sample analyzer 4000 in Embodiment 5 was a blood coagulation analyzer, it may also be an immunoassay analyzer, a biochemical analyzer, or a gene analyzer.

[0265] [Embodiment 6] Embodiment 6 shows an example configuration of a sample analyzer 4000 including a host processor and a parallel processing processor. In Embodiment 6, parallel processing is performed on waveform data in the parallel processing processor 3002, and information regarding each type of analyte is generated based on the results of the parallel processing.

[0266] According to Embodiment 6, even when analyzing a massive amount of data ranging from several hundred megabytes to several gigabytes per sample, the waveform data processing can be performed in parallel by a parallel processing processor provided separately from the host processor. Therefore, even when processing a massive amount of data using the AI ​​algorithm 60, the data processing is completed within the sample analyzer 4000. This eliminates the need to transmit data to an analysis server storing the AI ​​algorithm 60 via the internet or intranet. Consequently, according to Embodiment 6, there is no need to transmit large amounts of data from the sample analyzer 4000 to the analysis server and obtain analysis results returned from the analysis server, thus improving the classification accuracy of analytes in the sample while maintaining high processing capacity of the sample analyzer 4000.

[0267] The configuration of the sample analyzer 4000 of Embodiment 6 will be described with reference to Figures 42 and 43. In the configuration example shown in Figures 42 and 43, the measurement unit 400 includes an FCM detection unit 410 for measuring a sample (e.g., blood sample, urine sample, body fluid, bone marrow fluid).

[0268] Figure 42 is a block diagram showing the configuration of the sample analyzer 4000 according to Embodiment 6.

[0269] The sample analyzer 4000 of Embodiment 6 comprises a measurement unit 400 and an analysis unit 300 located inside the measurement unit 400. Compared to the measurement unit 400 of Embodiment 4 shown in Figure 24, the measurement unit 400 of Embodiment 6 omits the IF units 462, 464, 465 and the bus 463. The analysis unit 300 of Embodiment 6 is connected to the A / D conversion unit 461, the device mechanism unit 430, the sample preparation unit 440, and the sample aspiration unit 450 located inside the measurement unit 400, and to a computer 301 located outside the measurement unit 400.

[0270] Figure 43 is a block diagram showing the configuration of the analysis unit 300 according to Embodiment 6.

[0271] The analysis unit 300 of Embodiment 6, compared to the analysis unit 300 of Embodiment 4 shown in Figure 26, includes a parallel processing processor 3002, a bus controller 3005, and IF units 462 and 464.

[0272] The parallel processing processor 3002 is configured to be able to process calculations performed by the AI ​​algorithm 60 in place of the master processor. By using the parallel processing processor 3002, which is suitable for matrix operations performed by the AI ​​algorithm 60, it is possible to improve the turn-to-attachment time (TAT) required for AI analysis. However, although the TAT is improved by the parallel processing processor 3002, the computer load required for AI analysis increases as the amount of data to be analyzed increases. In contrast, as described above, by dividing the data analysis between computational processing analysis and AI analysis, the computer load can be reduced and inspection efficiency can be improved.

[0273] The processor 3001 uses the parallel processing processor 3002 to perform waveform data analysis processing using the AI ​​algorithm 60. Specifically, the processor 3001 performs AI analysis of waveform data based on the AI ​​algorithm 60 by executing the analysis software 3100. The analysis software 3100 is used to analyze waveform data corresponding to the analytes in the sample based on the AI ​​algorithm 60.

[0274] The analysis software 3100 may also be stored in the memory unit 3004. In this case, the processor 3001 executes the analysis software 3100 stored in the memory unit 3004 to perform AI analysis of the waveform data based on the AI ​​algorithm 60.

[0275] In this embodiment, for example, AI analysis is performed by processor 3001 and parallel processing processor 3002, while computational analysis is performed by processor 3001 without using parallel processing processor 3002.

[0276] Processor 3001 is, for example, a CPU (Central Processing Unit). Processor 3001 may be, for example, Intel's Core i9, Core i7, Core i5, or AMD's Ryzen 9, Ryzen 7, Ryzen 5, Ryzen 3, etc.

[0277] Processor 3001 controls the parallel processing processor 3002. The parallel processing processor 3002 performs parallel processing, for example, matrix operations, in response to the control by processor 3001. In other words, processor 3001 is the master processor of the parallel processing processor 3002, and the parallel processing processor 3002 is the slave processor of processor 3001. Processor 3001 is also called the host processor or main processor. Processor 3001 performs matrix operations by the AI ​​algorithm 60 in parallel processing by the parallel processing processor 3002.

[0278] The parallel processing processor 3002 executes multiple arithmetic operations in parallel, which are at least part of the processing related to the analysis of waveform data. The parallel processing processor 3002 is, for example, a GPU (Graphics Processing Unit), an FPGA (Field Programmable Gate Array), or an ASIC (Application Specific Integrated Circuit). If the parallel processing processor 3002 is an FPGA, it may have, for example, pre-programmed arithmetic operations related to a trained AI algorithm 60. If the parallel processing processor 3002 is an ASIC, it may have, for example, pre-built circuits for executing arithmetic operations related to a trained AI algorithm 60, or it may have programmable modules built in addition to such built-in circuits.

[0279] As the parallel processing processor 3002, for example, NVIDIA's GeForce, Quadro, TITAN, Jetson, etc. may be used. If using the Jetson series, for example, Jetson Nano, Jetson Tx2, Jetson Xavier, or Jetson AGX Xavier can be used.

[0280] The processor 3001 performs calculations related to the control of the measurement unit 400, for example. The processor 3001 also performs calculations related to the control signals transmitted and received between the device mechanism unit 430, the sample preparation unit 440, and the sample aspiration unit 450, for example. Furthermore, the processor 3001 performs calculations related to the transmission and reception of information with the computer 301, for example.

[0281] Computer 301 has, for example, the function of displaying the analysis results sent from the analysis unit 300 based on the processing of the processor 3001. Computer 301 also sends, for example, a measurement order to the analysis unit 300. The measurement order is sent to computer 301 from, for example, the host computer. The user can also input the measurement order via the input device of computer 301.

[0282] The processor 3001 performs, for example, the reading of program data from the storage unit 3004, the loading of the program into the RAM 3017, and the sending and receiving of data to and from the RAM 3017. Each of the above processes performed by the processor 3001 is required to be executed in a predetermined order. For example, if the processes required to control the device mechanism unit 430, the sample preparation unit 440, and the sample aspiration unit 450 are A, B, and C, respectively, they may be required to be executed in the order B, A, C. Because the processor 3001 often performs sequential processes that depend on such an order, increasing the number of arithmetic units (sometimes called "processor cores," "cores," etc.) does not necessarily increase the processing speed.

[0283] On the other hand, the parallel processing processor 3002 performs routine and large-scale computational processing, such as operations on matrix data containing a large number of elements. In this embodiment, the parallel processing processor 3002 performs parallel processing that parallelizes at least a portion of the process of analyzing waveform data according to the AI ​​algorithm 60. The AI ​​algorithm 60 includes, for example, a large number of matrix operations. The AI ​​algorithm 60 may include, for example, at least 100 matrix operations, or it may include at least 1000 matrix operations.

[0284] The parallel processing processor 3002 has multiple arithmetic units, each of which can perform matrix operations simultaneously. In other words, the parallel processing processor 3002 can perform matrix operations in parallel using each of its multiple arithmetic units. For example, the matrix operations included in the AI ​​algorithm 60 can be divided into multiple operations that are not dependent on each other's order. These divided operations can then be executed in parallel by each of the multiple arithmetic units. These arithmetic units are sometimes referred to as "processor cores" or "cores."

[0285] By performing such parallel processing, it is possible to speed up the overall computational processing of the sample analyzer 4000. Processing such as matrix operations included in the AI ​​algorithm 60 is sometimes called "Single Instruction Multiple Data" (SIMD). The parallel processing processor 3002 is suitable for such SIMD operations. Such a parallel processing processor 3002 is sometimes called a vector processor.

[0286] As described above, processor 3001 is suitable for performing diverse and complex processing. On the other hand, parallel processing processor 3002 is suitable for performing a large amount of standardized processing in parallel. By performing a large amount of standardized processing in parallel, the turnaround time (TAT) required for computation is shortened.

[0287] Furthermore, the parallel processing performed by the parallel processing processor 3002 is not limited to matrix operations. For example, when the parallel processing processor 3002 performs learning processing on the AI ​​algorithm 50, differential operations and other operations related to the learning process may also be subject to parallel processing.

[0288] The number of arithmetic units in processor 3001 can be, for example, dual-core (2 cores), quad-core (4 cores), or octa-core (8 cores). On the other hand, parallel processing processor 3002 has, for example, at least 10 arithmetic units (10 cores) and can perform 10 matrix operations in parallel. Some parallel processing processors 3002 have tens of arithmetic units. Also, some parallel processing processors 3002 have, for example, at least 100 arithmetic units (100 cores) and can perform 100 matrix operations in parallel. Some parallel processing processors 3002 have hundreds of arithmetic units. Also, some parallel processing processors 3002 have, for example, at least 1000 arithmetic units (1000 cores) and can perform 1000 matrix operations in parallel. Some parallel processing processors 3002 have thousands of arithmetic units.

[0289] Figure 44 is a block diagram showing another configuration of the sample analyzer 4000 according to Embodiment 6. The sample analyzer 4000 in Figure 44 performs counting and classification of blood cells in a blood sample.

[0290] The sample analyzer 4000 in Figure 44, compared to the sample analyzer 4000 in Figure 42, is equipped with the same RBC / PLT detection unit 4101, HGB detection unit 4102, analog processing units 4201 and 4202, and A / D conversion units 4611 and 4612 as in Figure 27. The sample preparation unit 440 in Figure 44 is configured similarly to the sample preparation unit 440 shown in Figure 28 or Figure 29.

[0291] Figure 45 shows an example configuration of the parallel processing processor 3002.

[0292] The parallel processing processor 3002 includes multiple arithmetic units 3200 and RAM 3201. Each of the arithmetic units 3200 performs matrix data arithmetic operations in parallel. RAM 3201 stores data related to the arithmetic operations performed by the arithmetic units 3200. RAM 3201 is memory with a capacity of at least 1 gigabyte. RAM 3201 may also be memory with a capacity of 2 gigabytes, 4 gigabytes, 6 gigabytes, 8 gigabytes, or 10 gigabytes or more. The arithmetic units 3200 retrieve data from RAM 3201 and perform arithmetic operations. The arithmetic units 3200 are sometimes referred to as "processor cores," "cores," etc.

[0293] Figures 46 to 48 schematically show examples of the installation of the parallel processing processor 3002.

[0294] In the example shown in Figure 46, the processor 3001 is mounted on the circuit board 3301. The parallel processing processor 3002 is mounted on the graphics board 3300, and the graphics board 3300 is connected to the circuit board 3301 via the connector 3310. The processor 3001 is connected to the parallel processing processor 3002 via the bus 3003. In the example shown in Figure 47, the parallel processing processor 3002 is directly mounted on the circuit board 3301 and connected to the processor 3001 via the bus 3003. In the example shown in Figure 48, the processor 3001 and the parallel processing processor 3002 are provided as a single unit. In this case, the parallel processing processor 3002 is built into the processor 3001 mounted on the circuit board 3301.

[0295] Figure 49 shows another example of the parallel processing processor 3002 being installed.

[0296] In the example shown in Figure 49, the measurement unit 400 is equipped with a parallel processing processor 3002 via an external device 3400 connected to the measurement unit 400. The parallel processing processor 3002 is implemented in the external device 3400, which is, for example, a USB device. The parallel processing processor 3002 is installed in the sample analyzer 4000 by connecting the external device 3400 to the bus 3003 via the IF unit 467. The USB device may be a small device such as a USB dongle. The IF unit 467 is, for example, a USB interface with a transfer speed of several hundred Mbps, and more preferably a USB interface with a transfer speed of several Gbps to several tens of Gbps or more. As the external device 3400 on which the parallel processing processor 3002 is implemented, for example, an Intel Neural Compute Stick 2 may be used.

[0297] Multiple parallel processing processors 3002 can be installed in the sample analyzer 4000 by connecting multiple USB devices, each equipped with a parallel processing processor 3002, to the IF unit 467. Since the number of arithmetic units 3200 in a single USB device may be smaller than that of a GPU, etc., the number of cores can be scaled up by increasing the number of USB devices connected to the measurement unit 400.

[0298] Next, with reference to Figures 50 to 52, an overview of the arithmetic processing performed by the parallel processing processor 3002 based on the control of the analysis software 3100 running on the processor 3001 will be described.

[0299] Figure 50 shows an example configuration of a parallel processing processor 3002 that performs arithmetic processing.

[0300] The parallel processing processor 3002 has multiple arithmetic units 3200 and RAM 3201. The processor 3001 that runs the analysis software 3100 instructs the parallel processing processor 3002 to perform at least some of the arithmetic processing required when analyzing waveform data with the AI ​​algorithm 60. The processor 3001 instructs the parallel processing processor 3002 to perform arithmetic processing related to the analysis of waveform data based on the AI ​​algorithm 60.

[0301] All or at least part of the waveform data is stored in RAM 3017. The data stored in RAM 3017 is transferred to RAM 3201 of the parallel processing processor 3002, for example, by DMA (Direct Memory Access). Each of the multiple arithmetic units 3200 of the parallel processing processor 3002 performs arithmetic processing on the data stored in RAM 3201 in parallel. Each of the multiple arithmetic units 3200 retrieves the necessary data from RAM 3201 and performs arithmetic processing. The data corresponding to the calculation result is stored in RAM 3201 of the parallel processing processor 3002. The data corresponding to the calculation result is transferred from RAM 3201 to RAM 3017, for example, by DMA.

[0302] Figure 51 shows an overview of the matrix operations performed by the parallel processing processor 3002.

[0303] When analyzing waveform data according to the AI ​​algorithm 60, matrix multiplication calculations (matrix operations) are performed. The parallel processing processor 3002, for example, executes multiple matrix operations in parallel.

[0304] The upper part of Figure 51 shows the formula for calculating matrix multiplication. This formula calculates matrix c by multiplying matrix a (n x n) and matrix b (n x n). As illustrated in the upper part of Figure 51, the formula is written using a multi-level loop construct. The lower part of Figure 51 shows an example of arithmetic operations that are executed in parallel by the parallel processing processor 3002. The formula illustrated in the lower part of Figure 51 can be divided into n x n operations, which is the number of combinations of the first-level loop variable i and the second-level loop variable j. Since each of these divided operations is independent of the others, they can be executed in parallel.

[0305] Figure 52 is a conceptual diagram showing that the multiple arithmetic processes exemplified in the lower part of Figure 51 are executed in parallel by the parallel processing processor 3002.

[0306] As shown in Figure 52, each of the multiple arithmetic operations is assigned to one of the multiple arithmetic units 3200 provided by the parallel processing processor 3002. Each of the arithmetic units 3200 executes its assigned arithmetic operations in parallel with the others. In other words, each of the arithmetic units 3200 executes the divided arithmetic operations simultaneously.

[0307] The parallel processing processor 3002, as illustrated in Figures 51 and 52, performs calculations to obtain, for example, information regarding the probability that the cells corresponding to the waveform data belong to each of several cell types. The processor 3001, which runs the analysis software 3100, performs an analysis on the cell type of the cells corresponding to the waveform data based on the results of the calculations.

[0308] The calculation of the probability that each analyte in the sample belongs to each of the multiple classification categories may be performed by a processor other than the parallel processing processor 3002. For example, the calculation results from the parallel processing processor 3002 may be transferred from RAM 3201 to RAM 3017, and the processor 3001 may calculate information regarding the probability that each analyte corresponding to each waveform data belongs to each of the multiple classification categories based on the calculation results read from RAM 3017. Alternatively, the calculation results from the parallel processing processor 3002 may be transferred from RAM 3201 to the analysis unit 300, and a processor mounted in the analysis unit 300 may calculate information regarding the probability that each analyte corresponding to each waveform data belongs to each of the multiple classification categories.

[0309] The processes shown in Figures 51 and 52 are applied, for example, to the computational processing (also called filtering) related to the convolutional layer in the AI ​​algorithm 60.

[0310] Figure 53 is a schematic diagram illustrating the overview of the computational processing related to the convolutional layer.

[0311] The upper part of Figure 53 shows waveform data obtained based on forward scattered light, which is input to the AI ​​algorithm 60. The waveform data in this embodiment is one-dimensional matrix data, as shown in Figure 32. More simply, the waveform data is array data with elements arranged in a single column. For the sake of explanation, the number of elements in the waveform data is n (where n is an integer greater than or equal to 1). The upper part of Figure 53 shows multiple filters. The filters are generated by the learning process of the AI ​​algorithm 50. Each of the filters is one-dimensional matrix data representing the features of the waveform data. The filters shown in the upper part of Figure 53 are matrix data with 1 row and 3 columns, but the number of columns is not limited to 3. By performing matrix operations on the waveform data input to the AI ​​algorithm 60 and each filter, features corresponding to the cell type related to the waveform data are calculated.

[0312] The lower part of Figure 53 shows an overview of the matrix operations on waveform data and filters. The matrix operations are performed by shifting each filter one position relative to each element of the waveform data. The matrix operations are calculated using (Equation 1) below.

[0313]

number

[0314] In (Equation 1), the subscript x is a variable indicating the row and column numbers of the waveform data. The subscript h is a variable indicating the row and column numbers of the filter. In the example shown in Figure 53, the waveform data is one-dimensional matrix data, and the filter is a 1x3 matrix data, so L=1, M=3, p=0, q=0,1,2, i=0, j=0,1,…,n-1.

[0315] The parallel processing processor 3002 performs matrix operations represented by (Equation 1) in parallel using each of the multiple arithmetic units 3200. Based on the arithmetic operations performed by the parallel processing processor 3002, classification information regarding the type of each analyte in the sample is generated. The generated classification information is used to generate and display the test results of the sample based on the classification information.

[0316] As shown in Figures 42 and 43, the computer 301 is connected to the processor 3001 via the IF unit 3006 and the bus 3003, and can receive analysis results from the processor 3001 and the parallel processing processor 3002. The IF unit 3006 is, for example, a USB interface. The computer 301 receives the analysis results from the analysis unit 300 via the IF unit 3006 and displays the analysis results on the computer 301's display device.

[0317] The computer 301 may include an operating unit comprising a pointing device including a keyboard, mouse, or touch panel. Users such as physicians and laboratory technicians can input measurement orders to the sample analyzer 4000 and input measurement instructions according to the measurement orders by operating the operating unit. Users can input instructions to display test results to the computer 301 via the operating unit. Users can operate the operating unit to view various information related to the test results, such as numerical results based on the analysis, graphs, charts, and flag information assigned to the samples.

[0318] <Operation of the specimen analyzer> Refer to Figures 54 to 56 to explain the sample analysis operation using the sample analyzer 4000.

[0319] Figure 54 is a flowchart showing the analysis operation of the analysis unit 300 and the measurement unit 400.

[0320] In step S200, when the processor 3001 of the analysis unit 300 receives a measurement order, it instructs the measurement unit 400 to perform the measurement. For example, the analysis unit 300 controls the operation of each detection unit (FCM detection unit 410, RBC / PLT detection unit 4101, HGB detection unit 4102), sample aspiration unit 450, and sample preparation unit 440 of the measurement unit 400 based on the instructions given to the measurement unit 400. The measurement unit 400 starts measuring the sample in response to the instructions from the analysis unit 300.

[0321] In step S300, the sample aspiration unit 450 aspirates a sample from the blood collection tube and discharges the aspirated sample into the reaction chamber in response to a measurement instruction from the analysis unit 300. The measurement instruction from the analysis unit 300 includes information on the measurement channel for which measurement is requested by the measurement order. Based on the measurement channel information included in the measurement instruction, the sample aspiration unit 450 discharges the sample into the reaction chamber of the corresponding measurement channel.

[0322] In step S301, the sample preparation unit 440 prepares the sample to be measured in accordance with the measurement instructions from the analysis unit 300. Specifically, based on the information of the measurement channel included in the measurement instructions, the sample preparation unit 440 supplies reagents (hemolytic agent and staining solution) to the reaction chamber from which the sample is discharged and mixes the sample and reagents. This prepares the sample to be measured (e.g., WDF measurement sample, RET measurement sample, WPC measurement sample, PLT-F measurement sample, WNR measurement sample). Furthermore, the sample preparation unit 440 supplies reagents to the reaction chamber from which the sample was discharged, mixes the sample and reagents to prepare the RBC / PLT measurement sample. The sample preparation unit 440 supplies reagents to the reaction chamber from which the sample was discharged, mixes the sample and reagents to prepare the hemoglobin measurement sample.

[0323] In step S302, the FCM detection unit 410 measures the prepared sample in response to a measurement instruction from the analysis unit 300. Specifically, the apparatus mechanism unit 430 delivers the sample from the reaction chamber of the sample preparation unit 440 to the FCM detection unit 410 in response to a measurement instruction from the analysis unit 300. The sample delivered from the reaction chamber flows into the flow cell 4113 and is irradiated with laser light by the light source 4111 (see Figure 25). As the analytes contained in the sample pass through the flow cell 4113, light is irradiated onto the analytes, and the forward scattered light, side scattered light, and fluorescence generated from the analytes are detected by the photodetectors 4116, 4121, and 4122, respectively, and an analog optical signal corresponding to the light detection intensity is output. The optical signal is processed by the analog processing unit 420 and then output to the A / D conversion unit 461.

[0324] Furthermore, the RBC / PLT detection unit 4101 measures blood cells based on the RBC / PLT measurement sample using the sheath flow DC detection method. The HGB detection unit 4102 measures hemoglobin based on the hemoglobin measurement sample using the SLS-hemoglobin method. The analog signal detected by the RBC / PLT detection unit 4101 is processed by the analog processing unit 4201 and then output to the A / D conversion unit 4611, and the analog signal detected by the HGB detection unit 4102 is processed by the analog processing unit 4202 and then output to the A / D conversion unit 4612 (see Figure 27). In step S303, the A / D conversion unit 461 generates digital data by sampling the analog optical signal at a predetermined rate, as described above, and generates waveform data corresponding to each of the analytes based on the digital data. The waveform data generated by the A / D conversion unit 461 is transferred directly to the RAM without going through the processor 3001 of the analysis unit 300, for example, by DMA transfer. As a result, waveform data based on the forward scattered light signal acquired from the analytes, waveform data corresponding to the side scattered light, and waveform data corresponding to fluorescence are taken into the RAM 3017.

[0325] Furthermore, the A / D conversion unit 4611 generates digital data by sampling the analog signal from the RBC / PLT detection unit 4101 at a predetermined rate. The A / D conversion unit 4612 generates digital data by sampling the analog signal from the HGB detection unit 4102 at a predetermined rate. This digital data may also be incorporated into the RAM 3017.

[0326] In step S201, the processor 3001 of the analysis unit 300 performs AI analysis on the waveform data using the AI ​​algorithm 60, and performs computational analysis on representative values ​​in the waveform data that correspond to the characteristics of the analytes. The division of labor between AI analysis and computational analysis is as described above. As a result, the analytes in the sample are classified. The processing of AI analysis in step S201 will be described later, but as a result of processing using the parallel processing processor 3002, the processor 3001 obtains, for example, classification information 82 of individual analytes in the sample, and obtains label values ​​83 and analysis results 84 (see Figure 35).

[0327] In step S202, the processor 3001 uses the program stored in the memory unit 3004 to analyze the label value 83 and the analysis result 84 and generate the test result for the sample. In step S202, for example, the number of each type of analyte is counted based on the label value 83 and analysis result 84 of each analyte.

[0328] For example, in the case of testing blood cells in a blood sample, if there are N classification pieces from a single sample that are assigned the label value "1" indicating neutrophils, then the test result for the sample will be a count result of neutrophil count = N. The processor 3001 obtains the count results for the measurement items corresponding to the measurement channel based on the analysis result 84 and stores them in the storage unit 3004 along with the sample identification information.

[0329] Here, the measurement items corresponding to the measurement channel are the items for which the counting result is required by the measurement order. For example, the measurement items corresponding to the WDF channel include the measurement items for the 5-part leukocyte classification, namely monocytes, neutrophils, lymphocytes, eosinophils, and basophils. The measurement items corresponding to the RET channel include the measurement item for the number of reticulocytes. The measurement items corresponding to PLT-F include the measurement item for the number of platelets. The measurement items corresponding to WPC include the measurement item for the number of hematopoietic progenitor cells. The measurement items corresponding to WNR include the measurement items for the number of leukocytes and nucleated erythrocytes.

[0330] The counting results may include not only the items for which measurement is required (also known as reportable items) as listed above, but also the counting results of other cells that can be measured by the same measurement channel. For example, in the case of the WDF channel, as shown in Figure 34, in addition to the five types of leukocytes, immature granulocytes (IG) and abnormal cells are also included in the counting results.

[0331] Furthermore, the processor 3001 generates the test results of the sample by analyzing the obtained counting results and stores them in the memory unit 3004. The analysis of the counting results includes, for example, determining whether the counting results are within the normal range, whether any abnormal cells have been detected, and whether the deviation from the previous test results is within an acceptable range.

[0332] In step S203, the computer 301 displays the test results generated by the analysis unit 300 on the display unit.

[0333] Figure 55 is a flowchart showing the details of the AI ​​analysis in step S201 of Figure 54.

[0334] Step S201 is executed by the processor 3001 in accordance with the operation of the analysis software 3100.

[0335] In step S2010, processor 3001 transfers the waveform data acquired in RAM 3017 in step S303 to parallel processing processor 3002. The waveform data is transferred from RAM 3017 to RAM 3201 via DMA transfer, as shown in Figure 50. At this time, processor 3001 controls, for example, bus controller 3005 to transfer the waveform data from RAM 3017 to RAM 3201 via DMA.

[0336] In step S2011, processor 3001 instructs parallel processing processor 3002 to perform parallel processing on the waveform data. Processor 3001 instructs parallel processing by, for example, calling a kernel function of parallel processing processor 3002. The processing performed by parallel processing processor 3002 will be explained later with reference to Figure 56. Processor 3001 instructs parallel processing processor 3002 to perform matrix operations related to AI algorithm 60, for example. Waveform data corresponding to each analyte in the sample is input to AI algorithm 60. The waveform data input to AI algorithm 60 is processed by parallel processing processor 3002.

[0337] In step S2012, processor 3001 receives the calculation results performed by parallel processing processor 3002. The calculation results are DMA-transferred from RAM 3201 to RAM 3017, as shown in Figure 50. In step S2013, processor 3001 generates analysis results for each type of analyte based on the calculation results from parallel processing processor 3002.

[0338] Figure 56 is a flowchart detailing step S2011 in Figure 55.

[0339] Step S2011 is executed by the parallel processing processor 3002 based on instructions from processor 3001.

[0340] In step S2100, the processor 3001, which executes the analysis software 3100, causes the parallel processing processor 3002 to assign calculations to the calculation unit 3200. The processor 3001 causes the parallel processing processor 3002 to assign calculations to the calculation unit 3200, for example, by calling the kernel function of the parallel processing processor 3002. As shown in Figure 52, for example, matrix operations related to the AI ​​algorithm 60 are divided into multiple calculations, and each divided calculation is assigned to the calculation unit 3200. Waveform data corresponding to each analyte in the sample is input to the AI ​​algorithm 60. Matrix operations corresponding to the waveform data are divided into multiple calculations and assigned to the calculation unit 3200.

[0341] In step S2101, each calculation process is processed in parallel by multiple calculation units 3200. The calculation process is performed on multiple waveform data. In step S2102, the calculation results generated by the parallel processing by the multiple calculation units 3200 are transferred from RAM 3201 to RAM 3017. The calculation results are transferred via DMA from RAM 3201 to RAM 3017, as shown in Figure 50.

[0342] In step S201 of Figure 54, the processor 3001 of the analysis unit 300 may obtain analysis results for measurement items corresponding to the RBC / PLT channel (e.g., red blood cell count, hematocrit value, etc.) by applying the AI ​​algorithm 60 to the digital data based on the analog signal from the RBC / PLT detection unit 4101. Alternatively, the processor 3001 may obtain analysis results for measurement items for the HGB channel (e.g., hemoglobin level, etc.) by applying the AI ​​algorithm 60 to the digital data based on the analog signal from the HGB detection unit 4102.

[0343] Next, with reference to Figures 57 and 58, other configuration examples of the sample analyzer 4000, which consists of a measurement unit 400 and an analysis unit 300, will be described.

[0344] Figure 57 is a block diagram showing other configurations of the measurement unit 400.

[0345] In the example shown in Figure 57, the analog optical signal processed by the analog processing unit 420 is transmitted to the analysis unit 300 via the connection port 421. A connection cable 4210 is connected to the connection port 421. The other configurations shown in Figure 57 have the same configuration and functions as the measurement unit 400 of the embodiment described above.

[0346] Figure 58 is a block diagram showing other configurations of the analysis unit 300.

[0347] In the example shown in Figure 58, the analysis unit 300 is connected to the measurement unit 400 via the IF unit 3006. The RAM 3017 and bus 3003 are, for example, transmission lines having a data transfer rate of several hundred MB / s or more. The bus 3003 may also be a transmission line having a data transfer rate of 1 GB / s or more. The bus 3003 performs data transfer based on, for example, PCI-Express or PCI-X standards. The configuration of the processor 3001, the parallel processing processor 3002, the storage unit 3004, and the RAM 3017, as well as the processing performed by them, are the same as the configuration and processing described above.

[0348] The analysis unit 300 includes a connection port 3007, an A / D conversion unit 3008, and an IF unit 3009.

[0349] The connection port 3007 is connected to the connection port 421 of the measurement unit 400 (see Figure 57) via the connection cable 4210. The connection cable 4210 has a number of transmission paths corresponding to the type of analog signal transmitted from the measurement unit 400 to the analysis unit 300. For example, the connection cable 4210 is made of twisted pair cable and has a number of pairs of wires corresponding to the type of analog signal transmitted to the analysis unit 300. The connection cable 4210 is preferably less than 1 meter in length to reduce noise during signal transmission.

[0350] The A / D converter 3008 is connected to the connection port 3007. As described above, the A / D converter 3008 samples the analog optical signal output from the measurement unit 400 and generates waveform data corresponding to each analyte in the sample. The generated waveform data is stored in the storage unit 3004 or RAM 3017 via the IF unit 3009 and the bus 3003. The transmission path from the connection port 3007 to the A / D converter 3008 may also have a number of wires corresponding to the types of optical signals transmitted to the analysis unit 300.

[0351] The processor 3001 and the parallel processing processor 3002 perform arithmetic processing on the waveform data stored in the memory unit 3004 or RAM 3017. The analysis software 3100 running on the processor 3001 is the same as the analysis software 3100 shown in Figure 50. By executing the analysis software 3100, the processor 3001 generates classification information regarding the type of analyte in the sample through the same operation as described above.

[0352] Next, with reference to Figures 59 and 60, other configuration examples of the sample analyzer 4000, which consists of a measurement unit 400 and an analysis unit 300, will be described.

[0353] Figure 59 is a block diagram showing other configurations of the measurement unit 400.

[0354] The measurement unit 400 shown in Figure 59 includes an IF unit 4631 for transmitting waveform data generated by the A / D conversion unit 461 to the analysis unit 300. A transmission line 4632 is connected to the IF unit 4631. Other configurations and functions are the same as those of the measurement unit 400 described above.

[0355] The IF unit 4631 is, for example, an interface as a dedicated line with a communication bandwidth of 1 gigabit / second or more. For example, the IF unit 4631 is an interface compliant with Gigabit Ethernet, USB 3.0, or Thunderbolt 3. If the IF unit 4631 is Gigabit Ethernet, the transmission path 4632 is a LAN cable. If the IF unit 4631 is USB 3.0, the transmission path 4632 is a USB cable compliant with USB 3.0. The transmission path 4632 is, for example, a dedicated transmission path for transmitting digital data between the measurement unit 400 and the analysis unit 300.

[0356] Figure 60 is a block diagram showing the other configurations of the analysis unit 300.

[0357] The analysis unit 300 shown in Figure 60 includes an IF unit 3010. Other configurations and functions are the same as those of the analysis unit 300 described above. The analysis unit 300 may be connected to multiple measurement units 400 via multiple IF units 3010 and multiple IF units 3006.

[0358] The analysis software 3100, which operates on the processor 3001, has the same functions as the analysis software 3100 described above. The analysis software 3100 analyzes the type of analyte in the sample by operating in the same manner as described above.

[0359] In the configurations shown in Figures 59 and 60, the A / D conversion unit 461 within the measurement unit 400 generates digital waveform data based on the analog optical signal generated in the FCM detection unit 410. The waveform data is sent to the analysis unit 300 via the IF unit 462, bus 463, IF unit 4631, and transmission line 4632.

[0360] The measurement unit 400 and the analysis unit 300 are connected one-to-one, for example, via a transmission line 4632. In this case, the transmission line 4632 is a transmission line that does not involve the transmission of data related to devices other than the components that make up the sample analyzer 4000 (e.g., the measurement unit 400 and the analysis unit 300). The transmission line 4632 is a transmission line separate from, for example, an intranet or the internet. This avoids bottlenecks in the communication speed of digital data transmission, even when waveform data generated in the measurement unit 400 is transmitted to the analysis unit 300.

[0361] Next, other configuration examples of the sample analyzer 4000 will be described with reference to Figures 61 to 65.

[0362] Figure 61 is a block diagram showing other configurations of the sample analyzer 4000.

[0363] In this configuration example, an analysis unit 600 is provided between the measurement unit 400 and the computer 301. That is, in the configurations shown in Figures 61 to 65, the sample analyzer 4000 comprises the measurement unit 400, the computer 301, and the analysis unit 600. The analysis unit 600 analyzes the type of cells measured. As will be described later, the parallel processing processor 6002 in this configuration example is mounted on the sample analyzer 4000 in a manner that it is integrated into the analysis unit 600.

[0364] Figure 62 is a block diagram showing other configurations of the measurement unit 400.

[0365] In the measurement unit 400 shown in Figure 62, compared to the configuration in Figure 59, a computer 301 is connected to the IF unit 465, and an analysis unit 600 is provided between the IF unit 4631 and the computer 301. The analysis unit 600 is connected to the IF unit 4631 and the computer 301 in a communication manner. Note that the analysis unit 600 may be connected to multiple measurement units 400. The analysis unit 600 may be connected to multiple computers 301.

[0366] Figure 63 is a block diagram showing the configuration of the analysis unit 600.

[0367] The analysis unit 600 comprises a processor 6001, a parallel processing processor 6002, a bus 6003, a storage unit 6004, a RAM 6005, and IF units 6006 and 6007. Each part of the analysis unit 600 is connected to the bus 6003.

[0368] Bus 6003 is, for example, a transmission line having a data transfer rate of several hundred MB / s or more. Bus 3003 may be a transmission line having a data transfer rate of 1 GB / s or more. Bus 3003 performs data transfer based on, for example, PCI-Express or PCI-X standards. The analysis unit 600 may be connected to multiple measurement units 400 via multiple IF units 6006. If multiple measurement units 400 are provided, each of the measurement units 400 may be connected to an analysis unit 600. In this case, for example, multiple measurement units 400 and multiple analysis units 600 are connected one-to-one.

[0369] Figure 64 shows an example configuration of a parallel processing processor 6002 that performs arithmetic processing.

[0370] Processor 6001 and parallel processing processor 6002 have the same configuration and functions as processor 3001 and parallel processing processor 3002 described above, respectively. Parallel processing processor 6002 includes multiple arithmetic units 6200 and RAM 6201. Analysis software 6100, which analyzes the type of analyte in the sample, runs on processor 6001. Analysis software 6100 running on processor 6001 has the same functions as analysis software 3100 shown in Figure 50. Analysis software 6100 analyzes the type of analyte in the sample in the same manner as described in Figure 50. Analysis software 6100 transmits classification information of the analyte in the sample to computer 301 via IF unit 6007.

[0371] Figure 65 is a block diagram showing the configuration of computer 301.

[0372] The computer 301 in Figure 65 has a configuration similar to the analysis unit 600 in Figure 63, but with the parallel processing processor 6002 omitted. The computer 301 comprises a processor 3501, a bus 3503, a storage unit 3504, a RAM 3505, and an IF unit 3506.

[0373] The analysis software 3100 does not necessarily need to be running on the processor 3501. The computer 301 receives the analysis results from the analysis unit 600 via the IF unit 3506. The IF unit 3506 is, for example, Ethernet or USB. The IF unit 3506 may also be a wireless communication interface.

[0374] In the configurations shown in Figures 62 to 65, the analog optical signals of cells generated in the FCM detection unit 410 are converted into digital waveform data in the A / D conversion unit 461 within the measurement unit 400. The waveform data is sent to the analysis unit 600 via the IF unit 462, bus 463, IF unit 4631, and transmission line 4632.

[0375] As described above, the IF unit 4631 is a dedicated interface for connecting the measurement unit 400 and the analysis unit 600, providing a one-to-one connection between them. In other words, the transmission path 4632 is a transmission path that does not involve the transmission of data related to devices other than, for example, the components that make up the sample analyzer 4000 (e.g., the measurement unit 400 and the analysis unit 300). The transmission path 4632 is a separate transmission path from the intranet or the internet. This makes it possible to avoid communication speed bottlenecks in the transmission of waveform data, even when waveform data generated in the measurement unit 400 is transmitted to the analysis unit 600.

[0376] In this case, steps S200 to S202 in Figure 54 are performed in the analysis unit 600, and step S203 is performed in the computer 301.

[0377] Next, with reference to Figures 66 and 67, other configuration examples of the sample analyzer 4000 of Figure 61 will be described. In this example, the sample analyzer 4000 comprises a measurement unit 400, a computer 301, and an analysis unit 600.

[0378] In the measurement unit 400 shown in Figure 66, compared to the configuration in Figure 57, a computer 301 is connected to the IF unit 465, and an analysis unit 600 is provided between the connection port 421 and the computer 301. The analysis unit 600 is communicatively connected to the connection port 421 and the computer 301. The measurement unit 400 transmits analog optical signals to the analysis unit 600 via the connection cable 4210.

[0379] Compared to the configuration in Figure 63, the analysis unit 600 in Figure 67 includes a connection port 6008 and an A / D conversion unit 6009 instead of the IF unit 6006.

[0380] The analog optical signal transmitted from the analysis unit 600 via the connection cable 4210 is input to the A / D converter 6009 via the connection port 6008. The A / D converter 6009 generates waveform data from the optical signal using the same processing as the A / D converter 461.

[0381] The analysis unit 600 may be connected to multiple measurement units 400 via multiple connection ports 6008. If multiple measurement units 400 are provided, each measurement unit 400 may be connected to an analysis unit 600. In this case, for example, multiple measurement units 400 and multiple analysis units 600 are connected one-to-one.

[0382] In the configurations shown in Figures 66 and 67, in step S303 of Figure 54, the analysis unit 600 generates waveform data based on the analog optical signal transmitted from the measurement unit 400. Steps S200 to S202 of Figure 54 are performed in the analysis unit 600, and step S203 is performed in the computer 301.

[0383] Next, with reference to Figures 68 and 69, other configuration examples of the measurement unit 400 and analysis unit 300 of the sample analyzer 4000 will be described.

[0384] The measurement unit 400 in Figure 68, compared to the configuration in Figure 27, includes connection ports 421, 4211, and 4212 in place of the A / D conversion units 461, 4611, and 4612 and the IF unit 462. The analog optical signals acquired by each detection unit are transmitted to the analysis unit 300 via the connection cable 4210.

[0385] The analysis unit 300 in Figure 69, compared to the configuration in Figure 58, has three sets consisting of connection ports 3007, A / D conversion units 3008, and IF units 3009. The three connection ports 3007 are connected to connection ports 421, 4211, and 4212 in Figure 68, respectively.

[0386] In the configurations shown in Figures 68 and 69, in step S303 of Figure 54, the analysis unit 300 generates waveform data based on the analog optical signal transmitted from the measurement unit 400.

[0387] Next, with reference to Figures 70 and 71, other configuration examples of the measurement unit 400 and analysis unit 300 of the sample analyzer 4000 will be described.

[0388] The measurement unit 400 in Figure 70 includes an IF unit 4631, compared to the configuration in Figure 27. The A / D conversion units 461, 4611, and 4612 each generate waveform data based on the analog optical signals acquired by the corresponding detection units. The waveform data corresponding to each detection unit is transmitted to the analysis unit 300 via the transmission line 4632.

[0389] The analysis unit 300 in Figure 71 has three IF units 3010, compared to the configuration in Figure 60. Each of the three IF units 3010 is connected to the transmission line 4632 in Figure 70.

[0390] Next, with reference to Figures 72 and 73, other configuration examples of the measurement unit 400 and analysis unit 300 of the sample analyzer 4000 will be described.

[0391] In the measurement unit 400 shown in Figure 72, compared to the configuration in Figure 68, the computer 301 is connected to the IF unit 465, and the analysis unit 600 is positioned between the connection ports 421, 4211, and 4212 and the computer 301. The analysis unit 600 is connected to the connection ports 421, 4211, 4212 and the computer 301 in a communicative manner. The analysis unit 600 and the computer 301 are connected in a manner that enables the transmission and reception of digital data.

[0392] The analysis unit 300 in Figure 73, compared to the configuration in Figure 67, has three sets of connection ports 6008 and A / D converters 6009. The three connection ports 6008 are connected to connection ports 421, 4211, and 4212 in Figure 72, respectively.

[0393] Next, we will explain the data size of waveform data and digital data.

[0394] In this embodiment, for example, sampling is performed for one analyte in the sample for each of the analog optical signals based on forward scattered light (FSC), analog optical signals based on side scattered light (SSC), and analog optical signals based on fluorescence (FL).

[0395] Examples of sampling rates include sampling 1024 points at 10 nanosecond intervals, sampling 128 points at 80 nanosecond intervals, or sampling 64 points at 160 nanosecond intervals. The amount of data is, for example, 2 bytes per sample. For each of FSC, SSC, and FL, an amount of data corresponding to the sampling rate is acquired (for a rate of 1024 points, 2 bytes × 1024 = 2048 bytes). This amount of data is the amount of data per analyte in the sample.

[0396] In a single measurement, for example, FSC, SSC, and FL are measured for at least 100 analytes. Alternatively, FSC, SSC, and FL may be measured for at least 1000 analytes in a single measurement. Furthermore, FSC, SSC, and FL may be measured for approximately 10,000 to 140,000 analytes in a single measurement. Therefore, if 100,000 analytes are measured in a single measurement and the sampling rate is 1024, the data size of the digital data for each of the FSC, SSC, and FL will be 2 bytes × 1024 × 100,000 = 204,800,000 bytes, and the total for FSC, SSC, and FL will be 614,400,000 bytes.

[0397] Furthermore, FSC, SSC, and FL are measured for each measurement channel. Therefore, if the number of analytes measured in one measurement is 100,000, the sampling rate is 1024, and the number of measurement channels is 5, the data volume for each of FSC, SSC, and FL will be 2 bytes × 1024 × 100,000 × 5 = 1,024,000,000 bytes, and the total for FSC, SSC, and FL will be 3,072,000,000 bytes.

[0398] As described above, the capacity of digital data can range from several hundred megabytes to several gigabytes per sample, and depending on the number of analytes, sampling rate, and number of measurement channels, it can be at least 1 gigabyte.

[0399] According to this embodiment, when analyzing a massive amount of digital data ranging from several hundred megabytes to several gigabytes per sample, the analysis process using the AI ​​algorithm 60 is completed within the sample analyzer 4000 as described above, and the digital data is not transmitted to an analysis server located outside the sample analyzer 4000 via the internet or intranet. Therefore, the reduction in processing capacity due to the increased communication load that occurs when transmitting digital data from the sample analyzer 4000 to the analysis server can be avoided.

[0400] [Embodiment 7] <Configuration of the waveform data analysis system> Figure 74 is a schematic diagram showing the configuration of the waveform data analysis system according to this embodiment.

[0401] The configuration of the measurement unit 400a is the same as that of the measurement unit 400 described above. The measurement unit 400a delivers the measurement sample prepared based on the sample to the flow cell 4113. The light source 4111 (see Figure 25) irradiates the measurement sample supplied to the flow cell 4113 with light, and the photodetectors 4116, 4121, and 4122 (see Figure 25) detect the forward scattered light, side scattered light, and fluorescence generated from the analytes in the measurement sample. The measurement unit 400a generates waveform data from the optical signals based on the forward scattered light, side scattered light, and fluorescence output from the photodetectors 4116, 4121, and 4122, and transmits the generated waveform data to the deep learning device 100.

[0402] The deep learning device 100 is a vendor-side device. The deep learning device 100 receives training waveform data acquired by the measurement unit 400a. The method for generating the training waveform data is as described above. The AI ​​algorithm 50 stored in the deep learning device 100 is a deep learning algorithm. The deep learning device 100 trains the AI ​​algorithm 50, which is composed of a neural network before training, using the training data, and provides the AI ​​algorithm 60 trained with the training data to the user. The AI ​​algorithm 60, composed of the trained neural network, is provided from the deep learning device 100 to the sample analyzer 4000 via the recording medium 98 or the communication network 99. The recording medium 98 is a computer-readable, non-temporary tangible recording medium such as a DVD-ROM or USB memory.

[0403] The deep learning device 100 is, for example, composed of a general-purpose computer and performs deep learning processing based on the flowchart described later.

[0404] The sample analyzer 4000 performs AI analysis on waveform data corresponding to the analyte using an AI algorithm 60 composed of a pre-trained neural network.

[0405] <Hardware configuration of the deep learning system> Figure 75 is a block diagram showing the configuration of the deep learning device 100.

[0406] The deep learning device 100 comprises a processing unit 10, an input unit 16, and an output unit 17.

[0407] The input unit 16 and the output unit 17 are connected to the processing unit 10 via the IF unit 15. The input unit 16 is, for example, an input device such as a keyboard or mouse. The output unit 17 is, for example, a display device such as a liquid crystal display.

[0408] The processing unit 10 comprises a CPU 11, memory 12, storage unit 13, bus 14, IF unit 15, and GPU 19.

[0409] The CPU 11 performs data processing as described later. The memory 12 is used as a workspace for data processing. The storage unit 13 records the program and processing data described later. The bus 14 transmits data between each unit. The IF unit 15 performs data input and output with external devices. The GPU 19 functions as an accelerator to assist the arithmetic processing (e.g., parallel processing) performed by the CPU 11. In other words, in the following description, the processing performed by the CPU 11 includes processing performed by the CPU 11 using the GPU 19 as an accelerator. The GPU 19 has the same functionality as the parallel processing processors 3002 and 6002 described above. Alternatively, a chip suitable for neural network calculations may be used instead of the GPU 19. Examples of such chips include FPGAs, ASICs, and Myriad X (Intel).

[0410] The processing unit 10 pre-records the AI ​​algorithm 50, which consists of the program and the neural network before training according to this embodiment, in the storage unit 13, for example in executable format, in order to perform the processing of each step described later with reference to Figure 77. The executable format is, for example, a format generated by converting from a programming language by a compiler. The processing unit 10 uses the program recorded in the storage unit 13 to perform the training processing of the AI ​​algorithm 50 before training.

[0411] In the following description, unless otherwise specified, the processing performed by the processing unit 10 refers to the processing performed by the CPU 11 based on the program and AI algorithm 50 stored in the storage unit 13 or memory 12. The CPU 11 temporarily stores necessary data (such as intermediate data during processing) in the memory 12, which it uses as a working area, and appropriately records data to be stored long-term, such as calculation results, in the storage unit 13.

[0412] <Hardware configuration of the analytical instrument> The sample analyzer 4000 (see Figure 74) has the same configuration as described above and processes waveform data based on the algorithm provided by the deep learning device 100. Alternatively, the sample analyzer 4000 may also incorporate the functions of the deep learning device 100 and train the AI ​​algorithm 50 using training data. In this case, the deep learning device 100 is not required.

[0413] The sample analyzer 4000 pre-records, for example in executable format, the program and the AI ​​algorithm 60, which consists of a trained neural network according to this embodiment, in the storage unit 3004 (see, for example, Figure 26) and the storage unit 6004 (see, for example, Figure 63) in order to perform the processing of each step described in the waveform data analysis processing below. The sample analyzer 4000 performs the processing using the program and the AI ​​algorithm 60 recorded in the storage unit 3004.

[0414] The AI ​​algorithms 60 recorded in the memory units 3004 and 6004 may be updated via a communication network. The deep learning device 100 transmits the AI ​​algorithms 60 to the sample analyzer 4000 via a communication network (e.g., the Internet, an intranet). The sample analyzer 4000 updates the AI ​​algorithms 60 already recorded in the memory units 3004 and 6004 with the received AI algorithms 60.

[0415] <Functional blocks and processing procedures> (Deep learning processing) Figure 76 is a functional block diagram of the deep learning device 100.

[0416] The processing unit 10A of the deep learning device 100 includes a training data generation unit 101, a training data input unit 102, and an algorithm update unit 103. A program that causes the computer to perform deep learning processing is installed in the storage unit 13 or memory 12 of the processing unit 10 shown in Figure 75, and each functional block of the processing unit 10A is realized when the CPU 11 and GPU 19 execute this program.

[0417] The training data database (DB) 104 and the algorithm database (DB) 105 are recorded in the storage unit 13 or memory 12 of the processing unit 10 shown in Figure 75. The training waveform data 72a, 72b, and 72c are acquired in advance by, for example, the measurement unit 400a and stored in the training data database 104. The AI ​​algorithm 50 is stored in the algorithm database 105.

[0418] Figure 77 is a flowchart showing the processing performed by the deep learning device 100.

[0419] Steps S401, S404, and S406 in Figure 77 are performed by the training data generation unit 101. Step S402 is performed by the training data input unit 102. Steps S403 and S405 are performed by the algorithm update unit 103.

[0420] First, the processing unit 10A acquires training waveform data 72a, 72b, and 72c. The training waveform data 72a, 72b, and 72c are waveform data based on forward scattered light, side scattered light, and fluorescence, respectively. The training waveform data 72a, 72b, and 72c may be acquired, for example, by operator operation from the measurement unit 400a, from the recording medium 98, or via the communication network 99. When acquiring the training waveform data 72a, 72b, and 72c, information on which cell type the training waveform data 72a, 72b, and 72c represent is also acquired. The cell type information may be linked to the training waveform data 72a, 72b, and 72c, or it may be input by the operator via the input unit 16.

[0421] In step S401, the processing unit 10A generates training data 75 from the training waveform data 72a, 72b, 72c and label values ​​77, as shown in Figure 33. In step S402, the processing unit 10A inputs the training data 75 into the AI ​​algorithm 50 and obtains trial results. The trial results are accumulated each time multiple sets of training data 75 are input into the AI ​​algorithm 50.

[0422] In the cell type analysis method according to this embodiment, a convolutional neural network is used, and stochastic gradient descent is employed. Therefore, in step S403, the processing unit 10A determines whether a predetermined number of training results have been accumulated. If a predetermined number of training results have been accumulated (S403: YES), the processing unit 10A proceeds to step S404. On the other hand, if a predetermined number of training results have not been accumulated (S403: NO), the processing unit 10A skips the processing in step S404.

[0423] When a predetermined number of training results have been accumulated (S403:YES), in step S404, the processing unit 10A updates the connection weights w of the neural network constituting the AI ​​algorithm 50 using the training results accumulated in step S402. In the cell type analysis method according to this embodiment, stochastic gradient descent is used, so the connection weights w of the neural network are updated when a predetermined number of training results have been accumulated. Specifically, the process of updating the connection weights w is a process that performs calculations using gradient descent as shown in (Equation 12) and (Equation 13) described later.

[0424] In step S405, the processing unit 10A determines whether the AI ​​algorithm 50 has been trained with a specified number of training data 75. If the AI ​​algorithm 50 has been trained with a specified number of training data 75 (S405: YES), the deep learning process ends. On the other hand, if the AI ​​algorithm 50 has not been trained with a specified number of training data 75 (S405: NO), in step S406, the processing unit 10A takes in other training waveform data 72a, 72b, and 72c and returns the process to step S401.

[0425] Through the above process, the processing unit 10A trains the AI ​​algorithm 50 and obtains the AI ​​algorithm 60.

[0426] (Structure of neural networks) The upper part of Figure 78 is a schematic diagram illustrating the structure of the neural network constituting the AI ​​algorithm 50. As described above, a convolutional neural network is used in this embodiment. The neural network of the AI ​​algorithm 50 comprises an input layer 50a, an output layer 50b, and an intermediate layer 50c between the input layer 50a and the output layer 50b, with the intermediate layer 50c being composed of multiple layers. The number of layers constituting the intermediate layer 50c is, for example, 5 or more, preferably 50 or more, and more preferably 100 or more.

[0427] In the neural network of AI algorithm 50, multiple nodes 89 arranged in layers are connected between layers. As a result, information propagates only in one direction, from the input layer 50a to the output layer 50b, as shown by arrow D in the figure.

[0428] (Calculations at each node) The middle section of Figure 78 is a schematic diagram showing the operations at each node 89. Each node 89 receives multiple inputs and calculates one output (z). In the example shown in the middle section of Figure 78, node 89 receives four inputs. The total input (u) received by node 89 is expressed, for example, by the following (Equation 2). Here, in this embodiment, since one-dimensional matrix data is used as the training data 75 and analysis data 85, if a variable in the operation formula corresponds to two-dimensional matrix data, a process is performed to convert the variable so that it corresponds to one-dimensional matrix data.

[0429]

number

[0430] Each input is multiplied by a different weight. In equation (2), b is a value called the bias. The node output (z) is the output of a predetermined function f on the total input (u) represented by equation (2), and is expressed by equation (3) below. The function f is called the activation function.

[0431]

number

[0432] The lower part of Figure 78 is a schematic diagram showing the operations between nodes. In a neural network, nodes 89 are arranged in layers, with each node 89 outputting a result (z) represented by (Equation 3) for a total input (u) represented by (Equation 2). The output of the node 89 in the previous layer becomes the input of the node 89 in the next layer. In the example shown in the lower part of Figure 78, the output of node 89a in the layer on the left becomes the input of node 89b in the layer on the right. Each node 89b receives an output from node 89a. Different weights are applied to each connection between each node 89a and each node 89b. If the outputs of multiple nodes 89a are x1 to x4, then the inputs for each of the three nodes 89b are represented by (Equations 4-1) to (Equations 4-3) below.

[0433]

number

[0434] Generalizing these equations (Equations 4-1) to (Equations 4-3), we obtain the following equation (Equation 4-4). Here, i = 1, ..., I and j = 1, ..., J. I is the total number of inputs, and J is the total number of outputs.

[0435]

number

[0436] Applying (Equation 4-4) to the activation function yields the output shown in (Equation 5) below.

[0437]

number

[0438] (Activation function) In the cell type analysis method according to this embodiment, a rectified linear unit function is used as the activation function. The rectified linear unit function is expressed by the following equation (Equation 6).

[0439]

number

[0440] Equation (6) is a function in which the part of the linear function z=u where u<0 is set to u=0. In the example shown in the lower part of Figure 78, the output of the node j=1 is expressed by the following Equation (6).

[0441]

number

[0442] (Neural network training) Let y(x:w) be a function represented using a neural network. The function y(x:w) changes when the parameters w of the neural network are changed. Adjusting the function y(x:w) so that the neural network selects the most suitable parameters w for a given input x is called training or learning the neural network. Suppose there are multiple pairs of inputs and outputs of a function represented using a neural network. If d is the desired output for a given input x, then the input / output pairs are given as {(x1, d1), (x2, d2), ..., (xn, dn)}. The set of each pair represented by (x, d) is called the training data. Specifically, as shown in Figure 33, the set of waveform data 72a, 72b, and 72c is the training data 75.

[0443] Learning a neural network means adjusting the weights w so that, for any input-output pair (xn, dn), the output y(xn:w) of the neural network, given an input xn, comes as close as possible to the output dn, as shown in the following equation.

[0444]

number

[0445] An error function is a measure of how closely a function represented using a neural network matches the training data. The error function is also called a loss function. In the cell type analysis method according to this embodiment, the error function E(w) is expressed by the following equation (Equation 7). Equation (7) is called cross-entropy.

[0446]

number

[0447] The method for calculating the cross-entropy in (Equation 7) will be explained. In the output layer 50b of the neural network used in the cell type analysis method according to the embodiment, that is, in the final layer of the neural network, an activation function is used to classify the input x into a finite number of classes according to its content. The activation function is called the softmax function and is expressed by (Equation 8) below. It is assumed that the output layer 50b has the same number of nodes as the number of classes k. The total input u of each node k (k=1,...,K) in the output layer L is obtained from the output of the previous layer L-1, uk (L) Let's assume that it is given by . Then the output of the k-th node of the output layer is expressed by (Equation 8) below.

[0448]

number

[0449] Equation (8) is the softmax function. The sum of the outputs y1, ..., yK determined by Equation (8) is always 1.

[0450] If we denote each class as C1, ..., CK, then the output yK (i.e., uk) of node k in the output layer L is (L) ) represents the probability that a given input x belongs to class CK. The input x is classified into the class that maximizes the probability expressed by (Equation 9) below.

[0451]

number

[0452] In neural network training, the function represented by the neural network is considered a model of the posterior probability of each class. Under such a probabilistic model, the likelihood of the weights w for the training data is evaluated, and the weights w that maximize the likelihood are selected.

[0453] The target output of the softmax function in (Equation 8) is set to 1 only if the output is in the correct class, and 0 otherwise. If we represent the target output in vector form dn=[dn1,…,dnK], then for example, if the correct class of input xn is C3, then only the target output dn3 will be 1, and all other target outputs will be 0. When encoded in this way, the posterior distribution is expressed by (Equation 10) below.

[0454]

number

[0455] The likelihood L(w) of the weights w for the training data {(xn, dn)} (n=1,…,N) is given by (Equation 11) below. Taking the logarithm of likelihood L(w) and inverting its sign yields the error function in (Equation 7).

[0456]

number

[0457] Learning means minimizing the error function E(w), which is calculated based on the training data, with respect to the neural network parameter w. In the cell type analysis method according to this embodiment, the error function E(w) is expressed by (Equation 7).

[0458] Minimizing the error function E(w) with respect to parameter w is equivalent to finding the local minimum of the error function E(w). Parameter w is the weight of the connections between nodes. The local minimum of weight w can be found by iterative calculations that repeatedly update parameter w, starting from an arbitrary initial value. One example of such calculation is the gradient descent method.

[0459] The gradient descent method uses the vector represented by the following (Equation 12).

[0460]

number

[0461] In gradient descent, the current parameter w is repeatedly moved in the direction of the negative gradient (i.e., -∇E). (t) And the weight after the move is w (t+1) Therefore, the calculation using gradient descent is expressed by the following (Equation 13). The value t represents the number of times the parameter w has been moved.

[0462]

number

[0463] The symbol shown in (Equation 14) below, which was used in (Equation 13), is a constant that determines the magnitude of the update amount of the parameter w, and is called the learning rate.

[0464]

number

[0465] By repeating the operation represented by (Equation 13), the error function E(w) increases as the value t increases. (t) The parameter w decreases, and the parameter w reaches a local minimum.

[0466] Note that the calculation using (Equation 13) may be performed on all training data (n=1,...,N) or on only a portion of the training data. Gradient descent performed on only a portion of the training data is called stochastic gradient descent. In the cell type analysis method according to this embodiment, stochastic gradient descent is used.

[0467] [Effects of the Embodiment] The sample analyzer 4000 includes a measurement unit 400 which includes an FCM detection unit 410 and a detection unit 470 (optical detection unit) for acquiring optical signals from a sample, and an analysis unit 300 and an analysis unit 600 which analyze first and second data corresponding to the optical signals. The analysis units 300 and 600 perform AI analysis (first analysis operation by artificial intelligence algorithm) on the first data from all acquired waveform data, and perform computational processing analysis (second analysis operation that processes representative values ​​corresponding to the characteristics of the analyte) on the second data from all acquired waveform data.

[0468] With this configuration, the analysis of data corresponding to optical signals acquired from samples is divided between AI analysis and computational analysis. This reduces the load on the analysis units 300 and 600, which are the computers that process the data, compared to analyzing all data corresponding to optical signals using only artificial intelligence algorithms.

[0469] If the sample analyzer 4000 is a blood cell analyzer or a urine analyzer, the first and second data are digital data (waveform data) corresponding to the intensity of the optical signal based on light generated from each analyte (cells or formed elements). In this case, the optical signal is an analog signal output from the photodetector based on forward scattered light, side scattered light, and fluorescence. The optical signal has a region corresponding to each analyte in the sample and is a signal that reflects the presence of the analyte in the sample. The waveform data (first and second data) is generated corresponding to the region of the optical signal. In other words, the waveform data corresponds to the optical signal acquired while the analyte passes through the irradiation position of light from the light source 4111. Representative values ​​corresponding to the characteristics of the analyte are, for example, values ​​such as peak value, area, and width obtained from the waveform data corresponding to the analyte (see Figure 3). The first and second analysis operations are operations that determine the type of analyte (cells or formed elements).

[0470] If the sample analyzer 4000 is a blood coagulation analyzer, the first and second data are digital data (coagulation waveform data) corresponding to the intensity of the optical signal based on transmitted or scattered light. In this case, the optical signal is an analog signal from the start of photometry to the end of photometry (e.g., 180 seconds after the start) based on the intensity of transmitted or scattered light. The optical signal may also be an analog signal from the start of the coagulation reaction (timing T2 in Figure 4) to the end of the coagulation reaction (timing T3). The coagulation waveform data (first and second data) is generated from the optical signal. A representative value corresponding to the characteristics of the analyte is, for example, the time (e.g., T-T2) (see Figure 4) obtained from the coagulation waveform data when the detected light intensity satisfies predetermined conditions (e.g., when the absorbance is 50%). The first analysis operation is the operation to determine whether or not there is suspicion of a nonspecific reaction, and the second analysis operation is the operation to determine the coagulation time.

[0471] The first and second data sets may be identical or completely different. For example, if AI analysis is performed on nucleated red blood cells and basophils based on waveform data obtained from a single measurement using a WDF channel, and computational analysis is performed on other white blood cells, then the first and second data sets are identical. If computational analysis is performed on waveform data obtained from two measurements, and AI analysis is performed on waveform data based on the first measurement, then the first and second data sets are different.

[0472] In computational analysis, representative values ​​of the waveform data (secondary data) to be processed are identified based on the size of the waveform data (secondary data). Specifically, representative values ​​such as peak value, area, width, and the time it takes for absorbance to reach 50% are identified based on the size of the secondary data. This allows for the smooth identification of representative values.

[0473] When the sample analyzer 4000 is a blood cell analyzer or a urine analyzer, the optical signal has regions corresponding to each analyte in the sample. In the computational analysis, the analysis units 300 and 600 identify representative values ​​to be analyzed based on waveform data (second data) corresponding to each region of the optical signal. Thus, since the optical signal contains regions corresponding to each analyte, representative values ​​such as peak values, area, and width corresponding to each analyte can be smoothly identified based on the waveform data corresponding to each region of the optical signal.

[0474] When the sample analyzer 4000 is a blood cell analyzer or a urine analyzer, the optical signal has regions corresponding to each analyte in the sample. In AI analysis, the analysis units 300 and 600 input waveform data (first data) corresponding to each region of the optical signal into the artificial intelligence algorithm. Thus, since the optical signal contains regions corresponding to each analyte, AI analysis can be smoothly performed by inputting the waveform data corresponding to each region of the optical signal into the artificial intelligence algorithm.

[0475] As described above, when the optical signal has regions corresponding to each of the analytes in the sample, the measurement unit 400 acquires waveform data (first and second data) based on signals greater than a predetermined threshold corresponding to the intensity of the optical signal, as shown in the upper diagram of Figure 3. With this configuration, waveform data corresponding to each of the analytes can be accurately acquired.

[0476] As shown in Figure 6, the analysis unit 300 identifies the data to be analyzed by AI (first analysis operation) and the data to be analyzed by computational processing (second analysis operation) based on rules for identifying the data to be analyzed by AI (first analysis operation) and computational processing analysis (second analysis operation), respectively. With this configuration, it is possible to smoothly decide whether to perform the analysis using the first analysis operation or the second analysis operation for data corresponding to optical signals.

[0477] As shown in Figure 7, the analysis unit 300 identifies data to be analyzed by AI (first data) and data to be analyzed by computational processing (second data) according to the measurement items included in the measurement order for the sample. With this configuration, for example, analysis of measurement items that are difficult to analyze with high accuracy using computational processing can be performed by AI analysis, and analysis of normal measurement items can be performed by computational processing analysis. This enables high-precision analysis and reduces the load on the analysis unit 300.

[0478] As shown in Figure 13, the analysis unit 300 identifies the data to be analyzed by AI (first data) and the data to be analyzed by computational processing (second data) according to the type of measurement order for the sample. With this configuration, it is possible to decide whether to perform AI analysis or computational processing analysis depending on the type of measurement order, such as normal measurement, rerun measurement (re-executing the same measurement order), and reflex measurement (resetting the measurement order), that is, according to the purpose of the measurement based on the measurement order.

[0479] As shown in Figure 11, the analysis unit 300 identifies the data to be analyzed using AI (first data) and the data to be analyzed using computational processing (second data) according to the analysis mode of the sample analyzer 4000. With this configuration, for example, by setting either the AI ​​analysis mode or the computational processing analysis mode in advance for the sample analyzer 4000, the effort of setting the analysis mode for each sample and measurement item can be eliminated.

[0480] As shown in Figures 17 and 22, the analysis unit 300 determines whether or not to perform AI analysis (first analysis operation) based on the analysis results of the computational processing analysis (second analysis operation). With this configuration, for example, if further detailed analysis is needed based on the analysis results of the computational processing analysis, AI analysis can be performed to enable highly accurate analysis.

[0481] As shown in Figures 17 and 22, the analysis unit 300 determines whether or not to perform AI analysis (first analysis operation) depending on whether or not a predetermined analyte is detected in the sample by computational processing analysis (second analysis operation). With this configuration, if, for example, blast cells, abnormal lymphocytes, and atypical lymphocytes are detected by computational processing analysis, further detailed examinations can be performed by AI analysis.

[0482] As shown in Figure 19, the analysis unit 300 analyzes waveform data (first data) corresponding to analytes classified into predetermined types by computational processing analysis (second analysis operation) using computational processing analysis (first analysis operation). According to the analysis results of computational processing analysis, for example, as shown in Figure 21, cells classified as monocytes and lymphocytes have closely spaced distribution areas. Therefore, by performing AI analysis on cells classified as monocytes and lymphocytes by computational processing analysis, highly accurate classification can be achieved.

[0483] The representative values ​​processed in computational analysis (second analysis operation) are smaller in volume than the waveform data (first data) input to AI algorithm 60 in AI analysis (first analysis operation). In other words, because the amount of data to be processed in computational analysis is smaller compared to AI analysis, the load on the computer performing the analysis is less than in AI analysis. This makes it possible to shorten the TAT (Turn Around Time) of the analysis of measurement results.

[0484] Embodiments of the present invention can be modified in various ways as appropriate within the scope of the technical idea set forth in the claims. [Explanation of Symbols]

[0485] 60 AI Algorithms (Artificial Intelligence Algorithms) 80a, 80b, 80c optical signal 82a, 82b, 82c Waveform data (1st data, 2nd data) 300, 600 analysis units 400 measuring units 410 FCM detection unit (optical detection unit) 440 Sample Preparation Section 470 Detection unit (optical detection unit) 471 Light source section (light source) 475 Light-receiving section (photodetector) 3001, 6001 Processors (Host Processors) 3002, 6002 parallel processing processors 4000 Sample Analyzers 4111 Light source 4113 Flow Cell 4116, 4121, 4122 Photodetector

Claims

1. A specimen analyzer for analyzing analytes in a specimen, A sample preparation unit that mixes the aforementioned sample, a staining reagent, and a hemolytic reagent to prepare a measurement sample, A measurement unit including an optical detection unit for optically measuring the aforementioned sample, The system includes an analysis unit that analyzes measurement data acquired by optical measurement using the aforementioned optical detection unit, The aforementioned analysis unit, A first analysis operation using a non-AI algorithm and a second analysis operation using an AI algorithm are performed on the aforementioned measurement data. (i) The first analytical operation includes classifying the analyte into groups of at least neutrophils, lymphocytes, monocytes and eosinophils, and obtaining count results for each classified group, Here, in the first analysis operation, the analysis unit calculates at least two representative values ​​corresponding to the characteristics of each analyte contained in the measurement sample based on the measurement data, generates a two-dimensional plot using the at least two representative values, classifies the analytes into groups of at least neutrophils, lymphocytes, monocytes, and eosinophils based on the two-dimensional plot, and obtains the counting results for each of the classified groups. (ii) Based on the classification of the analyte by the second analytical operation, the acquisition of information suggesting the presence of abnormal cells, The abnormal cells include at least one selected from the group consisting of nucleated red blood cells, immature granulocytes, blast cells, abnormal lymphocytes, atypical lymphocytes, reactive lymphocytes, plasma cells, megakaryocytes, and tumor cells. A specimen analyzer characterized by the following features.

2. The specimen analyzer according to claim 1, wherein the analysis unit analyzes the measurement data by matrix operations using the AI ​​algorithm in the second analysis operation.

3. The analysis unit provides measurement results in accordance with the measurement order for the sample, The sample analyzer according to claim 1, wherein the measurement results include a first result, which is the result of performing the first analysis operation, and a second result, which is the result of performing the second analysis operation.

4. The measurement order is associated with a sample number for identifying each of the samples, The sample analyzer according to claim 3, wherein the analysis unit provides the measurement results in association with the sample number.

5. The measurement unit prepares the measurement sample according to the measurement order, which includes leukocyte classification. The sample analyzer according to claim 3, wherein the analysis unit performs the first analysis operation and the second analysis operation on the measurement data obtained from one of the measurement samples.

6. The sample analyzer according to any one of claims 1 to 5, wherein the optical detection unit comprises a light source, a flow cell, and a photodetector, and detects light generated from the analyte in the sample flowing through the flow cell by irradiating the flow cell with light.

7. The specimen analyzer according to claim 6, wherein the light generated from the analyte in the specimen is scattered light or fluorescence.

8. The specimen analyzer according to claim 6, wherein the measurement data corresponds to waveform data acquired while the analyte passes through the light irradiation position.

9. The sample analyzer according to claim 8, wherein the at least two representative values ​​are selected from the group consisting of peak values, area and width calculated from the waveform data based on light generated from the analyte in the sample.

10. The sample analyzer according to any one of claims 1 to 9, wherein the analysis unit analyzes the measurement data by convolution operation using the AI ​​algorithm.

11. The sample analyzer according to claim 1, wherein the analysis unit performs matrix operations by the AI ​​algorithm in parallel processing by a parallel processing processor.

12. The sample analyzer according to claim 11, wherein the analysis unit performs the second analysis operation by the parallel processing processor and the first analysis operation by the host processor of the parallel processing processor.

13. The sample analysis apparatus according to any one of claims 1 to 12, wherein the AI ​​algorithm is a deep learning algorithm.

14. The sample analyzer according to any one of claims 1 to 13, wherein the analysis unit identifies the relevant measurement data from the measurement data based on rules for identifying the measurement data that is the target of the first analysis operation and the second analysis operation, respectively.

15. The sample analyzer according to claim 1, wherein the analysis unit performs a first analysis operation and a second analysis operation according to the measurement items included in the measurement order for the sample.

16. The sample analyzer according to claim 1, wherein the analysis unit determines whether or not to perform the second analysis operation on the measurement data according to the type of measurement order for the sample.

17. The sample analyzer according to any one of claims 1 to 16, wherein the analysis unit determines whether or not to perform the second analysis operation on the measurement data according to the analysis mode of the sample analyzer.

18. The specimen analyzer according to any one of claims 1 to 17, wherein the analysis unit determines whether or not to perform the second analysis operation according to the analysis results of the first analysis operation.

19. The sample analyzer according to any one of claims 1 to 18, wherein the analysis unit determines whether or not to perform the second analysis operation depending on whether or not a predetermined analyte has been detected in the sample by the first analysis operation.

20. The specimen analyzer according to any one of claims 1 to 19, wherein the analysis unit analyzes the measurement data corresponding to the analytes classified into a predetermined type by the first analysis operation in the second analysis operation.

21. The sample analyzer according to any one of claims 1 to 20, wherein the amount of data of the at least two representative values ​​processed in the first analysis operation is smaller than the amount of data of the measurement data input to the AI ​​algorithm in the second analysis operation.

22. The measurement data subject to the first analysis operation and the measurement data subject to the second analysis operation include data originating from the same analyte, The specimen analyzer according to claim 21.

23. The sample preparation unit prepares a plurality of measurement samples from the sample, and the plurality of measurement samples include (1) a first measurement sample which is the measurement sample, and (2) a second measurement sample. The measurement unit measures the second sample and obtains the corresponding measurement data. The analysis unit performs a third analysis operation on each of the measurement data obtained from the second measurement sample. The specimen analyzer according to claim 1.

24. The measurement unit further includes an electrical detection unit and an HGB detection unit, The specimen analyzer according to claim 23, wherein the second measurement sample is a measurement sample used for measuring red blood cells and platelets by the electrical detection unit or a measurement sample used for measuring hemoglobin by the HGB detection unit.

25. The sample analyzer according to claim 23, wherein multiple measurement samples are prepared using the same type of reagent.

26. The sample analyzer according to claim 23, wherein the multiple measurement samples are prepared using different reagents from each other.

27. A method for analyzing analytes in a sample, The process involves mixing the aforementioned sample, a staining reagent, and a hemolytic reagent to prepare a sample for measurement. The process involves optically measuring the sample being measured, This includes an analysis step of analyzing measurement data obtained by optical measurement, In the aforementioned analysis process, A first analysis operation using a non-AI algorithm and a second analysis operation using an AI algorithm are performed on the aforementioned measurement data. (i) The first analytical operation includes classifying the analyte into groups of at least neutrophils, lymphocytes, monocytes and eosinophils, and obtaining count results for each classified group, Here, based on the measurement data, at least two representative values ​​corresponding to the characteristics of each analyte contained in the measurement sample are calculated, a two-dimensional plot is generated using the at least two representative values, the analytes are classified into groups of at least neutrophils, lymphocytes, monocytes, and eosinophils based on the two-dimensional plot, and the counting results for each of the classified groups are obtained. (ii) Based on the classification of the analyte by the second analytical operation, the acquisition of information suggesting the presence of abnormal cells, The abnormal cells include at least one selected from the group consisting of nucleated red blood cells, immature granulocytes, blast cells, abnormal lymphocytes, atypical lymphocytes, reactive lymphocytes, plasma cells, megakaryocytes, and tumor cells. A method for analyzing a specimen characterized by the following features.

28. A program that causes a computer to perform the process of analyzing analytes in a sample, The process involves mixing the aforementioned sample with a staining reagent and a hemolytic reagent to prepare a sample for measurement. A process for optically measuring the aforementioned sample, This includes a process for analyzing measurement data obtained by optical measurement, The aforementioned analytical process is, A first analysis operation using a non-AI algorithm and a second analysis operation using an AI algorithm are performed on the aforementioned measurement data. (i) The first analytical operation includes classifying the analyte into groups of at least neutrophils, lymphocytes, monocytes and eosinophils, and obtaining count results for each classified group, Here, based on the measurement data, at least two representative values ​​corresponding to the characteristics of each analyte contained in the measurement sample are calculated, a two-dimensional plot is generated using the at least two representative values, the analytes are classified into groups of at least neutrophils, lymphocytes, monocytes, and eosinophils based on the two-dimensional plot, and the counting results for each of the classified groups are obtained. (ii) Based on the classification of the analyte by the second analytical operation, the acquisition of information suggesting the presence of abnormal cells, The abnormal cells include at least one selected from the group consisting of nucleated red blood cells, immature granulocytes, blast cells, abnormal lymphocytes, atypical lymphocytes, reactive lymphocytes, plasma cells, megakaryocytes, and tumor cells. A program characterized by the following features.

Images