Biological Fluid Analysis Data Dimensionality Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for analyzing biological liquids, such as flow cytometry, struggle to effectively visualize and automatically classify cell populations using more than two variables, leading to inadequate discrimination and counting of cellular classes, especially in detecting abnormal cells and pathologies.
Innovation Solution
A method that transforms n-tuples of biological fluid analysis data into m-tuples, allowing for classification and visualization in higher-dimensional spaces, using statistical knowledge to place cell classes in distinct zones, enabling better discrimination and representation of cell populations, including the use of linear or non-linear transformations and filters to enhance visualization and classification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If traditional two-dimensional matrix representation is used for visualizing cell populations, then the visualization is simple and easy to interpret, but the discrimination and classification of cell classes becomes inadequate when more than two variables are measured
Solution Approach 1:
The patent applies dimensionality change by transforming n-dimensional measurement data (where n>2) into m-dimensional composite spaces (where m<n) through mathematical transformations. This allows the system to preserve more measurement variables than traditional 2D matrices while still enabling visualization on standard displays. The composite space approach maintains discrimination precision by incorporating multiple physical parameters (optical, electrical, fluorescence) simultaneously, resolving the contradiction between simple visualization and adequate discrimination.
2Measurement precision
If more physical parameters are measured per cell (n>3), then the discrimination capability between cell classes improves, but the visualization and automatic classification becomes more complex
Solution Approach 1:
The patent transforms the measurement parameters from n-dimensional raw data into m-dimensional composite parameters through mathematical transformations. These transformations recombine multiple physical parameters (absorbance, resistivity, diffraction, fluorescence) into new composite variables that maintain the discriminatory power of the original measurements while reducing the dimensionality to a manageable level for automated classification algorithms.
Solution Approach 2:
The patent introduces composite spaces as intermediary representations between the raw n-dimensional measurements and the final classification output. These composite spaces act as mediators that simplify the complex multi-parameter data into a form that is more amenable to automated classification while preserving the essential discriminatory information needed to distinguish between different cell classes.
3Ease of operation
If orthogonal projection on two variables is used for visualization, then the display is simple, but the automatic discrimination of element classes is not always suitable
Solution Approach 1:
Instead of using traditional orthogonal projection onto two variables, the patent employs transformation into composite spaces of dimension m where 2≤m<n. This approach maintains visualization simplicity by still producing 2D or 3D displays that can be shown on standard screens, while simultaneously improving classification reliability by incorporating information from all n measurement variables through the mathematical transformations, rather than discarding n-2 variables through projection.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach enables precise discrimination and visualization of cell populations in multiple dimensions, improving the detection of abnormal cells and pathologies by integrating all measured physical parameters, providing a comprehensive representation of leukocyte subpopulations and facilitating the identification of pathological signatures.
Implementation Method 1
Flow cytometry is one of the techniques suitable for the statistical study of cell populations because the cells are studied one by one on a sample of several hundreds or thousands of cells
Implementation Method 2
A first detector, placed near the axis of the incident laser beam, measures the diffraction at small angles
Implementation Method 3
The use of molecular probes combining, according to the principle of surface antigens, an antibody and a luminophore makes it possible to reveal specific functions located on the surface of cell membranes
Implementation Method 4
Other detectors can be placed at 90° from the axis of the incident beam. The detected light is analyzed according to one or more spectral components corresponding to fluorescence or diffraction lights
Implementation Method 5
Electrical measurements are also carried out such as the measurement of resistivity by the electronic gate principle well known to those skilled in the art
Data Source
Figure 1
Figure 2A~2C
Figure 3
AI summary
The invention relates to a method for use in an automated biological fluid analysis system measuring at least four physical parameters (n>3) for each detected cell for classification by discrimination and counting into a set of at least three cell classes and their representation.According to this process, mathematical transformations of a plurality of n-tuples into m-tuples are stored and executed as needed, m <n, chaque transformation permettant de placer les classes cellulaires d'un liquide biologique présentant les caractéristiques statistiques moyennes dans des zones distinctes de l'espace composite à m dimensions, des filtres de discrimination et de reclassement en au moins deux classes cellulaires et au moins une transformation d'une pluralité de n-uplets en 3-uplets, 2-uplets, ou 1-uplets, pour afficher les classes cellulaires d'un liquide biologique présentant les caractéristiques statistiques moyennes dans des zones distinctes de l'espace à 3 dimensions, de la surface à 2 dimensions, ou de l'axe à une dimension.