Waveform Scaling for Low-Intensity Peak Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing peak detection methods using supervised machine learning, such as deep learning, face challenges in accurately detecting peaks in chromatograms due to limited teaching data, leading to missed detections, especially for peaks with low signal intensity, which requires significant time and cost to regenerate the learned model.
Innovation Solution
The analysis device and waveform processing program enhance peak detection accuracy by incorporating a waveform selection, enlargement, and reduction process, utilizing an existing learned model to detect low-intensity peaks without regenerating the model, through functions like waveform selection, enlargement, and reduction units.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If signal processing is performed by a single processor, then device complexity is reduced, but processing speed and capability deteriorate
Solution Approach 1:
The signal processing function is divided into multiple independent processors (first processor and second processor). The first processor performs initial signal processing to generate intermediate results, while the second processor performs additional processing on these intermediate results. This segmentation allows parallel processing operations, improving overall processing speed without requiring a single complex processor.
Solution Approach 2:
Multiple processors are combined to work together on the same signal processing task. The first processor and second processor are merged in a collaborative processing architecture where they share the load of processing the plurality of signals, achieving higher processing capability than a single processor could provide alone.
2Productivity
If multiple processors are used for signal processing, then processing capability is improved, but device complexity increases
Solution Approach 1:
The processing workload is segmented across multiple processors, with each processor handling specific portions of the signal processing task. This division allows the system to achieve high processing capability while keeping each individual processor relatively simple in structure.
Solution Approach 2:
The multiple processors are designed with universal processing capabilities to handle different types of signal processing operations. This multi-functionality allows them to work together efficiently on various signal processing tasks without requiring specialized complex hardware for each specific function.
3Measurement precision
If detailed waveform analysis is performed, then diagnostic accuracy is improved, but processing time increases
Solution Approach 1:
The first processor performs preliminary signal processing operations to generate intermediate results before the second processor conducts more detailed analysis. This preliminary action prepares the data in advance, allowing the second processor to focus on specific diagnostic features without having to process all raw signals from scratch, thereby reducing overall processing time while maintaining diagnostic accuracy.
Solution Approach 2:
The detailed waveform analysis is segmented into multiple processing stages handled by different processors. The first processor handles initial signal conditioning and feature extraction, while the second processor performs detailed diagnostic analysis on the pre-processed signals. This segmentation allows comprehensive analysis without requiring all processing to occur sequentially in a single processor.
Data Source
Figure 1
Figure 2~3
Figure 4A~4D
AI summary
An analysis device according to one mode of the present invention includes: a waveform deformation unit (112) configured to enlarge or reduce a target signal waveform obtained by analysis and indicating a change in signal intensity depending on a change in a value of a predetermined parameter by a scale factor of N (where N is a positive value other than 0 and 1) in the direction of the signal intensity axis and/or enlarge or reduce by a scale factor of M (where M may be a positive value other than 0 and 1 and may be the same value as N) in the direction of the predetermined parameter axis; a peak detection unit (113) configured to use a learned model generated in advance by machine learning using, as teaching data, a signal waveform and a start point and an end point of a correct solution, and use, as an input, a signal waveform after deformation by the waveform deformation unit, and output, as a detection result, the start point and the end point of the peak; and a waveform inverse deformation unit (114) configured to reduce or enlarge information on a start point and an end point of a peak, the information being output by the peak detection unit, by a scale factor inverse to that at the time of deformation by the waveform deformation unit, and obtain a peak detection result for the target signal waveform.