Method for detecting external vesicle marker through high-flux nano plasma exciting light immune color development

Abstract

The invention provides a method for detecting an external vesicle marker through high-flux nano plasma exciting light immune color development, and belongs to the technical field of human extracellular vesicles. After a serum sample is subjected to centrifugal treatment, the serum sample and a CD81 capture antibody substrate are incubated, and the particle size distribution of the vesicles is monitored in real time; a zwitterionic polymer modified gold nanoparticle LAM detection probe and a polyethylene glycol modified silver nanoparticle LprG detection probe are prepared to be specifically combined with a vesicle surface antigen, a chromogenic enhancement solution is adopted to induce a plasma resonance signal, and full-hole scanning imaging is carried out; and constructing a double-layer game optimization model to cooperatively optimize the detection sensitivity and the signal stability, carrying out weighted summation on normalized signals of the particle size subgroups to obtain comprehensive detection signal intensity, comparing the comprehensive detection signal intensity with a threshold value, and outputting a final judgment result. The technical problem that quantitative accuracy is affected by signal intensity deviation caused by vesicle particle size difference in outer vesicle marker detection is solved.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of human extracellular vesicles, and in particular relates to a method for high-throughput nano-plasmonic excitation light immuno-chromogenic detection of extracellular vesicle markers. BACKGROUND

[0002] In the early diagnosis of infectious diseases such as tuberculosis, immunodetection based on extracellular vesicle markers LAM and LprG is an important diagnostic means. In clinical practice, nano-plasmonic resonance technology is usually used to quantitatively analyze the surface antigens of extracellular vesicles in serum samples. The traditional extracellular vesicle marker detection method treats vesicles of different particle sizes as a homogeneous population during signal acquisition, without considering the influence of vesicle particle size difference on surface antigen density and plasmonic resonance signal intensity. Due to the significant difference in antigen loading between extracellular vesicles in the range of 30 to 200 nanometers due to the difference in membrane surface area to volume ratio, the signal intensity of small particle size vesicles is significantly lower than that of large particle size vesicles. This particle size-dependent systematic bias makes the detection signals of vesicles of different sizes cannot be directly compared, resulting in distortion of the quantitative results of antigen expression. That is, there is a technical problem in the prior art that the detection of extracellular vesicle markers is affected by the signal intensity bias caused by the difference in vesicle particle size, which affects the quantitative accuracy. SUMMARY

[0003] Therefore, the present application provides a method for high-throughput nano-plasmonic excitation light immuno-chromogenic detection of extracellular vesicle markers, which can solve the technical problem in the prior art that the detection of extracellular vesicle markers is affected by the signal intensity bias caused by the difference in vesicle particle size, which affects the quantitative accuracy.

[0004] The application is achieved as follows: the application provides a high-throughput nano-plasmonic excitation light immuno-chromogenic detection method for extracellular vesicle markers, comprising the following steps: adding a serum sample into a detection hole of a CD81 capture antibody substrate and incubating to allow extracellular vesicles in the serum to bind to the CD81 capture antibody on the surface of the substrate, and monitoring the vesicle particle size distribution in real time during the incubation by using dynamic light scattering technology; coupling amphoteric ion polymer modified gold nanoparticles with LAM detection antibodies to prepare LAM detection probes, and coupling polyethylene glycol modified silver nanoparticles with LprG detection antibodies to prepare LprG detection probes; adding the LAM detection probes and the LprG detection probes into the detection hole of the incubated substrate respectively to allow the detection probes to specifically bind to LAM antigens and LprG antigens on the surface of the vesicles to form immune complexes; adding a chromogenic enhancer solution into the washed detection hole of the substrate to induce plasmonic resonance signal enhancement on the surface of the nanoparticles; collecting the plasmonic resonance chromogenic signals in each detection hole and establishing a particle size correction signal data stream; establishing a particle size classification interval according to the vesicle particle size data obtained by the dynamic light scattering technology, and establishing a correction curve of signal intensity and antigen expression for different particle size subpopulations, dividing the original signal intensity value of each subpopulation by the correction coefficient of the corresponding particle size interval to obtain the normalized signal intensity; establishing a subpopulation weight coefficient according to the positive detection rate of different particle size subpopulations, summing up the weighted signal intensity of each subpopulation to obtain a comprehensive detection signal intensity, and outputting a final positive determination result when the comprehensive detection signal intensity exceeds a positive threshold, wherein the serum sample is supernatant obtained by centrifugation of human peripheral venous blood samples in a spot market area.

[0005] The amphoteric ion polymer modification refers to connecting a polymer molecule layer containing positive and negative charge groups on the surface of the gold nanoparticles by chemical bonding to form a protective layer structure with charge balance characteristics, and enhancing the dispersion stability of the nanoparticles in a high-salt and high-protein environment through the electrostatic interaction between the simultaneously existing positive and negative charge groups to prevent the particles from agglomeration and sedimentation due to electrostatic attraction.

[0006] The polyethylene glycol modification refers to grafting polyethylene glycol long-chain molecules on the surface of the silver nanoparticles, forming a dense hydrated protective layer on the particle surface by using the steric hindrance effect of the polyethylene glycol chain to prevent protein molecules and other biological macromolecules in the serum from approaching the particle surface and reduce signal interference and particle inactivation caused by non-specific adsorption.

[0007] In the coupling detection probe suspension, bovine serum albumin and Tween-20 are added to form a stabilization buffer system, and the surface potential change of the probes is continuously monitored by a Zeta potential detector.

[0008] The stable buffer system refers to a multi-component protective solution system composed of bovine serum albumin and non-ionic surfactant Tween-20 added in the suspension of nanoparticle detection probes. The bovine serum albumin reduces the false adsorption of serum proteins by occupying non-specific binding sites on the surface of nanoparticles, and the Tween-20 maintains the monodisperse state of nanoparticles by reducing the surface tension of the liquid and forming micellar structures.

[0009] The particle size correction signal data stream refers to a time sequence data sequence formed by synchronously matching the outer vesicle particle size information obtained in real time by the dynamic light scattering technology with the plasmonic resonance color development signal collected by the microscopic scanning imaging system. Each data point in the data stream contains two-dimensional information of the vesicle particle size value and the signal intensity value at the corresponding time.

[0010] The window characteristic vector sequence is a multi-dimensional vector set formed by arranging the statistical characteristic parameters extracted from the continuous multiple sliding windows in time sequence. Each characteristic vector contains four descriptive parameters of the mean, standard deviation, peak value and signal distribution skewness of the signal intensity in the corresponding window.

[0011] The window characteristic vector sequence is a multi-dimensional vector set formed by arranging the statistical characteristic parameters extracted from the continuous multiple sliding windows in time sequence. Each characteristic vector contains four descriptive parameters of the mean, standard deviation, peak value and signal distribution skewness of the signal intensity in the corresponding window.

[0012] The particle size classification interval is a classification standard for dividing the total particle size range of 30 to 200 nm into three discontinuous intervals according to the biophysical characteristics and light scattering behavior of the outer vesicle. The vesicle signal with a particle size range of 30 to 60 nm and containing 30 nm but not containing 60 nm is classified as a small particle size vesicle subpopulation, the vesicle signal with a particle size range of 60 to 120 nm and containing 60 nm but not containing 120 nm is classified as a medium particle size vesicle subpopulation, and the vesicle signal with a particle size range of 120 to 200 nm and containing 120 nm and 200 nm is classified as a large particle size vesicle subpopulation.

[0013] The double-layer game optimization model includes an upper model with the goal of maximizing detection sensitivity and a lower model with the goal of optimizing signal stability. The input of the objective function of the upper model includes normalized signal intensity, filter gain parameter and vesicle capture efficiency, and the input of the objective function of the lower model includes standard deviation of the window characteristic vector sequence, Zeta potential change rate and background signal intensity. The objective function of the upper model is associated with the objective function of the lower model through a coupling term signal quality factor.

[0014] The normalized signal intensity refers to a standardized value obtained by dividing the original plasmon resonance signal intensity value of different particle size subpopulations by the corresponding correction coefficient of the particle size subpopulation.

[0015] The frequency domain feature spectrum is obtained by performing a fast Fourier transform on the window feature vector sequence, when the proportion of low frequency components with a frequency range of 0.1 Hz to 1 Hz and containing 0.1 Hz but not containing 1 Hz in the frequency domain feature spectrum is more than 60%, it is determined that there is baseline drift interference and the cutoff frequency of the high-pass filter is increased, when the proportion of medium frequency components with a frequency range of 1 Hz to 10 Hz and containing 1 Hz but not containing 10 Hz is more than 50%, it is determined that there is environmental vibration interference and the center frequency of the band-stop filter is adjusted to the peak frequency of the interference, when the proportion of high frequency components with a frequency range of 10 Hz to 50 Hz and containing 10 Hz and 50 Hz is more than 40%, it is determined that there is electromagnetic interference and the attenuation slope of the low-pass filter is increased.

[0016] The signal quality factor is a comprehensive evaluation parameter for characterizing the signal-to-noise ratio of the detection signal and the stability of the system, which is determined by the upper-layer optimization of detection sensitivity and the lower-layer optimization of signal stability, and is used as a coupling term to connect the upper-layer model and the lower-layer model in the two-layer game optimization model. The signal quality factor is equal to the geometric mean of the upper-layer optimized detection sensitivity and the lower-layer optimized signal stability.

[0017] The normalized signal intensity after filtering is compared with the preset LAM positive threshold and LprG positive threshold, when the LAM signal intensity is greater than or equal to the LAM positive threshold for three consecutive detection results, it is determined that the LAM is in a sustained positive state and is marked as a strong positive result, when the LprG signal intensity is greater than or equal to the LprG positive threshold for three consecutive detection results, it is determined that the LprG is in a sustained positive state and is marked as a strong positive result, and when only a single detection result exceeds the threshold, an encryption detection process is started and five repeated scans are completed within 30 seconds.

[0018] The standard signal feature vector of the negative quality control and the positive quality control is established, and the cosine similarity between the window feature vector sequence of the sample to be tested and the standard signal feature vector is calculated. When the similarity between the sample and the negative quality control feature vector is greater than 0.95 and the similarity between the sample and the positive quality control feature vector is less than 0.3, it is determined as a negative result, when the similarity between the sample and the positive quality control feature vector is greater than 0.85 and the similarity between the sample and the negative quality control feature vector is less than 0.4, it is determined as a positive result, and when the similarity does not satisfy the above conditions, the background subtraction processing intensity is increased and the signal intensity is recalculated.

[0019] The subpopulation weight coefficient is a normalized weight factor given according to the detection rate and signal stability of different particle size vesicle subpopulations in a positive sample, the detection signal of small particle size vesicle subpopulation is multiplied by the weight coefficient 0.8, the detection signal of medium particle size vesicle subpopulation is multiplied by the weight coefficient 1.0, and the detection signal of large particle size vesicle subpopulation is multiplied by the weight coefficient 1.2.

[0020] The present application divides the outer vesicle population into three subpopulations of small particle size, medium particle size and large particle size by real-time monitoring of vesicle particle size distribution by dynamic light scattering technology and establishing particle size classification intervals, and establishes a correction curve of signal intensity and antigen expression for different particle size subpopulations, and divides the original signal intensity of each subpopulation by the correction coefficient of the corresponding particle size interval to obtain the normalized signal intensity. The systematic influence of vesicle size on the plasma resonance signal is eliminated through particle size correction processing, so that the detection signals of vesicles of different particle sizes have the same dimension and comparability, and can truly reflect the absolute expression level of the antigen rather than the apparent signal affected by the size of the vesicle. When comprehensive detection is performed, weight coefficients based on detection rate and signal stability are given to different particle size subpopulations, and the normalized signals of each subpopulation are integrated to obtain the comprehensive detection signal intensity, thereby realizing accurate quantitative detection of the outer vesicle marker and solving the technical problem that the signal intensity deviation caused by the difference in vesicle particle size affects the quantitative accuracy in the prior art. BRIEF DESCRIPTION OF DRAWINGS

[0021] Figure 1 The flowchart of the method of the present application.

[0022] Figure 2 The LAM and LprG signal intensity distribution graph of 36 samples in Example 1.

[0023] Figure 3 The signal comparison graph before and after filtering of a typical sample in Example 1. DETAILED DESCRIPTION

[0024] In order to make the purpose, technical scheme and advantages of the embodiments of the present application more clear, the technical scheme in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application.

[0025] As shown in Figure 1 The flowchart of the method of the present application provided is a high-throughput nano-plasmonic excitation light immunochromatographic detection method for an outer vesicle marker, and the method comprises the following steps:

[0026] S01, 50 μL of serum sample treated by centrifugation is added to the detection hole of the CD81 capture antibody substrate, and incubated at 37°C for 60 minutes to combine the extracellular vesicles in the serum with the CD81 capture antibody on the surface of the substrate, and the dynamic light scattering technology is used to monitor the vesicle particle size distribution in real time during the incubation process;

[0027] S02, coupling the zwitterionic polymer modified gold nanoparticles with LAM detection antibody to prepare LAM detection probe, coupling the polyethylene glycol modified silver nanoparticles with LprG detection antibody to prepare LprG detection probe, adding 0.5 g / L of bovine serum albumin and 0.05% of Tween-20 by volume fraction into the suspension of the coupled detection probe to form a stabilization buffer system, and continuously monitoring the surface potential change of the probe by a Zeta potential detector;

[0028] S03, adding 20 μL of LAM detection probe and LprG detection probe respectively into the incubated substrate detection holes, and oscillating and incubating at room temperature for 60 minutes to make the detection probe specifically bind to the LAM antigen and LprG antigen on the surface of the vesicle to form an immune complex, and washing 3 times with phosphate washing buffer containing Tween-20 to remove the unbound probe after the incubation;

[0029] S04, preparing color development enhancement solution, mixing 100 μL of color developer A cationic solution with 100 μL of color developer B cationic chelating agent solution, and adding 500 μL of color developer C dissolved in 1 mL of purified water into the washed substrate detection holes, and incubating at room temperature in the dark for 15 minutes to induce the enhancement of the surface plasmon resonance signal of the nanoparticles;

[0030] S05, using a full-automatic microscopic scanning imaging system to scan the substrate, collecting the plasmon resonance color development signal in each detection hole, and establishing a particle size correction signal data stream, and performing sliding window analysis on the continuously collected color development signal according to 100 frames as a unit, and extracting the mean and standard deviation of the signal intensity in each window to form a window feature vector sequence;

[0031] S06, establishing a particle size classification interval according to the vesicle particle size data obtained by dynamic light scattering technology, classifying the vesicle signal with a particle size range of [30, 60) nm as a small particle size vesicle subpopulation, classifying the vesicle signal with a particle size range of [60, 120) nm as a medium particle size vesicle subpopulation, and classifying the vesicle signal with a particle size range of [120, 200] nm as a large particle size vesicle subpopulation, and establishing a correction curve of signal intensity and antigen expression for different particle size subpopulations, and dividing the original signal intensity value of each subpopulation by the correction coefficient of the corresponding particle size interval to obtain the normalized signal intensity;

[0032] S07. Perform a fast Fourier transform on the window feature vector sequence to obtain the frequency domain feature spectrum. When the proportion of low-frequency components in the frequency domain feature spectrum with a frequency range of [0.1, 1) Hz exceeds 60%, it is determined to be baseline drift interference and the cutoff frequency of the high-pass filter is increased to 1.5 Hz. When the proportion of mid-frequency components in the frequency range of [1, 10) Hz exceeds 50%, it is determined to be environmental vibration interference and the center frequency of the band-stop filter is adjusted to the interference peak frequency. When the proportion of high-frequency components in the frequency range of [10, 50] Hz exceeds 40%, it is determined to be electromagnetic interference and the attenuation slope of the low-pass filter is increased to 80 dB / decibels.

[0033] S08. Construct a two-layer game optimization model to adaptively adjust the detection parameters. The two-layer game optimization model includes an upper-layer model with the goal of maximizing detection sensitivity and a lower-layer model with the goal of optimizing signal stability. The objective function input of the upper-layer model includes normalized signal strength, filter gain parameters, and vesicle capture efficiency, and the output is the upper-layer optimized detection sensitivity. The objective function input of the lower-layer model includes the standard deviation of the window feature vector sequence, the rate of change of Zeta potential, and the background signal strength, and the output is the lower-layer optimized signal stability. The objective functions of the upper-layer model and the lower-layer model are related through a coupling term signal quality factor.

[0034] S09, compare the normalized signal strength after filtering with the preset LAM positive threshold ( ) and LprG positive threshold ( The comparison was performed when the LAM signal strength was greater than or equal to the specified value in three consecutive tests. When the LprG signal intensity is determined to be persistently positive and marked as strongly positive, it is considered a positive result when the LprG signal intensity is greater than or equal to the LprG positive threshold for three consecutive tests. When the result is determined to be a persistent positive LprG and marked as a strong positive result, when only a single test result exceeds the threshold, the encrypted testing process is initiated and 5 repeated scans are completed within 30 seconds.

[0035] S10. Establish standard signal feature vectors for negative and positive quality control samples. Calculate the cosine similarity between the window feature vector sequence of the test sample and the standard signal feature vector. When the similarity between the sample and the negative quality control feature vector is greater than 0.95 and the similarity with the positive quality control feature vector is less than 0.3, it is judged as a negative result. When the similarity between the sample and the positive quality control feature vector is greater than 0.85 and the similarity with the negative quality control feature vector is less than 0.4, it is judged as a positive result. When the similarity does not meet the above conditions, the background subtraction processing intensity is increased and the signal intensity is recalculated.

[0036] S11. Establish subgroup weighting coefficients based on the positive detection rates of different particle size subgroups. Multiply the detection signal of the small particle size vesicle subgroup by a weighting coefficient of 0.8, the detection signal of the medium particle size vesicle subgroup by a weighting coefficient of 1.0, and the detection signal of the large particle size vesicle subgroup by a weighting coefficient of 1.2. Sum the weighted signal intensities of each subgroup to obtain the comprehensive detection signal intensity. When the comprehensive detection signal intensity exceeds the positive threshold, output the final positive judgment result.

[0037] Among them, zwitterionic polymer modification refers to connecting polymer molecular layers containing positive and negative charged groups on the surface of gold nanoparticles through chemical bonding to form a protective layer structure with charge balance characteristics. The electrostatic interaction between the coexisting positive and negative charged groups enhances the dispersion stability of nanoparticles in high-salt and high-protein environments, preventing the particles from agglomerating and settling due to electrostatic attraction.

[0038] Polyethylene glycol (PEG) modification refers to grafting long-chain PEG molecules onto the surface of silver nanoparticles. The steric hindrance effect of the PEG chains forms a dense hydration protective layer on the particle surface, preventing protein molecules and other biomolecules in serum from approaching the particle surface and reducing signal interference and particle inactivation caused by non-specific adsorption.

[0039] The stabilization buffer system refers to a multi-component protective solution system composed of bovine serum albumin and nonionic surfactant Tween-20 added to the nanoparticle detection probe suspension. Bovine serum albumin reduces the false adsorption of serum proteins by occupying non-specific binding sites on the surface of nanoparticles, while Tween-20 maintains the monodisperse state of nanoparticles by reducing the surface tension of the liquid and forming a micelle structure. The synergistic effect of the two ensures that the detection probe maintains stable optical properties and binding activity after being mixed with serum samples.

[0040] The particle size correction signal data stream refers to a time-series data sequence formed by synchronously matching the outer vesicle particle size information acquired in real time by dynamic light scattering technology with the plasma resonance colorimetric signal collected by the microscopic scanning imaging system. Each data point in the data stream contains two dimensions of information: the vesicle particle size value and the signal intensity value at the corresponding time, which are used to subsequently establish a particle size-dependent signal correction model.

[0041] Sliding window analysis refers to dividing a continuously acquired colorimetric signal sequence into partially overlapping time windows according to a fixed number of frames. Each window contains 100 frames of image data, and the sliding step between windows is 50 frames. Statistical characteristic parameters are calculated for the signal data within each window. By continuously sliding the windows, the dynamic tracking of signal change trends and timely identification of abnormal fluctuations can be achieved.

[0042] The window feature vector sequence is a multidimensional vector set formed by arranging statistical feature parameters extracted from multiple consecutive sliding windows in chronological order. Each feature vector contains four descriptive parameters: mean, standard deviation, peak value, and signal distribution skewness of the signal intensity within the corresponding window. The vector sequence reflects the evolution and statistical distribution characteristics of the detected signal over time and is used for signal quality assessment and abnormal pattern recognition.

[0043] Particle size grading is a classification standard that divides the total particle size range of 30 to 200 nm into three discontinuous intervals based on the biophysical characteristics and light scattering behavior of the outer vesicles. The surface density of CD81 antibody and LAM antigen or LprG antigen differs due to the different membrane surface area to volume ratio of vesicles in different particle size intervals, which in turn affects the capture efficiency and signal intensity. Grading process eliminates the systematic bias of particle size on quantitative detection.

[0044] Small-diameter vesicle subpopulation refers to extracellular vesicle populations with a diameter range of [30, 60) nm. Due to their smaller membrane surface area, these small-diameter vesicle subpopulations have lower antigen loading and weaker plasma resonance signal intensity.

[0045] The medium-sized vesicle subpopulation refers to the extracellular vesicle population with a particle size range of [60, 120) nm. The medium-sized vesicle subpopulation is the most widely distributed vesicle type in clinical samples, with moderate antigen loading and stable signal response characteristics.

[0046] Large-diameter vesicles refer to extracellular vesicle populations with a diameter range of [120, 200] nm. These large-diameter vesicles have a larger membrane surface area, resulting in higher antigen loading and stronger plasma resonance signal intensity.

[0047] Normalized signal intensity refers to the standardized value obtained by dividing the original plasma resonance signal intensity value of different particle size subgroups by the correction coefficient corresponding to the particle size subgroup. The correction coefficient is obtained by detecting standard vesicle samples with known antigen concentrations. Normalization makes the signal intensity of vesicles of different particle sizes have the same dimensions and comparability, and can truly reflect the absolute expression level of the antigen rather than the apparent signal affected by vesicle size.

[0048] The Fast Fourier Transform (FFT) is a mathematical algorithm that converts a time-domain signal into a frequency-domain representation. By performing the FFT operation on the signal intensity time series in the window feature vector sequence, the energy distribution spectrum of the signal at different frequency components is obtained. Frequency domain analysis can identify the sources of periodic interference and frequency-specific noise, providing frequency characteristic basis for adaptive filtering.

[0049] The frequency domain feature spectrum refers to the frequency-energy intensity correspondence obtained by converting the window feature vector sequence from the time domain to the frequency domain through fast Fourier transform. The horizontal axis of the frequency domain feature spectrum is the frequency unit Hz, and the vertical axis is the normalized energy intensity, which is used to identify the proportion of interference components in different frequency bands.

[0050] Baseline drift interference refers to the low-frequency fluctuation phenomenon in which the average level of the detected signal slowly rises or falls over time. It is mainly caused by factors such as temperature changes, light source aging, and incomplete chemical reactions. The presence of baseline drift interference can be determined by analyzing the energy proportion of low-frequency components in the frequency domain characteristic spectrum. When the proportion of low-frequency components exceeds 60%, the cutoff frequency of the high-pass filter is increased to attenuate the low-frequency interference components.

[0051] Environmental vibration interference refers to the mid-frequency periodic signal fluctuations caused by mechanical vibrations transmitted to the detection platform from building vibrations, equipment operation, and personnel movement. Vibration interference manifests as energy peaks in the frequency range of [1, 10) Hz in the frequency domain characteristic spectrum. Narrow-band suppression can be achieved by precisely aligning the center frequency of the band-stop filter with the interference peak frequency, thereby eliminating the impact of vibration on the clarity of microscopic imaging and signal stability.

[0052] Electromagnetic interference refers to high-frequency noise signals introduced by the coupling of high-frequency electromagnetic fields generated by power lines, motor drives, and wireless communication devices to the detection circuit. In the frequency domain characteristic spectrum, it is a broadband energy distribution in the frequency range of [10, 50] Hz. By increasing the attenuation slope of the low-pass filter to 80 dB / decade, the suppression capability of high-frequency components is enhanced, ensuring that high-frequency electromagnetic noise will not overwhelm biological signals.

[0053] The two-layer game optimization model refers to a hierarchical decision optimization framework consisting of an upper-layer model and a lower-layer model. The upper-layer model adjusts the filter gain parameters and vesicle acquisition time with the optimization objective of maximizing detection sensitivity, while the lower-layer model adjusts the background subtraction processing intensity and the number of repeated detections with the optimization objective of signal stability. The two models influence each other through the coupling term signal quality factor to form a game equilibrium, thereby achieving the synergistic optimization of detection sensitivity and signal stability.

[0054] The upper-level model refers to the decision layer responsible for optimizing detection sensitivity in the two-level game optimization model. The objective function of the upper-level model is used to calculate the response capability of the detection system to the target antigen. The inputs include normalized signal strength, filter gain parameter, and vesicle capture efficiency. The output is the upper-level optimized detection sensitivity. The objective function of the upper-level model is expressed as follows: The upper-level optimized detection sensitivity is equal to the product of normalized signal strength divided by the standard reference value of normalized signal strength and filter gain parameter divided by the standard reference value of filter gain parameter, multiplied by the square root of vesicle capture efficiency divided by the standard reference value of vesicle capture efficiency, and multiplied by the signal quality factor. The constraints of the upper-level model include the filter gain parameter being in the range of [0.5, 2.0] and the vesicle capture efficiency being in the range of (0, 1).

[0055] The lower-level model refers to the decision layer responsible for signal stability optimization in the two-level game optimization model. The objective function of the lower-level model is used to calculate the temporal fluctuation degree and noise suppression effect of the detected signal. The input includes the standard deviation of the window feature vector sequence, the rate of change of Zeta potential, and the background signal intensity. The output is the lower-level optimized signal stability. The objective function of the lower-level model is expressed as follows: The lower-level optimized signal stability is equal to 1 minus the product of the standard deviation of the window feature vector sequence divided by the standard reference value of the standard deviation of the window feature vector sequence and the rate of change of Zeta potential divided by the standard reference value of the rate of change of Zeta potential, minus the product of the absolute value of the natural logarithm function of the background signal intensity divided by the standard reference value of the background signal intensity, and multiplied by the signal quality factor. The constraints of the lower-level model include that the standard deviation of the window feature vector sequence is less than 20% of the normalized signal intensity and the absolute value of the rate of change of Zeta potential is less than 5mV / min.

[0056] The filter gain parameter refers to the amplification or attenuation factor of high-pass filter, band-stop filter and low-pass filter for signals in each frequency band. The filter gain parameter is dynamically adjusted by the upper-level model according to the detection sensitivity optimization requirements. The standard reference value of the filter gain parameter is 1.0.

[0057] Vesicle capture efficiency refers to the proportion of CD81 capture antibodies on the substrate surface that successfully bind to extracellular vesicles in a serum sample. The vesicle capture efficiency is determined by incubation time, temperature, and antibody density. The standard reference value for the vesicle capture efficiency is 0.75.

[0058] Zeta potential change rate refers to the rate at which the surface potential of the nanoparticle detection probe changes over time, reflecting the dispersion stability of the nanoparticles in the stabilized buffer system. The unit of the zeta potential change rate is mV / min. The zeta potential change rate is continuously monitored by the zeta potential detector in step S02, and the standard reference value of the zeta potential change rate is 1.0 mV / min.

[0059] Background signal intensity refers to the level of nonspecific signal detected in negative control samples that do not contain the target antigen. This background signal intensity primarily originates from the nonspecific adsorption of nanoparticles and the background noise of the detection system. The standard reference value for this background signal intensity is... .

[0060] The signal quality factor is a comprehensive evaluation parameter that characterizes the signal-to-noise ratio of the detected signal and the stability of the system. The signal quality factor is jointly determined by the upper-level optimized detection sensitivity and the lower-level optimized signal stability. As a coupling term in the two-level game optimization model, it connects the upper-level model and the lower-level model. The signal quality factor is expressed as follows: The signal quality factor is equal to the geometric mean of the upper-level optimized detection sensitivity and the lower-level optimized signal stability.

[0061] A sustained positive status refers to a stable detection result in which the LAM signal intensity or LprG signal intensity of the sample exceeds the preset positive judgment threshold in three consecutive independent tests. The consistency of multiple consecutive tests indicates that there is indeed a high concentration of the target antigen in the sample, rather than accidental signal fluctuations or single measurement errors. The determination of a sustained positive status reduces the false positive rate and improves the reliability of the diagnostic conclusion.

[0062] The encrypted detection process refers to a high-frequency repeated detection program that is automatically activated when a single test result exceeds the positive threshold but does not meet the confirmation condition of three consecutive positive results. It completes five independent scan measurements on the same sample within 30 seconds. By increasing the number of tests and shortening the test interval, statistical data is quickly accumulated, and boundary samples are reconfirmed to distinguish between true positive signals and false positive signals.

[0063] The standard signal feature vector is a set of representative feature parameters extracted through statistical analysis of a large number of negative and positive quality control test data. The standard signal feature vector of the negative quality control reflects the background signal pattern when no antigen is present, while the standard signal feature vector of the positive quality control reflects the signal response pattern of antigen with known concentration. The standard signal feature vector serves as a reference benchmark for evaluating the signal feature attribution of the test sample.

[0064] Cosine similarity is a mathematical index that measures the degree of directional consistency between two multidimensional vectors. It is calculated by dividing the inner product of the two vectors by the product of their respective magnitudes. The similarity value is between 0 and 1. The closer the value is to 1, the closer the directions of the two vectors are, that is, the more similar the feature patterns are. By calculating the cosine similarity between the sample feature vector and the standard signal feature vector, the degree of matching between the sample signal pattern and the negative or positive quality control product is determined.

[0065] The negative control feature vector refers to the standard signal feature vector extracted from the test data of negative controls that do not contain LAM antigen and LprG antigen. The negative control feature vector includes four parameters: mean signal intensity, standard deviation, peak value, and signal distribution skewness.

[0066] The positive control feature vector refers to the standard signal feature vector extracted from the test data of positive control samples containing known concentrations of LAM antigen and LprG antigen. The positive control feature vector includes four parameters: mean signal intensity, standard deviation, peak value, and signal distribution skewness.

[0067] Background subtraction intensity refers to the subtraction ratio used when subtracting the background signal intensity from the original signal intensity of the sample to be tested. The background subtraction intensity is dynamically adjusted by the lower-level model according to the signal stability optimization requirements. When the similarity between the sample signal features and the standard signal feature vector does not meet the judgment conditions, the background subtraction intensity is increased.

[0068] The subpopulation weighting coefficient is a normalized weighting factor assigned based on the detection rate and signal stability of different particle size vesicle subpopulations in positive samples. Small particle size vesicle subpopulations are assigned a weighting coefficient of 0.8 due to their lower surface antigen density and limited capture efficiency. Medium particle size vesicle subpopulations, as the main detection target, are assigned a baseline weighting coefficient of 1.0. Large particle size vesicle subpopulations are assigned an enhancement weighting coefficient of 1.2 due to their large surface area and high total antigen content. The weighting process balances the contribution of different particle size subpopulations to the final detection results.

[0069] The comprehensive detection signal intensity is a comprehensive evaluation index obtained by multiplying the normalized signal intensities of small-diameter vesicle subpopulations, medium-diameter vesicle subpopulations, and large-diameter vesicle subpopulations by their respective subpopulation weighting coefficients and then summing them. The comprehensive detection signal intensity integrates all the information about antigens carried by vesicles of different sizes, eliminating the one-sidedness and uncertainty of single-size subpopulation detection. The comprehensive detection signal intensity is compared with the positive judgment threshold to obtain the final negative or positive result.

[0070] The specific implementation methods of the above steps are described in detail below.

[0071] The specific implementation of step S01 is as follows: First, 50 μL of centrifuged serum sample is precisely added to the detection well of the substrate pre-fixed with CD81 capture antibody. The substrate is then placed in a 37°C constant temperature incubator and incubated for 60 minutes. During the incubation process, extracellular vesicles in the serum undergo a specific immune binding reaction with the CD81 capture antibody fixed on the substrate surface through CD81 molecules on their membrane surface. At the same time, a dynamic light scattering detection system is activated to monitor the changes in the particle size distribution of vesicles in the incubation solution in real time. The purpose of this step is to achieve selective capture of extracellular vesicles and synchronous acquisition of particle size information, providing basic data for subsequent particle size classification correction.

[0072] Among them, the centrifuged serum sample is obtained in advance. Optionally, the method of obtaining it is as follows: after the human peripheral venous serum sample is left to stand at room temperature for 30 minutes, it is centrifuged at 1000 g for 10 minutes, the supernatant is collected, and after centrifugation at 10000 g for 20 minutes at 4℃, the supernatant is taken as the sample to be tested.

[0073] The specific implementation of step S02 involves using a chemical bonding method to connect a zwitterionic polymer molecular layer containing positively and negatively charged groups to the surface of gold nanoparticles, forming a protective structure with charge balance characteristics. LAM detection antibodies are then immobilized on the modified gold nanoparticle surface using covalent coupling technology to prepare LAM detection probes. A similar method is used to graft polyethylene glycol long-chain molecules onto the surface of silver nanoparticles, forming a hydration protective layer using steric hindrance, followed by coupling with LprG detection antibodies to prepare LprG detection probes. Bovine serum albumin (BSA) at a mass concentration of 0.5 g / L and the nonionic surfactant Tween-20 at a volume fraction of 0.05% are added to the prepared detection probe suspension to construct a multi-component stabilizing buffer system to prevent probe aggregation and inactivation in a high-salt, high-protein environment. The rate of change of the surface potential of the probe suspension over time is continuously monitored using a Zeta potential detector. The purpose of this step is to prepare nanoparticle detection probes with high specificity and high stability, ensuring that the probes maintain good dispersibility and binding activity in subsequent immune reactions.

[0074] The specific implementation of step S03 involves taking the substrate incubated in step S01, adding 20 μL each of the LAM detection probe and LprG detection probe to each detection well, and placing the substrate on a horizontal shaking platform at room temperature and shaking at an appropriate frequency for 60 minutes. The shaking process promotes full contact between the detection probe and the LAM antigen and LprG antigen on the surface of the captured vesicles, resulting in a specific immune binding reaction and forming a sandwich-type immune complex composed of capture antibody, vesicle, and detection antibody. After incubation, the substrate is washed with phosphate washing buffer containing an appropriate amount of Tween-20. Each time, the substrate is fully covered on the surface of the detection wells, allowed to stand for a moment, and then the washing solution is aspirated. This washing is repeated 3 times to thoroughly remove free detection probes that have not bound to the target and non-specifically adsorbed nanoparticles. The purpose of this step is to establish a complete immune recognition complex and remove interfering signal sources, providing a clean reaction interface for subsequent plasma resonance signal detection.

[0075] The specific implementation of step S04 involves preparing a colorimetric enhancement solution. First, 100 μL of chromogenic agent A cationic solution and 100 μL of chromogenic agent B cationic chelating agent solution are thoroughly mixed in a clean container. Then, 500 μL of chromogenic agent C, which has been dissolved in 1 mL of purified water, is added and mixed thoroughly to obtain a stable colorimetric enhancement solution. 10 μL of the colorimetric enhancement solution is precisely added to each detection well of the washed substrate. The substrate is then incubated in a dark environment at room temperature for 15 minutes. During incubation, the cations in the colorimetric enhancement solution interact with the plasma on the surface of the gold and silver nanoparticles. By changing the dielectric environment and charge distribution around the nanoparticles, the surface plasmon resonance frequency is induced to redshift and the intensity of scattered light is enhanced. The purpose of this step is to amplify the optical signal of the nanoparticles through a chemical enhancement mechanism, thereby improving the sensitivity of the detection system to low-concentration antigens.

[0076] The specific implementation of step S05 involves starting a fully automated microscopic scanning imaging system to perform a full-field scanning of the substrate, one well at a time. The system uses a high numerical aperture objective lens and a high-sensitivity image sensor to continuously acquire the plasma resonance color development signal in each detection well. Simultaneously, the vesicle particle size data obtained by dynamic light scattering technology is time-stamped with the signal intensity data acquired by the imaging system to establish a synchronous time-series data stream containing particle size information and signal intensity information. The continuously acquired color development signal image sequence is processed using a sliding window analysis method. Each window contains 100 frames of image data, and the sliding step size between adjacent windows is 50 frames. The mean and standard deviation of the signal intensity are calculated for the signal data in each window. The feature parameters extracted from each window are combined into a multi-dimensional window feature vector sequence in chronological order. The purpose of this step is to obtain high-quality raw detection data and extract statistical features that reflect the temporal variation of the signal, providing a data foundation for subsequent signal processing and quality assessment.

[0077] The specific implementation of step S06 involves dividing the total particle size range of 30nm to 200nm into three non-overlapping particle size intervals based on the vesicle particle size distribution data measured by dynamic light scattering technology. Vesicles with a particle size range of 30nm to 60nm are classified as small-sized vesicles, those with a particle size range of 60nm to 120nm are classified as medium-sized vesicles, and those with a particle size range of 120nm to 200nm are classified as large-sized vesicles. For each particle size subgroup, a series of dilution detection experiments are performed using standard vesicle samples with known antigen concentrations to establish a quantitative calibration curve between the original signal intensity and antigen expression level of each particle size subgroup. The calibration coefficient corresponding to each particle size interval is extracted from the calibration curve. The original signal intensity value of each particle size subgroup in the test sample is divided by the corresponding calibration coefficient to obtain the normalized signal intensity. The purpose of this step is to eliminate the systematic influence of vesicle particle size differences on signal intensity, so that the detection signals of vesicles of different sizes are comparable and quantitatively accurate.

[0078] The specific implementation of step S07 involves applying the Fast Fourier Transform algorithm to the signal intensity time series data in the window feature vector sequence obtained in step S05, converting the time-domain signal into a frequency-domain representation, and obtaining the energy distribution spectrum of the signal at different frequency components, i.e., the frequency-domain feature spectrum. The energy proportion of each frequency band in the frequency-domain feature spectrum is statistically analyzed. When the energy of low-frequency components (frequency range 0.1Hz to 1Hz) accounts for more than 60% of the total energy, it is determined that baseline drift interference caused by temperature drift or light source aging exists. The cutoff frequency of the high-pass filter is increased from its initial value to 1.5Hz to enhance the attenuation effect on low-frequency components. When the energy of mid-frequency components (frequency range 1Hz to 10Hz) exceeds 60%, the cutoff frequency of the high-pass filter is increased. When the energy percentage exceeds 50%, environmental vibration interference caused by mechanical vibration transmission is identified. The center frequency of the interference is identified by the spectrum peak detection algorithm, and the center frequency of the band-stop filter is adjusted to the peak frequency of the interference to achieve narrowband suppression. When the energy percentage of high-frequency components in the frequency range of 10Hz to 50Hz exceeds 40%, electromagnetic interference caused by electrical equipment or wireless signal coupling is identified. The attenuation slope of the low-pass filter is increased from the initial design value to 80dB per ten octave to enhance the suppression capability of high-frequency noise. The purpose of this step is to identify different types of interference sources through frequency domain analysis and adaptively adjust the filter parameters to achieve targeted noise suppression and signal purification.

[0079] The specific implementation of step S08 involves constructing a two-layer game-theoretic optimization framework consisting of an upper-layer model and a lower-layer model. The upper-layer model aims to maximize detection sensitivity. Its objective function's input parameters include normalized signal strength, filter gain parameters, and vesicle capture efficiency. The output parameter is the optimized detection sensitivity. The upper-layer model finds the optimal parameter combination that maximizes detection sensitivity by adjusting the filter gain parameters within the range of 0.5 to 2.0. The lower-layer model aims to optimize signal stability. Its objective function's input parameters include the standard deviation of the window feature vector sequence, the rate of change of the Zeta potential, and the background signal strength. The output parameter is the lower-level optimized signal stability. The constraints of the lower-level model require that the standard deviation of the window feature vector sequence be less than 20% of the normalized signal intensity and the absolute value of the Zeta potential change rate be less than 5mV / min. The two-layer models are correlated with each other through the signal quality factor as a coupling term. The signal quality factor is defined as the geometric mean of the upper-level optimized detection sensitivity and the lower-level optimized signal stability. By iteratively solving the bi-layer optimization problem, a game equilibrium solution that makes both objective functions reach the optimal or near-optimal simultaneously is found. The purpose of this step is to maximize the detection sensitivity while ensuring signal stability, and to achieve global optimization of detection performance.

[0080] The specific implementation of step S09 involves comparing the normalized signal intensity after filtering and parameter optimization with a preset positive judgment threshold. The positive judgment threshold for LAM antigen is set as follows: The positive threshold for LprG antigen is set as follows: For the same sample, three consecutive independent tests are performed. When the LAM signal intensity is greater than or equal to the specified value in all three consecutive tests, the result is considered satisfactory. When the LprG signal intensity is greater than or equal to 1 in three consecutive tests, the sample is considered to be in a persistent LAM positive state and marked as a strong positive result. When a sample is determined to be in a persistently positive LprG state, it is marked as a strong positive result. For samples whose signal intensity exceeds the threshold only in a single test but does not meet the continuous positive standard, the encrypted detection process is immediately initiated. The sample is scanned five times rapidly within a 30-second time window. By increasing the detection frequency, more statistical data is accumulated to distinguish between true positive signals and random fluctuations. The purpose of this step is to improve the reliability of positive determination through consistency verification of multiple repeated tests and reduce the false positive rate caused by single measurement errors.

[0081] The specific implementation of step S10 involves establishing standard signal feature vectors for negative and positive quality controls. The negative control feature vector reflects the background signal pattern when the target antigen is absent, while the positive control feature vector reflects the typical signal response pattern of antigens at known concentrations. Each standard feature vector contains four descriptive parameters: mean signal intensity, standard deviation, peak value, and signal distribution skewness. The cosine similarity between the window feature vector sequence of the test sample and the two standard feature vectors is calculated. The cosine similarity is obtained by dividing the inner product of the two vectors by the product of their respective moduli, and the value ranges from 0 to 1. When the sample feature vector and the negative control... When the similarity between the sample feature vector and the positive control feature vector is greater than 0.95 and less than 0.3, the sample is judged as negative. When the similarity between the sample feature vector and the positive control feature vector is greater than 0.85 and less than 0.4, the sample is judged as positive. For ambiguous samples whose similarity does not meet the above judgment conditions, the intensity coefficient of background subtraction is increased, the signal intensity after background subtraction is recalculated, and the feature vector similarity is compared again. The purpose of this step is to improve the accuracy of positive and negative determination through signal pattern matching method and to perform refined analysis on boundary samples.

[0082] The specific implementation of step S11 involves determining the subpopulation weighting coefficients for different vesicle subpopulations based on statistical data from a large number of clinical samples. The small-sized vesicle subpopulation, due to its smaller surface area and relatively lower antigen load, has a weighting coefficient of 0.8. The medium-sized vesicle subpopulation, being the most widely distributed and signal-stable vesicle type in clinical samples, has a weighting coefficient set to a baseline value of 1.0. The large-sized vesicle subpopulation, due to its larger surface area and higher total antigen load, has a weighting coefficient of 1.2. The normalized signal of the small-sized vesicle subpopulation is then... The signal intensity is multiplied by 0.8, the normalized signal intensity of the medium-sized vesicle subpopulation is multiplied by 1.0, and the normalized signal intensity of the large-sized vesicle subpopulation is multiplied by 1.2. The sum of the three weighted signal intensity values ​​is then calculated to obtain the comprehensive detection signal intensity. The comprehensive detection signal intensity is compared with the positive judgment threshold. When the comprehensive detection signal intensity exceeds the threshold, the final positive judgment result is output. The purpose of this step is to integrate all the antigen information carried by vesicles of different sizes, balance the contribution of each size subpopulation through a weighted summation method, and improve the overall accuracy and robustness of the detection.

[0083] It should be noted that the key technical ideas of this invention include signal normalization processing based on particle size classification correction, adaptive frequency domain interference identification and filtering optimization, and a collaborative decision-making mechanism based on a two-layer game optimization model. The particle size classification correction technology establishes a particle size-dependent correction curve through the synchronous acquisition of dynamic light scattering and plasma resonance signals, eliminating the systematic bias of vesicle size differences in antigen quantitative detection. This makes the signals from vesicles of different sizes truly comparable. Compared to traditional methods that ignore the quantitative error caused by particle size influence, this technology significantly improves the accuracy and repeatability of antigen expression level determination. The adaptive frequency domain interference identification technology converts the time-domain signal to the frequency domain space through fast Fourier transform. Based on the energy proportion of different frequency bands, it automatically identifies three typical noise sources: baseline drift, environmental vibration, and electromagnetic interference. It then specifically adjusts the parameters of the high-pass filter, band-stop filter, and low-pass filter to achieve precise suppression. Compared to fixed-parameter filtering methods, this technology can adapt to complex and changing detection environments, effectively improving signal purification and detection stability. The two-layer game-theoretic optimization model solves the equilibrium through game theory, involving upper-layer sensitivity optimization and lower-layer stability optimization. It maximizes detection sensitivity while ensuring signal stability constraints, avoiding performance imbalances caused by single-objective optimization. Compared to traditional single-parameter adjustment strategies, this model achieves globally optimal detection performance. The synergistic effect of the three technical approaches lies in particle size correction providing accurate base signals, frequency domain filtering eliminating interference noise, and two-layer optimization achieving optimal parameter configuration. These three elements form a complete optimization chain from signal acquisition to processing to decision-making. Compared to the independent optimization of each stage in existing technologies, this synergistic mechanism achieves coordinated optimization of each stage through the coupling effect of signal quality factors, significantly improving the overall performance of the detection system and its reliability in clinical applications.

[0084] It should be noted that this invention also solves the following technical problem: existing nanoplasma detection systems suffer from signal quality degradation due to multi-band interference in complex biological sample environments. This invention extracts the window feature vector sequence through sliding window analysis of continuously acquired colorimetric signals, and obtains the frequency domain feature spectrum by performing a Fast Fourier Transform on the sequence. Based on the proportions of low-frequency, mid-frequency, and high-frequency components, baseline drift interference, environmental vibration interference, and electromagnetic interference are identified respectively. For different frequency band interferences, the high-pass filter cutoff frequency, band-stop filter center frequency, and low-pass filter attenuation slope are adaptively adjusted to achieve precise suppression of multi-band interference. Simultaneously, a two-layer game-theoretic optimization model is constructed. The upper-layer model adjusts the filter gain parameters with the goal of maximizing detection sensitivity, while the lower-layer model adjusts the background subtraction processing intensity with the goal of optimizing signal stability. The two models are coupled through a signal quality factor to form a synergistic optimization, effectively suppressing noise interference while ensuring detection sensitivity, thus solving the technical problem of signal quality degradation caused by multi-band interference.

[0085] Specifically, the principle of this invention is as follows: The solution to the signal deviation problem caused by vesicle size differences lies in the establishment of a particle size-dependent signal correction system. The plasmon resonance signal intensity of outer vesicles depends on the number of nanoprobes bound to the particle surface, and the number of bound nanoprobes is directly constrained by the vesicle membrane surface area. Even vesicles of different sizes with the same antigen surface density will produce different signal intensities due to differences in total surface area. Dynamic light scattering technology is used to acquire the particle size information of each vesicle, and the particle size data is synchronously matched with the colorimetric signal to form a correction data stream. Correction curves are established for each of the three particle size subgroups using standard vesicles with known antigen concentrations, determining the correction coefficients for each particle size range. Normalization converts the original signal intensity into a standardized value independent of vesicle size, reflecting the antigen density per unit surface area rather than the total signal amount. Subgroup weighting comprehensively considers the differences in capture efficiency and detection stability of vesicles of different sizes, ensuring that the final comprehensive signal intensity accurately reflects the true expression level of the antigen in the sample, thus meeting the logical requirements of quantitative detection.

[0086] 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.

[0087] In this embodiment, the specific implementation methods of steps S01-S04 and S09 are the same as those described above, and will not be repeated in detail here.

[0088] The specific implementation of step S05 involves starting a fully automated microscopic scanning imaging system to perform a full-field scanning of the substrate, one well at a time. The system uses a high numerical aperture objective lens and a high-sensitivity image sensor to continuously acquire the plasma resonance color development signal in each detection well. Simultaneously, the vesicle particle size data obtained by dynamic light scattering technology is time-stamped with the signal intensity data acquired by the imaging system to establish a synchronous time-series data stream containing particle size information and signal intensity information. The continuously acquired color development signal image sequence is processed using a sliding window analysis method. Each window contains 100 frames of image data, and the sliding step size between adjacent windows is 50 frames. The mean and standard deviation of the signal intensity are calculated for the signal data within each window. The formula for extracting the window feature vector sequence is expressed as follows:

[0089] ;

[0090] ;

[0091] In the formula, For the first The average signal strength of each window, in any unit; The number of frames contained in a single window, the default value is 100, dimensionless; For the first The first window The signal strength value of the frame, in any unit; The frame number within the window. Dimensionless; For the first The standard deviation of signal strength for each window, in any unit; For window number, Dimensionless; The total number of windows is dimensionless. The feature parameters extracted from each window are combined into a multi-dimensional window feature vector sequence in chronological order. The purpose of this step is to obtain high-quality raw detection data and extract statistical features that reflect the temporal variation of the signal, providing a data foundation for subsequent signal processing and quality assessment.

[0092] The specific implementation of step S06 involves dividing the total particle size range of 30nm to 200nm into three non-overlapping particle size intervals based on the vesicle particle size distribution data measured by dynamic light scattering technology. Vesicles with a particle size range of 30nm to 60nm are classified as small-sized vesicles, those with a particle size range of 60nm to 120nm as medium-sized vesicles, and those with a particle size range of 120nm to 200nm as large-sized vesicles. For each particle size subgroup, a series of dilution detection experiments are performed using standard vesicle samples with known antigen concentrations. A quantitative calibration curve is established between the original signal intensity and antigen expression level for each particle size subgroup. The calibration coefficients corresponding to each particle size interval are extracted from the calibration curves. The formula for calculating the normalized signal intensity is as follows:

[0093] ;

[0094] In the formula, For the first Normalized signal intensity of each particle size subgroup, in arbitrary units; For the first The original signal intensity of each particle size subgroup, in arbitrary units; For the first The correction coefficients corresponding to each particle size subgroup are dimensionless and obtained by fitting the test results of standard vesicle samples. The empirical values ​​range from 0.8 to 1.5. The particle size subgroup is numbered, dimensionless. Represents a small-sized vesicle subgroup. Representing the medium-sized vesicle subpopulation, This represents a subpopulation of large-sized vesicles. The purpose of this step is to eliminate the systematic influence of vesicle size differences on signal intensity, making the detection signals of vesicles of different sizes comparable and quantitatively accurate.

[0095] The specific implementation of step S07 involves applying the Fast Fourier Transform (FFT) algorithm to the signal intensity time series data in the window feature vector sequence obtained in step S05, converting the time-domain signal into a frequency-domain representation. The formula for the Fast Fourier Transform is as follows:

[0096] ;

[0097] In the formula, For frequency The complex amplitude in the frequency domain at a given point, in any unit; For the first The time-domain signal strength value at each moment, in any unit; The total length of the signal sequence undergoing Fast Fourier Transform is dimensionless. For the first Each frequency component is measured in Hz. Number the frequency components; they are dimensionless. ; The imaginary unit satisfies ; The time-domain sampling point number is dimensionless. ; is a natural constant, approximately equal to 2.71828. The energy distribution spectrum of the signal at different frequency components, i.e., the frequency domain characteristic spectrum, is obtained. The formula for calculating the energy proportion of each frequency band is as follows:

[0098] ;

[0099] In the formula, The energy percentage of a specified frequency band, dimensionless; A set of frequency component numbers corresponding to a frequency band interval; For frequency The frequency domain amplitude modulus at that point, in any unit. The energy proportion of each frequency band in the frequency domain characteristic spectrum is statistically analyzed. When the energy proportion of low-frequency components in the frequency range of 0.1Hz to 1Hz exceeds 60% of the total energy, baseline drift interference caused by temperature drift or light source aging is identified. The cutoff frequency of the high-pass filter is increased from the initial value to 1.5Hz to enhance the attenuation of low-frequency components. When the energy proportion of mid-frequency components in the frequency range of 1Hz to 10Hz exceeds 50%, environmental vibration interference caused by mechanical vibration transmission is identified. The center frequency of the interference is identified by the spectrum peak detection algorithm, and the center frequency of the band-stop filter is adjusted to the peak frequency of the interference to achieve narrowband suppression. When the energy proportion of high-frequency components in the frequency range of 10Hz to 50Hz exceeds 40%, electromagnetic interference caused by electrical equipment or wireless signal coupling is identified. The attenuation slope of the low-pass filter is increased from the initial design value to 80dB per decade to enhance the suppression capability of high-frequency noise. The purpose of this step is to identify different types of interference sources through frequency domain analysis and adaptively adjust the filter parameters to achieve targeted noise suppression and signal purification.

[0100] The specific implementation of step S08 is to construct a two-layer game optimization framework consisting of an upper-layer model and a lower-layer model. The upper-layer model aims to maximize detection sensitivity, and its objective function is expressed as follows:

[0101] ;

[0102] In the formula, To optimize detection sensitivity at the upper level, dimensionless; This is a normalized signal strength, with arbitrary units. This is a standard reference value for normalized signal strength, with arbitrary units, typically set to a value of [value missing]. ; The filter gain parameter is dimensionless. This is the standard reference value for the filter gain parameter, dimensionless, and the default value is 1.0; The vesicle capture efficiency is dimensionless. This is a standard reference value for vesicle capture efficiency, dimensionless, with a default value of 0.75; The signal quality factor is dimensionless. The upper-level model searches for the optimal combination of parameters that maximizes the detection sensitivity by adjusting the filter gain parameters within the range of 0.5 to 2.0. The constraints of the upper-level model include... and The lower-level model aims to optimize signal stability, and its objective function is expressed as follows:

[0103] ;

[0104] In the formula, Dimensionless for optimizing signal stability at lower levels; represents the standard deviation of the window feature vector sequence, in any unit; The standard reference value is the standard deviation of the window feature vector sequence, in any unit, and the empirical value is... ; The value represents the rate of change of Zeta potential, expressed in mV / min. This is the standard reference value for the rate of change of Zeta potential, in mV / min, with a default value of 1.0. The background signal strength is expressed in any unit. This is a standard reference value for background signal strength, with any unit, the default being [missing value]. ; It is the natural logarithm function; It is an absolute value function. The constraints of the lower-level model require... and The unit is mV / min. The two-layer model is interconnected through the signal quality factor as a coupling term. The formula for calculating the signal quality factor is as follows:

[0105] ;

[0106] In the formula, Let be the signal quality factor, which is dimensionless. The goal is to find a game equilibrium solution that simultaneously optimizes or nearly optimizes both objective functions by iteratively solving a two-level optimization problem. This step aims to maximize detection sensitivity while ensuring signal stability, thereby achieving global optimization of detection performance.

[0107] The specific implementation of step S10 involves establishing standard signal feature vectors for negative and positive quality control samples. Each standard feature vector contains four descriptive parameters: mean signal intensity, standard deviation, peak value, and signal distribution skewness. The formula for calculating cosine similarity is as follows:

[0108] ;

[0109] In the formula, The cosine similarity between the sample feature vector and the standard feature vector is dimensionless. is the window feature vector of the sample to be tested, and is a column vector containing 4 components; The standard signal feature vector is a column vector containing four components, and can be a negative quality control feature vector. or positive control sample feature vector ; The dot product of two vectors is calculated using the following formula: ; The magnitude of the sample feature vector is calculated using the following formula: ; The magnitude of the standard eigenvector is calculated using the following formula: ; The first feature vector of the sample One component; The first eigenvector of the standard feature vector One component; The eigenvector components are numbered, and are dimensionless. Represents the mean signal strength. Represents standard deviation, Represents the peak value. This represents the signal distribution skewness. When the sample feature vector and the negative control feature vector... similarity Greater than 0.95 and similar to the eigenvector of positive control samples similarity When the value is less than 0.3, the sample is considered a negative result. similarity Greater than 0.85 and similar to the feature vector of negative control products similarity When the similarity is less than 0.4, the sample is judged as a positive result. For fuzzy samples whose similarity does not meet the above judgment conditions, the intensity coefficient of background subtraction is increased, the signal intensity after background subtraction is recalculated, and the feature vector similarity is compared again. The purpose of this step is to improve the accuracy of positive and negative determination through signal pattern matching method and to perform refined analysis on boundary samples.

[0110] The specific implementation of step S11 involves determining the subpopulation weight coefficients of vesicle subpopulations of different particle sizes based on statistical data from a large number of clinical samples. The calculation formula for the comprehensive detection signal intensity is expressed as follows:

[0111] ;

[0112] In the formula, To comprehensively detect signal strength, the unit can be any unit; For the first The weighting coefficients of each particle size subgroup are dimensionless. For the first Normalized signal intensity of each particle size subgroup, in arbitrary units; , which is the weighting coefficient for the small-diameter vesicle subpopulation, dimensionless; , which is the weighting coefficient of the medium-sized vesicle subpopulation, dimensionless; , which is the weighting coefficient of the large-diameter vesicle subpopulation, dimensionless; The particle size subgroup is assigned a dimensionless number, with a value of 1, 2, or 3. The overall detection signal intensity is compared with the positive judgment threshold. When the overall detection signal intensity exceeds the threshold, the final positive judgment result is output. The purpose of this step is to integrate all antigen information carried by vesicles of different sizes and balance the contribution of each particle size subgroup through a weighted summation method, thereby improving the overall accuracy and robustness of the detection.

[0113] It needs further explanation that the objective function of the upper-level model is fully expressed as follows: This function achieves dimensionless processing by dividing the normalized signal intensity, filter gain parameter, and square root of vesicle capture efficiency by their respective standard reference values. The product of these three dimensionless terms is then coupled with a signal quality factor to characterize the detection system's response capability to the target antigen. The objective function of the lower-level model is fully expressed as follows: This function, obtained by subtracting the dimensionless product of two interference terms from 1, then subtracting the natural logarithm absolute value of the ratio of background signal intensities, and finally coupling it with the signal quality factor, characterizes the temporal fluctuation of the detected signal and the noise suppression effect. The first interference term... The second interference term reflects the combined effect of signal fluctuations and probe stability. It reflects the degree to which the background signal deviates from the standard value. The formula for calculating the signal quality factor is: , which represents the geometric mean of the upper-layer optimized detection sensitivity and the lower-layer optimized signal stability.

[0114] To better understand and implement this invention, a specific application scenario, Example 2, is provided below: A batch of serum samples from suspected tuberculosis patients in a high-incidence area, totaling 36 samples, required rapid screening for extracellular vesicle LAM and LprG antigens. The technical team used the high-throughput nanoplasma-excited photoimmunoassay detection method described in this invention to systematically detect and analyze this batch of samples.

[0115] First, the technical team centrifuged 36 pre-acquired human peripheral venous serum samples and added 50 μL of the supernatant to the detection wells of the CD81 capture antibody substrate. The samples were then incubated at 37°C for 60 minutes. During incubation, the team used dynamic light scattering technology to monitor the vesicle size distribution in real time. The monitoring data showed that the extracellular vesicles detected in the 36 samples were mainly distributed in the range of 45 nm to 185 nm, with the peak size concentrated in the range of 82 nm to 95 nm.

[0116] The technical team prepared the detection probes according to the established steps. They conjugated LAM detection probes with zwitterionic polymer-modified gold nanoparticles and LprG detection probes with polyethylene glycol-modified silver nanoparticles and LprG detection antibodies. A stabilizing buffer system was formed by adding 0.5 g / L bovine serum albumin and 0.05% Tween-20 to the conjugated probe suspension. The team continuously monitored the surface potential changes of the probes using a zeta potential detector. The monitoring results showed that the zeta potential of the LAM detection probe decreased from [value missing] within 60 minutes. Stable fluctuations to The rate of change was 0.038 mV / min, and the Zeta potential of the LprG detection probe changed from [value missing] to [value missing] during the same period. Change to The rate of change was 0.028 mV / min, and both probes exhibited good dispersion stability.

[0117] 20 μL each of LAM and LprG detection probes were added to the detection wells of the incubated substrate, and the substrate was incubated with shaking for 60 minutes at room temperature to allow the detection probes to specifically bind to the LAM and LprG antigens on the vesicle surface and form immune complexes. After incubation, the substrate was washed three times with phosphate wash buffer containing Tween-20 to remove unbound probes. The team prepared a colorimetric enhancement solution by mixing 100 μL of cationic solution of chromogenic agent A with 100 μL of cationic chelating agent B, and then adding 500 μL of chromogenic agent C dissolved in 1 mL of purified water. 10 μL of the colorimetric enhancement solution was added to the detection wells of the washed substrate, and the substrate was incubated at room temperature in the dark for 15 minutes to induce enhanced plasmon resonance signals on the nanoparticle surface.

[0118] The technical team used a fully automated microscopic scanning imaging system to perform full-well scanning of the substrate, acquiring plasmon resonance color development signals within each detection well and establishing a particle size correction signal data stream. The continuously acquired color development signals were analyzed using a sliding window approach, with 100 frames per window, and the mean and standard deviation of the signal intensity within each window were extracted to construct a window feature vector sequence. In 36 samples, the team detected LAM signal intensity data in 27 samples and LprG signal intensity data in 24 samples, such as... Figure 2 As shown, there are significant differences in signal intensity among different samples, with some samples exhibiting higher signal response levels.

[0119] The technical team established particle size classification intervals based on vesicle particle size data obtained from dynamic light scattering technology, categorizing the particle size range into... The vesicle signals were classified into a small-sized vesicle subgroup, with the particle size range being... The vesicle signals were classified into the medium-sized vesicle subgroup, with the particle size range being... The vesicle signals were classified into the large-size vesicle subgroup. Correction curves were established for signal intensity versus antigen expression levels for each subgroup, and the normalized signal intensity was obtained by dividing the original signal intensity value of each subgroup by the correction coefficient of the corresponding particle size interval. After particle size correction, the distribution ratio of the detected external vesicles in the three particle size subgroups in the 36 samples is shown in Table 1.

[0120] Table 1. Distribution of extravesicular vesicle size subpopulations in 36 samples

[0121]

[0122] The technical team performed a Fast Fourier Transform on the window feature vector sequence to obtain the frequency domain feature spectrum. Analysis revealed that in the detection of 36 samples, 8 samples had frequency ranges within the frequency domain feature spectrum. The low-frequency components accounted for more than 60%, which was determined to be baseline drift interference, and the high-pass filter cutoff frequency was increased to 1.5Hz. Five samples had frequency ranges in their frequency domain characteristic spectra within... The mid-frequency component accounted for more than 50%, indicating environmental vibration interference, and the center frequency of the band-stop filter was adjusted to the peak interference frequency. Three samples had frequency ranges in their frequency domain characteristic spectra within... If the high-frequency component accounts for more than 40%, it is judged as electromagnetic interference, and the attenuation slope of the low-pass filter is increased to 80dB / decibels. Figure 3 As shown, the signal after adaptive filtering significantly improves the signal-to-noise ratio and stability.

[0123] The technical team constructed a two-layer game-theoretic optimization model to adaptively adjust the detection parameters. The upper-layer model adjusts the filter gain parameters and vesicle acquisition efficiency with the goal of maximizing detection sensitivity, while the lower-layer model adjusts the background subtraction processing intensity and the number of repeated detections with the goal of optimizing signal stability. During the optimization process, the technical team set the standard reference value for the normalized signal strength as follows: The standard reference value for the filter gain parameter was set to 1.0, and the standard reference value for the vesicle capture efficiency was set to 0.75. After iterative calculations using the two-layer game optimization model, the average detection sensitivity of the 36 samples increased from the initial 1.82 to 2.35, the average signal stability increased from the initial 0.73 to 0.89, and the average signal quality factor reached 1.45.

[0124] The technical team compared the filtered, normalized signal strength with a preset LAM positive threshold. and LprG positive threshold A comparison was made. In 11 out of 36 samples, the LAM signal intensity was greater than or equal to [a certain value] in three consecutive tests. These were determined to be persistently positive for LAM and marked as strongly positive results. Nine samples had LprG signal intensities greater than or equal to [value missing] in three consecutive tests. The results were determined to be persistently positive for LprG and marked as strongly positive. Four other samples had only a single test result exceeding the threshold; the technical team initiated an encrypted testing process and completed five repeated scans within 30 seconds, ultimately confirming two of these as true positive results.

[0125] The technical team established standard signal feature vectors for negative and positive quality controls, and calculated the cosine similarity between the window feature vector sequence of the test sample and the standard signal feature vector. Of the 36 samples, 19 samples had a similarity greater than 0.95 with the negative control feature vector and less than 0.3 with the positive control feature vector, and were therefore judged as negative. 13 samples had a similarity greater than 0.85 with the positive control feature vector and less than 0.4 with the negative control feature vector, and were judged as positive. The remaining 4 samples did not meet the similarity criteria, so the technical team added background subtraction and recalculated the signal intensity to complete the final judgment.

[0126] The technical team established subgroup weighting coefficients based on the positive detection rates of different particle size subgroups. The detection signal of the small-sized vesicle subgroup was multiplied by a weighting coefficient of 0.8, the detection signal of the medium-sized vesicle subgroup by a weighting coefficient of 1.0, and the detection signal of the large-sized vesicle subgroup by a weighting coefficient of 1.2. The summation of the weighted subgroup signal intensities yielded the comprehensive detection signal intensity. When the comprehensive detection signal intensity exceeded the positive threshold, the final positive result was output. The final test results showed that out of 36 samples, 13 were double-positive for LAM and LprG, 2 were single-positive for LAM, 1 was single-positive for LprG, and the remaining 20 were double-negative.

[0127] This invention represents a significant technological advancement over traditional tuberculosis detection methods. Traditional methods, such as sputum smear microscopy and culture, suffer from low sensitivity and long detection cycles. This invention, however, utilizes nanoplasma resonance technology to achieve highly sensitive detection of extracellular vesicle surface antigens, enabling diagnosis without the need for direct isolation and culture of pathogens. Traditional immunological detection methods typically treat vesicles of different sizes as a homogeneous population, leading to underestimation of small-sized vesicles due to low antigen loading and overestimation of large-sized vesicles due to large surface area. This invention employs dynamic light scattering technology to monitor vesicle size distribution in real time and establishes a size grading correction model, eliminating the systematic bias of particle size in quantitative detection and ensuring true comparability of antigen expression levels among vesicles of different sizes. Traditional detection systems lack real-time identification and adaptive processing capabilities for signal interference. This invention uses Fast Fourier Transform (FFT) for frequency domain analysis, accurately identifying the frequency characteristics of baseline drift interference, environmental vibration interference, and electromagnetic interference, and dynamically adjusting filter parameters for precise suppression of different types of interference. Traditional detection methods often compromise between optimizing detection sensitivity and signal stability. This invention constructs a two-layer game-theoretic optimization model, treating sensitivity and stability optimization as interrelated objective functions for collaborative solution. By using a signal quality factor as a coupling term, it achieves a game-theoretic equilibrium between the two optimization objectives, avoiding performance imbalances caused by single-objective optimization. Traditional detection interpretation relies on a simple comparison of a single measurement result with a fixed threshold, making it susceptible to false positives or false negatives due to occasional signal fluctuations. This invention requires three consecutive test results exceeding a threshold to confirm a strong positive result. It employs an encrypted detection process for high-frequency repeated scanning of boundary samples and uses cosine similarity analysis to match sample signal patterns with the feature vectors of standard quality control samples. This multi-layered judgment logic significantly improves the reliability and accuracy of the detection results.

[0128] It should be noted that the variables involved in this invention are explained in detail in Table 2.

[0129] Table 2 Variable Explanation Table

[0130]

[0131] 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 high-throughput nanoplasma-excited photoimmunoassay to detect extravesicular biomarkers, characterized in that, Includes the following steps: Serum samples were added to the detection wells of a CD81 capture antibody substrate and incubated to allow extracellular vesicles in the serum to bind to the CD81 capture antibody on the substrate surface. Dynamic light scattering technology was used to monitor the vesicle size distribution in real time during incubation. LAM detection probes were prepared by conjugating zwitterionic polymer-modified gold nanoparticles with LAM detection antibodies, and LprG detection probes were prepared by conjugating polyethylene glycol-modified silver nanoparticles with LprG detection antibodies. The LAM and LprG detection probes were then added to the detection wells of the incubated substrate, respectively, to allow the detection probes to bind to the vesicles. LAM antigen and LprG antigen on the vesicle surface specifically bind to form immune complexes; a colorimetric enhancement solution is added to the detection wells of the washed substrate to induce the enhancement of the plasmonic resonance signal on the nanoparticle surface; the plasmonic resonance colorimetric signal in each detection well is collected and a particle size correction signal data stream is established; a particle size classification interval is established based on the vesicle particle size data obtained by dynamic light scattering technology, and a correction curve of signal intensity versus antigen expression level is established for different particle size subgroups; the original signal intensity value of each subgroup is divided by the correction coefficient of the corresponding particle size interval to obtain the normalized signal intensity; Subgroup weighting coefficients are established based on the positive detection rates of different particle size subgroups. The comprehensive detection signal intensity is obtained by summing the signal intensities of each weighted subgroup. When the comprehensive detection signal intensity exceeds the positive threshold, the final positive judgment result is output. The serum sample is the supernatant after centrifugation of human peripheral venous blood samples from the spot area.

2. The method according to claim 1, characterized in that, The zwitterionic polymer modification refers to the chemical bonding of polymer molecular layers containing positive and negative charged groups on the surface of gold nanoparticles to form a protective layer structure with charge balance characteristics. The electrostatic interaction between the coexisting positive and negative charged groups enhances the dispersion stability of nanoparticles in high-salt and high-protein environments, preventing the particles from agglomerating and settling due to electrostatic attraction.

3. The method according to claim 2, characterized in that, The polyethylene glycol modification refers to grafting long-chain polyethylene glycol molecules onto the surface of silver nanoparticles. The steric hindrance effect of the polyethylene glycol chains forms a dense hydration protective layer on the particle surface, preventing protein molecules and other biomolecules in serum from approaching the particle surface and reducing signal interference and particle inactivation caused by non-specific adsorption.

4. The method according to claim 3, characterized in that, Bovine serum albumin and Tween-20 were added to the coupled detection probe suspension to form a stabilizing buffer system, and the changes in probe surface potential were continuously monitored using a Zeta potential detector.

5. The method according to claim 4, characterized in that, The stabilization buffer system refers to a multi-component protective solution system composed of bovine serum albumin and nonionic surfactant Tween-20 added to the nanoparticle detection probe suspension. Bovine serum albumin reduces the false adsorption of serum proteins by occupying non-specific binding sites on the surface of nanoparticles, while Tween-20 maintains the monodisperse state of nanoparticles by reducing the surface tension of the liquid and forming a micelle structure.

6. The method according to claim 5, characterized in that, The particle size correction signal data stream refers to a time-series data sequence formed by synchronously matching the outer vesicle particle size information obtained in real time by dynamic light scattering technology with the plasma resonance colorimetric signal collected by the microscopic scanning imaging system. Each data point in the data stream contains two dimensions of information: the vesicle particle size value and the signal intensity value at the corresponding time.

7. The method according to claim 6, characterized in that, The continuously acquired colorimetric signals are analyzed in 100-frame units using a sliding window approach. The mean and standard deviation of the signal intensity within each window are extracted to form a sequence of window feature vectors.

8. The method according to claim 7, characterized in that, The window feature vector sequence is a multidimensional vector set formed by arranging statistical feature parameters extracted from multiple consecutive sliding windows in chronological order. Each feature vector contains four descriptive parameters: mean, standard deviation, peak value, and signal distribution skewness of the signal intensity within the corresponding window.

9. The method according to claim 8, characterized in that, The particle size classification intervals are based on the biophysical characteristics and light scattering behavior of external vesicles, dividing the total particle size range of 30 to 200 nm into three discontinuous intervals. Vesicles with a particle size range between 30 nm and 60 nm, including 30 nm but excluding 60 nm, are classified as small-sized vesicles; vesicles with a particle size range between 60 nm and 120 nm, including 60 nm but excluding 120 nm, are classified as medium-sized vesicles; and vesicles with a particle size range between 120 nm and 200 nm, including both 120 nm and 200 nm, are classified as large-sized vesicles.

10. The method according to claim 9, characterized in that, A two-layer game optimization model is constructed to adaptively adjust the detection parameters. The two-layer game optimization model includes an upper-layer model that aims to maximize detection sensitivity and a lower-layer model that aims to optimize signal stability. The objective function input of the upper-layer model includes normalized signal strength, filter gain parameters, and vesicle capture efficiency. The objective function input of the lower-layer model includes the standard deviation of the window feature vector sequence, the rate of change of Zeta potential, and the background signal strength. The objective functions of the upper-layer model and the lower-layer model are related through a coupling term signal quality factor.

Images