Analysis device and waveform processing program for analysis device
By amplifying and reducing the signal waveform and combining it with the learned model to detect peaks, the problem of missing low-intensity peaks is solved, and the peak detection accuracy is improved efficiently.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SHIMADZU SEISAKUSHO LTD
- Filing Date
- 2021-04-30
- Publication Date
- 2026-05-26
Smart Images
Figure CN117242345B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to an analysis apparatus and a program for processing signal waveforms obtained through analysis on a computer. Background Technology
[0002] In a liquid chromatography (LC) or gas chromatography (GC) system, a sample is introduced into a chromatographic column. During the sample's passage through the column, the various components contained in the sample are separated over time, and detected by a detector located at the column outlet. Peaks corresponding to the various components in the sample appear in the chromatographic waveform generated based on the detection signal obtained by the detector. Since the time at which the peak is observed (retention time) corresponds to the type of component, the component can be identified based on the retention time, enabling qualitative analysis. Furthermore, since the peak height or area corresponds to the concentration or content of the component, the concentration or content of the component can be determined based on the peak height or area value, enabling quantitative analysis.
[0003] For qualitative or quantitative analysis, it is necessary to accurately detect peaks on the chromatographic waveform and determine the start and end points (retention times) of the peaks. In the chromatographic waveform obtained through analysis, various noises overlap, baselines fluctuate, or peaks originating from multiple components overlap. Therefore, accurately detecting peaks from the chromatographic waveform is not always easy, and various algorithms have been used in the past to detect peaks using computers. As one such peak detection method, in recent years, methods utilizing AI (artificial intelligence) technologies, represented by deep learning, have been put into practical use (see Non-Patent Literature 1).
[0004] The peak detection method described in Non-Patent Document 1 utilizes a learned model generated through so-called supervised machine learning. During the generation of such a learned model, i.e., during the model's learning process, a large amount of teacher data is typically used, consisting of chromatographic waveforms collected by the actual analytical apparatus and information on peaks (start and end points) detected by a skilled operator. One of the major advantages of this AI-based peak detection method is that, unlike many other peak detection methods, it does not require the operator to pre-set cumbersome detection conditions (parameters) or make judgments during the peak detection process.
[0005] Existing technical documents
[0006] Non-patent literature
[0007] Non-Patent Document 1: "Software for Peakintelligence™ Ver.2 LabSolutions Insight™ Waveform Processing Options", [Online], [Searched April 7, 2021], Shimadzu Corporation, Internet<URL:https: / / www.an.shimadzu.co.jp / lcms / m_package / peakintelligence.htm> Summary of the Invention
[0008] The technical problem that the invention aims to solve
[0009] To improve the accuracy of learned models generated in supervised machine learning, such as deep learning, an effective approach is to utilize as much biased (variable) teacher data as possible in the learning process, encompassing the features of the target object. However, collecting teacher data is extremely time-consuming and labor-intensive. Furthermore, as the amount of teacher data increases, the processing time for generating the learned model also increases dramatically, thus raising the required cost. Consequently, the amount of prepared teacher data is limited, and when peak detection is performed using a learned model generated from such teacher data, peaks that should have been detected in the actual chromatographic waveform may sometimes be missed.
[0010] This invention was made to solve such a technical problem, and its purpose is to provide an analysis device and a waveform processing program for the analysis device, which can effectively detect peaks that were previously easily missed by using existing learned models.
[0011] Solution to the above technical problems
[0012] One solution of the analytical apparatus of the present invention, which is made to solve the above-mentioned technical problems, comprises:
[0013] The waveform deformation section amplifies or reduces the target signal waveform, obtained through analysis, which shows the change in signal strength corresponding to the change in the value of a specified parameter, by a factor of N (where N is a positive value other than 0 and 1) in the direction of the signal strength axis, and / or amplifies or reduces it by a factor of M (where M is a positive value other than 0 and 1, and may be the same value as N) in the direction of the axis of the specified parameter.
[0014] The peak detection unit uses a learned model, takes the signal waveform deformed by the waveform deformation unit as input, and outputs the start and end points of the peak as detection results. The learned model is pre-generated by using the signal waveform and the start and end points of the peak as teacher data for machine learning.
[0015] The waveform inverse deformation section reduces or amplifies the output of the peak detection section, i.e., the information of the start and end points of the peak, by a factor opposite to that of the deformation performed by the waveform deformation section, and calculates the peak detection result of the signal waveform with respect to the target.
[0016] Furthermore, one solution of the waveform processing program for the analysis apparatus of the present invention, which was completed to solve the above-mentioned technical problems, is a program for processing a target signal waveform obtained through analysis and showing a change in signal intensity corresponding to a change in the value of a specified parameter on a computer, causing the computer to execute:
[0017] The waveform deformation step involves magnifying or reducing the signal waveform of the target by a factor of N (where N is a positive value other than 0 and 1) in the direction of the signal strength axis, and / or magnifying or reducing it by a factor of M (where M is a positive value other than 0 and 1, and can be the same value as N) in the direction of the axis of the specified parameter.
[0018] In the peak detection step, a learned model is used. The deformed signal waveform from the waveform deformation step is taken as input, and the start and end points of the peak are taken as the detection results. The learned model is pre-generated by using the signal waveform and the start and end points of the peak as teacher data for machine learning.
[0019] The waveform inverse deformation step involves reducing or increasing the output of the peak detection step, i.e., the information of the start and end points of the peak, by a factor opposite to that used in the waveform deformation step, to obtain the peak detection result of the signal waveform for the target.
[0020] Invention Effects
[0021] According to the above-described scheme of the analytical apparatus and waveform processing program for the analytical apparatus of the present invention, using an existing learned model, it is possible to effectively detect peak waveforms that might have been missed in the past due to low peak signal intensity, such as in chromatographic waveforms. Therefore, it is not necessary to regenerate the learned model, thus improving the accuracy of peak detection. Attached Figure Description
[0022] Figure 1 This is a schematic diagram of an LC device according to an embodiment of the present invention.
[0023] Figure 2 It is generated Figure 1 The diagram shows a schematic configuration of the model generation device for the learned model used by the LC device.
[0024] Figure 3 This is a diagram showing an example of a chromatographic waveform.
[0025] Figure 4This is an explanatory diagram of the peak detection processing steps in the LC device of this embodiment.
[0026] Figure 5 This is an explanatory diagram illustrating the peak detection and processing steps in conventional LC devices. Detailed Implementation
[0027] The analytical apparatus of the present invention may include, for example, a liquid chromatograph (LC) apparatus including a liquid chromatograph mass analysis device, a gas chromatograph (GC) apparatus including a gas chromatograph mass analysis device, a spectrophotometer (infrared absorption spectrophotometer, visible ultraviolet spectrophotometer, fluorescence spectrophotometer, etc.), an X-ray analysis device (fluorescence X-ray analysis device, X-ray diffraction analysis device, etc.), etc.
[0028] Furthermore, the waveform processing program for the analysis device of the present invention can be provided to the user by storing it on a computer-readable non-transitory recording medium such as a CD-ROM, DVD-ROM, memory card, or USB memory (dongle). Additionally, the program can also be provided to the user via data transmission through communication lines such as the Internet. Moreover, the program can be pre-installed on the computer (strictly speaking, the storage device that is part of the computer) at the time the user purchases the system.
[0029] Furthermore, in this invention, as long as the machine learning is supervised machine learning, there is no particular limitation on its type; deep learning can generally be used.
[0030] Hereinafter, an LC apparatus according to one embodiment of the analytical apparatus of the present invention will be described with reference to the accompanying drawings.
[0031] [Configuration of the apparatus in this embodiment]
[0032] Figure 1 This is a schematic diagram of the LC device according to this embodiment. Figure 2 It is generated Figure 1 The diagram shows a schematic configuration of the learning-completed model generation device used by the LC device.
[0033] like Figure 1 As shown, the LC apparatus 1 of this embodiment includes an LC measurement unit 10, a data analysis unit 11, an input unit 12, and a display unit 13. Although not shown, the LC measurement unit 10 includes a mobile phase container, a delivery pump, a syringe, a chromatographic column, a column oven, a detector, etc., performs LC analysis on the provided sample, and outputs chromatographic data showing the time-varying changes in signal intensity obtained by the detector. The type or method of the detector is not particularly limited; for example, a mass analysis device, a photodiode array (PDA) detector, a fluorescence detector, etc., can be used.
[0034] The data analysis unit 11 includes functional blocks such as a data collection unit 110, a waveform selection unit 111, a waveform amplification unit 112, a peak detection unit 113, a waveform reduction unit 115, a qualitative and quantitative analysis unit 116, and a result display and processing unit 117. The peak detection unit 113 includes a learned model storage unit 114, which stores a learned model generated through pre-learning. This learned model is essentially a peak recognizer for identifying peak waveforms.
[0035] The entity of the data parsing unit 11 is a personal computer or a higher-performance workstation with prescribed software installed, or a computer system including a high-performance computer connected to the aforementioned computer via a communication line. That is, the functions of each functional block included in the data parsing unit 11 can be achieved by executing software mounted on a single computer or a computer system including multiple computers, and can be realized through processing using various data stored in that computer or computer system. Of course, by replacing a portion of the functions of the data parsing unit 11 with dedicated hardware circuitry such as a digital signal processor, high-speed processing can also be achieved.
[0036] like Figure 2 As shown, the model generation device 2, which is separate from the LC device 1, includes a teacher data input unit 20, a learning execution unit 21, and a model construction unit 22 as functional blocks. The learned model generated in this model generation device 2 is stored in the learned model storage unit 114 in the data parsing unit 11 of the LC device 1.
[0037] The physical entity of the model generation device 2 is a high-performance computer with the prescribed software installed. As is well known, learning models based on machine learning algorithms such as deep learning generally requires a massive amount of computation; therefore, a high-performance computer capable of high-speed computation is used. Thus, in most cases, the model learning operation is performed by the manufacturer of the LC device 1 or a company that develops the software. In this case, the model generation device 2 itself is located at such a manufacturer or company. Of course, if possible, the data parsing unit 11 of the LC device 1 can also have such model learning functionality.
[0038] [Steps to generate a learned model]
[0039] First, the actions taken when generating the learned model in the model generation device 2 will be explained.
[0040] When generating the learned model, a large amount of diverse chromatographic waveform data is prepared, and the retention times of the start and end points of one or more peaks appearing in each chromatographic waveform are accurately determined in advance. The diverse chromatographic waveform data mentioned here refers to the chromatographic waveforms that may appear in the chromatographic waveforms during actual peak detection, including various noise contamination, baseline changes (drift), overlap of multiple peaks, or deformation of peak shapes.
[0041] Figure 3 This is an example of a chromatographic waveform acquired by an LC apparatus. In this example, the starting and ending points are observed to be (t... s1 t e1 ), (t s2 t e2 ), (t s3 t e3 The three peaks of the chromatogram. The teacher data input unit 20 reads in a large amount of teacher data grouped together with such chromatographic waveform data and accurate peak information including the peak start and end points. Alternatively, the chromatographic waveform may not be... Figure 3 Instead of showing the entire waveform from the start time to the end time of the measurement, we select... Figure 3 The waveform of the peaks shown by the dotted lines.
[0042] The learning execution unit 21, for example, visualizes the chromatographic waveform data, enabling a model based on a prescribed method to learn the image and peak information as the positive solution. Based on this learning result, the model building unit 22 constructs a learning model for estimating the start and end points of peaks on the chromatographic waveform. As is well known, machine learning has various algorithms; deep learning, as one of the general object detection algorithms in image recognition, is used here, but the machine learning methods that can be used are not limited to this.
[0043] The model building unit 22 temporarily stores the learned model generated by learning using a large amount of teacher data. The learned model generated in the model generation device 2 as described above is transmitted and stored in the learned model storage unit 114 of the LC device 1, for example, via a communication line or via a storage medium.
[0044] [Peak detection processing for measured waveforms]
[0045] Next, refer to Figure 4 The operation of peak detection processing in LC device 1 is explained.
[0046] The LC measurement unit 10 performs LC analysis on the target sample. In the data analysis unit 11, the data collection unit 110 acquires and temporarily stores the chromatographic data obtained by the LC measurement unit 10. Here, it is assumed that the LC analysis yielded... Figure 4 The chromatographic waveform shown in (a) is as follows.
[0047] If an instruction to perform peak detection processing is issued from the input unit 12, the waveform selection unit 111 receives the chromatographic data of the object to be processed from the data collection unit 110. Then, the waveform selection unit 111 determines the retention time range of the object to be waveform amplified according to a predetermined standard.
[0048] Specifically, for example, by calculating the difference between the minimum and maximum signal intensity in the chromatographic data of the object being processed, and if this difference is below a predetermined threshold, the entire chromatogram (i.e., the entire retention time range) can be treated as a waveform amplification object. Furthermore, positions with small signal intensity changes along the time axis (e.g., Figure 4 Using position U in (a) as the boundary, the chromatogram is divided into multiple segments along the time axis. Each segmented waveform is then judged to be within the retention time range of the waveform amplification object according to the criteria described above.
[0049] The purpose of determining the retention time range for waveform amplification in the waveform selection unit 111 is to amplify the waveform only within retention time ranges that might be missed due to low peak signal strength, while not amplifying the waveform within retention time ranges that are impossible to miss due to high peak signal strength. This method is not limited to what has been described above, as long as the waveform shape can be determined in accordance with this objective. Furthermore, if the signal strength is high and amplification further increases the signal strength without causing substantial problems, the waveform selection unit 111 can be omitted, and the entire retention time range can be used as the target for waveform amplification.
[0050] exist Figure 4 In example (a), it is assumed that the entire chromatographic waveform is considered as the object of waveform magnification. Then, as... Figure 4 As shown in (b), the waveform amplification unit 112 amplifies the signal intensity by a factor of N (where N is a value greater than 1) (N times in the vertical direction) for the waveform contained within the retention time range to which the waveform amplification is intended. Two peaks, P1 and P2, present on the original chromatographic waveform (however, they were not detected as peaks at this time point) correspond to peaks P10 and P20, respectively, on the amplified chromatographic waveform. Furthermore, the value of N can be a predetermined value or can be set to be appropriately selected by the user.
[0051] The peak detection unit 113 inputs a chromatographic waveform with amplified signal intensity as described above into the learned model stored in the learned model storage unit 114, and receives peak information, including the start and end points of the peaks, as output. This peak information is information about peaks automatically detected by the peak recognizer in the input chromatographic waveform. Figure 4 (c) is shown for Figure 4The graph shown in (b) is a diagram of the automatically detected peak initiation and termination positions, and the dashed lines formed by connecting the paired peak initiation and termination points with straight lines. Here, the two peaks P10 and P20 are accurately detected.
[0052] After peak detection is performed in this manner, the waveform reduction unit 115 performs the opposite processing to the waveform amplification unit 112. That is, the waveform reduction unit 115 performs a reduction process on the chromatographic waveform encompassed by the retention time range designated for waveform amplification, reducing the signal intensity by a factor of 1 / N (1 / N in the vertical axis direction). Thus, compared to... Figure 4 The peak detection results in the amplified chromatographic waveform shown in (c), i.e., the signal intensities corresponding to the peak start and end points, are as follows: Figure 4 As shown in (d), it decreases along with the chromatographic waveform. The result is that it is possible to obtain [the desired effect]. Figure 4 The peak detection results are essentially the same as those obtained by performing peak detection on the original chromatographic waveform shown in (a).
[0053] Figure 5 This diagram illustrates the process of automatically detecting peaks in a chromatographic waveform that has not been amplified, within the peak detection unit 113. This is equivalent to peak detection processing in conventional LC devices.
[0054] In this case, the signal intensity of peak P2 on the chromatographic waveform is much smaller than that of peak P1. Generally, in machine learning such as deep learning, if appropriately labeled teacher data is not provided during model learning to identify such low-intensity peaks, the resulting learned model will lack the ability to recognize them. Therefore, as... Figure 5 As shown in (b), although the peak P1 with a large signal intensity was detected, the peak P2 with a small signal intensity was missed.
[0055] As mentioned above, while potential distortions or deviations in peak waveforms are considered during teacher data preparation, the larger the volume of teacher data, the greater the cost and time required to generate the learned model. Therefore, due to cost or time constraints, the amount of teacher data often needs to be limited. Furthermore, since the requirements of users of LC devices are diverse, there are sometimes users who also want to detect peaks with low signal strength that were not anticipated during the generation of the learned model. In the latter case, although it is possible to address this by adding new teacher data and regenerating the learned model, this is a very large undertaking and a significant cost burden.
[0056] In contrast, in the LC apparatus 1 of this embodiment, there is no need to change the learned-complete model; that is, using the existing learned-complete model, which lacks the ability to identify peaks with low signal strength, it is possible to detect peaks with low signal strength that have been missed until now. Adding the functions of a waveform selection unit 111, a waveform amplification unit 112, and a waveform reduction unit 115 to the data parsing unit 11 is particularly easy compared to rebuilding the learned-complete model using new teacher data, and requires no cost or time. Thus, with the LC apparatus 1 of this embodiment, the accuracy of peak detection can be improved while saving cost and time.
[0057] In the data analysis unit 11, the qualitative and quantitative analysis unit 116 identifies the components (compounds) corresponding to each peak based on the peak information provided by the peak detection unit 113. Alternatively, it calculates the peak height value or peak area value and calculates a quantitative value as the concentration or content of each component based on this value. The result display processing unit 117 receives the identification results or quantitative results of each component and displays them in a prescribed format on the display unit 13. Furthermore, the result display processing unit 117 may also be configured to display, for example... Figure 4 The chromatographic waveform, including the peak detection results, shown in (d) is displayed on the display unit 13.
[0058] [Variation Example]
[0059] In the LC apparatus 1 of the above embodiment, the waveform amplification unit 112 and the waveform reduction unit 115 amplify and reduce the chromatographic waveform only in the signal intensity axis direction, but they can also be configured to amplify and reduce the chromatographic waveform in the time axis direction. If the chromatographic waveform is amplified only in the signal intensity axis direction, the peak shape becomes elongated. That is, the value of half-width / peak height becomes smaller. With the learning method that uses teacher data during model learning, there may be cases where peaks deformed in the way of decreasing half-width / peak height values cannot be properly detected. Therefore, depending on the characteristics of the learned model used, the chromatographic waveform can be amplified and reduced in the signal intensity axis direction, and also amplified and reduced in the time axis direction.
[0060] In this case, the amplification factor along the signal intensity axis does not need to be the same as the amplification factor along the time axis. That is, as long as the chromatographic waveform is amplified to N times along the signal intensity axis and to M times along the time axis before peak detection (where M is a value greater than 1, and N = M), and after peak detection, the chromatographic waveform is reduced to 1 / N times along the signal intensity axis and to 1 / M times along the time axis, it is sufficient.
[0061] Furthermore, the waveform amplification section 112 and the waveform reduction section 115 can also be configured to amplify and reduce the chromatographic waveform only in the time axis direction, rather than in the signal intensity axis direction. This is useful, for example, in situations where the LC separation conditions differ from the intended state due to the high velocity of the mobile phase in actual LC analysis, and thus the chromatographic waveform is compressed in the time axis direction.
[0062] Furthermore, in the LC apparatus 1 of the above embodiment, the chromatographic waveform is amplified before peak detection and reduced after peak detection. However, depending on the characteristics of the learned model used, it may also be configured to reduce the chromatographic waveform before peak detection and amplify it after peak detection to return to the original chromatographic waveform. Contrary to the above description, this is useful when the signal intensity of the peak in the actual waveform to be processed is too large compared to the range of deviation of the peak signal intensity envisioned during model learning. Furthermore, it is useful when the LC separation conditions differ from the envisioned state due to factors such as a slow mobile phase velocity in actual LC analysis, resulting in a stretched chromatographic waveform in the time axis direction.
[0063] Furthermore, in the LC apparatus 1 of the above embodiment, multiple peaks are detected based on a single peak detection process for the chromatographic waveform. However, it is also possible to detect peaks with significantly different signal intensities as comprehensively as possible by repeatedly performing several peak detection processes, as described below.
[0064] That is, for example, for such Figure 4 For the chromatographic waveform of the object being processed as shown in (a), the first peak detection process does not involve waveform amplification or reduction, but rather peak detection is performed in the same manner as before. In this case, if using Figure 5 As explained, although peak P1 was detected, peak P2 might have been missed. Next, peaks detected in the first peak detection process are removed (e.g., peaks that were missed in the first peak detection process are removed). Figure 5 (The starting and ending points of peak P1 in (b) are connected by a straight line), and then peak detection with accompanying waveform amplification and reduction as described above is performed as a second peak detection process. In this case, because Figure 4 In (b), peak P10 is absent, so only the remaining peak P20 is detected. Since peak P20 has a high signal strength, it is detected without omission. In this way, by summarizing the peak information detected in multiple peak detection processes to generate the final peak detection result, it is possible to achieve peak detection with fewer omissions.
[0065] Of course, it is also possible to further change the amplification and reduction ratios in multiple stages. In addition, by performing different amplification and reduction ratios in the signal intensity axis direction and the time axis direction respectively, peak detection processing can be performed more than 3 times.
[0066] Furthermore, the above-described embodiments are LC devices. In addition to LC devices, the present invention can also be applied to GC devices, and further, it can also be applied to various other analytical devices.
[0067] For example, this invention can obviously also be applied to peak detection on spectral waveforms such as mass spectra obtained by mass analysis devices, optical spectra obtained by various spectrophotometers such as absorption spectrophotometers or fluorescence spectrophotometers, ion mobility spectra obtained by ion mobility analysis devices, and X-ray spectra obtained by X-ray analysis devices.
[0068] Furthermore, not limited to the above-described embodiments or variations, appropriate modifications, alterations, additions, etc., made within the scope of the spirit of this invention are obviously also included within the scope of the claims of this application.
[0069] [Various options]
[0070] Those skilled in the art will understand that the above exemplary embodiments and variations are specific examples of the following solutions.
[0071] (Item 1) One aspect of the analytical apparatus of the present invention comprises:
[0072] The waveform deformation section amplifies or reduces the target signal waveform, obtained through analysis, which shows the change in signal strength corresponding to the change in the value of a specified parameter, by a factor of N (where N is a positive value other than 0 and 1) in the direction of the signal strength axis, and / or amplifies or reduces it by a factor of M (where M is a positive value other than 0 and 1, and may be the same value as N) in the direction of the axis of the specified parameter.
[0073] The peak detection unit uses a learned model, takes the signal waveform deformed by the waveform deformation unit as input, and outputs the start and end points of the peak as detection results. The learned model is pre-generated by using the signal waveform and the start and end points of the peak as teacher data for machine learning.
[0074] The waveform inverse deformation section reduces or amplifies the output of the peak detection section, i.e., the information of the start and end points of the peak, by a factor opposite to that of the deformation performed by the waveform deformation section, and calculates the peak detection result of the signal waveform with respect to the target.
[0075] (Item 6) One aspect of the waveform processing program for the analysis apparatus of the present invention is a program for processing a target signal waveform obtained through analysis, which shows a change in signal strength corresponding to a change in the value of a specified parameter, on a computer, causing the computer to execute:
[0076] The waveform deformation step involves magnifying or reducing the signal waveform of the target by a factor of N (where N is a positive value other than 0 and 1) in the direction of the signal strength axis, and / or magnifying or reducing it by a factor of M (where M is a positive value other than 0 and 1, and can be the same value as N) in the direction of the axis of the specified parameter.
[0077] In the peak detection step, a learned model is used. The deformed signal waveform from the waveform deformation step is taken as input, and the start and end points of the peak are taken as the detection results. The learned model is pre-generated by using the signal waveform and the start and end points of the peak as teacher data for machine learning.
[0078] The waveform inverse deformation step involves reducing or increasing the output of the peak detection step, i.e., the information of the start and end points of the peak, by a factor opposite to that used in the waveform deformation step, to obtain the peak detection result of the signal waveform for the target.
[0079] According to the analytical apparatus described in item 1 and the waveform processing program for the analytical apparatus described in item 6, by using the existing learned model, it is possible to effectively detect peak waveforms that might have been missed in the past due to low peak signal intensity, such as in chromatographic waveforms. Therefore, it is not necessary to regenerate the learned model, thus improving the accuracy of peak detection.
[0080] (Item 2) In the analytical apparatus described in Item 1, it can be configured such that the specified parameter is time and the signal waveform is a chromatographic waveform.
[0081] Especially in chromatographic waveforms obtained by LC devices, baseline drift or incomplete separation of multiple components leading to overlapping peaks can easily occur. Therefore, it is difficult to detect all peaks using a specific algorithm, and parameter adjustment or setting can become cumbersome. In contrast, the analytical apparatus described in item 2 simplifies the troublesome parameter adjustment or setting and detects peaks with high accuracy.
[0082] (Item 3) In the analysis apparatus described in Item 1 or Item 2, the machine learning can be configured to be deep learning.
[0083] There are various methods in machine learning, among which deep learning, particularly in model learning, tends to struggle to fully realize its performance on data with features not included in the teacher's data. In contrast, the analysis apparatus described in item 3, using an existing fully learned model, can detect peaks with high accuracy.
[0084] (Item 4) In any of the analytical apparatuses described in items 1 to 3, the peak information can be configured to include the start and end points of the peak.
[0085] According to the analytical apparatus described in item 4, by determining the start and end points of the peak, the area value of the peak can be accurately calculated, enabling quantitative analysis with high precision.
[0086] (Item 5) In any of the analysis apparatuses described in items 1 to 4, it is possible to further include a determination unit that determines the magnitude of the change in the signal strength level in the signal waveform of the target, and the waveform deformation unit performs processing to amplify or reduce the signal waveform for a range of parameter values that the determination unit determines are less than or greater than a predetermined value.
[0087] In the analysis apparatus described in item 5, it is possible to amplify or reduce only a portion of the signal waveform, such as those with particularly small or large signal strengths, rather than processing the signal waveform corresponding to the entire range of parameter values obtained through analysis. Therefore, the range of peak signal strengths can be narrowed, enabling high-precision peak detection even when using a learned model with a narrow tolerance range for signal strength variations.
[0088] Explanation of reference numerals in the attached figures
[0089] 1 LC device
[0090] 10 LC Measurement Section
[0091] 11 Data Analysis Department
[0092] 110 Data Collection Department
[0093] 111 Waveform Selection Section
[0094] 112 Waveform Amplification Section
[0095] 113 Peak Testing Department
[0096] 114. Learning completed the model storage department.
[0097] 115 waveform reduction section
[0098] 116 Qualitative and Quantitative Analysis Department
[0099] 117 results show processing department
[0100] 12 Input Section
[0101] 13 Display Section
[0102] 2. Model Generation Device
[0103] 20 Teacher Data Input Department
[0104] 21 Learning Implementation Department
[0105] 22. Model Construction Department.
Claims
1. An analytical apparatus, characterized in that, have: The waveform deformation section amplifies or reduces the target signal waveform, which is obtained through analysis and shows the change in signal strength corresponding to the change in the value of the specified parameter, by a factor of N in the direction of the signal strength axis, and / or by a factor of M in the direction of the axis of the specified parameter, where N is a positive value other than 0 and 1, and M is a positive value other than 0 and 1, which may be the same value as N. The peak detection unit uses a learned model, takes the signal waveform deformed by the waveform deformation unit as input, and outputs the start and end points of the peak as detection results. The learned model is pre-generated by using the signal waveform and the start and end points of the peak as teacher data for machine learning. The waveform inverse deformation section reduces or amplifies the output of the peak detection section, i.e., the information of the start and end points of the peak, in the opposite direction to the deformation performed by the waveform deformation section, by a factor of 1 / N in the direction of the signal strength axis, and / or by a factor of 1 / M in the direction of the axis of the specified parameter, to obtain the peak detection result of the signal waveform with respect to the target. The determination unit determines the magnitude of the change in signal strength level in the signal waveform of the target. The waveform deformation unit amplifies or reduces the signal waveform for parameter values that the determination unit determines are less than or greater than a specified value.
2. The analytical apparatus as described in claim 1, characterized in that, The specified parameter is time, and the signal waveform is a chromatographic waveform.
3. The analytical apparatus as described in claim 1, characterized in that, The machine learning mentioned is deep learning.
4. The analytical apparatus as described in claim 1, characterized in that, It also includes a waveform selection unit, which determines the range of parameter values of the object whose waveform is amplified or reduced by the waveform deformation unit according to a predetermined reference for the signal waveform of the target.
5. The analytical apparatus as described in claim 1, characterized in that, The value of N and / or the value of M are predetermined or selected by the user.
6. A non-transient recording medium storing a waveform processing program for an analysis device, characterized in that it is a computer-readable non-transient recording medium storing a program for processing a target signal waveform that shows a change in signal intensity corresponding to a change in the value of a predetermined parameter on a computer, wherein the program is used to process the signal waveform on a computer. Make the computer perform: The waveform deformation step involves magnifying or reducing the signal waveform of the target by a factor of N in the direction of the signal strength axis, and / or magnifying or reducing it by a factor of M in the direction of the axis of the specified parameter, wherein N is a positive value other than 0 and 1, and M is a positive value other than 0 and 1, which can be the same value as N. In the peak detection step, a learned model is used. The deformed signal waveform from the waveform deformation step is taken as input, and the start and end points of the peak are taken as the detection results. The learned model is pre-generated by using the signal waveform and the start and end points of the peak as teacher data for machine learning. The waveform inverse deformation step involves reducing or increasing the output of the peak detection step, i.e., the information of the start and end points of the peak, by a factor opposite to that used in the waveform deformation step, to obtain the peak detection result of the signal waveform for the target. The determination step involves determining the magnitude of the change in signal strength level in the signal waveform of the target. The waveform deformation step amplifies or reduces the signal waveform for parameter values whose changes in signal strength level, as determined by the determination step, are less than or greater than a specified value.