Learning support method

Abstract

The present invention provides a learning support method, including: acquiring a first chromatogram (step S2), displaying an image of the first chromatogram (step S6), acquiring from a chromatogram DB a second chromatogram that is the same as or similar to the first chromatogram and second peak information for determining the peaks of the second chromatogram (step S4), displaying an image of the second chromatogram and an image of the second peak information (step S8), receiving an input of first peak information for determining the peaks of the first chromatogram by the user (step S10), and learning a prediction model based on the first chromatogram and the first peak information (step S14).

Description

Technical Field

[0001] The present disclosure relates to a learning support method. Background Art

[0002] International Publication No. 2017 / 040487 discloses a chromatography system. The chromatography system detects peaks by using an AI (Artificial Intelligence) of a presumption model, and performs qualitative analysis or quantitative analysis of a sample based on the peaks.

[0003] International Publication No. 2017 / 040487 discloses a technique in which a user can input training data for learning a presumption model. Specifically, the following technique is disclosed: the user visually observes an unseparated chromatogram in which peaks are not separated, and inputs information for specifying peaks as training data to the chromatography system. Summary of the Invention

[0004] In the generation of a presumption model, annotation of a large amount of training data is sometimes required, and it is desired to suppress deviation of the annotation to improve the quality of the presumption model.

[0005] The present invention has been made to solve such a technical problem, and an object thereof is to suppress deviation of annotation and improve the quality of a presumption model.

[0006] The learning support method of the present disclosure is a method for causing a computer to execute a process of supporting a learning operation of a presumption model for detecting peaks of a signal waveform acquired by an analysis device. The learning support method includes acquiring a first signal waveform generated by the analysis device. The learning support method includes displaying the first signal waveform on a display device. The learning support method includes acquiring, from a storage device storing a plurality of annotated signals, a second signal waveform having a high similarity to the first signal waveform and second peak information for determining peaks of the second signal waveform. The learning support method includes displaying the second signal waveform and a second peak information image showing the second peak information on the display device. The learning support method includes accepting an input of first peak information for determining peaks of the first signal waveform by the user. The learning support method includes learning the presumption model based on the first signal waveform and the first peak information.

[0007] The learning support program of the present disclosure is a program that causes a computer to execute processing for supporting a learning operation of a estimation model for detecting peaks of a signal waveform acquired by an analysis device. The learning support program causes the computer to execute acquiring a first signal waveform generated by the analysis device. The learning support program causes the computer to execute displaying the first signal waveform on a display device. The learning support program causes the computer to execute acquiring, from a storage device storing a plurality of labeled signals, a second signal waveform having a high similarity to the first signal waveform and second peak information for determining peaks of the second signal waveform. The learning support program causes the computer to execute displaying the second signal waveform and a second peak information image showing the second peak information on the display device. The learning support program causes the computer to execute accepting an input of first peak information for determining peaks of the first signal waveform by a user. The learning support program causes the computer to execute learning the estimation model based on the first signal waveform and the first peak information.

[0008] The above and other objects, features, aspects and advantages of the invention will be made apparent from the following detailed description of the invention in conjunction with the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Figure 1 is a diagram showing a configuration example of an analysis system.

[0010] Figure 2 is a diagram showing an example of assigning different training data to the same chromatogram.

[0011] Figure 3 is a block diagram showing a hardware configuration of the learning support device according to the present embodiment.

[0012] Figure 4 is a functional module diagram of the learning support device.

[0013] Figure 5 is an example of a chromatogram DB.

[0014] Figure 6 is an example of a screen displayed on a display device.

[0015] Figure 7 is an example of a screen displayed on a display device.

[0016] Figure 8 is an example of a screen displayed on a display device.

[0017] Figure 9 is a flowchart showing the processing of the learning support device.

[0018] Figure 10 is a flowchart showing the processing of the learning support device according to the second embodiment.

[0019] Figure 11It is a flowchart of the processing of the learning support device according to the third embodiment.

[0020] Figure 12 It is an example of a screen displayed on the display device. Specific embodiments

[0021] Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. In addition, the same or corresponding parts in the drawings are denoted by the same reference numerals, and the description thereof will not be repeated.

[0022] <First Embodiment>

[0023] [Analysis System]

[0024] The present disclosure relates to a technique for supporting learning of a estimation model for detecting peaks of a signal waveform generated by an analysis device. The analysis device is, for example, a gas chromatograph (GC) device, a liquid chromatograph (LC) device, a mass spectrometer, a spectrophotometer, an X-ray analysis device, or the like.

[0025] For example, the signal waveform can be a chromatogram waveform or a mass spectrum waveform. In addition, when the analysis device is a spectrophotometer, the signal waveform becomes an absorption spectrum waveform. When the analysis device is an X-ray analysis device, the signal waveform becomes an X-ray spectrum waveform.

[0026] In addition, the learning (learning process) of the estimation model (estimation model 121 described later) includes a process of newly generating (constructing) an estimation model that has not been constructed and a process of updating an already constructed estimation model. "Updating the estimation model" includes a process of updating the parameters of the estimation model. In addition, the estimation model updated (optimized) through the learning process is also referred to as a "learned model". In addition, the estimation model before learning and the learned estimation model are collectively referred to as an "estimation model".

[0027] In the present embodiment, an analysis device using liquid chromatography will be described. Figure 1 It is a diagram showing a configuration example of the analysis system 100. It includes an analysis device 10, a data analysis device 25, an input device 61, a display device 62, and a learning support device 30. In addition, the data analysis device 25 and the learning support device 30 are constituted by an information processing device (for example, a PC (personal computer: Personal Computer)), for example. In Figure 1 In the example of, the data analysis device 25 and the learning support device 30 are shown separately, but they may be integrated.

[0028] The input device 61 is, for example, a pointing device such as a keyboard or a mouse, and accepts instructions from the user. The display device 62 is constituted by, for example, a liquid crystal (LCD: Liquid Crystal Display) panel. The display device 62 displays various images. In the case where a touch panel is used as the user interface, the input device 61 is integrally formed with the display device 62. The input device 61 is connected to the data analysis device 25 and the learning support device 30. In addition, the display device 62 is connected to the data analysis device 25 and the learning support device 30.

[0029] The data analysis device 25 has a control unit 20. The control unit 20 controls the analysis device 10. The analysis device 10 includes a mobile phase container 11, a pump 12, a syringe 13, a chromatographic column 14, and a detector 15. A mobile phase is accommodated in the mobile phase container 11. The pump 12 sucks the mobile phase accommodated in the mobile phase container 11 and delivers it to the chromatographic column 14 at a substantially constant flow rate (or flow volume).

[0030] The syringe 13 injects a predetermined amount of a sample solution prepared in advance into the mobile phase at a predetermined timing corresponding to an instruction from the control unit 20. The injected sample solution is introduced into the chromatographic column 14 as the mobile phase flows. Various components contained in the sample solution are separated and eluted in the time direction during the passage through the chromatographic column 14. That is, the chromatographic column 14 separates the components contained in the sample solution according to the retention time.

[0031] The detector 15 detects the components in the eluate eluted from the chromatographic column 14. The detector 15 outputs a detection signal having an intensity corresponding to the amount of the component to the data analysis device 25. The detector 15 uses, for example, an optical detector such as a photodiode array (PDA: Photodiode Array) detector.

[0032] In addition to the above-described control unit 20, the data analysis device 25 further has a data collection unit 110, a peak detection processing unit 111, and an analysis unit 117.

[0033] The data collection unit 110 samples the detection signals output from the detector 15 at a predetermined time interval and converts them into digital data. The data collection unit 110 stores the digital data in a predetermined storage area (not shown). This digital data is data showing a chromatographic waveform (hereinafter also referred to as "chromatogram data").

[0034] The peak detection processing unit 111 uses AI (Artificial Intelligence) to estimate (derive) the peaks of the chromatogram based on the chromatogram data collected by the data collection unit 110.

[0035] In this embodiment, the peak detection processing unit 111 includes a model storage unit 114 and a peak determination unit 116. In addition, the model storage unit 114 stores, for example, a estimation model 121 (neural network) generated by machine learning. The estimation model 121 is represented by a prescribed function, for example. The prescribed function is, for example, an EMG (Exponentially Modified Gaussian) function.

[0036] In addition, the peak determination unit 116 inputs the chromatogram obtained based on the chromatogram data collected by the data collection unit 110 into the estimation model 121. The estimation model 121 outputs the peaks of the chromatogram. As described above, the peak detection processing unit 111 estimates the peaks of the chromatogram obtained based on the chromatogram data collected by the data collection unit 110 and outputs them to the analysis unit 117.

[0037] The time (retention time) when a peak is observed corresponds to the type of component. This chromatogram is sent to the data analysis device. The data analysis device determines the components based on the retention times of the peaks included in this chromatogram. This determination is also referred to as "qualitative analysis".

[0038] In addition, the height and area of the peaks of the chromatogram correspond to the concentration or content of the components of the sample. The data analysis device determines the concentration and content of the components of the sample based on the height or area value of the peaks included in this chromatogram. This determination is also referred to as "quantitative analysis".

[0039] The analysis unit 117 obtains the position (time) of the peak top and the area value (or height) of the peak among the peaks output from the peak determination unit 116. The analysis unit 117 determines the components based on the position information of each peak on the chromatogram. In addition, the analysis unit 117 calculates the content of each component based on the peak area value (or height value) using a pre-generated standard curve. In this way, the analysis unit 117 performs qualitative analysis and quantitative analysis of each component contained in the sample. The analysis unit 117 displays the qualitative analysis results and quantitative analysis results on the display device 62.

[0040] [Learning Support Device]

[0041] Next, the learning support device 30 will be described. As described above, in order to improve the accuracy of peak detection of the peak detection processing unit 111, the learning support device 30 optimizes the estimation model 121. In addition, in this embodiment, the estimation model 121 can be optimized by the manufacturer during the manufacturing stage of the analysis system 100. In addition, the analysis system 100 can be shipped to the user and the user can optimize the estimation model 121. In this case, the user prepares learning data for optimizing the estimation model 121 and performs the annotation process by the user himself / herself. Therefore, the user can generate the estimation model 121 that the user desires.

[0042] Generally, the performance of the presumptive model 121 obtained by machine learning is not perfect and is used on the premise that a certain degree of error is generated in peak detection. Thus, in the present embodiment, since the user himself / herself can learn the presumptive model 121, the convenience of the user can be improved.

[0043] The performance of the presumptive model 121 generally depends to a large extent on the quality of the learning data. In particular, it is preferable to be able to comprehensively cover various chromatograms and to be able to assign accurate training labels to the chromatograms.

[0044] In the present embodiment, there are two methods as methods for the user to optimize the presumptive model 121. The first method is a method in which the user performs a correction operation on the peaks detected by the peak detection processing unit 111. Specifically, the analysis system 100 displays a chromatogram image of the chromatogram and a peak information image of the peak information assigned to the chromatogram on the display device 62. This peak information image corresponds to the "detected peak image" described later. The chromatogram image is an image showing the chromatogram. The peak information image is an image showing the peak information. This peak information is information for determining the peaks of the chromatogram. The analysis system 100 accepts the correction of the displayed peak information by the user.

[0045] The second method is a method in which the user performs a peak designation operation on the chromatogram (chromatogram for which peaks have not yet been detected) newly collected by the data collection unit 110. Specifically, the analysis system 100 displays a chromatogram image on the display device 62 without displaying a peak information image. Then, the analysis system 100 accepts the input of peak information for the displayed chromatogram image. In addition, the input peak information becomes training labels or training data for optimizing the presumptive model 121.

[0046] Thus, in either the first method or the second method, the analysis system 100 accepts the input of peak information from the user as training data. As a result of the user inputting peak information, the parameters of the presumptive model 121 are updated with the input peaks and chromatograms as new learning data. The peak detection processing unit 111 can use the updated presumptive model 121. However, there is a case where the accuracy of the presumptive model 121 decreases when the user makes a label that is not consistent with the past annotation (a deviation occurs).

[0047] In addition, there is a case where a large amount of annotation of training data is required in the generation of the presumptive model 121. Figure 2 It is a diagram showing an example of assigning different training data (performing inconsistent annotation) to the same chromatogram. Figure 2 (A) shows a case where the user designates the peaks of the chromatogram widely. Figure 2(B) is a diagram showing a case where a user designates a peak of a chromatogram to be narrow. In addition, in the diagrams showing chromatograms described later, the horizontal axis represents time and the vertical axis represents signal intensity. Further, in the present embodiment, the peak information (training data) input by the user becomes peak information 92A and peak information 92B. Figure 2 In the diagrams showing chromatograms described later, the horizontal axis represents time and the vertical axis represents signal intensity. Further, in the present embodiment, the peak information (training data) input by the user becomes peak information 92A and peak information 92B.

[0048] For example, for the first chromatogram, the user inputs peak information 92A (refer to Figure 2 (A)), and for the second chromatogram, the user inputs peak information 92B (refer to Figure 2 (B)). In this case, different peak information (training data) is input for the same chromatogram, that is, a deviation in annotation occurs. Thus, if a deviation in annotation occurs, the quality of the estimation model 121 may sometimes decrease.

[0049] Here, the learning support device 30 of the present embodiment promotes the information input by the user to suppress the deviation in annotation (having consistency with past peak information). Thereby, the learning support device 30 can support the user's learning of the estimation model 121.

[0050] [Hardware configuration of the learning support device]

[0051] Figure 3 is a block diagram showing the hardware configuration of the learning support device 30 of the present embodiment. As Figure 3 shown, the learning support device 30 includes a control device 51, a storage device 52, a medium reading device 17, a display interface 18, and an input interface 26 as main hardware elements.

[0052] As described later, the control device 51 performs learning of the estimation model 121. The control device 51 is composed of, for example, a CPU (Central Processing Unit), an FPGA (Field Programmable Gate Array), and a GPU (Graphics Processing Unit). In addition, the control device 51 may be composed of at least one of a CPU, an FPGA, and a GPU, or may be composed of a combination of a CPU and an FPGA, an FPGA and a GPU, a CPU and a GPU, or all of a CPU, an FPGA, and a GPU. Further, the control device 51 may be composed of a processing circuitry.

[0053] The storage device 52 includes a volatile storage area (such as a work area) that temporarily stores program codes or working memories when the control device 51 executes any program. For example, the storage device 52 is composed of a volatile storage device such as a DRAM (Dynamic Random Access Memory) or an SRAM (Static Random Access Memory).

[0054] Furthermore, the storage device 52 includes a non-volatile storage area. For example, the storage device 52 is composed of a non-volatile storage device such as a hard disk or an SSD (Solid State Drive).

[0055] In addition, in the present embodiment, an example in which the volatile storage area and the non-volatile storage area are included in the same storage device 52 is shown, but the volatile storage area and the non-volatile storage area may also be included in different storage devices. For example, the control device 51 may include a volatile storage area, and the storage device 52 may include a non-volatile storage area. The learning support device 30 may also include a microcomputer including the control device 51 and the storage device 52.

[0056] The storage device 52 stores the estimation model 121, the control program 122, and the chromatogram DB (Data Base) 123. The estimation model 121 includes a neural network and parameters used in the processing in the neural network. The estimation model 121 is composed of a convolutional neural network (CNN: Convolution Neural Network) or the like. The control program 122 is a program executed by the control device 51.

[0057] The estimation model 121 includes at least a program capable of performing machine learning, and performs machine learning based on learning data (training data) to optimize (adjust) the parameters. The learning support device 30 sends the optimized estimation model 121 to the data analysis device 25. The data analysis device 25 updates the estimation model 121 stored in the model storage unit 114 to the sent estimation model 121 (optimized estimation model). In this way, the peak detection unit 111 uses the updated estimation model 121 to estimate peaks, thereby improving the estimation accuracy.

[0058] The medium reading device 17 receives a recording medium 130 such as a removable disk and acquires the data stored in the recording medium 130. This data is, for example, a control program. In addition, the control program 122 may also be stored in the recording medium 130 (such as a removable disk) and circulated as a program product. In addition, the control program 122 may be provided by an information provider as a so-called program product that can be downloaded via the Internet or the like. The control device 51 reads the program provided by the recording medium 130 or the Internet or the like. The control device 51 stores the read program in a prescribed storage area (the storage area of the storage device 52). The control device 51 executes the learning support process described later by executing the stored control program 122.

[0059] The recording medium 130 is not limited to DVD-ROM (Digital Versatile Disk Read Only Memory), CD-ROM (compact disc read-only memory), FD (Flexible Disk), hard disk, and may also be set as a medium that fixedly carries a program such as a magnetic tape, cassette tape, optical disc (MO (Magnetic Optical Disc) / MD (Mini Disc) / DVD (Digital Versatile Disc)), optical card, mask ROM, EPROM (Electrically Programmable Read-Only Memory), EEPROM (Electrically Erasable Programmable Read-Only Memory), flash ROM, or other semiconductor memories. In addition, the recording medium 130 is a non-transitory medium from which a computer can read the control program 122 and the like.

[0060] The display interface 18 is an interface for connecting the display device 62 and realizes the input / output of data between the learning support device 30 and the display device 62. The input interface 26 is an interface for connecting the input device 61 and realizes the input / output of data between the learning support device 30 and the input device 61.

[0061] [Function Modules of the Learning Support Device]

[0062] Figure 4It is a functional block diagram of the learning support device 30. As described above, the learning support device 30 includes a control device 51 and a storage device 52. The control device 51 further includes an acquisition unit 32, a processing unit 34, and an output unit 36.

[0063] The acquisition unit 32 acquires the information input by the user through the input device 61. This information is, for example, the above-mentioned peak information. In addition, every time a chromatogram is collected, the data collection unit 110 sends the chromatogram data to the learning support device 30. The acquisition unit 32 acquires the chromatogram data from the data collection unit 110. The information (peak information and chromatogram data) acquired by the acquisition unit 32 is output to the processing unit 34.

[0064] The processing unit 34 performs processing corresponding to the category of the information output from the acquisition unit 32. When peak information is input through the acquisition unit 32, the parameters of the estimation model 121 are updated based on the peak information. In addition, when chromatogram data is input through the acquisition unit 32, the chromatogram DB 123 is updated. The processing unit 34 also performs various other processes.

[0065] The output unit 36 outputs various signals or information. For example, the output unit 36 sends the image data of the image to be displayed on the display device 62 to the display device 62. The display device 62 displays an image based on the image data. In addition, every time the estimation model 121 is updated, the output unit 36 outputs the updated estimation model to the model storage unit 114. The model storage unit 114 stores the updated estimation model.

[0066] [Chromatogram DB]

[0067] Figure 5 It is an example of the chromatogram DB 123. As described above, the chromatogram DB 123 is stored in the storage device 52. The chromatogram DB 123 associates chromatogram data, stored peak information, characteristic quantities of the chromatogram shown by the chromatogram data, and analysis results in a chromatogram ID (identification). In addition, the chromatogram data and the stored peak information are also referred to as "fully annotated signals".

[0068] In addition, S (S is an integer of 1 or more) chromatogram IDs are stored in the chromatogram DB 123. In the chromatogram DB 123, chromatograms generated in the past by the analysis device 10 or a device equivalent to the analysis device 10, characteristic quantities of the chromatograms, analysis results derived from the chromatograms, etc. are stored. In other words, one or more (or multiple) fully annotated signals are stored in the chromatogram DB 123.

[0069] The chromatogram ID is information for identifying a chromatogram. Chromatogram data is data showing a chromatogram, which is digital data collected by the data collection unit 110. The chromatogram data corresponds to the "stored signal waveform" of the present disclosure.

[0070] The stored peak information is data for determining the peaks included in the chromatogram. The peak information is shown, for example, by the second peak information image 253 shown below. The stored peak information is information showing the peaks detected by the peak detection processing unit 111 or the peak information input by the user. Figure 6 The stored peak information is information showing the peaks detected by the peak detection processing unit 111 or the peak information input by the user.

[0071] The chromatogram feature quantity is a feature quantity of the chromatogram shown by the chromatogram data. In the example of Figure 5 , the chromatogram feature quantity includes the number of peaks in the chromatogram, the gradient between two points, and the area value. The gradient between two points refers to the gradient of the line segment connecting the start point and the end point of the peak. The area value is the area value of the region surrounded by the line segment forming the peak of the chromatogram and the line segment based on the gradient between two points. In addition, the peaks shown by the chromatogram feature quantity are the peaks shown by the stored peak information.

[0072] In addition, the gradient between two points and the area value are also feature quantities of the peak (hereinafter also referred to as "peak feature quantities"). The number of existing peak feature quantities corresponds to the number of peaks. For example, the chromatogram with the chromatogram ID of C1 shows that the number of peaks is 3. In addition, as the gradients between two points of the three peaks, E11, E12, and E13 are shown. In addition, as the area values of the three peaks, M11, M12, and M13 are shown.

[0073] The analysis result is a result derived based on the peaks detected by the peak detection processing unit 111 using the current estimation model 121 as the peaks of the chromatogram corresponding to the chromatogram data of the analysis result. The analysis result includes at least one of a qualitative analysis result and a quantitative analysis result. The qualitative analysis result is a result showing the components determined by the chromatogram. The quantitative analysis result is a result showing the amount of the component. In addition, as a modification example, the analysis result may be set to include the qualitative analysis result and not include the quantitative analysis result.

[0074] In the example of Figure 5 , the chromatogram data D1, the number of peaks G1, the gradient between two points E1, the area value M1, the qualitative analysis result P1, the quantitative analysis result Q1, and the stored peak information R1 are associated with the chromatogram ID of C1.

[0075] In this way, the chromatogram DB123 stores at least one stored signal waveform generated by the analysis device and at least one stored peak information for determining the peaks of each of the at least one stored signal waveform.

[0076] [Learning Support]

[0077] Next, a method for the learning support device 30 of the present embodiment to support a user's learning will be described. The learning support device 30 executes learning support for the user by displaying various images on the display area 62A of the display device 62. Figures 6 - 8 It is a diagram showing an example of such various images of the present embodiment. In the present embodiment, first, the Figure 6 image is displayed, and then the Figure 7 image is displayed, and then the Figure 8 image is displayed.

[0078] The user performs a predetermined operation on the input device 61, thereby transferring the mode of the analysis system 100 to the learning mode. In the learning mode, the user updates the estimation model 121 by inputting the first peak information described later.

[0079] As Figures 6 - 8 shown, the display area 62A includes an editing area 131A and a past result area 132A adjacent to the editing area 131A. In other words, the editing area 131A and the past result area 132A are set on the same screen. In the editing area 131A, an image of the chromatogram (hereinafter also referred to as the "first chromatogram") derived by the analysis device 10 (hereinafter also referred to as the "first chromatogram image 211") is displayed. The first chromatogram corresponds to the "first signal waveform" of the present disclosure. The first chromatogram image 211 corresponds to the "first waveform image" of the present disclosure. After the learning support device 30 acquires the chromatogram data collected by the data collection unit 110, it displays the image of the chromatogram of the chromatogram data as the first chromatogram image 211.

[0080] In addition, when acquiring the first chromatogram, the learning support device 30 determines a second chromatogram (second signal waveform) that is the same as or similar to the first chromatogram from the chromatogram DB ( Figure 5 ), and acquires the determined second chromatogram. Here, an example of a method for determining the second chromatogram that is the same as or similar to the first chromatogram will be described.

[0081] The learning support device 30 calculates the first similarity of each of the S stored chromatogram data stored in the chromatogram DB to the first chromatogram. That is, the learning support device 30 calculates S first similarities. The first similarity indicates the degree of similarity between one stored chromatogram data and the first chromatogram. The more similar one stored chromatogram data is to the first chromatogram, the larger the value of the first similarity becomes. In the present embodiment, the first similarity is expressed in %. In addition, the first similarity of the second chromatogram that is the same as the first chromatogram is 100%.

[0082] Calculate the first similarity according to the following two viewpoints. As the first viewpoint, the learning support device 30 calculates the first similarity based on the feature quantity of the first chromatogram (hereinafter also referred to as "the first feature quantity") and the feature quantity of the same category as this feature quantity (the first feature quantity) stored in the chromatogram (hereinafter also referred to as "the second feature quantity").

[0083] In the present embodiment, the first feature quantity is a feature quantity derived based on the peaks detected by the peak detection unit 111 using the current estimation model 121 as the peaks of the first chromatogram. In the present embodiment, it is a feature quantity of the same category as the Figure 5 second feature quantity of the stored chromatogram. Specifically, the first feature quantity and the second feature quantity are the number of peaks, the gradient between two points, and the area value. In addition, as long as the first feature quantity and the second feature quantity are of the same category, they may also be other categories. The other categories may be set to at least one of the peak width, peak resolution, peak fronting, peak tailing, etc.

[0084] As the second viewpoint, the learning support device 30 calculates the first similarity based on the analysis result shown (derived) by the first chromatogram and the analysis result shown by the stored chromatogram. Here, the analysis result shown by the stored chromatogram is as Figure 5 shown, and includes the qualitative analysis result and the quantitative analysis result corresponding to the stored chromatogram. In addition, the analysis result shown by the first chromatogram is the qualitative analysis result and the quantitative analysis result shown by the first chromatogram (that is, the analysis result of the same category as the analysis result shown by the stored chromatogram).

[0085] For example, the learning support device 30 compares the qualitative analysis result shown by the first chromatogram with the qualitative analysis result corresponding to the stored chromatogram. If the qualitative analysis results of both sides are different, the first similarity is set to "0". On the other hand, if the qualitative analysis results of both sides are the same, the learning support device 30 calculates the similarity of the quantitative analysis results of both sides (hereinafter also referred to as "the first provisional similarity"). In addition, the learning support device 30 calculates the second provisional similarity based on the above-mentioned first feature quantity and the above-mentioned second feature quantity. For example, the correlation coefficient is used to calculate the second provisional similarity. In addition, the second provisional similarity may also be calculated as the similarity between the shape of the first chromatogram and the shape of the stored chromatogram.

[0086] Then, the learning support device 30 calculates the first similarity based on the first provisional similarity and the second provisional similarity. In this way, the learning support device 30 comprehensively calculates S (the number of chromatogram IDs stored in the chromatogram DB) first similarities by combining the similarity (the first similarity) calculated through the first viewpoint and the similarity (the first similarity) calculated through the second viewpoint.

[0087] Further, as a modification example, the learning support device 30 may also calculate the first similarity degree from either the first viewpoint or the second viewpoint.

[0088] The learning support device 30 determines the first similarity degrees that are equal to or higher than the first threshold value based on the calculated S first similarity degrees. The learning support device 30 determines the stored chromatogram of the determined first similarity degree as the second chromatogram. Here, the first threshold value is a preset threshold value. For example, the first threshold value is 70%. That is, the learning support device 30 can determine, from the viewpoint of the first feature quantity and the analysis result, a chromatogram (second chromatogram) whose similarity degree to the first chromatogram is 70% or higher. Further, the learning support device 30 obtains peak information (second peak information) corresponding to the chromatogram ID of the second chromatogram by referring to the chromatogram DB123.

[0089] When the learning support device 30 obtains the first chromatogram, the second chromatogram, and the second peak information, as Figure 6 shown, it displays the first chromatogram image 211 of the first chromatogram, the second chromatogram image 212 of the second chromatogram, and the second peak information image 253 of the second peak information. In Figure 6 the example, there are displayed the second chromatogram images 212A, 212B, 212C, and 212D of the four second chromatograms respectively and the second peak information images 253 corresponding to the four second chromatogram images 212 respectively. Hereinafter, the four second chromatogram images 212A, 212B, 212C, and 212D will also be referred to as "second chromatogram images 212".

[0090] As described above, the second chromatogram image 212 is an image of the second chromatogram having a high similarity degree (equal to or higher than the first threshold value) to the first chromatogram. Therefore, by visually recognizing the second peak information image 253 of the chromatogram having a high similarity degree to the first chromatogram, the user can identify the peaks of the second chromatogram.

[0091] As Figure 6 shown, in a state where the first chromatogram image 211 is displayed, the learning support device 30 receives an input of peak information for determining the peaks of the first chromatogram shown in the first chromatogram image 211 from the user. This peak information corresponds to the "first peak information" in the present disclosure.

[0092] Figure 7 is an image when the user has input the first peak information. In Figure 7 the example, an image of the input first peak information is displayed as the first peak information image 220. The learning support device 30 updates the parameters of the estimation model 121 so that the peaks determined based on the input first peak information are detected as the peaks of the first chromatogram by the peak detection processing unit 111.

[0093] Thus, in the present embodiment, the user can input the first peak information that is consistent with the past second peak information (the second peak information displayed in the past result area 132A). That is, for example, it is possible to suppress a situation where different peak information is input for the same chromatogram (refer to Figure 2 ). And, the learning support device 30 can update the estimation model 121 based on the first peak information. Therefore, it is possible to improve the accuracy of updating the estimation model 121 from the perspective of improving the accuracy of determining the peak of the first signal waveform.

[0094] In addition, the learning support device 30 calculates the first similarity between the first chromatogram and each of the S stored chromatograms, and acquires the stored chromatograms with the first similarity being equal to or greater than the first threshold value as the second chromatograms. Then, as Figure 6 shown, the learning support device 30 displays the first chromatogram image 211, the second chromatogram image 212 of the acquired second chromatogram, and the second peak information image 253.

[0095] Then, as Figure 7 shown, the learning support device 30 accepts the input of the first peak information. Therefore, the user can visually recognize the past-generated second chromatogram that is the same as or similar to the first chromatogram and the second peak information image 253 showing the peaks of the second chromatogram, and input the first peak information of the first chromatogram. Therefore, the learning support device 30 can encourage the user to input training data (the first peak information) with suppressed annotation deviation. As a result, the learning support device 30 can suppress the annotation deviation and improve the quality of the estimation model.

[0096] In addition, the learning support device 30 calculates the first similarity based on the first feature amount of the first chromatogram and the second feature amount of the stored chromatogram that is the same category as the first feature amount. Therefore, the learning support device 30 can calculate the first similarity through relatively simple operations.

[0097] In addition, the learning support device 30 calculates the first similarity based on the analysis result shown by the first chromatogram and the analysis result shown by the stored chromatogram. Therefore, the learning support device 30 can calculate the first similarity through relatively simple operations.

[0098] In addition, in the examples of Figure 6 and Figure 7 , the learning support device 30 associates and displays the first similarity of the second chromatogram with the second chromatogram image 212 showing the second chromatogram. In the examples of Figure 6 and Figure 7 , a similarity image 271 showing the first similarity is displayed. In the examples of Figure 6 and Figure 7In the example, for instance, an image such as "98%" is displayed as the similarity image 271 in association with the second chromatogram image 212A. Thus, the user can visually recognize the first similarity of the second chromatogram.

[0099] Furthermore, the learning support device 30 displays the second feature amount image 275 in the past result area 132A in association with the second chromatogram image 212. Here, the second feature amount image 275 shows the feature amount of the peak determined based on the second peak information shown in the second peak information image 253. The second feature amount image 275 includes an image 272 showing the gradient between two points and an image 273 showing the area value. In addition, the gradient between two points and the area value are the feature amounts of each of one or more peaks included in the second chromatogram. In Figure 6 and Figure 7 the example, there are three peaks included in the chromatogram of the second chromatogram image 212A. The gradients between two points A1, A2, A3 of each of the three peaks are displayed in association with the second chromatogram image 212A, and the area values B1, B2, B3 of each of the three peaks are displayed.

[0100] In addition, in the present embodiment, the second chromatogram includes a highly similar chromatogram and a low-similar chromatogram with a first similarity lower than that of the highly similar chromatogram. For example, in Figure 7 the example, as an example of the highly similar chromatogram, the second chromatogram image 212A of the chromatogram with a first similarity of 98% is displayed. In addition, as an example of the low-similar chromatogram, the second chromatogram image 212D of the chromatogram with a first similarity of 80% is displayed. Furthermore, in the past result area 132A, the second chromatogram image 212A showing the highly similar chromatogram is displayed prior to the second chromatogram image 212B showing the low-similar chromatogram. Thus, the learning support device 30 displays the second chromatogram image 212 and the second peak information image 253 with priorities corresponding to the first similarity. In Figure 6 and Figure 7 the example, the highly similar chromatogram is displayed above the low-similar chromatogram.

[0101] Accordingly, the user can visually recognize more preferentially the second chromatogram with a higher similarity to the first chromatogram than the second chromatogram with a lower similarity to the first chromatogram. Therefore, the learning support device 30 can make the second peak information of the second chromatogram with a higher similarity to the first chromatogram easily visually recognizable.

[0102] In addition, in Figure 7 the example, the second peak information image 253 is a line-shaped image (hereinafter also referred to as "line image"). In addition, the area surrounded by the second chromatogram image 212 and the second peak information image 253 is the area showing the peaks of the second chromatogram.

[0103] In addition, while the learning support device 30 is displaying the first chromatogram image 211, the second chromatogram image 212, and the second peak information image 253, it accepts input (specification) of the first point 201 and the second point 202 by the user. In the present embodiment, the cursor 217 of the input device 61 (mouse) is displayed. The user can specify the first point 201 by aligning the cursor 217 with the desired position and clicking the mouse. In addition, the user can specify the second point 202 by aligning the cursor 217 with another desired position and clicking the mouse. If the first point 201 and the second point 202 are specified, the learning support device 30 displays a line image connecting the first point 201 and the second point 202 as the first peak information image 220. Then, the learning support device 30 identifies the region surrounded by the line connecting the first point 201 and the second point 202 and the lines included in the first chromatogram image 211 as the peak region (training data) determined by the first peak information of the first peak information image 220. With such a configuration, the first peak information can be input by specifying the first point 201 and the second point 202 of the first chromatogram image 211. Therefore, the user can intuitively input the first peak information, thus improving the convenience for the user. Also, the Figure 7 line image connecting the first point 201 and the second point 202 is referred to as the "baseline of the peak".

[0104] Figure 8 is a diagram showing the screen after the user inputs the first peak information. As Figure 8 shown, when the first peak information is input, the learning support device 30 calculates the first feature amount and displays the first feature amount image 231 of the first feature amount. Here, the first feature amount image 231 shows the feature amount of the peak determined according to the first peak information shown in the first peak information image 220. In Figure 8 the example, the first feature amount image 231 is an image showing the gradient between two points of the peak and the area value of the peak. In Figure 8 the example, "X1" is displayed as the value of the gradient between two points, and "Y1" is displayed as the area value.

[0105] As described above, as Figures 6 - 8 shown, the learning support device 30 displays the second feature amount image 275 in the past result area 132A. In this way, the learning support device 30 displays the first feature amount image 231 and the second feature amount image 275. Therefore, after the user inputs the first peak information, the user can compare the peak shown by the first peak information with the past peaks from the perspective of the feature amounts (in the present embodiment, the gradient between two points and the area value).

[0106] In addition, the first feature quantity image 231 is displayed in a corresponding number according to the number of peaks determined based on the first peak information input by the user. For example, when the number of peaks determined based on the first peak information input by the user is "3", the first feature quantity images 231 of the respective 3 peaks are displayed.

[0107] [Flowchart]

[0108] Figure 9 This is a flowchart showing the processing of the learning support device 30. In step S2, the learning support device 30 acquires the first chromatogram. Next, in step S4, the learning support device 30 acquires the second chromatogram and the second peak information. Step S4 includes: step S42, step S44 executed after this step S42, step S46 executed after this step S44, step S48 executed after this step S46, and step S50 executed after this step S48.

[0109] In step S42, the learning support device 30 extracts the first feature quantity of the first chromatogram and the second feature quantity of the stored chromatogram. As described above, the first feature quantity and the second feature quantity are the number of peaks, the gradient between two points, and the area value.

[0110] In step S44, the learning support device 30 extracts the first analysis result shown by the first chromatogram and the second analysis result shown by the stored chromatogram. The first analysis result and the second analysis result are qualitative analysis results and quantitative analysis results.

[0111] In step S46, the learning support device 30 calculates the first similarity between the first chromatogram and the stored chromatogram. The learning support device 30 calculates the first similarity based on the first feature quantity, the second feature quantity, the first analysis result, and the second analysis result.

[0112] In step S48, the learning support device 30 refers to the chromatogram DB123 and acquires the stored chromatogram with the first similarity being equal to or greater than the first threshold value as the second chromatogram.

[0113] In step S50, the learning support device 30 refers to the chromatogram DB123 and acquires the stored peak information corresponding to the second chromatogram as the second peak information.

[0114] If the processing of step S4 is completed, then in step S6, the learning support device 30 displays the first chromatogram image 211 of the first chromatogram acquired in step S2.

[0115] Next, in step S8, the learning support device 30 displays the second chromatogram image 212 and the second peak information image 253. Step S8 includes step S82 and step S84 executed after this step S82.

[0116] In step S82, the learning support device 30 displays the second chromatogram image 212 and the second peak information image 253 with priorities corresponding to the first similarity. In step S84, the similarity image 271 and the second feature quantity image 275 are displayed.

[0117] Thus, by performing the processes of steps S2 to S8, the following Figure 6 displayed images are shown.

[0118] Next, in step S10, the learning support device 30 determines whether the user has input the first peak information. Step S10 includes the process of step S102. In step S102, the learning support device 30 determines whether the user has input the first point 201 and the second point 202. The learning support device 30 repeatedly performs the process of step S102 until the first point 201 and the second point 202 are input. If it is determined as "Yes" in step S102, the process proceeds to step S12.

[0119] In step S12, the learning support device 30 displays the first peak information image of the first peak information determined to be input in step S10. Step S12 includes step S122. In step S122, the learning support device 30 displays the first peak information image 220 and the first feature quantity image 231. In addition, in step S122, the learning support device 30 displays the first peak information image 220, thereby displaying the following Figure 7 displayed images. Further, in step S122, the learning support device 30 displays the first feature quantity image 231, thereby displaying the following Figure 8 displayed images.

[0120] Next, in step S13, the learning support device 30 determines whether the user has performed an end operation. The end operation is an operation performed by the user on the input device 61. The end operation is, for example, an operation by the user on an end button (not shown) displayed on the Figures 6 - 8 displayed screen.

[0121] For the first chromatogram image 211, the user can input one or more pieces of the first peak information. When the input of the first peak information is completed, the user performs an end operation. If it is determined as "No" in step S13, the process returns to step S102. On the other hand, if it is determined as "Yes" in step S13, the process proceeds to step S14.

[0122] In step S14, the learning support device 30 learns the estimation model 121 based on the first chromatogram obtained in step S2 and the first peak information determined to be input in step S10. In addition, the estimation model 121 may learn the chromatogram and the peak information as a group in multiple groups.

[0123] AsFigure 9 As shown, after steps S6 and S8, the learning support device 30 receives the input of the first peak information by the user. As described above, step S6 is the process of displaying the first chromatogram image 211 on the display device 62. In addition, step S8 is the process of displaying the second chromatogram image 212 and the second peak information image 253 on the display device 62.

[0124] If it is Figure 9 such a configuration, the user can visually recognize the second chromatogram similar to the first chromatogram and previously generated and the second peak information showing the peaks of the second chromatogram, and at the same time input the first peak information of the first chromatogram. Therefore, the convenience of the user can be improved for the input of the first peak information.

[0125] <Second Embodiment>

[0126] Figure 10 is a flowchart of the process of the learning support device 30 of the second embodiment. In Figure 10 it, after the process of step S2 ends, the learning support device 30 executes the process of step S6. Then, the learning support device 30 executes the process of step S10. Here, the user inputs the first peak information in a state where the first chromatogram image 211 is displayed but the second chromatogram image 212 and the second peak information image 253 are not displayed. If the process of step S10 is executed, the processes of step S12 and step S13 are executed. It is determined as "Yes" in step S13. The process proceeds to step S4A. The difference between step S4A and step S4 is that step S48 is replaced by step S52 and step S54.

[0127] In step S52, the learning support device 30 calculates S similarities (hereinafter also referred to as "second similarities") between the first peak information input in step S10 and each of the S stored peak information R stored in the chromatogram DB123 (refer to Figure 5 ). The second similarity shows the degree of similarity between one stored peak information and the first peak information. The more similar one stored peak information and the first peak information are, the larger the second similarity becomes. For example, the learning support device 30 calculates parameters (such as a correlation coefficient) related to each of the S stored peak information and the first peak information as the second similarity.

[0128] Next, in step S54, the learning support device 30 acquires a stored chromatogram whose first similarity is above the first threshold and whose peak is determined by the stored peak information whose second similarity is above the second threshold as the second chromatogram. Here, the second threshold is a preset threshold. That is, in step S54, a second chromatogram similar in both the first peak information and the chromatogram is acquired. Next, in step S50, the second peak information corresponding to the second chromatogram acquired in step S54 is acquired.

[0129] In Figure 10 the example of, the learning support device 30 performs the process of step S4A after steps S6 and S10. Step S6 is a process of displaying the first chromatogram image 211 on the display device 62. Step S10 is a process of accepting the first peak information input by the user. Step S4A is a process of acquiring the second chromatogram and the second peak information from the chromatogram DB123. In addition, step S4A includes steps S46, S52, and S54.

[0130] Step S46 is a process of calculating the first similarity, and step S52 is a process of calculating the second similarity. Step S56 is a process of acquiring a stored chromatogram whose first similarity is equal to or greater than the first threshold and whose peak is determined by the stored peak information with the second similarity equal to or greater than the second threshold as the second chromatogram.

[0131] According to the second embodiment, the learning support device 30 can display a chromatogram whose first similarity is equal to or greater than the first threshold and a stored chromatogram whose peak is determined by the stored peak information with the second similarity equal to or greater than the second threshold as the second chromatogram. Therefore, the user can confirm such a second chromatogram.

[0132] <Third Embodiment>

[0133] Figure 11 is a flowchart of the process of the learning support device 30 according to the third embodiment. Figure 11 In Figure 9 after the process of step S6, the process of step S150 is performed. The process of step S150 is a process of displaying a detected peak image. The detected peak image is for determining the peak information (hereinafter also referred to as "detected peak information") of the peak of the first chromatogram (the first chromatogram acquired in step S2) detected using the current estimation model 121. That is, the temporary peak information of the first chromatogram acquired in step S2 is displayed as the detected peak image. In this way, by performing the processes of steps S6 and S150, the first chromatogram image and the detected peak image (the temporary peak image of the first chromatogram) are displayed. Then, the processes of steps S8 and S10 are performed.

[0134] According to the third embodiment, the user can input the first peak information while referring to the detected peak information image. Furthermore, the user can input the first peak information while referring to the detected peak information image and the second peak information image. Therefore, the convenience of the user can be improved.

[0135] In addition, in step S102 (step S10), when the user determines that the first peak information image is the detected peak image as it is, the user performs a specified operation (for example, an operation of pressing an OB button not shown), and thus it is determined as "Yes" in this step S102 and the next step S122 is entered. Further, in step S102 (step S10), when the user wants to correct the detected peak image, the first peak information is newly input (specifying the first point 201 and the second point 202), and thus it is determined as "Yes" in this step S102 and the next step S122 is entered.

[0136] [Variant Example]

[0137] (1) In the Figure 11 processing, the learning support device 30 may also execute the processing in the order of step S2, step S6, step S150, step S4A (refer to Figure 10 ), step S8, step S10, and step S12. In step S52 of step S4A, S second similarities between the detected peak information and each of the S stored chromatograms are calculated. Therefore, in step S54, the learning support device 30 acquires a chromatogram with the first similarity being equal to or greater than the first threshold value and a stored chromatogram in which the peak is determined by the stored peak information with the second similarity being equal to or greater than the second threshold value as the second chromatogram. That is, a second chromatogram image that is the same as or similar to the first chromatogram with the detected peak information added is displayed. By displaying such a second chromatogram image, the learning support device 30 can improve the convenience for the user.

[0138] (2) In the above-described embodiment, the configuration in which the first peak information is input by the user specifying the first point 201 and the second point 202 has been described. However, the input method of the first peak information may also be other methods. For example, the user may input coordinates for determining the desired peak to the displayed first chromatogram image 211.

[0139] (3) In the Figure 7 example, the configuration in which the user inputs the first point 201 and the second point 202 to input "the baseline of the peak (the first peak information image 220)" has been described. However, there are cases where the learning support device 30 cannot determine the peak based on the baseline.

[0140] Figure 12 is a diagram showing an example of the first chromatogram image 211 when the learning support device 30 cannot determine the peak based on the baseline. In the Figure 12 first chromatogram image 211, peaks Pa, Pb, and Pc are shown. In the Figure 12 example, peaks Pb and Pc are connected, the start point Pb1 of peak Pb is displayed, but the end point of peak Pb is not displayed (refer to Figure 12region S). In addition, although not particularly illustrated, there is also a case where a peak that does not display either the start point or the end point is shown. For such a peak that does not display at least one of the start point and the end point, it can be determined by a vertical line.

[0141] The vertical line is a line perpendicular or substantially perpendicular to the horizontal axis (time axis) of the first chromatogram image 211. In Figure 12 the example, by the user specifying the first point 201 and the second point 202, a vertical line is displayed as the first peak information image 220. Thus, the first peak information image 220 can include Figure 7 the baseline shown in Figure 12 and at least one of the vertical lines shown in

[0142] In addition, in Figure 12 the example, although the details of the past result region 132A are not described, for example, a second peak information image 253 such as a similarity image 271, a second feature quantity image 275, and a vertical line is shown.

[0143] [Solution]

[0144] Those skilled in the art can understand that the above-described multiple exemplary embodiments are specific examples of the following solutions.

[0145] (Item 1) A learning support method according to one solution is a method of causing a computer to execute a process of supporting a learning operation of a estimation model for detecting peaks of a signal waveform acquired by an analysis device. The learning support method includes acquiring a first signal waveform generated by the analysis device. The learning support method includes displaying the first signal waveform on a display device. The learning support method includes acquiring, from a storage device storing a plurality of labeled signals, a second signal waveform having a high similarity to the first signal waveform and second peak information for determining peaks of the second signal waveform. The learning support method includes displaying the second signal waveform and a second peak information image showing the second peak information on the display device. The learning support method includes accepting an input of first peak information for determining peaks of the first signal waveform by the user. The learning support method includes learning the estimation model based on the first signal waveform and the first peak information.

[0146] According to this configuration, the user can input first peak information that is consistent with past second peak information, and can update the estimation model based on the first peak information. Therefore, promoting user input suppresses the deviation of labeling in training data. As a result, it is possible to suppress the deviation of labeling and improve the quality of the estimation model.

[0147] (Item 2) In the learning support method described in Item 1, obtaining the second signal waveform and the second peak information from the storage device includes: calculating the first similarity between the first signal waveform and each of the multiple stored signal waveforms included in the multiple labeled signals, and obtaining the stored signal waveform with the first similarity greater than or equal to the first threshold as the second signal waveform. Further, after displaying the first signal waveform on the display device and displaying the second signal waveform and the second peak information image on the display device, an input of the first peak information by the user is received.

[0148] With such a configuration, the user can visually recognize the second signal waveform similar to the first signal waveform and previously generated, and the second peak information showing the peak of the second signal waveform, while inputting the first peak information of the first chromatogram. Therefore, the convenience of the user for inputting the first peak information can be improved.

[0149] (Item 3) In the learning support method described in Item 2, after displaying the first signal waveform on the display device and receiving the input of the first peak information by the user, the second signal waveform and the second peak information are obtained from the storage device. Obtaining the second signal waveform and the second peak information from the storage device includes: calculating the first similarity between the first signal waveform and each of the multiple stored signal waveforms included in the multiple labeled signals, calculating the second similarity between the first peak information and each of the multiple stored peak information included in the multiple labeled signals, and obtaining the stored signal waveform with the first similarity greater than or equal to the first threshold and having a peak determined by the stored peak information with the second similarity greater than or equal to the second threshold as the second signal waveform.

[0150] With such a configuration, the stored signal waveform with the first similarity greater than or equal to the first threshold and having a peak determined by the stored peak information with the second similarity greater than or equal to the second threshold can be displayed as the second signal waveform. Therefore, the user can confirm such a second signal waveform.

[0151] (Item 4) In the learning support method described in Item 2 or Item 3, the second signal waveform includes a highly similar signal waveform and a low-similarity signal waveform with a first similarity lower than that of the highly similar signal waveform. Displaying the second signal waveform and the second peak information image on the display device includes: displaying the waveform image showing the highly similar signal waveform prior to the waveform image showing the low-similarity signal waveform.

[0152] With this configuration, the user can visually recognize the second signal waveform with a higher similarity to the first signal waveform more preferentially than the second signal waveform with a lower similarity to the first signal waveform.

[0153] (Item 5) In the learning support method described in any one of Items 2 to 4, displaying the second signal waveform and the second peak information image on the display device includes: displaying the first similarity of the second signal waveform in association with the second signal waveform image showing the second signal waveform.

[0154] With such a configuration, the user can visually recognize the first similarity of the second signal waveform.

[0155] (Item 6) In the learning support method described in any one of Items 2 to 5, calculating the first similarity includes: calculating the first similarity based on the first feature amount of the first signal waveform and the second feature amount of the stored signal waveform that is of the same category as the first feature amount.

[0156] With such a configuration, the first similarity can be calculated by relatively simple operations.

[0157] (Item 7) In the learning support method described in any one of Items 2 to 6, calculating the first similarity includes calculating the first similarity based on the analysis result shown by the first signal waveform and the analysis result shown by the stored signal waveform.

[0158] With such a configuration, the first similarity can be calculated by relatively simple operations.

[0159] (Item 8) In the learning support method described in any one of Items 1 to 7, the learning support method further includes displaying the first peak information image showing the first peak information on the display device.

[0160] With such a configuration, the first peak information input by the user can be recognized on the display device.

[0161] (Item 9) In the learning support method described in Item 8, displaying the first peak information image on the display device includes displaying the first feature amount image, and the first feature amount image shows the feature amount of the peak determined by the first peak information shown by the first peak information image. In addition, displaying the second signal waveform and the second peak information image on the display device includes displaying the second feature amount image, and the second feature amount image shows the feature amount of the peak determined by the second peak information shown by the second peak information image.

[0162] With such a configuration, after the user inputs the first peak information, the user can recognize the feature amount of the peak determined by the first peak information and the feature amount of the peak determined by the second peak information. Therefore, the user can confirm whether the input first peak information is appropriate.

[0163] (Item 10) In the learning support method according to any one of Items 1 to 9, the learning support method further includes displaying a detected peak information image that shows detected peak information for determining the peak of the first signal waveform detected using the estimation model. After the detected peak information image and the first signal waveform are displayed, an input of the first peak information by the user is received.

[0164] With such a configuration, the user can input the first peak information while referring to the detected peak information.

[0165] (Item 11) In the learning support method according to any one of Items 1 to 10, the learning support method further includes receiving an input of a first point and a second point by the user while the first signal waveform is being displayed on the display device. The second peak information image is a line image. The area surrounded by the second signal waveform and the line image is an area showing the peak of the second signal waveform. The area of the peak determined by the first peak information is an area surrounded by the line connecting the first point and the second point and the line included in the first signal waveform.

[0166] With such a configuration, the first peak information can be input by specifying the first point and the second point of the first signal waveform. Therefore, the user can input the first peak information more easily, which can improve the convenience for the user.

[0167] (Item 12) A learning support program according to one aspect is a program that causes a computer to execute a process of supporting the learning operation of an estimation model for detecting the peak of a signal waveform acquired by an analysis device. The learning support program causes the computer to execute acquiring the first signal waveform generated by the analysis device. The learning support program causes the computer to execute displaying the first signal waveform on the display device. The learning support program causes the computer to execute acquiring, from a storage device storing a plurality of labeled signals, a second signal waveform having a high similarity to the first signal waveform and second peak information for determining the peak of the second signal waveform. The learning support program causes the computer to execute displaying the second signal waveform and a second peak information image showing the second peak information on the display device. The learning support program causes the computer to execute receiving an input of first peak information for determining the peak of the first signal waveform by the user. The learning support program causes the computer to execute learning the estimation model based on the first signal waveform and the first peak information.

[0168] With this configuration, the user can input first peak information that is consistent with the past second peak information, and can update the estimation model based on the first peak information. Therefore, promoting user input suppresses the deviation of annotation. As a result, it is possible to suppress the deviation of annotation and improve the quality of the estimation model.

[0169] In addition, for the above-described embodiments and variations, within the scope where no impropriety or contradiction occurs, appropriate combinations of the configurations described in the embodiments, including combinations not mentioned in the specification, are predetermined from the very beginning of the application.

[0170] The embodiments of the present invention have been described above, but it should be considered that the disclosed embodiments are illustrative in all respects and not restrictive. The scope of the present invention is shown by the claims, and it is also intended to include all modifications within the meaning and scope equivalent to the claims.

Claims

1. A learning support method is a learning support method for causing a computer to execute processing for supporting a learning operation of a estimation model for detecting peaks of a signal waveform acquired by an analysis device, characterized in that Comprising: Obtaining a first signal waveform generated by an analysis device; Displaying the first signal waveform on a display device; Obtaining, from a storage device storing a plurality of labeled signals, a second signal waveform having a high similarity to the first signal waveform and second peak information for determining peaks of the second signal waveform; Displaying the second signal waveform and a second peak information image showing the second peak information on the display device; Receiving an input of first peak information for determining peaks of the first signal waveform by a user; Learning the estimation model based on the first signal waveform and the first peak information.

2. The learning support method according to claim 1, wherein Obtaining the second signal waveform and the second peak information from the storage device includes: Calculating a first similarity between the first signal waveform and each of a plurality of stored signal waveforms included in the plurality of labeled signals, Obtaining a stored signal waveform with the first similarity being equal to or greater than a first threshold as the second signal waveform, After displaying the first signal waveform on the display device and displaying the second signal waveform and the second peak information image on the display device, Receiving an input of the first peak information by a user.

3. The learning support method according to claim 1, characterized in that, After displaying the first signal waveform on the display device and receiving an input of the first peak information by a user, Obtaining the second signal waveform and the second peak information from the storage device, Obtaining the second signal waveform and the second peak information from the storage device includes: Calculating a first similarity between the first signal waveform and each of a plurality of stored signal waveforms included in the plurality of labeled signals, Calculating a second similarity between the first peak information and each of a plurality of stored peak information included in the plurality of labeled signals, Obtaining a stored signal waveform with the first similarity being equal to or greater than a first threshold and having peaks determined by stored peak information with the second similarity being equal to or greater than a second threshold as the second signal waveform.

4. The learning support method according to claim 2, wherein The second signal waveform includes a high-similarity signal waveform and a low-similarity signal waveform with the first similarity lower than that of the high-similarity signal waveform, Displaying the second signal waveform and the second peak information image on the display device includes: preferentially displaying a waveform image showing the high-similarity signal waveform over a waveform image showing the low-similarity signal waveform.

5. The learning support method according to claim 2, wherein Displaying the second signal waveform and the second peak information image on the display device includes: displaying the first similarity of the second signal waveform in association with a second signal waveform image showing the second signal waveform.

6. The learning support method according to claim 2, wherein Calculating the first similarity includes: calculating the first similarity based on a first feature quantity of the first signal waveform and a second feature quantity of the stored signal waveform that is of the same category as the first feature quantity.

7. The learning support method according to claim 2, characterized in that Calculating the first similarity includes: calculating the first similarity based on an analysis result shown by the first signal waveform and an analysis result shown by the stored signal waveform.

8. The learning support method according to claim 1, characterized in that, The learning support method further includes displaying a first peak information image showing the first peak information on the display device.

9. The learning support method according to claim 8, characterized in that, Displaying the first peak information image on the display device includes displaying a first feature quantity image, and the first feature quantity image shows a feature quantity of a peak determined by the first peak information shown in the first peak information image. Displaying the second signal waveform and the second peak information image on the display device includes displaying a second feature quantity image, and the second feature quantity image shows a feature quantity of a peak determined by the second peak information shown in the second peak information image.

10. The learning support method according to claim 1, wherein The learning support method further includes displaying a detected peak information image, and the detected peak information image shows detected peak information for determining a peak of the first signal waveform detected using the estimation model. After the detected peak information image and the first signal waveform are displayed, an input of the first peak information by the user is accepted.

11. The learning support method according to claim 1, wherein, The learning support method further includes accepting an input of a first point and a second point by the user while the first signal waveform is being displayed on the display device. The second peak information image is a line image. The region surrounded by the second signal waveform and the line image is a region showing the peak of the second signal waveform. The region of the peak determined by the first peak information is a region surrounded by a line connecting the first point and the second point and a line included in the first signal waveform.

Images