Exosome multi-parameter joint quantitative analysis method and device based on fluorescence intensity of coded microspheres

Abstract

This invention relates to the field of exosome quantitative analysis technology, and discloses a multi-parameter joint quantitative analysis method and apparatus for exosomes based on the fluorescence intensity of encoded microspheres. The method includes: performing three-dimensional spatial attitude inversion to obtain a three-dimensional attitude characterization vector and constructing a feature point diffusion function; performing spatial deconvolution processing using the feature point diffusion function, and combining it with a visibility weight calculated from the longitudinal depth deviation for brightness gain compensation to obtain intrinsic feature intensity; constructing a channel intermodulation matrix to perform signal decoupling operations, obtaining decoupling intensity and mapping it to quantified abundance. This invention solves the problems of projection distortion, defocus blurring, and spectral leakage caused by random attitude, eliminates steric hindrance effects and system nonlinear interference, significantly improves the accuracy and repeatability of exosome detection, and achieves accurate analysis of high-fidelity biological quantitative information.

Description

Technical Field

[0001] This invention relates to the field of exosome quantitative analysis technology, and more specifically, to a method and apparatus for multi-parameter combined quantitative analysis of exosomes based on the fluorescence intensity of encoded microspheres. Background Technology

[0002] With the increasing demand for exosome detection in clinical quantitative analysis, high-throughput exosome quantitative analysis technology based on microfluidic chips has attracted much attention as an efficient detection method. However, traditional detection methods, such as simply improving sensor sensitivity or increasing sample volume and other macroscopic parameter optimization techniques, generally face bottlenecks due to signal fluctuations in dynamic flow environments. This results in high dispersion of detection results for low-abundance samples, failing to meet the technical requirements of high repeatability and high accuracy. Current research often focuses on the pixel brightness statistics of microsphere fluorescence signals or simplifies the coded microspheres into isotropic particle models for concentration conversion. This ignores the non-ideal coupling effect between microfluidic dynamics and optical imaging systems in high-throughput flow channel environments: the coded microspheres are affected by laminar shear forces and fluid disturbances within the flow channel, resulting in random three-dimensional spatial rotation and flipping. Because the characteristic dyes and surface loads inside the microspheres have a specific geometric distribution, this attitude shift causes anisotropic distortion in its feature projection on the two-dimensional image plane. Furthermore, the refractive and light-loss characteristics of the microsphere medium itself cause energy attenuation of the signal facing away from the imaging lens as it passes through the sphere, inducing a severe self-shielding effect. Traditional analysis algorithms, lacking awareness of random attitude angles, cannot distinguish between signal attenuation caused by steric hindrance and signal weakening due to low concentration of the target material, leading to a systematic bias in signal reconstruction. Therefore, shifting from passive, indiscriminate pixel statistics to spatially directional attitude recognition, and achieving inverse estimation of the real-time attitude of the microspheres and compensating for energy loss caused by physical shielding during digital processing, is a key challenge in overcoming the bottleneck of high-sensitivity exosome detection technology and achieving accurate reconstruction from three-dimensional physical features to two-dimensional reduced projection.

[0003] In the prior art, Chinese patent CN113358618A discloses an exosome capture method based on surface-spiked coded microspheres. This patent uses coded microspheres with a surface-spiked structure to improve exosome capture efficiency, achieves biomarker identification through fluorescence signal reading, and uses coded microspheres as carriers to complete exosome enrichment and preliminary characterization, optimizing the binding efficiency between microspheres and exosomes, and providing a technical foundation for the simultaneous capture of multiple biomarkers. Chinese patent CN111518668B discloses a microfluidic system for exosome extraction and detection. This technology uses a microfluidic chip as its core, integrating exosome extraction, fluorescence imaging, and signal acquisition modules. It achieves dynamic sample detection through optimized flow channel structure and directly calculates concentration using pixel brightness statistics, enabling integrated rapid detection of exosomes and adapting to clinical batch sample screening scenarios. However, while the two existing technologies mentioned above have some value in terms of exosome capture efficiency and detection integration, they fail to address the core pain points of three-dimensional posture distortion of coded microspheres in dynamic flow environments, fluorescence signal fluctuations caused by self-shielding effects, and quantitative bias in low-abundance samples. Chinese patent CN113358618A only focuses on improving the surface structure of microspheres to enhance capture capabilities, without addressing the two-dimensional projection distortion caused by random microsphere rotation or the energy attenuation due to medium self-shielding, and cannot distinguish between signal attenuation caused by steric hindrance and signal weakening caused by low concentration. Chinese patent CN111518668B only optimizes microfluidic hardware and basic imaging acquisition processes, without performing three-dimensional posture inversion, defocus signal repair, or handling multi-channel spectral cross-interference and sensor nonlinear response distortion. Both remain at the level of passive pixel brightness statistics, simplifying the coded microspheres into isotropic point mass models, lacking spatial directional posture recognition, and failing to compensate for energy loss caused by physical shielding. This results in high dispersion in the detection of low-abundance samples, making it difficult to meet the requirements of high repeatability and high accuracy for clinical quantitative analysis. Summary of the Invention

[0004] This invention is applicable to high-throughput quantitative analysis of exosomes based on microfluidic chips, meeting the technical requirements of high repeatability and high accuracy for multi-parameter joint detection. By inversely mapping the spatial distribution topology of feature points to the three-dimensional pose representation vector, it achieves accurate identification of random spatial rotation of coded microspheres, solving the problem of dimensional collapse during the projection of three-dimensional physical features onto a two-dimensional image plane. The feature point diffusion function utilizes the two-dimensional normal distribution characteristics to simulate the evolution of light field energy on the off-focal plane, transforming the qualitative description of pixel energy spillover caused by defocusing into quantitative compensation. Combined with visibility weights, it achieves energy realignment and repair in the pixel dimension, accurately correcting the self-shielding effect and energy loss caused by longitudinal depth deviation. The channel intermodulation matrix, combined with correction weight factors, provides a dynamic interference elimination benchmark for different detection channels, achieving algebraic decoupling of complex spectral leakage and sensor nonlinear response distortion. This invention comprehensively improves the signal restoration capability of exosome detection, significantly reducing systematic bias caused by pose variables while ensuring high-fidelity molecular distribution information.

[0005] To achieve the above objectives, the present invention provides the following technical solution: A multi-parameter combined quantitative analysis method for exosomes based on the fluorescence intensity of encoded microspheres includes: The original image of the coded microsphere is obtained, and edge enhancement and feature spot clustering analysis are performed on the original image of the microsphere to obtain a feature point spatial distribution topology map with spatial causal constraints. The feature point spatial distribution topology map is called to perform three-dimensional spatial pose inversion to obtain a three-dimensional pose representation vector. The longitudinal depth deviation is calculated based on the three-dimensional pose representation vector, and the feature point diffusion function is constructed using the longitudinal depth deviation. Spatial deconvolution processing is performed on the original microsphere image using the feature point diffusion function to generate a spatial consistency restoration image. Visibility weights are calculated based on the longitudinal depth deviation, and brightness gain compensation is performed on the spatial consistency restoration image using the visibility weights to obtain the intrinsic feature intensity. The channel intermodulation matrix is ​​constructed using the intrinsic feature strength. The signal decoupling operation is performed using the channel intermodulation matrix to obtain the decoupling strength. Based on the decoupling strength, digital encapsulation is performed to obtain the quantized output dataset.

[0006] Furthermore, the method for obtaining the spatial distribution topology map of the feature points includes: The original image of the microsphere is acquired through a high-speed imaging interface. A pre-set second-order differential convolution kernel is used to perform a convolution operation with the original image of the microsphere. The feature response matrix is ​​obtained by traversing all pixels of the original image of the microsphere, and the response value corresponding to each pixel is determined. The response values ​​in the feature response matrix that are greater than a preset background threshold are identified, and the corresponding set of pixel coordinates is defined as a candidate point set. The candidate point set is then subjected to pixel clustering processing using the neighborhood connection criterion to obtain connected regions. A weighted average calculation is performed on the connected regions to obtain the geometric center coordinates of each connected region, which are defined as the feature point mapping coordinates. A distance correlation matrix is ​​constructed by calculating the Euclidean distance between the mapped coordinates of each feature point. The feature point mapped coordinates, the distance correlation matrix, and the original image of the microsphere are logically encapsulated to generate a topological map of the spatial distribution of feature points.

[0007] Furthermore, the three-dimensional pose representation vector includes: Obtain the reference feature point set when the coded microsphere is in the zero position in the preset three-dimensional spatial coordinate system, construct the observation coordinate matrix using the feature point mapping coordinates, and construct the reference coordinate matrix using the reference feature point set; The mean of each column of the observation coordinate matrix and the reference coordinate matrix is ​​calculated separately, and then the centering process is performed to obtain the decentralized observation matrix and the decentralized reference matrix. The transpose of the decentralized reference matrix and the decentralized observation matrix are multiplied together to obtain the attitude covariance matrix. The singular value decomposition algorithm is used to perform matrix decomposition on the attitude covariance matrix, outputting a left singular vector matrix and a right singular vector matrix. The transpose of the left singular vector matrix and the right singular vector matrix are multiplied to generate a rotation transformation matrix. Angle decomposition is performed on the rotation transformation matrix to obtain a three-dimensional attitude representation vector composed of pitch angle, yaw angle and roll angle.

[0008] Furthermore, the three-dimensional pose representation vector also includes: Obtain the acquisition optical axis of the high-speed imaging interface, define an initial vector parallel to the direction of the acquisition optical axis of the high-speed imaging interface, and perform a rotation transformation operation on the initial vector using a rotation transformation matrix to obtain a spatial vector with a definite pointing offset in the three-dimensional spatial coordinate system, which is defined as the attitude vector.

[0009] Furthermore, the longitudinal depth deviation includes: Obtain the geometric radius of the coded microsphere and the ideal focal plane of the high-speed imaging interface, and use the high-speed imaging interface to determine the offset of the geometric center of the coded microsphere relative to the ideal focal plane; By performing trigonometric function calculations using the yaw angle and geometric radius, the local height deviation on the surface is obtained. The offset and the local height deviation on the surface are then summed to obtain the longitudinal depth deviation.

[0010] Furthermore, the feature point diffusion function includes: The optical system transfer constant is set, and the square of the longitudinal depth deviation is multiplied by the optical system transfer constant to obtain the energy diffusion radius. Centered on the feature point mapping coordinates, a matrix window containing multiple pixels is determined within the original image of the microsphere, and the region contained in the matrix window is defined as the support region; Each pixel within the support area is defined as a sampling pixel, and the two-dimensional pixel coordinate offset of the sampling pixel relative to the feature point mapping coordinates is calculated. Utilizing the probability density characteristics of a two-dimensional normal distribution, a second-order Gaussian kernel function is constructed based on the energy diffusion radius and the two-dimensional pixel coordinate offset, defined as the feature point diffusion function, where the integral value of the feature point diffusion function over the support region is one.

[0011] Furthermore, the consistency repair image includes: A local pixel matrix with the same size as the support region is extracted from the original image of the microsphere with the feature point mapping coordinates as the center. This matrix is ​​defined as the observation intensity matrix. A matrix with the same size as the observation intensity matrix and all initial element values ​​are one is constructed. This matrix is ​​defined as the correction intensity matrix. The energy ratio matrix is ​​obtained by performing a convolution operation between the correction intensity matrix and the feature point spread function, and then calculating the ratio between the resulting convolution and the observed intensity matrix. Perform spatial flipping on the feature point spread function to obtain the adjoint kernel matrix. Perform convolution operation on the adjoint kernel matrix and the energy ratio matrix to obtain the corrected gradient matrix. Using the maximum likelihood estimation criterion, the correction intensity matrix is ​​updated by dot product using the modified gradient matrix, and the correction intensity matrix after a preset number of iterations is determined as the spatial consistency repair image.

[0012] Furthermore, the intrinsic feature strength includes: Set the attenuation correction coefficient, and perform an arithmetic multiplication operation between the longitudinal depth deviation and the attenuation correction coefficient to obtain the phase compensation term; Perform an exponential operation on the phase compensation term to obtain the visibility weight; The intrinsic feature intensity is obtained by performing inverse multiplication compensation on the mapping coordinates of corresponding feature points in the spatial consistency repair image using the visibility weight.

[0013] Furthermore, the channel intermodulation matrix includes: The independent optical signal acquisition window in the high-speed imaging interface is identified and defined as the detection channel, and the fluorescent chemical groups used to label exosome markers are identified and defined as fluorescent probes; Obtain the transmission function of the filter corresponding to the detection channel and the emission spectral distribution function corresponding to the fluorescent probe, calculate the overlap area of ​​the transmission function of the filter and the emission spectral distribution function, obtain the spectral cross-induction coefficient, and form a spectral induction matrix based on the spectral cross-induction coefficient; A linear mapping operation is performed using the intrinsic characteristic intensity and the spectral sensing matrix to output the energy penetration amplitude, and a correction weighting factor is set. By performing a weighted dot product operation on the energy penetration amplitude using a modified weighting factor, the channel intermodulation matrix is ​​obtained.

[0014] Furthermore, the decoupling strength includes: Extract the values ​​corresponding to the preset matrix index positions in the channel intermodulation matrix, and define them as interference components; The intrinsic feature intensities corresponding to each detection channel are combined into a column vector. The corresponding interference components are then subtracted from the column vector to eliminate spurious brightness gains caused by spectral overlap and sensor nonlinear response, thus obtaining the decoupling intensity.

[0015] Furthermore, the quantized output dataset includes: The decoupling strength was obtained by using standard exosome samples with known concentration sequences, and regression fitting was performed to obtain a concentration mapping function consisting of the regression slope and intercept term. The decoupling strength corresponding to the sample to be tested is used as the input variable and substituted into the concentration mapping function. The multiplication and addition operation is performed using the regression slope and intercept term to obtain the quantitative abundance. The quantitative abundance and the identity information of the encoded microspheres are digitally encapsulated to obtain a quantitative output dataset.

[0016] A multi-parameter joint quantitative analysis device for exosomes based on encoded microsphere fluorescence intensity is used to implement the above-mentioned multi-parameter joint quantitative analysis method for exosomes based on encoded microsphere fluorescence intensity. The system includes: The attitude topology module is used to acquire the original image of the coded microsphere, perform edge enhancement and feature spot clustering analysis on the original image of the microsphere to obtain a feature point spatial distribution topology map with spatial causal constraints, call the feature point spatial distribution topology map to perform three-dimensional spatial attitude inversion to obtain a three-dimensional attitude representation vector, calculate the longitudinal depth deviation based on the three-dimensional attitude representation vector, and use the longitudinal depth deviation to construct the feature point diffusion function. Light intensity correction module: Used to perform spatial deconvolution processing on the original image of microspheres using feature point diffusion function to generate spatial consistency restoration image, calculate visibility weight based on longitudinal depth deviation, and call visibility weight to perform brightness gain compensation on spatial consistency restoration image to obtain intrinsic feature intensity; Decoupling Quantification Module: This module is used to construct a channel intermodulation matrix using the intrinsic feature strength, perform signal decoupling operations using the channel intermodulation matrix to obtain the decoupling strength, and perform digital encapsulation based on the decoupling strength to obtain the quantized output dataset.

[0017] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention achieves precise alignment between the digital image features of coded microspheres and their three-dimensional spatial orientation through the inverse mapping of the spatial distribution topology of feature points and the three-dimensional attitude representation vector. This solves the problem of projection distortion caused by neglecting rotational degrees of freedom in traditional isotropic point models. The feature point diffusion function transforms the energy dispersion caused by defocusing from a traditional blurring phenomenon into a quantitative degradation model based on Fresnel diffraction theory. Combined with visibility weights based on longitudinal depth deviation for brightness gain compensation, it achieves pixel-level energy realignment and accurately solves the problem of signal intensity asymmetry induced by feature points deviating from the ideal focal plane. The channel intermodulation matrix, combined with the correction weight factor of the sensor's nonlinear response range, achieves algebraic stripping of energy leakage from complex spectra, accurately restoring the intrinsic fluorescence emission level of the target exosome biomarker. This invention realizes a closed-loop analysis from the original digital image to high-fidelity bioquantitative information, significantly improving the repeatability and accuracy of exosome quantitative analysis. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart of a multi-parameter combined quantitative analysis method for exosomes based on the fluorescence intensity of encoded microspheres, provided in an embodiment of the present invention. Figure 2 This is a schematic diagram of the logic for generating a spatial distribution topology map of feature points, provided by an embodiment of the present invention. Figure 3 A logical schematic diagram illustrating the relationship between a three-dimensional posture representation vector and a spatial mapping, provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of a spatial deconvolution energy relocation logic provided in an embodiment of the present invention; Figure 5 A schematic diagram illustrating the definition of the nonlinear response interval provided in an embodiment of the present invention; Figure 6 This is a functional block diagram of the exosome multi-parameter combined quantitative analysis device based on the fluorescence intensity of encoded microspheres provided in an embodiment of the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Example 1 Please see Figure 1 As shown, this embodiment provides a multi-parameter joint quantitative analysis method for exosomes based on the fluorescence intensity of encoded microspheres, including: Step S10: Obtain the original image of the coded microspheres, perform edge enhancement and feature spot clustering analysis on the original image of the microspheres to obtain a feature point spatial distribution topology map with spatial causal constraints, call the feature point spatial distribution topology map to perform three-dimensional spatial pose inversion to obtain a three-dimensional pose representation vector, calculate the longitudinal depth deviation based on the three-dimensional pose representation vector, and use the longitudinal depth deviation to construct the feature point diffusion function.

[0022] Further, step S10 includes: Step S11: Obtain the original image of the encoded microspheres, perform edge enhancement and feature spot clustering analysis on the original image of the microspheres, and obtain a topological map of the spatial distribution of feature points with spatial causal constraints.

[0023] In high-throughput fluid control detection scenarios, the trajectory of coded microspheres within microchannels is influenced by complex laminar shear forces, leading to random rotation and flipping. Traditional detection techniques typically treat coded microspheres as isotropic point masses, neglecting their rotational degrees of freedom in three-dimensional space. This results in an inability to distinguish between signal attenuation due to steric hindrance and signal weakening due to low target concentration when extracting fluorescence signals, causing systematic bias in quantitative analysis results. To eliminate this dynamic response distortion caused by hidden physical space attitude variables, a reference framework is established that accurately maps the digital image features of coded microspheres to their three-dimensional spatial orientation, outputting a structured quantitative index. The aim is to transform passive, undifferentiated pixel brightness statistics into an active, spatially directional attitude recognition mechanism.

[0024] Specifically, the original image of the microsphere is acquired through a high-speed imaging interface. This high-speed imaging interface is a hardware communication protocol unit that captures a pixel matrix stream of coded microspheres moving within a microchannel at a preset sampling frequency. The sampling frequency is set based on the ratio of the flow velocity of the coded microsphere within the microchannel to its geometric diameter, aiming to ensure that at least ten complete images are captured within the time it takes for the coded microsphere to flow through the imaging field of view to eliminate motion artifacts. For example, the sampling frequency is set to 2000 frames per second. The original image of the microsphere is a two-dimensional matrix composed of pixels with different grayscale values. The original image includes background pixel noise values, grayscale step values ​​at the microsphere boundaries, and center intensity values ​​of internal feature points. The background pixel noise value refers to the pixel grayscale intensity generated by the dark current of the imaging sensor outside the physical contour boundary of the coded microsphere in the original image of the microsphere. The microsphere boundary grayscale step value refers to the pixel brightness change gradient caused by the sudden change in optical refractive index at the boundary between the edge region of the coded microsphere and the region corresponding to the background pixel noise value, used to lock the geometric closure range of the coded microsphere in digital space. The center intensity value of the internal feature point refers to the grayscale value of the center point with the highest local pixel brightness excited by the fluorescent coding dye inside the physical contour boundary of the coded microsphere, characterizing the digital identity feature carried by the coded microsphere. In order to convert the original image of the microsphere into a spatial attitude calculation for precise positioning, a two-dimensional rectangular coordinate system is constructed with the lower left corner vertex of the original image of the microsphere as the origin, defined as the pixel coordinate system. In the pixel coordinate system, the horizontal arrangement direction is defined as the horizontal axis, i.e., the U-axis, and the vertical arrangement direction is defined as the vertical axis, i.e., the V-axis. The function of the pixel coordinate system is to provide a unique spatial index coordinate for each pixel.

[0025] Edge enhancement processing is performed on the original image of the microspheres. Specifically, a pre-defined second-order differential convolution kernel is used to perform a convolution operation with the original microsphere image. This kernel is a 3x3 discrete matrix, set to ensure that the convolution result is zero in regions with uniform pixel brightness, while outputting a non-zero response value in regions with varying brightness. For example, the value of the central element of the discrete matrix is ​​set to -8, and the values ​​of its eight adjacent elements are all positive 1. The central element of the second-order differential convolution kernel is aligned with a target pixel in the original microsphere image. Arithmetic multiplication is performed between the nine elements of the second-order differential convolution kernel and the grayscale values ​​of the corresponding target pixel and its neighboring pixels. The resulting nine products are then summed to obtain the response value corresponding to the target pixel. Neighboring pixels refer to the eight adjacent pixel coordinates centered on the target pixel with a radius of one pixel in the pixel coordinate system. By traversing all pixels in the original microsphere image, a feature response matrix representing the strength of image edge features is obtained. The system identifies all pixel coordinates in the feature response matrix whose response values ​​are greater than a preset background threshold, and defines the set of pixels corresponding to these pixel coordinates as a candidate point set. The background threshold is set by superimposing three times the standard deviation of the statistical average of the background pixel noise values, with the aim of eliminating the interference of random noise at the statistical level. For example, it is set to 50.

[0026] For the candidate point set, pixel clustering is performed using the neighborhood connectivity criterion to obtain connected regions. A connected region is a set of pixels that are spatially adjacent in the candidate point set and have the same attribution label. Specifically, when two pixels in the candidate point set are adjacent in the horizontal, vertical, or diagonal direction in the pixel coordinate system, they are determined to belong to the same connected region. The role of connected regions is to logically aggregate discrete pixels belonging to the same internal feature point in the digital space, thereby transforming scattered pixel coordinates into feature blob entities with independent physical meaning. A weighted average calculation is performed on each connected region in the candidate point set. Specifically, the gray value of each pixel in the connected region is used as a weight coefficient. The gray value of each pixel is arithmetically multiplied by its horizontal and vertical coordinates in the pixel coordinate system and summed to obtain the weighted sum of the horizontal and vertical axes. The weighted sums of the horizontal and vertical axes are then divided by the sum of the gray values ​​of all pixels in the connected region to obtain the geometric center coordinates of each connected region, which are defined as the feature point mapping coordinates. The feature point mapping coordinates are binary real number pairs consisting of horizontal and vertical axis values. All feature point mapping coordinates are indexed to obtain feature point index values. These index values ​​are then combined to obtain a feature point index sequence. The feature point index sequence is a sequence of natural numbers from 1 to K, where K is the total number of feature point mapping coordinates. A numerical matrix, defined as a distance correlation matrix, is constructed based on the feature point index sequence. The rows and columns of the distance correlation matrix are set according to the total number of feature point mapping coordinates, i.e., the distance correlation matrix is ​​a K-row, K-column numerical matrix. The Euclidean distance between any two feature point mapping coordinates is calculated. Based on the feature point index values ​​corresponding to the two feature point mapping coordinates in the feature point index sequence, the matrix index in the distance correlation matrix is ​​determined, and the Euclidean distance is filled into the corresponding matrix index position. The feature point mapping coordinates, the distance correlation matrix, and the corresponding center strength values ​​of the internal feature points are logically encapsulated to generate a feature point spatial distribution topology map. The function of this feature point spatial distribution topology map is to transform the discrete pixels inside the encoded microsphere into a digital graph structure with spatial causal constraints. See also... Figure 2This is a schematic diagram illustrating the logic for generating a spatial distribution topology map of feature points according to an embodiment of the present invention. A pixel coordinate system is formed by the mutually perpendicular horizontal axis (U-axis) and vertical axis (V-axis), with the origin represented by O. The dashed circle represents the digital contour boundary of the coded microsphere identified by the grayscale step value of the microsphere boundary. Solid points P1, P2, and P3 represent the corresponding three feature point mapping coordinates. In the feature point index sequence, the feature point mapping coordinates P1, P2, and P3 are each assigned a unique feature point index value. The blue line segment L12 connecting P1 and P2, and the blue line segment L23 connecting P2 and P3, represent the topological connection relationship stored in the distance correlation matrix, characterizing the spatial distance constraints between the feature point mapping coordinates. The lengths of line segments L12 and L23 correspond to the Euclidean distance stored in the distance correlation matrix. By displaying the relative geometric configuration between the feature point mapping coordinates P1, P2, and P3, the projection state of the center intensity value of the internal feature points inside the coded microsphere under a specific rotational posture is simulated.

[0027] Step S12: Call the feature point spatial distribution topology map to perform three-dimensional spatial attitude inversion and obtain the three-dimensional attitude representation vector.

[0028] After generating the topological map of the spatial distribution of feature points, the random rotation of the coded microsphere in three-dimensional space causes nonlinear contraction or stretching of the relative distance between the mapped coordinates of feature points on the two-dimensional projection plane. This geometric distortion implicitly contains the spatial orientation information of the coded microsphere. To reversely recover the true spatial orientation of the coded microsphere within the microchannel and resolve the technical contradiction of inaccurate decoding of encoded information caused by dimensional collapse during the projection of a three-dimensional physical entity onto a two-dimensional image, this step establishes a computational model capable of calculating rotational degrees of freedom from the distorted topological configuration. The aim is to transform the geometric displacements in the topological map of the spatial distribution of feature points into physically deterministic three-dimensional attitude parameters.

[0029] Specifically, a reference feature point set is obtained. This reference feature point set refers to the set of three-dimensional physical coordinates formed by the standard coordinate points that correspond one-to-one with the feature point index values ​​in the feature point index sequence when the coded microsphere is in its zero-position attitude in the three-dimensional spatial coordinate system. The reference feature point set represents the theoretical topological parameters of the coded microsphere during the manufacturing stage. The three-dimensional spatial coordinate system consists of three mutually perpendicular coordinate axes. For example, the X-axis, Y-axis, and Z-axis are used to represent the three-dimensional spatial coordinate system, and the Z-axis is defined as parallel to the acquisition optical axis of the high-speed imaging interface. The acquisition optical axis is the physical central axis connecting the center of the optical lens system of the high-speed imaging interface and the center of the photosensitive surface of its image sensor, representing the direction of the central optical path from the detection area into the imaging device. An observation coordinate matrix is ​​constructed using the feature point mapping coordinates. The number of rows in the observation coordinate matrix is ​​equal to the total number of feature point index sequences, and the number of columns is 2. The first column vector of the observation coordinate matrix consists of the horizontal axis values ​​of the mapping coordinates of each feature point in the feature point index sequence, and the second column vector consists of the vertical axis values ​​of the mapping coordinates of each feature point in the feature point index sequence. A reference coordinate matrix is ​​constructed using a reference feature point set. The reference coordinate matrix has K rows and 3 columns. The first, second, and third column vectors of the reference coordinate matrix are each composed of the three physical coordinate values ​​of the corresponding standard coordinate points in the reference feature point set.

[0030] Calculate the arithmetic mean of the first and second columns of the observed coordinate matrix to obtain the mean components of the horizontal and vertical axes, respectively. Then calculate the arithmetic mean of the first, second, and third columns of the reference coordinate matrix to obtain the first, second, and third spatial mean terms, respectively. Perform subtraction operations on the observed coordinate matrix based on the horizontal and vertical mean components. Specifically, subtract the horizontal mean component from the first column vector of the observed coordinate matrix to obtain the decentralized horizontal axis component. Subtract the vertical mean component from the second column vector of the observed coordinate matrix to obtain the decentralized vertical axis component. Using the decentralized horizontal axis component as the first column and the decentralized vertical axis component as the second column, merge the column vectors to obtain the decentralized observed matrix. Perform subtraction operations on the reference coordinate matrix based on the first, second, and third spatial mean terms. Specifically, subtract the first spatial mean term from the first column vector of the reference coordinate matrix to obtain the reference first spatial decentralized component. Subtracting the second spatial mean term from the second column vector of the reference coordinate matrix yields the decentralized component of the second reference space. Subtracting the third spatial mean term from the third column vector of the reference coordinate matrix yields the decentralized component of the third reference space. Reconstructing the matrix using the decentralized components of the first, second, and third reference spaces according to their corresponding column indices yields the decentralized reference matrix. The subtraction operation is based on the principle of eliminating the interference of the translational displacement of the encoded microsphere within the microchannel on attitude calculation by translating the coordinate set to its respective geometric center, ensuring that subsequent calculations are unaffected by the translational position. Multiplying the transpose of the decentralized reference matrix with the decentralized observation matrix generates the attitude covariance matrix. The attitude covariance matrix quantifies the spatial rotation correlation between the decentralized feature point mapping coordinates and the reference feature point set.

[0031] The singular value decomposition (SVD) algorithm is invoked to perform matrix factorization on the attitude covariance matrix, outputting a left singular vector matrix, a singular diagonal matrix, and a right singular vector matrix. The rationale for invoking SVD is to extract orthogonal basis components from the attitude covariance matrix, thereby mathematically aligning the principal components of the observation space with the reference space. The left singular vector matrix is ​​an orthogonal matrix containing multiple column components obtained after SVD of the attitude covariance matrix. Each column component in the left singular vector matrix is ​​defined as a left singular vector, representing the spatial baseline distribution of the reference feature point set after decentralization. The singular diagonal matrix is ​​a diagonal matrix where the elements on the main diagonal are non-negative real numbers and all other elements are zero. Each non-negative real number on the main diagonal is defined as a singular value, used to quantify the similarity strength between the feature point spatial distribution topology and the reference feature point set along the principal component directions. The right singular vector matrix is ​​another orthogonal matrix containing multiple column components, obtained by performing singular value decomposition on the attitude covariance matrix. Each column component in the right singular vector matrix is ​​defined as a right singular vector, used to characterize the projection distribution characteristics of the feature point mapping coordinates after decentralization. Matrix multiplication is performed between the left singular vector matrix and the transpose of the right singular vector matrix to generate the rotation transformation matrix. The rotation transformation matrix is ​​a 3x3 square matrix satisfying orthogonality constraints, representing the spatial rotation transformation process experienced by the coded microsphere as it evolves from its zero-position attitude to its current observed projection state. The extraction of the rotation transformation matrix by calling the singular value decomposition algorithm is based on orthogonal Pluke analysis logic, used to determine the spatial mapping relationship under the constraint of minimizing the residual between the observed coordinate matrix and the reference coordinate matrix.

[0032] An angular decomposition of the rotation transformation matrix is ​​performed using the arctangent and arcsine functions to obtain the angular values ​​representing the spatial orientation of the coded microsphere. Specifically, the pitch angle is obtained by calculating the arctangent ratio of the matrix element value in the second row and third column to the matrix element value in the third row and third column; the yaw angle is obtained by calculating the arcsine of the negative value of the matrix element value in the first row and third column; and the roll angle is obtained by calculating the tangent ratio of the matrix element value in the first row and second column to the matrix element value in the first row and first column. The pitch, yaw, and roll angles are combined to generate a three-dimensional attitude representation vector. The pitch angle represents the rotational deflection of the coded microsphere about the U-axis in the pixel coordinate system, the yaw angle represents the rotational deflection of the coded microsphere about the V-axis in the pixel coordinate system, and the roll angle represents the rotational deflection of the coded microsphere about an axis perpendicular to the pixel coordinate plane. The three-dimensional attitude representation vector quantifies the real-time spatial attitude parameters of the coded microsphere at the instant of detection. To achieve a concrete representation of the spatial orientation of the coded microsphere, an attitude vector is constructed based on a three-dimensional attitude representation vector. Specifically, an initial vector is defined, parallel to the Z-axis and originating from the origin of the three-dimensional coordinate system. A rotation transformation matrix is ​​used to perform a rotation transformation operation on this initial vector, resulting in a spatial vector with a definite pointing offset in the three-dimensional coordinate system, defined as the attitude vector. The three-dimensional attitude representation vector establishes a coordinate transformation criterion between the reference feature point set and the mapped coordinates of the feature points through the rotation transformation matrix. When the angle values ​​in the three-dimensional attitude representation vector change, the rotation transformation matrix changes accordingly, causing varying degrees of distance contraction or stretching of the standard coordinate points in the reference feature point set when orthogonally projected onto the pixel coordinate plane, thus forming geometric distortion features in the spatial distribution topology of the feature points. By measuring the degree of deflection of the attitude vector relative to the Z-axis, the longitudinal deviation span of each feature region on the surface of the coded microsphere relative to the pixel coordinate plane can be quantified; see [link to relevant documentation]. Figure 3 This is a logical schematic diagram illustrating the relationship between a three-dimensional attitude representation vector and spatial mapping, provided by an embodiment of the present invention. A three-dimensional spatial coordinate system is exemplarily shown using the X, Y, and Z axes, with the sphere representing the coded microsphere. The red arrows emanating from the center of the sphere represent attitude vectors, and the spatial direction of these vectors is uniquely determined by the three-dimensional attitude representation vector. Below, a pixel coordinate system is shown. The dashed lines in the diagram represent specific red feature points on the surface of the coded microsphere, i.e., the paths through which standard coordinate points are vertically projected onto the pixel coordinate system plane. The bottom of the dashed lines terminates at a solid red dot inside the pixel coordinate system plane, representing the mapped coordinates of the feature points. This visually demonstrates how the three-dimensional attitude representation vector controls the positional distribution of the projection points through a rotation transformation matrix, thereby forming a topological map of the spatial distribution of feature points with specific geometric configurations in the digital space.

[0033] Step S13: Calculate the longitudinal depth deviation based on the three-dimensional pose representation vector, and use the longitudinal depth deviation to construct the feature point diffusion function.

[0034] After obtaining the three-dimensional attitude representation vector, due to the rotation of the coded microsphere, different feature regions on the surface of the microsphere will be displaced along the Z-axis in the three-dimensional spatial coordinate system, deviating from the ideal focal plane of the high-speed imaging interface. This spatial displacement will cause non-stationary diffusion of the optical signal in the pixel coordinate plane, resulting in edge blurring of pixels around the feature point mapping coordinates and inducing pixel energy statistical deviation. In order to capture and quantitatively describe the energy spillover characteristics caused by defocusing, a dynamic degradation model based on three-dimensional spatial position feedback is established. The aim is to simulate the spatial evolution of the optical signal after deviating from the focal plane.

[0035] Specifically, the geometric radius of the coded microsphere is obtained. This geometric radius is a constant determined based on the physical specifications of the coded microsphere during manufacturing, set according to its geometric outer radius in three-dimensional space; for example, it is set to 15 micrometers. The offset of the coded microsphere's center of gravity in the Z-axis direction is also obtained. This offset refers to the physical distance between the geometric center of the coded microsphere and the ideal focal plane, determined by autofocus feedback data from the high-speed imaging interface. The feature point index sequence is traversed, and for each feature point index value, the following processing is performed, defining the feature point mapping coordinates corresponding to the currently processed feature point index value as the target feature point mapping coordinates. Specifically, trigonometric function operations are performed using the yaw angle combined with the geometric radius to calculate the vertical displacement of different regions on the surface of the coded microsphere relative to the pixel coordinate system plane. The sine value of the yaw angle is calculated, and an arithmetic multiplication operation is performed between the sine value and the geometric radius to obtain the local surface height deviation. The offset and the local surface height deviation are summed to obtain the longitudinal depth deviation. The basis for performing the accumulation operation is as follows: Utilizing the geometric mapping principle in spherical coordinates, the rigid translation of the coded microsphere is linearly superimposed with the local deformation caused by rotation, thereby completely quantifying the true physical distance of a specific feature region on the coded microsphere from the ideal focal plane in the depth direction of the three-dimensional spatial coordinate system. The longitudinal depth deviation refers to the signed spatial distance of the feature region on the coded microsphere relative to the ideal focal plane, used to characterize the degree of defocusing of the feature region in the Z-axis direction of the three-dimensional spatial coordinate system. An optical system transfer constant is set. This optical system transfer constant is a physical scaling factor determined based on the numerical aperture of the high-speed imaging interface and the acquisition wavelength, set based on the rate of change of the blur spot size per unit defocusing distance; for example, it is set to 0.15. The square of the longitudinal depth deviation is calculated. Then, an arithmetic multiplication operation is performed between the squared value and the optical system transfer constant to obtain the energy diffusion radius R. The multiplication operation is performed based on Fresnel diffraction theory and the paraxial approximation condition, which states that the radius of the blur circle of a point source signal on the off-focal plane is linearly correlated with the square of the defocus distance. This quantifies the physical distribution scale of the energy of a single pixel affected by defocus, which evolves into a blurry spot within the pixel coordinate plane. The energy diffusion radius refers to the physical distribution scale of the lateral diffusion of the point source signal energy within the pixel coordinate plane caused by defocus, and is used to quantify the geometric radius of a single feature point after its mapped coordinates evolve into a blurry spot.

[0036] A second-order Gaussian kernel function, defined as the feature point diffusion function H, is constructed using the energy diffusion radius to characterize the degradation features of spatial energy distribution. Specifically, a window scaling factor is set. The window scaling factor is set to reduce computational complexity while ensuring the integrity of the energy envelope; for example, it is set to 3. An arithmetic multiplication operation is performed between the window scaling factor and the energy diffusion radius, followed by rounding up to obtain the radius half-width value. A local region containing multiple pixels is defined in the pixel coordinate plane, centered on the target feature point mapping coordinates; this region is defined as the matrix window. For example, a rectangular pixel region with a radius half-width value plus one, whose span in both the horizontal and vertical directions is twice the radius half-width value, is selected as the matrix window. The setting of twice the radius half-width value plus one ensures that the side length of the matrix window is odd, thus giving the matrix window a unique center pixel, achieving precise pixel-level alignment between the geometric center of the matrix window and the target feature point mapping coordinates. The pixel plane range contained in the matrix window is defined as the support region. The support region defines the computational boundary of the dynamic feature point diffusion function. Each pixel within the supported region is defined as a sampling pixel. The 2D pixel coordinate offset is then obtained. Specifically, the difference between the horizontal coordinate of the sampled pixel in the pixel coordinate system plane and the horizontal coordinate of the corresponding target feature point is calculated to obtain the two-dimensional pixel coordinate offset. Calculate the difference between the ordinate of the sampled pixel in the pixel coordinate plane and the ordinate of the corresponding target feature point's mapped coordinate to obtain the two-dimensional pixel coordinate offset. Using two-dimensional pixel coordinate offset as Example feature point spread function, then in Feature point spread function H The calculation formula is: H Where 2π is a constant, The term represents an exponential function; the logic behind the calculation formula is as follows: It utilizes the probability density characteristics of a two-dimensional normal distribution to simulate the energy evolution of the light field on the non-focal plane. Specifically, the numerator of the exponential term... The squared Euclidean distance used to calculate the mapped coordinates of the sampled pixel from the target feature point reflects the radial symmetry of energy decay with increasing distance; the denominator term in the formula... Used to control the rate of energy decay, it converts the defocusing caused by longitudinal depth deviation into the degree of blurring of feature points in lateral space, in the formula. The setting is based on ensuring that the integral value of the feature point diffusion function is one throughout the entire support region, thus following the law of conservation of energy at the mathematical level and ensuring that the total energy of the feature points remains consistent before and after diffusion.

[0037] Step S10 addresses the technical challenge of traditional detection techniques that treat the coded microsphere as an isotropic point model, neglecting the three-dimensional rotational degrees of freedom and thus failing to distinguish between steric hindrance attenuation and low-concentration reduction. This is achieved through a topological map of the spatial distribution of feature points, a three-dimensional attitude representation vector, longitudinal depth deviation, and a feature point diffusion function. This transforms passive, undifferentiated pixel statistics into active attitude recognition with spatial orientation. Specifically, the topological map of the spatial distribution of feature points transforms discrete pixels into a digital graph structure with spatial causal constraints, providing structured input for attitude inversion; the three-dimensional attitude representation vector quantifies the real-time spatial attitude parameters of the coded microsphere at the moment of detection; and the feature point diffusion function utilizes the two-dimensional normal distribution characteristics to simulate the energy evolution of the light field on the non-focal plane.

[0038] Step S20: Spatial deconvolution processing is performed on the original image of the microsphere using the feature point diffusion function to generate a spatial consistency restoration image. Visibility weights are calculated based on the longitudinal depth deviation, and brightness gain compensation is performed on the spatial consistency restoration image using the visibility weights to obtain the intrinsic feature intensity.

[0039] Further, step S20 includes: Step S21: Perform spatial deconvolution processing on the original image of the microsphere using the feature point diffusion function to generate a spatial consistency repair image.

[0040] After obtaining the feature point diffusion function corresponding to the mapping coordinates of each feature point, the defocusing effect causes the energy that should have converged at the target feature point mapping coordinates to be incorrectly distributed to the sampling pixels within the support region. This unexpected change in energy distribution directly weakens the central intensity value of the internal feature points. To achieve pixel-level energy realignment and eliminate the intensity attenuation artifacts caused by spatial rotation, an optical degradation model is used to perform inverse compensation on the original image of the microsphere.

[0041] Specifically, the original image of the microsphere is acquired. Centered on the target feature point mapping coordinates, a local pixel matrix with the same size as the supporting region is extracted from the original image of the microsphere, defined as the observation intensity matrix. A correction intensity matrix is ​​constructed based on the observation intensity matrix. Specifically, a blank matrix with the exact same number of rows and columns as the observation intensity matrix is ​​created, and the gray values ​​of each pixel in the observation intensity matrix are filled into the blank matrix according to their positional correspondence, resulting in the correction intensity matrix. To achieve reverse repair of the observation intensity matrix, a deconvolution algorithm based on the maximum likelihood estimation criterion is invoked. Specifically, a loop variable is set, increasing from 1 to a preset number of iterations, and the current value of the loop variable is defined as the iteration period. The number of iterations is set based on achieving a balance between the convergence speed of the deconvolution algorithm and the accuracy of feature repair; for example, it is set to 15 iterations. Within the current iteration period, an update operation is performed. Specifically, the correction intensity matrix and the feature point diffusion function are convolved to obtain a diffusion matrix. The observation intensity matrix and the diffusion matrix are then divided arithmetically to obtain an energy ratio matrix. The feature point spread function is rotated 180 degrees to obtain the adjoint kernel matrix. The energy ratio matrix and the adjoint kernel matrix are convolved to obtain the corrected gradient matrix. The corrected intensity matrix and the corrected gradient matrix are then multiplied arithmetically to obtain the updated corrected intensity matrix, defined as the iterative correction matrix, which is used as the input for the next iteration. The basis for this iterative calculation is: using the feature point spread function to simulate the formation process of the observed intensity matrix, and by continuously correcting the corrected intensity matrix, minimizing the residual between the diffusion matrix and the observed intensity matrix, thereby guiding the diffused energy to perform probabilistic regression to the target feature point mapping coordinates. After completing the required number of iterations for the repair, the pixel values ​​at the geometric center of the iterative correction matrix are extracted to replace the corresponding grayscale values ​​at the target feature point mapping coordinates in the original microsphere image, completing the intensity update for all feature point index sequences and obtaining a spatially consistent repaired image. The spatial consistency restoration image refers to a digital image that eliminates the defocus blur and energy diffusion characteristics caused by the random rotation of the coded microsphere, and restores the pixel energy of internal feature points in the pixel dimension. Its function is to eliminate signal representation deviations caused by spatial pose variables, ensuring that the coded microsphere possesses a consistent characteristic brightness response under different rotational postures in the three-dimensional spatial coordinate system. See also Figure 4This is a schematic diagram of spatial deconvolution energy repositioning logic provided by an embodiment of the present invention. The left side illustrates the spatial distribution of pixel energy in the observation intensity matrix. The solid black dot at the geometric center represents the target feature point mapping coordinates; the red shaded area surrounding the target feature point mapping coordinates represents the energy diffusion range caused by defocusing due to random rotation. The distribution of the observation intensity matrix intuitively reflects the technical defect of energy that should have converged at the target feature point mapping coordinates overflowing to the surrounding sampled pixels, resulting in blurred edges in the feature representation. The right side shows the local energy repositioning state corresponding to the observation intensity matrix in the spatial consistency restoration image. The solid black dot at the center represents the energy cluster formed after the diffused energy originally located in the red shaded area on the left side is re-probabilistically regressed to the target feature point mapping coordinates after the deconvolution algorithm is executed. By eliminating the red shaded area, the right image demonstrates that the spatial consistency restoration image has completed energy repositioning restoration in the pixel dimension, refocusing the pixel energy of the internal feature points to the target feature point mapping coordinates, thereby restoring the intrinsic physical intensity representation.

[0042] Step S22: Calculate the visibility weight based on the longitudinal depth deviation, and call the visibility weight to perform brightness gain compensation on the spatial consistency repair image to obtain the intrinsic feature intensity.

[0043] Spatial consistency restoration corrects the diffuse distribution of pixel energy at the geometric level. To further eliminate intensity acquisition differences in the Z-axis direction of the optical system in the three-dimensional spatial coordinate system, it is necessary to construct visibility weights to characterize the differences in luminous efficiency of feature points, in order to compensate for the signal intensity asymmetry caused by different longitudinal depths.

[0044] Specifically, the longitudinal depth deviation is obtained. An attenuation correction coefficient is set. This attenuation correction coefficient is a compensation constant determined based on the high-speed imaging interface, and its setting is based on the fluorescence photon escape loss rate caused by unit depth displacement; for example, it is set to 0.045. The arithmetic product of the longitudinal depth deviation and the attenuation correction coefficient is calculated to obtain a phase compensation term. An exponential operation is performed based on the phase compensation term to obtain the visibility weight W. The phase compensation term is used to quantify the intermediate physical parameter of the energy loss increment induced by the feature point deviating from the ideal focal plane, which conforms to the exponential attenuation characteristic. The visibility weight refers to the normalized coefficient for reverse gain compensation of the original luminous efficiency of each feature region based on the longitudinal depth deviation. The formula for calculating the visibility weight is: Where e is a constant, This is the attenuation correction factor. The longitudinal depth deviation is calculated using the inverse mapping logic of the Beer-Lambert law to compensate for energy loss during the penetration of excitation and emission light into the encoded microsphere polymer matrix. Specifically, the absolute value of the longitudinal depth deviation... The physical path length of photon propagation within the coded microsphere is characterized; the attenuation correction coefficient characterizes the scattering and absorption probability of photons by the medium per unit path length; by performing arithmetic multiplication of the attenuation correction coefficient with the absolute value of the longitudinal depth deviation, the total energy loss in logarithmic space, i.e., the phase compensation term, is obtained; by performing an exponential operation on the phase compensation term using the natural constant e, the energy loss in logarithmic space is transformed into a gain factor in linear space, thereby obtaining the visibility weight. Through the inverse compensation effect of the visibility weight, it is ensured that internal feature points at different depth positions can be mapped to a unified brightness reference scale, eliminating the interference of spatial pose variables on the quantitative accuracy of the signal.

[0045] Obtain the grayscale values ​​corresponding to the target feature point mapping coordinates in the spatial consistency restoration image. Perform an arithmetic multiplication operation between the grayscale values ​​corresponding to the target feature point mapping coordinates in the spatial consistency restoration image and the corresponding visibility weight to obtain the intrinsic feature intensity. The intrinsic feature intensity eliminates the effects of defocus blur and depth attenuation caused by the random rotation of the coded microsphere, serving as the digital identity identifier of the coded microsphere.

[0046] Step S20 addresses the challenges of defocus blurring, energy diffusion, and signal intensity asymmetry induced by feature points deviating from the ideal focal plane caused by the random rotation of the coded microsphere, through spatial consistency image restoration, visibility weight, and intrinsic feature intensity. This achieves pixel-level energy repositioning and consistent feature brightness response at different depths. Specifically, the spatial consistency image restoration eliminates energy spillover caused by defocusing, allowing pixel energy at internal feature points to be repositioned at the pixel level. Visibility weight, through inverse gain compensation, eliminates the energy loss increment induced by feature points deviating from the ideal focal plane. Intrinsic feature intensity eliminates the interference of spatial pose variables on signal quantitative accuracy, serving as a digital identifier for the coded microsphere's consistent response.

[0047] Step S30: Construct a channel intermodulation matrix using the intrinsic feature strength, perform signal decoupling operation using the channel intermodulation matrix to obtain the decoupling strength, and perform digital encapsulation based on the decoupling strength to obtain a quantized output dataset.

[0048] Further, step S30 includes: Step S31: Construct the channel intermodulation matrix using the intrinsic feature strengths.

[0049] After obtaining the intrinsic feature intensities corresponding to the mapped coordinates of all target feature points in the coded microsphere, the energy of the strong signal detection channel will interfere with the weak signal detection channel through spectral leakage due to the overlapping spectral sensitivity ranges of different detection channels. To accurately reconstruct the true signal values ​​of each detection channel, a channel intermodulation matrix is ​​constructed.

[0050] Specifically, the detection channel refers to an independent optical signal acquisition window in the high-speed imaging interface, defined by a specific band filter and a photodetector. The total number of detection channels is determined by the total number of sensor hardware configurations in the high-speed imaging interface, denoted by N.

[0051] A sequence of natural numbers, increasing from 1 to the total number of detection channels N, is defined as the detection channel index sequence, and i is used as the detection channel index value in the detection channel index sequence. That is, i ranges from 1 to N. A fluorescent probe is defined as a fluorescent chemical group loaded on the surface of an encoded microsphere to label specific exosome markers and emit a light signal of a specific wavelength upon stimulation. The total number of fluorescent probes is equal to the total number of types of target exosome markers to be detected, denoted by M. A sequence of natural numbers, increasing from 1 to the total number of fluorescent probes M, is defined as the fluorescent probe index sequence, and j is used as the fluorescent probe index value in the fluorescent probe index sequence, that is, j ranges from 1 to M. For each detection channel index value, the corresponding filter transmission function is obtained. This filter transmission function is an inherent optical parameter provided by the filter hardware manufacturer, characterizing the wavelength transmittance of the physical medium. For each fluorescent probe index value, the corresponding emission spectral distribution function is obtained. This emission spectral distribution function is determined by the chemical molecular structure of the fluorescent probe, characterizing the energy distribution with wavelength. A full-band range is defined, which refers to the spectral wavelength range that the high-speed imaging interface's optical detection system can effectively sense and completely cover the transmission bands of all detection channel filters and the emission bands of all fluorescent probes. This setting is based on ensuring that all potential spectral leakage energy is included in the integration calculation domain. For example, the starting wavelength of the full-band range is set to 350 nm, and the ending wavelength is set to 900 nm. The overlap integral value of the filter transmission function and the emission spectral distribution function over the full-band range is calculated using definite integral operations, and this value is defined as the spectral cross-induction coefficient of the j-th fluorescent probe to the i-th detection channel. All spectral cross-induction coefficients form an N x M matrix, defining the spectral sensing matrix.

[0052] The intrinsic feature intensities are multiplied with the spectral sensing matrix. Specifically, a column vector of length M, composed of the intrinsic feature intensities of each detection channel, is linearly mapped to the transpose of the spectral sensing matrix, outputting an energy penetration amplitude column vector of length N. This energy penetration amplitude column vector represents the interference intensity and quantifies the parasitic grayscale increment that migrates from strong signal detection channels to weak signal detection channels due to spectral leakage in the digital space. Since the sensor exhibits nonlinear response characteristics under strong light excitation, a correction criterion is obtained by analyzing the sensor's dynamic response curve to correct interference deviations in the high-intensity range. The dynamic response curve is a continuous function representing the physical mapping relationship between the sensor's output grayscale value and the input light radiation intensity. In the dynamic response curve, the response slope is defined as the first derivative of the output grayscale value with respect to the input light intensity, representing the detector's real-time sensing sensitivity. When the input light intensity continues to increase until the sensor reaches its full-scale saturation limit, the response slope will drop to zero. The full-scale saturation limit is the maximum digitized grayscale value that the imaging system can output, limited by the number of bits in the ADC. The nonlinear response range is defined by comparing the positional relationship between the pre-calibrated actual sampling points and the ideal linear extension line. The actual sampling points refer to the set of real digital grayscale values ​​output by the imaging sensor for standard light intensity inputs of different gradients under a controlled calibration environment. Each actual sampling point consists of a known input intrinsic feature intensity and its corresponding original output grayscale value of the sensor, which together constitute the original discrete dataset describing the physical sensing characteristics of the sensor. This represents the trajectory of the output grayscale value of the sensor increasing linearly with the intensity of the input intrinsic feature under ideal conditions where there is no energy loss or charge transfer distortion. Specifically, the nonlinear response interval refers to the deviation of the actual sampling point from the ideal linear extension line. The deviation is obtained by calculating the ratio of the absolute value of the difference between the actual output grayscale value and the ideal linear output value under the same input light intensity to the ideal linear output value, and is presented as a percentage. The nonlinear response interval is a continuous numerical region where the deviation is greater than a preset linear threshold and the rate of change of the response slope is no longer zero. The rate of change of the response slope is the second derivative of the output grayscale value with respect to the input light radiation intensity. The linear threshold is set as the maximum allowable linear deviation of the sensor. After defining the nonlinear response interval, a correction weighting factor is set according to the nonlinear response interval in which the intrinsic feature intensity value is located. The correction weighting factor refers to the proportional coefficient used to compensate for the amplitude of nonlinear distortion in the detection channel. It represents the energy correction gain required to restore the nonlinear response to the ideal linear state. Its setting is based on the quotient of the theoretical output value on the ideal linear extension line and the actual output gray value in the corresponding actual sampling point. The channel intermodulation matrix is ​​obtained by performing a weighted dot product operation on the energy penetration amplitude and the correction weighting factor.The channel intermodulation matrix achieves precise stripping of complex spectral leakage signals under high illumination by providing each detection channel with an interference cancellation benchmark that dynamically evolves with signal intensity. See also. Figure 5 This diagram illustrates the definition of the nonlinear response interval provided in this embodiment of the invention. The horizontal axis represents the true value of the input intrinsic feature intensity, and the vertical axis represents the sensor output grayscale value. It shows a series of actual sampling points represented by discrete solid circles, and an ideal linear extension line extending beyond the origin. This ideal linear extension line exhibits a linear growth trajectory, while the actual sampling points gradually bend downwards after crossing a critical point. The critical point is the starting position where the actual sampling trajectory deviates downwards relative to the ideal linear reference. The shaded area in the diagram represents the nonlinear response interval, defined by the deviation amplitude corresponding to the longitudinal displacement between the actual sampling point and the ideal linear extension line, and a preset linear threshold. The horizontal line at the top of the diagram represents the upper limit of full-scale saturation where the response slope drops to zero, indicating the physical limit of the sensor's charge collection capability and illustrating the evolution from linear response to nonlinear distortion.

[0053] Step S32: Perform signal decoupling operation using the channel intermodulation matrix to obtain the decoupling strength. The channel intermodulation matrix is ​​used as a physical quantization benchmark to perform signal decoupling operations. Specifically, the values ​​corresponding to the index positions of each matrix in the channel intermodulation matrix are defined as interference components. The column vector composed of the intrinsic characteristic intensities of each detection channel is subtracted from the corresponding interference components in the channel intermodulation matrix to obtain the corrected intrinsic characteristic intensities, which are defined as the decoupling intensity. The decoupling intensity eliminates the spurious brightness gain induced by spectral overlap and hardware nonlinear response, and can accurately reflect the intrinsic fluorescence emission level of the target exosome marker. The signal decoupling operation is constructed based on the principle of linear energy conservation and the inverse mapping compensation logic of nonlinear response. In a multicolor fluorescence detection system, due to the overlap of spectral sensitivities of each detection channel, the original signal value captured by a single detector is a coupling amount formed by the linear superposition of the intrinsic emission energy of the target fluorescent probe and the spectral leakage energy from adjacent detection channels. By using arithmetic subtraction, the parasitic energy generated by spectral leakage is removed from the original signal containing spectral interference, thereby achieving algebraic stripping of the physically superimposed energy in the digital space.

[0054] Based on the decoupling strength, a concentration mapping function is constructed to perform quantitative analysis. Specifically, a biochemical reaction is performed using standard exosome samples with known concentration sequences and coded microspheres, and raw images of the microspheres at each concentration gradient are acquired through a high-speed imaging interface. The decoupling strength corresponding to the raw images of the microspheres at each concentration gradient is obtained, and the least squares method is used to regress the decoupling strength and the concentration values ​​of the standard exosome samples with the concentration sequence, thereby obtaining the concentration mapping function. The concentration mapping function is a mathematical expression representing the evolutionary relationship between decoupling strength and quantified abundance, and its geometric form is a standard response curve. The quantified abundance refers to a quantitative indicator representing the true physical content of the target exosome biomarker loaded on the surface of the coded microspheres. The concentration mapping function includes model feature parameters used to describe the geometric characteristics of the standard response curve, which consist of a regression slope and an intercept term. The regression slope refers to the inclination of the standard response curve in a rectangular coordinate system, and physically represents the sensitivity of the high-speed imaging interface to a specific target exosome biomarker, i.e., the concentration change increment corresponding to a unit intensity change. The intercept term refers to the cutoff value of the standard response curve on the vertical axis, which is used to characterize the inherent bias of the system caused by the background of biochemical reactions or the dark current of the optical detection system.

[0055] In the actual detection process, the regression slope and intercept term matching the current fluorescent probe index value are retrieved. Based on the decoupling intensity corresponding to each fluorescent probe index value, calculations are performed using a concentration mapping function to obtain the quantitative abundance of each target exosome biomarker encoded on the microsphere surface. Specifically, the decoupling intensity and regression slope are arithmetically multiplied, and the resulting product is arithmetically added to the intercept term to output the quantitative abundance. The quantitative abundance corresponding to all fluorescent probes is digitally encapsulated to generate a quantitative output dataset containing exosome molecular expression characteristics. This quantitative output dataset refers to a structured digital dataset that integrates the encoded microsphere identity information and the expression levels of multiple biomarkers. As the final carrier reflecting the biological state of the sample under test, it contains high-fidelity molecular distribution information after physical attitude correction, optical defocus compensation, and spectral nonlinear decoupling, and is output in the form of a structured report to achieve accurate typing and content analysis of exosome biomarkers, completing the entire analytical process from digital raw images to high-fidelity biological quantitative information.

[0056] Step S30 addresses the challenges of energy penetration caused by channel spectral overlap in multicolor fluorescence detection systems, and incomplete interference removal due to nonlinear response distortion of the sensor under high-intensity excitation, through channel intermodulation matrices, decoupling strength, and concentration mapping functions. This achieves precise removal of complex parasitic energies and a closed-loop transformation from digital grayscale features to high-fidelity biological quantitative information. Specifically, the channel intermodulation matrix provides a dynamic interference elimination benchmark for each detection channel, evolving with signal intensity; the decoupling strength, based on the principle of linear energy conservation, algebraically removes the physically superimposed energy, accurately reflecting the intrinsic fluorescence emission level of the target analyte; and the concentration mapping function establishes a physical source mapping between digital grayscale values ​​and biomolecular abundance, ensuring the accuracy and repeatability of the quantitative output dataset through regression fitting.

[0057] Example 2 This embodiment, based on Embodiment 1, provides a multi-parameter combined quantitative analysis device for exosomes based on the fluorescence intensity of encoded microspheres, such as... Figure 6 As shown, it includes: The attitude topology module is used to acquire the original image of the coded microsphere, perform edge enhancement and feature spot clustering analysis on the original image of the microsphere to obtain a feature point spatial distribution topology map with spatial causal constraints, call the feature point spatial distribution topology map to perform three-dimensional spatial attitude inversion to obtain a three-dimensional attitude representation vector, calculate the longitudinal depth deviation based on the three-dimensional attitude representation vector, and use the longitudinal depth deviation to construct the feature point diffusion function. Light intensity correction module: Used to perform spatial deconvolution processing on the original image of microspheres using feature point diffusion function to generate spatial consistency restoration image, calculate visibility weight based on longitudinal depth deviation, and call visibility weight to perform brightness gain compensation on spatial consistency restoration image to obtain intrinsic feature intensity; Decoupling Quantification Module: This module is used to construct a channel intermodulation matrix using the intrinsic feature strength, perform signal decoupling operations using the channel intermodulation matrix to obtain the decoupling strength, and perform digital encapsulation based on the decoupling strength to obtain the quantized output dataset.

[0058] In addition, the parts of the technical solutions provided in the embodiments of this application that are consistent with the implementation principles of the corresponding technical solutions in the prior art have not been described in detail, so as to avoid excessive elaboration.

[0059] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the invention. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A multi-parameter combined quantitative analysis method for exosomes based on the fluorescence intensity of encoded microspheres, characterized in that, The method includes: The original image of the coded microsphere is obtained, and edge enhancement and feature spot clustering analysis are performed on the original image of the microsphere to obtain a feature point spatial distribution topology map with spatial causal constraints. The feature point spatial distribution topology map is called to perform three-dimensional spatial pose inversion to obtain a three-dimensional pose representation vector. The longitudinal depth deviation is calculated based on the three-dimensional pose representation vector, and the feature point diffusion function is constructed using the longitudinal depth deviation. Spatial deconvolution processing is performed on the original microsphere image using the feature point diffusion function to generate a spatial consistency restoration image. Visibility weights are calculated based on the longitudinal depth deviation, and brightness gain compensation is performed on the spatial consistency restoration image using the visibility weights to obtain the intrinsic feature intensity. The channel intermodulation matrix is ​​constructed using the intrinsic feature strength. The signal decoupling operation is performed using the channel intermodulation matrix to obtain the decoupling strength. Based on the decoupling strength, digital encapsulation is performed to obtain the quantized output dataset.

2. The multi-parameter joint quantitative analysis method for exosomes based on the fluorescence intensity of encoded microspheres according to claim 1, characterized in that, The method for obtaining the spatial distribution topology map of the feature points includes: The original image of the microsphere is acquired through a high-speed imaging interface. A pre-set second-order differential convolution kernel is used to perform a convolution operation with the original image of the microsphere. The feature response matrix is ​​obtained by traversing all pixels of the original image of the microsphere, and the response value corresponding to each pixel is determined. The response values ​​in the feature response matrix that are greater than a preset background threshold are identified, and the corresponding set of pixel coordinates is defined as a candidate point set. The candidate point set is then subjected to pixel clustering processing using the neighborhood connection criterion to obtain connected regions. A weighted average calculation is performed on the connected regions to obtain the geometric center coordinates of each connected region, which are defined as the feature point mapping coordinates. A distance correlation matrix is ​​constructed by calculating the Euclidean distance between the mapped coordinates of each feature point. The feature point mapped coordinates, the distance correlation matrix, and the original image of the microsphere are logically encapsulated to generate a topological map of the spatial distribution of feature points.

3. The multi-parameter combined quantitative analysis method for exosomes based on the fluorescence intensity of encoded microspheres according to claim 1, characterized in that, The three-dimensional attitude representation vector includes: Obtain the reference feature point set when the coded microsphere is in the zero position in the preset three-dimensional spatial coordinate system, construct the observation coordinate matrix using the feature point mapping coordinates, and construct the reference coordinate matrix using the reference feature point set; The mean of each column of the observation coordinate matrix and the reference coordinate matrix is ​​calculated separately, and then the centering process is performed to obtain the decentralized observation matrix and the decentralized reference matrix. The transpose of the decentralized reference matrix and the decentralized observation matrix are multiplied together to obtain the attitude covariance matrix. The singular value decomposition algorithm is used to perform matrix decomposition on the attitude covariance matrix, outputting a left singular vector matrix and a right singular vector matrix. The transpose of the left singular vector matrix and the right singular vector matrix are multiplied to generate a rotation transformation matrix. Angle decomposition is performed on the rotation transformation matrix to obtain a three-dimensional attitude representation vector composed of pitch angle, yaw angle and roll angle.

4. The multi-parameter combined quantitative analysis method for exosomes based on the fluorescence intensity of encoded microspheres according to claim 3, characterized in that, The three-dimensional attitude representation vector also includes: Obtain the acquisition optical axis of the high-speed imaging interface, define an initial vector parallel to the direction of the acquisition optical axis of the high-speed imaging interface, and perform a rotation transformation operation on the initial vector using a rotation transformation matrix to obtain a spatial vector with a definite pointing offset in the three-dimensional spatial coordinate system, which is defined as the attitude vector.

5. The multi-parameter combined quantitative analysis method for exosomes based on the fluorescence intensity of encoded microspheres according to claim 1, characterized in that, The longitudinal depth deviation includes: Obtain the geometric radius of the coded microsphere and the ideal focal plane of the high-speed imaging interface, and use the high-speed imaging interface to determine the offset of the geometric center of the coded microsphere relative to the ideal focal plane; By performing trigonometric function calculations using the yaw angle and geometric radius, the local height deviation on the surface is obtained. The offset and the local height deviation on the surface are then summed to obtain the longitudinal depth deviation.

6. The multi-parameter combined quantitative analysis method for exosomes based on the fluorescence intensity of encoded microspheres according to claim 5, characterized in that, The feature point diffusion function includes: The optical system transfer constant is set, and the square of the longitudinal depth deviation is multiplied by the optical system transfer constant to obtain the energy diffusion radius. Centered on the feature point mapping coordinates, a matrix window containing multiple pixels is determined within the original image of the microsphere, and the region contained in the matrix window is defined as the support region; Each pixel within the support area is defined as a sampling pixel, and the two-dimensional pixel coordinate offset of the sampling pixel relative to the feature point mapping coordinates is calculated. Utilizing the probability density characteristics of a two-dimensional normal distribution, a second-order Gaussian kernel function is constructed based on the energy diffusion radius and the two-dimensional pixel coordinate offset, defined as the feature point diffusion function, where the integral value of the feature point diffusion function over the support region is one.

7. The multi-parameter combined quantitative analysis method for exosomes based on the fluorescence intensity of encoded microspheres according to claim 6, characterized in that, The consistency-repaired image includes: A local pixel matrix with the same size as the support region is extracted from the original image of the microsphere with the feature point mapping coordinates as the center. This matrix is ​​defined as the observation intensity matrix. A matrix with the same size as the observation intensity matrix and all initial element values ​​are one is constructed. This matrix is ​​defined as the correction intensity matrix. The energy ratio matrix is ​​obtained by performing a convolution operation between the correction intensity matrix and the feature point spread function, and then calculating the ratio between the resulting convolution and the observed intensity matrix. Perform spatial flipping on the feature point spread function to obtain the adjoint kernel matrix. Perform convolution operation on the adjoint kernel matrix and the energy ratio matrix to obtain the corrected gradient matrix. Using the maximum likelihood estimation criterion, the correction intensity matrix is ​​updated by dot product using the modified gradient matrix, and the correction intensity matrix after a preset number of iterations is determined as the spatial consistency repair image.

8. The multi-parameter combined quantitative analysis method for exosomes based on the fluorescence intensity of encoded microspheres according to claim 1, characterized in that, The intrinsic feature strength includes: Set the attenuation correction coefficient, and perform an arithmetic multiplication operation between the longitudinal depth deviation and the attenuation correction coefficient to obtain the phase compensation term; Perform an exponential operation on the phase compensation term to obtain the visibility weight; The intrinsic feature intensity is obtained by performing inverse multiplication compensation on the mapping coordinates of corresponding feature points in the spatial consistency repair image using the visibility weight.

9. The multi-parameter combined quantitative analysis method for exosomes based on the fluorescence intensity of encoded microspheres according to claim 1, characterized in that, The channel intermodulation matrix includes: The independent optical signal acquisition window in the high-speed imaging interface is identified and defined as the detection channel, and the fluorescent chemical groups used to label exosome markers are identified and defined as fluorescent probes; Obtain the transmission function of the filter corresponding to the detection channel and the emission spectral distribution function corresponding to the fluorescent probe, calculate the overlap area of ​​the transmission function of the filter and the emission spectral distribution function, obtain the spectral cross-induction coefficient, and form a spectral induction matrix based on the spectral cross-induction coefficient; A linear mapping operation is performed using the intrinsic characteristic intensity and the spectral sensing matrix to output the energy penetration amplitude, and a correction weighting factor is set. By performing a weighted dot product operation on the energy penetration amplitude using a modified weighting factor, the channel intermodulation matrix is ​​obtained.

10. The multi-parameter combined quantitative analysis method for exosomes based on the fluorescence intensity of encoded microspheres according to claim 9, characterized in that, The decoupling strength includes: Extract the values ​​corresponding to the preset matrix index positions in the channel intermodulation matrix, and define them as interference components; The intrinsic feature intensities corresponding to each detection channel are combined into a column vector. The corresponding interference components are then subtracted from the column vector to eliminate spurious brightness gains caused by spectral overlap and sensor nonlinear response, thus obtaining the decoupling intensity.

11. The multi-parameter combined quantitative analysis method for exosomes based on the fluorescence intensity of encoded microspheres according to claim 10, characterized in that, The quantized output dataset includes: The decoupling strength was obtained by using standard exosome samples with known concentration sequences, and regression fitting was performed to obtain a concentration mapping function consisting of the regression slope and intercept term. The decoupling strength corresponding to the sample to be tested is used as the input variable and substituted into the concentration mapping function. The multiplication and addition operation is performed using the regression slope and intercept term to obtain the quantitative abundance. The quantitative abundance and the identity information of the encoded microspheres are digitally encapsulated to obtain a quantitative output dataset.

12. A multi-parameter joint quantitative analysis device for exosomes based on encoded microsphere fluorescence intensity, used to implement the multi-parameter joint quantitative analysis method for exosomes based on encoded microsphere fluorescence intensity as described in any one of claims 1-11, characterized in that, The system includes: The attitude topology module is used to acquire the original image of the coded microsphere, perform edge enhancement and feature spot clustering analysis on the original image of the microsphere to obtain a feature point spatial distribution topology map with spatial causal constraints, call the feature point spatial distribution topology map to perform three-dimensional spatial attitude inversion to obtain a three-dimensional attitude representation vector, calculate the longitudinal depth deviation based on the three-dimensional attitude representation vector, and use the longitudinal depth deviation to construct the feature point diffusion function. Light intensity correction module: Used to perform spatial deconvolution processing on the original image of microspheres using feature point diffusion function to generate spatial consistency restoration image, calculate visibility weight based on longitudinal depth deviation, and call visibility weight to perform brightness gain compensation on spatial consistency restoration image to obtain intrinsic feature intensity; Decoupling Quantification Module: This module is used to construct a channel intermodulation matrix using the intrinsic feature strength, perform signal decoupling operations using the channel intermodulation matrix to obtain the decoupling strength, and perform digital encapsulation based on the decoupling strength to obtain the quantized output dataset.

Images