Waveform analysis method, waveform analysis device, and analyzer

Abstract

To provide more accurate peak information when performing peak detection automatically.SOLUTION: Provided is a method for analyzing the signal waveform of a chromatograph, etc. The method includes: a step for creating a trained model that identifies a peak section included in an inputted waveform, by machine learning of a set of a plurality of partial waveforms created by dividing a reference waveform; a step for dividing the waveform to be analyzed into a plurality of partial waveforms, determining whether or not each of the plurality of partial waveforms is a peak section, and estimating a plurality of different types of regions in the waveform to be analyzed, including an overlapping peak region, etc.; and a step for determining whether or not an overlapping peak included in a region estimated to be the overlapping peak region is a multicrested peak, by using the height of one of a plurality of peaks or the plurality of peaks, the depth of a trough between adjacent two peaks, and the width in lateral direction between the bottom of the trough and the apex of one of peaks sandwiching the trough.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to a method and device for analysing a signal waveform acquired by an analytical device, and to an analytical device equipped with such a waveform analyser. [Background technology]

[0002] In a chromatogram obtained by an analytical device such as a gas chromatograph (GC) or a liquid chromatograph (LC), peaks originating from components contained in a sample appear. In a data device provided in such an analytical device, peaks are generally detected by performing waveform processing on the chromatogram obtained by analysis, and peaks of a target compound are identified by performing identification processing on the detected multiple peaks. In addition, the concentration or content of the compound corresponding to the identified peak is calculated from the area or height of the identified peak.

[0003] To date, various methods have been put into practical use as methods for detecting peaks. In recent years, a new peak detection method that utilizes machine learning has been proposed and put into practical use (see Patent Document 1 and Non-Patent Document 1).

[0004] In the waveform analysis method described in Patent Document 1, a set including a plurality of partial waveforms created by finely dividing a reference waveform, whose peak portion positions are known, in the time axis direction is prepared for each reference waveform, and a trained model is created to identify partial waveforms corresponding to peak portions included in an input waveform by machine learning using the partial waveforms for each of the many sets of reference waveforms. The analysis target waveform is also divided into a plurality of partial waveforms in the same manner as the reference waveform, and a trained model is used to determine whether each of the plurality of partial waveforms is a peak portion. Then, based on the determination result, the peak region and other regions of the analysis target waveform are determined. By performing machine learning using partial waveforms corresponding to peak start points and peak end points in addition to peak portions, partial waveforms corresponding to peak start points and peak end points can also be found among the plurality of partial waveforms in the analysis target waveform. [Prior art documents] [Patent documents]

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

[0006] [Non-Patent Document 1] "Peakintelligence for LCMS LabSolutions LCMS, optional waveform processing software for LabSolutions Insight," [online], [searched March 16, 2023], Shimadzu Corporation, Internet <URL:https: / / www.an.shimadzu.co.jp / products / liquid-chromatograph-mass-spectrometry / lc-ms-software / peakintelligence / index.html> [Non-Patent Document 2] Olaf Ronneberger and 2 others, "U-Net: Convolutional Networks for Biomedical Image Segmentation", [online], [Submitted on 18 May 2015], arXiv.org, Internet.<URL: https: / / arxiv.org / pdf / 1505.04597.pdf> Summary of the Invention [Problem to be solved by the invention]

[0007] In waveforms such as chromatograms, peaks derived from multiple components are often observed overlapping. Therefore, in the waveform analysis method described in Patent Document 1, tailing processing, complete separation, vertical division, and other processing can be performed to separate overlapping peaks. In peak detection using machine learning, the waveform shapes and peak start and end points of various overlapping peaks are learned, and it is possible to determine which separation method will most appropriately separate the overlapping peaks to be analyzed.

[0008] In particular, in chromatograms obtained by analyzing samples with low component concentrations using a gas chromatograph mass spectrometer (GC-MS) or liquid chromatograph mass spectrometer (LC-MS), the number of ions derived from the target component is small, and therefore the peaks may appear multimodal. A multimodal peak (hereinafter referred to as a "multimodal peak") is apparently difficult to distinguish from overlapping peaks. Therefore, in conventional peak detection methods using machine learning, there are cases where the valleys of a multimodal peak are judged to be the end and start points of the peak, which is one of the causes of false peak detection.

[0009] An object of the present invention is to provide a waveform analysis method and a waveform analysis device that can reduce erroneous determinations of superimposed peaks and accurately recognize multi-modal peaks in a peak detection process using machine learning, particularly when the peak has a multi-modal shape due to reasons such as low component concentrations. [Means for solving the problem]

[0010] One aspect of the waveform analysis method according to the present invention is a waveform analysis method for analyzing a signal waveform that is a chromatogram or a spectrum, comprising: a model creation step of creating a trained model for identifying peak portions included in an input waveform by machine learning using a plurality of sets of partial waveforms created by dividing a reference waveform in which the positions of the peak portions are known; a region estimation step of dividing the waveform to be analyzed into a plurality of partial waveforms, determining whether each of the plurality of partial waveforms is a peak portion using the trained model, and estimating a plurality of different types of regions in the waveform to be analyzed, including a single peak region, a superimposed peak region, and a non-peak region, based on the result of the determination; a multimodal determination step for determining whether or not a superimposed peak included in the region estimated to be a superimposed peak region in the region estimation step is a multimodal peak due to one component, using at least one piece of information of the height of one or more of the multiple peaks, the depth of a valley between two adjacent peaks among the multiple peaks, or the width in the horizontal axis direction between the bottom of the valley and the top of one of the peaks sandwiching the valley; has.

[0011] One aspect of the waveform analysis device according to the present invention is a waveform analysis device that analyzes a signal waveform that is a chromatogram or a spectrum, comprising: a region estimation unit that divides a waveform to be analyzed into a plurality of partial waveforms, and determines whether each of the plurality of partial waveforms of the waveform to be analyzed is a peak portion using a trained model created by machine learning using a plurality of sets of the plurality of partial waveforms created by dividing a reference waveform whose peak portion positions are known, and estimates a plurality of different types of regions in the waveform to be analyzed, including a single peak region, a superimposed peak region, and a non-peak region, based on the result of the determination; a multimodal determination unit that determines whether or not a superimposed peak included in a region estimated by the region estimation unit to be a superimposed peak region is a multimodal peak due to one component, using at least one piece of information of the height of one or more of the multiple peaks, the depth of a valley between two adjacent peaks among the multiple peaks, or the width in the horizontal axis direction between the bottom of the valley and the top of one of the peaks sandwiching the valley; Equipped with.

[0012] Furthermore, one aspect of the analysis device according to the present invention is any one of a chromatography device, a mass spectrometer, and an optical measurement device, which includes any aspect of the waveform analysis device according to the present invention as a data analysis section. Effect of the Invention

[0013] According to the above aspects of the waveform analysis method and waveform analysis device of the present invention, even if a peak detection process using machine learning determines that a superimposed peak is a multimodal peak derived from a single component, the waveform processing performed after the peak detection process determines whether the superimposed peak is a multimodal peak derived from a single component, and the determination result can be notified to the user, or the estimated result that the peak is a multimodal peak can be corrected without manual intervention. This makes it possible to improve the accuracy of automated peak detection, including multimodal peaks that tend to appear when, for example, the component concentrations are low. [Brief description of the drawings]

[0014] [Figure 1] 1 is a schematic diagram showing the configuration of an embodiment of an analysis device for carrying out a waveform analysis method according to the present invention. [Diagram 2] 11 is a flowchart showing the steps of a trained model creation process in the analysis device of this embodiment. [Diagram 3] 4 is a flowchart showing the procedure of a peak detection process on a chromatogram waveform to be analyzed in the analyzer of the present embodiment. [Figure 4] FIG. 11 is a waveform diagram showing an example of a multi-modal peak determination process. [Diagram 5] FIG. 11 is a waveform diagram showing an example of a multi-modal peak determination process. [Figure 6] FIG. 11 is a waveform diagram showing an example of a multi-modal peak determination process. [Figure 7] FIG. 11 is a waveform diagram for explaining a method of dividing a superimposed peak. [Figure 8] FIG. 13 is a diagram for explaining an example of a method for integrating multimodal peaks. [Figure 9] FIG. 13 is a diagram for explaining an example of a method for integrating multimodal peaks. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0015] In the waveform analysis method and waveform analysis device according to the above aspects of the present invention, the signal waveform includes a chromatogram obtained by a GC device including a GC-MS, a chromatogram obtained by an LC device including an LC-MS, an electropherogram obtained by an electrophoresis device, etc. In a GC device or LC device in which the detector is a mass spectrometer, the chromatogram includes a total ion (total ion current) chromatogram and an extracted ion chromatogram. In addition, the spectrum includes a mass spectrum obtained by a mass spectrometer (a profile spectrum not subjected to centroid processing), a time-of-flight spectrum obtained by a time-of-flight mass spectrometer before being converted into a mass spectrum, a light intensity spectrum obtained by an optical measurement device such as a spectrometer or a fluorescence measurement device, an X-ray intensity spectrum obtained by an X-ray analysis device, etc.

[0016] Hereinafter, an LC system, which is one embodiment of an analytical device including a waveform analyzer for carrying out a waveform analysis method according to the present invention, will be described with reference to the accompanying drawings. FIG. 1 is a schematic configuration diagram of an LC system A of this embodiment and a system B that creates a trained model used in the LC system A.

[0017] The LC system A in this embodiment includes an LC measurement unit 1, a data analysis unit 2, an input unit 24, and a display unit 25. Although not shown, the LC measurement unit 1 includes a liquid delivery pump, an injector, a column, a column oven, a detector, etc., and performs LC analysis on a given sample to obtain chromatogram data indicating the temporal change in signal intensity obtained by the detector. There is no particular limit to the type or type of detector, and for example, a mass spectrometer, a photodiode array (PDA) detector, etc. can be used.

[0018] The data analysis unit 2 includes functional blocks such as a data collection unit 20, a peak detection processing unit 21, a qualitative / quantitative analysis unit 22, and a display processing unit 23. The peak detection processing unit 21 includes functional blocks such as a waveform preprocessing unit 210, a determination unit 211, a trained model storage unit 212, a region determination unit 213, a multimodal peak candidate extraction unit 214, a multimodal peak determination unit 215, and a multimodal peak integrating unit 216.

[0019] In the data analysis section 2, the data collection section 20 collects and stores the chromatogram data obtained in the LC measurement section 1. The peak detection processing section 21 automatically detects peaks in a chromatogram waveform formed by the collected chromatogram data in response to an instruction received from a user at the input section 24, and outputs peak information including the start and end positions (retention time) of the detected peaks and the range of the peak region. The qualitative and quantitative analysis section 22 identifies components (compounds) corresponding to each peak based on the peak information provided by the peak detection processing section 21, calculates peak height values ​​and peak area values, and calculates quantitative values, which are the concentrations or content amounts of each component, from the values. The display processing section 23 displays the peak detection results and calculated values ​​such as quantitative values ​​calculated based on them on the display section 25 in a predetermined format.

[0020] In general, the entity of the data analysis unit 2 is a personal computer or a more powerful workstation in which a specific software (computer program) is installed, or a computer system including a high-performance computer connected to such a computer via a communication line. That is, the functions of each block included in the data analysis unit 2 are realized by executing software installed on a single computer or a computer system including multiple computers on the computer. Of course, some of these functions may be executed using hardware circuits specialized for specific calculations, such as a digital signal processor.

[0021] The computer program can be provided to the user in a form stored in a computer-readable non-transitory storage medium, such as a CD-ROM, a DVD-ROM, a memory card, a USB memory (dongle), etc. The program can also be provided to the user in the form of data transfer via a communication line such as the Internet. Furthermore, the program can be pre-installed in a computer that is part of the system (strictly speaking, a storage device that is part of the computer) at the time the user purchases the system.

[0022] 1, system B provided separately from LC system A includes a model creation unit 3, and the model creation unit 3 includes, as functional blocks, a learning data storage unit 30, a learning execution unit 31, and a model storage unit 32. The trained model created in this model creation unit 3 is stored in a trained model storage unit 212 in the data analysis unit 2 of LC system A.

[0023] Although the details will be described later, in general, the task of creating a trained model requires a huge amount of calculation. Therefore, the entity of system B is a high-performance computer, and the functions of each block are realized by executing software installed on the computer. Of course, system B may be integrated with LC system A.

[0024] Next, the peak detection process executed mainly by the peak detection processing unit 21 of the data analysis unit 2 will be described. In very general terms, the peak detection processing unit 21 visualizes a chromatogram waveform composed of chromatogram data, and then detects the position or range of each of the multiple types of regions described below by using a semantic segmentation technique based on deep learning, which is a machine learning technique that detects the category and position of objects present in the image.

[0025] [Creating a trained model] As is well known, in the machine learning method, it is necessary to construct a trained model in advance using a large amount of learning data (training data and validation data). As described above, the work of constructing this trained model is not performed in the data analysis unit 2 that is a part of the LC system A, but is performed in the model creation unit 3 that is configured by a separate computer system, and the result is stored in the trained model storage unit 212. FIG. 2 is a flowchart showing an outline of the procedure of the trained model creation process performed in the model creation unit 3.

[0026] When creating a trained model, first, training data based on a reference waveform is prepared (step S1). Here, a large number of diverse chromatogram waveforms are used as reference waveforms. The diverse chromatogram waveforms referred to here are preferably chromatogram waveforms including elements such as various noise contamination, baseline fluctuation (drift), overlapping of multiple peaks, or deformation of peak shape, which may appear in a chromatogram waveform when actually performing peak detection. However, this chromatogram waveform data may not be data collected by actual LC analysis, but may be data created by simulation.

[0027] Peak detection is performed in advance on the chromatogram waveform, which is the reference waveform, and the start and end points of one or more peaks on the waveform are accurately determined. This chromatogram waveform is visualized with the signal intensity, i.e., the vertical axis of the graph, normalized, and further divided into a predetermined number of partial original waveforms in the horizontal axis, i.e., the time axis direction. This division number is determined so that the width (length in the time axis direction) of each partial waveform is at least smaller than the peak width. Therefore, the division number can be appropriately set according to the expected minimum value of the peak width.

[0028] A chromatogram waveform is composed of many partial waveforms. Data constituting each partial waveform is associated with characteristic information indicating which of multiple types of regions the partial waveform belongs to. Here, the multiple types of regions are seven types: a single peak region that corresponds to a peak portion on the chromatogram waveform and is a single peak that is not overlapped with other peaks, a tailing peak region that corresponds to a peak portion on the chromatogram waveform and is overlapped with other peaks, and tailing is suitable as a method for dividing the peak, a complete separation peak region that corresponds to a peak portion on the chromatogram waveform and is overlapped with other peaks, and complete separation is suitable as a method for dividing the peak, a vertical division peak region that corresponds to a peak portion on the chromatogram waveform and is overlapped with other peaks, and vertical division is suitable as a method for dividing the peak, a peak start region that includes the start point of the peak, a peak end region that includes the end point of the peak, and a non-peak region that is not a peak portion (usually a baseline). In addition, the tailing processed peak region, the completely separated peak region, and the vertically separated peak region are regions in which a plurality of peaks overlap, and therefore are collectively referred to as overlapping peak regions.

[0029] Here, the method of dividing the peak will be briefly explained with reference to FIG. Tailing is a method of dividing a target superimposed peak into two peaks by treating the start point to the end point of the target superimposed peak as one peak and superimposing another peak on the first peak, as shown in Figure 7(A). Complete separation is a method of dividing a target superimposed peak into two peaks by connecting the start point, minimum point, and end point of the target superimposed peak in order with straight lines, as shown in Figure 7(B). Vertical division is a method of dividing a target superimposed peak into two peaks by a perpendicular line passing through the minimum point of the target superimposed peak, as shown in Figure 7(C). These methods are not limited to peak detection methods using machine learning, but are used in conventional general peak detection.

[0030] In tailing processing, the start and end points of two peaks are arranged in the order of the first peak start and second peak start from the side with the shortest retention time, followed by the first peak end and second peak end. On the other hand, in complete separation and vertical division, the first peak start and end are arranged in the order of the first peak start and end from the side with the shortest retention time, followed by the second peak start and end. In addition to these division methods, peaks may be separated by fitting using a model function such as a Gaussian function.

[0031] In order to classify into the seven types of regions as described above, a reference waveform including a single peak and a reference waveform including overlapping peaks separated by each of the methods of tailing processing, complete separation, and vertical division are prepared as reference waveforms, and a plurality of sets of reference waveforms each consisting of a plurality of partial waveforms created by dividing the reference waveform are prepared. Each partial waveform is added with characteristic information indicating which region it corresponds to. The learning data storage unit 30 stores partial waveform data constituting a large number of chromatogram waveforms and characteristic information associated therewith together. The learning data may be divided into training data and verification data in advance, or may be used as either training data or verification data as appropriate during learning without such a division.

[0032] When the start of creating a learning model is instructed, the learning execution unit 31 prepares an unlearned learning model (step S2). Various learning models capable of executing semantic segmentation can be used for this learning model. Semantic segmentation is generally used to analyze an image composed of pixel data distributed two-dimensionally, but here, it is applied to the analysis of a chromatogram waveform composed of data arranged one-dimensionally along the time axis. Here, U-Net (see Non-Patent Document 2) is used as a learning model capable of executing semantic segmentation, but other learning models such as SeGNet and PSPNet may also be used.

[0033] Next, the learning execution unit 31 reads the learning data (partial waveform data and characteristic information) from the learning data storage unit 30 (step S3). The learning execution unit 31 performs machine learning using the read learning data, and constructs a learning model for estimating to which region a given partial waveform corresponds (step S4). The learning procedure will not be described in detail here, but a trained model can be constructed according to the procedure described in Patent Document 1, for example.

[0034] The model storage unit 32 stores the trained model created by machine learning using a large amount of training data (step S5). The trained model stored in the model storage unit 32 is transmitted to the trained model storage unit 212 in the LC system A via, for example, a communication line and stored therein.

[0035] [Peak detection processing for the waveform to be analyzed] Next, a process for detecting peaks on a chromatogram waveform obtained for a target sample, which is executed in the data analysis unit 2 in the LC system A, will be described. Fig. 3 is a flow chart showing an outline of the flow of the peak detection process performed in the peak detection processing unit 21.

[0036] First, the waveform preprocessing unit 210 reads the chromatogram waveform data to be analyzed from the data collecting unit 20 (step S11). The waveform preprocessing unit 210 normalizes the signal intensity of the read data and visualizes it, and further divides the visualized chromatogram waveform into a predetermined number of partial waveforms in the horizontal axis (time axis) direction (step S12). This number of divisions may be the same as the number of divisions in the training data, but may be a different number as long as the width of the partial waveform is smaller than the peak width.

[0037] Next, the determination unit 211 reads out the trained model from the trained model storage unit 212 and sequentially inputs the partial waveforms to the trained model. The trained model determines whether the input partial waveform corresponds to each of the seven types of regions, namely, a single peak region, a tailing processed peak region, a completely separated peak region, a vertically divided peak region, a peak start region, a peak end region, and a non-peak region. Specifically, the determination unit 211 uses the trained model to calculate probability information indicating the possibility that each partial waveform corresponds to each region as a numerical value (step S13). The higher the numerical value of this probability, the more likely the partial waveform is to be that region. In this way, the determination unit 211 outputs all partial waveforms constituting the input chromatogram waveform with the probability information for each region, such as a single peak region and a peak start region, added thereto.

[0038] The region determining unit 213 receives the output from the judging unit 211, judges the region showing the highest probability for each partial waveform as the region corresponding to that partial waveform, and determines the type of region corresponding to each partial waveform (step S14). In this way, all partial waveforms constituting the entire chromatogram waveform are classified into one of the regions.

[0039] Although peaks on a chromatogram are usually detected quite accurately by judgment using a trained model, a multimodal peak in which a peak derived from one component (compound) is split into multiple peaks appears similar to a superimposed peak, and therefore the multimodal peak may be erroneously judged to be a superimposed peak. In order to improve the detection accuracy of the peak, the system of this embodiment has a function of judging whether or not a peak is a multimodal peak as a post-processing of the judgment described above using machine learning, and if a peak is a multimodal peak, integrating the peaks to correct it into a single peak.

[0040] Multimodal peaks tend to appear, for example, when a mass spectrometer is used as a detector and the component concentration in the sample is low. This is because when the component concentration is low, the number of ions generated in the ion source of the mass spectrometer is originally small, and the influence of fluctuations in the ion generation efficiency, passing efficiency, and detection efficiency in the mass spectrometer tend to be prominent in the fluctuation of the detection signal. In addition, the peak waveform shape tends to be a shape in which a part of a peak that is originally one peak is missing (concave). Therefore, there are few cases in which the peak is erroneously determined to be a tailing processing peak region or a completely separated peak region in steps S13 and S14, and in many cases it tends to be erroneously determined to be a vertically divided peak region.

[0041] First, the multimodal peak candidate extraction unit 214 extracts peak portions (usually multiple continuous partial waveforms) determined to be vertically divided peak regions in response to the region determination result by the region determination unit 213. Furthermore, for each of the extracted peak portions, it determines whether the height (height from the baseline to the peak top in the vertical division) of a shoulder peak (a peak that sits on the bottom of another peak with a larger crest value) is equal to or less than a predetermined threshold, and extracts peak portions including shoulder peaks that are equal to or less than the threshold as multimodal peak candidates (step S15). The latter is for the purpose of extracting peaks corresponding to components with a relatively low concentration.

[0042] Next, the multimodal peak determining unit 215 determines whether each of the multimodal peak candidates satisfies all of the following three conditions, and determines that a peak that satisfies the three conditions is a multimodal peak (step S16). <Condition 1> The height Hs of a shoulder peak Ps in one multimodal peak is equal to or less than a certain ratio T of the height Hm of the highest main peak Pm (see FIG. 4). Here, the ratio T can be determined appropriately, but can be set in the range of about 60 to 95%, for example, to 90%.

[0043] <Condition 2> The depth of the valley between two peaks adjacent in the time direction, i.e. on the horizontal axis, is equal to or less than a certain ratio of the height of one of the peaks. Here, as shown in Fig. 5, it is determined whether the depth V of the valley between the main peak Pm and the shoulder peak Ps (the difference in height between the bottom of the valley and the peak top of the shoulder peak Ps) is equal to or less than a certain ratio M of the height Hs of the shoulder peak Ps. This ratio M can also be determined appropriately, but can be set in the range of, for example, about 5 to 30%, and can be set to 10%, for example.

[0044] <Condition 3> The number of data points obtained during the time W between the bottom of the valley and the top of one of the peaks (here, the shoulder peak Ps) in the above condition 2 is equal to or less than a predetermined threshold value N (see FIG. 6). This threshold value N can also be determined appropriately, but can be set in the range of, for example, about 3 to 10, and can be set to 5 as an example.

[0045] When the detector is a mass spectrometer, as described above, the multimodal peak often appears due to a temporary decrease in the number of ions detected due to fluctuations in the ion generation efficiency, passage efficiency, etc. in the device. Therefore, in a multimodal peak, it is rare for the heights of the multiple peaks to be approximately the same. Therefore, by imposing condition 1, it is possible to extract peaks with a certain degree of difference in height between the multiple peaks. In most cases, the drop in signal intensity between the multiple peaks is small in a multimodal peak. Therefore, by imposing condition 2, it is possible to extract peaks with a small drop in signal intensity between the multiple peaks. In addition, the drop in signal intensity due to the above factors usually occurs suddenly and is quickly restored. Therefore, by imposing condition 3, it is possible to extract peaks with a sudden drop in signal intensity between the multiple peaks and a sudden recovery.

[0046] It is acceptable to impose only one or two of the above conditions 1 to 3, but it is preferable to impose all three. In addition, any method other than the one adopted here can be used as long as it can capture the above-mentioned features by using at least one of the peak height, the valley depth between multiple peaks, and the time between the bottom of the valley and the peak top.

[0047] The values ​​of T, M, and N, which are the criteria in the above conditions 1 to 3, may be changed by the user through the input unit 24.

[0048] Next, the multimodal peak integrating unit 216 executes a process of integrating the multiple peaks determined to be multimodal peaks into one peak (step S17). Various methods can be used to integrate multiple peaks.

[0049] FIG. 8 is a diagram showing an example of the integration process. When the peak shown in FIG. 8(A) is determined to be a vertically divided peak in step S14, the regions are usually determined in the order of retention time, as shown in FIG. 8(B), as follows: peak start region → vertically divided peak region → peak end region → peak start region → vertically divided peak region → peak end region. Now, if this peak is determined to be a multimodal peak in step S16, the multimodal peak integration unit 216 deletes the peak start region and peak end region sandwiched between two vertically divided peak regions, and replaces the entire region between the first peak start region and the second peak end region with a single peak region (see FIG. 8(C)). As a result, although the peak shape itself looks like a superimposed peak, this peak is treated as a single peak, and a user who sees this can recognize that this is a multimodal peak derived from one component.

[0050] Of course, in the peak integration process, in addition to changing the region, the peak waveform may be shaped by appropriate waveform processing such as smoothing processing as shown in FIG.

[0051] After the multimodal peaks are integrated as necessary, the display processor 23 displays the peak detection results by the peak detection processor 21 on the screen of the display unit 25 (step S18). If the setting is such that the qualitative analysis is automatically performed based on the peak detection results, the qualitative and quantitative analysis unit 22 obtains, for example, the retention time of the peak top for each detected peak, and identifies the component corresponding to the peak based on the retention time. If the setting is such that the quantitative analysis is automatically performed based on the peak detection results, the qualitative and quantitative analysis unit 22 obtains the peak area value or height value for each detected peak, and calculates the concentration (content) of the component corresponding to the peak by referring to the area value or height value against a calibration curve created in advance. The display processor 23 displays the results of the qualitative or quantitative analysis on the screen of the display unit 25 together with the peak detection results.

[0052] As described above, in the LC system of this embodiment, when a multimodal peak is erroneously determined to be a superimposed peak among peaks automatically detected using machine learning, this can be detected, and peak information, etc. can be appropriately corrected as necessary and provided to the user.

[0053] In the above description, the integration process is automatically performed on peaks determined to be multimodal, but there may be cases where the user wants to confirm whether a peak is truly multimodal and then perform integration process or other processes as necessary. Therefore, instead of performing the integration process automatically, when the multimodal peak determination unit 215 determines that a peak is a multimodal peak, the peak waveform may first be displayed on the display unit 25 to notify the user. Then, according to an instruction from the user who has confirmed the result, the peak integration process may be performed, or the result of the determination that the peak is a multimodal peak may be deleted and the peak may be treated as a superimposed peak.

[0054] Next, some preferred additional configurations will be described. The configurations described below can be appropriately combined with those described in the above embodiment. In addition, multiple additional configurations can be combined as long as they do not perform contradictory processes.

[0055] For example, in the above embodiment, for a peak having multiple apexes that is determined to be a vertically divided peak, which is one of the superimposed peaks, it is determined whether or not it is a multimodal peak based on information reflecting the waveform shape, such as the peak height, the depth of the valley between the multiple peaks, or the width between the bottom of the valley and the peak, etc. It may also be possible to further refer to other information to determine whether or not it is a multimodal peak. Specifically, numerical information reflecting the shape and characteristics of the peak, such as the signal-to-noise ratio, resolution, symmetry coefficient, area, or peak width in the vertically divided peak region, may be calculated, and one of the conditions for determining that a peak is a multimodal peak may be whether the numerical value exceeds or falls below a predetermined threshold value.

[0056] For example, as described above, multimodal peaks are likely to appear when the component concentrations are low, and in such cases, even if there is a peak, the S / N ratio is often relatively low. Therefore, it may be possible to determine whether the S / N ratio calculated from the signal intensities in the vertically divided peak region and non-peak region exceeds a predetermined threshold, and if it does, to determine that the peak is not a multimodal peak.

[0057] The symmetry coefficient is an index showing the symmetry of a peak, and for example, if it is greater than 1, it is a tailing peak. For a compound whose peaks are known to have poor symmetry, if the symmetry coefficient of the detected peak is high, it is assumed that the unresolved peaks that occur in the tailing part of the peak are likely to be peaks derived from one component rather than other components. Therefore, it is possible to use the symmetry coefficient as a criterion for determining that such peaks are multimodal peaks and integrating them.

[0058] Mass spectrometers can usually observe multiple types of ions (called quantitative ions and confirmatory ions) with different m / z values ​​generated from one component. Therefore, in GC-MS and LC-MS, it is possible to create an extracted ion chromatogram of the quantitative ions and one or more confirmed ions for the same component. Since these chromatograms are derived from the same component, the waveforms should be similar. Therefore, it is also possible to compare the area estimation results in the chromatograms of the quantitative ions and the confirmatory ions for the same component, and use the comparison results to determine whether a peak has multiple peaks.

[0059] For example, even if a peak is determined to be a multimodal peak in either the peak determination result for the chromatogram of the quantitative ion or the peak determination result for the chromatogram of the confirmation ion of the same component, if the peak is determined not to be a multimodal peak in the other result, the result determined to be a multimodal peak may be corrected.

[0060] In addition, compound information on the target compound in the sample that can be known in advance can be additionally used as a condition for multimodal peak judgment. The compound information here includes, for example, the concentration of the compound, as well as structural information such as whether or not the compound has an isomer, and information such as whether or not derivatives generated in the pretreatment process are present.

[0061] In a mass spectrometer, some compounds are easily ionized and some are not, depending on the characteristics of the compound, so even if the concentration is the same, some compounds may have a multi-modal peak and some may not, depending on the type of compound. Therefore, it is possible to add a process for changing the criteria for determining whether a peak has a multi-modal shape depending on the type of compound.

[0062] Furthermore, when the presence of isomers is known from compound information, even if the isomers have the same molecular weight, the peaks on the chromatogram often shift in time due to differences in structure, etc. Therefore, it is possible to determine that the peak presumed to correspond to the target compound is unlikely to be a multimodal peak.

[0063] Furthermore, when smoothing processing based on a predetermined algorithm is performed, a clear difference may occur in the magnitude and appearance of the difference in waveform shape before and after smoothing processing between a multimodal peak and an original vertically divided peak (and other overlapping peaks). Therefore, it is possible to use the magnitude and appearance of this difference in the judgment of a multimodal peak.

[0064] In addition, the waveform analysis process for the above-mentioned multimodal peaks can be performed not only on chromatograms obtained by measuring unknown samples, but also on chromatograms obtained by measuring standard samples with known concentrations or blank samples that do not contain the target component.

[0065] In this way, the LC system of this embodiment can improve the accuracy of multimodal peak determination by combining not only information reflecting the waveform shape of the acquired peak, but also various other additional information, thereby providing the user with more accurate peak information.

[0066] The above-mentioned embodiment and modified examples are merely examples and can be modified as appropriate in accordance with the spirit of the present invention. In the above-mentioned embodiment, the waveform analyzer for detecting peaks is incorporated into the same measurement system A as the measurement unit, but it can be configured as a waveform analyzer independent of the LC system A. In that case, it is sufficient to read the chromatogram data previously acquired by the LC measurement unit 1 and perform the analysis.

[0067] Furthermore, while the above embodiment has been described with reference to an example in which a chromatogram waveform is processed, it is clear that the present invention can be applied to the waveform analysis of signal waveforms whose signal intensity can change in response to changes in the value of a specified parameter, which are obtained by various analytical instruments, such as electropherograms obtained by an electrophoresis instrument, mass spectra (profile spectra) obtained by a mass spectrometer, spectroscopic spectra obtained by a spectrometer, fluorescence spectra obtained by a fluorescence measuring instrument, and X-ray intensity spectra obtained by an X-ray analyzer.

[0068] [Various aspects] It will be appreciated by those skilled in the art that the exemplary embodiments described above are examples of the following aspects.

[0069] (Item 1) One aspect of the waveform analysis method according to the present invention is a waveform analysis method for analyzing a signal waveform that is a chromatogram or a spectrum, comprising: a model creation step of creating a trained model for identifying peak portions included in an input waveform by machine learning using a plurality of sets of partial waveforms created by dividing a reference waveform in which the positions of the peak portions are known; a region estimation step of dividing the waveform to be analyzed into a plurality of partial waveforms, determining whether each of the plurality of partial waveforms is a peak portion using the trained model, and estimating a plurality of different types of regions in the waveform to be analyzed, including a single peak region, a superimposed peak region, and a non-peak region, based on the result of the determination; a multimodal determination step for determining whether or not a superimposed peak included in the region estimated to be a superimposed peak region in the region estimation step is a multimodal peak due to one component, using at least one piece of information of the height of one or more of the multiple peaks, the depth of a valley between two adjacent peaks among the multiple peaks, or the width in the horizontal axis direction between the bottom of the valley and the top of one of the peaks sandwiching the valley; has.

[0070] (Item 11) One aspect of the waveform analysis device according to the present invention is a waveform analysis device that analyzes a signal waveform that is a chromatogram or a spectrum, comprising: a region estimation unit that divides a waveform to be analyzed into a plurality of partial waveforms, and determines whether each of the plurality of partial waveforms of the waveform to be analyzed is a peak portion using a trained model created by machine learning using a plurality of sets of the plurality of partial waveforms created by dividing a reference waveform whose peak portion positions are known, and estimates a plurality of different types of regions in the waveform to be analyzed, including a single peak region, a superimposed peak region, and a non-peak region, based on the result of the determination; a multimodal determination unit that determines whether or not a superimposed peak included in a region estimated by the region estimation unit to be a superimposed peak region is a multimodal peak due to one component, using at least one piece of information of the height of one or more of the multiple peaks, the depth of a valley between two adjacent peaks among the multiple peaks, or the width in the horizontal axis direction between the bottom of the valley and the top of one of the peaks sandwiching the valley; Equipped with.

[0071] According to the waveform analysis method described in paragraph 1 and the waveform analyzer described in paragraph 11, even if a peak detection process using machine learning determines that a superimposed peak is a multimodal peak resulting from overlapping peaks derived from different components, it is possible to determine whether the superimposed peak is a multimodal peak derived from one component or not by waveform processing performed after the peak detection process, and to inform the user of the determination result or to correct the estimated result that the peak is a multimodal peak without manual labor. This makes it possible to improve the accuracy of automated peak detection, including multimodal peaks that tend to appear when, for example, component concentrations are low.

[0072] (Item 2) The waveform analysis method described in item 1 may further include an integration step of integrating a multimodal peak that is determined to be a multimodal peak in the multimodal determination step so that the multimodal peak can be treated as a single peak.

[0073] (Item 12) The waveform analysis device described in item 11 may further include an integration unit that, when the multimodal determination unit determines that a multimodal peak is a multimodal peak, integrates the multimodal peak so that the multimodal peak can be treated as a single peak.

[0074] According to the waveform analysis method described in paragraph 2 and the waveform analysis device described in paragraph 12, even if a peak is erroneously determined to be a superimposed peak by peak detection using machine learning, the peak information can be provided to the user as if it were an actual single peak.

[0075] (Item 3) In the waveform analysis method according to item 2, execution of integration in the integration step may be switched on or off in response to a user's selection.

[0076] (Item 13) The waveform analysis device according to item 12 may further include an operation unit that receives a selection operation by a user to switch between on and off execution of integration by the integration unit.

[0077] According to the waveform analysis method described in paragraph 3 and the waveform analysis device described in paragraph 13, the user can decide whether or not to correct the estimation result using the trained model for a peak determined to be a multimodal peak. This allows the user to visually check the waveform shape of a peak determined to be a multimodal peak, and further to determine whether the determination that the peak is a multimodal peak is correct by taking into account various other information, and as a result, to integrate the multimodal peak or to leave it as it is without integrating it.

[0078] (Item 4) In the waveform analysis method described in any one of items 1 to 3, the superimposed peak region is subdivided into multiple types of regions including a vertically divided peak region depending on a method of dividing the superimposed peak, and in the multimodal determination step, it may be determined whether or not a peak corresponding to a vertically divided peak region is a multimodal peak.

[0079] (Item 14) In the waveform analysis device described in any one of items 11 to 13, the superimposed peak region may be subdivided into a plurality of types of regions including a vertically divided peak region depending on a method for dividing the superimposed peak, and the multimodal determination unit may determine whether or not a peak corresponding to a vertically divided peak region is a multimodal peak.

[0080] The method of dividing the overlapping peaks includes, in addition to vertical division, tailing processing, complete separation, etc. A multimodal peak resulting from a peak derived from one component being divided into multiple peaks is likely to be erroneously determined as a vertically divided peak because a shoulder peak with a relatively high signal intensity is observed near a main peak with the maximum signal intensity. According to the waveform analysis method described in paragraph 4 and the waveform analyzer described in paragraph 14, whether or not a peak is a multimodal peak can be determined after excluding overlapping peaks that are unlikely to be multimodal peaks, thereby improving the accuracy of the determination of a multimodal peak.

[0081] (Item 5) In the waveform analysis method described in any one of items 1 to 4, in the multimodal determination step, one of the conditions for a multimodal peak may be that the height of a main peak or a shoulder peak is equal to or less than a predetermined threshold value.

[0082] (Item 15) In the waveform analysis device described in any one of items 11 to 14, the multimodal determination section may determine that one of the conditions for a peak to be a multimodal peak is that the height of a main peak or a shoulder peak is equal to or less than a predetermined threshold value.

[0083] Moreover, multimodal peaks are likely to occur when the concentration of components contained in a sample is relatively low, that is, when the height of the peaks in a chromatogram is low. According to the waveform analysis method described in paragraph 5 and the waveform analyzer described in paragraph 15, the determination of multimodal peaks is performed after narrowing down to components with relatively low concentrations, so that the accuracy of the determination of multimodal peaks can be further improved.

[0084] (Item 6) In the waveform analysis method described in any one of items 1 to 5, the multimodal peak determination step may determine whether or not a peak is multimodal using at least one of the following: the height ratio between the main peak and a shoulder peak; the ratio between the depth of a valley between two adjacent peaks and the height of one of the peaks; or the width between the bottom of the valley and the top of one of the peaks.

[0085] (Item 16) In the waveform analysis device described in any one of items 11 to 15, the multimodal determination unit may determine whether a peak is multimodal using at least one of the following: the height ratio between a main peak and a shoulder peak; the ratio between the depth of a valley between two adjacent peaks and the height of one of the peaks; or the width between the bottom of the valley and the top of one of the peaks.

[0086] According to the waveform analysis method described in item 6 and the waveform analysis device described in item 16, multi-modal peaks can be recognized more accurately.

[0087] (Item 7) In the waveform analysis method described in any one of items 1 to 6, the multimodal determination step may calculate one or more of the signal-to-noise ratio, resolution, symmetry coefficient, area, or peak width in the overlapping peak region, and the calculation results may be used in conjunction with the determination of the multimodal peak.

[0088] (Item 17) In the waveform analysis device described in any one of items 11 to 16, the multimodal judgment unit may calculate one or more of the S / N ratio, resolution, symmetry coefficient, area, or peak width in the overlapping peak region, and use the calculation results in conjunction with the judgment of the multimodal peak.

[0089] According to the waveform analysis method described in paragraph 7 and the waveform analysis device described in paragraph 17, the accuracy of determining multi-modal peaks can be improved by utilizing information reflecting characteristics of the waveform shape other than the peak heights and valley depths.

[0090] (Item 8) In the waveform analysis method described in any one of items 1 to 7, the signal waveform may be a chromatogram obtained by chromatography mass spectrometry, and in the multimodal determination step, a multimodal peak may be determined by utilizing both a determination result for a chromatogram of a quantitative ion for the same component and a determination result for a chromatogram of a confirmation ion.

[0091] (Item 18) In the waveform analysis device described in any one of items 11 to 17, the signal waveform may be a chromatogram obtained by chromatography mass spectrometry, and the multimodal judgment unit may judge a multimodal peak by utilizing both the judgment result for a chromatogram of a quantitative ion for the same component and the judgment result for a chromatogram of a confirmation ion.

[0092] Quantitative ions and confirmatory ions derived from the same component should appear as peaks of roughly similar shapes in extracted ion chromatograms. Therefore, according to the waveform analysis method described in paragraph 8 and the waveform analyzer described in paragraph 18, even if a peak is not properly detected in one chromatogram due to some factor, highly accurate peak information can be obtained by using the peak detection result in the other chromatogram.

[0093] (Item 9) In the waveform analysis method according to any one of items 1 to 8, in the multimodal determination step, compound information on the target compound may be used to determine the multimodal peak.

[0094] (Item 19) In the waveform analyzer according to any one of items 11 to 18, the multimodal determination unit may utilize compound information on the target compound in determining a multimodal peak.

[0095] The "compound information" referred to here can include not only information about the compound itself contained in the sample, such as concentration values, but also information about isomers of the compound, and information about derivatives generated by pretreatment, etc.

[0096] As described above, when the concentration of a component in a sample is low, the peaks derived from the component are likely to have a multimodal shape. Therefore, for example, if the concentration value is known as compound information, the multimodal peak determination is performed only when the concentration value is lower than a predetermined threshold value, thereby avoiding unnecessary multimodal peak determination and improving the accuracy of the multimodal peak determination.

[0097] (Item 10) In the waveform analysis method described in any one of items 1 to 9, the multimodal determination step may utilize the difference between the waveforms before and after a smoothing process is performed on the superimposed peak waveform to determine whether the peak is multimodal.

[0098] (Item 20) In the waveform analysis device described in any one of items 11 to 19, the multimodal determination section may utilize the difference between the waveform before and after a smoothing process is performed on the superimposed peak waveform to determine whether the peak is multimodal.

[0099] According to the waveform analysis method described in paragraph 10 and the waveform analysis device described in paragraph 20, it is possible to identify multi-modal peaks whose shape is changed only slightly by smoothing, thereby making it possible to more accurately distinguish multi-modal peaks.

[0100] (Item 21) Furthermore, one aspect of the analytical device of the present invention can be any one of a chromatograph device, a mass spectrometer, and an optical measurement device, which includes a waveform analysis device described in any one of items 11 to 20 as a data analysis unit.

[0101] According to the analytical device described in paragraph 21, high-precision peak information can be utilized to achieve high qualitative and quantitative performance. [Explanation of symbols]

[0102] 1...LC measurement section 2. Data analysis section 20…Data collection section 21...Peak detection processing unit 210...Waveform preprocessing unit 211…Judgment section 212…Model memory section 213...Region determination section 214...Multimodal peak candidate extraction unit 215...Multimodal peak judgment unit 216…Multimodal peak integration section 22...Qualitative / Quantitative Analysis Department 23...Display processing unit 24...Input section 25...Display section 3. Model Creation Section 30…Learning data storage unit 31…Learning Execution Department 32…Model memory section

Claims

1. A waveform analysis method for analyzing a signal waveform that is a chromatogram or a spectrum, comprising: a model creation step of creating a trained model for identifying peak portions included in an input waveform by machine learning using a plurality of sets of partial waveforms created by dividing a reference waveform in which the positions of the peak portions are known; a region estimation step of dividing the waveform to be analyzed into a plurality of partial waveforms, determining whether each of the plurality of partial waveforms is a peak portion using the trained model, and estimating a plurality of different types of regions in the waveform to be analyzed, including a single peak region, a superimposed peak region, and a non-peak region, based on the result of the determination; a multimodal determination step for determining whether or not a superimposed peak included in the region estimated to be a superimposed peak region in the region estimation step is a multimodal peak due to one component, using at least one piece of information of the height of one or more of the multiple peaks, the depth of a valley between two adjacent peaks among the multiple peaks, or the width in the horizontal axis direction between the bottom of the valley and the top of one of the peaks sandwiching the valley; The waveform analysis method includes the steps of:

2. 2. The waveform analysis method according to claim 1, further comprising an integration step of integrating a multimodal peak, when the multimodal peak is determined to be a multimodal peak in the multimodal determination step, so that the multimodal peak can be treated as a single peak.

3. The waveform analysis method according to claim 2 , further comprising the step of switching between execution and non-execution of integration in the integration step in response to a user's selection.

4. 2. The waveform analysis method according to claim 1, wherein the superimposed peak region is subdivided into a plurality of types of regions including a vertically divided peak region according to a division method of the superimposed peak, and in the multimodal determination step, it is determined whether or not a peak corresponding to a vertically divided peak region is a multimodal peak.

5. 2. The waveform analysis method according to claim 1, wherein in the multimodal determination step, one of the conditions for a multimodal peak is that the height of a main peak or a shoulder peak is equal to or less than a predetermined threshold value.

6. 2. The waveform analysis method according to claim 1, wherein in the multi-modal determination step, the multi-modal peak is determined using at least one of the following: a height ratio between a main peak and a shoulder peak; a ratio between a depth of a valley between two adjacent peaks and a height of one of the peaks; or a width between a bottom of the valley and an apex of one of the peaks.

7. 2. The waveform analysis method according to claim 1, wherein in the multimodal determination step, one or more of an S / N ratio, a resolution, a symmetry coefficient, an area, or a peak width in a superimposed peak region is calculated, and the calculation result is used in conjunction with determining whether or not the peak is multimodal.

8. 2. The waveform analysis method according to claim 1, wherein the signal waveform is a chromatogram obtained by chromatography mass spectrometry, and in the multimodal determination step, a multimodal peak is determined by utilizing both a determination result for a chromatogram of a quantitative ion for the same component and a determination result for a chromatogram of a confirmation ion.

9. 2. The waveform analysis method according to claim 1, wherein in said multimodal determination step, compound information on the target compound is used to determine whether a peak is a multimodal peak.

10. 2. The waveform analysis method according to claim 1, wherein in said multimodal determination step, a difference between waveforms before and after a smoothing process is performed on the superimposed peak waveform is used to determine whether the superimposed peak is a multimodal peak.

11. A waveform analyzer for analyzing a signal waveform that is a chromatogram or a spectrum, comprising: a region estimation unit that divides a waveform to be analyzed into a plurality of partial waveforms, and determines whether each of the plurality of partial waveforms of the waveform to be analyzed is a peak portion using a trained model created by machine learning using a plurality of sets of the plurality of partial waveforms created by dividing a reference waveform whose peak portion positions are known, and estimates a plurality of different types of regions in the waveform to be analyzed, including a single peak region, a superimposed peak region, and a non-peak region, based on the result of the determination; a multimodal determination unit that determines whether or not a superimposed peak included in a region estimated by the region estimation unit to be a superimposed peak region is a multimodal peak due to one component, using at least one piece of information of the height of one or more of the multiple peaks, the depth of a valley between two adjacent peaks among the multiple peaks, or the width in the horizontal axis direction between the bottom of the valley and the top of one of the peaks sandwiching the valley; A waveform analysis device comprising:

12. 12. The waveform analyzer according to claim 11, further comprising an integration unit that, when the multimodal determination unit determines that the multimodal peak is a multimodal peak, integrates the multimodal peak so that the multimodal peak can be treated as a single peak.

13. The waveform analyzing apparatus according to claim 12 , further comprising an operation unit configured to accept a selection operation by a user for switching between execution and non-execution of integration by the integration unit.

14. The overlapping peak region is divided into a plurality of types of regions including a vertically divided peak region according to a method for dividing the overlapping peak, The waveform analyzing device according to claim 11 , wherein the multimodal determination section determines whether or not a peak corresponding to a vertically divided peak region is a multimodal peak.

15. 12. The waveform analyzer according to claim 11, wherein the multimodal determination section determines that a peak is a multimodal peak when the height of a main peak or a shoulder peak is equal to or less than a predetermined threshold value.

16. 12. The waveform analysis device according to claim 11, wherein the multimodal determination unit determines whether a peak is multimodal using at least one of the following: a ratio of the heights of a main peak and a shoulder peak; a ratio of the depth of a valley between two adjacent peaks to the height of one of the peaks; or a width between the bottom of the valley and the top of one of the peaks.

17. 12. The waveform analysis device according to claim 11, wherein the multimodal determination unit calculates one or more of an S / N ratio, a resolution, a symmetry coefficient, an area, or a peak width in a superimposed peak region, and uses the calculation results in combination with the determination of a multimodal peak.

18. 12. The waveform analysis device according to claim 11, wherein the signal waveform is a chromatogram obtained by chromatography mass spectrometry, and the multimodal determination unit determines a multimodal peak by using both a determination result for a chromatogram of a quantitative ion and a determination result for a chromatogram of a confirmation ion for the same component.

19. The waveform analyzer according to claim 11 , wherein the multimodal determination unit uses compound information on the target compound to determine whether a peak is a multimodal peak.

20. 12. The waveform analyzer according to claim 11, wherein the multimodal determination section uses a difference between waveforms before and after a smoothing process is performed on the superimposed peak waveform to determine whether the superimposed peak is a multimodal peak.

21. An analysis device, which is any one of a chromatography device, a mass spectrometer, and a spectrometer, comprising the waveform analysis device according to any one of claims 11 to 20 as a data analysis unit.

Images