US-12625119-B2 - Training method

US12625119B2US 12625119 B2US12625119 B2US 12625119B2US-12625119-B2

Abstract

A training method includes: acquiring a pseudo noise waveform indicating an assumed noise; acquiring a pseudo peak waveform not including the pseudo noise; generating a pseudo signal waveform by adding the pseudo noise waveform and the pseudo peak waveform; and updating an estimation model based on the pseudo signal waveform, in which the acquiring the pseudo noise waveform includes: causing an analysis device to execute noise measurement generating the pseudo noise waveform a plurality of times; calculating a similarity degree of a plurality of pseudo noise waveforms generated by the noise measurement performed the plurality of times; and executing prescribed processing according to the similarity degree.

Inventors

Kenta CHINOMI
Kaori SUGIMURA
Shinji KANAZAWA

Assignees

SHIMADZU CORPORATION

Dates

Publication Date: 20260512
Application Date: 20230314
Priority Date: 20220315

Claims (8)

1 . A training method for training an estimation model used to detect a peak of a signal waveform output by an analysis device that analyzes a sample, the training method comprising: acquiring a noise waveform that may be generated when the analysis device executes analysis processing; and training the estimation model based on the noise waveform, wherein the acquiring the noise waveform includes: acquiring first training data, wherein the first training data includes a first noise waveform and a second noise waveform generated by a noise measurement performed a plurality of times by the analysis device; calculating a similarity degree between the first noise waveform and the second noise waveform; and executing a prescribed processing according to the similarity degree, wherein the prescribed processing includes generating second training data by removing the first noise waveform and the second noise waveform from the first training data when the similarity degree is equal to or greater than a threshold value, wherein the training the estimation model includes training the estimation model based on the second training data.
2 . The training method according to claim 1 , wherein the prescribed processing includes processing for notifying a user that the similarity degree is greater than or equal to the threshold value.
3 . The training method according to claim 1 , wherein the noise measurement is measurement analyzed by the analysis device while the sample is not disposed in the analysis device.
4 . The training method according to claim 1 , wherein the noise measurement is measurement in which an actual peak waveform is generated in a first section of a signal waveform by the analysis device analyzing a known sample in which a compound is known, and the acquiring the noise waveform includes acquiring an actual noise waveform generated in a second section in which the actual peak waveform is not detected as the noise waveform.
5 . The training method according to claim 4 , wherein the training the estimation model includes training the estimation model based on the actual peak waveform and the noise waveform.
6 . The training method according to claim 1 , wherein the similarity degree is value of correlation coefficient between the first noise waveform and the second noise waveform.
7 . The training method according to claim 1 , further comprising: acquiring a peak waveform that does not include the noise waveform; and generating a signal waveform by adding the noise waveform and the peak waveform, wherein the training the estimation model includes training the estimation model based on the signal waveform.
8 . The training method according to claim 7 , wherein the peak waveform is acquired by a generative adversarial network.

Description

BACKGROUND OF THE INVENTION Field of the Invention The present disclosure relates to a training method. Description of the Background Art WO 2020/070786 discloses a chromatographic system. This chromatographic system separates and detects a peak of an unseparated peak of a chromatogram by artificial intelligence (AI) using an estimation model. The chromatographic system performs qualitative analysis or quantitative analysis of a sample based on the peak. WO 2020/070786 discloses that a computer executes training updating an estimation model. This computer acquires a plurality of chromatograms each of which has a peak, and prepares the chromatogram of the unseparated peak by adding the plurality of chromatograms. The computer updates an estimation model using the plurality of chromatograms as training data and using the prepared chromatogram as training data. SUMMARY OF THE INVENTION Generally, in an analysis device such as a chromatograph, a signal waveform (for example, the chromatogram) including a noise waveform of noise that can be generated when a sample is analyzed is sometimes generated. WO 2020/070786 does not disclose training in which the noise waveform is reflected. The present disclosure has been made to solve such a problem, and an object of the present disclosure is to perform training reflecting the noise waveform that can be generated when the analysis device analyzes the sample. A training method of the present disclosure is a training method for training an estimation model used to detect a peak of a signal waveform output by an analysis device that analyzes a sample. The training method includes acquiring a noise waveform that may be generated when an analysis device executes analysis processing and training the estimation model based on the noise waveform. The acquiring the noise waveform includes: acquiring a plurality of noise waveforms by noise measurement executed a plurality of times by the analysis device; calculating a similarity degree of the plurality of noise waveforms; and executing prescribed processing according to the similarity degree. A training program of the present disclosure is a training program that causes a computer to update an estimation model used to detect a peak of a signal waveform output by an analysis device that analyzes a sample. The training program causes a computer to execute acquiring a noise waveform that may be generated when an analysis device executes analysis processing and training an estimation model based on the noise waveform. The acquiring the noise waveform includes: acquiring a plurality of noise waveforms by noise measurement executed a plurality of times by the analysis device; calculating a similarity degree of the plurality of noise waveforms; and executing prescribed processing according to the similarity degree. The foregoing and other objects, features, aspects and advantages of the present invention will become more apparent from the following detailed description of the present invention when taken in conjunction with the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a view illustrating a configuration example of an analysis system. FIG. 2 is a block diagram illustrating a hardware configuration of a training device 30. FIG. 3 is a view illustrating pseudo chromatogram. FIG. 4 illustrates an example of blank chromatogram that does not include a non-assumed noise. FIG. 5 illustrates an example of the blank chromatogram including a non-assumed noise M. FIG. 6 is a functional block diagram illustrating a training device. FIG. 7 is a functional block diagram illustrating a GAN execution unit and an update unit. FIG. 8 is a functional block diagram illustrating a noise generation unit. FIG. 9 is a flowchart illustrating a processing procedure of the analysis system. FIG. 10 is a flowchart illustrating processing in step S2. FIG. 11 is a view illustrating an example of a result of prescribed measurement. FIG. 12 is a flowchart illustrating processing in step S2 according to a second embodiment. FIG. 13 illustrates an example of a prescribed image. DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. In the drawings, the same or corresponding part is denoted by the same reference numeral, and the description thereof will not be repeated. First Embodiment (Analysis System) The present disclosure relates to a training technique for updating an estimation model used to detect a peak of a signal waveform output by an analysis device. Examples of the analysis device include a gas chromatograph (GC) device, a liquid chromatography (LC) device, a mass spectrometer, a spectrophotometer, and an X-ray analyzer. For example, the signal waveform may be a chromatogram waveform or a mass spectrum waveform. When the analysis device is the spectrophotometer, the signal waveform is an absorption spectrum waveform. When the analysis devic