Data analyzing device

Abstract

One aspect of a data analyzing device according to the present invention comprises a model storage unit (31) for storing a plurality of trained models for extracting prescribed information using machine learning from analysis data acquired by analysis or observation of a sample, a script storage unit (32) for storing a plurality of scripts for executing rule-based calculation processing, an analysis recipe creating unit (23) for accepting a selection of at least one trained model and a script by a user, and creating an analysis recipe by combining at least one trained model and at least one script, and an analysis execution unit (25) for executing data processing with respect to the analysis data in accordance with the created analysis recipe to obtain a calculation processing result, wherein: each script is managed using a name and version information subordinate to the name; and the analysis recipe creating unit includes a selection acceptance processing unit for displaying a list of all the scripts that have version information differing from that of a designated script, and for accepting a selection of any script among the displayed scripts.

Description

Data analysis equipment

[0001] The present invention relates to an apparatus for analyzing data such as image data obtained from a sample or specimen.

[0002] In research and development of regenerative medicine using pluripotent stem cells such as iPS cells, ES cells, and mesenchymal stem cells, cell quality is confirmed and evaluated using cell images obtained with a cell observation device. Cell image analysis systems such as those disclosed in Patent Document 1 and Non-Patent Documents 1 and 2 are known for appropriately and efficiently managing cell image data obtained with the cell observation device and analyzing the image data to obtain information useful for cell evaluation, etc.

[0003] The cell image analysis system described in Non-Patent Document 1 ("Cell Pocket" manufactured by Shimadzu Corporation) TM ") allows the storage of multiple trained models generated by machine learning using a dataset consisting of cell images and labeled images that accurately indicate the areas occupied by cells on the cell images, as well as multiple rule-based analysis algorithms for calculating quantitative information such as cell count and cell area. A user selects an appropriate trained model and analysis algorithm depending on the analysis they wish to perform, creates an "analysis recipe" that combines these models, and performs image analysis using the analysis recipe. This makes it possible to obtain useful quantitative information about cell images. Non-Patent Documents 1 and 2 exemplify rule-based analysis algorithms, such as processes for quantitative calculations such as calculating coverage, measuring cell count, and calculating the area ratio of multiple extracted regions.

[0004] Processing on a computer in accordance with such an analysis algorithm is usually realized by a program with a relatively small amount of data (size) called a script. Therefore, in this specification, a program for realizing the rule-based processing or calculation algorithm described above will be called a "script," and will be distinguished from a "trained model," which may also be composed of a program with a relatively small amount of data.

[0005] In the conventional cell image analysis systems described above, scripts for performing the above-described routine calculations are generally created by the manufacturer of the system and pre-registered in the system. However, some users have a strong desire to obtain unique, less common quantitative information in order to gain new knowledge. For this reason, the systems also include a function that allows users to freely plug in (upload) their own scripts into the system and use them.

[0006] Japanese Patent Application Laid-Open No. 2023-69333

[0007] Shuhei Yamamoto and seven others, "Development and Application of a Web System for Cell Image Analysis," Shimadzu Review, March 20, 2022, Shimadzu Corporation, Vol. 78, No. 3-4, 2021. "Cell Pocket: A Web Application to Support Cell Observation," [Online], [Retrieved July 22, 2024], Shimadzu Corporation, Internet <URL: https: / / www.an.shimadzu.co.jp / sites / an.shimadzu.co.jp / files / pim / pim_document_file / an_jp / brochures / 20329 / c297-0523.pdf>

[0008] In the quantitative calculation process described above, in addition to cases where the user wishes to change the calculation algorithm itself, there are also cases where the user wishes to change only numerical parameters (part of the program) such as the calculation range, initial value, and threshold value included in the program while keeping the calculation algorithm itself completely or almost the same. In the conventional cell image analysis system described above, with regard to a script created by the user himself, it is possible to register a script in which only a part of the program has been modified or changed by the user, while the basic calculation program that is the core of the script remains the same, as a script completely different from the original script.

[0009] However, in conventional cell image analysis systems, when selecting a trained model and a script to create an analysis recipe, it is not easy for a user to determine whether another script that performs a quantitative calculation similar to or equivalent to the desired quantitative calculation, but that has been partially modified or changed, is registered in the system. This is particularly time-consuming when creating an analysis recipe using a script created by the user himself, which is one factor that reduces the efficiency of the analysis work. Furthermore, because it is difficult for users to recognize that multiple scripts perform the same type of quantitative calculation, this can lead to user errors, such as accidentally selecting a script different from the one they want to use to create an analysis recipe.

[0010] The present invention has been made to solve these problems, and its main object is to provide a data analysis device that allows a user to easily and reliably determine whether there is another script in which only a portion of the program of a certain script has been modified or changed, and that allows the user to accurately select the script they want to use with simple operations and create an analysis recipe.

[0011] One aspect of the data analysis device according to the present invention comprises: a model storage unit in which a plurality of trained models for extracting specified information from analytical data obtained by analyzing or observing a sample using machine learning are stored; a script storage unit in which a plurality of scripts for performing rule-based computational processing based on the specified information are stored; an analysis recipe creation unit that accepts a user's selection of at least one trained model stored in the model storage unit and at least one script stored in the script storage unit, and creates an analysis recipe indicating a procedure for data processing on the analytical data by combining the accepted at least one trained model and script; and a processing execution unit that performs data processing on the analytical data in accordance with the analysis recipe created by the analysis recipe creation unit to obtain computational processing results, wherein each of the scripts is managed by a name and version information subordinate to the name, and the analysis recipe creation unit includes a selection reception processing unit that displays a list of all scripts that have the same name as a specified script or are associated with the specified script and have different version information, and accepts the user's selection of any script from the displayed scripts.

[0012] In the data analysis device of the above aspect of the present invention, when a user creates an analysis recipe by combining a desired trained model and a script, other scripts that have substantially the same type of calculation processing as the selected script but differ only in part of the program, such as analysis processing parameters, and have different version information can be displayed in a list. This allows the user to easily and reliably recognize that other scripts registered in the system have been modified or changed only in part of the program, and that differ only in version information, and to create an analysis recipe by selecting the script that best suits the analysis purpose. This improves work efficiency when creating analysis recipes and prevents the user from performing an analysis using an inappropriate script due to a work error.

[0013] 1 is a schematic block diagram of an embodiment of an analysis system including an example of a data analysis apparatus according to the present invention. FIG. 1 is a flowchart showing an example of an analysis procedure in the analysis system of this embodiment. FIG. 2 is an explanatory diagram of a script management method in the analysis system of this embodiment. FIG. 3 is a diagram showing an example of a script registration setting screen in the analysis system of this embodiment. FIG. 4 is a diagram showing an example of an analysis recipe creation screen in the analysis system of this embodiment. FIG. 5 is a diagram showing a state in which a list of scripts with different version numbers is displayed when an analysis recipe is created in the analysis system of this embodiment. FIG. 6 is a diagram showing an example of a screen used when adjusting parameters during analysis execution in the analysis system of this embodiment. FIG. 7 is a diagram showing an example of a screen used when an analysis recipe is downloaded to a local environment in the analysis system of this embodiment. FIG. 8 is a diagram showing an example of a screen used when an analysis recipe is downloaded to a local environment in the analysis system of this embodiment. FIG. 9 is a diagram showing an example of a screen used when a learning dataset is created in the analysis system of this embodiment. FIG. 10 is a diagram showing an example of a screen used when a learning dataset is created in the analysis system of this embodiment. FIG. 11 is a diagram showing an example of a screen used when input data to be analyzed is confirmed in the analysis system of this embodiment. FIG. 12 is a diagram showing an example of a screen used when confirming input data to be analyzed in the analysis system of this embodiment. FIG. 13 is a diagram showing an example of a screen used to explain a method of allocating input data for an analysis recipe in the analysis system of this embodiment. 10A and 10B are diagrams showing an example of a screen for explaining a method of allocating input data of an analysis recipe in the analysis system of the present embodiment;

[0014] [Examples of the above aspects of the present invention] In the above aspects, "analysis data" refers to data obtained as a result of various analyses and measurements of samples or specimens, as well as by photography, observation, etc. Typically, analysis data is image data constituting an image, but is not limited to this and may also include text data, audio data, etc. Furthermore, when the analysis data is image data, for example, the "predetermined information" extracted from the analysis data may be information that is characteristic of the image or indicates a part or area that the user should pay attention to or is interested in.

[0015] Furthermore, the "rule-based calculation processing" performed by the script is typically calculation processing for obtaining some quantitative information, and for example, if the above-mentioned "predetermined information" is information indicating a specific area on the image, it may include the number of areas, the average value and variance of the area occupied by the area, the average value and variance of the perimeter of the area, the coverage rate, etc.

[0016] Although the functions performed by each unit in the above aspects can be realized by hardware circuits, it is more practical to realize at least a part of the functions by a computer. Therefore, each unit can be considered to be a functional unit that is realized when a predetermined computer program is executed on a computer.

[0017] [Configuration of an Example of Analysis System] A data analysis device as an example of the present invention will be described in detail below with reference to the accompanying drawings.

[0018] 1 is a block diagram of an embodiment of an analysis system including a data analysis device according to the present invention. This analysis system includes an image acquisition unit 1, an information processing unit 2, an input unit 4, and a display unit 5. As an example, this analysis system will be described as a cell image analysis system that analyzes the state of cells, etc. However, as will be described later, the images to be processed in the information processing unit 2 are not limited to cell images, and the data to be processed is not limited to image data. In this example of a cell image analysis system, the image acquisition unit 1 is a cell observation device that captures images of a subject, for example, a cell being cultured.

[0019] The information processing unit 2 is an example of a data analysis device according to the present invention, and includes, as functional blocks, a memory unit 3, a trained model creation and registration unit 20, a script creation and registration unit 21, a dataset creation unit 22, an analysis recipe creation unit 23, an analysis condition setting unit 24, an analysis execution unit 25, and a display processing unit 26. The memory unit 3 includes a data memory unit 30, a model storage unit 31, a script storage unit 32, and an analysis recipe storage unit 33.

[0020] The information processing unit 2 is actually a computer such as a personal computer or a more advanced workstation, or a computer system consisting of multiple computers. The storage unit 3 is composed of a storage device such as a hard disk drive (HDD) or a solid state drive (SSD) that is built into the computer or connected to the outside of the computer. The other functional blocks in the information processing unit 2 are realized by executing one or more dedicated software programs (computer programs) installed on the computer.

[0021] Such a computer program can be provided to the user by being stored on a computer-readable non-transitory recording medium such as a CD-ROM, DVD-ROM, memory card, USB memory (dongle), etc. Alternatively, it can be provided to the user in the form of data transfer via a communication line such as the Internet. Alternatively, it can be pre-installed on a computer that is part of the system when the user purchases the system.

[0022] The input unit 4 and display unit 5 connected to the information processing unit 2 are a user interface. Specifically, the input unit 4 is a keyboard and a pointing device (such as a mouse) attached to the computer, and the display unit 5 is a monitor display attached to the computer that is capable of displaying images.

[0023] The data storage unit 30 stores image data constituting a large number of images (cell images) acquired by the image acquisition unit 1, data to be analyzed such as image data read from other devices, and analysis result data obtained by the analysis execution unit 25. The model storage unit 31 stores models (source code files for generating trained models and making inferences using trained models) previously created by a manufacturer or the like that provides the system, and data constituting one or more trained models created based on the models in the trained model creation and registration unit 20. The script storage unit 32 stores scripts newly created in the script creation and registration unit 21 and multiple scripts for executing standard or general arithmetic processing previously created by a manufacturer or the like that provides the system. The analysis recipe storage unit 33 stores one or more analysis recipes created in the analysis recipe creation unit 23.

[0024] The trained model creation and registration unit 20 creates a trained model by performing machine learning using multiple training datasets, each of which is a pair of a given image (not limited to an image acquired by the image acquisition unit 1) and a label image corresponding to that image. In this example, the label image is an image that indicates the area occupied by a target cell (e.g., an undifferentiated cell) in a cell image, and is typically created manually by a specialized engineer. A trained model created by machine learning using a training dataset including such a label image is intended to detect the area in a given cell image where the target cell exists.

[0025] When a user wishes to perform analysis using a script unique to the user, the script creation and registration unit 21 creates a script in accordance with the user's instructions, and registers the created script and stores it in the script storage unit 32. The specific procedure for creating a script will be described later.

[0026] The dataset creation unit 22 creates a dataset to be used when performing machine learning in the trained model creation / registration unit 20 to create a trained model, in accordance with the user's selections and instructions, and provides the dataset to the trained model creation / registration unit 20.

[0027] The analysis recipe creation unit 23 creates an analysis recipe indicating the procedure for performing data processing on the image acquired by the image acquisition unit 1 using one or more trained models and one or more scripts, or using only one or more trained models, or only one or more scripts, in accordance with the user's selection or instructions, and stores the created analysis recipe in the analysis recipe storage unit 33.

[0028] Prior to executing the analysis, the analysis condition setting unit 24 determines various analysis conditions, such as the data to be analyzed, the analysis recipe to be used for the analysis, the analysis processing parameters to be used during the analysis, and the output format of the analysis results, in accordance with the selections and instructions of the user.

[0029] The analysis execution unit 25 executes the analysis in accordance with the analysis conditions set by the analysis condition setting unit 24. The display processing unit 26 performs display processing such as displaying the analysis results on the display unit 5 in a predetermined format.

[0030] The data handled by the information processing unit 2 is not limited to image data constituting the image captured by the image acquisition unit 1. For example, a chromatogram obtained by liquid chromatography analysis of a sample or a spectrum obtained by mass spectrometry or optical analysis can be treated as an image when drawn. Data obtained by surface analysis of a sample is also image data, and graphs such as heat maps obtained by various analyses of a sample are essentially image data. Therefore, the information processing unit 2 can also process such image data. Furthermore, data having a three-dimensional spatial extent can be considered as an extension of the dimensions of image data, and the information processing unit 2 can also handle such data. Furthermore, the information processing unit 2 can also handle various types of data other than image data.

[0031] [Outline of Analysis Procedure] Next, the procedure of the analysis process in the analysis system of this embodiment will be described in outline. Fig. 2 is a flowchart showing the outline of the procedure of the analysis performed in the information processing unit 2. The cell images obtained by the image acquisition unit 1 are sent to the information processing unit 2, and are temporarily stored in the data storage unit 30 in the information processing unit 2. The information processing unit 2 performs a predetermined analysis on the images stored in the data storage unit 30.

[0032] First, in response to a predetermined operation by a user via the input unit 4, the dataset creation unit 22 creates a training dataset required to create a trained model (step S1). The dataset includes, for example, a plurality of pairs of a cell image designated by the user and a label image corresponding to the cell image.

[0033] Next, the trained model creation and registration unit 20 creates a trained model by performing machine learning on the dataset based on a model previously stored in the model storage unit 31. The created trained model is then given an appropriate name and registered (stored) in the model storage unit 31 (step S2). Trained models are created according to the type of region to be detected in a cell image, or for each cell image containing different cell types. Therefore, the model storage unit 31 generally stores multiple trained models that differ in accordance with the purpose of analysis, etc. Note that, when performing analysis using a trained model already stored in the trained model storage unit 31, steps S1 and S2 are not necessary.

[0034] Furthermore, if a user wishes to perform analysis using a user-specific script, the script creation and registration unit 21 creates and registers a new script based on the user's instructions (step S3). If a script provided by the manufacturer and pre-stored in the script storage unit 32 is to be used for analysis, the operation of step S3 is unnecessary. Note that in this analysis system, the trained model is created using a model pre-provided by the manufacturer, rather than by the user. However, similar to the script, the user may also be able to create and register the model. In that case, the model creation and registration may be performed before the trained model is created in step S2, i.e., either before step S1 or between steps S1 and S2.

[0035] Next, in response to a predetermined operation by the user using the input unit 4, the analysis recipe creation unit 23 creates an analysis recipe and stores it in the analysis recipe storage unit 33 (step S4). Typically, an analysis recipe is a serial and / or parallel combination of one or more trained models and one or more scripts that are rule-based analysis algorithms. As will be described later, an analysis recipe for performing a desired analysis on an image can be created by the user selecting one or more of the trained models stored in the model storage unit 31 and one or more of the scripts stored in the script storage unit 32. Note that, if an analysis is performed using an analysis recipe already stored in the analysis recipe storage unit 33, the operation of step S4 is not necessary.

[0036] Next, when performing analysis, the user selects the data to be analyzed (step S5). That is, when the user performs a predetermined operation on the input unit 4, the analysis condition setting unit 24 reads out part of the image data stored in the data storage unit 30 and displays information specifying the image, thumbnail images, etc. on the display unit 5. After checking this, the user selects the image he or she wants to analyze, and in response, the analysis condition setting unit 24 specifies the image data to be analyzed.

[0037] Next, the user selects the analysis recipe to be used for analysis (step S6). That is, when the user performs a predetermined operation on the input unit 4, the analysis condition setting unit 24 reads out information specifying the analysis recipe stored in the analysis recipe storage unit 33 and displays this on the display unit 5. After checking this, the user selects the analysis recipe he or she wants to use, and the analysis condition setting unit 24 then specifies the analysis recipe.

[0038] Next, the user checks the settings of the analysis parameters, including the model and script, in the selected analysis recipe and performs an operation to change them as necessary. In response to this operation, the analysis condition setting unit 24 modifies the content of the analysis recipe (step S7).

[0039] Next, when the user instructs execution of the analysis, the analysis execution unit 25 executes data processing on the selected image data based on the trained model and script specified in the selected analysis recipe (step S8). As a result, for example, a region in the cell image where the target cell exists is extracted, and appropriate quantitative information about that region, such as coverage, is calculated. The display processing unit 26 displays, as the analysis results, for example, an image showing the extracted region on the cell image and text information and graphs showing the calculated quantitative information on the display unit 5 (step S9).

[0040] Through the above-described series of operations and processes, the user can obtain highly reliable quantitative information about the cell images obtained by the image acquisition unit 1.

[0041] [Script Management and Registration Method] Prior to describing the method for creating an analysis recipe, the script management and registration method (step S3) in this analysis system will be described with reference to Fig. 3 and Fig. 4. Fig. 3 is a conceptual diagram showing an example of management information for managing individual scripts. Fig. 4 is a schematic diagram showing an example of a script registration setting screen 110 for updating an existing script or adding a new script.

[0042] As described above, a script is a computer program with a relatively small amount of data for executing a predetermined computation. Each script is assigned a "name" and a "version number," as shown in FIG. 3 as examples, to identify the individual script. The "name" is the program name and can be arbitrarily set by the user under certain restrictions. The "version number," on the other hand, is identification information subordinate to the "name" and is represented by three numbers, such as "Ver. 1.0.0" in FIG. 3. The "name" and "version number" have a superior-subordinate (or parent-child) relationship. Typically, the "name" indicates the content of the computation executed by the script. When a part of the program, such as the calculation range or threshold value, included in the script, is changed, a new "version number" is assigned while the "name" remains the same. This version number can also be arbitrarily set by the user; however, when a user assigns a version number, only numbers greater than the existing version number for the script with the same name are permitted.

[0043] Furthermore, in order to group and manage multiple scripts used for the same purpose, a category named "Project" is provided above the name, and information indicating which project each script belongs to is recorded in the properties, etc., of each script. In the example of Figure 3, projects such as "Microscopic Image Analysis" for the purpose of cell analysis and "Raman Data Analysis" for analyzing image data obtained by Raman spectroscopy analysis are shown, but scripts can be grouped under any project. By managing multiple scripts together on a project-by-project basis in this way, even if the total number of scripts is large, their management is simplified, and necessary scripts can be easily checked and selected, for example, when creating an analysis recipe.

[0044] When a user wants to create a new script or wants to register a new script in which some of the analysis processing parameters, etc., included in an existing script have been rewritten, the user performs a predetermined operation on the input unit 4, which causes the script creation / registration unit 21 to display a script registration setting screen 110 as shown in FIG. 4 on the display unit 5. This script registration setting screen 110 is arranged with a project selection area 111, a script list table 112, an "Add" button 113, and the like. The project selection area 111 displays a list of the names of projects that are currently set, and provides a toggle switch for selecting / deselecting each project. Only the scripts included in the project selected by this toggle switch are displayed in the script list table 112.

[0045] The script list table 112 displays information about each script, such as its name, project name, version number, creator, and creation date and time. As described above, even if there are scripts with the same name but different version numbers, the script list table 112 displays only the script with the most recent version number. If the user wants to register a completely new script that is not in the script list table 112, the user clicks the "Add" button 113. In response to this operation, the script creation and registration unit 21 displays a registration screen (not shown) for registering a new script, superimposed on the script registration setting screen 110. The user enters necessary information such as the name and version number on the registration screen, and then performs an operation to load, for example, a newly created and temporarily saved script. As a result, the entered name and version number are assigned to the loaded script, and the script is registered as a new script.

[0046] On the other hand, if the user wishes to modify or change a portion of a script that has already been registered (i.e., is listed in the script list table 112) and register it as a script with a different version number, the user clicks on the version upgrade instruction button 115 in the row of the desired script in the script list table 112. In response to this operation, the script creation and registration unit 21 displays a registration screen (not shown) for registering the script with a newly assigned version number, superimposed on the script registration setting screen 110. The user inputs necessary information, such as the name and version number, on the registration screen, and then performs an operation to load a script that has been temporarily saved with a partially modified program, for example. As a result, the loaded script is registered as having a different version number from the original script. It is also possible to change the name of a script when upgrading, but even in this case, the name information before the change is maintained, and the script can be treated as a script with a different version number corresponding to the original script.

[0047] As described above, the version number can be changed by user operation only for scripts created and registered by the user himself / herself. For scripts provided by the manufacturer, the version upgrade instruction button 115 is disabled and clicking on this button 115 is not accepted. This is to ensure the reliability of scripts provided by the manufacturer, but it is also possible to allow version upgrades for scripts provided by the manufacturer. In that case, however, it is advisable to impose restrictions on version number changes, for example, so that it is easy to distinguish between program changes made by the user and script upgrades made by the manufacturer itself for the purpose of improving performance, etc.

[0048] In the analysis system of this embodiment, model versions are also managed using the same mechanism and method as the scripts described above. Meanwhile, multiple trained models that can be generated from an original model (a model with the same name and version) are managed under different version numbers. That is, unlike scripts, all trained models are created for each user, but like scripts, each trained model is managed by its name and its version number. Therefore, models and trained models can also be upgraded, just like scripts.

[0049] [Analysis Recipe Creation Method] Next, an example of an analysis recipe creation method (step S4) will be described with reference to Fig. 5 and Fig. 6. Fig. 5 is a schematic diagram showing an example of an analysis recipe creation screen 120, and Fig. 6 is a schematic diagram showing a state in which a list of scripts with different versions is displayed when creating an analysis recipe.

[0050] The analysis system of this embodiment allows for the creation of analysis recipes in a visually easy-to-understand manner through graphical operations on the display screen. When a user performs a predetermined operation on the input unit 4, the analysis recipe creation unit 23 responds by displaying an analysis recipe creation screen 120, as shown in FIG. 5, on the display unit 5. As shown in FIG. 5, the analysis recipe creation screen 120 has a node list area 121 on the left and a flow creation area 122 on the right. The node list area 121 displays a list of nodes (components) for creating an analysis recipe. The flow creation area 122 is used to construct a data processing flow by appropriately arranging nodes and connecting multiple nodes with lines. The "data processing" shown in the node list area 121 in FIG. 5 refers to a script. Note that FIG. 5 shows a state in which an analysis recipe is being created; at the start of analysis recipe creation, the flow creation area 122 is blank.

[0051] On the analysis recipe creation screen 120, the user selects one or more input nodes, one or more output nodes, one or more data processing (script) nodes, and one or more model nodes (nodes for source code files used to generate trained models and perform inference using trained models) displayed in the node list area 121 by clicking, placing them in the flow creation area 122 by, for example, drag-and-drop, and drawing connections between these nodes. Furthermore, the user sets trained models and inference parameters that can be used in the selected model node (i.e., registered) and for the script node, sets parameters in the script. This allows the user to create an analysis recipe to be used for analysis. The nodes can be connected in series or in parallel.

[0052] 5 , a symbol 123 indicating an input node named “Input Data,” a symbol 124 indicating an output node named “Output Data,” and a symbol 126 indicating a data processing (i.e., script) node named “Particle Number Measurement” are arranged in the flow creation area 122, and a connection line 127 connecting the symbols 123 and 126 and a connection line 128 connecting the symbols 126 and 124 are depicted. That is, “Input Data,” “Particle Number Measurement,” and “Output Data” are connected in series, which means that input data provided as “Input Data” is subjected to arithmetic processing by the script named “Particle Number Measurement,” and the results are output as “Output Data.” For example, if a user wants to change the script currently selected in the flow creation area 122 (the script named “Particle Number Measurement” in FIG. 5 ) to a script with a different name, the user can simply click to select a script with a different name displayed in the node list area 121.

[0053] On the other hand, if a version number is not assigned to the symbol 126 indicating a script node placed in the flow creation area 122, but the user wishes to check whether another version of this script exists, the user clicks on the information display instruction button 126a indicated by a circled letter "i" placed in the symbol 126. In response to this operation, the analysis recipe creation unit 23 displays a script detail information display screen 130 as shown in Fig. 6, superimposed on the analysis recipe creation screen 120. In Fig. 6, the script detail information display screen 130 is placed on the right edge of the analysis recipe creation screen 120, but the display position and size of the script detail information display screen 130 are not limited to this. However, it is preferable to display the script detail information display screen 130 in a position that does not overlap the symbol 126 indicating the script node.

[0054] The script detail information display screen 130 displays the script name 131 at the top, with an information type selection section 132 below it for selecting the type of detail information to display. To check the version information, simply select "Version" in the information type selection section 132, as shown in FIG. 6 . A display field 133 below displays detailed information about other versions of scripts registered in association with the script indicated by the name 131. This detailed information includes the name, version number, ID (identification information), creator, and text information. In the example of FIG. 6 , it can be seen that two scripts with the name "Particle Number Counting" and the version numbers "ver. 1.1.0" and "ver. 1.0.0" have been registered. The text information is text information that can be entered by the user (or manufacturer) when registering a new script or upgrading a script. Typically, the text information can be used to enter a brief explanation of the purpose of the upgrade and the changes made.

[0055] Among the circles 134 arranged in front of the names of the multiple scripts shown in the display field 133, circles shown in a dark color (which can actually be an appropriate display color such as green) in FIG. 5 (circles in front of the lower scripts) indicate scripts that are currently selected as nodes of the analysis recipe. On the other hand, circles shown in a neutral color (which can actually be an appropriate display color such as blue) in FIG. 5 indicate scripts that are not currently selected as nodes of the analysis recipe. From this display, the user can at a glance ascertain the version that is currently selected. Note that even if the name of a script is changed when upgrading the script, as described above, the information about the name before the change is maintained in a hidden state, and therefore such scripts are also displayed in the display field 133.

[0056] As described above, the user checks the information displayed in the display field 133 of the script detail information display screen 130, and if the user wants to select another script with a different version number (or name and version number) instead of the currently selected script, the user clicks the "Change to this version" button 135 in the display field 133 that corresponds to the script to be newly selected. In response to this operation, the analysis recipe creation unit 23 pops up a change confirmation screen (not shown). In response to the change OK operation being performed on this change confirmation screen, the analysis recipe creation unit 23 replaces the script displayed in the flow creation area 122, for example, from one corresponding to the script with the pre-change version number "ver. 1.0.0" to one corresponding to the script with the post-change version number "ver. 1.1.0".

[0057] In this way, with this analysis system, when creating an analysis recipe, the user can perform a simple operation to check all other scripts that have different version numbers, etc., associated with a certain script that is to be used in the analysis recipe, and then, if necessary, easily replace that script with a script of a different version.

[0058] The above explanation concerns operations and processing related to scripts, but as mentioned above, trained models created based on models are also managed by their names and their subordinate version numbers. Therefore, similar to scripts, when creating an analysis recipe, it is possible to check in a list whether a different version of the trained model to be used exists, and then, if necessary, easily replace the trained model to be used in the analysis recipe. Furthermore, similar to scripts, it is also possible to easily replace the model to be used in an analysis recipe when creating the analysis recipe.

[0059] [Procedure for Executing Analysis Using Analysis Recipe] Next, the characteristic operations and processes when executing an analysis using the analysis recipe created as described above (steps S5 to S8) will be described. When executing an analysis, the user performs an operation to select the data to be analyzed and the analysis recipe to be used for the analysis. Specifically, when the user performs a predetermined operation on the input unit 4, the analysis condition setting unit 24 displays an analysis condition setting screen 190 on the display unit 5, examples of which are shown in Figures 12 and 13. Figure 12 shows approximately the left half of one analysis condition setting screen 190, and Figure 13 shows approximately the right half of the same analysis condition setting screen 190.

[0060] The analysis condition setting screen 190 includes an analysis target data table 191, an analysis recipe selection area 192, a flow display area 193, a parameter set table 194, and the like. The user specifies data, such as images, to be analyzed in a data tree (not shown). The analysis condition setting unit 24 displays a list of the specified data, such as images, in the analysis target data table 191. The user then selects an analysis recipe to be used for analysis in the analysis recipe selection area 192. The analysis condition setting unit 24 displays a data processing flow (a diagram simply showing the connection state of multiple nodes) for the selected analysis recipe in the flow display area 193. The analysis condition setting unit 24 also displays, in the parameter set table 194, the names and IDs of the data files and program files selected corresponding to each node included in the analysis recipe displayed in the flow display area 193, each in one line.

[0061] That is, in this analysis system, the flow of the analysis recipe selected by the user and a list of data such as input images selected by the user are displayed on the same screen. In conventional cell image analysis systems, after selecting an image to be analyzed, when attempting to select an analysis recipe to be used for the analysis, it is not possible to confirm what image was selected as input. Therefore, the user must either remember the selected image or go back and check the selected image. This is particularly true when handling large amounts of data, and is prone to human error, such as not being able to select the appropriate analysis recipe for the image to be analyzed.

[0062] In contrast, as described above, in the analysis system of this embodiment, when selecting an analysis recipe, the data such as images selected as input can be confirmed in a list. This has the advantage of allowing the user to quickly select an appropriate analysis recipe for the input data. The analysis target data table 191 may also display attribute information such as image type and measurement target together with the file name. Furthermore, when displaying image data and the like in a list format in the analysis target data table 191, an input function for specifying data to be excluded from the analysis target may be further provided, so that some input data can be excluded when selecting an analysis recipe after being selected as input data.

[0063] Furthermore, in conventional cell image analysis systems, the trained model and its version used in the analysis recipe, as well as the analysis processing parameters specified therein, had to be set when the analysis recipe was created. Therefore, for example, if you wanted to change the version of the trained model included in an analysis recipe that had already been created, or if you wanted to change the analysis processing parameters, you had to go back to the analysis recipe creation process, create a new analysis recipe that reflected those changes, and then select the newly created analysis recipe when performing the analysis, which was a time-consuming process. In contrast, the analysis system of this embodiment makes it possible to change the trained model and its version and the analysis processing parameters even when attempting to perform an analysis using an analysis recipe.

[0064] Specifically, when a user performs a predetermined operation on the input unit 4 before executing an analysis, the analysis condition setting unit 24, in response to this operation, displays an analysis condition setting screen 140 on the display unit 5, an example of which is shown in FIG. 7 . The analysis condition setting screen 140 has a flow display area 141 arranged at the top and a parameter set table 142 arranged at the bottom. The flow display area 141 displays the processing flow of the analysis recipe currently selected. In this example, a model named "image region classification model" is selected as the trained model, and no script is used. The parameter set table 142 displays, in one line each, the names and IDs of data files and program files selected corresponding to each node included in the analysis recipe displayed in the flow display area 141. When the user clicks on a node symbol representing a model or script included in the analysis recipe displayed in the flow display area 141, the parameter set table 142 displays a list of parameters if the operation target is a script, or a list of trained models and inference parameters that can be used in that model if the operation target is a model.

[0065] In the example of Figure 7, as a result of clicking on the symbol of the node corresponding to the model named "Image Region Classification Model," one trained model that can be used here, "Ver. 1.0.0 for Cell Images," which is one version of the training model "for Cell Images," is displayed in a list in the selection display area 143 in the parameter set table 142. The user can use the list display in this selection display area 143 to check whether there are trained models generated from models selected as components of the analysis recipe and whether there are different versions of those trained models, and can change the trained model and its version to be used in the analysis as necessary.

[0066] Furthermore, as described above, if the trained model or script specified in the flow display area 141 includes an analysis processing parameter that can be changed by the user, a text box 144 for specifying the name and value of the analysis processing parameter, and / or controls such as a combo box, radio button, check box, or slider are displayed in the parameter set table 142. In the example of FIG. 7 , the specified trained model includes a batch size as an analysis processing parameter, so a text box 144 for specifying the batch size is displayed. The displayed value "2" is the default value or a value set most recently. If the user wishes to change this value, they can enter an appropriate value in the text box 144. Note that the analysis processing parameter referred to here is a value indicating the calculation range, threshold, constraint range, etc., when processing data in a trained model or script. Naturally, this affects the results of data processing, but does not involve any change in the essential processing, such as a change in the algorithm.

[0067] In this way, when the user changes the version of the trained model or script included in the analysis recipe or changes the analysis processing parameters as needed, and then clicks the "Analyze" button 145, the analysis condition setting unit 24 confirms the analysis conditions set at that time in response to this operation. Then, the analysis execution unit 25 executes the analysis according to the set conditions.

[0068] As described above, the analysis system of this embodiment not only allows for easy selection of the type and version of the model and script to be used in the analysis recipe when creating (designing) the analysis recipe, but also allows for easy confirmation of other versions of the trained models and scripts used in the selected analysis recipe when attempting to perform an analysis. If necessary, the user can then change the parameter values ​​or trained models to perform the analysis. Similarly, analysis parameters included in trained models and scripts can also be changed when attempting to perform an analysis. This eliminates the need to return to the analysis recipe creation stage, allowing for more efficient analysis than conventional cell image analysis systems. Of course, since the versions of the models and scripts used in the actual analysis are recorded, even if the versions or analysis parameters are changed before the analysis is performed, the user can easily confirm and verify the conditions under which the analysis was performed later.

[0069] Furthermore, the analysis target data table 191 arranged on the analysis condition setting screen 190 described above has an item called "file type," and by utilizing this file type, input data can be automatically distributed to multiple input nodes in the analysis recipe. This function will be described with reference to FIGS. 14 and 15.

[0070] 14 and 15 correspond to FIGS. 12 and 13. As shown in FIG. 15, the analysis recipe in this example has two input nodes arranged in parallel. The input data for both input nodes is image data. The file type can be selected for each symbol corresponding to the input node displayed in the flow display area 193. Here, as shown in FIG. 15, "Image A" is selected as the file type for the symbol of the upper input node, and "Image B" is selected as the file type for the symbol of the lower input node. Meanwhile, for the data to be input to the upper input node in the analysis recipe, "Image A" is set as the file type when the data is registered in the system, and similarly, for the data to be input to the lower node, "Image B" is set as the file type when the data is registered in the system. The analysis recipe is then invoked after selecting the data.

[0071] Whether the file type of each input data is "Image A" or "Image B" can be confirmed in the analysis target data table 191 arranged on the analysis condition setting screen 190, as shown in Fig. 14. By defining the file type in this way, when an analysis is performed, the data listed in the analysis target data table 191 is sorted as specified by the file type, and the analysis process is executed so that the data is input to the two input nodes, respectively.

[0072] In the examples shown in Figures 14 and 15, of the six data items named "imagen_training.png" (where n is 1 to 6) listed in the analysis target data table 191, the top three data items (191A in Figure 14) are input to the upper input node in the analysis recipe, and the bottom three data items (191B in Figure 14) are input to the lower input node in the analysis recipe. By utilizing file type settings in this way, input data can be sorted within a single analysis recipe as intended by the user. This allows a wider variety of analysis processes to be performed using this system.

[0073] [Analysis Recipe Download Function] As described above, analysis recipes created by a user are stored in the analysis recipe storage unit 33 and used when performing analysis. To accurately perform a desired analysis of data such as images, it is important to create an appropriate analysis recipe that matches that purpose. Typically, a person with relatively high levels of specialized knowledge is responsible for creating the analysis recipe. For example, within an organization (such as a company, university, or research institute), an analysis recipe created at one location may be desired to be used at another location, or an analysis recipe created at one company may be desired to be used at another company. However, in conventional cell image analysis systems, it was not possible to simply copy an analysis recipe created and stored on the main system to a local environment and use it in that local environment. Therefore, it was necessary to prepare an environment in which analysis recipes similar to those in the main system could be created in the local environment, and to create the analysis recipe in that local environment.

[0074] In contrast to this, the analysis system of this embodiment allows an analysis recipe created and saved in the main system to be copied to a local environment through a simple operation. The operation and processing for this purpose will be described below with reference to FIGS. 8 and 9.

[0075] When the user performs a predetermined operation on the input unit 4, including an operation to specify the analysis recipe that the user wants to copy to the local environment, the analysis recipe creation unit 23 responds by displaying an analysis recipe download setting screen 150, an example of which is shown in Figure 8, on the display unit 5.

[0076] The analysis recipe download setting screen 150 has a flow display area 152 in which the nodes of the selected analysis recipe are displayed, and a "Download recipe" button 151 is provided above it. When the user clicks the "Download recipe" button 151 after checking the contents of the analysis recipe, the analysis recipe creation unit 23, in response to this operation, obtains the necessary files corresponding to each node constituting the analysis recipe from the analysis recipe storage unit 33, the model storage unit 31, the script storage unit 32, etc., and creates a file in a predetermined compressed format (a ZIP file in this example) containing these files. Figure 9 shows an example of a file list screen 160 for checking the files corresponding to each node. The IDs of each node are listed in the name column of this file list screen 160.

[0077] When a compressed file is created, it is displayed in the browser's standard download function and becomes available for download. The compressed file downloaded to the local computer by the user's operation is decompressed on the computer to obtain multiple files related to the analysis recipe. These multiple files can be viewed by performing file operations on the local computer.

[0078] Furthermore, the analysis recipe creation unit 23 pops up, for example, a browser-standard download dialogue screen (not shown). When the user specifies a download destination (i.e., a file transfer destination) on this download dialogue screen and instructs execution of the download, the above-mentioned compressed file is downloaded and copied to the specified download destination computer, etc. The copied file thus contains all files necessary for performing analysis according to the analysis recipe, so that analysis according to the analysis recipe can be performed by batch processing on the download destination computer, independent of the main system.

[0079] In the analysis system of this embodiment, as described above, the files necessary for executing an analysis recipe can be copied from the main system to other computers with a simple operation, eliminating the need for the tedious task of creating an analysis recipe for each local computer. This improves user work efficiency. Furthermore, for example, since analyses can be performed using the same appropriate analysis recipe at multiple locations, the accuracy and reliability of comparisons and evaluations of analysis results obtained at the multiple locations are improved.

[0080] [Dataset Creation Function During Training] The analysis system of this embodiment has a function for creating a dataset to be used for training and validation when creating a trained model. Conventional cell image analysis systems also have similar functions, but the data that can be registered in the dataset is limited to image data, and they cannot support training of data other than images or multimodal training. In contrast, the analysis system of this embodiment provides greater flexibility in setting the data format and dataset structure that can be used in training datasets, thereby broadening the range of models that can be used in this system, expanding the targets and fields of data analysis, and also expanding the types of training and analysis processing algorithms that can be executed in this system.

[0081] Next, a method for creating a dataset for learning in this analysis system (step S1) will be described with reference to Figures 10 and 11. Figures 10 and 11 are both schematic diagrams showing an example of a dataset setting screen 170.

[0082] In the analysis system of this embodiment, the user pre-sets the structure of a dataset and the data format that can be used for the dataset using master data named "dataset category." That is, when the user performs a predetermined operation on the input unit 4, the dataset creation unit 22 displays a dataset category setting screen in a predetermined format on the display unit 5. On this dataset category setting screen, the user sets items such as the number of columns of input data that make up the dataset, the data format of each input column, the number of columns of label data that represent correct answers, and the data format of the label data, and then assigns an appropriate name to the dataset category. Then, when the user specifies the dataset category to be used when creating the dataset and performs a predetermined operation, a dataset setting screen 170, for example, as shown in FIG. 10 or 11, is displayed on the display unit 5.

[0083] The dataset setting screen 170 includes a category display area 171, an input data format setting area 172, and a dataset table 173. FIG. 10 shows an example of the dataset setting screen 170 corresponding to a dataset category with one input column and one label column, where both the input and label are in image data format. As shown in the figure, in this dataset category, the dataset table 173 includes items such as one vertical column of input (Input 1), one vertical column of labels (Label 1), and a mode. The user selects desired measured images and label images from those stored in advance in the data storage unit 30, and enters them into the Input 1 and Label 1 columns to create a dataset. For each dataset, the user also selects a mode for use—training, validation, or testing—in machine learning.

[0084] FIG. 11 shows an example of a dataset setting screen 170 corresponding to a dataset category with two input columns and one label column, where Input 1 is in the form of an image and Input 2 and Label 1 are in the form of text. As shown in the figure, in this dataset category, the dataset table 173 has two vertical columns of inputs (Input 1, Input 2), one vertical column of labels (Label 1), and a mode as items. The user then selects desired measured images and text information from those stored in advance in the data storage unit 30, and enters them into the Input 1 and 2 and Label 1 columns, respectively, to create a dataset. For each dataset, the user also selects a mode to use the dataset for training (Train), validation (Valid), or test (Test) in machine learning.

[0085] In all of the selectable modes, namely, training, validation, and testing, whether or not both measured images and labeled images are required can be set by an item called dataset type when creating a dataset. This dataset type is information indicating the purpose and use of the dataset, such as whether the dataset format is for use in supervised learning or semi-supervised learning, and the details of validation related to the pairing of measured images and labeled images included in the dataset are determined according to this information.

[0086] Here, the image data format is PNG, but other appropriate file formats such as JPEG and GIF can be used. The text data format may also be CSV or other formats other than TXT. Data in formats other than images and text, such as audio data, may also be handled.

[0087] In the analysis system of this embodiment, the number of columns of inputs and labels and the data format can be flexibly set according to the dataset to be created. Therefore, unlike conventional cell image analysis systems in which the number of inputs and the data format in a dataset are fixed, the types of models that can be learned or inferred in the system are expanded, and for example, learning targeting objects other than images and multimodal learning are also supported, making it possible to use the system for analyzing a wider variety of analytical data.

[0088] [Setting the Imported Data Format] The analysis system of this embodiment can handle not only captured image data such as cell images, but also data constituting externally imported graphs (e.g., chromatograms, mass spectra, etc.) as data to be analyzed. However, as is well known, there are a variety of file formats for graphs, etc., and it is not easy to support all imported graph formats. However, if the user cannot confirm the imported data on the display, it is difficult for the user to determine which data should be analyzed. In response to this, the analysis system of this embodiment can display imported graphs regardless of their file format, allowing the user to reliably confirm the data to be analyzed.

[0089] FIG. 16 shows an example of a screen used when using a function that enables such display. When a user performs a predetermined operation on the input unit 4 before importing (or uploading) a data file representing a desired graph into the system, the display processing unit 26 displays a graph format setting screen 200, such as that shown in FIG. 16, on the display unit 5. The user sets a name for the graph format in the name setting area 201 of the graph format setting screen 200, and selects "Data Viewer" from the options "Data Viewer," "Learning Curve," and "Accuracy Evaluation" in the graph format type selection box 202. The user then inputs a json (JavaScript Object Notation) file to be registered as a graph format into the json file registration area 203 by dragging and dropping it. This allows the graph format to be used on the data viewer screen to be registered in the system.

[0090] After performing these operations, if you set the json file data of the numerical data you want to register to use the above graph format when uploading the data, when you check the data on this data viewer screen, the numerical data will be displayed in a graph that conforms to the above graph format.

[0091] In this way, the analysis system of this embodiment can display and check data in a graph even if the data is in a file format that could not be displayed in a graph in conventional cell image analysis systems. This reduces the likelihood of mistakes when selecting data to be analyzed, and improves the efficiency of the analysis process.

[0092] As described above, the analysis system of this embodiment can improve the efficiency of analyzing various types of data, including image data, and also improve the accuracy and reliability of the analysis through various improvements and innovations. Furthermore, while conventional systems have been able to analyze only image data, such as cell images acquired by photography, the analysis system of this embodiment can also analyze data that has been converted into images, such as graphs, and data derived from text information, audio information, and the like, using models, trained models, and scripts, thereby obtaining accurate and precise quantitative information from the original information.

[0093] [Modifications] It goes without saying that the various functions achieved in the analysis system of the above embodiment do not necessarily need to be employed in combination, and only some of the functions may be employed. Furthermore, it goes without saying that the configurations and display screens shown in the drawings used in the above explanation are merely schematic or conceptual, and can be modified, added to, or deleted as appropriate.

[0094] Furthermore, the above-described embodiment and the various modified examples are merely examples of the present invention, and it goes without saying that any appropriate modifications, corrections, additions, etc. made within the spirit of the present invention will also be encompassed within the scope of the claims of the present application.

[0095] Various Aspects It will be apparent to those skilled in the art that the exemplary embodiments described above are examples of the following aspects.

[0096] (Item 1) One aspect of the data analysis device according to the present invention comprises: a model storage unit in which a plurality of trained models for extracting predetermined information from analytical data obtained by analyzing or observing a sample using machine learning are stored; a script storage unit in which a plurality of scripts for performing rule-based computational processing based on the predetermined information are stored; an analysis recipe creation unit that accepts a user's selection of at least one trained model stored in the model storage unit and at least one script stored in the script storage unit, and creates an analysis recipe indicating a procedure for data processing on the analytical data by combining the accepted at least one trained model and script; and a processing execution unit that executes data processing on the analytical data in accordance with the analysis recipe created by the analysis recipe creation unit to obtain computational processing results, wherein each of the scripts is managed by a name and version information subordinate to the name, and the analysis recipe creation unit displays a list of all scripts that have the same name as a specified script or are associated with the specified script but have different version information, and includes a selection reception processing unit that accepts the user's selection of any script from the displayed scripts.

[0097] In the data analysis device described in paragraph 1, when a user creates an analysis recipe by combining a desired trained model and a script, other scripts that have substantially the same type of calculation processing as the selected script but differ only in part of the program, such as analysis processing parameters, and have different version information can be displayed in a list. This allows the user to easily and reliably recognize that other scripts registered in the system have been modified or changed only in part of the program, and that differ only in version information, and allows the user to select the script that best suits the analysis purpose and create the analysis recipe. This improves work efficiency when creating analysis recipes and prevents the user from performing an analysis using an inappropriate script due to a work error.

[0098] (Clause 2) In the data analysis device described in paragraph 1, the analysis recipe creation unit arranges symbols representing input data, a trained model, a script, and output data on an analysis recipe creation screen and connects the symbols on the screen to graphically design an analysis recipe, and the selection reception processing unit, in response to a predetermined operation by a user on the symbol of the script, can display a screen displaying a list of all scripts that have the same name as the script or are associated with the script but have different version information, as part of the analysis recipe creation screen or superimposed on the analysis recipe creation screen.

[0099] According to the data analysis device described in paragraph 2, when creating an analysis recipe, the user can check the scripts with different versions by a very simple and easy-to-understand operation.

[0100] (Clause 3) In the data analysis device described in paragraph 2, the selection reception processing unit can display an operator on a screen displaying a list of all scripts that have the same name as the script or are associated with the script but have different version information, suggesting replacement of each of the other scripts except for the currently selected script with the script, and in response to a selection operation of the operator, perform an operation to replace the script displayed on the analysis recipe creation screen with the script having the selected version information.

[0101] According to the data analysis device described in paragraph 3, after checking the scripts of different versions, the user can replace the script used in the analysis recipe with a different version through a simple and easy-to-understand operation, thereby further improving the user's work efficiency and reducing work errors.

[0102] DESCRIPTION OF SYMBOLS 1... Image acquisition section 2... Information processing section 20... Model creation and registration section 21... Script creation and registration section 22... Data set creation section 23... Analysis recipe creation section 24... Analysis condition setting section 25... Analysis execution section 26... Display processing section 3... Storage section 30... Data storage section 31... Model saving section 32... Script saving section 33... Analysis recipe saving section 4... Input section 5... Display section

Claims

1. A data analysis device comprising: a model storage unit in which a plurality of trained models for extracting specified information from analytical data obtained by analyzing or observing a sample using machine learning are stored; a script storage unit in which a plurality of scripts for performing rule-based computational processing based on the specified information are stored; an analysis recipe creation unit that accepts a user's selection of at least one trained model stored in the model storage unit and at least one script stored in the script storage unit, and creates an analysis recipe indicating a procedure for data processing on the analytical data by combining the accepted at least one trained model and script; and a processing execution unit that executes data processing on the analytical data in accordance with the analysis recipe created by the analysis recipe creation unit to obtain computational processing results, wherein each of the scripts is managed by a name and version information subordinate to the name, and the analysis recipe creation unit displays a list of all scripts that have the same name as a specified script or are associated with the specified script but have different version information, and includes a selection reception processing unit that accepts the user's selection of any script from the displayed scripts.

2. The data analysis device according to claim 1, wherein the analysis recipe creation unit graphically designs an analysis recipe by arranging symbols representing input data, trained models, scripts, and output data on an analysis recipe creation screen and connecting the symbols on the screen, and the selection reception processing unit, in response to a predetermined user operation on the symbol of the script, displays a screen displaying a list of all scripts that have the same name as the script or are associated with the script but have different version information, as part of the analysis recipe creation screen or superimposed on the analysis recipe creation screen.

3. The data analysis device according to claim 2, wherein the selection reception processing unit displays an operator on a screen displaying a list of all scripts that have the same name as the script or are associated with the script but have different version information, suggesting replacement of each of the other scripts except for the currently selected script with the script, and in response to a selection operation of the operator, performs an operation to replace the script displayed on the analysis recipe creation screen with the script having the selected version information.

Images