Neural Network Unmixing for Overlapping Fluorescent Biomolecule Imaging
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional fluorescent imaging techniques for biological samples are limited by the need for non-overlapping emission spectra of fluorescent substances, restricting the number of simultaneously usable substances to 3-5, and chemical or physical treatments to deactivate these substances can damage the sample.
Innovation Solution
An image processing method using an artificial neural network model and an unmixing matrix to generate unmixed images from mixed images, allowing for the separation of multiple biomolecules labeled with overlapping fluorescent substances without the need for deactivation or removal processes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If fluorescent substances with non-overlapping emission spectra are used, then the number of simultaneously observable biomolecules is limited to 3-5, but if fluorescent substances with overlapping emission spectra are used, then the emission spectra must be carefully selected and the number of simultaneously usable substances is restricted
Solution Approach 1:
The patent changes the parameter of emission spectrum overlap from forbidden to allowed, using fluorescent substances with overlapping spectra. The unmixing algorithm mathematically separates the mixed signals by solving the linear equation system, where the emission spectra of multiple fluorescent substances form the basis matrices and the measured mixed spectra are decomposed to quantify individual component contributions.
2Productivity
If chemical or physical treatment is applied to deactivate fluorescent substances, then the same fluorescent substances can be reused for labeling other biomolecules, but the sample or biomolecules inside the sample may be distorted or damaged
Solution Approach 1:
The patent replaces the mechanical/chemical deactivation process with a computational unmixing algorithm. Instead of physically removing or deactivating fluorescent substances through chemical treatments that may damage the sample, the system uses mathematical decomposition of the mixed emission spectra to separate and quantify individual fluorescent substance signals, allowing direct observation without sample manipulation.
3Quantity of substance
If a process of deactivating fluorescent substances or removing labeled biomolecules is repeated to observe 10 or more biomolecules, then the limited number of fluorescent substances can be reused, but the imaging process becomes complex and time-consuming
Solution Approach 1:
The patent merges multiple fluorescent substances with overlapping emission spectra into a single imaging experiment. Instead of performing sequential imaging with deactivation steps, the system captures mixed emission spectra simultaneously and uses unmixing algorithms to resolve individual component signals, combining multiple observations into one integrated measurement process.
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
Improves the accuracy of image unmixing and enables the simultaneous observation of multiple biomolecules using a greater number of fluorescent substances, reducing sample damage and simplifying the imaging process.
Implementation Method 1
A value of at least one element included in the unmixing matrix may be determined based on training of an artificial neural network model
Implementation Method 2
generating a unmixed image for at least one biomolecule among the plurality of biomolecules from the plurality of mixed images using an unmixing matrix
Data Source
Figure 1
Figure 2
Figure 3
AI summary
As one aspect disclosed herein, an image processing method may be proposed. The method is executed in an electronic device comprising one or more processors and one or more memories for storing instructions to be executed by the one or more processors, and may comprise the steps of: acquiring a plurality of mixed images of a sample including a plurality of biological molecules; and generating unmixed images of at least one of the plurality of biological molecules from the plurality of mixed images by using an unmixing matrix. The value of at least one element included in the unmixing matrix may be determined on the basis of artificial neural network model training.