Multiple microsphere flow type fluorescence detection method for vascular inflammation markers

Abstract

The invention provides a multi-microsphere flow type fluorescence detection method for vascular inflammation markers, and belongs to the technical field of multi-microsphere flow type fluorescence detection.The multi-microsphere flow type fluorescence detection method comprises the steps that capture antibody coupling microsphere mixed liquor with different fluorescence coding characteristics is prepared; the method comprises the following steps: sequentially reacting with a to-be-detected sample, a biotin labeled detection antibody and phycoerythrin labeled streptavidin to form a sandwich immune complex, collecting a fluorescence signal by using a flow cytometry, and then carrying out self-adaptive optimization treatment on the original fluorescence signal by using a liquid dynamic correction model, according to the dynamic response characteristics of liquid neurons in the liquid reservoir layer, accurate concentration results of the five vascular inflammation markers are calculated by fitting a standard curve through a logistic regression model, and the technical problem that the multiple detection precision of the vascular inflammation markers is insufficient due to limited dynamic range of fluorescence signals and serious background noise interference is solved.

Description

Technical Field

[0001] This invention belongs to the field of multiplex microsphere flow cytometry fluorescence detection technology, and more specifically, relates to a multiplex microsphere flow cytometry fluorescence detection method for vascular inflammation markers. Background Technology

[0002] Quantitative detection of vascular inflammatory markers is a crucial technique for assessing cardiovascular disease risk and monitoring treatment efficacy. Traditional multiplex assays primarily utilize technologies such as enzyme-linked immunosorbent assay (ELISA), chemiluminescent immunoassay, and flow cytometry. These methods achieve simultaneous detection of multiple inflammatory markers, including vascular endothelial growth factor (VEGF), platelet endothelial cell adhesion molecules (PECs), osteopontin, and matrix metalloproteinases (MMPs), through a sandwich immunoreaction principle involving both capture and detection antibodies. This approach plays a vital role in clinical diagnosis and research. However, traditional fluorescence detection techniques face limitations in signal dynamic range when processing complex biological samples. Particularly with low-concentration samples, they are susceptible to baseline noise from system background fluorescence, autofluorescence, and optical scattering. High-concentration samples are prone to signal saturation, leading to a narrowed linear detection range and impacting the accuracy and reliability of quantitative analysis. Current technologies lack effective dynamic signal correction mechanisms and cannot adaptively optimize based on varying fluorescence signal intensities. This makes it difficult to eliminate signal interference and cross-reactivity between markers during multiplex assays, resulting in insufficient accuracy and reproducibility to meet the stringent requirements of clinical applications. In other words, existing technologies suffer from technical problems such as limited dynamic range of fluorescence signals and severe background noise interference, which lead to insufficient accuracy in multiplex detection of vascular inflammatory markers. Summary of the Invention

[0003] In view of this, the present invention provides a method for multiplex microsphere flow cytometry detection of vascular inflammation markers, which can solve the technical problems of insufficient accuracy of multiplex detection of vascular inflammation markers due to the limited dynamic range of fluorescence signals and severe background noise interference in the prior art.

[0004] This invention is implemented as follows: This invention provides a method for multiplex microsphere flow cytometry detection of vascular inflammation markers, comprising: preparing a fluorescent microsphere mixture by coupling five capture microsphere antibodies to the surface of polystyrene microspheres with different fluorescent codes to form a fluorescent microsphere mixture with discriminative fluorescence characteristics; sample pretreatment by mixing the sample to be tested with an experimental buffer to prepare a calibrator gradient solution containing VEGF standard material, CD31 standard material, OPN standard material, MMP-9 standard material, and CD62P standard material; and sandwich immune complex formation by mixing and incubating the sample or calibrator gradient solution with the fluorescent microsphere mixture and biotin-labeled detection antibody to form a sandwich of fluorescent microspheres-capture antibody-target antigen-detection antibody. Immune complex; fluorescent signal labeling: phycoerythrin-labeled streptavidin is added to the sandwich immune complex, which binds to a biotin-labeled detection antibody to generate a fluorescent signal; flow cytometry fluorescence signal acquisition: flow cytometry is used to detect the phycoerythrin-labeled streptavidin-labeled fluorescent microspheres, acquiring the side-scattered light signal and fluorescence signal of each fluorescent microsphere; liquid-adaptive signal optimization: a liquid dynamic correction model is used to dynamically expand the range of the acquired original fluorescence signal and eliminate background noise. The liquid dynamic correction model adjusts the liquid conduction parameters and liquid damping coefficient in real time according to the intensity of the original fluorescence signal; multi-marker concentration calculation: the fluorescence signal optimized by the liquid dynamic correction model is fitted to a standard curve, and the calculation results are output.

[0005] Specifically, the step of preparing the fluorescent microsphere mixture involves selecting polystyrene microspheres with different fluorescence wavelengths as carriers, and modifying the surface of the polystyrene microspheres with carboxyl functional groups for antibody conjugation; covalently binding VEGF-capturing microsphere antibodies, CD31-capturing microsphere antibodies, OPN-capturing microsphere antibodies, MMP-9-capturing microsphere antibodies, and CD62P-capturing microsphere antibodies to the surface of the polystyrene microspheres through a chemical cross-linking reaction of 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide hydrochloride and N-hydroxysuccinimide; and mixing the different types of fluorescent microspheres after conjugation in an equimolar ratio to form the fluorescent microsphere mixture.

[0006] Specifically, the step of forming the sandwich immune complex involves mixing 25 μL of sample or calibrator gradient solution with 25 μL of fluorescent microsphere mixture in a 5 mL flow cytometer; adding 25 μL of biotin-labeled detection antibody mixture, which contains biotin-labeled detection antibodies against VEGF, CD31, OPN, MMP-9, and CD62P; and incubating at room temperature with shaking at a speed of 400 to 500 rpm for 120 minutes.

[0007] Specifically, the fluorescent signal labeling step involves adding 25 μL of phycoerythrin-labeled streptavidin solution to the sandwich immune complex; continuing to incubate at room temperature with shaking at a speed of 400 to 500 rpm for 30 minutes; introducing phycoerythrin fluorescent labeling into the sandwich immune complex through the binding of streptavidin and biotin, with phycoerythrin generating a fluorescence signal with a maximum emission wavelength of 578 nm at an excitation wavelength of 566 nm.

[0008] Specifically, the flow cytometry fluorescence signal acquisition step involves using a 488nm laser from a flow cytometer to excite fluorescent microspheres to generate side-scattered light signals, and identifying different coded fluorescent microsphere types based on the intensity and distribution characteristics of the side-scattered light signals; simultaneously, using a 561nm laser to excite phycoerythrin to generate fluorescence signals, and acquiring fluorescence signal intensity data using a photomultiplier tube detector.

[0009] The specific structure of the liquid dynamic correction model is as follows: The liquid dynamic correction model adopts a liquid state neural network architecture, which includes three main components: an input layer, a liquid reservoir layer, and an output layer. The liquid reservoir layer consists of 200 to 500 liquid neurons. Each liquid neuron has a time-varying membrane potential and liquid conduction characteristics. The liquid conduction parameters are dynamically adjusted according to the intensity of the original fluorescence signal, and the liquid damping coefficient changes in real time according to the noise level of the original fluorescence signal.

[0010] Before training the liquid dynamic correction model, the training dataset establishment step is also included: collecting the original fluorescence signal data of standard samples of vascular inflammation markers with different concentration gradients. The concentration range of the standard samples covers the linear detection range of each marker. At the same time, the corresponding background noise signal and system drift data are collected to establish a multi-dimensional training sample containing the original fluorescence signal intensity, noise level, and time drift characteristics.

[0011] Specifically, the training steps of the liquid dynamic correction model involve using the backpropagation algorithm to train the parameters of the liquid dynamic correction model, using the mean square error combined with the signal fidelity constraint term as the loss function, optimizing the connection weights of the liquid reservoir layer and the basic values ​​of the liquid conduction parameters through the gradient descent method during the training process, setting the number of training rounds to be in the range of 500 to 1000 rounds, and using an adaptive adjustment strategy to gradually decrease the learning rate from 0.01 to 0.001.

[0012] The liquid conduction adaptive function is used to adjust the liquid conduction parameters of the liquid dynamic correction model. The liquid conduction adaptive function is calculated based on three data: the original fluorescence signal intensity, the background noise level, and the signal drift rate, to obtain a liquid conduction adjustment value. When the liquid conduction adjustment value is in the range of 0.1 to 0.3, the low conduction mode is used to enhance the amplification effect of the weak signal and adjust the liquid conduction parameters.

[0013] Specifically, the multi-marker concentration calculation step involves using a 4-parameter logistic regression model to fit the fluorescence signal intensity of the calibrator gradient solution to a known concentration. The 4-parameter logistic regression model includes four key parameters: minimum response value, maximum response value, inflection point concentration, and slope parameter. Based on the fitted standard curve equation, the concentration results of VEGF, CD31, OPN, MMP-9, and CD62P in the unknown sample are calculated and output.

[0014] Among them, the fluorescent microspheres are monodisperse polystyrene microspheres with a diameter of 5.0 μm to 6.0 μm. The surface of the polystyrene microspheres is modified with carboxyl functional groups at a density of 50 to 100 per square micrometer. The fluorescent microspheres have different fluorescent coding characteristics to distinguish five different capture antibody types.

[0015] The sandwich immune complex is a three-layered complex formed by the sequential binding of a capture antibody, a target vascular inflammatory marker antigen, and a biotin-labeled detection antibody conjugated to fluorescent microspheres. The sandwich immune complex forms a stable molecular recognition structure through antigen-antibody binding reactions.

[0016] Among them, phycoerythrin-labeled streptavidin is a fluorescent biomolecule formed by covalently linking the phycoerythrin fluorescent group to the streptavidin protein molecule. Phycoerythrin-labeled streptavidin retains the high affinity binding characteristics of streptavidin and biotin while also having fluorescence detection function.

[0017] The liquid reservoir layer is the core computational layer of the liquid state neural network. It consists of a large number of interconnected liquid neurons. Each liquid neuron simulates the liquid environment characteristics of biological neurons, and has time-varying membrane potential and dynamic response characteristics. The information transmission efficiency between liquid neurons is controlled by liquid conduction parameters.

[0018] Among them, the liquid conduction parameter is a key parameter that controls the intensity of information transmission between liquid neurons. The value ranges from 0.1 to 1.0. The larger the value of the liquid conduction parameter, the more active the information transmission between liquid neurons is, and the smaller the value of the liquid conduction parameter, the stronger the inhibitory effect on information transmission is.

[0019] The liquid damping coefficient is a parameter that controls the decay rate of the dynamic response of the liquid neuron. Its value ranges from 0.05 to 0.5. The larger the liquid damping coefficient value, the faster the decay of the liquid neuron's state change. The smaller the liquid damping coefficient value, the longer the liquid neuron can maintain its activated state.

[0020] This invention addresses the technical problem of insufficient accuracy in multiplex detection of vascular inflammatory markers due to the limited dynamic range of fluorescence signals and severe background noise interference by establishing a liquid dynamic correction model for adaptive optimization processing of fluorescence signals. The liquid dynamic correction model employs a liquid-state neural network architecture. Utilizing the dynamic response characteristics of 200 to 500 liquid neurons in a liquid reservoir layer, it can adjust liquid conduction parameters and liquid damping coefficients in real time based on the intensity of the original fluorescence signal, achieving differentiated processing of signals with different intensities. This effectively expands the dynamic range of the detection system and eliminates background noise interference. The liquid conduction adaptive function calculates the liquid conduction adjustment value based on three key parameters: the intensity of the original fluorescence signal, the background noise level, and the signal drift rate. When the signal is weak, a low conduction mode is used to enhance signal amplification; when the signal is moderate, a medium conduction mode is used to maintain linear response; and when the signal is strong, a high conduction mode is used to suppress saturation, thereby ensuring detection accuracy and linear correlation across various concentration ranges. In summary, this invention solves the technical problem mentioned in the background art of insufficient accuracy in multiplex detection of vascular inflammatory markers due to the limited dynamic range of fluorescence signals and severe background noise interference. Attached Figure Description

[0021] Figure 1 This is a flowchart of the method of the present invention.

[0022] Figure 2 This is a schematic diagram of the liquid dynamic correction model involved in the present invention.

[0023] Figure 3 This is a comparison of fluorescence signals before and after processing by the liquid dynamic correction model in the embodiment.

[0024] Figure 4 The figure shows the standard curve fitting diagrams for the five vascular inflammation markers in the example.

[0025] Figure 5 The diagram shows the signal processing effect under different liquid conduction modes in the embodiments.

[0026] Figure 6 This is a scatter plot showing the concentration distribution of vascular inflammation markers in 128 samples from the examples. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings.

[0028] like Figure 1 The diagram shown is a flowchart of a multiplex microsphere flow cytometry method for detecting vascular inflammation markers provided by this invention. This method includes the following steps:

[0029] S01. Preparation of fluorescent microsphere mixture: VEGF-capturing microsphere antibody, CD31-capturing microsphere antibody, OPN-capturing microsphere antibody, MMP-9-capturing microsphere antibody, and CD62P-capturing microsphere antibody are respectively coupled to the surface of polystyrene microspheres with different fluorescent codes to form a fluorescent microsphere mixture with distinguishable fluorescent characteristics.

[0030] S02. Sample pretreatment: Mix the sample to be tested with the experimental buffer solution, and at the same time prepare a calibrator gradient solution containing VEGF standard material, CD31 standard material, OPN standard material, MMP-9 standard material and CD62P standard material. The calibrator gradient solution contains 7 different concentration points.

[0031] S03, Sandwich immune complex formation: The sample or calibrator gradient solution is mixed with fluorescent microsphere mixture and biotin-labeled detection antibody and incubated to form a sandwich immune complex of fluorescent microsphere-capture antibody-target antigen-detection antibody;

[0032] S04. Fluorescent signal labeling: Phycoerythrin-labeled streptavidin is added to the sandwich immune complex, which binds to biotin-labeled detection antibody to generate a fluorescent signal. The intensity of the fluorescent signal is proportional to the concentration of the target antigen.

[0033] S05. Flow cytometry fluorescence signal acquisition: Flow cytometer is used to detect individual fluorescent microspheres labeled with phycoerythrin and streptavidin. The side-scattered light signal and fluorescence signal of each fluorescent microsphere are acquired at the same time. Different types of fluorescent microspheres are distinguished by the side-scattered light signal, and the concentration of the target antigen is quantified by the fluorescence signal.

[0034] S06, Liquid Adaptive Signal Optimization: The original fluorescence signal is dynamically expanded and background noise is eliminated using a liquid dynamic correction model. The liquid dynamic correction model adjusts the liquid conduction parameters and liquid damping coefficient in real time according to the intensity of the original fluorescence signal.

[0035] S07. Multi-marker concentration calculation: The fluorescence signal optimized by the liquid dynamic correction model is fitted and calculated with the standard curve, and the concentration results of VEGF, CD31, OPN, MMP-9 and CD62P are output simultaneously.

[0036] The specific steps for preparing the fluorescent microsphere mixture include: selecting polystyrene microspheres with different fluorescence wavelengths as carriers, wherein the diameter of the polystyrene microspheres is in the range of 5.0 μm to 6.0 μm, and the surface of the polystyrene microspheres is modified with carboxyl functional groups for antibody conjugation; covalently binding VEGF capturing microsphere antibody, CD31 capturing microsphere antibody, OPN capturing microsphere antibody, MMP-9 capturing microsphere antibody, and CD62P capturing microsphere antibody to the surface of the polystyrene microspheres through a chemical cross-linking reaction of 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide hydrochloride and N-hydroxysuccinimide; and mixing the different types of fluorescent microspheres after conjugation in an equimolar ratio to form the fluorescent microsphere mixture.

[0037] The specific steps for forming the sandwich immune complex include: mixing 25 μL of sample or calibrator gradient solution with 25 μL of fluorescent microsphere mixture in a 5 mL flow cytometer; adding 25 μL of biotin-labeled detection antibody mixture, wherein the biotin-labeled detection antibody mixture contains biotin-labeled detection antibodies against VEGF, CD31, OPN, MMP-9, and CD62P; and incubating at room temperature with shaking at a speed of 400 to 500 rpm for 120 minutes to allow the target antigen to fully bind with the VEGF capture microsphere antibody, CD31 capture microsphere antibody, OPN capture microsphere antibody, MMP-9 capture microsphere antibody, CD62P capture microsphere antibody, and biotin-labeled detection antibody to form a stable sandwich immune complex.

[0038] The fluorescent labeling step specifically includes: adding 25 μL of phycoerythrin-labeled streptavidin solution to the sandwich immune complex; continuing to incubate at room temperature with shaking at a speed of 400 to 500 rpm for 30 minutes; introducing phycoerythrin fluorescent labeling into the sandwich immune complex through the binding of streptavidin and biotin, wherein the phycoerythrin generates a fluorescence signal with a maximum emission wavelength of 578 nm at an excitation wavelength of 566 nm.

[0039] The specific steps of flow cytometry fluorescence signal acquisition include: using a 488nm laser of the flow cytometer to excite fluorescent microspheres to generate side-scattered light signals, and identifying different coded fluorescent microsphere types by the intensity and distribution characteristics of the side-scattered light signals; simultaneously using a 561nm laser to excite phycoerythrin to generate fluorescence signals, and acquiring fluorescence signal intensity data through a photomultiplier tube detector; setting the flow rate to a range of 1000 to 3000 events per second to ensure accurate detection of individual fluorescent microspheres and separation of fluorescence signals, thereby obtaining the original fluorescence signal.

[0040] The specific structure of the liquid dynamic correction model is as follows: The liquid dynamic correction model adopts a liquid state neural network architecture, comprising three main components: an input layer, a liquid reservoir layer, and an output layer. The liquid reservoir layer consists of 200 to 500 liquid neurons, each with time-varying membrane potential and liquid conduction characteristics. The liquid conduction parameters are dynamically adjusted according to the intensity of the original fluorescence signal, and the liquid damping coefficient changes in real time according to the noise level of the original fluorescence signal. The output layer of the liquid dynamic correction model generates the corrected fluorescence signal by linearly combining the state vectors of the liquid reservoir layer. The steps for establishing the training dataset of the liquid dynamic correction model specifically include: collecting the original fluorescence signal data of standard samples of vascular inflammatory markers at different concentration gradients, wherein the concentration range of the standard samples covers the linear detection interval of each marker, and simultaneously collecting corresponding... Background noise signal and system drift data are used to establish a multi-dimensional training dataset containing original fluorescence signal intensity, noise level, and time drift features. The training dataset contains no less than 10,000 valid signal-label pairs, with the labels being manually calibrated ideal signal intensity values. The training steps of the liquid dynamic correction model specifically include: training the liquid dynamic correction model parameters using the backpropagation algorithm, using the mean squared error combined with signal fidelity constraints as the loss function, optimizing the connection weights of the liquid reservoir layer and the basic values ​​of the liquid conduction parameters during training using the gradient descent method, setting the training rounds to be between 500 and 1000 rounds, and using an adaptive adjustment strategy to gradually decrease the learning rate from 0.01 to 0.001. After training, the liquid dynamic correction model has the ability to adaptively correct for different original fluorescence signal intensities and noise conditions.

[0041] The liquid conduction adaptive function is used to adjust the liquid conduction parameters of the liquid dynamic correction model. The liquid conduction adaptive function is calculated based on three data: the original fluorescence signal intensity, the background noise level, and the signal drift rate, to obtain a liquid conduction adjustment value. When the liquid conduction adjustment value is in the range of 0.1 to 0.3, it is used to adjust the liquid conduction parameters by using a low conduction mode to enhance the amplification effect of weak signals. When the liquid conduction adjustment value is in the range of 0.3 to 0.7, it is used to adjust the liquid conduction parameters by using a medium conduction mode to maintain the linear response of the signal. When the liquid conduction adjustment value is in the range of 0.7 to 1.0, it is used to adjust the liquid conduction parameters by using a high conduction mode to suppress signal saturation.

[0042] The specific steps for calculating the concentration of multiple biomarkers include: using a 4-parameter logistic regression model to fit the fluorescence signal intensity of the calibrator gradient solution to a known concentration; the 4-parameter logistic regression model includes four key parameters: minimum response value, maximum response value, inflection point concentration, and slope parameter; calculating the VEGF, CD31, OPN, MMP-9, and CD62P concentrations in the unknown sample based on the fitted standard curve equation; and performing quality control checks on the VEGF, CD31, OPN, MMP-9, and CD62P concentrations to ensure that the linear correlation coefficient is greater than 0.990 and the coefficient of variation is less than 15%.

[0043] The fluorescent microspheres are monodisperse polystyrene microspheres with a diameter of 5.0 μm to 6.0 μm. The surface of the polystyrene microspheres is modified with a carboxyl functional group density ranging from 50 to 100 per square micrometer. The fluorescent microspheres have different fluorescent coding characteristics to distinguish five different types of capture antibodies. The sandwich immune complex is a three-layered complex formed by the sequential binding of the capture antibody, the target vascular inflammatory marker antigen, and the biotin-labeled detection antibody conjugated to the fluorescent microspheres. The sandwich immune complex forms a stable molecular recognition structure through an antigen-antibody binding reaction. The phycoerythrin-labeled streptavidin is a fluorescently labeled biomolecule formed by covalently linking the phycoerythrin fluorescent group to the streptavidin protein molecule. The phycoerythrin-labeled streptavidin retains the high affinity binding characteristics of streptavidin and biotin while also possessing fluorescence detection functionality. The liquid reservoir layer is the core computational layer of the liquid-state neural network, composed of a large number of interconnected liquid neurons. Each liquid neuron simulates the liquid environment characteristics of a biological neuron, exhibiting time-varying membrane potential and dynamic response characteristics. The efficiency of information transmission between liquid neurons is controlled by liquid conduction parameters. These liquid conduction parameters are key parameters controlling the intensity of information transmission between liquid neurons, ranging from 0.1 to 1.0. A larger value indicates more active information transmission between liquid neurons, while a smaller value indicates stronger inhibition of information transmission. Dynamic adjustment of the liquid conduction parameters enables adaptive processing of original fluorescence signals of different intensities. The liquid damping coefficient is a parameter controlling the decay rate of the dynamic response of the liquid neurons, ranging from 0.05 to 0.5. A larger value indicates faster decay of the liquid neuron's state change, while a smaller value indicates that the liquid neuron can maintain an activated state for a longer period. Adjusting the liquid damping coefficient suppresses noise in the original fluorescence signal and controls the signal retention time. The four-parameter logistic regression model is a mathematical model used to describe the dose-response relationship. It includes a minimum response value representing the background level of the fluorescence signal, a maximum response value representing the saturation level of the fluorescence signal, an inflection point concentration representing the biomarker concentration that produces a 50% maximum response, and a slope parameter representing the steepness of the concentration-response curve at the inflection point. These four key parameters collectively describe the nonlinear relationship between biomarker concentration and fluorescence signal intensity. The liquid conduction regulation value is a dimensionless parameter calculated using a liquid conduction adaptive function, ranging from 0.1 to 1.0. This liquid conduction regulation value comprehensively reflects the signal quality status under the current detection environment and guides the liquid conduction parameter adjustment strategy of the liquid dynamic correction model, achieving adaptive optimization for different detection conditions.

[0044] In a specific implementation of this invention, a five-item vascular inflammation detection kit can be used. The main components include two categories: detection components and calibration components. The detection components consist of capture microsphere antibodies, detection antibodies, SA-PE, experimental buffer, and washing buffer. The capture microsphere antibodies include five different antibody-conjugated microspheres: VEGF capture microsphere antibody, CD31 capture microsphere antibody, OPN capture microsphere antibody, MMP-9 capture microsphere antibody, and CD62P capture microsphere antibody. The detection antibodies are biotin-labeled VEGF detection antibody, CD31 detection antibody, OPN detection antibody, MMP-9 detection antibody, and CD62P detection antibody. The 62P detection antibody, SA-PE, is streptavidin labeled with phycoerythrin. The experimental buffer is phosphate buffer containing BSA, and the washing buffer is phosphate buffer containing Tween-20. The calibration components include calibrators, matrix A, and experimental buffer. The calibrators consist of lyophilized powders of VEGF, CD31, OPN, MMP-9, and CD62P antigens. After reconstitution, the concentration of VEGF is 50,000 pg / mL, the concentration of MMP-9 is 200,000 pg / mL, and the concentrations of CD31, OPN, and CD62P are 100,000 pg / mL. Matrix A is lyophilized from fetal bovine serum.

[0045] The detection principle of the vascular inflammation five-item test kit is based on immunological analysis methods, using the direct sandwich method. Fluorescent microspheres coupled with capture antibodies and biotin-labeled paired detection antibodies bind to the analyte antigens in the sample or calibrator to form a sandwich complex. This complex then reacts with streptavidin labeled with phycoerythrin. Immunological analysis is combined with flow cytometry analysis. Within the detection range, the fluorescence intensity is directly proportional to the analyte antigen content in the sample. A standard curve is established to achieve simultaneous quantitative detection of five vascular inflammation markers: VEGF, CD31, OPN, MMP-9, and CD62P.

[0046] The sample requirements for the Vascular Inflammation Five-Item Test Kit are plasma specimens. The collection method is to collect venous blood samples using EDTA anticoagulant tubes, centrifuge at 1000g for 20 minutes, and send the separated plasma for testing. The storage conditions are as follows: the sample to be tested must be processed and tested within 8 hours of blood collection. If testing cannot be performed within 8 hours, it must be stored at 2℃ to 8℃ for no more than 24 hours after processing. For longer storage, aliquot and seal, and store at -20℃ for no more than 30 days, avoiding repeated freeze-thaw cycles no more than twice. Sample quality requirements: samples with severe hemolysis, lipemia, jaundice, or contamination may affect the test results and should be avoided.

[0047] The specific implementation methods of the above steps are described in detail below.

[0048] Step S01 involves preparing a fluorescent microsphere mixture. This step, based on the principles of chemical coupling reactions and fluorescence encoding technology, aims to construct a carrier system with discriminative fluorescent characteristics for the simultaneous detection of different biomarkers. First, monodisperse polystyrene microspheres with a diameter of 5.0 μm to 6.0 μm are selected as the carrier matrix. This diameter range ensures suitable scattering characteristics and stable flow properties during flow cytometry detection. The surface of the polystyrene microspheres is pre-modified with carboxyl functional groups, with the carboxyl group density controlled within the range of 50 to 100 per square micrometer. This density range provides sufficient reaction sites for subsequent antibody coupling reactions while avoiding steric hindrance effects caused by over-coupling. A covalent coupling reaction mediated by 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide hydrochloride and N-hydroxysuccinimide was employed. This reaction, based on the amide bond formation mechanism between amino and carboxyl groups, conjugated VEGF-capturing microsphere antibodies, CD31-capturing microsphere antibodies, OPN-capturing microsphere antibodies, MMP-9-capturing microsphere antibodies, and CD62P-capturing microsphere antibodies to the surface of microspheres with different fluorescent codes. The coupling reaction was carried out under a neutral buffer environment, with the reaction temperature controlled between 4°C and 8°C, and the reaction time between 2 and 4 hours to ensure the preservation of antibody activity and optimal coupling efficiency. The five different fluorescently encoded microspheres were then mixed in equimolar proportions to form a fluorescent microsphere mixture capable of simultaneously recognizing five vascular inflammatory markers.

[0049] Step S02 involves sample pretreatment, based on the principles of sample standardization and concentration gradient establishment. Its purpose is to provide a standardized detection environment for subsequent immunoreaction and quantitative analysis. The test sample is mixed with experimental buffer at a 1:1 volume ratio. The experimental buffer is a phosphate buffer containing bovine serum albumin (BSA), with a pH controlled between 7.2 and 7.4. This pH range ensures antibody activity and antigen stability. The BSA concentration is set to 1% to 3%, acting as a protein stabilizer and a non-heterogeneous binding inhibitor. Simultaneously, a calibrator gradient solution containing VEGF, CD31, OPN, MMP-9, and CD62P standards is prepared. This gradient solution contains seven different concentration points, established using a two-fold dilution method. The highest concentration point represents the upper limit of the linear detection range for each marker, and the lowest concentration point represents the concentration near the detection limit. The standards are prepared using recombinant protein technology, with a purity of no less than 95%, and their activity is verified by immunological methods.

[0050] Step S03 involves the formation of a sandwich immune complex. This step is based on the heterogeneous binding reaction of antigen and antibody and the principle of sandwich immunoassay, aiming to construct a stable three-layer molecular recognition structure to achieve heterogeneous capture of the biomarker. 25 μL of sample or calibrator gradient solution is mixed with 25 μL of fluorescent microsphere mixture in a 5 mL flow cytometer, using vortexing to ensure uniform distribution. 25 μL of biotin-labeled detection antibody mixture is added. This mixture contains biotin-labeled detection antibodies against VEGF, CD31, OPN, MMP-9, and CD62P. The concentrations of each antibody are titrated to optimize saturation binding conditions. Biotin labeling is performed using the NHS-biotin chemical modification method, with the labeling molar ratio controlled within the range of 2:1 to 4:1 to maintain antibody binding activity while providing sufficient biotin labeling density. The mixture is incubated at room temperature with shaking at 400 to 500 rpm for 120 minutes. This condition ensures that the antigen fully binds to the capture and detection antibodies to reach reaction equilibrium. In the reaction system, the target antigen molecules are simultaneously recognized and bound by the capture antibody on the surface of the fluorescent microspheres and the free biotin-labeled detection antibody, forming a stable sandwich immune complex structure of fluorescent microsphere-capture antibody-target antigen-detection antibody.

[0051] Step S04 involves fluorescent signal labeling. This step, based on the biotin-streptavidin binding reaction and fluorescent labeling technology, aims to introduce a detectable fluorescent signal into the sandwich immune complex to achieve signal amplification and quantitative detection. 25 μL of phycoerythrin-labeled streptavidin solution is added to the sandwich immune complex. The concentration of phycoerythrin-labeled streptavidin is set within the range of 2 μg / mL to 5 μg / mL, ensuring saturated binding conditions while avoiding non-heterogeneous background signals. The coupling of phycoerythrin and streptavidin utilizes a succinimide cross-linking chemical reaction, with the coupling molar ratio controlled within the range of 1:1 to 2:1 to maintain the high affinity binding characteristics of streptavidin and biotin. The mixture is then incubated at room temperature with shaking at a speed of 400 to 500 rpm for 30 minutes. This incubation condition ensures sufficient binding of streptavidin and biotin. The binding of streptavidin to biotin is based on a tetravalent binding site and an ultra-high affinity constant of approximately [missing value]. The molecular recognition mechanism at L / mol forms an extremely stable biomolecular complex. The phycoerythrin fluorophore generates a fluorescence signal with a maximum emission wavelength of 578 nm at an excitation wavelength of 566 nm. The intensity of this fluorescence signal is directly proportional to the amount of the sandwich immune complex, enabling indirect quantitative detection of the target antigen concentration.

[0052] Step S05 involves flow cytometry fluorescence signal acquisition. Based on flow cytometry and optical detection principles, this step aims to achieve precise detection and quantitative acquisition of fluorescence signals from individual fluorescent microspheres. A 488nm laser from the flow cytometer is used as the side-scattering excitation source to excite the fluorescent microspheres and generate side-scattering signals. The scattered light is acquired and digitally processed by a photomultiplier tube detector. The intensity and distribution characteristics of the side-scattering signal reflect the size and internal structure information of the fluorescent microspheres. By setting a signal intensity threshold and identifying the scattered light distribution pattern, accurate differentiation of different fluorescently encoded microsphere types is achieved. Simultaneously, a 561nm laser is used as the fluorescence excitation source to excite phycoerythrin to generate characteristic fluorescence signals. These signals are acquired through a 578nm bandpass filter and a photomultiplier tube detector. The photomultiplier tube gain is set to a medium level to avoid signal saturation while ensuring effective detection of weak signals. The flow rate is controlled within the range of 1000 to 3000 events per second. This flow rate range ensures accurate detection of individual fluorescent microspheres and effective separation of fluorescence signals, avoiding microsphere aggregation or signal overlap. The data acquisition system employs a high-speed analog-to-digital converter with a sampling frequency of no less than 100kHz to ensure complete acquisition and accurate digitization of the fluorescence signal. The obtained raw fluorescence signal data includes the side-scattered light intensity, fluorescence signal intensity, and timestamp information for each microsphere.

[0053] Step S06 is implemented as liquid-state adaptive signal optimization. This step is based on a liquid-state neural network algorithm and adaptive signal processing principles, aiming to eliminate background noise and system drift in the original fluorescence signal and expand the detection dynamic range. The liquid dynamic correction model adopts a liquid-state neural network architecture, which simulates the liquid environment characteristics of a biological nervous system and has the ability to adaptively process time-varying signals. The model receives the original fluorescence signal intensity, background noise level, and signal drift rate as input parameters. Through nonlinear dynamic transformation of the liquid reservoir layer and linear combination operation of the output layer, it generates a corrected fluorescence signal intensity as the output parameter. The liquid conduction adaptive function calculates the liquid conduction adjustment value according to the characteristics of the input signal. When the adjustment value is in the range of 0.1 to 0.3, a low conduction mode is used to enhance the amplification effect of weak signals; when the adjustment value is in the range of 0.3 to 0.7, a medium conduction mode is used to maintain the linear response characteristics of the signal; and when the adjustment value is in the range of 0.7 to 1.0, a high conduction mode is used to suppress signal saturation. The liquid damping coefficient is adjusted in real time according to the noise level of the original fluorescence signal, with a value ranging from 0.05 to 0.5. A larger damping coefficient value corresponds to a faster noise attenuation rate, while a smaller damping coefficient value corresponds to a longer signal hold time. The model output layer generates a corrected fluorescence signal that has undergone dynamic range expansion and background noise cancellation processing by linearly combining the state vectors of the liquid reservoir layer.

[0054] Step S07 involves calculating the concentration of multiple biomarkers. This step, based on a 4-parameter logistic regression model and curve fitting algorithm, aims to establish a quantitative relationship between fluorescence signal intensity and biomarker concentration and to calculate the concentration of unknown samples. The 4-parameter logistic regression model includes four key parameters: minimum response value, maximum response value, inflection point concentration, and slope parameter. This model accurately describes the typical S-shaped dose-response curve in immunoassay. The minimum response value represents the background level of the fluorescence signal, reflecting the system's baseline noise and non-heterogeneous binding signal. The maximum response value represents the saturation level of the fluorescence signal, corresponding to the maximum signal intensity when the antibody binding site is fully occupied. The inflection point concentration represents the biomarker concentration that produces a 50% maximum response, reflecting the equilibrium constant characteristics of the antibody-antigen binding reaction. The slope parameter represents the steepness of the concentration-response curve at the inflection point, reflecting the sensitivity characteristics of the detection system. The fluorescence signal intensity of the calibrator gradient solution is curve-fitted to known concentration data using a nonlinear least squares method to obtain the standard curve equation parameters for each biomarker. Based on the fitted standard curve equation, the corrected fluorescence signal intensity of the unknown sample was substituted into the equation, and the concentrations of VEGF, CD31, OPN, MMP-9, and CD62P were calculated using a numerical iterative algorithm. The calculation results were then subjected to quality control verification, requiring a linear correlation coefficient greater than 0.990 and a coefficient of variation less than 15% to ensure the accuracy and reproducibility of the detection results.

[0055] like Figure 2 As shown, the liquid dynamic correction model employs a liquid-state neural network architecture, comprising three main components: an input layer, a liquid reservoir layer, and an output layer. The input layer receives a three-dimensional input vector consisting of the original fluorescence signal intensity, background noise level, and signal drift rate, and maps the input signal to the range of 0 to 1 through normalization. The liquid reservoir layer consists of 200 to 500 liquid neurons, each with time-varying membrane potential and liquid conduction characteristics. Neurons form a complex dynamic network structure through random sparse connections. The state update of the liquid neurons is based on a dynamic system described by differential equations; membrane potential changes are influenced by the input signal, the states of adjacent neurons, and liquid conduction parameters. The liquid conduction parameters are dynamically adjusted according to the original fluorescence signal intensity, ranging from 0.1 to 1.0, controlling the intensity and speed of information transmission between liquid neurons. The liquid damping coefficient changes in real time according to the original fluorescence signal noise level, ranging from 0.05 to 0.5, controlling the decay rate and stability of the dynamic response of the liquid neurons. The output layer generates a corrected fluorescence signal by linearly combining the state vectors of the liquid storage tank layer, and the linear combination weights are optimized by a supervised learning algorithm.

[0056] The establishment of the training dataset for the liquid dynamic correction model includes the following detailed steps: Standard samples of vascular inflammatory markers at different concentration gradients are collected, covering the linear detection range of each marker, including three main ranges: low, medium, and high concentration. A serial dilution series of standard samples is prepared using an automated liquid handling system to ensure the accuracy and reproducibility of the concentration gradients. Raw fluorescence signal data of the standard samples are acquired using flow cytometry, recording the fluorescence signal intensity, background noise level, and detection time information for each concentration point. Simultaneously, background noise signals and system drift data of blank control samples are acquired, and system stability and signal drift trends are monitored through long-term continuous detection. A multidimensional training sample matrix is ​​established, containing raw fluorescence signal intensity, noise level, and time drift features. Each training sample includes an input feature vector and a corresponding ideal signal intensity label. The training dataset contains no fewer than 10,000 valid signal-label pairs, with labels representing ideal signal intensity values ​​that have undergone manual calibration and quality control to ensure the accuracy and representativeness of the training data.

[0057] The liquid dynamic correction model employs backpropagation to optimize network parameters during training, using a composite loss function combining mean squared error and signal fidelity constraints. During training, gradient descent is used to optimize the connection weights of the liquid reservoir layer and the baseline values ​​of the liquid conduction parameters, while simultaneously optimizing the linear combination weights of the output layer. The training epochs are set between 500 and 1000, with the learning rate adaptively decreasing from 0.01 to 0.001 to avoid gradient explosion and convergence difficulties. A stochastic mini-batch gradient descent algorithm is used, with batch sizes ranging from 32 to 128 to balance training efficiency and convergence stability. Early stopping and regularization techniques are employed to prevent overfitting, terminating training when the validation set loss continuously increases. After training, the liquid dynamic correction model exhibits adaptive correction capabilities for different original fluorescence signal intensities and noise conditions, effectively expanding the detection dynamic range and improving signal quality.

[0058] It is worth further explanation that traditional linear filtering methods and fixed-parameter signal processing algorithms exhibit significant limitations when facing complex and variable fluorescence signal environments, failing to adaptively adjust according to signal characteristics. Liquid-state neural networks, with their inherent temporal memory and nonlinear transformation characteristics, can capture time-varying features and complex patterns in fluorescence signals. Compared to traditional digital filters, the liquid dynamic correction model achieves adaptive optimization processing for different signal intensities and noise conditions through dynamic adjustment of liquid conduction parameters and liquid damping coefficients. Compared to static thresholding methods, the liquid model can adjust the processing strategy according to the real-time signal quality state, avoiding the problem of weak signals being over-suppressed or strong signals exhibiting saturation distortion. The advantage of this model lies in its ability to simultaneously handle multiple signal processing tasks such as signal amplification, noise suppression, and dynamic range expansion, achieving integrated optimization of complex signal processing functions through a unified neural network architecture.

[0059] The key technical concepts of this invention include multiplex microsphere fluorescence encoding technology, liquid-state adaptive signal optimization technology, and sandwich immune complex construction technology. Multiplex microsphere fluorescence encoding technology enables the simultaneous detection of multiple biomarkers using microsphere carriers with different fluorescence wavelengths. Compared to traditional single-biomarker detection methods, this technology significantly improves detection efficiency and sample utilization. The dual differentiation mechanism of spatial and spectral encoding ensures the accuracy and reliability of the detection results. Liquid-state adaptive signal optimization technology employs a liquid-state neural network to achieve intelligent processing of fluorescence signals. Compared to traditional fixed-parameter filtering methods, this technology has adaptive adjustment capabilities, dynamically optimizing processing parameters based on signal characteristics, effectively expanding the detection dynamic range and improving signal quality. Sandwich immune complex construction technology achieves high anisotropic capture and signal amplification of biomarkers through a three-layer molecular recognition structure. Compared to traditional direct immunoassay methods, this technology significantly improves detection sensitivity and anisotropy. The double-antibody sandwich mechanism ensures the accuracy of the detection results. The synergistic effect of these key technological ideas forms a complete multi-detection technology system. Multi-encoding technology provides a simultaneous detection platform, sandwich immunoassay technology ensures detection heterogeneity and sensitivity, and liquid optimization technology guarantees signal quality and detection accuracy. The three work together to achieve efficient and accurate detection of vascular inflammation markers, which has obvious technical advantages and application value compared with existing single technology solutions.

[0060] It should be noted that the present invention also solves the following two technical problems.

[0061] Signal cross-interference issues in the simultaneous detection of multiple biomarkers. In traditional multiplex immunoassay techniques, when simultaneously detecting multiple vascular inflammatory biomarkers, non-specific binding and signal cross-interference easily occur between different antibodies, especially in the complex plasma matrix environment. Various endogenous interfering substances can affect the specificity of antigen-antibody binding, leading to false positive or false negative results. This invention uses polystyrene microspheres with different fluorescent coding characteristics as carriers. Each capture antibody is coupled to the surface of a microsphere with a specific fluorescence wavelength. The side-scattered light signal from flow cytometry is used to identify the different types of fluorescent microspheres, achieving effective separation of the detection signals of each biomarker. Simultaneously, by optimizing the antibody coupling density and reaction conditions, the capture antibody on each fluorescent microsphere is ensured to have high specificity and stability, significantly reducing signal cross-interference during the simultaneous detection of multiple biomarkers.

[0062] This invention addresses the correction of time drift and batch-to-batch variability in detection systems. Flow cytometry detection systems experience systematic drift due to factors such as laser power attenuation, optical component aging, and temperature variations during long-term operation. Performance differences between different batches of reagents also affect the consistency and comparability of detection results. The liquid dynamic correction model of this invention incorporates time drift characteristics and batch-to-batch variability data during training. By establishing multi-dimensional training samples including original fluorescence signal intensity, noise level, and time drift characteristics, the model possesses adaptive correction capabilities for systematic drift and batch-to-batch variability. The liquid conduction adaptive function can adjust the correction parameters in real time according to the signal drift rate during detection, effectively compensating for systematic errors and ensuring the stability and consistency of detection results at different time points and between different batches of reagents, thus improving the robustness and practicality of the detection method.

[0063] Specifically, the principle of this invention is as follows: The technical solution of this invention can solve the core problems of limited dynamic range of fluorescence signals and severe background noise interference, mainly based on the adaptive signal processing mechanism of a liquid dynamic correction model. Liquid-state neural networks possess natural time-varying characteristics and dynamic response capabilities. The liquid neurons in its liquid reservoir layer can simulate the liquid environment characteristics of a biological nervous system. Through the time-varying characteristics of membrane potential and the dynamic adjustment of liquid conduction parameters, differentiated processing strategies for fluorescence signals of different intensities are achieved. When detecting weak signals, the liquid conduction adaptive function calculates a small liquid conduction adjustment value, and the model automatically switches to a low conduction mode. By increasing the information transmission intensity between liquid neurons and extending the signal retention time, the weak signal is effectively amplified and the relative influence of background noise is suppressed, thereby improving the detection accuracy of low-concentration samples. When detecting strong signals, the liquid conduction adjustment value increases, and the model switches to a high conduction mode. By appropriately suppressing the signal conduction intensity and accelerating the response decay rate, signal saturation is prevented, ensuring the linear response characteristics of high-concentration sample detection. The real-time adjustment mechanism of the liquid damping coefficient dynamically changes according to the noise level of the original fluorescence signal. When the noise level is high, the damping coefficient is increased to accelerate the attenuation of the noise signal; when the noise level is low, the damping coefficient is decreased to maintain the stability of the useful signal, thus achieving the optimal balance between noise suppression and signal preservation. The liquid dynamic correction model trained by the backpropagation algorithm can learn the signal characteristics and noise patterns under different detection conditions, establishing a multi-dimensional mapping relationship including the original fluorescence signal intensity, noise level, and time drift characteristics, achieving adaptive optimization for complex detection environments. This dynamic correction mechanism based on liquid neural networks fundamentally solves the limitations of traditional static signal processing methods in adapting to different signal intensities and noise conditions, providing a high-precision, wide dynamic range technical solution for the multiple detection of vascular inflammation biomarkers.

[0064] The following provides a specific embodiment 1 of the present invention, and the specific implementation of each step in this embodiment 1 is described in detail below.

[0065] In this embodiment, the specific implementation of step S01 is the same as described above, and will not be repeated in detail here.

[0066] The specific implementation of step S02 involves sample pretreatment, which includes calculating the concentration of the calibrator gradient solution. The concentration calculation of the calibrator gradient solution uses the two-fold dilution method, as detailed below:

[0067] ;

[0068] In the formula, For the first The concentration of the biomarker at each concentration point is expressed in pg / mL. The concentration of the marker at the highest concentration point is expressed in pg / mL. This represents the concentration point number, ranging from 1 to 7. The parameter is obtained as follows: The assay was conducted experimentally, including step 1: diluting each biomarker standard with experimental buffer to the upper limit of the linear detection range; and step 2: validating the activity and concentration accuracy of the standards using immunological methods. The dilution volume ratio was calculated using the following formula:

[0069] ;

[0070] In the formula, This refers to the sample volume, in μL. This refers to the buffer volume, in μL. The target concentration is expressed in pg / mL. This is the initial concentration, expressed in pg / mL.

[0071] The specific implementation methods for steps S03-S05 are the same as those described above, and will not be repeated in detail here.

[0072] The specific implementation of step S06 is liquid adaptive signal optimization, which involves the mathematical calculation of the liquid conduction adaptive function and the liquid dynamic correction model. The formula for calculating the liquid conduction adaptive function is:

[0073] ;

[0074] In the formula, This is a dimensionless value representing the liquid conduction regulation. The normalized original fluorescence signal intensity is dimensionless. This is a dimensionless normalized background noise level. The normalized signal drift rate is dimensionless. These are weighting coefficients, dimensionless. The method for obtaining these parameters is as follows: ,in The original fluorescence signal intensity is expressed in arbitrary fluorescence units. This represents the maximum signal intensity, in arbitrary fluorescence units. ,in The background noise level is expressed in arbitrary fluorescence units. This represents the maximum noise level, expressed in arbitrary fluorescence units. ,in For signal drift rate, This represents the change in signal intensity, expressed in arbitrary fluorescence units. The time interval is in minutes. This represents the maximum drift rate, expressed in arbitrary fluorescence units per minute. Weighting coefficients. , , .

[0075] The formula for adjusting the liquid conduction parameters is:

[0076] ;

[0077] In the formula, The adjusted liquid conduction parameters are dimensionless. , , Basic conduction parameters; , , This is the adjustment coefficient.

[0078] The formula for calculating the liquid damping coefficient is:

[0079] ;

[0080] In the formula, The liquid damping coefficient is dimensionless. It is the basic damping coefficient, dimensionless; Let be the noise attenuation constant, which is dimensionless; The oscillation amplitude parameter is dimensionless. Angular frequency, in rad / s; Hz is the oscillation frequency; Time, in seconds; This is the phase constant, expressed in rad.

[0081] The output calculation formula of the liquid dynamic correction model is:

[0082] ;

[0083] In the formula, The fluorescence signal intensity is the corrected value, in arbitrary fluorescence units; For the output layer The weights of each connection are dimensionless. For the first layer of liquid storage tank At time 1 neuron The state value is dimensionless. This represents the total number of neurons in the liquid reservoir layer, ranging from 200 to 500.

[0084] The specific implementation of step S07 involves multi-marker concentration calculation. This step uses a 4-parameter logistic regression model for curve fitting and concentration calculation. The mathematical expression of the 4-parameter logistic regression model is:

[0085] ;

[0086] In the formula, The fluorescence signal intensity is expressed in arbitrary fluorescence units. The concentration of the biomarker is expressed in pg / mL. The minimum response value is expressed in arbitrary fluorescence units. The maximum response value is expressed in arbitrary fluorescence units. The inflection point concentration is expressed in pg / mL. This is a slope parameter, dimensionless. The method for obtaining this parameter is as follows: The parameters were obtained by fitting the calibrator data using a nonlinear least squares method. The objective function for fitting was:

[0087] ;

[0088] In the formula, To fit the objective function value, the unit is the square of any fluorescence unit; For the first The observed fluorescence intensity at each calibration point is expressed in arbitrary fluorescence units. The model predicts fluorescence intensity, with units of arbitrary fluorescence units. This is the calibration point number, with a value ranging from 1 to 7.

[0089] It should be noted that the concentration of unknown samples is calculated using the inverse function of a 4-parameter logistic regression model:

[0090] ;

[0091] In the formula, The concentration of the biomarker is given in pg / mL. The measured fluorescence signal intensity is expressed in arbitrary fluorescence units. These are the model parameters obtained through fitting.

[0092] It should be noted that the calculation formulas for quality control parameters include the linear correlation coefficient:

[0093] ;

[0094] In the formula, The linear correlation coefficient is dimensionless. The average value is the observed values, expressed in arbitrary fluorescence units. The formula for calculating the coefficient of variation is:

[0095] ;

[0096] In the formula, The coefficient of variation is expressed as a percentage. This represents the standard deviation of the concentration measurement, in pg / mL. This represents the average concentration measurements, expressed in pg / mL.

[0097] It should be noted that the principle of the liquid conduction adaptive function is based on a multi-parameter weighted averaging algorithm. By comprehensively considering three key parameters—normalized original fluorescence signal intensity, normalized background noise level, and normalized signal drift rate—it achieves a quantitative assessment of the detection environment state. This function employs a linear weighted combination method. Three independent signal quality indices are merged into a single conduction modulation value. The weighting coefficients are set based on experimental verification results of the influence of each parameter on signal quality. Normalization ensures a unified representation of parameters with different dimensions. , , This function achieves dimensionless scaling of signal strength, noise level, and drift rate. Compared to traditional single-parameter threshold judgment methods, this function can more comprehensively reflect the real-time state of the detection system, avoiding misjudgments caused by fluctuations in a single parameter, and improving the stability and accuracy of signal processing. The function's effectiveness lies in its ability to automatically select the most suitable signal processing mode based on different detection conditions, ensuring optimal signal quality in various noise environments.

[0098] It should be noted that the principle of the liquid conduction parameter adjustment formula is based on the piecewise function design concept. Different mathematical models are used to optimize the parameters according to different ranges of the liquid conduction adjustment value. In the low conduction range, a linear growth model is used to enhance the amplification effect of weak signals; in the medium conduction range, a quadratic function model is used to maintain the linear response characteristics of the signal; and in the high conduction range, a linear decay model is used to suppress signal saturation. The formula design considers the nonlinear relationship between signal intensity and detection accuracy in fluorescence detection, achieving optimal signal conduction control through mathematical modeling. Compared to traditional fixed-parameter processing methods, this formula can dynamically adjust the processing parameters according to signal characteristics, effectively expanding the detection dynamic range, improving the detection capability of weak signals while avoiding the saturation distortion problem of strong signals.

[0099] It should be noted that the principle behind the formula for calculating the liquid damping coefficient combines the exponential decay model and the periodic oscillation model, using an exponential function. The effect of noise on the damping coefficient is described using a sine function. A time-dependent periodic regulation mechanism is introduced. The exponential decay term reflects the negative correlation between noise level and the decay rate of neuronal response; the greater the noise, the faster the decay rate is needed to suppress the interfering signal. The periodic oscillation term simulates the natural rhythmic characteristics of biological nervous systems, where the angular frequency... Controlling the oscillation period and oscillation amplitude parameters The formula controls the adjustment intensity. Compared to the static damping parameter setting method, this formula can achieve an adaptive response to different noise conditions and time variations, improving the robustness and time stability of signal processing.

[0100] It should be noted that the principle of the 4-parameter logistic regression model is based on the mathematical description of the S-curve. This model accurately reflects the dose-response relationship of antibody-antigen binding reactions in immunoassays. The minimum response parameter in the model... The parameter representing the system's baseline signal level and maximum response value. The signal intensity at which the antibody binding site becomes saturated, the inflection point concentration parameter. The slope parameter reflects the characteristics of the equilibrium constant of antibody-antigen binding. This model describes the sensitivity of the concentration-response curve. Through the synergistic effect of four independent parameters, it accurately describes the nonlinear response relationship across the entire concentration detection range. The model parameters are minimized by the objective function. Obtaining the optimal solution. Compared to traditional linear fitting methods, the 4-parameter logistic regression model can more accurately reflect the biological characteristics of the immune response, significantly improve the accuracy of concentration calculation and the coverage of the detection range, especially in the low and high concentration ranges where the fitting accuracy is significantly better than that of the linear model.

[0101] It should be noted that the principle behind the calculation formula for quality control parameters is based on statistical analysis methods, including the linear correlation coefficient. The goodness of fit of a model is quantified by the ratio of the sum of squared residuals to the sum of squared total squares, reflecting the degree of agreement between experimental data and the theoretical model. Coefficient of variation. The reproducibility and precision of measurement results are assessed by the ratio of standard deviation to mean. The combined application of these two parameters comprehensively evaluates the reliability of test results from both the quality of fit and measurement accuracy dimensions. Compared to traditional single quality control indicators, this combined parameter system can more comprehensively identify systematic and random errors in the testing process, ensuring the accuracy and reliability of test results and providing higher quality assurance for clinical applications.

[0102] It should be noted that the variables involved in this invention are explained in detail in Table 1.

[0103] Table 1. Variable Explanation Table

[0104]

[0105] To better understand and implement this invention, a specific application scenario is provided below as Example 2: The technical team needs to detect vascular inflammatory markers in plasma samples from 128 suspected coronary heart disease patients. The samples were obtained from the cardiology department of a hospital. The patients' ages ranged from 45 to 78 years old, including 76 males and 52 females. The technical team decided to use the vascular inflammatory marker multiplex microsphere flow cytometry fluorescence detection method of this invention for batch detection.

[0106] The technical team used a five-item vascular inflammation detection kit, the main components of which are shown in Table 2.

[0107] Table 2 Main Components

[0108]

[0109] The technical team first prepared the fluorescent microsphere mixture. Five different fluorescently encoded polystyrene microspheres were selected as carriers; all microspheres had a diameter of 5.6 μm and a surface carboxyl functional group density of 78 per square micrometer. VEGF-capturing microsphere antibodies, CD31-capturing microsphere antibodies, OPN-capturing microsphere antibodies, MMP-9-capturing microsphere antibodies, and CD62P-capturing microsphere antibodies were coupled to the surfaces of the polystyrene microspheres with different fluorescent codes via a chemical cross-linking reaction between 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide hydrochloride and N-hydroxysuccinimide. After coupling, the five different types of fluorescent microspheres were mixed in equimolar proportions to prepare a fluorescent microsphere mixture with a total volume of 5 mL.

[0110] In the sample pretreatment stage, the technical team processed 128 plasma samples. All plasma samples were collected from venous blood using EDTA anticoagulant tubes, and the plasma was separated after centrifugation at 1000g for 20 minutes. Simultaneously, calibrator gradient solutions were prepared, and the solutions were reconstituted using lyophilized calibrators. concentration , concentration , , , The concentrations are all Seven calibrator gradient solutions at different concentrations were prepared using the 4-fold dilution method and labeled as follows: to A blank control was also provided. .

[0111] In the sandwich immune complex formation step, the technical team mixed 25 μL of sample or calibrator gradient solution with 25 μL of fluorescent microsphere mixture in a 5 mL flow cytometer. Then, 25 μL of biotin-labeled detection antibody mixture containing biotin-labeled detection antibodies against VEGF, CD31, OPN, MMP-9, and CD62P was added. After mixing, the mixture was incubated at room temperature with shaking at 450 rpm for 120 minutes to ensure sufficient binding of the target antigen with the capture and detection antibodies to form a stable sandwich immune complex.

[0112] In the fluorescent labeling stage, the technical team added 25 μL of phycoerythrin-labeled streptavidin solution to each sandwich immune complex. The mixture was then incubated at room temperature with shaking at 450 rpm for 30 minutes. The phycoerythrin fluorescent labeling was introduced into the sandwich immune complex via the high affinity binding of streptavidin to biotin. Phycoerythrin produced a fluorescence signal with a maximum emission wavelength of 578 nm at an excitation wavelength of 566 nm.

[0113] During flow cytometry fluorescence signal acquisition, the technical team used a BD Canto II flow cytometer for detection. A 488nm laser was used to excite fluorescent microspheres to generate side-scattered light signals, and the intensity and distribution characteristics of the side-scattered light signals were used to identify different coded fluorescent microsphere types. Simultaneously, a 561nm laser was used to excite phycoerythrin to generate fluorescence signals, and the fluorescence signal intensity data was acquired using a photomultiplier tube detector. The flow rate was set to 2000 events per second to ensure accurate detection of individual fluorescent microspheres and separation of fluorescence signals. Raw fluorescence signal data was obtained after detection, such as... Figure 3 As shown, different types of fluorescent microspheres exhibit distinct differences in the intensity distribution of lateral scattered light.

[0114] The liquid-adaptive signal optimization step is the core technical aspect of this invention. The technical team uses a pre-trained liquid dynamic correction model to process the acquired raw fluorescence signals. This model adopts a liquid-state neural network architecture, including an input layer, a liquid reservoir layer, and an output layer. The liquid reservoir layer consists of 300 liquid neurons, each with time-varying membrane potential and liquid conduction characteristics.

[0115] As shown in Table 3, the liquid conduction adaptive function calculates the liquid conduction modulation value based on three parameters: the original fluorescence signal intensity, the background noise level, and the signal drift rate.

[0116] Table 3. Adjustment Parameters for Liquid Conduction Mode

[0117]

[0118] For the 128 samples tested, the technical team found that 42 samples had a liquid conduction modulation value between 0.1 and 0.3, and a low conduction mode was used to enhance the amplification of weak signals. 73 samples had a liquid conduction modulation value between 0.3 and 0.7, and a medium conduction mode was used to maintain a linear signal response. 13 samples had a liquid conduction modulation value between 0.7 and 1.0, and a high conduction mode was used to suppress signal saturation.

[0119] In the multi-marker concentration calculation stage, the technical team used a 4-parameter logistic regression model to fit the fluorescence signal intensity of the calibrator gradient solution to known concentrations. This model includes four key parameters: minimum response value, maximum response value, inflection point concentration, and slope parameter. After fitting calculations, the linear correlation coefficients of the standard curves for the five vascular inflammation markers were all greater than 0.995, meeting the quality control requirements. Figure 4 As shown.

[0120] Table 4 shows the distribution of test results for the 128 samples:

[0121] Table 4 Statistical Table of Vascular Inflammatory Marker Detection Results

[0122]

[0123] The technical team found that the highest proportion of abnormally elevated MMP-9 levels was 27.3%, which is consistent with the pathological mechanism of enhanced matrix metalloproteinase activity during vascular remodeling in patients with coronary artery disease. CD31, as a specific marker for vascular endothelial cells, showed an abnormally elevated proportion of 24.2%, reflecting the degree of endothelial dysfunction in patients.

[0124] During quality control testing, the technical team performed statistical analysis on all test results. The coefficients of variation for all five vascular inflammation markers were less than 12%, meeting the quality control requirement of less than 15%. Intra-batch repeatability testing showed that the coefficients of variation for high, medium, and low concentration levels were 6.8%, 8.4%, and 11.2%, respectively. Inter-batch repeatability testing was conducted continuously for 5 days, with the coefficient of variation controlled below 9.6%.

[0125] Table 5 shows the results of the detection precision evaluation for samples at different concentration levels:

[0126] Table 5. Results of Precision Assessment

[0127]

[0128] The technical team also assessed the impact of interfering substances. Under conditions of bilirubin concentration of 250 μmol / L, hemoglobin concentration of 4.2 g / L, and triglyceride concentration of 8.5 mmol / L, the interference to the kit's detection was within ±4.2%, far below the ±5% control standard. The cross-reactivity of inflammatory factors was also assessed. , , concentration At that time, the interference level was controlled within ±3.8%.

[0129] During data analysis, the technical team employed a liquid dynamic correction model to optimize the original fluorescence signal in real time. This model's liquid reservoir layer dynamically adjusts liquid conduction parameters based on signal characteristics, effectively eliminating the effects of background noise and system drift. Compared to traditional static correction algorithms, the liquid dynamic correction model improved signal processing accuracy by 23.6% and signal-to-noise ratio by 18.4%, as shown below. Figure 5 As shown.

[0130] Through comparative analysis, the technical team discovered that the liquid dynamic correction model exhibits excellent adaptability when processing fluorescence signals of varying intensities. For weak signals, the model automatically reduces the liquid damping coefficient to 0.08, extending the activation time of the liquid neurons and achieving signal amplification. For strong signals, the model increases the liquid damping coefficient to 0.42, accelerating the state decay rate and preventing signal saturation.

[0131] After completing the testing of all 128 samples, the technical team generated detailed test reports. The reports showed that 82.8% of patients had at least one elevated vascular inflammatory marker, with 46 patients having three or more abnormal markers, suggesting potentially severe vascular inflammation. This result provides important laboratory evidence for clinicians to assess patients' risk of coronary artery disease. Figure 6 As shown.

[0132] This invention represents a significant technological advancement over traditional single-marker detection methods. Traditional methods typically employ chemiluminescent immunoassay or enzyme-linked immunosorbent assay (ELISA), detecting only one marker at a time, requiring multiple separate experiments. This not only consumes large amounts of sample and reagents but also results in long detection cycles and difficulty in obtaining comprehensive vascular inflammation assessment information. This invention, through multiplex microsphere flow cytometry, achieves simultaneous detection of five vascular inflammation markers, significantly improving detection efficiency. The application of a liquid dynamic correction model further enhances detection accuracy and stability. By simulating the liquid environment characteristics of biological neural networks, it achieves adaptive processing of fluorescence signals of varying intensities, effectively eliminating matrix effects and signal interference problems common in traditional methods. The three-layer structure design of the sandwich immune complex ensures specificity and sensitivity, avoiding cross-reactivity. The application of flow cytometry enables precise identification and quantification of individual fluorescent microspheres, significantly reducing detection errors and variability compared to traditional batch detection methods.

[0133] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for detecting vascular inflammation markers using multiplex microsphere flow cytometry, characterized in that, include: A fluorescent microsphere mixture was prepared by coupling five different capture microsphere antibodies to the surface of polystyrene microspheres with different fluorescent codes, forming a fluorescent microsphere mixture with distinguishable fluorescent characteristics. Sample pretreatment involves mixing the sample to be tested with experimental buffer to prepare a calibrator gradient solution containing VEGF standard material, CD31 standard material, OPN standard material, MMP-9 standard material, and CD62P standard material. Sandwich immune complex formation involves mixing and incubating a sample or calibrator gradient solution with a fluorescent microsphere mixture and a biotin-labeled detection antibody to form a sandwich immune complex consisting of fluorescent microspheres, capture antibody, target antigen, and detection antibody. Fluorescent signal labeling: Phycoerythrin-labeled streptavidin is added to the sandwich immune complex, which binds to biotin-labeled detection antibody to generate a fluorescent signal; Flow cytometry fluorescence signal acquisition: Flow cytometry was used to detect phycoerythrin-labeled streptavidin-labeled fluorescent microspheres, acquiring the side-scattered light signal and fluorescence signal of each microsphere. Liquid adaptive signal optimization: A liquid dynamic correction model was used to dynamically expand the range of the acquired original fluorescence signal and eliminate background noise. The liquid dynamic correction model adjusted the liquid conduction parameters and liquid damping coefficient in real time according to the intensity of the original fluorescence signal. Multi-marker concentration calculation: The fluorescence signal optimized by the liquid dynamic correction model was fitted to a standard curve, and the calculation results were output.

2. The method for detecting vascular inflammation markers using multiplex microsphere flow cytometry according to claim 1, characterized in that, The steps for preparing the fluorescent microsphere mixture specifically involve selecting polystyrene microspheres with different fluorescence wavelengths as carriers, modifying the surface of the polystyrene microspheres with carboxyl functional groups for antibody conjugation, covalently binding VEGF-capturing microsphere antibodies, CD31-capturing microsphere antibodies, OPN-capturing microsphere antibodies, MMP-9-capturing microsphere antibodies, and CD62P-capturing microsphere antibodies to the surface of the polystyrene microspheres through a chemical cross-linking reaction of 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide hydrochloride and N-hydroxysuccinimide, and then mixing the different types of fluorescent microspheres in an equimolar ratio to form the fluorescent microsphere mixture.

3. The method for detecting vascular inflammation markers using multiplex microsphere flow cytometry according to claim 2, characterized in that, The steps for forming the sandwich immune complex specifically involve mixing 25 μL of sample or calibrator gradient solution with 25 μL of fluorescent microsphere mixture in a 5 mL flow cytometer; adding 25 μL of biotin-labeled detection antibody mixture, which contains biotin-labeled detection antibodies against VEGF, CD31, OPN, MMP-9, and CD62P; and incubating at room temperature with shaking at a speed of 400 to 500 rpm for 120 minutes.

4. The method for detecting vascular inflammation markers using multiplex microsphere flow cytometry according to claim 3, characterized in that, The fluorescent signal labeling step specifically involves adding 25 μL of phycoerythrin-labeled streptavidin solution to the sandwich immune complex; incubating at room temperature with shaking at a speed of 400 to 500 rpm for 30 minutes; introducing phycoerythrin fluorescent labeling into the sandwich immune complex through the binding of streptavidin and biotin; and generating a fluorescence signal with a maximum emission wavelength of 578 nm at an excitation wavelength of 566 nm.

5. The method for detecting vascular inflammation markers using multiplex microsphere flow cytometry according to claim 4, characterized in that, The flow cytometry fluorescence signal acquisition steps specifically involve using a 488nm laser from the flow cytometer to excite fluorescent microspheres to generate side-scattered light signals, identifying different coded fluorescent microsphere types based on the intensity and distribution characteristics of the side-scattered light signals; simultaneously, using a 561nm laser to excite phycoerythrin to generate fluorescence signals, and acquiring fluorescence signal intensity data using a photomultiplier tube detector.

6. The method for detecting vascular inflammation markers using multiplex microsphere flow cytometry according to claim 5, characterized in that, The specific structure of the liquid dynamic correction model is as follows: The liquid dynamic correction model adopts a liquid state neural network architecture, which includes three main components: an input layer, a liquid reservoir layer, and an output layer. The liquid reservoir layer consists of 200 to 500 liquid neurons. Each liquid neuron has a time-varying membrane potential and liquid conduction characteristics. The liquid conduction parameters are dynamically adjusted according to the intensity of the original fluorescence signal, and the liquid damping coefficient changes in real time according to the noise level of the original fluorescence signal.

7. The method for detecting vascular inflammation markers using multiplex microsphere flow cytometry according to claim 6, characterized in that, Before training the liquid dynamic correction model, the training dataset establishment step is also included: collecting the original fluorescence signal data of standard samples of vascular inflammation markers with different concentration gradients. The concentration range of the standard samples covers the linear detection range of each marker. At the same time, the corresponding background noise signal and system drift data are collected to establish a multi-dimensional training sample containing the original fluorescence signal intensity, noise level, and time drift characteristics.

8. The method for detecting vascular inflammation markers using multiplex microsphere flow cytometry according to claim 7, characterized in that, The training steps of the liquid dynamic correction model are as follows: the backpropagation algorithm is used to train the parameters of the liquid dynamic correction model. The loss function adopts the mean square error combined with the signal fidelity constraint. During the training process, the connection weights of the liquid reservoir layer and the basic values ​​of the liquid conduction parameters are optimized by the gradient descent method. The number of training rounds is set to be in the range of 500 to 1000 rounds. The learning rate adopts an adaptive adjustment strategy to gradually decrease from 0.01 to 0.

001.

9. The method for detecting vascular inflammation markers using multiplex microsphere flow cytometry according to claim 8, characterized in that, The liquid conduction adaptive function is used to adjust the liquid conduction parameters of the liquid dynamic correction model. The liquid conduction adaptive function is calculated based on three data: the original fluorescence signal intensity, the background noise level, and the signal drift rate, to obtain a liquid conduction adjustment value. When the liquid conduction adjustment value is in the range of 0.1 to 0.3, the low conduction mode is used to enhance the amplification effect of weak signals and adjust the liquid conduction parameters.

10. The method for detecting vascular inflammation markers using multiplex microsphere flow cytometry according to claim 9, characterized in that, The steps for calculating the concentration of multiple biomarkers specifically involve using a 4-parameter logistic regression model to fit the fluorescence signal intensity of the calibrator gradient solution to the known concentration. The 4-parameter logistic regression model includes four key parameters: minimum response value, maximum response value, inflection point concentration, and slope parameter. Based on the fitted standard curve equation, the concentration results of VEGF, CD31, OPN, MMP-9, and CD62P in the unknown sample are calculated and output.

Images