Chromatogram Waveform Simulation for Peak Detection Training
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing peak detection methods in chromatography face challenges in precision due to factors like baseline drift and peak overlap, and machine learning approaches require extensive labor and time for collecting teacher data, especially for samples from living bodies, with simulated data failing to accurately reflect variations in actual measurement data.
Innovation Solution
A data generation method using a generative adversarial network (GAN) to simulate chromatogram waveforms, focusing on shape parameters like peak height, width, and noise distribution, to create teacher data for training discriminators, thereby improving peak detection precision.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional peak detection methods are used, then the detection process is simple, but the detection precision deteriorates when noise is large or signal is unstable
Solution Approach 1:
The patent introduces a baseline correction line as an intermediary element to separate overlapping peaks. This baseline acts as a mediator that divides the complex overlapping peak signal into individual peak components, enabling accurate detection even when peaks overlap significantly. The baseline correction line is dynamically adjusted to fit the actual signal characteristics, providing a reference for peak integration and area calculation.
2Measurement precision
If machine learning methods are used to improve detection precision, then peak detection accuracy improves, but the time and labor required for data collection increases significantly
Solution Approach 1:
The patent performs baseline correction and peak separation operations in advance, before the actual peak detection process. By pre-establishing the baseline correction line and separating overlapping peaks beforehand, the system prepares the data in a format that enables faster and more accurate subsequent detection, reducing the computational burden and time required during actual analysis.
Solution Approach 2:
The patent creates a simplified model or representation of the baseline behavior through the baseline correction line. This copied baseline model is then used to subtract from the original signal, effectively separating the baseline drift from the peak signals. This copying approach allows the system to handle baseline variations without requiring complex real-time calculations.
3Manufacturing precision
If baseline correction is applied to separate overlapping peaks, then peak separation improves, but the complexity of the processing method increases
Solution Approach 1:
The patent dynamically adjusts parameters of the baseline correction line, such as its slope and intercept, to fit the actual signal characteristics. By changing these baseline parameters based on the local signal conditions, the system adapts to different baseline drift patterns and overlapping peak configurations, improving separation accuracy without requiring fundamentally complex processing architecture.
Data Source
AI summary
A data generation device according to the present invention is a data generation device configured to simulatively generate data used when creating, by machine learning, a discriminator configured to detect a peak observed in a signal waveform, the data generation device including: a parameter frequency information acquisition unit configured to acquire information on frequency of a predetermined shape parameter which characterizes a shape of a signal waveform from a plurality of signal waveforms collected only in a target field of machine learning for creating the discriminator; and a simulated waveform generation unit configured to generate a simulated signal waveform which is able to include overlapping of a plurality of peaks and noise using the information on frequency of the shape parameter, in which the simulated signal waveform is provided as data for training or evaluating machine learning.