Three-Dimensional Spectral Data Processing Without Peak Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional three-dimensional spectral data processing methods require peak detection, which can lead to inconsistent results due to varying peak detection algorithms and incorrect threshold settings, resulting in inaccurate multivariate analysis and the need for manual alignment processing, especially in systems like LC and GC, where retention time shifts occur.
Innovation Solution
A three-dimensional spectral data processing device and method that analyze similarities or differences in spectral data without peak detection, using multivariate analysis to identify characteristic spectra and calculate representative similarity values, eliminating the need for peak detection and alignment processing by treating spectral data as a single spectrum independent of the second parameter, such as retention time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If peak detection is performed on three-dimensional spectral data to extract characteristic data, then a two-dimensional characteristic data table can be created for multivariate analysis, but the results become inconsistent due to varying peak detection algorithms and incorrect threshold settings
Solution Approach 1:
The patent extracts only the essential spectral information needed for analysis by directly utilizing the spectral data without performing peak detection. This extraction approach removes the problematic peak detection step that causes inconsistency, while still obtaining the necessary characteristic data for multivariate analysis through direct spectral comparison and processing.
Solution Approach 2:
Instead of the conventional approach of detecting peaks first and then creating characteristic data tables, the patent inverts the process by performing multivariate analysis directly on the three-dimensional spectral data or by using spectral moments and other integrated parameters that do not require peak detection. This inversion eliminates the source of inconsistency.
2Productivity
If conventional peak detection methods are used to create characteristic data tables, then multivariate analysis can be performed, but manual alignment processing is required due to retention time shifts in LC and GC systems
Solution Approach 1:
The patent implements self-alignment capabilities through multivariate analysis methods that can automatically handle retention time shifts. By using techniques such as correlation optimization, dynamic time warping, or other alignment algorithms integrated into the multivariate analysis process, the system performs alignment automatically without requiring manual intervention, thus maintaining productivity while reducing operational complexity.
Solution Approach 2:
The patent replaces the manual mechanical alignment process with automated computational methods. Instead of manually adjusting and aligning peaks across different samples, the system uses computer-based multivariate analysis and alignment algorithms to automatically correct retention time shifts, eliminating the need for manual operation while maintaining or improving analysis quality.
3Loss of information
If peak detection with predetermined conditions is performed on spectral data, then characteristic data can be extracted, but the process becomes complex and requires careful setting of detection conditions
Solution Approach 1:
The patent changes the parameters used for spectral analysis from peak-specific parameters (requiring detection conditions) to integrated spectral parameters such as spectral moments, area-under-curve values, or other holistic descriptors. This parameter change allows extraction of meaningful characteristic data without the complexity of peak detection condition setting, while retaining comprehensive spectral information for analysis.
Data Source
AI summary
When performing an analysis of the difference between a specific sample group and a nonspecific sample group, a principle component analysis processing unit (33) performs principle component analysis on a collection of a plurality of mass spectrums created from data obtained for a single specific sample, and a characteristic spectrum acquisition unit (34) acquires a characteristic spectrum for each of a plurality of principle components using factor loadings. A spectrum similarity calculation unit (35) calculates the similarities between all mass spectrums and the characteristic spectrum for each sample, and obtains a representative value for the same. The similarity representative value for each sample is obtained for all the characteristic spectrums. A difference determination unit (36) checks whether there is a significant difference between the distribution of the similarity representative values of the specific sample group and the distribution of the similarity representative values of the nonspecific sample group and determines that the characteristic spectrum which is the source of the similarities having a significant difference is a difference spectrum. The difference spectrum reflects component information characterizing a sample group difference, so a component identification unit (37) searches for the difference spectrum in a library to identify a component. This makes it possible to perform different analysis without performing spectrum peak detection.