Cell analysis method and cell analyzer

Abstract

To provide a cell analysis method and a cell analyzer that analyze data obtained from a plurality of cells included in a specimen, and can process data of cells with a largely increased information amount with required throughput.SOLUTION: A cell analyzer comprises: measurement units (400, 400a, 500, 500a, 700) that each measure a plurality of cells included in a sample; first processors (3001, 4831, 6001, 8111) that each perform information processing related to analysis of the cells; and second processors (3002, 4833, 6002, 8112) that each perform parallel processing. The measurement unit acquires matrix data including, as elements, values indicating an analog signal level corresponding to the intensities of light emitted from the cells at a plurality of time points of each of the cells in a flow cytometer. The second processor executes the parallel processing for processing the matrix data according to an artificial intelligence algorithm including a matrix operation. The first processor classifies the respective cell classes of the cells on the basis of a result of the parallel processing.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a cell analysis method and a cell analysis device. [Background technology]

[0002] Patent Document 1 describes a method for classifying cells according to type by analyzing data obtained by measuring blood cells using a flow cytometer in a data processing system equipped with a processor. Patent Document 1 also describes that when cells cannot be classified using optical information measured by the flow cytometer, information on cell volume and electrical conductivity is also used. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Special Publication No. 2012-519848 Summary of the Invention [Problem to be solved by the invention]

[0004] If the amount of information used for cell classification is increased to improve the accuracy of cell classification, the amount of information obtained from each cell increases in samples such as blood and urine that contain multiple cells, resulting in a huge amount of data per sample. For example, to classify individual cells using a deep learning algorithm, it is necessary to significantly increase the amount of information obtained from each cell in order to extract the characteristics of each cell. Patent Document 1 does not disclose a system that can process such a significantly increased amount of information within the required processing capacity.

[0005] One aspect of the present invention aims to provide a cell analysis method and cell analysis device that can satisfy the required processing power requirements for cell data with a significantly increased amount of information, in a configuration for analyzing data obtained from multiple cells contained in a specimen. [Means for solving the problem]

[0006] In order to solve the above-mentioned problems, a cell analysis device according to one aspect of the present invention includes a flow cytometer and is provided with a measurement unit (400, 400a, 500, 500a, 700) configured to measure a plurality of cells contained in a sample, a first processor (3001, 4831, 6001, 8111) configured to process information related to the analysis of the plurality of cells, and a second processor (3002, 4833, 6002, 8112) configured to perform parallel processing, and the measurement unit (400, 400a, 500, 500a, 700) flows the sample through a flow cell of the flow cytometer, and detects light emitted from the irradiated cells. By obtaining an analog signal corresponding to the intensity of light emitted from each of the plurality of cells, matrix data is obtained whose elements are values that digitally indicate the analog signal level of each of the plurality of cells at multiple points in time, and the second processor (3002, 4833, 6002, 8112) is configured to perform parallel processing to process the matrix data according to an artificial intelligence algorithm including multiple matrix operations, and the first processor (3001, 4831, 6001, 8111) is configured to classify the cell type of each of the plurality of cells based on the results of the parallel processing by the second processor (3002, 4833, 6002, 8112).

[0007] In order to solve the above problem, another embodiment of the cell analysis method of the present invention is a cell analysis method using a cell analysis device including a first processor (3001, 4831, 6001, 8111), a second processor (3002, 4833, 6002, 8112), and a flow cytometer, comprising: flowing a sample through a flow cell of the flow cytometer; and acquiring, for each of the plurality of cells, an analog signal corresponding to the intensity of light emitted from the cells when irradiated with light, thereby acquiring matrix data for each of the plurality of cells in the sample, wherein the matrix data has elements that digitally indicate the analog signal levels at multiple points in time; performing parallel processing by the second processor to process the matrix data according to an artificial intelligence algorithm including multiple matrix operations; and classifying, by the first processor, the cell type of each of the plurality of cells based on the results of the parallel processing by the second processor. [Effects of the Invention]

[0008] It is possible to analyze multiple cells based on large amounts of data while still meeting throughput requirements. [Brief explanation of the drawings]

[0009] [Figure 1] (a) shows an example of white blood cell differentiation using the conventional method, and (b) shows an example of white blood cell differentiation using the present method. [Figure 2] (a) shows an example of irradiating light onto cells flowing through a flow cell. (b) shows an example of sampling forward scattered light signals, side scattered light signals, and fluorescent light signals. (c) shows an example of waveform data obtained by sampling. [Figure 3] An example of a method for generating training data is shown below. [Figure 4] Examples of label values are: [Figure 5] An example of a method for analyzing analytical data is shown below. [Figure 6] 1 shows an example of the appearance of a cell analyzer. [Figure 7] 1 shows an example of a block diagram of a measurement unit. [Figure 8] An example of the specimen aspiration section and sample preparation section is shown. [Figure 9] An example of the optical system configuration of the FCM detection unit is shown below. [Figure 10] 1 shows an example of the configuration of a processing unit. [Figure 11] 1 shows an example of the configuration of a parallel processing processor. [Figure 12] An example of implementing a parallel processing processor in a measurement unit is shown. [Figure 13] 10 shows another example of implementing a parallel processor in a measurement unit. [Figure 14] 10 shows another example of implementing a parallel processor in a measurement unit. [Figure 15] 10 shows another example of implementing a parallel processor in a measurement unit. [Figure 16] 1 shows an overview of the operation of a processor that performs arithmetic processing on matrix data using a parallel processing processor. [Figure 17] (a) shows the matrix multiplication formula, and (b) shows an example of the calculation process executed in parallel by a parallel processor. [Figure 18] This shows how the arithmetic processing is executed by a parallel processor. [Figure 19] (a) shows an example of waveform data of forward scattered light as waveform data input to the deep learning algorithm, and (b) shows an overview of the matrix operation between the waveform data and the filter. [Figure 20] 10 shows an example of the operation of analyzing a sample by a cell analyzer. [Figure 21] An example of a cell analysis process is shown. [Figure 22] An example of parallel processing is shown below. [Figure 23] 10 shows another example of a block diagram of a measurement unit. [Figure 24] 1 shows an example of a block diagram of a processing unit. [Figure 25] 1 shows an overview of the operation of a processor that performs arithmetic processing on matrix data using a parallel processing processor. [Figure 26] 10 shows another example of a block diagram of a measurement unit. [Figure 27] FIG. 10 shows another example of a block diagram of a processing unit. [Figure 28] 1 shows an example of the configuration of a measurement unit, a processing unit, and an analysis unit. [Figure 29] 10 shows another example of a block diagram of a measurement unit. [Figure 30] 1 shows an example of a block diagram of an analysis unit. [Figure 31] 1 shows an overview of the operation of a processor that performs arithmetic processing on matrix data using a parallel processing processor. [Figure 32] FIG. 10 shows another example of a block diagram of a processing unit. [Figure 33] 10 shows another example of a block diagram of a measurement unit. [Figure 34] 10 shows another example of a block diagram of an analysis unit. [Figure 35] 1 shows an example of a block diagram of a measurement unit. [Figure 36] 1 shows a schematic example of the optical system of a flow cytometer. [Figure 37] 1 shows an example of a schematic diagram of a sample preparation section of a measurement unit. [Figure 38] 1 shows an example of a schematic of a waveform data analysis system. [Figure 39] 1 shows an example of a block diagram of a vendor-side device. [Figure 40] 1 shows an example of a block diagram of a measurement unit. [Figure 41] 10 shows another example of a block diagram of an analysis unit. [Figure 42] 1 shows an example of a functional block diagram of a deep learning device. [Figure 43] 10 shows an example flowchart of the operation of a processing unit for generating training data. [Figure 44] The following diagrams are used to explain neural networks. (a) shows a schematic diagram outlining a neural network. (b) shows a schematic diagram illustrating the operations at each node. (c) shows a schematic diagram illustrating the operations between nodes. [Figure 45]The confusion matrix between the judgment results using the reference method and the judgment results using the deep learning algorithm is shown. [Figure 46] (a) shows the ROC curve for neutrophils, (b) shows the ROC curve for lymphocytes, and (c) shows the ROC curve for monocytes. [Figure 47] (a) shows the ROC curve for eosinophils, (b) shows the ROC curve for basophils, and (c) shows the ROC curve for control blood (CONT). [Figure 48] 1 shows an example of the configuration of a cell analyzer as an image analyzer. [Figure 49] 1 shows an example of the configuration of a processing unit. [Figure 50] An example of a method for generating training data is shown below. [Figure 51] Examples of label values are: [Figure 52] An example of a method for analyzing an image is shown below. [Figure 53] 1 shows an embodiment of the analysis results. DETAILED DESCRIPTION OF THE INVENTION

[0010] 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 will denote the same or similar components, and therefore, descriptions of the same or similar components will be omitted.

[0011] [1. Cell analysis method] This embodiment discloses a cell analysis method in a cell analysis device including a host processor and a parallel processing processor, which includes acquiring data regarding each cell based on control by the host processor, performing parallel processing on the data in the parallel processing processor, and generating information regarding the cell type for each of the cells based on the results of the parallel processing.

[0012] According to this analysis method, even when analyzing huge volumes of data, ranging from hundreds of megabytes to several gigabytes per sample, processing of cell data can be performed in parallel using a parallel processor separate from the host processor. Therefore, even when classifying cells using a deep learning algorithm that uses huge volumes of data, data processing can be completed using the cell analyzer alone. For example, there is no need to transmit cell data to an analysis server storing the deep learning algorithm via the Internet or intranet. Therefore, according to this analysis method, there is no need to transmit large volumes of data from the cell analyzer to the analysis server and then obtain analysis results from the analysis server. This allows the processing power of the cell analyzer to be maintained while improving the accuracy of cell classification.

[0013] An example of the outline of this embodiment will be described using FIG. 1. FIG. 1(a) is a diagram schematically illustrating white blood cell classification according to a conventional method, and FIG. 1(b) is a diagram schematically illustrating white blood cell classification according to the present method. In FIGS. 1(a) and 1(b), FSC represents an analog signal indicating the signal intensity of forward scattered light, SSC represents an analog signal of side scattered light, and SFL represents an analog signal indicating the signal intensity of side fluorescence. As shown in FIG. 1(a), in the conventional method, individual cells contained in a specimen are measured using a flow cytometer, and the peak heights of the pulses of the analog signals of forward scattered light, side scattered light, and side fluorescence are obtained as the forward scattered light intensity, side scattered light intensity, and side fluorescence intensity, respectively. Next, the cells are classified into specific types based on the forward scattered light intensity, side scattered light intensity, and side fluorescence intensity. The results of cell classification are displayed as a scattergram as shown in FIG. 1(a). In the scattergram in FIG. 1(a), the horizontal axis represents side scattered light intensity, and the vertical axis represents side fluorescence intensity.

[0014] Conventional white blood cell classification methods determine the type of blood cell based solely on the peak height information of an analog waveform, as shown in FIG. 1(a). In contrast, the method of the present embodiment classifies cells by analyzing the entire waveform of an analog signal acquired from a single cell by a flow cytometer, as data related to the cells in a specimen, as shown in FIG. 1(b). While FIG. 1(b) shows a waveform plot of the analog signal acquired by the flow cytometer, as will be described later, in this embodiment, the data related to the cells in a specimen is intended to be digital data (waveform data, as will be described later) whose elements are values indicating signal intensities at multiple time points obtained by A / D conversion of this analog signal. This digital data group is matrix data, and in this embodiment, for example, it is matrix data with one row and multiple columns (i.e., one-dimensional array data).

[0015] In this embodiment, waveform data for each cell type is learned by an untrained deep learning algorithm 50 shown in FIG. 1(b). Then, waveform data for cells of unknown cell types contained in a sample is input to a trained deep learning algorithm 60, which then derives a cell type determination result for each cell. The deep learning algorithms 50 and 60 are artificial intelligence algorithms and are composed of a neural network including multiple intermediate layers. In this embodiment, when processing related to waveform data analysis is performed according to the trained deep learning algorithm 60, a large number of matrix operations included in the deep learning algorithm 60 are executed in parallel using a parallel processing processor installed in the cell analyzer. The cell analyzer includes a parallel processing processor capable of executing parallel processing and an execution instruction processor (hereinafter simply referred to as a processor) that causes the parallel processing processor to execute parallel processing.

[0016] Hereinafter, individual cells in a biological sample subjected to analysis for the purpose of determining cell type will also be referred to as "cells to be analyzed." In other words, a biological sample may contain multiple cells to be analyzed. The multiple cells may include multiple types of cells to be analyzed.

[0017] Examples of biological samples include those collected from subjects. For example, biological samples may include peripheral blood such as venous blood and arterial blood, urine, and body fluids other than blood and urine. Body fluids other than blood and urine may include bone marrow fluid, ascites, pleural effusion, cerebrospinal fluid, etc. Hereinafter, body fluids other than blood and urine may be simply referred to as "body fluids." There are no limitations on blood samples, as long as they are in a state that allows cell counting and cell type determination. The blood is preferably peripheral blood. For example, the blood may be peripheral blood collected using an anticoagulant such as ethylenediaminetetraacetic acid (sodium salt or potassium salt) or heparin sodium. Peripheral blood may be collected from an artery or a vein.

[0018] The cell type to be determined in this embodiment is based on a morphological classification and varies depending on the type of biological sample. When the biological sample is blood and the blood is collected from a healthy individual, the cell type to be determined in this embodiment includes, for example, nucleated cells such as red blood cells and white blood cells, platelets, etc. Nucleated cells include, for example, neutrophils, lymphocytes, monocytes, eosinophils, and basophils. Neutrophils include, for example, segmented neutrophils and band-shaped neutrophils. On the other hand, when the blood is collected from a non-healthy individual, the nucleated cells may include, for example, at least one type selected from the group consisting of immature granulocytes and abnormal cells. Such cells are also included in the cell type to be determined in this embodiment. Immature granulocytes may include, for example, metamyelocytes, myelocytes, promyelocytes, myeloblasts, etc.

[0019] Furthermore, nucleated cells may include, in addition to normal cells, abnormal cells that are not found in the peripheral blood of healthy individuals. Examples of abnormal cells are cells that appear when a patient is affected with a specific disease, such as tumor cells. In the case of the hematopoietic system, the specific disease may be, for example, a disease selected from the group consisting of leukemias such as myelodysplastic syndrome, acute myeloblastic leukemia, acute myeloblastic leukemia, acute promyelocytic leukemia, acute myelomonocytic leukemia, acute monocytic leukemia, erythroleukemia, acute megakaryoblastic leukemia, acute myeloid leukemia, acute lymphocytic leukemia, lymphoblastic leukemia, chronic myelogenous leukemia, or chronic lymphocytic leukemia, malignant lymphomas such as Hodgkin's lymphoma and non-Hodgkin's lymphoma, and multiple myeloma.

[0020] Furthermore, abnormal cells may include cells that are not normally found in the peripheral blood of healthy individuals, such as lymphoblasts, plasma cells, atypical lymphocytes, reactive lymphocytes, proerythroblasts, basophilic erythroblasts, polychromatic erythroblasts, normochromatic erythroblasts, promegaloblasts, basophilic megaloblasts, polychromatic megaloblasts, and normochromatic megaloblasts; and megakaryocytes, including micromegakaryocytes.

[0021] Furthermore, when the biological sample is urine, the cell types to be determined in this embodiment may include, for example, red blood cells, white blood cells, epithelial cells such as transitional epithelium and squamous epithelium, etc. Abnormal cells may include, for example, bacteria, fungi such as filamentous fungi and yeast, tumor cells, etc.

[0022] When the biological sample is a body fluid that does not normally contain blood components, such as ascites, pleural effusion, or cerebrospinal fluid, the cell types may include, for example, red blood cells, white blood cells, and large cells. Here, "large cells" refer to cells that are detached from the lining of a body cavity or the peritoneum of an internal organ and are larger than white blood cells, such as mesothelial cells, histiocytes, and tumor cells.

[0023] When the biological sample is bone marrow fluid, the cell types to be determined in this embodiment may include, as normal cells, mature blood cells and immature blood cells. Mature blood cells include, for example, nucleated cells such as red blood cells and white 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-shaped neutrophils. Immature blood cells include, for example, hematopoietic stem cells, immature granulocytic cells, immature lymphocytic cells, immature monocytic cells, immature erythroid cells, megakaryocytic cells, and mesenchymal cells. Immature granulocytes may include, for example, metamyelocytes, myelocytes, promyelocytes, and myeloblasts. Immature lymphocytic cells include, for example, lymphoblasts. Immature monocytic cells include, for example, monoblasts. Immature erythroid cells include nucleated erythrocytes such as proerythroblasts, basophilic erythroblasts, polychromatic erythroblasts, normochromatic erythroblasts, promegaloblasts, basophilic megaloblasts, polychromatic megaloblasts, and normochromatic megaloblasts. Megakaryocytic cells include megakaryoblasts, etc.

[0024] Examples of abnormal cells that may be contained in bone marrow include hematopoietic tumor cells selected from the group consisting of leukemias such as the above-mentioned myelodysplastic syndrome, acute myeloblastic leukemia, acute myeloblastic leukemia, acute promyelocytic leukemia, acute myelomonocytic leukemia, acute monocytic leukemia, erythroleukemia, acute megakaryoblastic leukemia, acute myeloid leukemia, acute lymphocytic leukemia, lymphoblastic leukemia, chronic myelogenous leukemia, and chronic lymphocytic leukemia; malignant lymphomas such as Hodgkin's lymphoma and non-Hodgkin's lymphoma; and multiple myeloma; and metastatic tumor cells of malignant tumors that have developed in organs other than the bone marrow.

[0025] Figure 1 shows examples of signals obtained from cells, including forward scattered light signals, side scattered light signals, and side fluorescent light signals, which are optical signals obtained by irradiating light onto cells flowing through a flow cell. However, there are no particular limitations on the signals that can be obtained as long as they represent the characteristics of the cells and can be used to classify the cells by type.

[0026] The signal obtained from the cell may be any of a signal representing a morphological characteristic, a signal representing a chemical characteristic, a signal representing a physical characteristic, and a signal representing a genetic characteristic of the cell, but is preferably a signal representing a morphological characteristic of the cell. The signal representing a morphological characteristic of the cell is preferably an optical signal obtained from the cell.

[0027] The optical signal is preferably an optical signal obtained as an optical response when a cell is irradiated with light, and may include at least one signal selected from a signal based on light scattering, a signal based on light absorption, a signal based on transmitted light, and a signal based on fluorescence.

[0028] The signal based on light scattering may include a scattered light signal generated by light irradiation and a light loss signal generated by light irradiation. The scattered light signal becomes a parameter that indicates the characteristics of a cell depending on the angle at which the scattered light is received relative to the traveling direction of the irradiated light. The forward scattered light signal is used as a parameter that indicates the size of the cell. The side scattered light signal is used as a parameter that indicates the complexity of the cell nucleus.

[0029] The term "forward" in forward scattered light refers to the direction in which light is emitted from the light source. "Forward" may include a low forward angle where the light receiving angle is approximately 0 to 5 degrees, and / or a high forward angle where the light receiving angle is approximately 5 to 20 degrees, when the angle of the irradiating light is 0 degrees. "Side" is not limited as long as it does not overlap with "forward." "Side" may include a light receiving angle of approximately 25 to 155 degrees, preferably approximately 45 to 135 degrees, and more preferably approximately 90 degrees, when the angle of the irradiating light is 0 degrees.

[0030] A signal based on light scattering may include polarization or depolarization as a signal component. For example, by irradiating a cell with light and receiving the resulting scattered light through a polarizing plate, it is possible to receive only scattered light polarized at a specific angle. Alternatively, by irradiating a cell with light through a polarizing plate and receiving the resulting scattered light through a polarizing plate that transmits only light polarized at an angle different from the polarizing plate used for irradiation, it is possible to receive only depolarized scattered light.

[0031] The light loss signal represents the loss in the amount of light received, which occurs when light is irradiated onto cells and scattered, resulting in a decrease in the amount of light received at the light receiving unit. The light loss signal is preferably obtained as the light loss in the optical axis direction of the irradiated light (axial light loss). The light loss signal can be expressed as the ratio of the amount of light received when cells flow through the flow cell, to the amount of light received at the light receiving unit when no cells are flowing through the flow cell, which is taken as 100%. Like the forward scattered light signal, the axial light loss is used as a parameter representing the size of the cells, but the signal obtained differs depending on whether the cells are translucent or not.

[0032] The fluorescence-based signal may be fluorescence excited by irradiating light onto cells labeled with a fluorescent substance, or autofluorescence generated from unstained cells. The fluorescent substance may be a fluorescent dye that binds to nucleic acids or membrane proteins, or a labeled antibody in which an antibody that binds to a specific protein in a cell is modified with a fluorescent dye.

[0033] The optical signal may be acquired in the form of image data obtained by irradiating cells with light and capturing an image of the irradiated cells. Image data can be obtained by using a so-called imaging flow cytometer to capture images of individual cells flowing through a flow channel using an imaging device such as a TDI camera or a CCD camera. Alternatively, cell image data may be obtained by applying, spraying, or spotting a specimen or measurement sample containing cells onto a glass slide and capturing an image of the glass slide using an imaging device.

[0034] The signal obtained from the cells is not limited to an optical signal, but may also be an electrical signal obtained from the cells. For example, the electrical signal may be obtained by applying a direct current to a flow cell and using the change in impedance caused by the cells flowing through the flow cell as the electrical signal. The electrical signal obtained in this manner is a parameter that reflects the volume of the cells. Alternatively, the electrical signal may be the change in impedance caused by applying a radio frequency to the cells flowing through the flow cell. The electrical signal obtained in this manner is a parameter that reflects the conductivity of the cells.

[0035] The signal obtained from a cell may be a combination of at least two or more of the above-mentioned signals obtained from a cell. Combining multiple signals allows for multifaceted analysis of cell characteristics and enables more accurate cell classification. For example, the combination may be a combination of at least two of multiple optical signals, such as a forward scattered light signal, a side scattered light signal, and a fluorescent signal, or a combination of scattered light signals at different angles, such as a low-angle scattered light signal and a high-angle scattered light signal. Alternatively, an optical signal may be combined with an electrical signal, and the types and number of signals to be combined are not particularly limited.

[0036] In the cell analysis method of this embodiment, the determination of cell type is not limited to a method using a deep learning algorithm. For each cell passing through a predetermined position in a flow channel, signal intensities may be acquired at multiple time points while the cell is passing through the predetermined position, and the cell type may be determined based on the results of recognizing the acquired signal intensities at multiple time points for each cell as a pattern. The pattern may be recognized as a numerical pattern of signal intensities at multiple time points, or as a shape pattern obtained by plotting the signal intensities at multiple time points as a graph. When recognizing the pattern as a numerical pattern, the cell type can be determined by comparing the numerical pattern of the cell to be analyzed with a numerical pattern of a known cell type. The comparison of the numerical pattern of the cell to be analyzed with the numerical pattern of a control can be performed using, for example, Spearman's rank correlation, z-score, etc. The cell type can be determined by comparing the graph shape pattern of the cell to be analyzed with a graph shape pattern of a known cell type. The comparison of the graph shape pattern of the cell to be analyzed with the graph shape pattern of a cell whose type is already known may be performed using, for example, geometric shape pattern matching or a feature descriptor such as the SIFT Descriptor.

[0037] <Overview of cell analysis methods> Next, a method for generating training data 75 and a method for analyzing waveform data will be described using examples shown in FIGS.

[0038] <Waveform data> FIG. 2 is a schematic diagram illustrating waveform data used in this analysis method. As shown in FIG. 2(a), when a sample containing cells C is flowed through a flow cell FC and light is irradiated onto the cells C flowing through the flow cell FC, forward scattered light FSC is generated in the direction forward relative to the direction of light propagation. Similarly, side scattered light SSC and side fluorescent light SFL are generated to the side relative to the direction of light propagation. The forward scattered light is received by the first light-receiving element D1, and a signal corresponding to the amount of received light is output. The side scattered light is received by the second light-receiving element D2, and a signal corresponding to the amount of received light is output. The side fluorescent light is received by the third light-receiving element D3, and a signal corresponding to the amount of received light is output. As a result, analog signals representing changes in the signals over time are output from the light-receiving elements D1 to D3. The analog signal corresponding to the forward scattered light is called the "forward scattered light signal," the analog signal corresponding to the side scattered light is called the "side scattered light signal," and the analog signal corresponding to the side fluorescent light is called the "fluorescence signal." One pulse of each analog signal corresponds to one cell.

[0039] The analog signal is input to the A / D converter and converted into a digital signal. Figure 2(b) is a diagram showing the conversion to a digital signal by the A / D converter. For simplicity, the diagram shows the analog signal being directly input to the A / D converter. The analog signal level may be converted directly into a digital signal, or, if necessary, processing such as noise removal, baseline correction, and normalization may be performed. As shown in Figure 2(b), the A / D converter samples the forward scattered light signal, side scattered light signal, and fluorescence signal from the analog signals input from the light-receiving units D1 to D3, starting from the point when the level of the forward scattered light signal reaches a predetermined threshold level. The A / D converter samples each analog signal at a predetermined sampling rate (e.g., 1024 sampling points at 10 nanosecond intervals, 128 sampling points at 80 nanosecond intervals, or 64 sampling points at 160 nanosecond intervals).

[0040] Figure 2(c) is a schematic diagram showing waveform data obtained by sampling. Sampling generates waveform data corresponding to a single cell, which consists of matrix data whose elements represent analog signal levels at multiple time points in digital form. In this way, the A / D converter generates digital signals for forward scattered light, side scattered light, and side fluorescent light corresponding to a single cell. A / D conversion is repeated until a predetermined number of digitized cells are detected, or until a predetermined time has elapsed since the sample began flowing through the flow cell. This results in a digital signal combining the waveform data of N cells contained in a single sample, as shown in Figure 2(c). The collection of sampling data for each cell (in the example of Figure 2, a collection of 1024 digital values every 10 nanoseconds from t = 0 ns to t = 10240 ns) is called waveform data, and the collection of waveform data obtained from a single sample is called a digital signal.

[0041] Each piece of waveform data generated by the A / D conversion unit may be assigned an index to identify each cell. For example, the index is assigned an integer from 1 to N in the order in which the waveform data is generated, and the same index is assigned to waveform data of forward scattered light, waveform data of side scattered light, and waveform data of side fluorescent light obtained from the same cell.

[0042] Since one piece of waveform data corresponds to one cell, the index corresponds to the measured cell. By assigning the same index to waveform data corresponding to the same cell, the deep learning algorithm described below can analyze the waveform data of forward scattered light, waveform data of side scattered light, and waveform data of fluorescence corresponding to each cell as a set and classify the cell type.

[0043] <Generating training data> 3 is a schematic diagram showing an example of a method for generating training data used to train a deep learning algorithm for determining cell types. The training data 75 is waveform data generated based on a forward scattered light (FSC) analog signal 70a, a side scattered light (SSC) analog signal 70b, and a side fluorescent light (SFL) analog signal 70c obtained for cells contained in a sample measured using a flow cytometer. The waveform data is acquired as described above.

[0044] For example, the training data 75 can be waveform data of cells determined to be likely to belong to a specific cell type after measuring a sample using a flow cytometer and analyzing the cells contained in the sample based on a conventional scattergram. Taking a hemocytometer as an example, a blood sample is first measured using a flow cytometer, and waveform data of forward scattered light, side scattered light, and fluorescence of each cell contained in the sample is accumulated. Based on the side scattered light intensity (pulse height of the side scattered light signal) and fluorescence intensity (pulse height of the fluorescence signal), the cells are classified into groups of neutrophils, lymphocytes, monocytes, eosinophils, basophils, immature granulocytes, and abnormal cells. Training data can be obtained by assigning label values corresponding to the classified cell types to the waveform data of the cells. For example, the mode, mean, or median of the side scattered light intensity and side fluorescence intensity of cells contained in a neutrophil group can be calculated, and representative cells can be identified based on these values. The waveform data of these cells can then be assigned a label value of "1" corresponding to neutrophils, thereby obtaining training data. The method of generating training data is not limited to this. For example, training data can be obtained by collecting only specific cells using a cell sorter, measuring those cells using a flow cytometer, and assigning the label values of the cells to the obtained waveform data.

[0045] Analog signals 70a, 70b, and 70c represent the forward scattered light signal, side scattered light signal, and side fluorescent light signal, respectively, when neutrophils are measured by a flow cytometer. When these analog signals are A / D converted as described above, waveform data 72a of the forward scattered light signal, waveform data 72b of the side scattered light signal, and waveform data 72c of the side fluorescent light signal are obtained. Adjacent cells within each of the waveform data 72a, 72b, and 72c store signal levels at intervals corresponding to the sampling rate, e.g., 10 nanosecond intervals. The waveform data 72a, 72b, and 72c are combined with label values 77 that indicate the type of cell from which the data originated, and a set of three waveform data corresponding to each cell—in other words, three signal intensities (forward scattered light signal intensity, side scattered light signal intensity, and side fluorescent light signal intensity)—is input to the deep learning algorithm 50 as training data 75. In the example of Figure 3, the cells that formed the training data are neutrophils, so a label value 77 of "1" indicating that the data are neutrophils is assigned to waveform data 72a, 72b, and 72c, and training data 75 is generated. Figure 4 shows an example of the label value 77. Since training data 75 is generated for each cell type, different label values 77 are assigned depending on the cell type.

[0046] <Deep Learning Overview> An overview of neural network training will be explained using FIG. 3 as an example. The neural network 50 is preferably a convolutional neural network with a convolutional layer. The number of nodes in the input layer 50a of the neural network 50 corresponds to the number of elements in the array contained in the input waveform data 75. The number of elements in the array is equal to the sum of the number of elements in the waveform data 72a, 72b, and 72c of forward scattered light, side scattered light, and side fluorescent light corresponding to one cell. In the example of FIG. 3, 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 50. The label values 77 of each waveform data in the training data 75 are input to the output layer 50b of the neural network to train the neural network 50. The reference symbol 50c in FIG. 3 indicates a hidden layer.

[0047] <Waveform data analysis method> An example of a method for analyzing waveform data of cells to be analyzed is shown in Figure 5. In the waveform data analysis method, analysis data 85 consisting of waveform data obtained by the above-mentioned method is generated from analog signals 80a of forward scattered light, 80b of side scattered light, and 80c of side fluorescent light obtained by a flow cytometer from the cells to be analyzed.

[0048] It is preferable that the analytical data 85 and the training data 75 have at least the same acquisition conditions. The acquisition conditions include conditions for measuring cells contained in a specimen using a flow cytometer, such as preparation conditions for the measurement sample, the flow rate when the measurement sample is passed through the flow cell, the intensity of light irradiated onto the flow cell, and the amplification factor of the light receiving unit that receives scattered light and fluorescent light. The acquisition conditions also include the sampling rate when analog signals are A / D converted.

[0049] When the cells to be analyzed flow through the flow cell, an analog signal 80a of forward scattered light, an analog signal 80b of side scattered light, and an analog signal 80c of side fluorescent light are obtained. When these analog signals 80a, 80b, and 80c are A / D converted as described above, the signal intensity acquisition times for each cell are synchronized, and waveform data 82a of the forward scattered light signal, waveform data 82b of the side scattered light signal, and waveform data 82c of the side fluorescent light signal are generated. The waveform data 82a, 82b, and 82c are combined to form a set of data on the three signal intensities for each cell (the signal intensity of the forward scattered light, the signal intensity of the side scattered light, and the signal intensity of the side fluorescent light), and the combined data is input to the deep learning algorithm 60 as analysis data 85.

[0050] When analysis data 85 is input to the input layer 60a of the neural network 60 constituting the trained deep learning algorithm 60, an analysis result 83 is output from the output layer 60b as classification information regarding the type of cell corresponding to the analysis data 85. Reference symbol 60c in FIG. 5 indicates an intermediate layer. The classification information regarding the cell type is, for example, the probability that a cell belongs to each of multiple cell types. Furthermore, the analysis result 83 may include a label value 82, which is an identifier representing the cell type, by determining that the cell to be analyzed from which the analysis data 85 was obtained belongs to the classification with the highest probability. The analysis result 83 may include not only the label value itself, but also data in which the label value is replaced with information indicating the cell type (e.g., a character string). FIG. 5 shows an example in which, based on the analysis data 85, the deep learning algorithm 60 outputs the label value "1," which has the highest probability of belonging to the cell to be analyzed from which the analysis data 85 was obtained, and further outputs the character data "neutrophil" corresponding to this label value as the analysis result 83. The label value may be output by the deep learning algorithm 60, or another computer program may output the most preferred label value based on the probability calculated by the deep learning algorithm 60.

[0051] [2. Cellular analyzers and measurement of biological samples using cell analyzers] The cellular waveform data of this embodiment, or the analog cellular signal that is the source of the data, can be acquired by the first cellular analyzer 4000 or the second cellular analyzer 4000′. FIG. 6(a) shows an example of the external appearance of the cellular analyzer 4000. FIG. 6(b) shows an example of the external appearance of the cellular analyzer 4000′. In FIG. 6(a), the cellular analyzer 4000 includes a measurement unit 400 and a processing unit 300 that controls the measurement and setting of sample measurement conditions in the measurement unit 400, and analyzes the results of the analysis of the cellular data using the deep learning algorithm 60. In FIG. 6(b), the cellular analyzer 4000′ includes a measurement unit 500 and a processing unit 300 that controls the measurement and setting of sample measurement conditions in the measurement unit 500, and analyzes the results of the analysis of the cellular data using the deep learning algorithm 60. The measurement units 400 and 500 and the processing unit 300 can be connected to each other via wire or wirelessly for mutual communication. Below, configuration examples of the measurement units 400 and 500 are shown, but the present embodiment is not limited to the following examples. The processing unit 300 may be shared with the vendor-side device 100 described later.

[0052] <First cell analyzer and preparation of measurement samples> (Configuration of measurement unit and processing unit) An example of the configuration will be described in which the measurement unit 400 is a blood analyzer equipped with an FCM detection unit, which is a flow cytometer for detecting cells in a blood sample, more specifically, a blood cell counter.

[0053] (Configuration example 1) 7 to 19, exemplary configurations of the measurement unit 400 and processing unit 300 will be described. Fig. 7 shows an exemplary block diagram of the measurement unit 400. As shown in Fig. 7, the measurement unit 400 includes an FCM detection section 410 that detects blood cells, an analog processing section 420 that processes analog signals output from the FCM detection section 410, a measurement unit control section 480, a sample preparation section 440, an apparatus mechanism section 430, and a specimen aspirating section 450.

[0054] 8 is a schematic diagram illustrating the specimen aspirating unit 450 and the sample preparing unit 440. The specimen aspirating unit 450 includes a nozzle 451 for aspirating a blood specimen (whole blood) from a blood collection tube T, and a pump 452 for applying negative / positive pressure to the nozzle. The nozzle 451 is inserted into the blood collection tube T by being moved up and down by the device mechanism unit 430. When the pump 452 applies negative pressure while the nozzle 451 is inserted into the blood collection tube T, the blood specimen is aspirated through the nozzle 451. The device mechanism unit 430 may also include a hand member that inverts and agitates the blood collection tube T before aspirating blood from the blood collection tube T.

[0055] The sample preparation unit 440 includes five reaction chambers 440a to 440e. The reaction chambers 440a to 440e are used in the DIFF, RET, WPC, PLT-F, and WNR measurement channels, respectively. A hemolytic agent container containing a hemolytic agent and a staining solution container containing a staining solution are connected to each reaction chamber via a flow path. A measurement channel is formed by one reaction chamber and the reagents (hemolytic agent and staining solution) connected thereto. For example, the DIFF measurement channel is formed by the DIFF hemolytic agent and DIFF staining solution, which are DIFF measurement reagents, and the DIFF reaction chamber 440a. The other measurement channels are similarly configured. Note that, although the configuration in which one measurement channel contains one hemolytic agent and one staining solution is illustrated here, one measurement channel does not necessarily have to contain both a hemolytic agent and a staining solution; one reagent may be shared by multiple measurement channels.

[0056] After aspirating the blood sample, nozzle 451 accesses from above one of reaction chambers 440a-440e that corresponds to the measurement channel corresponding to the order by horizontal and vertical movement using device mechanism unit 430, and dispenses the aspirated blood sample. Sample preparation unit 440 supplies the corresponding hemolytic agent and staining solution to the reaction chamber into which the blood sample has been dispensed, and prepares a measurement sample by mixing the blood sample, hemolytic agent, and staining solution in the reaction chamber. The prepared measurement sample is supplied from the reaction chamber via a flow path to FCM detection unit 410, where cells are measured by flow cytometry.

[0057] Fig. 9 shows an example of the configuration of the optical system of the FCM detection unit 410. As shown in Fig. 9, in measurement using a flow cytometer, when cells contained in a measurement sample pass through a flow cell (sheath flow cell) 4113 provided in the flow cytometer, a light source 4111 irradiates the flow cell 4113 with light, and scattered light and fluorescence emitted from the cells in the flow cell 4113 are detected by this light.

[0058] In FIG. 9, light emitted from a laser diode, which is a light source 4111, is irradiated onto cells passing through a flow cell 4113 via an illumination lens system 4112.

[0059] In this embodiment, the light source 4111 of the flow cytometer is not particularly limited, and a light source 4111 having a wavelength suitable for exciting the fluorescent dye is selected. As such a light source 4111, for example, a semiconductor laser light source including a red semiconductor laser light source and / or a blue semiconductor laser light source, an argon laser light source, a gas laser light source such as a helium-neon laser, a mercury arc lamp, etc. are used. In particular, a semiconductor laser light source is preferable because it is much cheaper than a gas laser light source.

[0060] As shown in Fig. 9, forward scattered light emitted from particles passing through a flow cell 4113 is received by a forward scattered light receiving element 4116 via a condenser lens 4114 and a pinhole portion 4115. The forward scattered light receiving element 4116 is a photodiode. The side scattered light is received by a side scattered light receiving element 4121 via a condenser lens 4117, a dichroic mirror 4118, a bandpass filter 4119, and a pinhole portion 4120. The side scattered light receiving element 4121 is a photodiode. The side fluorescent light is received by a side fluorescent light receiving element 4122 via the condenser lens 4117 and a dichroic mirror 4118. The side fluorescent light receiving element 4122 is an avalanche photodiode. It should be noted that photomultiplier tubes may be used as the forward scattered light receiving element 4116, the side scattered light receiving element 4121, and the side fluorescent light receiving element 4122.

[0061] The light receiving signals output from the light receiving elements 4116, 4121 and 4122 are input to the analog processing unit 420 via amplifiers 4151, 4152 and 4153, respectively.

[0062] Returning to FIG. 7, the analog processing section 420 performs processing such as noise removal and smoothing on the analog signal input from the FCM detection section 410, and outputs the processed analog signal to the measurement unit control section 480.

[0063] The measuring unit control section 480 includes an A / D conversion section 482, a processor 4831, a RAM 4834, a storage section 4835, a bus controller 4850, a parallel processing processor 4833, and an interface section 489 that connects to the processing unit 300. The measuring unit control section 480 further includes an interface section 484 that is interposed between the measuring unit control section 480 and the A / D conversion section 482, and an interface section 488 that is interposed between the measuring unit control section 480 and the device mechanism section 430. As shown in FIG. 7, in this configuration example, the parallel processing processor 4833 is mounted on the cell analyzer 4000 by being incorporated inside the measuring unit 400.

[0064] The processor 4831 is connected to the interface unit 489, the interface unit 488, the interface unit 484, the RAM 4834, and the storage unit 4835 via the bus 485. The processor 4831 is connected to the parallel processing processor 4833 via the bus 485. The processing unit 300 is connected to each component of the measuring unit 400 via the interface unit 489 and the bus 485. The bus 485 is, for example, a transmission path having a data transfer rate of several hundred MB / s or more. The bus 485 may be, for example, a transmission path having a data transfer rate of 1 GB / s or more. The bus 485 transfers data based on, for example, the PCI-Express or PCI-X standard.

[0065] The A / D conversion unit 482 converts the analog signals output from the analog processing unit 420 into digital signals. The A / D conversion unit 482 converts the analog signals from the start of sample measurement to the end of measurement into digital signals. When multiple types of analog signals (for example, analog signals corresponding to forward scattered light intensity, side scattered light intensity, and fluorescence intensity) are generated by measurement in a certain measurement channel, the A / D conversion unit 482 converts each analog signal from the start of measurement to the end of measurement into a digital signal. For example, as described with reference to FIG. 9, three types of analog signals (i.e., forward scattered light signal, side scattered light signal, and fluorescence signal) are input to the A / D conversion unit 482 via multiple corresponding signal transmission paths 421. The A / D conversion unit 482 converts each of the analog signals input from the multiple signal transmission paths 421 into a digital signal. Each signal transmission path 421 is configured, for example, to transmit the analog signal as a differential signal.

[0066] As described with reference to FIG. 2, the A / D conversion unit 482 samples analog signals at a predetermined sampling rate (for example, sampling at 1024 points at 10 nanosecond intervals, sampling at 128 points at 80 nanosecond intervals, or sampling at 64 points at 160 nanosecond intervals). The A / D conversion unit 482 performs sampling processing on the three types of analog signals corresponding to each cell, thereby generating waveform data of a forward scattered light signal, waveform data of a side scattered light signal, and waveform data of a fluorescent signal for each cell. The A / D conversion unit 482 assigns an index to each of the generated waveform data. The generated waveform data becomes a digital signal consisting of consecutive waveform data of N cells contained in one specimen, as shown in FIG. 2(c). As a result, three digital signals corresponding to the three types of analog signals (forward scattered light signal, side scattered light signal, and fluorescent signal) obtained from the N cells are generated.

[0067] In addition to generating waveform data from the analog signal, the A / D conversion section 482 may also calculate peak values from the pulses of the analog signal.

[0068] The A / D conversion unit 482 inputs the generated digital signal to the bus 485. The bus controller 4850 transmits the digital signal output from the A / D conversion unit 482 to the RAM 4834 by, for example, DMA (Direct Memory Access) transfer. The RAM 4834 stores the digital signal.

[0069] The processor 4831 uses the parallel processing processor 4833 to perform analysis processing of waveform data included in the generated digital signal in accordance with the deep learning algorithm 60. That is, the processor 4831 is programmed to perform analysis processing of waveform data included in the digital signal in accordance with the deep learning algorithm 60. Analysis software 4832 for analyzing cell data based on the deep learning algorithm 60 may be stored in the storage unit 4835. In this case, the processor 4831 executes the analysis software 4832 stored in the storage unit 4835 to perform data analysis processing based on the deep learning algorithm 60. The processor 4831 is, for example, a CPU (Central Processing Unit). The processor 4831 may be, for example, a Core i9, Core i7, or Core i5 manufactured by Intel Corporation, or a Ryzen 9, Ryzen 7, Ryzen 5, or Ryzen 3 manufactured by AMD.

[0070] The processor 4831 controls the parallel processing processor 4833. The parallel processing processor 4833 executes parallel processing relating to, for example, matrix operations, according to the control of the processor 4831. In other words, the processor 4831 is the master processor of the parallel processing processor 4833, and the parallel processing processor 4833 is a slave processor of the processor 4831. The processor 4831 is also called a host processor or a main processor.

[0071] The parallel processing processor 4833 executes multiple arithmetic processes in parallel, which are at least a part of the processing related to the analysis of waveform data. The parallel processing processor 4833 is, for example, a graphics processing unit (GPU), a field programmable gate array (FPGA), or an application specific integrated circuit (ASIC). When the parallel processing processor 4833 is an FPGA, the parallel processing processor 4833 may be pre-programmed with arithmetic processes related to the trained deep learning algorithm 60, for example. When the parallel processing processor 4833 is an ASIC, the parallel processing processor 4833 may be pre-programmed with a circuit for executing arithmetic processes related to the trained deep learning algorithm 60, for example, or may have a programmable module built in addition to such a built-in circuit. The parallel processing processor 4833 may be, for example, NVIDIA GeForce, Quadro, TITAN, Jetson, or the like. For the Jetson series, for example, Jetson Nano, Jetson Tx2, Jetson Xavier, or Jetson AGX Xavier may be used.

[0072] The processor 4831 executes, for example, calculations related to the control of the measurement unit 400. The processor 4831 executes, for example, calculations related to control signals transmitted and received between the device mechanism 430, the sample preparation unit 440, and the specimen aspirating unit 450. The processor 4831 executes, for example, calculations related to the transmission and reception of information to and from the processing unit 300. The processor 4831 executes, for example, processes related to reading program data from the storage unit 4835, loading the program into the RAM 4834, and transmitting and receiving data to and from the RAM 4834. The above-described processes executed by the processor 4831 are required to be executed in a predetermined order, for example. For example, if processes A, B, and C are required to control the device mechanism 430, the sample preparation unit 440, and the specimen aspirating unit 450, they may be required to be executed in the order B, A, and C. The processor 4831 often executes sequential processes that depend on such an order, and therefore, increasing the number of arithmetic units (sometimes called "processor cores," "cores," etc.) does not necessarily increase the processing speed.

[0073] On the other hand, the parallel processing processor 4833 executes routine, large-volume computations, such as operations on matrix data containing a large number of elements. In this embodiment, the parallel processing processor 4833 executes parallel processing, which parallelizes at least a portion of the processing for analyzing waveform data according to the deep learning algorithm 60. The deep learning algorithm 60 may include, for example, a large number of matrix operations. The deep learning algorithm 60 may include, for example, at least 100 matrix operations, or may include at least 1,000 matrix operations. The parallel processing processor 4833 has multiple arithmetic units, each of which can simultaneously execute a matrix operation. In other words, the parallel processing processor 4833 can execute matrix operations in parallel by each of the multiple arithmetic units as parallel processing. For example, the matrix operation included in the deep learning algorithm 60 can be divided into multiple arithmetic operations that are independent of each other in order. The divided arithmetic operations can then be executed in parallel by each of the multiple arithmetic units. These arithmetic units may be referred to as "processor cores," "cores," or the like.

[0074] By performing such parallel processing, it is possible to speed up the calculation processing of the measurement unit 400 as a whole. Processing such as matrix operations included in the deep learning algorithm 60 is sometimes called, for example, "Single Instruction Multiple Data" (SIMD). The parallel processing processor 4833 is suitable for such SIMD operations, for example. Such a parallel processing processor 4833 is sometimes called a vector processor.

[0075] As described above, the processor 4831 is suitable for executing a variety of complex processes. On the other hand, the parallel processing processor 4833 is suitable for executing a large amount of standardized processes in parallel. By executing a large amount of standardized processes in parallel, the TAT (Turn Around Time) required for calculation processing is shortened.

[0076] Note that the target of parallel processing executed by the parallel processing processor 4833 is not limited to matrix operations. For example, when the parallel processing processor 4833 executes learning processing according to the deep learning algorithm 50, differential operations and the like related to the learning processing can be targets of parallel processing.

[0077] The number of arithmetic units of the processor 4831 is, for example, a dual-core (number of cores: 2), a quad-core (number of cores: 4), or an octa-core (number of cores: 8). On the other hand, the parallel processing processor 4833 has, for example, at least 10 arithmetic units (number of cores: 10) and can execute 10 matrix operations in parallel. Some parallel processing processors 4833 have, for example, several tens of arithmetic units. Other parallel processing processors 4833 have, for example, at least 100 arithmetic units (number of cores: 100) and can execute 100 matrix operations in parallel. Some parallel processing processors 4833 have, for example, several hundred arithmetic units. Other parallel processing processors 4833 have, for example, at least 1,000 arithmetic units (number of cores: 1,000) and can execute 1,000 matrix operations in parallel. The parallel processing processor 4833 may have, for example, several thousand arithmetic units.

[0078] FIG. 11 shows an example configuration of the parallel processing processor 4833. The parallel processing processor 4833 includes a plurality of arithmetic units 4836 and a RAM 4837. Each of the arithmetic units 4836 executes arithmetic processing of matrix data in parallel. The RAM 4837 stores data related to the arithmetic processing executed by the arithmetic units 4836. The RAM 4837 is a memory having a capacity of at least 1 gigabyte. The RAM 4837 may also have a capacity of 2 gigabytes, 4 gigabytes, 6 gigabytes, 8 gigabytes, 10 gigabytes, or more. The arithmetic units 4836 obtain data from the RAM 4837 and execute arithmetic processing. The arithmetic units 4836 may be referred to as a "processor core," a "core," or the like.

[0079] 12 to 14 show an example of mounting the parallel processing processor 4833 in the measuring unit 400. FIGS. 12 to 14 show an example in which the parallel processing processor 4833 is mounted in the cell analyzer 4000 so as to be incorporated inside the measuring unit 400. FIGS. 12 and 13 show an example in which the processor 4831 and the parallel processing processor 4833 are provided as separate units. As shown in FIG. 12, the processor 4831 is mounted on, for example, a substrate 4838. The parallel processing processor 4833 is mounted on, for example, a graphic board 4830, which is connected to the substrate 4838 via a connector 4839. The processor 4831 is connected to the parallel processing processor 4833 via a bus 485. As shown in FIG. 13, the parallel processing processor 4833 may be mounted directly on the substrate 4838 and connected to the processor 4831 via the bus 485. 14 shows an example of a mounting arrangement in which a processor 4831 and a parallel processing processor 4833 are integrally provided. As shown in FIG. 14, the parallel processing processor 4833 may be built into the processor 4831 mounted on a substrate 4838, for example.

[0080] FIG. 15 shows another example of mounting the parallel processing processor 4833 on the measuring unit 400. FIG. 15 shows an example in which the parallel processing processor 4833 is mounted on the measuring unit 400 by an external device 4800 connected to the measuring unit 400. The parallel processing processor 4833 is mounted on the external device 4800, which is, for example, a USB (Universal Serial Bus) device, and the USB device is connected to the bus 485 via an interface unit 487, thereby mounting the parallel processing processor 4833 on the cellular analyzer 4000. The USB device may be, for example, a small device such as a USB dongle. The interface unit 487 is, for example, a USB interface having a transfer speed of several hundred Mbps, more preferably a USB interface having a transfer speed of several Gbps to several tens of Gbps or more. For example, the external device 4800 on which the parallel processing processor 4833 is mounted may be a Neural Compute Stick 2 manufactured by Intel Corporation.

[0081] A plurality of parallel processing processors 4833 may be mounted on the cell analysis device 4000 by connecting a plurality of USB devices each equipped with a parallel processing processor 4833 to the interface unit 487. Since the parallel processing processor 4833 mounted on one USB device may have a smaller number of arithmetic units 4836 than a GPU or the like, the number of cores can be increased by adding a plurality of USB devices connected to the measurement unit 400.

[0082] 16, 17, and 18 show an overview of the arithmetic processing executed by the parallel processing processor 4833 under the control of analysis software 4832 running on the processor 4831. FIG. 16 shows an example configuration of the parallel processing processor 4833 that executes the arithmetic processing. The parallel processing processor 4833 includes multiple arithmetic units 4836 and a RAM 4837. The processor 4831, which executes the analysis software 4832, can instruct the parallel processing processor 4833 to execute at least a portion of the arithmetic processing required to analyze waveform data using the deep learning algorithm 60. The processor 4831 instructs the parallel processing processor 4833 to execute arithmetic processing related to waveform data analysis based on the deep learning algorithm. All or at least a portion of the waveform data corresponding to the signal detected by the FCM detection unit 410 is stored in the RAM 4834. The data stored in the RAM 4834 is transferred to the RAM 4837 of the parallel processing processor 4833. The data stored in the RAM 4834 is transferred to the RAM 4837 by, for example, DMA (Direct Memory Access). Each of the multiple arithmetic units 4836 of the parallel processing processor 4833 executes arithmetic processing on the data stored in the RAM 4837 in parallel. Each of the multiple arithmetic units 4836 obtains necessary data from the RAM 4837 and executes arithmetic processing. Data corresponding to the arithmetic results is stored in the RAM 4837 of the parallel processing processor 4833. The data corresponding to the arithmetic results is transferred from the RAM 4837 to the RAM 4834 by, for example, DMA.

[0083] FIG. 17 shows an overview of matrix operations executed by the parallel processing processor 4833. When analyzing waveform data according to the deep learning algorithm 60, matrix multiplication (matrix operation) is performed. The parallel processing processor 4833 executes, for example, multiple arithmetic processes related to matrix operations in parallel. (a) of FIG. 17 shows a formula for calculating matrix multiplication. In the formula shown in (a), matrix c is obtained by multiplying matrix a with n rows and n columns by matrix b with n rows and n columns. As shown in FIG. 17, the formula is written using a multi-layer loop syntax. (b) of FIG. 17 shows an example of arithmetic processes executed in parallel by the parallel processing processor 4833. The formula shown in FIG. 17(a) can be divided into n×n arithmetic processes, where n is the number of combinations of loop variable i in the first layer and loop variable j in the second layer. Each of the divided arithmetic processes is independent of each other and can therefore be executed in parallel.

[0084] Fig. 18 is a conceptual diagram showing that the multiple arithmetic processes illustrated in Fig. 17(b) are executed in parallel by the parallel processor 4833. As shown in Fig. 18, each of the multiple arithmetic processes is assigned to one of the multiple arithmetic units 4836 included in the parallel processor 4833. Each of the arithmetic units 4836 executes the assigned arithmetic processes in parallel with each other. In other words, each of the arithmetic units 4836 executes the divided arithmetic processes simultaneously.

[0085] 17 and 18, calculations are performed by the parallel processing processor 4833 to determine, for example, information regarding the probability that a cell corresponding to waveform data belongs to each of a plurality of cell types. Based on the calculation results, the processor 4831 executing the analysis software 4832 performs an analysis regarding the cell type of the cell corresponding to the waveform data. The calculation results are stored in the RAM 4837 of the parallel processing processor 4833 and transferred from the RAM 4837 to the RAM 4834. The processor 4831 transmits the results of the analysis based on the calculation results stored in the RAM 4834 to the processing unit 300 via the bus 485 and the interface section 489.

[0086] The calculation of the probability that a cell belongs to each of the plurality of cell types may be performed by a processor other than the parallel processing processor 4833. For example, the calculation results by the parallel processing processor 4833 may be transferred from the RAM 4837 to the RAM 4834, and the processor 4831 may calculate information regarding the probability that a cell corresponding to each waveform data belongs to each of the plurality of cell types based on the calculation results read from the RAM 4834. Alternatively, the calculation results by the parallel processing processor 4833 may be transferred from the RAM 4837 to the processing unit 300, and a processor mounted on the processing unit 300 may calculate information regarding the probability that a cell corresponding to each waveform data belongs to each of the plurality of cell types.

[0087] In this embodiment, the processing shown in FIGS. 17 and 18 is applied to, for example, calculation processing (also called filtering processing) related to a convolution layer in a deep learning algorithm 60.

[0088] FIG. 19 shows an overview of the computational processing related to the convolutional layer. (a) of FIG. 19 shows an example of waveform data of forward scattered light (FSC) as waveform data input to the deep learning algorithm 60. The waveform data in this embodiment is one-dimensional matrix data as shown in FIG. 2. More simply, the waveform data is an array in which elements are arranged in a row. For ease of explanation, the number of elements in the waveform data is assumed to be n (n is an integer equal to or greater than 1). (a) shows multiple filters. The filters are generated by the learning process of the deep learning algorithm 50. Each of the multiple filters is one-dimensional matrix data representing the characteristics of the waveform data. The filter shown in (a) is matrix data with one row and three columns, but the number of columns is not limited to three. By performing a matrix operation on the waveform data input to the deep learning algorithm 60 and each filter, characteristics of the waveform data corresponding to the cell type are calculated. (b) of FIG. 19 shows an overview of the matrix operation between the waveform data and the filters. As shown in (b), the matrix operation is performed while shifting each filter by one for each element of the waveform data. The matrix operation is calculated using the following equation 1.

number

[0089] The parallel processing processor 4833 executes the matrix operation expressed by Equation 1 in parallel using each of the multiple calculation units 4836. Classification information regarding the cell type of each cell is generated based on the calculation processing executed by the parallel processing processor 4833. The generated classification information is transmitted to the processing unit 300, which uses the classification information to generate and display test results for the specimen.

[0090] The measuring unit 400 can process the waveform data and the identification information in association with each other. Specifically, the measuring unit 400 can generate the analysis results of the waveform data (i.e., classification information regarding the cell type of each cell) in association with the identification information. The measuring unit 400, for example, associates the classification information regarding the cell type of each cell with the identification information and transmits them to the processing unit 300. Examples of the identification information include: (1) identification information of the biological sample corresponding to the waveform data; (2) identification information of the cell corresponding to the waveform data (e.g., the index described above); (3) identification information of the patient corresponding to the waveform data; (4) identification information of the test corresponding to the waveform data; (5) identification information of the cell analyzer that measured the waveform data; and (6) identification information of the facility, such as a hospital or testing facility, where the waveform data was measured (hereinafter referred to as "test-related facility"). The (1) biological sample identification information corresponding to the waveform data may include information for determining the priority of parallel processing, such as information regarding the time when the measurement order for the biological sample was registered, information regarding the time when the analyzer identified the biological sample, information regarding the time when the analyzer started measuring the biological sample, information for identifying whether the biological sample is an urgent sample or a regular sample, and information for identifying whether the biological sample is a remeasurement or a new measurement. For example, when receiving a test order from an LIS (Laboratory Information System) or the processing unit 300, the measuring unit 400 can acquire at least one or a combination of the above identification information (1) to (6) from the LIS or the processing unit 300. For example, at least one of the exemplified identification information (1) to (6) is associated with classification information and transmitted to the processing unit 300, and the result is provided to the user via the processing unit 300. A combination of the exemplified identification information (1) to (6) may also be associated with classification information and transmitted to the processing unit 300. The measuring unit 400 can also generate at least one or a combination of the above identification information (1) to (6) itself.

[0091] The processing unit 300 may associate the waveform data with the above-mentioned identification information (at least one of the above (1) to (6) or a combination thereof). For example, when the processing unit 300 receives a test order from the LIS, it acquires the identification information of the biological sample and the identification information of the patient. The processing unit 300 instructs the measurement unit 400 to perform a measurement corresponding to the above-mentioned test order. When the processing unit 300 acquires a result (i.e., waveform data) corresponding to the measurement instruction from the measurement unit 400, it associates the result (i.e., waveform data) with the above-mentioned identification information.

[0092] 6, the processing unit 300 is connected to the processor 4831 via the interface unit 489 and the bus 485, and can receive the analysis results from the processor 4831 and the parallel processing processor 4833. The interface unit 489 is, for example, a USB interface.

[0093] 10 is a diagram showing the configuration of the processing unit 300. The processing unit 300 includes a processor 3001, a bus 3003, a storage unit 3004, an interface unit 3006, a display unit 3015, and an operation unit 3016. The processing unit 300 is configured from a general personal computer as hardware, and functions as a processing unit of the cell analyzer 4000 by executing a dedicated program stored in the storage unit 3004.

[0094] The processor 3001 is a CPU and is capable of executing programs stored in the storage unit 3004 .

[0095] The storage unit 3004 includes a hard disk drive. The storage unit 3004 stores at least a program 60 for processing the cell classification information transmitted from the measurement unit 400 and generating specimen test results. The specimen test results refer to the results of counting blood cells contained in the specimen based on the analysis results 83 of the classification information 82 of individual cells obtained by the measurement unit 400, as will be described later.

[0096] The display unit 3015 includes a computer screen. The display unit 3015 is connected to the processor 3001 via the interface unit 3006 and the bus 3003. The display unit 3015 receives image signals input from the processor 3001 and can display the analysis results 83 received from the measurement unit 400 and the test results obtained by the processor 3001 analyzing the analysis results 83.

[0097] The operation unit 3016 is equipped with a pointing device including a keyboard, a mouse, or a touch panel. By operating the operation unit 3016, users such as doctors and laboratory technicians can input measurement orders to the cell analyzer 4000 and input measurement instructions according to the measurement orders. The operation unit 3016 can also accept instructions from the user to display test results. By operating the operation unit 3016, the user can view various information related to the test results, such as graphs, charts, and flag information assigned to samples.

[0098] The above-mentioned measuring unit 400 is connected to the processing unit 300 via the interface unit 3006. The processor 4831 of the measuring unit 400 can associate the classification information of each cell generated by the deep learning algorithm 60 with the specimen identification information and transmit it to the processor 3001 of the processing unit 300. The processor 3001 stores the cell analysis results 83 received from the measuring unit 400 in the memory unit 3004, associating it with the specimen identification information.

[0099] <Cell analyzer operation> The operation of analyzing a sample by the cell analyzer 4000 will be described with reference to FIGS.

[0100] When the processor 3001 of the processing unit 300 receives a measurement order and a measurement instruction from a user via the operation unit 3016, it transmits a measurement command to the measurement unit 400 (step S1).

[0101] Upon receiving the measurement command, the processor 4831 of the measurement unit 400 starts measuring the sample. The processor 4831 causes the sample aspirating section 450 to aspirate the sample from the blood collection tube T (step S10). Next, the processor 4831 causes the sample aspirating section 450 to dispense the aspirated sample into one of the reaction chambers 440a to 440e of the sample preparation section 440. The measurement command sent from the processing unit 300 in step S1 includes information about the measurement channel for which measurement is requested by the measurement order. The processor 4831 controls the sample aspirating section 450 to discharge the sample into the reaction chamber of the corresponding measurement channel based on the measurement channel information included in the measurement command.

[0102] Processor 4831 causes sample preparation unit 440 to prepare a measurement sample (step S11). Specifically, upon receiving a command from processor 4831, sample preparation unit 440 supplies reagents (hemolytic agent and staining solution) to the reaction chamber into which the specimen has been discharged, and mixes the specimen with the reagents. As a result, a measurement sample is prepared in which red blood cells are hemolyzed by the hemolytic agent in the reaction chamber and cells targeted by the measurement channel, such as white blood cells and reticulocytes, are stained by the staining agent.

[0103] The processor 4831 causes the FCM detection unit 410 to measure the prepared measurement sample (step S12). Specifically, the processor 4831 controls the device mechanism unit 430 to send the measurement sample in the reaction chamber of the sample preparation unit 440 to the FCM detection unit 410. The reaction chamber and the FCM detection unit 410 are connected by a flow path, and the measurement sample sent from the reaction chamber flows through the flow cell 4113 and is irradiated with laser light by the light source 4111 (see FIG. 9). When cells contained in the measurement sample pass through the flow cell 4113, the cells are irradiated with light, and forward scattered light, side scattered light, and side fluorescent light generated from the cells are detected by the light receiving elements 4116, 4121, and 4122, respectively, and an analog signal corresponding to the received light intensity is output. The analog signal is output to the A / D conversion unit 482 via the analog processing unit 420.

[0104] The A / D conversion unit 482 samples the analog signal at a predetermined rate to generate a digital signal including waveform data of each cell (step S13). The method of generating waveform data and digital signals has already been described. The processor 4831 loads the digital signal generated by the A / D conversion unit 482 into the RAM 4834. For example, the processor 4831 controls the bus controller 4850 to load the digital signal generated by the A / D conversion unit 482 into the RAM 4834 by DMA transfer. By DMA transfer, the digital signal is transferred directly to the RAM 4834 without going through the processor 4831. Specifically, the processor 4831 loads into the RAM 4834 digital signals of the forward scattered light signal, the side scattered light signal, and the fluorescent light signal, which are acquired from cells contained in the specimen to be tested. The digital signals are stored in the RAM 4834.

[0105] The processor 4831 performs cell classification based on the waveform data included in the generated digital signal using the deep learning algorithm 60 (step S14). The processing of S14 will be described later.

[0106] The processor 4831 transmits the analysis results 83, including the classification information 82 of each cell obtained as a result of S14, to the processing unit 300 in association with the specimen identification information (step S15). For example, the analysis results 83 of multiple cells contained in one specimen are each linked to the specimen identification information and transmitted to the processing unit 300.

[0107] When processor 3001 of processing unit 300 receives analysis result 83 from the measurement unit (step S2), it analyzes analysis result 83 using a program stored in memory 3004 and generates test results for the specimen (step S3). In the process of S3, for example, the number of cells for each cell type is counted based on the label value included in analysis result 83 for each cell. For example, if there are N pieces of classification information assigned with a label value "1" indicating neutrophils from one specimen, a count result in which the number of neutrophils = N is obtained as the test result for the specimen.

[0108] The processor 3001 acquires count results for the measurement items corresponding to the measurement channel based on the analysis results 83 and stores them in the memory unit 3004 along with the specimen identification information. The measurement items corresponding to the measurement channel are those for which count results are requested by the measurement order. For example, the measurement items corresponding to the DIFF channel are the five-part differential white blood cell count, i.e., the counts of monocytes, neutrophils, lymphocytes, eosinophils, and basophils. The measurement item corresponding to the RET channel is the count of reticulocytes. The measurement item corresponding to the PLT-F channel is the count of platelets. The measurement item corresponding to the WPC channel is the count of hematopoietic progenitor cells. The measurement item corresponding to the WNR channel is the count of white blood cells and nucleated red blood cells. The count results are not limited to the items for which measurement is requested (also called reportable items) listed above, but may also include count results for other cells that can be measured using the same measurement channel. For example, if the measurement channel is DIFF, as shown in FIG. 4, the count results include immature granulocytes (IG) and abnormal cells in addition to the five-part differential white blood cell count. Furthermore, the processor 3001 generates test results for the specimen by analyzing the obtained counting results and stores them in the memory unit 3004. Analysis of the counting results includes determining, for example, whether the counting results are within the normal range, whether abnormal cells have been detected, whether the deviation from the previous test results is within an acceptable range, and so on.

[0109] Processor 3001 displays the generated test results on display unit 3015 (step S4).

[0110] Next, the cell classification process of step S14 will be described with reference to Fig. 21. The cell classification process of step S14 is a process performed by the processor 4831 in accordance with the operation of the analysis software 4832. The processor 4831 transfers the digital signal taken into the RAM 4834 in step S13 to the parallel processing processor 4833 (S101). The processor 4831 transfers the digital signal from the RAM 4834 to the RAM 4837 by DMA transfer, as shown in Fig. 16. The processor 4831 controls, for example, the bus controller 4850 to cause the digital signal to be DMA transferred from the RAM 4834 to the RAM 4837.

[0111] The processor 4831 instructs the parallel processing processor 4833 to execute parallel processing on the waveform data included in the digital signal (S102). The processor 4831 instructs the parallel processing processor 4833 to execute parallel processing, for example, by calling a kernel function of the parallel processing processor 4833. The processing executed by the parallel processing processor 4833 will be described later using the flowchart illustrated in FIG. 22. The processor 4831 instructs the parallel processing processor 4833 to execute a matrix operation related to the deep learning algorithm 60, for example. The digital signal is decomposed into multiple waveform data, which are sequentially input to the deep learning algorithm 60. The indexes corresponding to each cell included in the digital signal are not input to the deep learning algorithm 60. The waveform data input to the deep learning algorithm 60 is operated on by the parallel processing processor 4833.

[0112] The processor 4831 receives the results of the calculations executed by the parallel processing processor 4833 (S103). The results of the calculations are transferred by DMA from the RAM 4837 to the RAM 4834, for example, as shown in FIG.

[0113] The processor 4831 generates analysis results for each cell type of each measured cell based on the calculation results by the parallel processing processor 4833 (S104).

[0114] FIG. 22 shows an example of the operation of the parallel processing processor 4833 which is executed based on instructions from the processor 4831 in response to the operation of the analysis software 4832.

[0115] The processor 4831 executing the analysis software 4832 causes the parallel processing processor 4833 to assign arithmetic operations to the arithmetic units 4836 (S110). The processor 4831 causes the parallel processing processor 4833 to assign arithmetic operations to the arithmetic units 4836, for example, by calling a kernel function of the parallel processing processor 4833. As shown in FIG. 18 , for example, a matrix operation related to the deep learning algorithm 60 is divided into multiple arithmetic operations, and each of the divided arithmetic operations is assigned to the arithmetic units 4836. Multiple waveform data are input sequentially to the deep learning algorithm 60. A matrix operation corresponding to the waveform data is divided into multiple arithmetic operations, and the divided arithmetic operations are assigned to the arithmetic units 4836.

[0116] Each arithmetic process is performed in parallel by a plurality of arithmetic units 4836 (S111). The arithmetic process is performed on a plurality of waveform data.

[0117] The operation results generated by parallel processing by the multiple operation units 4836 are transferred from the RAM 4837 to the RAM 4834 (S112). For example, the operation results are transferred from the RAM 4837 to the RAM 4834 by DMA.

[0118] (Configuration example 2) 23 and 24, another example of the configuration of a cell analyzer 4000 configured by a measurement unit 400 and a processing unit 300 will be described. In this example, the parallel processor is provided in the processing unit 300.

[0119] Fig. 23 shows another example of a block diagram of the measurement unit 400. The measurement unit 400 shown in Fig. 23 does not include the A / D conversion unit 482, the processor 4831, the RAM 4834, the storage unit 4835, or the parallel processing processor 4833, and has the same configuration and functions as the measurement unit 400 described in Fig. 6, Fig. 7 and the related descriptions, except that it is provided with a connection port 4201. A connection cable 4202 is connected to the connection port 4201.

[0120] FIG. 24 shows an example of a block diagram of the processing unit 300. As shown in FIG. 24, the processing unit 300 includes a processor 3001, a parallel processing processor 3002, a storage unit 3004, a RAM 3005, an interface unit 3006, an A / D conversion unit 3008, a bus controller 4850, and an interface unit 3009, all of which are connected to a bus 3003. In other words, in the example of FIG. 24, the parallel processing processor 3002 is incorporated into the processing unit 300 and is mounted on the cell analyzer 4000. The bus 3003 is, for example, a transmission line 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 transfers data based on, for example, the PCI-Express or PCI-X standard. The configurations of the processor 3001, parallel processing processor 3002, storage unit 3004, and RAM 3005, and the processes executed therein, are similar to those of the processor 4831, parallel processing processor 4833, storage unit 4835, and RAM 4834 described above in FIGS. 11 to 19. The A / D conversion unit 3008 samples the analog signal output from the measurement unit 400 as described above and generates a digital signal containing waveform data of the cell. The method of generating the digital signal is as described above. In the example of FIGS. 23 and 24, the connection cable 4202 has, for example, a number of transmission paths corresponding to the types of analog signals transmitted from the measurement unit 400 to the processing unit 300. For example, the connection cable 4202 is formed of a twisted pair cable and has a number of wire pairs corresponding to the types of analog signals transmitted to the processing unit 300. The transmission path from the connection port 3007 to the A / D conversion unit 3008 may also have a number of wires corresponding to the types of analog signals transmitted to the processing unit 300. In the transmission path from the connection port 3007 to the A / D conversion unit 3008, for example, the analog signal is transmitted as a differential signal.

[0121] 25, the processor 3001 and the parallel processing processor 3002 have the same configuration and functions as the above-described processor 4831 and parallel processing processor 4833. The parallel processing processor 3002 includes a plurality of arithmetic units 3200 and a RAM 3201. Analysis software 3100 that analyzes the cell type of measured cells runs on the processor 3001. Note that the parallel processing processor 3002 does not need to be directly connected to the bus 3003. For example, it may be implemented in a USB device, and this USB device may be connected to the bus 3003 via an interface unit (not shown), thereby being installed in the cell analysis apparatus 4000 as part of the processing unit 300. This USB device may be, for example, a small device such as a USB dongle.

[0122] 23 and 24, the processing unit 300 is connected to the interface section 489 of the measurement unit 400 via the interface section 3006. The processing unit 300 transmits control signals for the device mechanism section 430 and the sample preparation section 440 to the measurement unit 400 via the interface section 3006. The interface section 3006 is, for example, a USB interface.

[0123] 23 and 24 , the processing unit 300 is connected to a connection port 4201 of the measurement unit 400 via an interface unit 3006, a connection port 3007 connected to an A / D conversion unit 3008, and a connection cable 4202 connected to the connection port 3007. The connection port 4201 is connected to an analog processing unit 420. As described above, the analog signal output from the connection port 4201 to the processing unit 300 is a signal obtained by processing the output of the FCM detection unit 410 of the measurement unit 400 by the analog processing unit 420. The analog processing unit 420 performs processing, including noise removal, on the analog signal input from the FCM detection unit 410. The analog signal processed by the analog processing unit 420 is transmitted to the processing unit 300 via the connection port 4201 and the connection cable 4202. The connection cable 4202 is configured to be, for example, one meter or less in length to reduce noise during signal transmission. The analog signal is transmitted to the processing unit 300, for example as a differential signal, via a connection cable 4202. The processing unit 300 may include multiple connection ports 3007. The processing unit 300 may acquire analog signals from multiple measurement units 400 via the multiple connection ports 3007.

[0124] The analog signal transmitted from the measurement unit 400 via the connection cable 4202 is converted into a digital signal by the A / D converter 3008 of the processing unit 300. The A / D converter 3008 samples the transmitted analog signal at a predetermined sampling rate (e.g., sampling of 1024 points at 10 nanosecond intervals, sampling of 128 points at 80 nanosecond intervals, or sampling of 64 points at 160 nanosecond intervals), as described with reference to FIG. 2, to generate waveform data for each cell. The waveform data is stored in the memory unit 3004 or RAM 3005 via the bus 3003. The waveform data is transferred to the RAM 3005 by, for example, DMA. The processor 3001 and the parallel processing processor 3002 perform arithmetic processing on the waveform data stored in the memory unit 3004 or RAM 3005.

[0125] Analysis software 3100 running on processor 3001 has the same functions as analysis software 4832 shown in Fig. 16. By executing analysis software 3100, processor 3001 generates classification information regarding the cell type of the measured cells through operations similar to those shown in Figs. 16, 17, 18, 21, 22 and their related descriptions.

[0126] In the case of the cell analyzer 4000 of Configuration Example 2 shown in Figures 23 and 24, step S13 (digital signal generation) and step S14 (cell classification) of the flowchart shown in Figure 20 are performed in the processing unit 300. Step S15 (transmission of classification information) is omitted. The processing of Figures 21 and 22 is performed by the processor 3001 and parallel processing processor 3002 of the processing unit 300.

[0127] The processor 3001 can process the waveform data in association with the identification information. Specifically, the processor 3001 can associate the analysis results of the waveform data (i.e., classification information regarding the cell type of each cell) with the identification information. Examples of the identification information include: (1) identification information of the biological sample corresponding to the waveform data; (2) identification information of the cell corresponding to the waveform data; (3) identification information of the patient corresponding to the waveform data; (4) identification information of the test corresponding to the waveform data; (5) identification information of the cell analyzer that measured the waveform data; and (6) identification information of the test-related facility that measured the waveform data. For example, when receiving a test order from the LIS, the processor 3001 can acquire at least one or a combination of the above identification information (1) to (6) from the LIS. For example, at least one of the exemplified identification information (1) to (6) is associated with classification information and provided to the user as the test result. A plurality of combinations of the exemplified identification information (1) to (6) may be associated with classification information and provided to the user as the test result.

[0128] (Configuration example 3) 26 and 27, another configuration example of a cell analyzer 4000 configured by a measurement unit 400 and a processing unit 300 will be described. In this configuration example, the parallel processing processor 3002 is also installed in the cell analyzer 4000 by being incorporated inside the processing unit 300.

[0129] FIG. 26 shows another example of a block diagram of the measurement unit 400. The measurement unit 400 shown in FIG. 26 does not include a processor 4831, a RAM 4834, a storage unit 4835, or a parallel processing processor 4833. However, the measurement unit 400 has the same configuration and functions as the measurement unit 400 described in FIGS. 6 and 7 and their related descriptions, except that it includes an interface unit 4851 and a transmission path 4852 for transmitting a digital signal generated by the A / D conversion unit 482 to the processing unit 300. The interface unit 4851 is, for example, an interface serving as a dedicated line with a communication bandwidth of 1 gigabit / second or more. For example, the interface unit 4851 is an interface compliant with Gigabit Ethernet, USB 3.0, or Thunderbolt 3. If the interface unit 4851 is Gigabit Ethernet, the transmission path 4852 is, for example, a LAN cable. If the interface unit 4851 is USB 3.0, the transmission path 4852 is a USB cable compliant with USB 3.0. The transmission line 4852 is, for example, a dedicated transmission line for transmitting digital signals between the measurement unit 400 and the processing unit 300 .

[0130] Fig. 27 shows another example of a block diagram of the processing unit 300. The processing unit 300 shown in Fig. 27 has the same configuration and function as the processing unit 300 described in Fig. 24 and its related description, except that it does not include an A / D conversion unit 3008 and a connection port 3007 and includes an interface unit 3010. The processing unit 300 may be connected to a plurality of measurement units 400 via a plurality of interface units 3010 and a plurality of interface units 3006.

[0131] The processor 3001 and the parallel processing processor 3002 have the same configuration and functions as the processor 3001 and the parallel processing processor 3002 described in Fig. 25 and its related description. In Fig. 27, the parallel processing processor 3002 does not necessarily need to be directly connected to the bus 3003, but may be implemented in, for example, a USB device, and this USB device may be connected to the bus 3003 via an interface unit (not shown). This USB device may be, for example, a small device such as a USB dongle.

[0132] Analysis software 3100 running on processor 3001 has the same functions as analysis software 4832 shown in Fig. 16. Analysis software 3100 analyzes the cell type of the measured cells by performing operations similar to those shown in Fig. 16, Fig. 17, Fig. 18, Fig. 21, Fig. 22 and their related descriptions.

[0133] In the case of the cell analyzer 4000 of Configuration Example 3 shown in Figures 26 and 27, step S14 (cell classification) in the flowchart shown in Figure 20 is performed in the processing unit 300. Step S15 (transmission of classification information) is omitted. The processes in Figures 21 and 22 are performed by the processor 3001 and parallel processing processor 3002 of the processing unit 300.

[0134] In the configurations of FIGS. 26 and 27 , analog signals (forward scattered light signal, side scattered light signal, side fluorescent light signal) of cells generated in the FCM detection unit 410 are converted into digital signals in the A / D conversion unit 482 in the measuring unit 400. The digital signals are sent to the processing unit 300 via the interface unit 484, bus 485, interface unit 4851, and transmission path 4852. As described above, the transmission path 4852 is a dedicated transmission path for transmitting digital signals between the measuring unit 400 and the processing unit 300. For example, the measuring unit 400 and the processing unit 300 are connected one-to-one via the transmission path 4852. In other words, the transmission path 4852 is a transmission path that does not involve transmission of data related to devices other than the components (e.g., the measuring unit 400 and the processing unit 300) that make up the cell analysis device 4000. The transmission path 4852 is a transmission path that is separate from, for example, an intranet or the Internet. This makes it possible to avoid bottlenecks in the communication speed of digital signal transmission even when digital signals generated by A / D conversion within the measurement unit 400 are transmitted to the processing unit 300.

[0135] (Configuration Example 4) Another example of the configuration of the cell analyzer 4000 will be described with reference to FIGS. 28, 29, 30, 31, and 32. FIG.

[0136] In this configuration example, as illustrated in Fig. 28, an analysis unit 600 is provided between the measurement unit 400 and the processing unit 300. That is, in the configurations of Figs. 28, 29, 30, 31, and 32, the cell analyzer 4000 is configured to include the measurement unit 400, the processing unit 300, and the analysis unit 600. The analysis unit 600 analyzes the cell type of the measured cells. As will be described later, in this configuration example, the parallel processing processor 6002 is installed in the cell analyzer 4000 by being incorporated into the analysis unit 600.

[0137] The configuration of the measurement unit 400 illustrated in FIG. 29 has the same configuration and functions as the measurement unit 400 described in FIG. 26 and its related description. An analysis unit 600 is provided between the measurement unit 400 and the processing unit 300. The analysis unit 600 may be connected to multiple measurement units 400. The analysis unit 600 may be connected to multiple processing units 300. The interface unit 4851 is, for example, an interface having a communication bandwidth of 1 gigabit / second or more. For example, the interface unit 4851 is an interface compliant with Gigabit Ethernet, USB 3.0, or Thunderbolt 3. If the interface unit 4851 is Gigabit Ethernet, the transmission path 4852 is, for example, a LAN cable. If the interface unit 4851 is USB 3.0, the transmission path 4852 is a USB cable compliant with USB 3.0. As described above, the transmission path 4852 is a dedicated transmission path for transmitting digital signals between the measurement unit 400 and the processing unit 300. For example, the measurement unit 400 and the processing unit 300 are connected one-to-one via a transmission path 4852 .

[0138] FIG. 30 shows a configuration example of the analysis unit 600. The analysis unit 600 includes, for example, a processor 6001, a parallel processing processor 6002, a bus 6003, a storage unit 6004, a RAM 6005, and interface units 6006 and 6007, which are connected to the bus 6003. The bus 6003 is, for example, a transmission path having a data transfer rate of several hundred MB / s or more. The bus 6003 may also be a transmission path having a data transfer rate of 1 GB / s or more. The bus 3003 performs data transfer based on, for example, the PCI-Express or PCI-X standard. The analysis unit 600 may be connected to multiple measurement units 400 via multiple interface units 6006. When multiple measurement units 400 are provided, the analysis unit 600 may be connected to each of the measurement units 400 (for example, multiple measurement units 400 and multiple analysis units 600 are connected one-to-one).

[0139] As shown in FIG. 31 , the processor 6001 and the parallel processing processor 6002 have the same configuration and functions as the processor 4831 and the parallel processing processor 4833 described above. The parallel processing processor 6002 includes multiple arithmetic units 6200 and a RAM 6201. Analysis software 6100 runs on the processor 6001, analyzing the cell type of measured cells. The analysis software 6100 running on the processor 6001 has the same functions as the analysis software 4832 shown in FIG. 16 . The analysis software 6100 analyzes the cell type of measured cells by performing operations similar to those shown in FIGS. 16 , 17 , 18 , 21 , 22 , and their related descriptions. The analysis software 6100 transmits classification information of the measured cells to the processing unit 300 via an interface unit 6007. The interface unit 6007 is, for example, Ethernet (registered trademark) or USB. The interface unit 6007 may also be a wireless communication interface (for example, Wi-Fi (registered trademark) or Bluetooth (registered trademark)).

[0140] FIG. 32 shows an example configuration of a processing unit 300. The processing unit 300 shown in FIG. 32 does not need to include a parallel processing processor 3002 like the processing unit 300 shown in FIGS. 24 and 27. Furthermore, the analysis software 3100 shown in FIGS. 24 and 27 does not need to run on the processor 3001 shown in FIG. 32. The processing unit 300 receives the analysis results from the analysis unit 600 via an interface unit 3006. The interface unit 3006 is, for example, Ethernet or USB. The interface unit 3006 may also be an interface capable of wireless communication (for example, WiFi or Bluetooth).

[0141] In the configurations of Figures 29, 30, 31, and 32, analog signals (forward scattered light signal, side scattered light signal, side fluorescent light signal) of cells generated in the FCM detection unit 410 are converted into digital signals in the A / D conversion unit 482 in the measurement unit 400. Waveform data is sent to the analysis unit 600 via the interface unit 484, bus 485, interface unit 4851, and transmission path 4852. As described above, the interface unit 4851 is a dedicated interface for connecting the measurement unit 400 and the analysis unit 600, and connects the measurement unit 400 and the analysis unit 600 on a one-to-one basis. In other words, the transmission path 4852 is a transmission path that does not involve the transmission of data related to devices other than the components constituting the cell analysis device 4000 (e.g., the measurement unit 400 and the processing unit 300). The transmission path 4852 is a transmission path that is separate from, for example, an intranet or the Internet. This makes it possible to avoid bottlenecks in the communication speed of digital signal transmission even when digital signals generated by A / D conversion within the measurement unit 400 are transmitted to the processing unit 300.

[0142] In the case of the cell analyzer 4000 of configuration example 4 shown in Figures 29, 30, 31, and 32, step S14 (cell classification) and step S15 (transmission of classification information) of the flowchart shown in Figure 20 are performed in the analysis unit 600. That is, the digital signal generated in step S13 is transmitted from the measurement unit 400 to the analysis unit 600, and in step S15, the classification information is transmitted to the processing unit 300. The processing in Figures 21 and 22 is performed by the processor 6001 and parallel processing processor 6002 of the analysis unit 600.

[0143] (Configuration Example 5) 28, 33, and 34, another configuration example of the cell analyzer 4000 will be described. Like the above-mentioned configuration example 4, the cell analyzer of configuration example 5 also includes a measurement unit 400, a processing unit 300, and an analysis unit 600. The measuring unit 400 shown in FIG. 33 has the same function and configuration as the measuring unit 400 described in FIG. 23 and its related description. The measuring unit 400 shown in FIG. 33 is connected to the analyzing unit 600 via a connection cable 4202. For example, the connection cable 4202 is configured as a twisted pair cable, and has a number of wire pairs corresponding to the type of analog signal transmitted to the processing unit 300. The connection cable 4202 is configured with a length of, for example, 1 meter or less to reduce noise during signal transmission. The measuring unit 400 transmits analog signals to the analyzing unit 600 via the connection cable 4202.

[0144] The analysis unit 600 shown in Fig. 34 has the same functions and configuration as the analysis unit 600 described in Fig. 30 and its related description. That is, in the example of Fig. 34, a parallel processing processor 6002 is mounted on the cell analyzer 4000 so as to be incorporated into the analysis unit 600. The analysis unit 600 shown in Fig. 34 further includes a connection port 6008 and an A / D conversion unit 6009. An analog signal transmitted from the analysis unit 600 via the connection cable 4202 is input to the A / D conversion unit 6009 via the connection port 6008. The A / D conversion unit 6009 converts the analog signal into a digital signal by processing similar to that of the A / D conversion unit 482.

[0145] The analysis unit 600 may be connected to a plurality of measurement units 400 via a plurality of connection ports 6008. When a plurality of measurement units 400 are provided, the analysis unit 600 may be connected to each of the measurement units 400 (for example, the plurality of measurement units 400 and the plurality of analysis units 600 are connected one-to-one).

[0146] As shown in FIG. 31 , the processor 6001 and the parallel processing processor 6002 have the same configuration and functions as the processor 4831 and the parallel processing processor 4833 described above. Analysis software 6100 runs on the processor 6001, analyzing the cell type of the measured cells. The analysis software 6100 running on the processor 6001 has the same functions as the analysis software 4832 shown in FIG. 16 . The analysis software 6100 analyzes the cell type of the measured cells by performing operations similar to those shown in FIGS. 16 , 17 , 18 , 21 , 22 , and their related descriptions. The analysis software 6100 transmits the analysis results of the cell type of the measured cells to the processing unit 300 via an interface unit 6007. The interface unit 6007 is, for example, Ethernet or USB. The interface unit 6007 may also be a wireless communication interface (for example, Wi-Fi or Bluetooth).

[0147] In the case of the cell analyzer 4000 of Configuration Example 5 shown in Figures 33 and 34, step S13 (generation of digital signals), step S14 (cell classification), and step S15 (transmission of classification information) of the flowchart shown in Figure 20 are performed in the analysis unit 600. That is, generation of digital signals (step S13) is performed in the A / D conversion section 6009 of the analysis unit 600, cell classification based on the digital signals (step S14) is performed by the processor 6001 and the parallel processing processor 6002, and the classification information is transmitted from the analysis unit 600 to the processing unit 300 (step S15). The processing of Figures 21 and 22 is performed by the processor 6001 and the parallel processing processor 6002 of the analysis unit 600.

[0148] (Data size of waveform data and digital signals) Next, the data sizes of waveform data and digital signals will be described. In this embodiment, for example, sampling is performed at multiple time points at regular intervals for each of the forward scattered light analog signal (FSC), the side scattered light analog signal (SSC), and the side fluorescent light analog signal (SFL) for one cell. Examples of sampling rates include 1024 sampling points at 10 nanosecond intervals, 128 sampling points at 80 nanosecond intervals, or 64 sampling points at 160 nanosecond intervals. The amount of data is, for example, 2 bytes per sampling. For each of FSC, SSC, and SFL, an amount of data corresponding to the sampling rate (for a 1024-point rate, 2 bytes x 1024 = 2048 bytes) is acquired. This data amount is the amount of data per cell. In one measurement, for example, FSC, SSC, and SFL are measured for at least 100 cells. Furthermore, in one measurement, for example, FSC, SSC, and SFL may be measured for at least 1000 cells. Furthermore, in one measurement, for example, FSC, SSC, and SFL may be measured for approximately 10,000 to approximately 140,000 cells. Therefore, for example, if the number of cells measured in one measurement is 100,000 and the sampling rate is 1024, the data volume of each digital signal for FSC, SSC, and SFL will be 2 bytes × 1024 × 100,000 = 204,800,000 bytes, and the total data volume for FSC, SSC, and SFL will be 614,400,000 bytes.

[0149] Furthermore, FSC, SSC, and SFL are measured for each measurement channel. Therefore, for example, if the number of cells 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 SFL is 2 bytes × 1024 × 100,000 × 5 = 1,024,000,000 bytes, and the total data volume for FSC, SSC, and SFL is 3,072,000,000 bytes.

[0150] As a result, the capacity of the digital signal per specimen is, for example, several hundred megabytes to several gigabytes, and depending on the number of cells, sampling rate, and number of measurement channels, it can reach at least 1 gigabyte.

[0151] According to this embodiment, when analyzing a huge amount of digital signals ranging from several hundred megabytes to several gigabytes per sample, the analysis process using the deep learning algorithm 60 is completed within the cell analyzer 4000 or 4000' as described above, and the digital signals are not transmitted via the Internet or an intranet to an analysis server installed outside the cell analyzer 4000 or 4000'. This makes it possible to avoid a decrease in processing capacity that would occur due to an increase in communication load when transmitting digital signals from the cell analyzer 4000 or 4000' to the analysis server.

[0152] <Second cell analyzer and measurement of biological samples on the second cell analyzer> As a configuration example of the second cell analyzer 4000', an example of a block diagram is shown in which the measuring unit 500 is a urine particle analyzer or a body fluid analyzer equipped with a flow cytometer for measuring urine samples or body fluid samples.

[0153] FIG. 35 is an example block diagram of a measurement unit 500. In FIG. 35, the measurement unit 500 includes a specimen distribution section 501, a sample preparation section 502, an optical detection section 505, an amplifier circuit 550 that amplifies the output signal of the optical detection section 505 (the output signal amplified by the preamplifier), a filter circuit 506 that performs filtering on the output signal from the amplifier circuit 550, an A / D conversion section 507 that converts the output signal (analog signal) of the filter circuit 506 into a digital signal, a processor 4831, a parallel processing processor 4833, a RAM 4834, a bus controller 4850, a storage device 511a, and a LAN (Local Area Network) adapter 512. The processor 4831, the parallel processing processor 4833, the RAM 4834, the bus controller 4850, the storage device 511a, and the LAN adapter 512 are connected to a bus 508. A deep learning algorithm 60 and analysis software based on the deep learning algorithm 60 are stored in the storage device 511a.

[0154] As explained based on the example of the first cell analyzer 4000, the processor 4831 performs predetermined arithmetic processing on the digital signal output from the A / D conversion unit 507. The processor 4831 analyzes waveform data using the parallel processing processor 4833. Configuration examples of the processor 4831 and the parallel processing processor 4833 and the operation of the parallel processing processor 4833 are the same as those shown in the above-mentioned Figures 11 to 19 and their related descriptions.

[0155] The processing unit 300 is connected to the measurement unit 500 by, for example, a LAN cable via a LAN adapter 512, and the processing unit 300 generates inspection results based on the measurement data acquired by the measurement unit 500. The optical detection unit 505, the amplifier circuit 550, the filter circuit 506, the A / D conversion unit 507, the processor 4831, the parallel processing processor 4833, and the storage device 511a constitute an optical measurement unit 510 that measures the measurement sample and generates measurement data.

[0156] FIG. 36 shows the configuration of the optical detection section 505 of the measurement unit 500. In FIG. 36, a condenser lens 552 focuses laser light emitted from a semiconductor laser light source 553, which serves as a light source, onto a flow cell 551, and a focusing lens 554 focuses forward scattered light emitted from the solid components in the measurement sample onto a forward scattered light receiving section 555. Another focusing lens 556 focuses side scattered light and fluorescence emitted from the solid components onto a dichroic mirror 557. The dichroic mirror 557 reflects the side scattered light to a side scattered light receiving section 558 and transmits the fluorescence to a fluorescence receiving section 559. These optical signals reflect the characteristics of the solid components in the measurement sample. The forward scattered light receiving section 555, the side scattered light receiving section 558, and the fluorescence receiving section 559 then convert the optical signals into electrical signals and output a forward scattered light signal, a side scattered light signal, and a fluorescence signal, respectively. These outputs are amplified by a preamplifier and then provided to the next stage of processing. Furthermore, the forward scattered light receiving unit 555, the side scattered light receiving unit 558, and the fluorescent light receiving unit 559 can each be switched between low-sensitivity output and high-sensitivity output by switching the drive voltage. This sensitivity switching is performed by a processor 4831, which will be described later. In this embodiment, a photodiode is used as the forward scattered light receiving unit 555, and photomultiplier tubes may be used as the side scattered light receiving unit 558 and the fluorescent light receiving unit 559, or photodiodes may be used as the side scattered light receiving unit 558 and the fluorescent light receiving unit 559. The fluorescent light signal output from the fluorescent light receiving unit 559 is amplified by a preamplifier and then supplied to two branched signal channels. The two signal channels are each connected to the amplifier circuit 550 described above in FIG. 35. The fluorescent light input to one signal channel is amplified to high sensitivity by the amplifier circuit 550.

[0157] (Preparation of measurement samples) Figure 37 is a diagram showing the schematic functional configuration of the sample preparation unit 502 and optical detection unit 505 shown in Figure 35. The sample distribution unit 501 shown in Figures 35 and 37 includes an aspirating tube 517 and a syringe pump. The sample distribution unit 501 aspirates a sample (urine or body fluid) 00b through the aspirating tube 517 and dispenses it into the sample preparation unit 502. The sample preparation unit 502 includes a reaction tank 512u and a reaction tank 512b. The sample distribution unit 501 distributes a quantified amount of measurement sample to each of the reaction tanks 512u and 512b.

[0158] In the reaction vessel 512u, the dispensed biological sample is mixed with a first reagent 519u as a diluent and a third reagent 518u containing a dye. The pigment contained in the third reagent 518u stains the formed elements in the specimen. If the biological sample is urine, the sample prepared in this reaction vessel 512u is used as a first measurement sample for analyzing relatively large formed elements in urine, such as red blood cells, white blood cells, epithelial cells, and tumor cells. If the biological sample is a body fluid, the sample prepared in the reaction vessel 512u is used as a third measurement sample for analyzing red blood cells in the body fluid.

[0159] Meanwhile, in reaction chamber 512b, the distributed biological sample is mixed with second reagent 519b as a diluent and fourth reagent 518b containing a dye. As will be described later, second reagent 519b has a hemolytic effect. The pigment contained in fourth reagent 518b stains formed elements in the specimen. If the biological sample is urine, the sample prepared in reaction chamber 512b serves as a second measurement sample for analyzing bacteria in the urine. If the biological sample is a body fluid, the sample prepared in reaction chamber 512b serves as a fourth measurement sample for analyzing nucleated cells (white blood cells and large cells) and bacteria in the body fluid.

[0160] A tube extends from reaction chamber 512u to flow cell 551 of optical detection unit 505, allowing the measurement sample prepared in reaction chamber 512u to be supplied to flow cell 551. A solenoid valve 521u is provided at the outlet of reaction chamber 512u. A tube also extends from reaction chamber 512b, and this tube is connected to the middle of the tube extending from reaction chamber 512u. This allows the measurement sample prepared in reaction chamber 512b to be supplied to flow cell 551. A solenoid valve 521b is provided at the outlet of reaction chamber 512u.

[0161] A tube extending from the reaction chambers 512u and 512b to the flow cell 551 branches off just before the flow cell 551, and the branched end is connected to a syringe pump 520a. An electromagnetic valve 521c is provided between the syringe pump 520a and the branched point.

[0162] The tubes extend from the reaction vessels 512u and 512b, and further branch from the connection point to the branch point, which is connected to the syringe pump 520b. An electromagnetic valve 521d is provided between the branch point of the tube extending to the syringe pump 520b and the connection point.

[0163] Furthermore, a sheath fluid storage section 522 that stores sheath fluid is connected to the sample preparation section 502, and this sheath fluid storage section 522 is connected to the flow cell 551 by a tube. A compressor 522a is connected to the sheath fluid storage section 522, and when the compressor 522a is driven, compressed air is supplied to the sheath fluid storage section 522, and the sheath fluid is supplied from the sheath fluid storage section 522 to the flow cell 551.

[0164] Of the two types of suspensions (measurement samples) prepared in the reaction vessels 512u and 512b, the suspension in the reaction vessel 512u (the first measurement sample when the biological sample is urine, or the third measurement sample when the biological sample is a body fluid) is first introduced to the optical detection unit 505, where it forms a thin stream surrounded by sheath liquid in the flow cell 551, and is then irradiated with laser light. Thereafter, the suspension in the reaction vessel 512b (the second measurement sample when the biological sample is urine, or the fourth measurement sample when the biological sample is a body fluid) is similarly introduced to the optical detection unit 505, where it forms a thin stream in the flow cell 551, and is then irradiated with laser light. These operations are automatically performed by operating the solenoid valves 521a, 521b, 521c, 521d, the drive unit 503, etc. under the control of the processor 4831 (control unit).

[0165] The first to fourth reagents will be described in detail. The first reagent 519u is a reagent whose main component is a buffer, and contains an osmotic pressure compensator so that a stable fluorescent signal can be obtained without hemolyzing red blood cells, and is adjusted to 100 to 600 mOsm / kg so that the osmotic pressure is suitable for classification and measurement. It is preferable that the first reagent 519u does not have a hemolytic effect on red blood cells in urine.

[0166] Unlike the first reagent 519u, the second reagent 519b has a hemolytic effect. This is to increase the permeability of the fourth reagent 518b (described later) to the bacterial cell membrane, thereby accelerating the staining process. It also serves to shrink impurities such as mucus threads and red blood cell fragments. The second reagent 519b contains a surfactant to achieve the hemolytic effect. Various surfactants are used, including anionic, nonionic, and cationic surfactants, but cationic surfactants are particularly suitable. Because the surfactant can damage the bacterial cell membrane, the dye contained in the fourth reagent 518b can efficiently stain the bacterial nucleic acids. As a result, bacterial measurement can be performed with a short staining process.

[0167] In yet another embodiment, the second reagent 519b may achieve hemolysis by being adjusted to an acidic or low pH rather than using a surfactant. A low pH means a pH lower than that of the first reagent 19u. When the first reagent 519u is neutral or in the range of weakly acidic to weakly alkaline, the second reagent 19b is acidic or strongly acidic. When the pH of the first reagent 519u is 6.0 to 8.0, the pH of the second reagent 519b is lower, preferably 2.0 to 6.0.

[0168] The second reagent 519b may contain a surfactant and may be adjusted to a low pH.

[0169] In yet another embodiment, the second reagent 519b may have a lower osmotic pressure than the first reagent 19u, thereby achieving a hemolytic effect.

[0170] On the other hand, first reagent 519u does not contain a surfactant. In another embodiment, first reagent 519u may contain a surfactant, but the type and concentration of the surfactant must be adjusted so as not to hemolyze red blood cells. Therefore, first reagent 519u preferably does not contain the same surfactant as second reagent 519b, or, even if it does contain the same surfactant, it preferably has a lower concentration than second reagent 519b.

[0171] The third reagent 518u is a staining reagent used to measure urinary formed elements (red blood cells, white blood cells, epithelial cells, casts, etc.). The dye contained in the third reagent 518u is a membrane-staining dye selected to stain even formed elements that do not contain nucleic acids. The third reagent 518u preferably contains an osmotic pressure compensator to prevent hemolysis of red blood cells and to obtain stable fluorescence intensity, and its osmotic pressure is adjusted to 100 to 600 mOsm / kg to provide an appropriate osmotic pressure for classification and measurement. The cell membranes and nuclei (membranes) of urinary formed elements are stained by the third reagent 18u. A condensed benzene derivative, such as a cyanine dye, is used as a staining reagent containing a membrane-staining dye. The third reagent 18u is designed to stain not only cell membranes but also nuclear membranes. When the third reagent 518u is used, the staining intensity in the cytoplasm (cell membrane) of nucleated cells such as white blood cells and epithelium is combined with the staining intensity in the nucleus (nuclear membrane), resulting in a staining intensity higher than that of urinary formed elements that do not contain nucleic acids. This allows nucleated cells such as white blood cells and epithelium to be distinguished from urinary formed elements that do not contain nucleic acids, such as red blood cells. The reagent described in U.S. Pat. No. 5,891,733 can be used as the third reagent. U.S. Pat. No. 5,891,733 is incorporated herein by reference. The third reagent 518u is mixed with urine or a body fluid together with the first reagent 519u.

[0172] The fourth reagent 518b is a staining reagent that can accurately measure bacteria even in a sample containing contaminants of similar size to bacteria and fungi. A detailed description of the fourth reagent 518b is given in European Patent Publication No. 1,136,563. A dye that stains nucleic acids is preferably used as the dye contained in the fourth reagent 518b. For example, the cyanine dyes described in U.S. Patent No. 7,309,581 can be used as a staining reagent containing a dye that stains nuclei. The fourth reagent 518b is mixed with urine or a sample together with the second reagent 519b. European Patent Publication No. 1,136,563 and U.S. Patent No. 7,309,581 are incorporated herein by reference.

[0173] Therefore, it is preferable that the third reagent 518u contains a dye that stains cell membranes, while the fourth reagent 518b contains a dye that stains nucleic acids. Because urinary formed elements include those without nuclei, such as red blood cells, the third reagent 518u contains a dye that stains cell membranes, making it possible to detect urinary formed elements, including those without nuclei. Furthermore, because the second reagent can damage bacterial cell membranes, the dye contained in the fourth reagent 518b can efficiently stain the nucleic acids of bacteria and fungi. As a result, bacteria can be measured with a short staining process.

[0174] [3. First Waveform Data Analysis System] <Configuration of the first waveform data analysis system> Another embodiment relates to a waveform data analysis system. Referring to Fig. 38, a first waveform data analysis system includes a deep learning device 100A. A vendor-side device 100 operates as the deep learning device 100A. The deep learning device 100A trains a neural network 50 using training data and provides a deep learning algorithm 60 trained by the training data to a user. The deep learning algorithm 60, which is configured from the trained neural network, is provided from the deep learning device 100A to a measurement unit 400 via a recording medium 98 or a communication network 99. The measurement unit 400 analyzes the waveform data of the cells to be analyzed using the deep learning algorithm 60, which is configured from the trained neural network.

[0175] Deep learning device 100A is configured, for example, by a general-purpose computer, and performs deep learning processing based on a flowchart described later. Measurement unit 400 performs waveform data analysis processing based on a flowchart described later. Recording medium 98 is a computer-readable, non-transitory, tangible recording medium, such as a DVD-ROM or USB memory.

[0176] The deep learning device 100A is connected to a measurement unit 400a or a measurement unit 500a. The configuration of the measurement unit 400a or the measurement unit 500a is similar to that of the measurement unit 400 or the measurement unit 500 described above. The deep learning device 100A acquires training waveform data acquired by the measurement unit 400a or the measurement unit 500a. The method for generating the training waveform data is as described above.

[0177] 38 , the measuring unit 400a or the measuring unit 500a includes a flow cell 4113 or 551, respectively. The measuring unit 400a or the measuring unit 500a delivers a biological sample to the flow cell 4113 or 551. The biological sample supplied to the flow cell 4113 or 551 is irradiated with light from the light source 4112 or 553, and forward scattered light, side scattered light, and side fluorescent light emitted from cells in the biological sample are detected by light detection units (4116, 4121, 4122, 555, 558, 559). The measuring unit 400a or the measuring unit 500a generates waveform data from forward scattered light signals, side scattered light signals, and side fluorescent light signals obtained by light detection by the light detection units (4116, 4121, 4122, 555, 558, 559), and transmits the waveform data to the vendor-side device 100.

[0178] <Hardware configuration of deep learning device> 39 illustrates a block diagram of the vendor-side device 100 (deep learning device 100A). The vendor-side device 100 includes a processing unit 10 (10A), an input unit 16, and an output unit 17.

[0179] The processing unit 10 includes, for example, a CPU 11 that performs data processing (described later), a memory 12 used as a work area for data processing, a storage unit 13 that stores programs and processing data (described later), a bus 14 that transmits data between the various units, an interface unit (I / F unit) 15 that inputs and outputs data to and from external devices, and a GPU 19. The input unit 16 and output unit 17 are connected to the processing unit 10 via the interface unit 15. For example, the input unit 16 is an input device such as a keyboard or a mouse, and the output unit 17 is a display device such as a liquid crystal display. The GPU 19 functions as an accelerator that assists the arithmetic processing (e.g., parallel arithmetic processing) performed by the CPU 11. In other words, in the following description, the processing performed by the CPU 11 also includes processing performed by the CPU 11 using the GPU 19 as an accelerator. The GPU 19 has the same functions as the parallel processing processor 4833 described above. Here, a chip suitable for neural network calculations may be installed instead of the GPU 19. Examples of such chips include FPGA, ASIC, Myriad X (Intel), and the like.

[0180] Furthermore, in order to perform the processing of each step described below in FIG. 43, the processing unit 10 pre-records a deep learning algorithm 50, which is composed of a program according to this embodiment and a pre-trained neural network, in, for example, an executable format in the storage unit 13. The executable format is, for example, a format generated by conversion from a programming language by a compiler. The processing unit 10 uses the program recorded in the storage unit 13 to perform training processing of the pre-trained neural network 50.

[0181] In the following description, unless otherwise specified, the processing performed by the processing unit 10 means the processing performed by the CPU 11 based on the programs and neural network 50 stored in the storage unit 13 or memory 12. The CPU 11 uses the memory 12 as a working area to temporarily store necessary data (intermediate data during processing, etc.), and records data to be stored for a long period of time, such as calculation results, in the storage unit 13 as appropriate.

[0182] <Hardware configuration of the analyzer> The configuration of the measurement unit 400 or 500 that processes waveform data based on the algorithm provided by the vendor-side device 100 is equivalent to the configuration described above. The measurement unit 400 or 500 may also have the functions of the vendor-side device 100 (deep learning device 100A) and train the neural network 50 using training data. In this case, the vendor-side device 100 (deep learning device 100A) is not required.

[0183] Furthermore, in order to perform the processing of each step described in the waveform data analysis process below, the measurement unit control unit 480 has a deep learning algorithm 60, which is composed of a program according to this embodiment and a trained neural network, pre-recorded in, for example, an executable format in the storage unit 4835. The executable format is, for example, a format generated by conversion from a programming language using a compiler. The measurement unit control unit 480 performs processing using the program and deep learning algorithm 60 recorded in the storage unit 4835.

[0184] As shown in FIG. 40, the measurement unit control unit 480 may update the program and deep learning algorithm 60 recorded in the memory unit 4835, for example, via a LAN adapter 481 via a communication network. As shown in FIG. 41 , the above-described analysis unit 600 may transmit the digital signal and analysis results received from the measurement unit 400 to the deep learning device 100A via the interface unit 6011. The analysis unit 600 transmits waveform data and classification information corresponding to the waveform data to the deep learning device 100A, for example, via the Internet. The deep learning device 100A performs a learning process based on the waveform data transmitted from the analysis unit 600 and the classification information corresponding to the waveform data, and updates the deep learning algorithm 60. The deep learning device 100A transmits the updated deep learning algorithm 60 to the analysis unit 600. The analysis unit 600 updates the algorithm stored in the memory unit 6004 using the deep learning algorithm 60 transmitted from the deep learning device 100A. For example, while the processor 6001 and the parallel processing processor 6002 are processing the waveform data transmitted from the measurement unit 400, the analysis unit 600 transmits the waveform data to the deep learning device 100A in parallel. Upon completing the analysis of the waveform data, i.e., the generation of the classification information, the analysis unit 600 transmits the classification information to the deep learning device 100A.

[0185] In the following description, unless otherwise specified, the processing performed by the measurement unit control unit 480 means processing actually performed by the processor 4831 and the parallel processing processor 4833 based on the program and deep learning algorithm 60 stored in the memory unit 4835 or RAM 4834. The processor 4831 temporarily stores necessary data (intermediate data during processing, measurement results, analysis results, etc.) using the RAM 4834 as a working area, and appropriately records data to be stored for a long period of time, such as calculation results, in the memory unit 4835.

[0186] <Function blocks and processing procedures> (Deep learning processing) FIG. 42 shows an example of functional blocks of a deep learning device 100A that performs deep learning. Referring to FIG. 42, a processing unit 10A of the deep learning device 100A according to this embodiment includes a training data generation unit 101, a training data input unit 102, and an algorithm update unit 103. These functional blocks are realized by installing a program that causes a computer to execute deep learning processing in the storage unit 13 or memory 12 of the processing unit 10A shown in FIG. 39 and executing this program with the CPU 11 and GPU 19. A training data database (DB) 104 and an algorithm database (DB) 105 are recorded in the storage unit 13 or memory 12 of the processing unit 10A.

[0187] The training waveform data 76a, 76b, and 76c are acquired in advance by, for example, the measurement units 400 and 500, and are stored in advance in the storage unit 13 or the memory 12 of the processing unit 10A. The deep learning algorithm 50 is stored in the algorithm database 105 in advance.

[0188] The processing unit 10A of the deep learning device 100A performs the processing shown in Fig. 43. Explaining this using the functional blocks shown in Fig. 42, the processing of steps S211, S214, and S216 shown in Fig. 43 is performed by the training data generation unit 101. The processing of step S212 is performed by the training data input unit 102. The processing of steps S213 and S215 is performed by the algorithm update unit 103.

[0189] An example of deep learning processing performed by the processing unit 10A will be described using FIG. 43 . First, the processing unit 10A acquires training waveform data 72a, 72b, and 72c. The training waveform data 72a is waveform data of forward scattered light, the training waveform data 72b is waveform data of side scattered light, and the training waveform data 72c is waveform data of side fluorescent light. The training waveform data 72a, 72b, and 72c are acquired, for example, by an operator, by being imported from the measurement units 400 and 500, by being imported from the recording medium 98, or by being acquired via a communication network through the I / F unit 15. When acquiring the training waveform data 72a, 72b, and 72c, information regarding which cell type the training waveform data 72a, 72b, and 72c indicate is also acquired. The information regarding which cell type the training waveform data 72a, 72b, and 72c indicate may be linked to the training waveform data 72a, 72b, and 72c, or may be input by the operator through the input unit 16.

[0190] In step S211, the processing unit 10A generates training data 75 from the acquired waveform data 72a, 72b, and 72c and label values 77.

[0191] In step S212, the processing unit 10A inputs the training data 75 to the neural network 50 and acquires trial results. The trial results are accumulated every time a plurality of training data 75 are input to the neural network 50.

[0192] In the cell type analysis method according to this embodiment, a convolutional neural network is used, and the stochastic gradient descent method is employed, so in step S213, the processing unit 10A determines whether a predetermined number of trial results have been accumulated. If the predetermined number of trial results have been accumulated (YES), the processing unit 10A proceeds to step S214, and if the predetermined number of trial results have not been accumulated (NO), the processing unit 10A proceeds to step S215.

[0193] Next, when a predetermined number of trial results have been accumulated, in step S214, processing unit 10A updates the connection weights w of neural network 50 using the training results accumulated in step S212. Since the cell-type analysis method according to this embodiment uses stochastic gradient descent, the connection weights w of neural network 50 are updated when a predetermined number of trial results have been accumulated. Specifically, the process of updating the connection weights w is a process of performing calculations using the gradient descent method shown in (Equation 12) and (Equation 13) described below.

[0194] In step S215, the processing unit 10A determines whether the neural network 50 has been trained with a prescribed number of training data 75. If training has been performed with the prescribed number of training data 75 (YES), the deep learning process ends.

[0195] If the neural network 50 has not been trained with the specified number of training data 75 (NO), the processing unit 10A proceeds from step S215 to step S216, and performs the processes from step S211 to step S215 on the next training waveform data.

[0196] According to the process described above, the neural network 50 is trained to obtain the deep learning algorithm 60.

[0197] (Neural network structure) As described above, in this embodiment, a convolutional neural network is used. An example of the structure of a neural network 50 is shown in FIG. 44(a). The neural network 50 includes an input layer 50a, an output layer 50b, and a middle layer 50c between the input layer 50a and the output layer 50b, and the middle layer 50c is composed of multiple layers. The number of layers constituting the middle layer 50c can be, for example, 5 or more, preferably 50 or more, and more preferably 100 or more.

[0198] In the neural network 50, multiple nodes 89 are arranged in layers and connected between layers, allowing information to propagate in only one direction, as indicated by arrow D in the figure, from the input-side layer 50a to the output-side layer 50b.

[0199] (operation at each node) (b) of Figure 44 is a schematic diagram showing the calculations at each node. Each node 89 receives multiple inputs and calculates one output (z). In the example shown in (b) of Figure 44, 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, one-dimensional matrix data is used as the training data 75 and analysis data 85, so if the variables of the calculation formula correspond to two-dimensional matrix data, processing is performed to convert the variables so that they correspond to one-dimensional matrix data.

number

number

number

number

number

number

number

[0200] Learning a neural network means adjusting the weights w so that the output y(xn:w) of the neural network when given an input xn is as close as possible to the output dn for any input / output pair (xn, dn). The error function is the closeness of the function expressed using a neural network to the training data.

number

number

number

[0201] If each class is represented as C1,...,CK, the output yK of node k in the output layer L (i.e., uk (L) ) represents the probability that a given input x belongs to class CK. See (Equation 9) below. Input x is classified into the class that maximizes the probability expressed by (Equation 9).

number

[0202] The target output dn by the softmax function in (Equation 8) is set to 1 only if the output is the correct class, and to 0 otherwise. If the target output is expressed in vector form as dn = [dn1, , dnK], for example, if the correct class of the input xn is C3, 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 as follows (Equation 10).

number

number

[0203] Minimizing the error function E(w) with respect to the parameter w is equivalent to finding a local minimum of the function E(w). The parameter w is the weight of the connection between nodes. The minimum of the weight w is found by iterative calculations that start with an arbitrary initial value and repeatedly update the parameter w. An example of such calculations is the gradient descent method.

[0204] The gradient descent method uses a vector expressed by the following (Equation 12).

number

number

number

[0205] The calculation using (Equation 13) may be performed on all training data (n=1, . . . , N) or on only a portion of the training data. A gradient descent method performed on only a portion of the training data is called a stochastic gradient descent. The cell type analysis method according to the embodiment uses the stochastic gradient descent method.

[0206] [4. Building a Deep Learning Model] Blood samples collected from healthy individuals were measured, and XN CHECK Lv2 (Streck's control blood, which had been fixed and processed) was used as the unhealthy blood sample. Fluorocell WDF (Sysmex Corporation) was used as the fluorescent staining reagent. Lysercell WDF (Sysmex Corporation) was used as the hemolytic agent. For each cell in each biological sample, waveform data for forward scattered light, side scattered light, and side fluorescent light were acquired at 10 nanosecond intervals from the start of forward scattered light measurement. For the healthy blood samples, waveform data from blood cells collected from eight healthy individuals were pooled as digital signals. The waveform data for each cell was manually classified into neutrophils (NEUT), lymphocytes (LYMPH), monocytes (MONO), eosinophils (EO), basophils (BASO), and immature granulocytes (IG), and each waveform data was annotated (labeled) by cell type. The measurement start point was the point when the signal strength of forward scattered light exceeded a threshold, and the waveform data acquisition points for forward scattered light, side scattered light, and side fluorescent light were synchronized to generate training data. Control blood samples were also annotated as control blood-derived cells (CONT). The training data was input into a deep learning algorithm and trained.

[0207] Analytical waveform data was obtained using the Sysmex XN-1000 from blood cells from healthy individuals, separate from the cell data used for learning. Waveform data from control blood was mixed to create analytical data. This analytical data was then input into the deep learning algorithm, and data for each individual cell type was obtained.

[0208] The results are shown as a confusion matrix in Figure 45. The horizontal axis shows the judgment results using the constructed deep learning algorithm, and the vertical axis shows the manual judgment results (reference method) by a human. Although there was some confusion between basophils and lymphocytes, and between basophils and ghosts, the judgment results using the constructed deep learning algorithm showed a 98.8% agreement rate with the judgment results using the reference method.

[0209] Next, ROC analysis was performed to evaluate sensitivity and specificity for each cell type. Figure 46(a) shows the ROC curves for neutrophils, Figure 46(b) shows the ROC curves for lymphocytes, Figure 46(c) shows the ROC curves for monocytes, Figure 47(a) shows the ROC curves for neutrophils, Figure 47(b) shows the ROC curves for basophils, and Figure 47(c) shows the ROC curves for control blood (CONT). The sensitivity and specificity were 99.5% and 99.6%, respectively, for neutrophils, 99.4% and 99.5%, respectively, for lymphocytes, 98.5% and 99.9%, respectively, for monocytes, 97.9% and 99.8%, respectively, for eosinophils, 71.0% and 81.4%, respectively, for basophils, and 99.8% and 99.6%, respectively, for control blood (CONT), demonstrating good results.

[0210] These results demonstrate that it is possible to determine cell types with high classification accuracy by using deep learning algorithms based on signals obtained from cells contained in biological samples based on waveform data.

[0211] Furthermore, when unhealthy blood cells, such as those in control blood, are mixed with healthy blood cells, it is sometimes difficult to determine the type of individual cells using conventional scattergram methods. However, by using the deep learning algorithm of this embodiment, it has been shown that it is possible to determine the type of individual cells even when unhealthy blood cells are mixed with healthy blood cells.

[0212] [5. Analysis system using image analysis equipment] An embodiment will be described in which an image analyzer is used as the cell analyzer. The third cell analyzer 4000″, which is an image analyzer, analyzes captured image data to estimate the cell type of the captured cells.

[0213] FIG. 48 shows an example of the configuration of a cell analyzer 4000″. The cell analyzer 4000″ shown in FIG. 48 includes a measurement unit 700 and a processing unit 800, and measures and analyzes a sample 901 prepared by pretreatment using a pretreatment device 900.

[0214] The measurement unit 700 includes a flow cell 710, light sources 720 to 723, condenser lenses 730 to 733, dichroic mirrors 740 to 741, a condenser lens 750, an optical unit 751, a condenser lens 752, and an imaging section 760. A sample 701 flows through a flow path 711 of the flow cell 710.

[0215] Light sources 720 to 723 irradiate light onto sample 701 flowing through flow cell 710. Light sources 720 to 723 are configured by, for example, semiconductor laser light sources. Light sources 720 to 723 emit light of wavelengths λ11 to λ14, respectively. The condenser lenses 730 to 733 condense the light of wavelengths λ11 to λ14 emitted from the light sources 720 to 723, respectively. The dichroic mirror 740 transmits the light of wavelength λ11 and refracts the light of wavelength λ12. The dichroic mirror 741 transmits the light of wavelengths λ11 and λ12 and refracts the light of wavelength λ13. In this way, the light of wavelengths λ11 to λ14 is irradiated onto the sample 701 flowing through the flow path 711 of the flow cell 710. The number of semiconductor laser light sources provided in the measurement unit 700 is not limited as long as it is one or more. The number of semiconductor laser light sources can be selected from, for example, 1, 2, 3, 4, 5, or 6.

[0216] If the sample 701 flowing through the flow cell 710 is stained with a fluorescent dye, when the sample 701 is irradiated with light of wavelengths λ11 to λ13, the fluorescent dye that stains the cells will produce fluorescence. For example, fluorescence of wavelengths λ21, λ22, and λ23 corresponding to wavelengths λ11, λ12, and λ13, respectively, will be produced. When the sample 701 flowing through the flow cell 710 is irradiated with light of wavelength λ14, this light will pass through the cells. The transmitted light of wavelength λ14 that has passed through the cells will be used to generate a bright-field image.

[0217] The condensing lens 750 condenses the fluorescence generated from the sample 701 flowing through the flow path 711 of the flow cell 710 and the transmitted light that has passed through the sample 701 flowing through the flow path 711 of the flow cell 710. The optical unit 751 has a configuration in which, for example, four dichroic mirrors are combined. The four dichroic mirrors of the optical unit 751 reflect the fluorescence and the transmitted light at slightly different angles, causing them to be separated on the light receiving surface of the imaging section 760. The condensing lens 752 condenses the fluorescence and the transmitted light.

[0218] The imaging section 760 is configured with a TDI (Time Delay Integration) camera. The imaging section 760 captures fluorescence and transmitted light, and can output a fluorescence image corresponding to the fluorescence and a bright-field image corresponding to the transmitted light to the processing unit 800 as imaging signals.

[0219] The processing unit 800 includes, as its hardware configuration, a processing unit 811, a storage unit 812, an interface unit 816, and a bus 815. The processing unit 811, the storage unit 812, and the interface unit 816 are connected to the bus 815. Image data (e.g., a fluorescent image or a bright-field image) formed from an imaging signal captured by the imaging unit 760 of the measurement unit 700 is stored in the storage unit 812 via the interface unit 816. The processing unit 811 reads out the image data from the storage unit 812 and analyzes the image data.

[0220] Fig. 49 shows an example configuration of the processing unit 811. The processing unit 811 includes, for example, a processor 8111, a parallel processing processor 8112, and a RAM 8113. The processor 8111, the parallel processing processor 8112, and the RAM 8113 have the same configurations and functions as the above-described processor 4831, the parallel processing processor 4833, and the RAM 4834, respectively. The processor 8111 and the parallel processing processor 8112 analyze image data captured by the imaging unit 760. The parallel processing processor 8112 executes arithmetic processing of matrix data related to image data in parallel using multiple arithmetic units, for example, by the processing exemplified in Figs. 16, 17, and 18.

[0221] The functions of the processing unit 811 (processor 8111, parallel processing processor 8112, RAM 8113) do not necessarily have to be provided in the processing unit 800, but may be provided in the measurement unit 700.

[0222] <Generating training data> An example of generating training data in this embodiment will be described below.

[0223] Training images used to train deep learning algorithms are preferably captured in RGB and CMY colors, etc. Color images preferably represent the shade or brightness of each primary color, such as red, green, and blue, or cyan, magenta, and yellow, using 24-bit values (8 bits x 3 colors). Training images may contain at least one hue and its shade or brightness, but more preferably at least two hues and their respective shades or brightness. Information including a hue and its shade or brightness is also referred to as color tone.

[0224] The color tone information of each pixel in the training images is converted, for example, from RGB color to a format including brightness information and hue information. Examples of formats including brightness information and hue information include YUV (YCbCr, YPbPr, YIQ, etc.). Here, conversion to YCbCr format will be described as an example. Training images captured in RGB color are converted into image data based on brightness, image data based on a first hue (e.g., blue-based), and image data based on a second hue (e.g., red-based). Conversion from RGB to YCbCr can be performed using a known method. For example, conversion from RGB to YCbCr can be performed in accordance with the international standard ITU-R BT.601. The image data based on brightness, image data based on a first hue, and image data based on a second hue can be represented as matrix data of grayscale values, as shown in FIG. 50 (hereinafter also referred to as color tone matrix data 72y, 72cb, and 72cr). The image data based on the luminance, the image data based on the first hue, and the image data based on the second hue are each expressed, for example, in 256 gradations ranging from 0 to 255. Here, instead of the luminance, the first hue, and the second hue, the training images may be converted using the three primary colors of red R, green G, and blue B, or the three primary colors of cyan C, magenta M, and yellow Y.

[0225] Next, based on the color tone matrix data 72y, 72cb, 72cr, color tone vector data 74 is generated for each pixel by combining three tone values, ie, brightness 72y, first hue 72cb, and second hue 72cr.

[0226] Next, for example, if segmented neutrophils are captured in the training images, each piece of color tone vector data 74 generated from the training images is assigned a label value 77 of "1" indicating that it is a segmented neutrophil, resulting in training data 75. For convenience, in Figure 50, training data 75 is represented as 3 pixels x 3 pixels, but in reality, there is as much color tone vector data as there are pixels when the training images were captured.

[0227] 51 shows an example of the label value 77. Different label values 77 are assigned depending on the cell type and the presence or absence of characteristics of each cell.

[0228] <Deep Learning Overview> An overview of neural network training will be explained using Figure 50 as an example. The neural network 50 is preferably a convolutional neural network. The number of nodes in the input layer 50a of the neural network 50 corresponds to the product of the number of pixels in the input training data 75 and the number of luminances and hues contained in the image (for example, in the above example, there are three: luminance 72y, first hue 72cb, and second hue 72cr). Hue vector data 74 is input as a set 76 to the input layer 50a of the neural network 50. The neural network 50 is trained using label values 77 of each pixel in the training data 75 as the output layer 50b of the neural network.

[0229] The neural network 50 extracts feature quantities for morphological cell types and cell characteristics based on the training data 75. The output layer 50b of the neural network outputs results that reflect these feature quantities.

[0230] Reference numeral 50c in FIG. 50 denotes an intermediate layer.

[0231] The deep learning algorithm 60 having the thus trained neural network 60 is used as a classifier to identify which of multiple cell types the analyzed cell belongs to and is morphologically classified as belonging to a predetermined cell group.

[0232] <Image analysis method> An example of an image analysis method is shown in Figure 52. In the image analysis method, analysis data 81 is generated from an analysis image obtained by capturing an image of cells to be analyzed. The analysis image is an image obtained by capturing an image of cells to be analyzed.

[0233] For example, in this embodiment, the imaging device preferably captures images in RGB color, CMY color, or the like. A color image preferably represents the shading or brightness of each primary color, such as red, green, and blue, or cyan, magenta, and yellow, as a 24-bit value (8 bits x 3 colors). The analysis image may contain at least one hue and its shading or brightness, but more preferably contains at least two hues and their respective shading or brightness. Information including a hue and its shading or brightness is also referred to as color tone.

[0234] For example, RGB color is converted into a format including luminance information and hue information. Examples of formats including luminance information and hue information include YUV (YCbCr, YPbPr, YIQ, etc.). Here, conversion to YCbCr format is described as an example. Here, RGB color training images are converted into image data based on luminance, image data based on a first hue (e.g., blue-based), and image data based on a second hue (e.g., red-based). Conversion from RGB to YCbCr can be performed using a known method. For example, conversion from RGB to YCbCr can be performed in accordance with the international standard ITU-R BT.601. Image data corresponding to the luminance, first hue, and second hue can be represented as matrix data of gradation values (hereinafter also referred to as color matrix data 79y, 79cb, and 79cr), as shown in FIG. 52. The luminance, first hue, and second hue 72Cr are each represented in 256 gradations ranging from 0 to 255. Here, instead of the luminance, first hue, and second hue, the training images may be converted using the three primary colors of red R, green G, and blue B, or the three primary colors of cyan C, magenta M, and yellow Y.

[0235] Next, based on the color tone matrices 79y, 79cb, and 79cr, color tone vector data 80 is generated for each pixel by combining three gradation values: luminance 79y, first hue 79cb, and second hue 79cr. A collection of color tone vector data 80 generated from one analysis image is generated as analysis data 81.

[0236] It is preferable that the generation of the analysis data 81 and the generation of the training data 75 be the same at least in terms of the imaging conditions and the conditions for generating the vector data to be input from each image to the neural network.

[0237] Analysis data 81 is input to the input layer 60a of the neural network 60 that constitutes the trained deep learning algorithm 60. The deep learning algorithm extracts features from the analysis data 81 and outputs the results from the output layer 60b of the neural network 60. The values output from the output layer 60b are the probabilities that the cells to be analyzed, contained in the analysis image, belong to each of the morphological cell classifications and features input as training data.

[0238] Among these probabilities, it is determined that the cell to be analyzed contained in the image for analysis belongs to the morphological classification with the highest value, and a label value associated with the morphological cell type or cell characteristics is output. The label value itself, or data in which the label value has been replaced with information (e.g., terminology) indicating the presence or absence of the morphological cell type or cell characteristics, is output as the analysis result 83 regarding the cell morphology. In FIG. 52, from the analysis data 81, the classifier outputs the label value "1" as the most likely label value 82, and the text data "segmented neutrophil" corresponding to this label value is output as the analysis result 83 regarding the cell morphology.

[0239] Reference numeral 60c in FIG. 52 denotes an intermediate layer.

[0240] [6. Other forms] Although the present invention has been described above with reference to the outline and specific embodiments, the present invention is not limited to the outline and each embodiment described above.

[0241] In the above embodiment, the functional blocks of the training data generation unit 101, the training data input unit 102, the algorithm update unit 103, the analysis data generation unit 201, the analysis data input unit 202, and the analysis unit 203 are executed by a single processor 4831 and a single parallel processing processor 4833, but these functional blocks do not necessarily have to be executed by a single processor and a parallel processing processor, and may be executed in a distributed manner across multiple processors and multiple parallel processing processors.

[0242] In the above embodiment, a program for performing the processing of each step described in Fig. 43 is pre-recorded in the storage unit 4835. Alternatively, the program may be installed in the measurement unit control unit 480 from a computer-readable, non-transitory, tangible recording medium 98, such as a DVD-ROM or a USB memory. Alternatively, the measurement unit control unit 480 may be connected to a communication network 99, and the program may be downloaded from, for example, an external server (not shown) via the communication network 99 and installed.

[0243] FIG. 53 shows an embodiment of the analysis results. FIG. 53 shows the cell types and cell counts of each cell type assigned the label values shown in FIG. 4 contained in a biological sample measured by flow cytometry. Instead of or in addition to displaying the cell count, the percentage (e.g., %) of each cell type in the total number of counted cells may be output. The cell count can be calculated by multiplying the number of label values (the number of cells with the same label value) corresponding to each output cell type. The output results may also include a warning indicating the presence of abnormal cells in the biological sample. While FIG. 53 shows an example in which an exclamation mark is added to the abnormal cell column as a warning, this is not limiting. Furthermore, the distribution of each cell type may be output as a scattergram. When outputting as a scattergram, the highest value obtained from the signal intensity may be plotted, for example, with the side fluorescence intensity on the vertical axis and the side scattered light intensity on the horizontal axis.

[0244] [Additional Notes] The present invention also includes the following aspects.

[0245] [Aspect 1] a sample preparation unit that mixes a specimen with a reagent to prepare a measurement sample; a detection unit that passes the measurement sample through a flow cell and acquires matrix data having elements that indicate signal intensities at multiple time points for each of multiple cells contained in the measurement sample that pass through the flow cell; a parallel processor that executes parallel processing of matrix operations included in a deep learning algorithm that outputs information about the cell type of each of the plurality of cells in response to input of the matrix data; a processor that causes the sample preparation unit to prepare the measurement sample, causes the detection unit to acquire the matrix data, and causes the parallel processing processor to execute the parallel processing; A cell analysis device comprising:

[0246] [Aspect 2] the detection unit includes a light source that irradiates light onto the cells passing through the flow cell, and a light receiving unit that receives the light emitted from the cells, and acquires matrix data whose elements are values indicating signal intensities at multiple points in time obtained by the light receiving unit receiving the light emitted from the cells passing through the flow cell. 2. The cell analysis device of embodiment 1.

[0247] [Aspect 2] the light receiving unit includes a first light receiving unit that receives a first type of light from the cell and a second light receiving unit that receives a second type of light from the cell; The matrix data is (1) first data having values indicating signal intensities at multiple points in time obtained by the first light receiving unit receiving the first type of light emitted from cells passing through the flow cell as elements; and (2) second data including values indicating signal intensities at multiple points in time obtained by the second light receiving unit receiving the second type of light emitted from cells passing through the flow cell; 3. The cell analysis device of embodiment 2.

[0248] [Aspect 4] The deep learning algorithm determines a cell type of a cell corresponding to the input matrix data from among a plurality of cell types including monocytes, neutrophils, lymphocytes, eosinophils, basophils, and abnormal cells. 2. The cell analysis device of embodiment 1.

[0249] [Aspect 5] transmitting the matrix data to the parallel processor via a transmission path provided in the cell analysis device; 2. The cell analysis device of embodiment 1.

[0250] [Aspect 6] 2. The cell analysis device according to aspect 1, wherein the matrix data is transmitted to the parallel processor via a transmission path different from the Internet or an intranet.

[0251] [Aspect 7] 6. The cell analysis device according to aspect 5, wherein the transmission path has a communication bandwidth of 1 gigabit / second or more.

[0252] [Aspect 8] 6. The cell analysis device according to aspect 5, wherein the transmission path is a bus, and the matrix data is transmitted to the parallel processor via the bus.

[0253] [Aspect 9] the matrix data converted from analog signals relating to the plurality of cells measured by the flow cytometer is transmitted to the parallel processor via the transmission path; A cell analysis device according to any one of aspects 5 to 7.

[0254] [Aspect 10] The information includes an identifier for identifying the cell type. A cell analysis device according to any one of aspects 1 to 9.

[0255] [Aspect 11] The information includes a probability that the cell belongs to each of the plurality of cell types. A cell analysis device according to any one of aspects 1 to 10.

[0256] [Aspect 12] transmitting the analysis results, including the identifiers for identifying the cell types, to a processing unit that analyzes the analysis results; A cell analysis device according to any one of aspects 1 to 11.

[0257] [Aspect 13] transmitting the analysis results, including the probability that the cell belongs to each of the plurality of cell types, to a processing unit that analyzes the analysis results; 13. The cell analysis device of any one of aspects 1 to 12.

[0258] [Aspect 14] the parallel processing processor executes, in parallel, a plurality of arithmetic processes related to the analysis of the matrix data, as the parallel processing; A cell analysis device according to any one of aspects 1 to 13.

[0259] [Aspect 15] The parallel processing processor includes: a plurality of arithmetic units capable of executing arithmetic processing relating to the analysis of the matrix data; the parallel processing includes performing the arithmetic processing in parallel by each of the arithmetic units; A cell analysis device according to any one of aspects 1 to 14.

[0260] [Aspect 16] the parallel processor executes, in parallel, a filtering process for extracting features of the matrix data as the parallel processing; 16. A cell analysis device according to any one of aspects 1 to 15.

[0261] [Aspect 17] The parallel processing processor executes, in parallel, a plurality of arithmetic operations in a convolution layer in the deep learning algorithm as the parallel processing. 17. A cell analysis device according to any one of aspects 1 to 16.

[0262] [Aspect 18] The parallel processing processor executes the parallel processing in accordance with a single instruction. 18. A cell analysis device according to any one of aspects 1 to 17.

[0263] [Aspect 19] The matrix data is data based on signals detected by irradiating the cells with light. 19. A cell analysis device according to any one of aspects 1 to 18.

[0264] [Aspect 20] sampling data obtained by sampling signals obtained by measuring the cells at a predetermined rate is input to the parallel processor as the matrix data; the parallel processing processor analyzes the sampling data by executing the parallel processing; 20. A cell analysis device according to any one of aspects 1 to 19.

[0265] [Aspect 21] inputting image data acquired by irradiating the cells with light into the parallel processor as the matrix data; the parallel processing processor analyzes the image data by executing the parallel processing; 21. A cell analysis device according to any one of aspects 1 to 20.

[0266] [Aspect 22] The parallel processor performs at least 100 matrix operations for each of the cells. 22. A cell analysis device according to any one of aspects 1 to 21.

[0267] [Aspect 23] The parallel processor performs at least 1000 matrix operations for each of the cells. 23. A cell analysis device according to any one of aspects 1 to 22.

[0268] [Aspect 24] the parallel processor analyzes the matrix data corresponding to each of at least 100 of the cells. 24. A cell analysis device according to any one of aspects 1 to 23.

[0269] [Aspect 25] the parallel processor analyzes the matrix data corresponding to each of at least 1000 of the cells. 25. A cell analysis device according to any one of aspects 1 to 24.

[0270] [Aspect 26] the parallel processors each analyze the matrix data having a capacity of at least 1 gigabyte; 26. A cell analysis device according to any one of aspects 1 to 25.

[0271] [Aspect 27] The parallel processing processor includes: At least 10 calculation units capable of performing calculation processing related to the analysis of the matrix data; the parallel processing includes performing the arithmetic processing in parallel by each of the arithmetic units; 27. A cell analysis device according to any one of aspects 1 to 26.

[0272] [Aspect 28] The parallel processing processor includes: having at least 100 arithmetic units capable of executing arithmetic processing related to the analysis of the matrix data; the parallel processing includes performing the arithmetic processing in parallel by each of the arithmetic units; 28. A cell analysis device according to any one of embodiments 1 to 27.

[0273] [Aspect 29] The parallel processing processor includes: At least 1000 calculation units capable of performing calculation processing related to the analysis of the matrix data; the parallel processing includes performing the arithmetic processing in parallel by each of the arithmetic units; 29. A cellular analysis device according to any one of aspects 1 to 28.

[0274] [Aspect 30] the parallel processor receives the matrix data read from a memory having a capacity of at least 1 gigabyte as an input and executes the parallel processing; 30. A cellular analysis device according to any one of embodiments 1 to 29.

[0275] [Aspect 31] 31. The cell analyzer according to any one of aspects 1 to 30, wherein the matrix data regarding cells in a specimen is a combination of matrix data based on multiple types of signals obtained from the cells.

[0276] [Aspect 32] In a cell analysis device, Mix the specimen and the reagent to prepare a measurement sample; Passing the measurement sample through a flow cell, and acquiring matrix data whose elements are values indicating signal intensities at multiple time points for each of multiple cells contained in the measurement sample passing through the flow cell; executes parallel processing of matrix operations included in a deep learning algorithm that outputs information about the cell type for each of the plurality of cells in response to input of the matrix data, using a parallel processing processor; causing a processor to prepare the measurement sample, obtain the matrix data, and cause the parallel processing processor to perform the parallel processing; A cell analysis method comprising: [Explanation of symbols]

[0277] 50 Pre-trained Deep Learning Algorithms 60 Pre-trained Deep Learning Algorithms 400, 400a, 500, 500a, 700 measurement units 410 FCM detector 450 Sample suction unit 482, 507, 3008, 6009 A / D conversion section 3001, 4831, 6001, 8111 processor (host processor) 3002, 4833, 6002, 8112 parallel processing processors 4000, 4000', 4000'' cell analyzer 3200, 6200, 4836 computing units

Claims

1. a measurement unit including a flow cytometer and configured to measure a plurality of cells contained in the sample; a first processor configured to process information related to the analysis of the plurality of cells; a second processor configured to perform parallel processing; Equipped with the measurement unit causes the sample to flow through a flow cell of the flow cytometer, and acquires, for each of a plurality of cells, an analog signal corresponding to the intensity of light emitted from the cells irradiated with light, thereby acquiring matrix data whose elements are values digitally indicating the analog signal levels of each of the plurality of cells at a plurality of time points; the second processor is configured to perform parallel processing for processing matrix data according to an artificial intelligence algorithm including a plurality of matrix operations; the first processor is configured to classify the cell type of each of the plurality of cells based on a result of the parallel processing by the second processor; Cell analysis device.

2. The flow cytometer includes a light source that irradiates the flow cell with light, and a detection unit that detects light emitted from each of the cells in response to the light irradiated by the light source. The cell analysis device according to claim 1 .

3. The detection unit is configured to detect a plurality of types of light emitted from each of the cells. The cell analysis device according to claim 2 .

4. The artificial intelligence algorithm is a deep learning algorithm. The cell analysis device according to claim 1 .

5. The deep learning algorithm determines a cell type of a cell corresponding to the input matrix data from among a plurality of cell types including monocytes, neutrophils, lymphocytes, eosinophils, basophils, and abnormal cells. The cell analysis device according to claim 4 .

6. The matrix data is transmitted to the second processor via a transmission path provided in the cell analysis device. The cell analysis device according to claim 1 .

7. the matrix data is transmitted to the second processor via a transmission path other than the Internet or an intranet; The cell analysis device according to claim 6 .

8. The transmission path has a communication bandwidth of 1 gigabit / second or more. The cell analysis device according to claim 6 or 7.

9. the transmission path is a bus, and the matrix data is transmitted to the second processor via the bus; The cell analysis device according to any one of claims 6 to 8.

10. the measurement unit includes an A / D converter configured to convert the analog signal for each of the plurality of cells measured by the flow cytometer into a digital signal; the A / D converter samples each analog signal at a predetermined sampling rate to acquire the matrix data of the cell as waveform data corresponding to the cell; The cell analysis device according to any one of claims 1 to 9.

11. generating information including an identifier for identifying the cell type; The cell analysis device according to any one of claims 1 to 10.

12. generating information including a probability that the cell belongs to each of the plurality of cell types; The cell analysis device according to any one of claims 1 to 11.

13. transmitting the analysis results, including the identifiers for identifying the cell types, to a processing unit that analyzes the analysis results; The cell analysis device according to any one of claims 1 to 12.

14. transmitting the analysis results, including the probability that the cell belongs to each of the plurality of cell types, to a processing unit that analyzes the analysis results; The cell analysis device according to any one of claims 1 to 13.

15. the second processor executes, in parallel, a plurality of arithmetic processes related to the analysis of the matrix data as the parallel processing; The cell analysis device according to any one of claims 1 to 14.

16. The second processor a plurality of arithmetic units capable of executing arithmetic processing relating to the analysis of the matrix data; the parallel processing includes performing the arithmetic processing in parallel by each of the arithmetic units; The cell analysis device according to any one of claims 1 to 15.

17. the second processor executes, in parallel, a filtering process for extracting features of the matrix data as the parallel processing; The cell analysis device according to any one of claims 1 to 16.

18. the artificial intelligence algorithm is a deep learning algorithm, and the second processor executes, as the parallel processing, a plurality of arithmetic operations in a convolution layer of the deep learning algorithm in parallel; The cell analysis device according to any one of claims 1 to 17.

19. the second processor executes the parallel processing in accordance with a single instruction; The cell analysis device according to any one of claims 1 to 18.

20. The matrix data is data based on signals detected by irradiating the cells with light. The cell analysis device according to any one of claims 1 to 19.

21. sampling data obtained by sampling signals obtained by measuring the cells at a predetermined rate, and inputting the sampled data into the second processor as the matrix data; the second processor analyzes the sampling data by executing the parallel processing; The cell analysis device according to any one of claims 1 to 20.

22. inputting image data acquired by irradiating the cell with light into the second processor as the matrix data; the second processor analyzes the image data by performing the parallel processing; The cell analysis device according to any one of claims 1 to 21.

23. the second processor performs at least 100 matrix operations for each of the cells. The cell analysis device according to any one of claims 1 to 22.

24. the second processor performs at least 1000 matrix operations for each of the cells. The cell analysis device according to any one of claims 1 to 23.

25. the second processor analyzes the matrix data corresponding to each of at least 100 of the cells. The cell analysis device according to any one of claims 1 to 24.

26. the second processor analyzes the matrix data corresponding to each of at least 1000 of the cells. The cell analysis device according to any one of claims 1 to 25.

27. the second processors each analyze the matrix data having a capacity of at least 1 gigabyte; The cell analysis device according to any one of claims 1 to 26.

28. The second processor at least 10 arithmetic units capable of executing arithmetic processing related to the analysis of the matrix data; the parallel processing includes performing the arithmetic processing in parallel by each of the arithmetic units; The cell analysis device according to any one of claims 1 to 27.

29. The second processor At least 100 calculation units are provided that can perform calculations related to the analysis of the matrix data; the parallel processing includes performing the arithmetic processing in parallel by each of the arithmetic units; The cell analysis device according to any one of claims 1 to 28.

30. The second processor having at least 1000 arithmetic units capable of executing arithmetic processing related to the analysis of the matrix data; the parallel processing includes performing the arithmetic processing in parallel by each of the arithmetic units; The cell analysis device according to any one of claims 1 to 29.

31. the second processor receives the matrix data read from a memory having a capacity of at least 1 gigabyte as an input and executes the parallel processing; The cell analysis device according to any one of claims 1 to 30.

32. The matrix data regarding the cells in the specimen is a combination of matrix data based on multiple types of signals obtained from the cells. The cell analysis device according to any one of claims 1 to 31.

33. A method of cell analysis using a cell analysis device including a first processor, a second processor, and a flow cytometer, comprising: obtaining matrix data for each of a plurality of cells in the sample by flowing the sample through a flow cell of the flow cytometer and obtaining, for each of the plurality of cells, an analog signal corresponding to the intensity of light emitted from the illuminated cells; wherein the matrix data has, as elements, values that digitally indicate analog signal levels at a plurality of points in time; performing, by the second processor, parallel processing for processing the matrix data according to an artificial intelligence algorithm including a plurality of matrix operations; classifying, by the first processor, the cell type of each of the plurality of cells based on the results of the parallel processing by the second processor; A cell analysis method comprising:

Images