Waveform analyzing method and waveform analyzing device

Abstract

To correctly detect a peak from a chromatogram.SOLUTION: A method for analyzing a measurement waveform of a sample, a waveform analyzing device includes: a trained model storage section (44) that stores a trained model, which is constructed by machine learning using reference waveform data having a different baseline shape and a known position of a peak portion including an overlapping peak with tailing processing, complete separation, or vertical separation related to the overlapping peak, as teacher data, and outputs, for an input of measurement data, an index which represents a single peak portion, an overlapping peak portion, or a non-peak portion and to which the tailing processing, the complete separation, or the vertical separation is related as a peak separation technique ; and an index output section (55 to 57) inputs analysis target data into the trained model to output the index which represents the single peak portion, the overlapping peak portion, or the non-peak portion and to which the tailing processing, the complete separation, or the vertical separation is related as a separation technique of the overlapping peak portion.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a method and apparatus for analyzing waveforms obtained by measuring a sample using an analytical device. [Background technology]

[0002] Liquid chromatographs and gas chromatographs are used to identify and quantify the components in a sample. In a chromatograph, the components in the sample are separated using a column, and the components that flow out of the column are detected in order. A chromatogram is then created, with the horizontal axis representing time and the vertical axis representing detection intensity, to detect peaks, and the concentration or amount of the compound corresponding to the peak can be determined from its area and height.

[0003] Various methods have been put into practical use to detect peaks in chromatograms. In recent years, new peak detection methods that utilize machine learning have been proposed and put into practical use (e.g., Patent Document 1, Non-Patent Documents 1 and 2).

[0004] Patent Document 1 describes a waveform analysis technology that constructs a trained model by performing machine learning using data of multiple reference waveforms whose peak positions are known as training data, and then uses the trained model to estimate peaks contained in data of a waveform to be analyzed. As an example, the technology describes a trained model constructed by performing machine learning on a training model using semantic segmentation, which is used in the field of image analysis, using data of multiple extracted ion chromatograms (EICs) whose peak positions are known and obtained by selected ion monitoring (SIM) or multiple reaction monitoring (MRM) measurement as training data, and inputting a predetermined number of measurement data extracted from the extracted ion chromatogram of the analysis target obtained by SIM or MRM measurement into the trained model, which outputs an index (label) indicating whether the data belongs to a peak portion or a non-peak portion. It also describes that when there are multiple overlapping peaks (superimposed peaks) on a chromatogram, the system outputs an indicator (label) indicating which of the following separation methods is more appropriate: tailing (the region from the start point to the end point of the superimposed peaks is considered to be one peak, and then another peak is superimposed on the tailing part of that peak), complete separation (peaks are separated by a line connecting the start point, minimum point, and end point of the superimposed peak in order), or vertical separation (two peaks are separated by a perpendicular line passing through the minimum point of the superimposed peak). [Prior art documents] [Patent documents]

[0005] [Patent Document 1] International Publication No. 2021 / 064924 [Non-patent literature]

[0006] [Non-Patent Document 1] "Peakintelligence for GCMS LabSolutions Insight waveform processing software," [online], [searched March 14, 2024], Shimadzu Corporation, Internet <URL:https: / / www.an.shimadzu.co.jp / products / gas-chromatograph-mass-spectrometry / gc-ms-software / peakintelligence-for-gcms / index.html> [Non-patent document 2] "Peakintelligence for LCMS: Optional peak processing software for LabSolutions LCMS and LabSolutions Insight," [online], [Retrieved March 14, 2024], Shimadzu Corporation, Internet <URL: https: / / www.an.shimadzu.co.jp / products / liquid-chromatograph-mass-spectrometry / lc-ms-software / peakintelligence / index.html> [Non-patent document 3] "Reducing pesticide data analysis time with Peakintelligence for GCMS," [online], [Retrieved March 14, 2024], Shimadzu Corporation, Internet <URL: https: / / www.an.shimadzu.co.jp / sites / an.shimadzu.co.jp / files / pim / pim_document_file / an_jp / applications / application_note / 21749 / an_01-00585-jp.pdf> [Non-patent document 4] "High-speed pharmaceutical impurity analysis in accordance with the European Pharmacopoeia using Nexera-i MT," [online], [searched March 14, 2024], Shimadzu Corporation, Internet <URL: https: / / www.an.shimadzu.co.jp / sites / an.shimadzu.co.jp / files / pim / pim_document_file / an_jp / applications / application_note / 17727 / an_l518.pdf> Summary of the Invention [Problem to be solved by the invention]

[0007] To improve the accuracy of peak determination using a trained model, a large amount of training data is required. To prepare such a large amount of training data, chromatograms acquired under standard measurement conditions or chromatograms created by simulations assuming standard measurement conditions are often used as training data.

[0008] However, the shape of a chromatogram varies depending on the configuration of the analytical instrument and the measurement conditions. For example, when a mass spectrometer is used as a chromatograph detector, narrow, sharp peaks are likely to be obtained, whereas when a PDA detector or UV detector is used, peaks of various widths appear (e.g., Non-Patent Documents 3 and 4). Furthermore, when gradient analysis is performed in a liquid chromatograph or temperature-programmed analysis is performed in a gas chromatograph, baseline drift is likely to occur. When multiple overlapping peaks (superimposed peaks) appear on a chromatogram, known methods for separating them include tailing, complete separation, and vertical separation. However, the method suitable for separating superimposed peaks varies depending on the shape of the baseline. Because the shape of the baseline varies depending on the configuration of the analytical instrument and the measurement conditions, it was sometimes difficult for conventional trained models to correctly detect peaks.

[0009] Although the example described here is one in which peaks are detected from a chromatogram obtained using a chromatograph, the same problems as those described above also exist when peaks are detected from other types of waveform data.

[0010] The problem to be solved by the present invention is to provide a technique that can correctly detect peaks from a chromatogram regardless of the configuration of the analyzer or the measurement conditions. [Means for solving the problem]

[0011] One aspect of the present invention, which has been made to solve the above problems, is a waveform analysis method for analyzing a waveform constituted by analysis target data, which is data acquired by measuring a sample using an analytical device, comprising: a trained model construction step for constructing a trained model by machine learning using a plurality of reference waveform data as training data, the reference waveform data being data constituting a plurality of reference waveforms each having a different baseline shape, with the positions of peak portions including superimposed peaks known and associated with any of tailing processing, complete separation, and vertical separation as a method for separating the superimposed peaks, and for which data elements constituting the measurement data are input, outputting an index that represents a single peak portion, a superimposed peak portion, or a non-peak portion and associates whether tailing processing, complete separation, or vertical separation should be used as a method for separating the superimposed peaks; and an index output step of inputting the analysis target data into the trained model, and outputting, from the trained model, an index representing a single peak portion, a superposed peak portion, or a non-peak portion for each of a plurality of analysis target data elements constituting the analysis target data, and associating the superposed peak with which of tailing processing, complete separation, and vertical separation should be used as a method for separating the superposed peaks; Equipped with.

[0012] Another aspect of the present invention, which has been made to solve the above-mentioned problems, is a waveform analysis device that analyzes a waveform configured from analysis target data, which is data acquired by measuring a sample using an analytical device, comprising: a trained model storage unit in which trained models are stored, the trained models being constructed by machine learning using as training data a plurality of reference waveform data, the reference waveform data being data constituting a plurality of reference waveforms each having a different baseline shape, with the positions of peak portions including superimposed peaks known, and associated with any of tailing processing, complete separation, and vertical separation as a method for separating the superimposed peaks, and which, when measurement data is input, outputs an index representing a single peak portion, a superimposed peak portion, or a non-peak portion for each data element constituting the measurement data, and associating it with whether tailing processing, complete separation, or vertical separation should be used as a method for separating the superimposed peaks; and an index output unit that inputs the analysis target data into the trained model, and outputs, for each of a plurality of analysis target data elements that make up the analysis target data, an index that represents a single peak portion, a superposed peak portion, or a non-peak portion from the trained model, and that associates an index with which of tailing processing, complete separation, and vertical separation should be used as a method for separating the superposed peaks; Equipped with. [Effects of the Invention]

[0013] In the present invention, machine learning is performed using multiple reference waveform data as training data when constructing a trained model through machine learning. The multiple reference waveform data is data constituting each of multiple reference waveforms, each of which has known positions of peak portions including overlapping peaks, which are multiple peaks with different baseline shapes and overlapping each other, and is associated with one of tailing, complete separation, and vertical separation as a method for separating the overlapping peaks. Therefore, a trained model is constructed that estimates peak portions based on the baseline shape and also estimates an appropriate peak separation method for the overlapping peaks based on the baseline shape. In the present invention, when data to be analyzed is input into the trained model, the peak portions are correctly estimated based on the baseline shape, which appears uniquely depending on the detector and measurement conditions, and an index indicating a peak separation method appropriate for separating the overlapping peaks is output. Therefore, peaks can be correctly detected from a chromatogram regardless of the device configuration or measurement conditions. [Brief explanation of the drawings]

[0014] [Figure 1] 1 is a diagram showing the configuration of a main part of a liquid chromatograph system including an embodiment of a waveform analyzer according to the present invention. [Figure 2] Example chromatogram showing a drifting baseline and a column cleaning period. [Figure 3] An example of separating overlapping peaks using tailing, complete separation, and vertical separation. [Figure 4] An example of overlapping peaks that cannot be properly separated using conventional techniques. [Figure 5] Examples of a single peak on the tail, vertically separated peaks on the tail, peaks on the tail, and peaks with associated peak labels (indicators) on the leading edge. [Figure 6] 1 is a flowchart showing the procedure for creating a trained model in one embodiment of the waveform analysis method according to the present invention. [Figure 7] FIG. 10 is a diagram illustrating the state in which the first window, second window, and third window are applied to training data. [Figure 8] 1 is a flowchart showing a procedure for estimating peak portions contained in unanalyzed chromatogram data in one embodiment of the waveform analysis method according to the present invention. [Figure 9] 1 shows an example of chromatogram analysis using the mass spectrometer and mass analysis method of the present embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0015] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A waveform analysis method and a waveform analysis device according to embodiments of the present invention will be described below with reference to the accompanying drawings.

[0016] 1 shows the main configuration of a liquid chromatograph system 1 including a waveform analyzer according to this embodiment. The liquid chromatograph system 1 includes a liquid chromatograph unit 10 and a control and processing unit 40. A part of the control and processing unit 40 corresponds to the waveform analyzer according to the present invention.

[0017] The liquid chromatograph unit 10 includes a mobile phase container 11 containing a mobile phase, a liquid delivery pump 12 that delivers the mobile phase contained in the mobile phase container 11, an injector 13 that injects a liquid sample, a column 14 that separates components contained in the liquid sample, and a detector 15 that detects the components that flow out of the column 14 in sequence. The liquid chromatograph unit 10 also includes an autosampler 16 that is set with sample containers containing multiple liquid samples and introduces the multiple liquid samples into the injector 13 in the order specified in the measurement conditions. The detector 15 can be a type appropriate for the components to be detected, such as a mass spectrometer, ultraviolet absorbance detector (UV detector), photodiode array detector (PDA detector), refractive index detector (RID), or electrical conductivity detector.

[0018] The control and processing unit 40 includes a memory unit 41. The memory unit 41 includes a reference waveform data memory unit 42, a measurement data memory unit 43, and a trained model memory unit 44. The reference waveform data memory unit 42 stores measurement data (reference waveform data) that has been acquired by measurement using a detector 15 such as a mass spectrometer, an ultraviolet absorbance detector (UV detector), a photodiode array detector (PDA detector), a refractive index detector (RID), or an electrical conductivity detector, and that has undergone peak detection and the like, together with information such as the measurement conditions (including the sampling rate) and the type of detector.

[0019] In liquid chromatography, reference waveform data is typically two-dimensional data with time or sampling interval on the horizontal axis and intensity on the vertical axis. However, the reference waveform data may also be a one-dimensional data sequence in which the output signal from the detector is arranged in time series, with the known sampling interval information removed. Such reference waveform data may also contain identified peaks. Reference waveform data has different baseline shapes, known peak positions including overlapping peaks, and is associated with labels (indicators) for tailing, complete separation, and vertical separation as a method for separating the overlapping peaks. Specifically, reference waveform data includes, for example, chromatograms obtained by gradient analysis in which the baseline increases (or decreases) throughout the entire measurement time, and chromatograms obtained by continuous measurement of multiple samples in which a column cleaning period occurs during the continuous measurement. Figure 2 shows an example of such a chromatogram.

[0020] The reference waveform data also includes superimposed peaks (associated with tailing processing) in which another peak is superimposed on a leading or tailing peak, and / or superimposed peaks (associated with either complete separation or vertical separation) in which the tails of multiple peaks overlap. Figure 3 shows examples of separating superimposed peaks by tailing processing, complete separation, and vertical separation.

[0021] Generally, in superimposed peaks, peak start point A, peak start point B, peak end point A, and peak end point B are located in this order from the side with the shortest retention time (the origin side). If more than half of the entire section of the superimposed peaks is tailing of one peak, tailing processing is suitable for separating the superimposed peaks. If the end point of one peak (with a short retention time) coincides with the start point of the other peak (with a long retention time), vertical separation is suitable. However, this only applies to standard chromatograms. If a method for separating superimposed peaks in a chromatogram with a fluctuating baseline is determined based on the above-mentioned rule-based method, an appropriate separation method may not necessarily be obtained. For example, a prior patent application by the present applicant (Patent Application No. 2023-065939) describes a technique for determining multimodality based on the intensity ratio (peak height, peak valley depth) of adjacent peaks or the peak distance. The technology described in this prior application (Patent Application No. 2023-065939) can be applied to determine multimodality by comparing peaks with each other instead of comparing peaks with noise, but these rule-based methods may not be able to determine an appropriate separation method.

[0022] An example of such a chromatogram is shown in Figure 4. The upper part of Figure 4 shows overlapping peaks on a chromatogram. Conventional techniques that determine separation methods based on rules, as described above, would recommend vertical separation for these overlapping peaks, as shown in the middle part of Figure 4. However, as shown in the enlarged view in the lower part of Figure 4, these overlapping peaks are small peaks that appear on the tailing of a larger peak located on the shorter retention time side, and are overlapping peaks that should be separated by tailing processing. A rule-based approach is not sufficient to correctly separate these overlapping peaks; therefore, using a trained model constructed by machine learning, as in this embodiment, is effective.

[0023] The labels for the tailing process are further subdivided into a single peak on the tailing, a vertically separated peak on the tailing, a peak on the tailing, and a peak on the leading edge. Figure 5 shows a schematic example of peaks associated with peak labels (indicators) for a single peak on the tailing, a vertically separated peak on the tailing, a peak on the tailing, and a peak on the leading edge.

[0024] The control and processing unit 40 also includes, as functional blocks, a trained model creation unit 51, a measurement condition setting unit 52, a measurement execution unit 53, a window setting unit 54, a first index output processing unit 55, a second index output processing unit 56, a third index output processing unit 57, a peak portion estimation unit 58, and an analysis result output unit 59. The control and processing unit 40 is actually a general personal computer, and each of the above functional blocks is realized by executing a pre-installed waveform analysis program on the computer's processor. The control and processing unit 40 is also connected to an input unit 6 consisting of a keyboard, a mouse, etc., and a display unit 7 consisting of a liquid crystal display, etc.

[0025] Next, we will explain a method for analyzing a chromatogram using the chromatography mass spectrometry system of this embodiment. In the chromatography mass spectrometry system of this embodiment, when the waveform analysis program is executed, a screen is displayed on the display unit 7, which allows the user to select either creating a trained model or analyzing chromatogram data.

[0026] First, the procedure for creating a trained model will be described with reference to the flowchart in Figure 6.

[0027] When the user selects the creation of a trained model, the trained model creation unit 51 prepares an untrained trained model (Step 1). Various trained models capable of performing semantic segmentation can be suitably used for this trained model. Semantic segmentation is generally used to analyze images composed of pixel data distributed two-dimensionally, but in this embodiment, it is applied to the analysis of chromatogram waveform data composed of data arranged one-dimensionally along the time axis, for example. As trained models capable of performing semantic segmentation, for example, U-Net, SeGNet, PSPNet, etc. can be used (for example, Patent Document 1). In this embodiment, U-Net is used.

[0028] Next, the trained model creation unit 51 reads the reference waveform data from the reference waveform data storage unit 42. Then, using the reference waveform data as training data, the trained model performs machine learning, creating a trained model that, when measurement data is input, outputs labels representing the attributes of each data element constituting the measurement data. In this embodiment, the labels may be, for example, baseline, single (single peak), fully resolved peak, vertically resolved peak, peak start point, peak end point, single peak on the tail, vertically resolved peak on the tail, or column wash section. The types of labels can be changed as appropriate depending on the purpose of the analysis to be performed later. Only some of the above labels may be used, or additional labels may be added. An example of a label to be added is the uneluted section (the time until the component with the shortest retention time in the sample elutes from the column, i.e., a time period during which peak detection is not necessary).

[0029] The number of data points that can be input to the learning model can be any value, but if the number of data points is large, processing will take time. Therefore, in this embodiment, the number of data points input to U-Net is set to 1,024.

[0030] First, the trained model creation unit 51 inputs 1,024 measurement data elements from the beginning (the side with the shortest time; the same applies below) of the measurement data into the U-Net as a set, and performs machine learning on the training data. The range (frame) for extracting a set of partial measurement data from the measurement data is called a window. The first window used here is set to a width equivalent to the sampling rate x 1,024. Then, as schematically shown in the upper part of Figure 7, the first window is moved while overlapping adjacent first windows by 1 / 3 to 1 / 2 of the width, and machine learning is performed on the entire reference waveform data. This allows almost all peaks appearing on the chromatogram to be entirely contained within one of the first windows. Although it is possible to distinguish peak and non-peak portions even if a portion of a peak is located outside the first window, configuring the peak to be entirely located within the window improves the accuracy of peak identification.

[0031] Next, the trained model creation unit 51 applies a second window having a width that is pre-associated with the type of detector, and performs machine learning similar to that described above.

[0032] As mentioned above, various types of detectors are used in liquid chromatographs depending on the components to be detected, such as mass spectrometers, ultraviolet absorbance detectors (UV detectors), photodiode array detectors (PDA detectors), refractive index detectors (RIDs), and electrical conductivity detectors. The shape and width of the peaks that appear in the chromatogram vary depending on these detectors. Therefore, it is effective to determine the width of the second window for each detector type so that the widest peak expected for each detector is completely contained within a single second window.

[0033] For example, when the detector is a mass spectrometer, the expected peak width is approximately 1.5 minutes at most, whereas when the detector is a PDA detector or UV detector, peaks with widths of 5 minutes or 10 minutes may appear. Therefore, the width of the second window is predetermined according to the type of detector. For example, when the detector is a mass spectrometer, the width of the second window is predetermined to 3 minutes, and when the detector is a PDA detector or UV detector, the width of the second window is predetermined to 15 minutes. The width of the second window is set to, for example, 1.5 to 2 times the maximum expected peak width. When performing sliding window analysis, the window can be moved so that 1 / 3 to 1 / 2 of the width of the second window overlaps between adjacent windows, as schematically shown in the middle of Figure 7.

[0034] When using the second window described above, the number of points of the measurement data elements present within the second window will be greater than the number of points of the data elements input to U-Net. Therefore, when using the second window described above, it is recommended to adjust the number of points of the measurement data elements present within the second window to 1,024 by summing or averaging multiple measurement data elements or thinning out the measurement data elements before inputting them into U-Net and performing machine learning.

[0035] Furthermore, the trained model creation unit 51 may perform the same machine learning as described above, applying a third window having a width equivalent to the entire measurement time period, on all reference waveform data.

[0036] Liquid chromatographs sometimes perform gradient analysis, in which the mixing ratio of multiple mobile phases is gradually changed during measurement. Gradient analysis can cause a so-called drift, where the baseline gradually increases (or decreases) throughout the entire measurement time period. Machine learning of only a portion of the reference waveform data makes it difficult to obtain a trained model that can correctly identify such drifts and peaks. Therefore, in this embodiment, as shown schematically in the lower part of Figure 7, machine learning is performed using a single third window that corresponds to the entire measurement time period. Even when using the third window, the number of measurement data elements within the third window will be greater than the number of data elements input to U-Net. Therefore, when using the third window, the number of measurement data elements within the third window is adjusted to 1,024 by averaging multiple measurement data elements or thinning out measurement data elements, as described above, before inputting the data into U-Net and performing machine learning.

[0037] The trained model creation unit 51 performs the above processing to construct a trained model and stores it in the trained model storage unit 44. In this embodiment, as described above, the reference waveform data includes, for example, a chromatogram obtained by gradient analysis in which the baseline increases (or decreases) over the entire measurement time, and a chromatogram obtained by continuous measurement of multiple samples in which a column cleaning period occurs during the continuous measurement. The reference waveform data also includes superimposed peaks (associated with tailing processing) in which a leading or tailing peak is superimposed on another peak, and / or superimposed peaks (associated with either complete separation or vertical separation) in which the tails of multiple peaks overlap. Therefore, a trained model capable of identifying various baseline configurations and superimposed peaks of various shapes is constructed.

[0038] Regarding the column washing interval, it is conceivable to read information about the column washing time period from a method file containing the measurement conditions and reflect that time period. However, there is no guarantee that the washing time period set as the measurement conditions will exactly match the time period during which the column is washed during the actual measurement, and a discrepancy may occur between the two. In particular, when measuring multiple samples consecutively, such discrepancies may accumulate toward the end of a series of measurements. As a result, if the column washing time read from the measurement conditions is treated as the washing interval, the chromatogram during the time period during which the components in the sample are flowing out of the column may be mistakenly identified as a washing interval, i.e., a period in which no peaks are detected. In contrast, in this embodiment, the washing interval is estimated based on chromatogram data, so that the washing interval of the chromatogram can be correctly identified even if there is a discrepancy between the column washing time period set in the measurement conditions and the column washing interval in the actual measurement.

[0039] In this embodiment, three trained models were created as described above, but only one trained model may be constructed using any of the three windows. However, as in this embodiment, it is preferable to construct multiple trained models using windows of different widths. The three windows described above are each suitable for identifying peaks and backgrounds of different widths, and using these multiple trained models allows for more accurate peak detection.

[0040] Next, the procedure for analyzing the waveform of an unanalyzed chromatogram will be described with reference to the flowchart in Figure 8. The following describes an example in which three types of trained models have been constructed by machine learning, each using the three types of windows described above. In the following description, the trained model constructed by machine learning using a window (first window) with a width determined based on the sampling rate will be referred to as the first trained model, the trained model constructed by machine learning using a window (second window) with a width determined according to the type of detector will be referred to as the second trained model, and the trained model constructed by machine learning using a window (third window) with a width equivalent to the entire measurement range will be referred to as the third trained model.

[0041] When the user places a sample in the autosampler 16 and instructs the start of analysis, the measurement condition setting unit 52 reads out the measurement conditions stored in the measurement data storage unit 43 and displays them on the screen of the display unit 7. These measurement conditions include the type of detector to be used for the measurement and information on its sampling rate. When the user selects the measurement conditions to be used from the displayed measurement conditions (or makes appropriate changes) and instructs the start of measurement, the measurement condition setting unit 52 creates a batch file that executes the measurement under the selected conditions and saves it in the measurement data storage unit 43.

[0042] When the user instructs the execution of a measurement, the measurement execution unit 53 executes a batch file stored in the measurement data storage unit 43 to perform chromatographic analysis of the sample, obtains chromatogram data, and stores it in the measurement data storage unit 43. Like the reference waveform data, this chromatogram data is one-dimensional data in which, for example, output signals from a detector are arranged in time series, and corresponds to the analysis target data in the present invention. Here, an example has been described in which a chromatogram is obtained by measuring a sample using the measurement execution unit 53, but chromatogram data may also be obtained by reading chromatogram data obtained in advance.

[0043] After acquiring chromatogram data by measuring a sample or reading out acquired data (step 11), when the user issues an instruction to analyze the chromatogram data, the window setting unit 54 creates a chromatogram from the read-out data and displays it on the screen of the display unit 7 (step 12). Based on the sampling rate, detector type, and total measurement time listed in the measurement conditions, the window setting unit 54 determines the corresponding window width values ​​and displays these values ​​on the display unit 7. The width of the first window is the sampling rate x 1,024, the width of the second window is the value associated with the detector type, and the width of the third window is the total measurement time. The user checks the values ​​of each window displayed on the display unit 7 and determines these values ​​by performing a predetermined input operation (step 13).

[0044] Once the user determines the width of each window, the first index output processing unit 55 reads 1,024 measurement data elements from the beginning of the chromatogram data and inputs them into the first trained model. The first trained model outputs one of the following labels for each input measurement data element: baseline, single (single peak), fully separated peak, vertically separated peak, peak start point, peak end point, single peak on the tail, vertically separated peak on the tail, or wash interval (step 14). The position of the first window is shifted so that the ranges of adjacent windows overlap, and 1,024 data elements are input. A label for each data element is output. This process is performed over the entire measurement range. In this way, one or more labels are output for all measurement data elements (multiple labels are output for measurement data located in the overlapping window area).

[0045] The second index output processor 56 applies a second window to the chromatogram and reduces the number of data elements contained therein to 1,024. Specifically, similar to when the second window was applied to the training data, multiple measurement data elements are summed, averaged, or thinned. The second index output processor 56 then reads 1,024 measurement data elements from the beginning of the chromatogram data and inputs them into the second trained model. The second trained model outputs one of the following labels for each input measurement data element: peak start point, peak end point, single peak, tailing peak, fully separated peak, vertically separated peak, or non-peak portion. The position of the second window is shifted so that the ranges of adjacent windows overlap, and the 1,024 data elements are input. A label for each data element is output. This process is performed over the entire measurement range. In this way, one or more labels are output for all measurement data elements (step 15). Here, multiple labels are also output for measurement data elements located in the overlapping portion of the windows.

[0046] The third indicator output processing unit 57 performs a process to reduce the total number of measurement points to 1,024. Specifically, similar to when the third window is applied to the training data, multiple measurement data elements are summed, averaged, or the measurement data is thinned out. The third indicator output processing unit 57 then inputs the 1,024 measurement data elements into the third trained model. For each input measurement data element, the third trained model outputs one of the following labels: baseline, single (single peak), fully separated peak, vertically separated peak, peak start point, peak end point, single peak on the tailing, vertically separated peak on the tailing, or wash interval. In this way, one label is output for each measurement data element (step 16).

[0047] When processing is complete, applying all windows to the chromatogram data to be analyzed, the peak portion estimation unit 58 determines a label for each measurement data element. If multiple labels are output for the same measurement data element (measurement point), they are integrated. Then, the peak portion is estimated based on the label for each measurement data element (step 17). If different labels are output for one measurement data element, the label for that measurement data element (measurement point) is determined based on a predetermined priority. Specifically, for example, if labels for peak portions and non-peak portions are output, the peak portion is prioritized. Furthermore, regarding single peaks and overlapping peaks (tailing-processed peaks, fully separated peaks, and vertically separated peaks), the overlapping peaks are prioritized. This makes it possible to avoid overlooking the existence of a peak or erroneously estimating that an overlapping peak that requires peak separation is a single peak.

[0048] Furthermore, regarding the label of a cleaning interval, it is advisable to determine the interval as a cleaning interval only if the measurement data elements to which the label is output are consecutive and the duration of the consecutive time is equal to or greater than a predetermined time, or if the proportion of the consecutive time to the total time range of the chromatogram to be analyzed is equal to or greater than a predetermined value. Because it is impossible to perform column cleaning in a time equivalent to one or several measurement data elements, performing this processing makes it possible to correct an error even if the trained model erroneously outputs a label for a cleaning interval. The column cleaning time is predetermined as a measurement condition, and the length of the time and its proportion to the total measurement time can be read from the measurement conditions.

[0049] Furthermore, even when using a trained model that outputs a label for an uneluted interval (the time until the component contained in the sample with the shortest retention time flows out of the column, i.e., identifies the time period during which peak detection is not necessary), similarly to the above, the interval can be determined as an uneluted interval only when the measurement data elements for which the label is output are continuous and the length of this continuous time is equal to or greater than a predetermined time, or when the proportion of this continuous time to the entire time range of the chromatogram to be analyzed is equal to or greater than a predetermined value. Note that the above-mentioned washing interval and uneluted interval can also be collectively referred to as a non-detection interval.

[0050] Finally, the analysis result output unit 59 displays the analysis results (labels for each measurement data element) together with the chromatogram of the analysis target on the display unit 7 (step 18), thereby enabling the user to confirm the peaks estimated to be present on the chromatogram of the analysis target.

[0051] When analyzing a known target component contained in a sample (target analysis), for example, a mass spectrometer is used as a detector, and SIM or MRM measurements are performed using ions generated from the target component as target ions to create an extracted ion chromatogram. In targeted analysis, peaks are detected from the waveform during a limited time period (e.g., 1.5 minutes) corresponding to the retention time of the target component out of the total measurement time of the chromatograph (see, for example, Non-Patent Document 3). Furthermore, SIM and MRM measurements are highly selective for the target component, resulting in narrow, sharp peaks. The waveform analysis technology described in Patent Document 1 was developed with the detection of peaks from such waveforms in mind.

[0052] On the other hand, when comprehensively analyzing unknown components contained in a sample (non-target analysis) or when using a PDA detector, UV detector, etc. as a detector, component selectivity is lower than in targeted analysis using a mass spectrometer, and multiple peaks are likely to overlap on the chromatogram.

[0053] Additionally, gradient analysis in liquid chromatographs and temperature-programmed analysis in gas chromatographs are prone to baseline drift. When multiple overlapping peaks (superimposed peaks) appear on a chromatogram, tailing, complete separation, and vertical separation are known methods for separating them. However, the appropriate method for separating overlapping peaks depends on the shape of the baseline. Therefore, conventional trained models sometimes have difficulty correctly detecting peaks due to baseline shapes that appear differently depending on the configuration of the analytical instrument and measurement conditions.

[0054] In contrast, in this embodiment, as described above, the reference waveform data includes chromatograms obtained by gradient analysis and showing an increase (or decrease) in the baseline over the entire measurement time, chromatograms obtained by continuous measurement of multiple samples and including a column cleaning period during the continuous measurement, superimposed peaks (associated with tailing treatment) in which a leading or tailing peak is superimposed on another peak, and / or superimposed peaks in which the tails of multiple peaks overlap (associated with either complete separation or vertical separation), and further, chromatograms in which the tailing treatment labels are subdivided into single peaks on the tail, vertically separated peaks on the tail, peaks on the tail, and peaks on the leading. Therefore, a trained model capable of identifying various baseline forms and superimposed peaks of various shapes is constructed. Furthermore, because such trained models are used to detect peaks from the chromatogram data to be analyzed, even when baseline drift occurs or superimposed peaks appear, peaks can be correctly detected and the separation method to be used to separate the superimposed peaks can be correctly estimated.

[0055] In this embodiment, a first trained model, a second trained model, and a third trained model are constructed by machine learning using three windows of different widths. This makes it possible to use the first trained model, which can correctly detect narrow peaks, and the second trained model, which can correctly detect wide peaks. Furthermore, the third trained model, which uses the entire measurement range as one window, can correctly detect the baseline that fluctuates over the entire measurement time range and correctly distinguish between baseline fluctuations and peaks.

[0056] It is possible to enlarge minute peaks in the time direction (and intensity direction) to create measurement data that artificially represents broad peaks, and then train a learning model on this data. However, in this case, the machine learning will be performed on data in which not only the peaks but also the noise level are stretched in the time direction. Because such time-stretched noise is not detected in actual measurements, such machine learning will construct a trained model that has learned waveforms that do not appear in actual measurement data. As a result, such a trained model will not be able to correctly distinguish between non-peak portions (noise portions) and peak portions contained in the data to be analyzed obtained in actual measurements.

[0057] Next, an example will be described in which the inventor analyzed an actual chromatogram using the waveform analysis device and waveform analysis method of the above embodiment. In this example, a trained model was constructed using the same processing as in the above embodiment, outputting labels (indicators) indicating 0 (baseline), 1 (single: independent single peak), 2 (superimposed peaks to be vertically separated), 3 (peak start point), 4 (peak end point), 5 (single peak on the tail), 6 (vertically separated peak on the tail), or 7 (washing section). Then, of the chromatogram data obtained by sequentially performing gradient analysis on multiple samples, chromatogram data corresponding to the measurement of one sample was used as the analysis target data. This analysis target data was input into the trained model, and one of the above labels was output for each measurement data element constituting the measurement data.

[0058] The analysis results are shown in Figure 9. The top panel of Figure 9 shows the waveform of a chromatogram created from the data to be analyzed (the same chromatogram as Figure 2), the middle panel of Figure 9 plots the labels for each measurement data element output from the trained model, and the bottom panel of Figure 9 provides an explanation of the labels. Comparing the chromatogram waveform in the top panel of Figure 9 with the output labels in the middle panel of Figure 9 reveals that by using the waveform analysis device and waveform analysis method of this embodiment, the baseline fluctuates significantly due to gradient analysis and column cleaning. Furthermore, the waveform shows peaks of various sizes appearing on the drifted baseline, and the peaks are correctly detected. While the chromatogram used in this example does not show superimposed peaks, the inventors output similar labels for other chromatograms and confirmed that the detection of superimposed peaks and the estimation of separation methods suitable for peak separation were also performed correctly.

[0059] The above embodiment is merely an example and can be modified as appropriate in accordance with the spirit of the present invention.

[0060] In the above embodiment, three trained models were created by machine learning using three windows of different widths, but the number of window types and trained models may be one or two, or may be four or more.

[0061] In the above embodiment, the data to be analyzed is chromatogram data obtained by measurement using a liquid chromatograph. However, the waveform analysis method and waveform analysis device according to the present invention can be used to analyze waveforms composed of various types of data. For example, chromatogram data obtained by measurement using a gas chromatograph, and waveforms composed of measurement data other than chromatogram data, can also be analyzed in a similar manner. Furthermore, for example, spectroscopic spectra (waveforms representing changes in detected intensity on the wavelength or wavenumber axis) obtained by measurement using a spectrophotometer, and mass spectra obtained by measurement using a mass analyzer can also be analyzed in a similar manner.

[0062] In the above embodiment, the second trained model was constructed by applying a second window with a width previously associated with the type of detector to training data and performing machine learning. However, multiple trained models may be created by applying windows of different widths to training data for a single detector, and information regarding the window width (and detector type) may be associated and stored in the trained model storage unit 44. In this case, the user can change the width of the second window by checking the shape of the chromatogram created from the chromatogram data to be analyzed, displayed on the display unit 7. When the user changes the width of the second window, the second indicator output processing unit 56 reads out the second trained model corresponding to the changed width of the second window from the trained model storage unit 44 and outputs a label for each measurement data element in the same manner as described above. Furthermore, if it is clear that no fluctuations, such as drift, have occurred in the baseline of the chromatogram to be analyzed over the entire measurement time, the peak portion may be estimated without using the third window.

[0063] Alternatively, in the above configuration, instead of changing the window width, the user may input the value of the expected peak width. In that case, the value obtained by multiplying the input peak width by a predetermined constant (e.g., 1.5 or 2) is set as the width of the second window, and processing is performed in the same manner as above.

[0064] Furthermore, in the above embodiment, U-Net was used for the first trained model, the second trained model, and the third trained model, but different trained models may be used for each trained model. These trained models can be preferably neural networks, including those that perform semantic segmentation, those that perform object detection (SSD), those that use a stochastic model, recurrent neural networks (RNN), transformers, and various other architectures. By constructing trained models using these multiple types of architectures appropriately, the accuracy of peak detection can be further improved.

[0065] In the above embodiment, one label is output for each measurement data element and the analysis results are displayed on the display unit 7. However, depending on the architecture of the trained model, multiple labels and the accuracy of each label can be output as an inference result for one measurement data element. The U-Net described in the above embodiment is one such architecture. In the above embodiment, only one label with the highest accuracy is output. However, when using such a trained model, other labels and their accuracy may be displayed on the display unit 7 in addition to the label with the highest accuracy. This allows the user to more accurately estimate the peak, even if the label determined to be the most accurate is incorrect, by changing it to the label with the next highest accuracy. Alternatively, labels with accuracy greater than or equal to a predetermined value may be displayed on the display unit 7. In this case, the user only needs to check the labels with the highest accuracy, thereby efficiently progressing the analysis.

[0066] [Aspect] It will be apparent to those skilled in the art that the above-described exemplary embodiments are examples of the following aspects.

[0067] (Section 1) One aspect of the present invention is a waveform analysis method for analyzing a waveform configured from analysis target data, which is data acquired by measuring a sample using an analytical device, comprising: a trained model construction step for constructing a trained model by machine learning using a plurality of reference waveform data as training data, the reference waveform data being data constituting a plurality of reference waveforms each having a different baseline shape, with the positions of peak portions including superimposed peaks known and associated with any of tailing processing, complete separation, and vertical separation as a method for separating the superimposed peaks, and for which data elements constituting the measurement data are input, outputting an index that represents a single peak portion, a superimposed peak portion, or a non-peak portion and associates whether tailing processing, complete separation, or vertical separation should be used as a method for separating the superimposed peaks; and an index output step of inputting the analysis target data into the trained model, and outputting, from the trained model, an index representing a single peak portion, a superposed peak portion, or a non-peak portion for each of a plurality of analysis target data elements constituting the analysis target data, and associating the superposed peak with which of tailing processing, complete separation, and vertical separation should be used as a method for separating the superposed peaks; Equipped with.

[0068] (Section 2) Another aspect of the present invention is a waveform analysis device that analyzes a waveform configured from analysis target data, which is data acquired by measuring a sample using an analytical device, comprising: a trained model storage unit in which trained models are stored, the trained models being constructed by machine learning using as training data a plurality of reference waveform data, the reference waveform data being data constituting a plurality of reference waveforms each having a different baseline shape, with the positions of peak portions including superimposed peaks known, and associated with any of tailing processing, complete separation, and vertical separation as a method for separating the superimposed peaks, and which, when measurement data is input, outputs an index representing a single peak portion, a superimposed peak portion, or a non-peak portion for each data element constituting the measurement data, and associating it with whether tailing processing, complete separation, or vertical separation should be used as a method for separating the superimposed peaks; and an index output unit that inputs the analysis target data into the trained model, and outputs, for each of a plurality of analysis target data elements that make up the analysis target data, an index that represents a single peak portion, a superposed peak portion, or a non-peak portion from the trained model, and that associates an index with which of tailing processing, complete separation, and vertical separation should be used as a method for separating the superposed peaks; Equipped with.

[0069] In the waveform analysis method according to paragraph 1 and the waveform analysis device according to paragraph 2, machine learning is performed using, as training data for constructing a trained model through machine learning, multiple reference waveform data, each of which is data constituting multiple reference waveforms, in which the positions of peak portions, including overlapping peaks (multiple overlapping peaks) with different baseline shapes are known, and which are associated with tailing, complete separation, or vertical separation as a method for separating the overlapping peaks. Thus, a trained model is constructed that estimates peak portions based on the baseline shapes and estimates an appropriate peak separation method for the overlapping peaks based on the baseline shapes. In the waveform analysis method according to paragraph 1 and the waveform analysis device according to paragraph 2, when data to be analyzed is input into the trained model, the peak portions are correctly estimated based on the baseline shapes that appear uniquely depending on the detector and measurement conditions, and an index indicating a peak separation method appropriate for separating the overlapping peaks is output. Therefore, peaks can be correctly detected from a chromatogram regardless of the device configuration or measurement conditions.

[0070] (Section 3) The waveform analyzer according to paragraph 3 is the waveform analyzer according to paragraph 2, The tailing treatment further includes a single peak on the tail and vertically separated peaks on the tail.

[0071] The waveform analyzer according to paragraph 3 can use more specific methods for separating superimposed peaks, where another peak appears above a tailing peak, and therefore can separate superimposed peaks more appropriately.

[0072] (Section 4) The waveform analyzer according to paragraph 4 is a waveform analyzer according to paragraph 2 or 3, The reference waveform includes a non-detection section in which peak detection is not required, The trained model further outputs an index representing a non-detection interval.

[0073] (Section 5) The waveform analyzer according to paragraph 5 is the waveform analyzer according to paragraph 4, The non-detection section includes a chromatogram of the time period until the component with the shortest retention time among the components contained in the sample flows out of the column, and / or a chromatogram of the time period during which the column is washed.

[0074] The waveform analyzer according to paragraph 4 uses a trained model that has been machine-learned on the non-detection intervals contained in the reference waveform to estimate the non-detection intervals contained in the data to be analyzed, making it possible to efficiently estimate peaks by excluding the non-detection intervals from the waveform data to be analyzed. As described in paragraph 5, the non-detection intervals include, for example, a chromatogram of the time period until the component with the shortest retention time contained in the sample flows out of the column, and a chromatogram of the time period during which the column is washed in measurements in which multiple samples are continuously subjected to gradient analysis.

[0075] (Section 6) The waveform analyzer according to paragraph 6 is a waveform analyzer according to paragraph 4 or 5, The index output unit determines a section to be a non-detection section if the time period during which the non-detection section is continuous is longer than a predetermined time, or if the ratio of the time period during which the non-detection section is continuous to the time during which the data to be analyzed was acquired exceeds a predetermined value.

[0076] In the waveform analysis device according to paragraph 6, even if the trained model erroneously outputs an index of a non-detection section for measurement data in a short time period, the error can be corrected.

[0077] (Section 7) The waveform analyzer according to paragraph 7 is a waveform analyzer according to any one of paragraphs 2 to 6, The trained model is configured with an architecture that outputs multiple indicators and the accuracy of each indicator for one measurement data element, and outputs multiple indicators and the accuracy of each indicator from the trained model for each of the data elements to be analyzed.

[0078] When a waveform with a complex shape is input into a trained model, multiple indices with similar accuracy may be output for a single data element. In such cases, the waveform analyzer according to paragraph 7 allows the user to check multiple indices with similar accuracy and select the appropriate one from among them.

[0079] (Section 8) The waveform analyzer according to paragraph 8 is the waveform analyzer according to paragraph 7, The trained model outputs an indicator for each of the data elements to be analyzed that indicates that the probability is equal to or greater than a predetermined value.

[0080] In the waveform analysis device according to paragraph 8, the user only needs to check labels with high accuracy, so analysis can proceed efficiently. [Explanation of symbols]

[0081] 1...Liquid chromatograph system 10...Liquid chromatograph section 11...Mobile phase container 12...Liquid transfer pump 13...Injector 14...Column 15...Detector 16...Autosampler 40...Control and processing section 41...Storage section 42...Reference waveform data storage section 43...Measurement data storage unit 44…Trained model memory section 51...Trained model creation unit 52...Measurement condition setting section 53...Measurement execution unit 54...Window settings section 55...First index output processing unit 56...Second index output processing unit 57...Third index output processing unit 58...Peak part estimation section 59...Analysis result output section 6...Input section 7…representation section

Claims

1. A waveform analysis method for analyzing a waveform configured from analysis target data, which is data acquired by measuring a sample using an analytical device, comprising: a trained model construction step for constructing a trained model by machine learning using a plurality of reference waveform data as training data, the reference waveform data being data constituting a plurality of reference waveforms each having a different baseline shape, with the positions of peak portions including superimposed peaks known and associated with any of tailing processing, complete separation, and vertical separation as a method for separating the superimposed peaks, and for which data elements constituting the measurement data are input, outputting an index that represents a single peak portion, a superimposed peak portion, or a non-peak portion and associates whether tailing processing, complete separation, or vertical separation should be used as a method for separating the superimposed peaks; and an index output step of inputting the analysis target data into the trained model, and outputting, from the trained model, an index representing a single peak portion, a superposed peak portion, or a non-peak portion for each of a plurality of analysis target data elements constituting the analysis target data, and associating the superposed peak with which of tailing processing, complete separation, and vertical separation should be used as a method for separating the superposed peaks; A waveform analysis method comprising:

2. A waveform analysis device that analyzes a waveform configured from analysis target data, which is data acquired by measuring a sample using an analytical device, a trained model storage unit in which trained models are stored, the trained models being constructed by machine learning using as training data a plurality of reference waveform data, the reference waveform data being data constituting a plurality of reference waveforms each having a different baseline shape, with the positions of peak portions including superimposed peaks known, and associated with any of tailing processing, complete separation, and vertical separation as a method for separating the superimposed peaks, and which, when measurement data is input, outputs an index representing a single peak portion, a superimposed peak portion, or a non-peak portion for each data element constituting the measurement data, and associating it with whether tailing processing, complete separation, or vertical separation should be used as a method for separating the superimposed peaks; and an index output unit that inputs the analysis target data into the trained model, and outputs, for each of a plurality of analysis target data elements that make up the analysis target data, an index that represents a single peak portion, a superposed peak portion, or a non-peak portion from the trained model, and that associates an index with which of tailing processing, complete separation, and vertical separation should be used as a method for separating the superposed peaks; A waveform analysis device comprising:

3. The waveform analyzer according to claim 2 , wherein the tailing process further includes a single peak on the tailing and a vertically separated peak on the tailing.

4. The reference waveform includes a non-detection section in which peak detection is not required, The waveform analysis device according to claim 2 , wherein the trained model further outputs an index representing a non-detection section.

5. The waveform analysis device according to claim 4, wherein the non-detection section includes a chromatogram of a time period until the component with the shortest retention time among the components contained in the sample flows out of the column, and / or a chromatogram of a time period during which the column is washed.

6. 5. The waveform analysis device according to claim 4, wherein the index output unit determines a section to be a non-detection section when the time period during which the non-detection section is continuous is longer than a predetermined time, or when the ratio of the time period during which the non-detection section is continuous to the time during which the analysis target data is acquired exceeds a predetermined value.

7. 3. The waveform analysis device of claim 2, wherein the trained model is configured with an architecture that outputs multiple indices and the accuracy of each index for one measurement data element, and the trained model outputs multiple indices and the accuracy of each index for each of the data elements to be analyzed.

8. The waveform analysis device according to claim 7 , wherein the trained model outputs an index for each of the analysis target data elements, the index being equal to or greater than a predetermined value.

Images