Data analysis methods, data analysis systems, and computers

Abstract

Provided is a technology that makes it possible to improve accuracy, etc. when determining the deviation of measurement data from a regression line, etc. and detect abnormalities. According to the present invention, a data analysis method includes a first step in which a computer system acquires reaction process data that includes device information, reagent information, and measurement data from a plurality of automatic analysis systems as reference data, a second step in which the computer system acquires label data that includes, for each piece of reaction process data, an abnormality determination and an abnormality source that have been inputted by an operator, a third step in which, on the basis of an analysis of the reference data and the included label data relative to a distribution map for evaluation parameters for the reaction process data, the computer system calculates overall regression line information 91 that can be commonly applied to the plurality of automatic analysis systems and overall deviation determination reference line information 92 for determining deviation from the overall regression line information 91, and a fourth step in which the computer system displays the overall deviation determination reference line information 92 on a screen 90.

Description

Technical Field

[0001] The present invention relates to data analysis technology, and particularly to data analysis technology related to an automatic analyzer.

Background Art

[0002] An automatic analyzer for clinical tests (also referred to as an automatic analysis system) is a device that dispenses a certain amount of a sample and a reagent, stirs and reacts them, and analyzes the liquid (also referred to as a reaction solution) obtained after the reaction. Using an automatic analyzer, a user (also referred to as an operator) such as a clinical laboratory technician measures the absorbance of the reaction solution over a certain period of time and determines the concentration of the component to be analyzed based on the measurement results. For example, in clinical tests, devices such as automatic analyzers, reagents for each analysis item, standard solutions for calibrating the reagents, the device during analysis, and accuracy control samples measured to check the state of the reagents are required. And the final analysis performance is obtained by combining these and other elements. These other elements include, for example, the dispensing volume accuracy of the device, the uniformity inside the reagent bottle, the stability during storage, the degree of chemical reaction, particularly the stirring efficiency of the reagent and the sample, the cleanliness of the reaction vessel, and the stability of the standard solution.

[0003] Thus, there are multiple factors that govern the analysis performance. Also, the elements in the device that affect the analysis performance (in other words, the elements that directly affect the analysis performance) include configurations such as a sample dispensing mechanism, a reagent dispensing mechanism, a stirring mechanism, an optical system, a reaction vessel, and a thermostat. Furthermore, factors that affect outside the device include the properties of the reagent, the sample, and the control sample.

[0004] Thus, the analysis performance is affected by various factors. Therefore, when using an automatic analyzer, it is necessary to check these factors (in other words, the influencing factors) to confirm whether normal clinical tests are possible.

[0005] Calibration in automated analyzers is performed for each reagent bottle using standard solutions. Specifically, blank and standard solutions are measured to determine the origin, the absorbance per unit concentration is calculated, and a conversion factor (also known as the K-factor) is determined. Generally, a technician checks the magnitude of the absorbance and the change in the K-factor over time to determine the quality of the calibration results.

[0006] Furthermore, after calibration, quality control is performed by measuring quality control samples of known concentration, and the difference from the reference value is used for verification. Typically, when measuring patient samples, quality control samples are measured periodically at regular intervals to check for deviations from the acceptable value. If the acceptable value is exceeded, it is assumed that there is a problem with either the reagent or the equipment, and an inspection is performed. In addition, for verification of data in routine tests, verification is performed using reaction process data. The method of verification varies depending on the analytical method.

[0007] Clinical laboratory measurement methods can be classified into two types based on the analytical method: rate methods and endpoint methods. Endpoint methods mainly concern methods for measuring components such as proteins and lipids contained in a sample. When components in a sample react with a reagent, if the binding reaction is fast, the reaction will be completed in a short time, and the concentration of the reaction product will reach a constant value. If the reaction time is long, it will take time for the reaction product to reach a constant concentration. The relationship between time and reaction product can be schematically shown in Figure 1 as a logistic curve (ABS=A0+A1(1-exp). kt )) is the result. k is the reaction rate constant. The endpoint method is a method for accurately determining the concentration of a substance to be measured in biochemical analysis. In the endpoint method (and similarly in the rate method), the relationship between the absorbance of the color reaction and time is determined using the least squares method: y = A + (BA) / e Kt The concentration of the substance being measured is determined by approximation. A conventional method for detecting data anomalies during measurement in the endpoint method is prozone checking.

[0008] Furthermore, with reagents used in immunoturbidimetry, such as IgA (immunoglobulin A) and CRP (C-reactive protein), the salt concentration in the reagent composition can cause proteins to precipitate. This precipitate can disrupt the reaction process, often appearing in the latter half of the reaction time. If this fluctuation occurs at the photometric point used for concentration calculation, accurate measurements cannot be obtained. Methods to check for this include antibody re-addition and reaction rate ratio methods, both of which issue an alarm when a parameter-specified limit is exceeded.

[0009] As an example of prior art, International Publication No. 2020 / 195783 (Patent Document 1) describes how to improve the accuracy of discriminating deviations from the regression line of measurement data, how the data analysis system generates composite regression line information that can be commonly applied to multiple automatic analysis units based on reference data, and how to display the composite regression line information on a display screen. [Prior art documents] [Patent Documents]

[0010] [Patent Document 1] International Publication No. 2020 / 195783 [Overview of the Initiative] [Problems that the invention aims to solve]

[0011] Improved performance of automated analyzers has made it possible to perform highly accurate analysis of various parameters even with minute amounts of samples and reagents. On the other hand, slight malfunctions in various parts of the instrument, or subtle changes in the quality of samples and reagents, can sometimes prevent accurate analysis. Automated analyzers for clinical testing measure the absorbance of the reaction solution, obtained by reacting a sample and reagent, at regular intervals as time-series absorbance, and then measure the rate of change in absorbance and the final absorbance from this time-series absorbance. From this data, the concentration of the target substance and the activity value of the enzyme are calculated.

[0012] During the monitoring of the reaction process, the automated analyzer performs sample dispensing (sampling), reagent dispensing, and mixing, and these processes include multiple error factors, such as errors due to differences in the degree of mixing and errors due to the amount of bubbles generated during sampling. In particular, conventionally, it has not been possible to quantitatively evaluate whether or not mixing is performed or the level of mixing, and there are no criteria for judgment. As a result, evaluations such as the quality of reproducibility and the presence or absence of measurements that clearly indicate some kind of problem, such as discontinuity in measured values, have been ambiguous. Furthermore, in cases where the reagent is diluted by the washing water of the reagent probe, or when the user accidentally mixes another solution with the reagent, the automated analyzer needs to detect abnormalities and notify the user of the abnormality, prompting retesting or maintenance of the equipment.

[0013] Clinical laboratory technologists, who use automated analyzers, find it difficult to visually check the entire reaction process during their daily testing work. In particular, when the measured values ​​are within the normal range, they are prone to overlooking reaction abnormalities, which can lead to inaccurate results.

[0014] While Patent Document 1, a prior art example, performs analysis using measurement data, there is room for improvement in terms of accuracy, such as when discriminating deviations from the regression line of the measurement data.

[0015] The object of the present invention is to provide a technology that can improve accuracy and enable anomaly detection when performing data analysis techniques, such as discriminating deviations from regression lines of measured data. [Means for solving the problem]

[0016] A typical embodiment of this disclosure has the configuration shown below. The data analysis method of the embodiment includes: a first step in which a computer system acquires, as reference data, reaction process data including instrument information, reagent information, and measurement data from each of the multiple automated analysis systems; a second step in which the computer system acquires, input by an operator, label data including whether there is an abnormality and the cause of the abnormality for each of the reaction process data; and a third step in which the computer system calculates, with respect to a distribution diagram of evaluation parameters of the reaction process data, comprehensive regression line information applicable to the multiple automated analysis systems in common, and comprehensive deviation discrimination reference line information for determining deviations from the comprehensive regression line information, based on the analysis of the reference data including the label data. [Effects of the Invention]

[0017] According to a representative embodiment of this disclosure, data analysis techniques can improve accuracy and enable anomaly detection when performing tasks such as discriminating deviations from regression lines of measurement data. Other issues, configurations, and effects are shown in the embodiments for carrying out the invention. [Brief explanation of the drawing]

[0018] [Figure 1] This is a diagram illustrating the outline of the logistic curve. [Figure 2] This figure shows an example configuration of the automated analysis system in Embodiment 1. [Figure 3] This figure shows an example configuration of the automated data analysis system in Embodiment 1. [Figure 4] This figure shows an example configuration of an automated data analysis system in a modified version of Embodiment 1. [Figure 5] This figure shows an example of the functional block configuration of the comprehensive analysis server in Embodiment 1. [Figure 6] This figure shows an example of the functional block configuration of a remote terminal in Embodiment 1. [Figure 7]It is a diagram showing a functional block configuration example of the automatic analysis system in Embodiment 1. [Figure 8] It is a diagram showing a configuration example of the reference data information table of the remote terminal in Embodiment 1. [Figure 9] It is a diagram showing a configuration example of the data configuration of the comprehensive analysis server in Embodiment 1. [Figure 10] It is a diagram showing a configuration example of the comprehensive divergence discrimination screen displayed on, for example, a remote terminal in Embodiment 1. [Figure 11] It is a diagram showing a detailed example of the GUI in the comprehensive divergence discrimination screen in Embodiment 1. [Figure 12] It is a diagram showing another configuration example of the comprehensive divergence discrimination screen in Embodiment 1. [Figure 13] It is a diagram showing an example of the processing sequence between devices in Embodiment 1. [Figure 14] It is a diagram showing an example of the processing flow of the comprehensive analysis server in Embodiment 1. [Figure 15] It is a diagram showing an example of the processing flow regarding the details of the comprehensive analysis process in step S1005 of FIG. 14.

Mode for Carrying Out the Invention

[0019] Hereinafter, embodiments of the present disclosure will be described in detail with reference to the drawings. In the drawings, the same components are generally denoted by the same reference numerals, and repeated explanations are omitted. In the drawings, the representation of components may not represent the actual position, size, shape, and range, etc. in order to facilitate understanding of the invention. In the drawings, only a part of the lines representing communication lines and the like is shown, but it is not limited thereto. For example, all components may be interconnected. The system and the like of the embodiment may be mainly implemented by software operating on a general-purpose computer, or may be implemented by dedicated hardware, or a combination of software and hardware.

[0020] In explanations, when describing program-based processing, the focus may sometimes be on the program, functions, or processing units. However, the core hardware component is the processor, or a controller, device, computer, or system composed of such a processor. The computer, using its processor, executes processing according to the program read into memory, utilizing resources such as memory and communication interfaces as appropriate. This realizes the specified functions and processing units. The processor is composed of semiconductor devices such as a CPU or GPU. Processing is not limited to software program processing; it can also be implemented using dedicated circuits. FPGAs, ASICs, CPLDs, etc., can be used as dedicated circuits.

[0021] The program may be pre-installed as data on the target computer, or it may be distributed as data to the target computer from the program source. The program source may be a program distribution server on a communication network, or it may be a non-transient computer-readable storage medium (e.g., a memory card). The program may consist of multiple modules. The computer system may consist of multiple devices. The computer system may consist of a client-server system, a cloud computing system, an IoT system, etc.

[0022] Various types of data and information can be structured in ways such as tables, lists, queues, and databases (DBs), but are not limited to these. Therefore, tables and similar structures are sometimes simply referred to as information or data. Furthermore, expressions such as identification information, identifiers, IDs, names, and numbers are interchangeable.

[0023] [Summary etc.] The system described in Patent Document 1, a prior art example, performs analysis (particularly the creation of a composite regression line) using measurement data from an automated analyzer, as well as basic information such as instrument information and reagent information. The system in Patent Document 1 does not utilize information such as whether or not a malfunction occurred in the automated analyzer, or whether or not there was an anomaly in the measurement data discovered by the operator, or the cause of that anomaly (label data recording this information). Furthermore, in recent years, many effective analysis methods using label data have been proposed, such as those represented by neural networks.

[0024] Therefore, the data analysis system of this embodiment utilizes not only measurement data and basic information, but also information such as whether or not a malfunction occurred in the automated analyzer, and whether or not an anomaly in the measurement data discovered by the operator and the cause of that anomaly (label data recording this information) for analysis. The system of this embodiment improves the accuracy of discriminating deviations from the regression line (especially the composite regression line) of the measurement data by performing analysis and learning using the label data, and not only that, it also estimates and detects anomalies and malfunctions. Furthermore, the data analysis system of this embodiment efficiently acquires label data from the operator. The data analysis system of this embodiment displays an interface on the display screen for inputting and assisting with label data, and displays the input label data, etc. (see Figure 10 below).

[0025] The data analysis method of the embodiment includes: a first step in which a computer system (e.g., a first computer and a second computer) acquires, as reference data, reaction process data including instrument information, reagent information, and measurement data from each of the multiple automated analysis systems; a second step in which the computer system acquires label data, including the presence or absence of abnormalities and the cause of abnormalities for each reaction process data, which is input by an operator; and a third step in which the computer system calculates, with respect to the distribution diagram of evaluation parameters of the reaction process data, comprehensive regression line information applicable to multiple automated analysis systems in common, and comprehensive deviation discrimination reference line information for determining deviations from the comprehensive regression line information, based on the analysis of the reference data including the label data. The data analysis method also includes a fourth step in which the computer system displays the comprehensive deviation discrimination line information on a screen.

[0026] The data analysis method of this embodiment includes the steps of: each of a plurality of first computers (remote terminals in Figure 3) acquires information (reference data in Figure 6) including measurement data and label data from a plurality of automated analysis systems (automated analyzers including automated analysis units); a second computer (integrated analysis server in Figure 3) connected to the plurality of first computers acquires reference data from the plurality of first computers; and the second computer performs an analysis (integrated analysis) of the plurality of automated analysis systems based on the reference data and generates analysis result information. The label data is input information such as the presence or absence of abnormalities and the cause of abnormalities discovered or judged by the operator regarding reaction process data including measurement data of the automated analysis system.

[0027] The second computer, through the first computer, acquires information about the automated analysis system (referred to as "Automated Analysis Unit Information" or "Instrument Information") and information about reagents (also referred to as "Reagent Information"), as well as reaction process data including corresponding measurement data, as reference data. The second computer also acquires label data entered by the operator on the display screen of the first computer as part of the reference data. Based on this reference data, the second computer generates and updates composite regression line information applicable to multiple automated analysis units, related to the automated analysis unit and reagents, as well as composite deviation discrimination reference line information for determining deviations from the composite regression line information.

[0028] Furthermore, the data analysis method of the embodiment includes the step of a second computer transmitting analysis result information (including overall regression line information and overall deviation discrimination criterion line information) to each of the multiple first computers, each first computer acquiring the analysis result information, and displaying a screen on its display screen that includes a distribution diagram of the evaluation parameters of the reaction process data and the overall deviation discrimination criterion line information. The first computer displays an interface (GUI / screen) on its screen for inputting and assisting label data for the reaction process data, and acquires label data input by an operator through that interface.

[0029] The data analysis system (first or second computer) compares the measured data points in the distribution map with the overall deviation discrimination criteria line information to determine whether there is a deviation from the overall regression line and the degree of deviation, or to estimate the possibility of anomalies, and displays an interface for inputting label data. The second computer updates the overall regression line and overall deviation discrimination criteria line information based on the analysis and learning of reference data including the acquired label data.

[0030] The data analysis system (second or first computer) searches for measurement data points in the distribution map that do not have label data entered and that meet predetermined conditions, as measurement data for which it is difficult or impossible to determine whether or not there is an anomaly. In particular, for first measurement data points that have label data entered and are considered to have an anomaly, or for first measurement data points that are outside the overall deviation discrimination reference line information, the data analysis system searches for second measurement data points that do not have label data entered as other measurement points in the neighboring region. The data analysis system displays an interface on the screen for the second measurement data point, prompting the operator to enter label data in relation to the content of the label data of the first measurement data. In other words, this interface is for confirming and inquiring about the presence or absence of anomalies and the causes of anomalies, and for presenting estimation results.

[0031] <Embodiment 1> As one embodiment of this disclosure, a data analysis system and method of Embodiment 1 will be described. The data analysis system of Embodiment 1 is a system having a plurality of automatic analysis systems (in other words, automatic analysis devices equipped with automatic analysis units), remote terminals connected thereto, and an integrated analysis server connected to the remote terminals, as shown in Figure 3. The automatic analysis systems or remote terminals correspond to the first computer, and the integrated analysis server corresponds to the second computer. The data analysis method of Embodiment 1 is a method having steps executed in the first computer and the second computer of the data analysis system of Embodiment 1. The computers of the embodiment are the first computer and the second computer, and they perform processing based on a program.

[0032] Embodiment 1 primarily describes, as an example, a function for monitoring reactions during clinical laboratory analysis in an automated analysis system that qualitatively or quantitatively analyzes blood, urine, and other biological samples (also referred to as specimens). This function improves the accuracy of reaction process approximation methods and enables anomaly detection.

[0033] [Automated Analysis System] Figure 2 shows a theoretical example of the overall configuration of the automated analysis system 3, which is an element of the data analysis system of Embodiment 1. This automated analysis system 3 corresponds to one automated analysis system 302, etc., in Figure 3, which will be described later. This automated analysis system 3 includes a reaction disk 109, a reagent disk 112A, a sample dispensing mechanism 105, a reagent dispensing mechanism 110, a stirring device 113, a washing device 119, a light source 114, a multi-wavelength photometer 115, a rack transport device 123, a sample rack 102, an interface 104, a computer 103, and the like.

[0034] The reaction disk 109 is a reaction vessel holding mechanism, on which multiple reaction vessels 106 are arranged concentrically. The reaction vessels 106 are containers in which the reaction liquid is stored. The reaction disk 109 is equipped with a rotation drive mechanism (not shown) and is mounted to be rotatable. The reaction disk 109 is maintained at a predetermined temperature by a heat-retaining tank 126 connected to a constant temperature bath 108.

[0035] The reagent disk 112A is a reagent container holding mechanism, on which multiple reagent bottles (in other words, reagent containers) 112 are placed concentrically. The multiple reagent bottles 112 are containers that hold various reagents. Around the reaction disk 109 and the reagent disk 112A are a sample dispensing mechanism 105, a reagent dispensing mechanism 110, a stirring device 113, a washing device 119, a light source 114, and a multi-wavelength photometer 115.

[0036] A rack transport device 123 is installed on the rotational circumference of the sample dispensing mechanism 105 and along the tangential direction of the reaction disk 109. A rack number reader 124 and a sample ID reader 125 are also positioned along the transport line. The operation of all these mechanisms is controlled by a computer 103 via an interface 104.

[0037] One or more sample containers 101 containing the samples are placed in a sample rack 102. The sample rack 102 is transported along a transport line by a rack transport device 123. Each sample rack 102 is assigned a serial number, and as it is being transported along the transport line, this serial number is first read by a rack number reader 124.

[0038] Subsequently, if each sample container held in the sample rack 102 has been assigned an ID number, the sample ID number is read by the sample ID reader 125. Then, the sample rack 102 moves until the first sample container 101 held on the rack is directly below the sample dispensing probe 105A of the sample dispensing mechanism 105. All information read by the rack number reader 124 and the sample ID reader 125 is sent to the computer 103 via the interface 104.

[0039] The sample dispensing mechanism 105, based on control by the computer 103, uses the sample dispensing probe 105A to dispense a predetermined amount of sample from the sample container 101 into the reaction vessel 106. After dispensing is completed for one sample container 101, the sample rack 102 moves so that the next sample container 101 is positioned directly below the sample dispensing probe 105A.

[0040] The reaction vessel 106, into which the sample has been dispensed, rotates on the reaction disk 109 due to the rotational movement of the reaction disk 109. During this time, reagents from the reagent bottle 112 are dispensed into the sample in the reaction vessel 106 by the reagent dispensing mechanism 110, the reaction solution is stirred by the stirrer 113, and the absorbance is measured by the light source 114 and the multi-wavelength photometer 115. Afterwards, the reaction vessel 106, which has finished its analysis, is washed by the washing device 119. The measured absorbance signal is input to the computer 103 via the interface 104 through the A / D converter 116 and converted into the concentration of the target component in the sample. The concentration-converted data is displayed on the CRT (display) 118 via the interface 104, or printed out by the printer 117 and stored in the storage device 122.

[0041] The sample dispensing mechanism 105, reaction disk 109, rack transport device 123, reagent dispensing mechanism 110, stirring device 113, and washing device 119 are driven by pulse motors or the like (not shown). Although not shown, it is also possible to connect multiple automated analyzers and operate them as a single automated analysis system.

[0042] [Automated Data Analysis System] Figure 3 shows an example configuration of an automated data analysis system 300, which is a data analysis system of Embodiment 1 and includes multiple automated analysis systems 3. The automated data analysis system 300 in Figure 3 includes multiple automated analysis systems 302 (four in this example) in the first laboratory (Laboratory A) and multiple automated analysis systems 307 (four in this example) in the second laboratory (Laboratory B). The automated data analysis system 300 also includes two remote terminals 2 connected to the multiple automated analysis systems 3: a first remote terminal, remote terminal 303, and a second remote terminal, remote terminal 310. The automated data analysis system 300 also includes an integrated analysis server 309 connected to the two remote terminals 2.

[0043] Remote terminal 303 is an information processing terminal device connected to multiple automated analysis systems 302 (referred to as A1 to A4 in this example) in laboratory A via communication lines (e.g., LAN) 301 and 304. Similarly, remote terminal 310 is an information processing terminal device connected to multiple automated analysis systems 307 (referred to as B1 to B4 in this example) in laboratory B via communication lines.

[0044] The integrated analysis server 1 is a server device connected to remote terminals 303 and 310 via communication lines 305 and 308. The communication means and network, such as communication line 308, can be, for example, a dedicated line or an internet connection.

[0045] The number of automated analysis systems 3, remote terminals 2, and integrated analysis servers 1 is not limited to those illustrated in Figure 3. Preferably, there should be at least one automated analysis system 3 in each laboratory, and two or more remote terminals 2.

[0046] Each automated analysis system 302 in Laboratory A transmits reaction process data, including measurement data from the automated analysis unit, to a remote terminal 303. The remote terminal 303 also obtains item codes, instrument lot numbers, and reagent lot numbers from the automated analysis systems 302. Based on the reaction process data received from each automated analysis system 302, the remote terminal 303 creates reference data (reference data 701 in Figure 8, described later) for each automated analysis system 302. The remote terminal 303 transmits the created reference data to the integrated analysis server 309 via a communication line 305, etc. Similarly, each automated analysis system 307 in Laboratory B transmits reaction process data, including measurement data from the automated analysis unit, to a remote terminal 310. The remote terminal 310 also obtains item codes, instrument lot numbers, and reagent lot numbers from the automated analysis systems 307. Based on the reaction process data received from each automated analysis system 307, the remote terminal 310 creates reference data for each automated analysis system 307. The remote terminal 310 sends the created reference data to the comprehensive analysis server 309.

[0047] The integrated analysis server 1 stores reference data, including item codes, instrument lot numbers, reagent lot numbers, and reaction process data, acquired from each remote terminal 2, as data in the data configuration of 801 units as shown in Figure 9 below. The integrated analysis server 1 may also acquire reference data from each automated analysis system 3. The integrated analysis server 1 also acquires reference data, including label data, from each remote terminal 2. Using the reference data, including label data, the integrated analysis server 1 calculates and generates integrated regression line information 91 and integrated deviation discrimination criterion line information 92 (Figure 10 below), which are information for determining overall deviation and abnormality. The integrated analysis server 1 transmits this calculated information to remote terminals 303 and 310 (at least one of them) via a communication line 308 or the like.

[0048] The remote terminal 303 displays, on the display screen, a distribution map of reference data (particularly evaluation parameters of reaction process data), overall regression line information 91, and overall deviation discrimination reference line information 92 for at least one of the multiple automated analysis systems 302 installed in laboratory A.

[0049] The integrated analysis server 1 and the remote terminal 2 are each composed of a computer system. The computer system includes, for example, a processor, memory or storage devices, input devices, output devices, communication devices, etc. Input and output devices may be externally connected devices or internally implemented devices. As storage devices, external storage devices may be included. Various programs, various parameters, and various data and information are stored in the memory or storage devices. The processor realizes various functions and processing units by executing processing according to various programs. Input devices are devices for inputting data and information such as instructions and settings from the operator, such as a mouse or keyboard. Output devices are devices that output calculation results, such as displaying or printing, such as a monitor display or printer. Communication devices are devices on which a communication interface is implemented.

[0050] Furthermore, each automated analysis system 3 is also partially composed of a computer system (such as computer 103 in Figure 2). Certain functions in the data analysis system may be realized through communication between the computer systems of the automated analysis system 3, the remote terminal 2, and the integrated analysis server 1. Each computer system may also be realized through communication between multiple computers, such as a client-server system.

[0051] The computer 103 of the automated analysis system 3 and / or the remote terminal 2 can be referred to as the first computer. The integrated analysis server 1 can be referred to as the second computer. The remote terminal 2 and the integrated analysis server 1 are not limited to these names and may be any computer system with the specified functions. For example, it is possible to configure the computer 103 of the automated analysis system 3 to execute the processing of the first computer and one remote terminal 2 to execute the processing of the second computer. Alternatively, the remote terminal 2 may be omitted, and the automated analysis data analysis system 300 may be configured with the integrated analysis server 1 and multiple automated analysis systems 3. In other words, the remote terminal 2 and the integrated analysis server 1 may be merged into one.

[0052] Furthermore, while each automated analysis system 3 and each remote terminal 2 shares a common basic configuration, their detailed configurations may differ. Additionally, a portable information processing device (such as a tablet or smartphone) carried by the operator may be connected via communication to the automated analysis system 3, remote terminal 2, or integrated analysis server 1. The processor of this portable information processing device may then perform processing as a first or second computer. The information may also be displayed on the display screen of the portable information processing device.

[0053] [Differences in data analysis systems] Figure 4 shows a modified version of the data analysis system shown in Figure 3. The data analysis system in Figure 4 includes a data analysis system 10, which is a computer system connected via communication to multiple automatic analysis systems 3, and a portable information processing terminal device 20 for operators, which is connected via communication to the data analysis system 10. The data analysis system 10, being a computer system, includes a processor 11 and memory 12, etc. The processor 11 performs individual analysis processing for each individual automatic analysis system 3 and comprehensive analysis processing for multiple automatic analysis systems 3. The memory 12 stores various data and information, including comprehensive analysis information including label data as a data configuration 801 (Figure 9), comprehensive regression line information 91 and comprehensive deviation discrimination criterion line information 92 generated as a result of the comprehensive analysis, and screen data 15 for displaying the comprehensive analysis results, etc.

[0054] The data analysis system 10 acquires equipment information, reaction process data, and other data from each of the multiple automated analysis systems 3. The data analysis system 10 also acquires label data entered by the operator from each automated analysis system 3 or the operator's portable information processing terminal 20. The operator's portable information processing terminal 20 is, for example, a tablet terminal with a touchscreen display. The portable information processing terminal 20 accesses the data analysis system 10 (particularly its server function) based on the operator's operation. The portable information processing terminal 20 acquires screen data 15 containing comprehensive analysis result information from the data analysis system 10, and based on this screen data 15, displays a screen for comprehensive deviation anomaly discrimination, as shown in Figure 10, on the touchscreen display. The operator inputs label data on this screen. The portable information processing terminal 20 transmits the information, including the input label data, to the data analysis system 10. The data analysis system 10 updates the comprehensive deviation discrimination reference line information 92 and other data using the label data acquired from the portable information processing terminal 20.

[0055] Another variation involves transmitting information such as the overall deviation discrimination reference line information 92 from the data analysis system 10 to the portable information processing terminal 20, which then uses this information to create screen data and displays it on the display screen. The portable information processing terminal 20 may also estimate and determine the possibility of anomalies by comparing the measurement data points on the screen with the overall deviation discrimination reference line information 92.

[0056] As another variation of the system, in a configuration with multiple remote terminals 2 as shown in Figure 3, a function equivalent to the integrated analysis server 1 (second computer) may be implemented in only one specific remote terminal 2, and that specific remote terminal 2 may communicate with the other remote terminals 2 as appropriate.

[0057] [Integrated Analysis Server] Figure 5 shows an example of the functional block configuration of the integrated analysis server 1 (309). The integrated analysis server 1 includes, for example, a processor 3091, memory 3092, storage device 3093, input device 3094, output device 3095, communication device 3096, etc. These elements are interconnected, for example, via a bus.

[0058] The processor 3091 executes processing according to the various installed programs. The memory 3092 stores data and information such as various programs and various parameters. The storage device 3093 stores calculation results and acquired information. The input device 3094 is one or more devices such as a keyboard, touch panel, various buttons, and microphone. The output device 3095 is one or more devices such as a display device, printer, and speaker. The communication device 3096 is a device connected to the communication line 308 and has a communication interface implemented for communicating with multiple remote terminals 2. In a configuration where the integrated analysis server 1 and multiple automatic analysis systems 3 can communicate directly, the communication device 3096 has a communication interface with the automatic analysis systems 3.

[0059] The processor 3091 reads various programs from memory 3092, loads these programs into internal memory (not shown), and executes program processing as appropriate. These programs include a comprehensive regression line information processing program and a comprehensive deviation discrimination reference line information processing program. Figure 5 illustrates the state in which the comprehensive regression line information processing unit 30911 and the comprehensive deviation discrimination reference line information processing unit 30912, which are realized through processing according to these programs, are implemented within the processor 3091. The comprehensive analysis server 1 has at least the comprehensive regression line information processing unit 30911 and the comprehensive deviation discrimination reference line information processing unit 30912. The details of these processes will be described later.

[0060] [Remote device] Figure 6 shows an example of the functional block configuration of remote terminal 2 (303, 310). Note that the configurations of the two remote terminals 303 and 310 are the same, so sometimes only remote terminal 303 is described. Remote terminal 2 in Figure 6 includes, for example, a processor 3031, memory 3032, storage device 3033, input device 3034, output device 3035, etc.

[0061] The processor 3031 executes processing according to the various installed programs. The memory 3032 stores data and information such as various programs and various parameters. The storage device 3033 stores calculation results and acquired information. The input device 3034 is one or more devices such as a keyboard, touch panel, various buttons, and microphone. The output device 3035 is one or more devices such as a display device, printer, and speaker. The communication device 3036 is connected to the communication lines 304 and 305 and has a communication interface for communicating with the integrated analysis server 1 and multiple automated analysis systems 3 within the same laboratory (e.g., laboratory A).

[0062] The processor 3031 reads various programs from memory 3032, loads those programs into internal memory (not shown), and executes program processing as appropriate. The various programs include a screen processing program for overall deviation detection and a label data processing program. Figure 6 shows the state in which the screen processing unit 30311 for overall deviation detection and the label data processing unit 30312, which are realized by processing according to these programs, are realized within the processor 3031. The remote terminal 2 has at least the screen processing unit 30311 for overall deviation detection and the label data processing unit 30312.

[0063] The screen processing unit 30311 of the remote terminal 2 processes the display of a screen for overall deviation discrimination, as shown in Figure 10 below, on the display screen of the display device. At that time, the screen processing unit 30311 displays the overall deviation discrimination screen by overlaying the overall regression line information 91 and the overall deviation discrimination reference line information 92 onto a distribution map of reference data (particularly evaluation parameters of reaction process data) for multiple automated analysis systems 3 installed in the laboratory.

[0064] Furthermore, the label data processing unit 30312 displays an interface (GUI / screen) for inputting label data for the reaction process data points in the distribution map and the overall deviation discrimination reference line information 92 on its overall deviation discrimination screen, and acquires the label data entered by the operator through that interface. At that time, the label data processing unit 30312 may also estimate and determine the degree of deviation or the possibility of abnormality for a point by comparing the reaction process data point (i.e., the measurement result from an automated analysis system 3) with the overall deviation discrimination reference line information 92. Based on the estimation and determination result, the label data processing unit 30312 may then decide whether to display the interface for inputting label data for that point, and determine the type and content of that interface.

[0065] Furthermore, the estimation and determination of the degree of deviation and the possibility of abnormality may be performed in advance on the comprehensive analysis server 1 as part of the comprehensive analysis. In other words, a function equivalent to the label data processing unit 30312 may be implemented on the comprehensive analysis server 1. In that case, the comprehensive analysis result information transmitted from the comprehensive analysis server 1 to the remote terminal 2 may include information for controlling the interface for label data input for each measurement data point.

[0066] [Computer system for automated analysis systems] Figure 7 shows an example of the functional block configuration of the automated analysis system 3 as a computer system. The automated analysis system 3 comprises a computer 103 and an analysis unit 3027. The analysis unit 3027 is, in other words, an automated analysis unit. The computer 103 includes, for example, a processor 3021, memory 3022, storage device 3023, input device 3024, output device 3025, communication device 3026, etc.

[0067] The processor 3021 executes processing according to the various installed programs. The memory 3022 stores data and information such as various programs and various parameters. The storage device 3023 (corresponding to the storage device 122 in Figure 2) stores calculation results and acquired information. The input device 3024 is one or more devices such as a keyboard, touch panel, various buttons, and microphone. The output device 3025 is one or more devices such as a display device, printer, and speaker (corresponding to the printer 117 and CRT 118 in Figure 2). The communication device 3026 is a device connected to the communication line 301 and has a communication interface for communicating with the corresponding remote terminal 2. In the modified version, each automated analysis system 3 may communicate with other automated analysis systems in the laboratory through the communication device 3026, or it may communicate directly with the integrated analysis server 1.

[0068] The analysis unit 3027 is equipped with a sample dispensing mechanism 105, a reagent dispensing mechanism 110, a stirring device 113, a washing device 119, a light source 114, a multi-wavelength photometer 115, and the like, as explained in Figure 2.

[0069] The processor 3021 reads various programs from memory 3022, loads these programs into internal memory (not shown), and executes them as appropriate. These programs include analysis calculation processing programs and analysis unit operation control programs. Figure 7 illustrates the state in which the analysis calculation processing unit 30211 and the analysis unit operation control unit 30212 are realized within the processor 3021 through processing according to these programs. The analysis calculation processing unit 30211 performs, for example, the processing of data acquired by the analysis unit 3027. The analysis unit operation control unit 30212 controls the analysis unit 3027 to operate according to instructions input from the input device 3024, for example.

[0070] Alternatively, the automated analysis system 3 may be configured by separating the computer 103 and the analysis unit 3027 and installing them in different locations. Furthermore, the automated analysis system 3 may be configured by connecting multiple analysis units 3027 to a single computer 103.

[0071] [Reference data] Figure 8 shows an example of the configuration of the reference data information table (also simply referred to as reference data 701), which is an information table that stores the reference data held by each remote terminal 2. Each remote terminal 2 stores the information obtained from the automated analysis system 3 in this reference data information table. The reference data 701 in Figure 8 has a row number (No) and a reference data column, and a data type column is also shown for clarity in the explanation.

[0072] The reference data 701 in Figure 8 includes the following in numerical order: 1. Approximate code, 2. Analysis method, 3. Specimen type, 4. Sample identification mode, 5. Specimen type, 6.Item name, 7. Item code, 8. Analytical dispensing volume, 9. Measured values, 10. Data alarm, 11. Analysis Unit, 12. Lineage, 13. Equipment Lot, 14. Reagent Lot, 15. Dispensing date and time, 16.Residue sum of squares: Err, 17. Reaction rate constant: k, 18. Change in absorbance: A1, 19. Change in reaction absorbance: A0, 20. Slope of the asymptotic line: p, 21. Intercept of the asymptotic line: q, 22. Lag phase size: D0, 23. Lag phase length: T1, 24. Absorbance (main / minor) 1~28, 25. Approximate values ​​1-28, 26. Deviation presence / absence: 0 (absent), 1 (present), 27. Abnormality: 0 (absent), 1 (present), 28. Abnormal cause code, 29. Comments, 30. Deviation Judgment Model Parameters, 31. Anomaly detection model parameters.

[0073] The reference data 701 in Figure 8 is created by storing reaction process data, including measurement data, transmitted from the automated analysis system 3. The reference data 701 includes item code 703, instrument Lot 704, reagent Lot 705, evaluation parameter 702, label data 706, and model parameter 707 as configuration information.

[0074] The reaction process data, including measurement data, specifically refers to the data from Reference Data 701, including 9 measured values ​​and 24 absorbance values. Item code 703 is a code that identifies the analytical item (or test item) assigned to the reagent. Regarding item code 703, if there are multiple chemical manufacturers for the same type of reagent, a different code may be assigned because the codes may differ between companies. Instrument Lot 704 is instrument information (in other words, automated analysis unit information) for the automated analysis unit of automated analysis system 3, indicating the model name, production unit, etc. Reagent Lot 705 is reagent information indicating the production unit for the reagent.

[0075] The approximation code (1) is a code that identifies the approximate formula of the reaction process. Depending on the combination of the analysis item (item code) and the reagent (reagent information), the combination of the approximate formula and evaluation parameters to be used is selected. The correspondence between this information may be stored in a table in advance. The data alarm (10) is alarm information that is added and output as data when the automated analysis system detects various errors, etc.

[0076] The evaluation parameter 702 is a parameter used to evaluate the reaction process of the measured data, and is a parameter that constitutes an approximate formula for the reaction process. In this example, the evaluation parameter 702 has multiple parameters ranging from 16 residue sum of squares Err to 23 lag phase length T1. An example of the evaluation parameter 702 is described below. 18 absorbance change A1 (in other words, final reaction absorbance) is the absorbance change of the reaction process data in the endpoint method. For example, as an approximate formula, x = a0 - a1 * exp(-k t When using the formula, the absorbance change A1 is the same as parameter a1 in the approximation formula. The residue sum Err is the mean squared error of the difference between the approximate absorbance value calculated by the approximation formula and the actually measured absorbance value for each time point. For example, a distribution plot can be created using these two evaluation parameters (A1, ERR) as two axes, as shown in Figure 10 below.

[0077] Furthermore, for example, in the rate method, there are evaluation parameters that indicate the shape of the curve representing the reaction process. By plotting time from the start of the reaction on the horizontal axis and absorbance on the vertical axis, an approximate curve of the absorbance change can be obtained using an approximation formula. An asymptotic line can be obtained for this approximate curve. The time at which the approximate curve asymptotically approaches the asymptotic line is calculated, and the time from 0 to that time (in other words, the lag time) corresponds to the length of the lag phase T1. Also, the slope and intercept of the asymptotic line correspond to parameters 20 and 21. The evaluation parameters are not limited to the examples above; they can be defined arbitrarily.

[0078] Label data 706 contains 27 anomaly status items, 28 anomaly cause codes, and 29 comments. Label data 706 is primarily data entered by the operator. The data analysis system pre-defines the selection information for label data 706 and provides it on the screen as shown in Figure 10 below, so that the operator can select and input it. Label data 706 is used as training information in supervised machine learning (Figure 15) below.

[0079] Note that the label data 706 in Figure 8 is an example of a data storage instance where information entered by the operator is mainly stored. However, it is not limited to this, and information regarding the presence or absence of a malfunction in the automated analysis system 3 may also be included as part of the label data 706. For example, if the automated analysis system 3 or another device such as the remote terminal 2 detects a malfunction in the automated analysis system 3, it may output the malfunction information to the remote terminal 2, for example, and the remote terminal 2 may store this malfunction information as part of the label data within the reference data. Alternatively, the operator may input the malfunction information of the automated analysis system 3 as part of the label data on the screen of the remote terminal 2 or similar device.

[0080] The presence or absence of deviation in 26 represents the result of a computer system (first computer or second computer) determining the presence or absence of deviation from the overall regression line information 91 using the overall deviation discrimination criterion line information 92. Furthermore, the presence or absence of deviation in 26 may be made available to the operator as part of the label data 706.

[0081] Model parameters 707 are parameter information for the computational model used in the learning process described later (Figure 15) using reference data 701, which includes label data 706. One example is the parameter information for a neural network model (CNN). Note that model parameters 707 are not limited to learning; parameter information for algorithms used in processing such as deviation detection may also be applied. Initially, the integrated analysis server 1 creates initial values ​​for model parameters 707. The integrated analysis server 1 updates the model parameters 707 as needed in accordance with the learning process.

[0082] [Example of data configuration for a comprehensive analysis server] Figure 9 shows an example of the configuration of data configuration 801, which includes a reference data information table stored on the integrated analysis server 1. The data in data configuration 801 is all the data that the integrated analysis server 1 acquires and holds from the remote terminals 2 (303, 310), in other words, it is the integrated analysis data. The integrated analysis server 1 acquires reference data 701 as shown in Figure 8 from each remote terminal 2 and creates and holds the data in data configuration 801 that includes this data. Note that various types of data may be managed using a file system or database.

[0083] In the example of data structure 801 in Figure 9, multiple reference data sets are shown, specifically reference data 1 to N, which are, for example, data for each sample. Each of these reference data sets contains various data and information, including new data elements such as label data 706 and model parameters 707, as shown in Figure 8 above.

[0084] The data structure 801 stores and manages, for example, each reference data 804 corresponding to each reagent Lot 803 on a per-device Lot 802 basis. In Figure 9, the data structure 801 is structured in a hierarchy of device Lot 802 and reagent Lot 803, on a per-item code 703 basis of the reference data 701 in Figure 8. The data structure is not limited to this example, as long as links are established between the information of device Lot 802, etc.

[0085] [Information on the overall regression line and the overall deviation discrimination criteria line] The comprehensive regression line information 91 generated by the comprehensive analysis server 1 through comprehensive analysis is information such as regression lines generated by comprehensively analyzing reference data (particularly reaction process data and evaluation parameters) from multiple automated analysis systems 3 that are commonly applied. The comprehensive deviation discrimination reference line information 92 generated by the comprehensive analysis server 1 is reference line information that serves as a threshold for determining the degree of deviation from the comprehensive regression line information 91. The further a point representing a combination of evaluation parameters in the reaction process data is from the comprehensive regression line information 91, the greater the degree of deviation. In particular, if that point is outside the comprehensive deviation discrimination reference line information 92, it is estimated that the reaction process data is highly likely to be abnormal.

[0086] [Example screen for determining overall deviation] Figure 10 shows an example of the configuration of a screen (also referred to as the screen for determining overall deviation anomalies) that is displayed on the display device of the remote terminal 2 (or the automated analysis system 3) based on the analysis result information from the comprehensive analysis server 1. This screen 90 includes a distribution diagram of the evaluation parameters of the reaction process data, comprehensive regression line information 91, and comprehensive deviation discrimination reference line information 92. In this example, it is assumed that this screen 90 is displayed on the display screen of the remote terminal 303.

[0087] Screen 90 in Figure 10 includes a distribution map of combinations of evaluation parameters 702 for reaction process data, as a reference data distribution map. In this example, the horizontal axis of this distribution map is the final reaction absorbance A1 (absorbance change A1 in Figure 8), and the vertical axis is the mean squared error Err (residue sum Err in Figure 8). Each point, such as measurement data 99, is a plot of each reference data (in other words, measurement data, reaction process data). Each point can be distinguished and displayed by changing, for example, color or shape, for each target automated analysis system 3 or sample. In this example, the combination of evaluation parameters 702 is shown as the combination of A1 and Err, but it is not limited to this, and multidimensional distribution maps using two or more combinations of evaluation parameters 702 are similarly possible.

[0088] Furthermore, although the distribution map contains multiple points, these point clusters represent examples using historical data sets for one or more samples within a specified period from one or more automated analysis systems 3. When plotting data for multiple samples, each point can be distinguished and displayed using color or shape. For example, the data for the first sample could be represented by a green point, and the data for the second sample by a blue point. In another example, the distribution map could represent, for example, the data set from laboratory A in green, the data set from laboratory B in green, and so on.

[0089] In addition to the example in Figure 10, the display screen of the remote terminal 2 will show at least the overall deviation discrimination criterion line information 92. In the example in Figure 10, the overall regression line information 91 is also displayed, but the display of the overall regression line information 91 can be omitted.

[0090] In the overall deviation anomaly discrimination screen 90 of Figure 10, the overall regression line information 91 and overall deviation discrimination reference line information 92 are displayed overlaid on the distribution map of reference data (particularly combinations of evaluation parameters for reaction process data). The overall regression line information 91, shown as a solid line, represents the overall regression line (in other words, the regression function) for the measurement data set obtained from multiple automatic analysis systems 3 of the target, shown as a point cloud. The overall deviation discrimination reference line information 92, shown as a dashed line, is a reference line for discriminating deviations from the overall regression line information 91. In other words, it represents the boundary lines on both sides that constitute the reference range, and the lines that constitute the threshold. In this example, these are two curves set on both sides of the overall regression line information 91. In other words, the overall deviation discrimination reference line information 92 is overall anomaly detection information for comprehensively detecting anomalies by considering multiple automatic analysis systems 3.

[0091] The data analysis system (which may be either the integrated analysis server 1 or the remote terminal 2; for example, the integrated analysis server 1) can estimate that there is no deviation for points within the range between the two curves of the integrated deviation discrimination reference line information 92, and that there is a deviation for points outside the range. For example, measurement data such as point a, indicated by a black circle, which is outside one of the reference lines (the upper left side in the diagram), is estimated to be data with a deviation, in other words, abnormal data.

[0092] Furthermore, if the operator specifies a point for measurement data, the screen (e.g., a pop-up screen) may display detailed information about the reaction process data and reference data, including the measurement data corresponding to the specified point, such as a curve representing the reaction process. It is also possible to display regression lines and deviation discrimination criteria lines for each individual automated analysis system 3, and this is described in Patent Document 1.

[0093] Furthermore, in the screen 90 of Figure 10, a GUI / screen (also referred to as the label data input screen 93) is displayed that allows label data to be entered for the measurement data (abnormal data) at point a where there is a discrepancy. This label data input screen 93 is a GUI / screen that allows the operator to input label data (label data 706 in Figure 8) including whether or not there is an abnormality, the cause code of the abnormality, and comments. This label data input screen 93 may be a GUI component that is superimposed on the distribution map of reference data or the composite regression line information 91 as shown in the figure, or it may be a multi-window or a separate screen that transitions to separately, or it may be a separate GUI component that is displayed in parallel rather than superimposed. In this example, the label data input screen 93 is displayed as a callout-shaped GUI component for the measurement data at point a. GUI components such as list boxes and text boxes are included in the label data input screen 93.

[0094] Furthermore, the data analysis system (which may be either the integrated analysis server 1 or the remote terminal 2; for example, the integrated analysis server 1) calculates a neighboring region 94 (illustrated by a dashed circle) for the measurement data of point a, which is abnormal data with a deviation. The data analysis system detects measurement data that is within the neighboring region 94 of point a, inside the integrated deviation discrimination reference line 92, and for which no label data has been entered. In this example, the measurement data of point b is measurement data that meets these conditions.

[0095] The neighborhood region 94 can be defined by methods such as clustering, L1 norm, L2 norm, or Mahalanobis distance. Furthermore, the neighborhood region 94 is a subspace within a multidimensional space, combining the evaluation parameters 702 shown in Figure 8. That is, the neighborhood region 94 in this example is a projection of a multidimensionally defined subspace onto a two-dimensional plane. In this example, the neighborhood region 94 is a circle centered at point a and extending to a predetermined distance, and is a two-dimensional region.

[0096] The data analysis system (which may be either the integrated analysis server 1 or the remote terminal 2; for example, the integrated analysis server 1) displays a GUI / screen (also referred to as the label data input prompt screen 95) prompting the operator to input additional label data for the measurement data of point b, based on the neighboring region 94. The label data input prompt screen 95 may be a GUI component superimposed on the relevant measurement data as shown in the figure, or it may be a separate GUI component displayed in parallel rather than superimposed. In this example, the label data input prompt screen 95 is displayed as a callout-shaped GUI component for the measurement data of point b. The label data input prompt screen 95 contains a text message prompting the input of label data for point b.

[0097] Similarly, within screen 90, point c is an example of measurement data with a deviation (abnormal data) that lies outside the overall deviation discrimination reference line information 92. The label data input screen 96 is displayed for point c. Point d is located in the vicinity of point c, outside the overall deviation discrimination reference line 92, and is measurement data for which no label data has been entered. For measurement data like point d, the data analysis system displays the label data input prompt screen 97.

[0098] Furthermore, in the screen 90 of Figure 10, each measurement data point may be displayed with identification codes such as a, b, etc., as shown in the figure. In addition, for neighboring regions such as the neighboring region 94, information representing the neighboring region may be displayed, such as the dashed circle shown in the figure. When the cursor approaches a point or a point is selected, the neighboring region and other points contained within that neighboring region may be highlighted.

[0099] Details of screen 90 in Figure 10 are as follows. In the modified example, screen 90 may be similarly displayed on the display screen of the display device of each automated analysis system 3. Alternatively, as shown in Figure 4 above, it may be similarly displayed on the display screen of the data analysis system 10 or the operator's portable information processing terminal device 20.

[0100] Each point in the measurement data may be displayed with a different color or shape based on the deviation judgment made by the data analysis system in relation to the overall deviation discrimination reference line information 92. For example, since point a is outside the overall deviation discrimination reference line information 92, the data analysis system may estimate that there is a deviation and that it is abnormal, and display it with a predetermined color and shape (e.g., a red circular dot). Also, for points that are inside the overall deviation discrimination reference line information 92 and within a predetermined distance from the overall deviation discrimination reference line information 92, the data analysis system may estimate that there is a possibility of abnormality and display them with a predetermined color and shape. Furthermore, point b is within the vicinity region 94 of point a and inside the overall deviation discrimination reference line information 92, and for such points, the data analysis system may estimate that there is a possibility of abnormality in relation to point a, which is abnormal data, and display it with a predetermined color and shape (e.g., an orange circular dot).

[0101] Figure 11 shows details of screen 90 in Figure 10. Figure 11(A) shows the label data input screen 93 for point a, and includes an abnormality field 931 for inputting whether there is an abnormality, an abnormality cause field 932 for inputting the cause of the abnormality, and a comment field 933 for inputting comments. The abnormality field 931 is composed of a list box, and the operator can select and input from the options "abnormal" and "no abnormality". In this example, the abnormality field 931 has two options, abnormality or not, but in modified versions, it may have three or more options. For example, a value for abnormality not determined (when it is not possible or difficult to determine whether there is an abnormality) may be added, or the degree or probability of the abnormality may be classified into three or more values. The abnormality cause field 932 is composed of a list box, and the operator can select and input from a number of predefined abnormality cause codes. For example, "Abnormality cause code 4: Insufficient sample dispensing due to pipette clogging" is selected. Comment field 933 consists of a text box where operators can freely enter comments as text. For example, in comment field 933, "ZZ malfunction" is entered. Comments can include supplementary information about the cause of the malfunction.

[0102] Furthermore, initially, the label data input screen 93 may display information determined by the data analysis system as default information (in other words, initial values). For example, if point a is estimated or determined to be abnormal, the abnormality status column 931 on the label data input screen 93 will display "1: Abnormal" as the initial value. The operator can check this initial value, and if it is appropriate, leave it as is; if they determine that there is no abnormality, they can select and input "0: No abnormality".

[0103] Furthermore, in the abnormal cause field 932, if the data analysis system can automatically determine the cause of the abnormality, the automatically determined abnormal cause information may be automatically displayed as the initial value. For example, if the automatic analysis system 3 outputs an abnormal cause code or data alarm 10 in Figure 8, the remote terminal 2 or the integrated analysis server 1 may estimate and determine the cause of the abnormality based on the output information. The operator can leave the initial value as is if it is appropriate, or select and input a different abnormal cause code if they determine that there is another cause of the abnormality.

[0104] Furthermore, if an operator believes there is a cause for an anomaly other than the pre-defined options, they can create a new anomaly cause code. The data analysis system has a function for this purpose. Figure 11(B) shows an example of entering a new anomaly cause code in the anomaly cause field 932. The operator selects the "New Anomaly Cause Code A" item from the options displayed in the list box of the anomaly cause field 932, excluding the existing anomaly cause codes. This item may be displayed as "Create New Anomaly Cause," etc. Then, the operator creates and enters the content of "New Anomaly Cause Code A" as text in the anomaly cause field 932 and registers it. Separate registration and cancellation buttons may be displayed. After this new registration, the newly registered "New Anomaly Cause Code A" will be displayed as one of the options in the list box of the anomaly cause field 932, so the operator can reuse the new anomaly cause code.

[0105] Furthermore, various information entered by the operator through GUI components on screen 90 in Figure 10 is automatically saved and updated as data during background processing. Alternatively, a confirmation button or the like may be provided on screen 90, and the corresponding data may be saved and updated when that button or the like is pressed. The remote terminal 2 may communicate with the comprehensive analysis server 1 as appropriate and send the input data on screen 90, or the updated reference data 701 (including label data 706), to the comprehensive analysis server 1.

[0106] Figure 11(C) shows details of the label data input prompt screen 95 for point b. Within this label data input prompt screen 95, an example message is displayed: "Measurement point a had insufficient sample. Is it possible that nearby measurement point b also had insufficient sample?" The data analysis system automatically displays the label data input prompt screen 95 for point b, which is inside the overall deviation discrimination reference line information 92, taking into account the content of the label data entered on the label data input screen 93 for point a. For the measurement data of point a, the label data is entered as "abnormal cause code 4," indicating insufficient sample dispensing. Based on this, the data analysis system can infer that point b, which is in the vicinity of point a 94, may also have the same abnormal cause as point a.

[0107] Therefore, the data analysis system prompts the user to input label data for point b on the label data input prompt screen 95, conveying the possibility that the nearby point b may also have the same abnormal cause as point a, and prompts the user to input label data. As in the example above, for measurement data such as point b which is inside the overall deviation discrimination reference line information 92, it is more preferable that the content of the label data input prompt screen 95 is created by considering the content of the label data for point a (a point in the neighboring region 94) which is outside the overall deviation discrimination reference line information 92.

[0108] The operator can view and confirm the contents of the label data input prompt screen 95, determine the cause of the anomaly at point b, and input a corresponding affirmative or negative response. In this example, when the operator places the cursor within the label data input prompt screen 95, the YES / NO buttons shown in the figure are displayed. If the operator determines that the message content is correct, that is, that the cause of the anomaly estimated by the data analysis system is appropriate, they press the YES button. If it is incorrect, that is, that they determine it is a different cause of the anomaly, they press the NO button. When either button is pressed, the label data input screen for point b is displayed.

[0109] First, if the YES button is pressed, a label data input screen 95B like (D) is displayed. On this label data input screen 95B, the same content as the label data input screen 93 for point a (especially the abnormality status field 931 and the abnormality cause field 932) is automatically displayed as the initial value. The same content may also be displayed in the comment field 933. The operator checks the content of the label data input screen 95B for point b and confirms that the initial value is appropriate. If the operator determines that the abnormality status or abnormality cause is different from that of point a, they should enter the different abnormality status and abnormality cause in the abnormality status field and abnormality cause field.

[0110] Furthermore, if the NO button is pressed, a label data input screen will automatically be displayed, which will contain predetermined initial values ​​(for example, it may be empty), and will be different from the contents of the label data input screen 93 for point a. The operator will then enter the information they have determined into each field on that label data input screen.

[0111] Furthermore, the GUI can be anything other than the YES / NO buttons mentioned above. For example, a button such as "Undetermined" could be provided to handle cases where a YES / NO decision cannot be made. In the label data input screen displayed in response to such an "Undetermined" button, it may be possible to input a label data indicating that the cause of the anomaly is undetermined or unknown.

[0112] In this way, for measurement data without label data, such as point b, a GUI is provided in relation to the neighboring point a, and auxiliary information is provided to determine whether there is an anomaly and the cause of the anomaly, thereby reducing the effort required for operators to input label data. The data analysis system can efficiently acquire label data for multiple measurement data points overall.

[0113] Points c and d in Figure 10 are generally similar to the examples of points a and b above, but from a different perspective, an example of a false positive is shown. First, in the label data input screen 96 for point c, the fields for abnormality presence, abnormality cause, and comments are displayed, similar to the label data input screen 93 for point a. The operator enters information in each field on the label data input screen 96. For example, in the abnormality presence field, "0: No abnormality" is entered. Point c is outside the overall deviation discrimination reference line information 92, but the operator judges that there is no abnormality (in other words, a system erroneous detection, a false positive) and selects and enters "0: No abnormality". Also, in the abnormality cause field, "Abnormality cause code 1: False detection due to incorrect threshold setting" is selected and entered. In the comments field, for example, "Because the value of coefficient zz is xx, a false detection occurs in the case of yy" is entered. The threshold here also corresponds to the overall deviation discrimination reference line information 92. If the current overall deviation discrimination reference line information 92 is updated to expand outwards, point c will be judged as having no abnormalities.

[0114] Point d is located within the vicinity of point c, outside the overall deviation discrimination reference line information 92, and has no label data entered. The data analysis system can simply estimate that point d is abnormal, as it is outside the overall deviation discrimination reference line information 92, similar to points a and c. However, it further considers its relationship with the neighboring point c. The data analysis system, considering the content of the label data of the neighboring point c (indicating false detection or false positive), can estimate the possibility of a false detection or false positive for point d. Therefore, the data analysis system automatically displays a label data input prompt screen 97 for point d. The label data input prompt screen 97 includes a message such as, "Measurement point c was a false positive. Is the neighboring measurement point d also a false positive?" The operator can review the content of the label data input prompt screen 97 and, as in the example of the label data input prompt screen 95, input affirmative / negative responses as needed. For example, if the YES button is pressed on the label data input prompt screen 97, a label data input screen with the same content as point c will be displayed. The operator can then review the label data input screen and modify the content as needed.

[0115] As shown in the example above, in screen 90 of Figure 10, as a basic function, the operator can input label data for each measurement data point on the GUI / screen. Furthermore, in Embodiment 1, as a more intelligent function, the data analysis system automatically performs estimations and judgments and presents information such as initial values ​​for label data input, prompts for label data input, and estimated results of abnormalities and their causes. As a result, the operator can easily perform label data input work for a large number of measurement data through the screen.

[0116] Another modification of screen 90 in Figure 10 is that for each measurement data point, the label data input screen and the label data input prompt screen may be displayed as a single unit. Furthermore, for multiple points on the distribution map, a display may be provided to distinguish between points where label data has not been entered and those where it has. When a point with entered label data is selected, the displayed label data input screen allows the operator to review the entered label data and make corrections as needed.

[0117] Figure 12 shows an example of the display of screen 90 in a modified example. In this example, screen 90 is a window corresponding to a web page. Within this screen 90, a label data input field 90B is displayed in parallel with the distribution map 90A (details omitted) similar to that in Figure 10. The label data input field 90B contains items that display and allow selection of the target measurement point (reference data corresponding to the measurement point; for example, the point specified by the cursor in distribution map 90A), a message prompting for label data input, and fields for abnormality presence / absence, abnormality cause, and comments, similar to those described above.

[0118] The display of the comprehensive deviation anomaly discrimination screen 90, as shown in Figure 10, is not limited to display on each remote terminal 2 or each automated analysis system 3, but may be similarly applied to any computer or other device related to them. The screen 90, as shown in Figure 10, may be created and displayed by the remote terminal 2 based on the analysis result data received from the comprehensive analysis server 1. Alternatively, the comprehensive analysis server 1 may send screen data in the form of a web page or the like with content similar to screen 90 to the remote terminal 2, and the remote terminal 2 may display it as a web page or the like.

[0119] [Processing Sequence] Figure 13 shows an example of the processing sequence in the automated analysis data analysis system 300 of Figure 3. In step S101, each automated analysis system 3, for example, automated analysis system 302 in laboratory A, transmits instrument information, reagent information, and reaction process data to a remote terminal 2 (for example, remote terminal 303). In step S102, the remote terminal 2 creates reference data 701 (however, initially, this reference data does not include label data 706, etc.) as shown in Figure 8, based on the reaction process data, etc., obtained from each automated analysis system 3. In step S103, the remote terminal 2 transmits the reference data at the level of the automated analysis system 3 or laboratory to the integrated analysis server 1 (309).

[0120] In step S104, the comprehensive analysis server 1 creates and stores data (comprehensive analysis data) with a data structure 801 as shown in Figure 9, based on reference data acquired from each remote terminal 2. In step S105, the comprehensive analysis server 1 performs a comprehensive analysis based on the data with data structure 801 for common application to multiple target automatic analysis systems 3, and obtains and stores comprehensive analysis result information. This comprehensive analysis process includes comprehensive regression line information processing to calculate comprehensive regression line information 91 and comprehensive deviation discrimination reference line information processing to calculate comprehensive deviation discrimination reference line information 92. In step S106, the comprehensive analysis server 1 transmits information including the comprehensive regression line information 91 and comprehensive deviation discrimination reference line information 92 as comprehensive analysis result information to the target remote terminal 2.

[0121] In step S107, the remote terminal 2 displays a screen like the one in Figure 10 (screen for discriminating abnormal deviations 90) to the operator based on the comprehensive analysis result information received from the comprehensive analysis server 1. This screen also automatically displays interfaces such as the label data input screen 93 mentioned above. In step S108, the operator looks at the screen and checks the distribution map of the reaction process data, the comprehensive regression line information 91, and the comprehensive deviation discrimination reference line information 92. The operator also inputs or modifies label data as appropriate on the screen, following the label data input screen 93 and the label data input prompt screen 95. In step S109, the remote terminal 2 stores the label data entered on the screen in the reference data 701 in Figure 8 and sends the reference data containing that label data (in other words, the updated reference data) to the comprehensive analysis server 1. In a modified example, only the entered label data may be sent from the remote terminal 2 to the comprehensive analysis server 1.

[0122] In step S110, the comprehensive analysis server 1 updates the data in the data configuration 801 as shown in Figure 9, based on reference data including label data acquired from the remote terminal 2. In step S111, the comprehensive analysis server 1 performs a comprehensive analysis using the label data based on the updated data in the data configuration 801 and stores the comprehensive analysis result information. In step S112, the comprehensive analysis server 1 transmits the comprehensive analysis result information from step S111 to the remote terminal 2. In step S113, the remote terminal 2 displays a screen with updated content based on the comprehensive analysis result information from the comprehensive analysis server 1, similar to Figure 10. The steps from label data input in step S108 can be repeated as needed.

[0123] The data analysis system of Embodiment 1, similar to the system in Patent Document 1, targets multiple automated analysis systems and performs comprehensive deviation discrimination based on comprehensive regression line information 91. In the case of deviation discrimination based only on reference data from a single automated analysis system, the accuracy of deviation discrimination may be insufficient due to factors such as sample bias or differences in equipment. Therefore, the data analysis system of Embodiment 1 creates comprehensive regression line information 91 and comprehensive deviation discrimination reference line information 92 based on reference data from multiple automated analysis systems 3, and performs comprehensive deviation discrimination based on this information. At that time, the data analysis system of Embodiment 1 performs comprehensive analysis using the acquired label data and updates the comprehensive deviation discrimination reference line information 92, etc.

[0124] In the data analysis system of Embodiment 1, similar to the system in Patent Document 1, the evaluation process of the reaction process for each individual automatic analysis system 3 can be performed, for example, by a remote terminal 2. In contrast, the integrated analysis server 1 performs an integrated evaluation process of the reaction process as a unit that combines multiple automatic analysis systems 3, based on the reference data obtained as processing results from each remote terminal 2. The integrated analysis server 1 adjusts evaluation parameters 702 and learning model parameters 707 as a result of this processing. This is not limited to this, and as a modification, for example, the integrated analysis server 1 (data analysis system 10 in the case of Figure 4) may perform all evaluation and analysis processes, including individual analysis and integrated analysis.

[0125] [Processing by the integrated analysis server] Figure 14 shows the processing flow by the integrated analysis server 1 (second computer) in the data analysis system of Embodiment 1. The processing of this flow is mainly implemented by software program processing, for example, by the processor 3091 in Figure 5.

[0126] In step S1001, the integrated analysis server 1 periodically activates the integrated regression line information processing unit 30911 and the integrated deviation discrimination reference line information processing unit 30912, etc., when it communicates with each remote terminal 2 (step S103 in Figure 13). Periodically means, for example, once a day, or every time a predetermined number of new reference data (e.g., 1000) are accumulated. The trigger for communication connection between the remote terminal 2 and the integrated analysis server 1 may come from either the remote terminal 2 or the integrated analysis server 1.

[0127] Then, the processor of the integrated analysis server 1 collects reference data from the remote terminal 2. Specifically, the processor obtains reference data 701, as shown in Figure 8, from a connected remote terminal, for example, 303. The reference data 701 includes item code 703, device Lot 704 representing the automated analysis system 302 connected to the remote terminal 303, and reagent Lot 705 representing the reagents used in the automated analysis system 302.

[0128] In step S1002, the processor of the integrated analysis server 1 checks item code 703 included in the acquired reference data. Specifically, the processor checks whether item code 703 is registered in the data of the data configuration 801 of the integrated analysis server 1. If item code 703 is registered in the data configuration 801 (S1002-Yes), the process proceeds to step S1003; if it is not registered (S1002-No), the process proceeds to step S1006.

[0129] In step S1003, the processor of the integrated analysis server 1 checks whether device Lot 704 is already registered in the data of data configuration 801 with respect to item code 703, which is already registered. If the device Lot is already registered (S1003-Yes), the process proceeds to step S1004; otherwise, the process proceeds to step S1006.

[0130] In step S1004, the processor of the integrated analysis server 1 checks whether reagent Lot 705 is already registered in the data of data configuration 801 for the already registered instrument Lot 704. If reagent Lot 705 is already registered (S1004-Yes), the process proceeds to step S1005; otherwise, the process proceeds to step S1006.

[0131] In step S1005, the processor of the integrated analysis server 1 adds the reference data 701 acquired from the remote terminal 2 to the reference data 804 in the data of the data configuration 801. Based on the data configuration 801, the processor of the integrated analysis server 1 generates integrated regression line information 91 and integrated deviation discrimination criterion line information 92. Specifically, the processor generates integrated regression line information 91 and integrated deviation discrimination criterion line information 92 using the reference data 1 to N in Figure 9, on a unit basis of item code 703, instrument Lot 803, and reagent Lot 804. Details of the calculation process for this information will be explained in Figure 15 below. The processor stores the calculated integrated regression line information 91 and integrated deviation discrimination criterion line information 92 in a storage device 3093 or the like.

[0132] In step S1006, the processor of the integrated analysis server 1 adds information for item code 703, instrument lot 704, or reagent lot 705, which are not registered in the data configuration 801, to the data configuration 801 as new item codes.

[0133] In step S1007, the processor of the comprehensive analysis server 1 checks whether there is a remote terminal 2 to which the comprehensive analysis result information, including the comprehensive regression line information 91 and comprehensive deviation discrimination reference line information 92 calculated in step S1005, is distributed. The presence or absence of the remote terminal 2 to which the distribution is distributed is determined, for example, by whether the device Lot 803 and reagent Lot 804, which are the source data for generating the comprehensive deviation discrimination reference line information 92, are registered with the remote terminal 2 (303 or 310) connected to the comprehensive analysis server 1.

[0134] If it is determined that there is no remote terminal 2 for distribution (S1007-None), the process proceeds to step S1008. If it is determined that there is a remote terminal 2 for distribution (S1007-Present), the process proceeds to step S1009. Note that even if the device Lot 803 and reagent Lot 804 for generating the overall deviation discrimination reference line information 92 etc. are registered in the data configuration 801, there may be cases where there is no remote terminal 2 for distribution. For example, this could be the case when the reagent corresponding to the reagent Lot registered in the data configuration 801 was previously used in the automated analysis system 3 of the device Lot, but is no longer being used. Alternatively, this could be the case when the automated analysis system 3 corresponding to the device Lot 803 no longer exists among the automated analysis systems 3 currently managed by the remote terminal 2. The check in step S1007 is useful as a countermeasure for cases where previously used reagents etc. are no longer being used.

[0135] In step S1008, the processor of the integrated analysis server 1 displays a notification of an anomaly (that the destination remote terminal 2 does not exist) on the display screen of the display device, which is output device 3095 in Figure 5, and terminates the flow in Figure 14.

[0136] In step S1009, the processor of the comprehensive analysis server 1 distributes the comprehensive analysis result information, including the comprehensive regression line information 91 and the comprehensive deviation discrimination reference line information 92 generated in step S1005, to the remote terminal 2, and the flow in Figure 14 ends. The remote terminal 2, having received the comprehensive analysis result information from the comprehensive analysis server 1, then displays the aforementioned comprehensive deviation anomaly discrimination screen and performs label data input processing based on the comprehensive analysis result information. The remote terminal 2 may also perform deviation discrimination and anomaly detection processing based on the comprehensive analysis result information.

[0137] Regarding the destination check in step S1007, for example, if a new automated analysis system 3 is added as a new destination within the system shown in Figure 3, the same process as shown in Figure 14 should be performed for that new automated analysis system 3. If the type (instrument lot) and the reagents used (reagent lot) of the newly added automated analysis system 3 are the same as those used in the existing automated analysis system group, the existing comprehensive analysis result information can also be applied to the newly added automated analysis system 3.

[0138] [Details of the comprehensive analysis process] Figure 15 shows the detailed processing flow for the comprehensive analysis process (calculation of comprehensive regression line information 91 and comprehensive deviation discrimination criterion line information 92) in step S1005 of Figure 14. This comprehensive analysis process is performed, for example, when a sufficient amount of reference data has been collected and stored in step S1001 mentioned above. A sufficient amount is the amount necessary and sufficient for learning, etc.

[0139] Furthermore, the reference data used in step S1005 is split into two sets: one for training in step S1102 and another for evaluation in steps S1103 and S1104. Methods for splitting the reference data include, for example, holdout validation and k-cross-validation. In the two resulting sets of data (in other words, two sets of data), the evaluation parameters 702 are the same, but different sets of measurement data are used for training and evaluation. For example, two sets of data may be created by randomly selecting multiple points from a reference data set relating to the same combination of evaluation parameters 702 (e.g., the combination of A1 and Err). The first set of data is used for training, and the second set of data is used for evaluation.

[0140] In step S1101, the processor of the integrated analysis server 1 reads the reference data. This operation includes obtaining the evaluation parameters 702 and label data 706 from all the reference data (Figure 8) that have been stored. The processor then divides the read reference data into data for training in step S1102 and data for evaluation in step S1103, as described above.

[0141] In step S1102, the processor first performs a regression equation criterion calculation process (in other words, a regression line creation process) using the evaluation parameter 702 in Figure 8, and expresses the distribution of combinations of evaluation parameter 702 (for example, the combination of A1 and Err) using a regression equation. This regression equation criterion calculation process is always performed. Five types of regression equations can be used to express the distribution of evaluation parameter 702, for example, a zero-degree function, a linear function, a quadratic function, a logarithmic function, and an exponential function.

[0142] Each function has the following regression equation, where X and Y are combinations of evaluation parameters. Ck (k=1~4) represents the coefficients of the regression equation. Zero-degree function: Y = C1 Linear function: Y = C1X + C2 Quadratic function: Y = C1X² + C2X + C3 Logarithmic function: Y = C1 log(C3X + C4) + C2 Exponential function: Y = C1 exp(C3X + C4) + C2

[0143] The calculated regression equation corresponds to the composite regression line information 91 (Figure 10).

[0144] Furthermore, in step S1102, if the system is configured to perform learning, learning is performed in addition to creating the regression equation described above. Whether or not to perform learning, and the learning method, can be configured by the system settings or by the user through the operator settings. This learning is performed using the evaluation parameters 702 and label data 706 shown in Figure 8. In the example of Embodiment 1, this learning is supervised machine learning that performs a multi-class classification task. A multi-class classification task is performed because there can be multiple causes of anomalies. This supervised machine learning can be applied to, for example, multi-class logistic regression analysis or a multi-class neural network. The processor performs this machine learning using data with labels (label data 706) (also referred to as labeled data) as training information. Using this learning, the processor estimates the presence or absence of anomalies and the causes of anomalies in data without labels (label data 706) (also referred to as unlabeled data).

[0145] Furthermore, if there is insufficient labeled data available for the supervised machine learning described above, the processor may use clustering (e.g., K-means, k-mean++, etc.) to assign pseudo-labels to unlabeled data in the vicinity of the labeled data, and then perform supervised learning.

[0146] Furthermore, the processor may utilize the model parameters 707 shown in Figure 8 as initial parameters for the learning computation model during training. Model parameters 707 are parameter information applied to the learning computation model. The deviation judgment model parameters 30 are parameter information for the model used to determine deviations from the overall regression line information 91. The anomaly judgment model parameters 31 are parameter information for the model used to determine anomalies based on the overall deviation judgment criterion line information 92.

[0147] If training is performed in step S1102, the generalization performance evaluation in step S1103 is also performed. In step S1103, the processor of the integrated analysis server 1 evaluates the generalization performance of the multi-class classification model using labeled data for evaluation of the regression equation (in other words, regression line, regression function, especially the integrated regression line information 91) generated in step S1102 and the multi-class classification model. The labeled data used in this evaluation is the evaluation data that was not used in training in step S1102 (the other data from the aforementioned split).

[0148] Examples of generalization evaluation methods include plotting F-scores, AUC, ROC curves, or combinations thereof. Furthermore, the overall regression line information 91, which serves as the criterion for deviation, and the overall deviation discrimination criterion line information 92, which is used to determine the presence or absence of anomalies, are adopted as the values ​​that yield the highest generalization performance.

[0149] In step S1104, the processor of the integrated analysis server 1 calculates and evaluates an information criterion to prevent overfitting of the above model. Methods for calculating the information criterion include, for example, AIC or BIC.

[0150] In step S1105, the processor determines whether the calculations for all models related to the learning and evaluation described above are complete. If the calculations are complete (S1105-Y), the processor proceeds to step S1106; otherwise, it proceeds to step S1102, and steps S1102 to S1104 are repeated in the same manner.

[0151] In step S1106, the processor checks whether there is any performance improvement in all the above models. For example, the processor determines whether the generalization performance and information threshold of the model are better or worse than the model that was previously applied. If there is performance improvement (S1106-Y), the processor proceeds to step S1107. If there is no performance improvement (S1106-N), for example, if the generalization performance and information threshold of the model are worse than the model that was previously applied, the flow terminates. In step S1107, the processor selects a predetermined number of optimal or suitable models (for example, one optimal model) from among the models that have performance improvements. For example, the processor selects the optimal model with high generalization performance and a low information threshold, and updates the training and evaluation models to this model. The processor adopts the updated information, including the overall regression line information 91 and the overall deviation discrimination criterion line information 92, which will be the optimal threshold information, that are aligned with the updated optimal model.

[0152] [Effects, etc.] As described above, the data analysis method of Embodiment 1 improves accuracy and enables anomaly detection when performing deviation discrimination from the regression line of measurement data from an automated analysis system. In particular, Embodiment 1 improves accuracy when performing deviation discrimination and anomaly detection from the combined regression line information 91 targeting multiple automated analysis systems 3.

[0153] (1) According to Embodiment 1, the integrated analysis server 1, which is the second computer, acquires label data that was not previously used from the remote terminal 2, which is the first computer, and generates and updates suitable integrated regression line information 91 and integrated deviation discrimination reference line information 92 through machine learning and analysis using the label data. As a result, the accuracy of integrated deviation discrimination etc. targeting multiple automated analysis systems 3 can be improved compared to the conventional method.

[0154] (2) Furthermore, according to Embodiment 1, label data can be efficiently acquired from the operator through a screen like the one shown in Figure 10. The computer system (first computer and second computer) provides a function that allows the operator to input label data, including whether there is an abnormality and the cause of the abnormality, on the screen. As part of its function, the computer system provides the operator with a screen that has a GUI for inputting label data. In particular, the computer system estimates whether there is an abnormality in the measurement data near the overall deviation discrimination reference line information 92 on the screen, and provides a label data input screen or a label data input prompt screen for the measurement data that is estimated to have an abnormality. In particular, for the first measurement data that has already had label data entered and has been determined to have an abnormality, the computer system provides a label data input prompt screen to prompt, confirm, and inquire about label data input for the second measurement data that is in the vicinity, has not had label data entered, and is estimated to have an abnormality. These measures make the acquisition of label data more efficient.

[0155] The functions described in the data analysis system and method of Embodiment 1 can also be realized by a program (in other words, software program code) and a storage medium on which that program is recorded. The program can be written in various programming languages ​​or scripts, such as assembler, C / C++, or Perl. The program or storage medium is provided to the computer system of the embodiment. The processor of the computer system reads the program stored on the storage medium into memory, for example, and executes processing according to the program. This realizes the functions of the embodiment. Examples of storage media include flexible disks, CD-ROMs, DVD-ROMs, hard disks, optical disks, magneto-optical disks, CD-Rs, magnetic tapes, non-volatile memory cards, and ROMs.

[0156] The technology of the embodiments of this disclosure can be implemented using various general-purpose devices (e.g., PCs and servers on a communication network) without being limited to specific devices, or it can be implemented in dedicated devices (e.g., as part of an automated analysis system). Furthermore, the functions of the embodiments may be realized by an operating system (OS) or middleware running on a computer system performing part of the actual processing based on program instructions.

[0157] Although embodiments of this disclosure have been described in detail above, the present invention is not limited to these embodiments and can be modified in various ways without departing from the gist of the invention. Except for essential components, the embodiments can be modified by adding, deleting, or replacing components. Unless otherwise specified, each component may be singular or plural. Combinations of various configuration examples are also possible. [Explanation of Symbols]

[0158] 90...Screen (Screen for determining overall deviation anomaly), 91...Overall regression line information, 92...Overall deviation discrimination reference line information, 93...Label data input screen, 94...Nearby region, 95...Label data input prompt screen, 96...Label data input screen, 97...Label data input prompt screen, 99...Measurement data.

Claims

1. The first step involves the computer system acquiring, as reference data, instrument information, reagent information, and reaction process data including measurement data from each of the multiple automated analysis systems. The computer system performs a second step of acquiring label data, which includes whether there is an abnormality and the cause of the abnormality for each reaction process data, input by the operator. The computer system performs a third step of calculating, with respect to the distribution diagram of evaluation parameters of the reaction process data, comprehensive regression line information applicable to the multiple automated analysis systems, and comprehensive deviation discrimination reference line information for determining deviations from the comprehensive regression line information, based on the analysis of the reference data including the label data. The computer system performs a fourth step in which it displays the overall deviation discrimination reference line information on the screen, The computer system provides the screen with a distribution diagram of the evaluation parameters for the reaction process data and the overall deviation discrimination reference line information. The computer system estimates the possibility of an anomaly for the first point of the reaction process data based on its relationship to the overall deviation discrimination reference line information, displays a first interface on the screen for inputting the label data for the first point, and acquires the label data input by the operator to the first interface. It has, The computer system calculates the neighborhood region for the first point, displays information representing the neighborhood region on the screen, and highlights other points included within the neighborhood region. Data analysis methods.

2. In the data analysis method according to claim 1, The first interface includes an abnormality presence field for inputting whether or not there is an abnormality, and an abnormality cause field for inputting the cause of the abnormality, The computer system displays initial values ​​as options in the first interface, based on the estimation of whether an abnormality exists and the cause of the abnormality for the first point, in the abnormality presence / absence column and the abnormality cause column. Data analysis methods.

3. In the data analysis method according to claim 1, The first interface includes an abnormality presence field for inputting whether or not there is an abnormality, and an abnormality cause field for inputting the cause of the abnormality, The computer system includes a step of registering a new cause of abnormality, other than one of the multiple selectable causes of abnormality, in the cause of abnormality field of the first interface, based on the operator's operation. Data analysis methods.

4. The first step involves the computer system acquiring, as reference data, instrument information, reagent information, and reaction process data including measurement data from each of the multiple automated analysis systems. The computer system performs a second step of acquiring label data, which includes whether there is an abnormality and the cause of the abnormality for each reaction process data, input by the operator. The computer system performs a third step of calculating, with respect to the distribution diagram of evaluation parameters of the reaction process data, comprehensive regression line information applicable to the multiple automated analysis systems, and comprehensive deviation discrimination reference line information for determining deviations from the comprehensive regression line information, based on the analysis of the reference data including the label data. The computer system performs a fourth step in which it displays the overall deviation discrimination reference line information on the screen, The computer system provides the screen with a distribution diagram of the evaluation parameters for the reaction process data and the overall deviation discrimination reference line information. The computer system estimates the possibility of an anomaly for the first point of the reaction process data based on its relationship to the overall deviation discrimination reference line information, displays a first interface on the screen for inputting the label data for the first point, and acquires the label data input by the operator to the first interface. The computer system includes the step of calculating a neighborhood region for the first point, and for a second point located within the neighborhood region for which label data has not yet been entered, displaying a second interface on the screen to prompt input of label data by inquiring about the presence or absence of an anomaly and the cause of the anomaly for the second point. Data analysis methods.

5. In the data analysis method according to claim 4, The computer system includes the steps of displaying a third interface for inputting the label data for the second point based on an operation by the operator on the second interface, and acquiring the label data input by the operator on the third interface. Data analysis methods.

6. In the data analysis method according to claim 4, The computer system displays a message in the second interface that uses the same information as the content of the label data for the first point to inquire about the presence or absence of an abnormality and the cause of the abnormality. Data analysis methods.

7. In the data analysis method according to claim 5, The third interface includes an abnormality presence field for inputting whether or not there is an abnormality, and an abnormality cause field for inputting the cause of the abnormality, The computer system, in the third interface, displays initial values ​​as options in the abnormality presence / absence column and the abnormality cause column that are the same as the content of the label data for the first point. Data analysis methods.

8. A data analysis system that performs analysis related to an automated analysis system, Equipped with a computer system, The aforementioned computer system, From each of the multiple automated analysis systems, instrument information, reagent information, and reaction process data including measurement data are acquired as reference data. The operator inputs label data for each reaction process, including whether there is an abnormality and the cause of the abnormality. With respect to the distribution diagram of evaluation parameters of the reaction process data, based on the analysis of the reference data including the label data, comprehensive regression line information applicable to the multiple automated analysis systems and comprehensive deviation discrimination reference line information for determining deviations from the comprehensive regression line information are calculated. The aforementioned overall deviation discrimination reference line information is displayed on the screen. On the aforementioned screen, the distribution diagram of the evaluation parameters of the reaction process data and the overall deviation discrimination reference line information are displayed. With respect to the first point of the reaction process data, the possibility of an anomaly is estimated based on its relationship with the overall deviation discrimination reference line information, a first interface for inputting the label data for the first point is displayed on the screen, and the label data input by the operator to the first interface is acquired. The system calculates the neighborhood region for the first point, displays information representing the neighborhood region on the screen, and highlights other points included within the neighborhood region. Data analysis system.

9. A computer that performs analysis related to an automated analysis system, From each of the multiple automated analysis systems, instrument information, reagent information, and reaction process data including measurement data are acquired as reference data. The operator inputs label data for each reaction process, including whether there is an abnormality and the cause of the abnormality. With respect to the distribution diagram of evaluation parameters of the reaction process data, based on the analysis of the reference data including the label data, comprehensive regression line information applicable to the multiple automated analysis systems and comprehensive deviation discrimination reference line information for determining deviations from the comprehensive regression line information are calculated. The aforementioned overall deviation discrimination reference line information is displayed on the screen. On the aforementioned screen, the distribution diagram of the evaluation parameters of the reaction process data and the overall deviation discrimination reference line information are displayed. With respect to the first point of the reaction process data, the possibility of an anomaly is estimated based on its relationship with the overall deviation discrimination reference line information, a first interface for inputting the label data for the first point is displayed on the screen, and the label data input by the operator to the first interface is acquired. The system calculates the neighborhood region for the first point, displays information representing the neighborhood region on the screen, and highlights other points included within the neighborhood region. calculator.

10. In the data analysis method according to claim 1, The computer system estimates that there is a possibility of an anomaly in relation to the first point, even if other points included within the neighboring region of the first point are within the range of the comprehensive deviation discrimination reference line information. Data analysis methods.

11. A data analysis system that performs analysis related to an automated analysis system, Equipped with a computer system, From each of the multiple automated analysis systems, instrument information, reagent information, and reaction process data including measurement data are acquired as reference data. The operator inputs label data for each reaction process, including whether there is an abnormality and the cause of the abnormality. With respect to the distribution diagram of evaluation parameters of the reaction process data, based on the analysis of the reference data including the label data, comprehensive regression line information applicable to the multiple automated analysis systems and comprehensive deviation discrimination reference line information for determining deviations from the comprehensive regression line information are calculated. The aforementioned overall deviation discrimination reference line information is displayed on the screen. On the aforementioned screen, the distribution diagram of the evaluation parameters of the reaction process data and the overall deviation discrimination reference line information are displayed. With respect to the first point of the reaction process data, the possibility of an anomaly is estimated based on its relationship with the overall deviation discrimination reference line information, a first interface for inputting the label data for the first point is displayed on the screen, and the label data input by the operator to the first interface is acquired. The system calculates the neighborhood region for the first point, and for the second point located within the neighborhood region for which the label data has not yet been entered, it displays a second interface on the screen to prompt the user to input the label data by inquiring about the presence or absence of an anomaly and the cause of the anomaly for the second point. Data analysis system.

12. A computer that performs analysis related to an automated analysis system, From each of the multiple automated analysis systems, instrument information, reagent information, and reaction process data including measurement data are acquired as reference data. The operator inputs label data for each reaction process, including whether there is an abnormality and the cause of the abnormality. With respect to the distribution diagram of evaluation parameters of the reaction process data, based on the analysis of the reference data including the label data, comprehensive regression line information applicable to the multiple automated analysis systems and comprehensive deviation discrimination reference line information for determining deviations from the comprehensive regression line information are calculated. The aforementioned overall deviation discrimination reference line information is displayed on the screen. On the aforementioned screen, the distribution diagram of the evaluation parameters of the reaction process data and the overall deviation discrimination reference line information are displayed. With respect to the first point of the reaction process data, the possibility of an anomaly is estimated based on its relationship with the overall deviation discrimination reference line information, a first interface for inputting the label data for the first point is displayed on the screen, and the label data input by the operator to the first interface is acquired. The system calculates the neighborhood region for the first point, and for the second point located within the neighborhood region for which the label data has not yet been entered, it displays a second interface on the screen to prompt the user to input the label data by inquiring about the presence or absence of an anomaly and the cause of the anomaly for the second point. calculator.

Images