Full-automatic non-treponema pallidum serological experiment detection system

Abstract

The invention discloses a full-automatic non-treponema pallidum serological experiment detection system, and relates to the technical field of image processing, the system is composed of a plurality of functional modules, and the system comprises: a data acquisition module for synchronously acquiring image data and thermal imaging data by controlling a camera and a thermal imager through a PLC when a PVC hard sheet enters a detection area; the three-dimensional graph synthesis module is used for generating a three-dimensional model according to image data and superposition contour recognition; based on the thermal imaging data, thermal imaging data position information is obtained through affine transformation, space registration with the three-dimensional model is carried out, and a temperature gradient field is generated; the defect detection and identification module is used for equally dividing the surface of the three-dimensional model, numbering equally divided areas, analyzing connected domains based on morphological operation, overlapping multi-scale feature fusion and identifying defects; and according to the temperature gradient field after region segmentation, temperature uniformity evaluation is carried out, and the temperature standard deviation between the regions is calculated.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, specifically to a fully automated non-Treponema pallidum serological laboratory detection system. Background Technology

[0002] Syphilis is a Class B notifiable infectious disease in China, primarily transmitted through sexual contact, mother-to-child transmission, and blood. If left untreated, it can damage multiple organs, including the skin, mucous membranes, cardiovascular system, and nervous system, causing irreversible damage. Early symptoms of syphilis are often subtle, such as the chancre, which is easily overlooked. Non-treponemal antibody testing is highly sensitive and can be the preferred method for large-scale population screening, such as prenatal checkups, blood donors, and high-risk groups. Combined with specific antibody testing, such as TPPA, it enables early diagnosis of syphilis. This allows for early detection and treatment, preventing the disease from progressing to later stages, such as neurosyphilis and cardiovascular syphilis, which can cause organ damage. Furthermore, monitoring treatment effectiveness ensures complete treatment and reduces the risk of recurrence. For pregnant women, it directly protects maternal and infant health and prevents congenital syphilis.

[0003] Traditional detection methods, such as the rapid plasma reagin ring card test and the toluidine red unheated serum test, suffer from drawbacks such as low sensitivity, weak anti-interference ability, susceptibility to hemolysis, lipemia, and sample turbidity, insufficient automation, numerous manual operation steps, and low efficiency. Furthermore, they lack quality control collaboration mechanisms between regional laboratories, making it difficult to meet the needs of precise clinical testing and large-scale screening. While chemiluminescence immunoassay has demonstrated high sensitivity in other detection fields, its specificity optimization for non-treponemal antibody detection is insufficient, and it has not effectively addressed sample interference issues such as hemolysis or lipemia. Its interference correction largely relies on simple linear algorithms, which cannot cope with complex nonlinear interference scenarios. Summary of the Invention

[0004] (a) Technical problems to be solved To address the shortcomings of existing technologies, this invention provides a fully automated non-Treponema pallidum serological laboratory detection system, which solves the problems mentioned in the background art.

[0005] (II) Technical Solution To achieve the above objectives, the present invention provides the following technical solution: The sample and magnetic bead processing module mixes samples with magnetic beads and uses dynamic magnetic flux control to process the magnetic beads and construct a reaction system. The luminescence signal acquisition module acquires the chemiluminescence photon signal in the reaction system, combines it with a wavelet threshold algorithm to suppress noise, and converts the noise-suppressed photon signal into a quantifiable original signal. The AI ​​multimodal interference correction module pre-acquires multi-dimensional interference features, including hemoglobin absorbance and lipid scattering characteristics. It removes multi-dimensional interference features from the original signal through a deep residual network, calculates the antibody concentration based on the original signal after interference removal, and corrects it through kinetic fit. The adaptive carrier system iteration module interacts with the LIS system in real time with the corrected antibody concentration through a dual-link network, performs self-calibration using a regional quality control collaborative algorithm, and outputs a standardized test report.

[0006] Furthermore, the process of mixing the sample with the magnetic beads is as follows: The sample and magnetic beads are mixed at a volume ratio of 10:1 to form a sample flow and a magnetic bead flow. The sample flow and magnetic bead flow are introduced into a microfluidic bifurcation channel. A micropillar array is set in the microfluidic bifurcation channel to make the sample flow and magnetic bead flow form a laminar flow state, so that they can collide fully.

[0007] Furthermore, the process of constructing the reaction system is as follows: The magnetic flux is adjusted and controlled by the dynamic current of a variable air gap electromagnet to capture, wash and separate magnetic beads and release them. The magnetic bead capture method uses high-intensity magnetic flux to rapidly deposit and enrich magnetic beads that bind antigen-antibody complexes to non-Treponema pallidum in the direction of the magnetic field. The washing and separation process switches to a medium-intensity magnetic flux while simultaneously introducing pulsed washing liquid. The release of magnetic beads is regulated by using a low-intensity magnetic flux to ensure that the magnetic beads, antigen and antibody complex are uniformly dispersed in the reaction system.

[0008] Furthermore, the process of collecting chemiluminescent photon signals in the reaction system is as follows: A single-photon avalanche diode array is used to capture individual chemiluminescent photons. The arrival time of each photon is recorded using time-correlated single-photon counting to form a photon signal with the arrival time of the photon. A wavelet thresholding algorithm is then used to suppress noise.

[0009] Furthermore, the process of converting it into a quantifiable original signal is as follows: A calibration curve was established in advance using non-treponemal antibody standards of known concentrations. A four-parameter logistic model was used, and the noise-suppressed photon signal was substituted into the model to calculate the corresponding original signal.

[0010] Furthermore, the process of removing multi-dimensional interference features using deep residual networks is as follows: S301: Collect multi-dimensional feature data as input signals for deep residual networks. Features include hemoglobin absorbance and lipid scattering characteristics, as well as photon signals after noise suppression. S302: The deep residual network is composed of multiple residual blocks connected in series. Each residual block contains core components such as convolutional layers and activation layers. It splits the input signal into two parts: the real effective signal and the interference residual signal, and captures the interference residual signal. S303: The deep residual network adopts a composite learning objective to correct the interfering residual signal to obtain the corrected signal. By measuring the deviation between the corrected signal and the true effective signal, the correction result is made to converge to the photon signal corresponding to the actual antibody concentration of the sample; at the same time, the deep residual network removes the interfering components in the interfering residual signal. S304: During the deep residual network training phase, a large-scale dataset containing samples with different levels of interference is used for model training. The dataset includes samples with different hemoglobin concentrations corresponding to different degrees of hemolysis and different degrees of lipemia corresponding to different degrees of scattering polarization. Each sample is paired with a non-treponemal antibody standard of known concentration to learn the signal patterns under different interference intensities and interference modes.

[0011] Furthermore, the process of calculating antibody concentration based on the original signal after removing interference is as follows: When the trained deep residual network processes a new sample, it receives preprocessed multi-dimensional feature data and calculates the corresponding interference residual signal for the sample. This residual signal includes spectral interference from hemoglobin absorbance and scattering interference from lipid scattering characteristics. The original photon signal after noise suppression is then processed to remove the interference residual signal output by the deep residual network, and the antibody concentration is output.

[0012] Furthermore, the process of correction using the dynamic fit degree is as follows: A standard kinetic model based on the secondary reaction of non-treponemal antigen and antibody was established and the basic parameters were calibrated. Photon signals of the chemiluminescent reaction were collected, fitted with the standard kinetic model, and the goodness-of-fit threshold was determined. When the goodness-of-fit did not meet the standard, the deviation was analyzed and the parameters were adjusted for a second fitting. After the goodness-of-fit was met, the validity of the corrected photon signal was confirmed and the correction was completed.

[0013] Furthermore, the process of real-time interaction with the LIS system via dual-link networking is as follows: It adopts a dual-link redundancy architecture of 5G and wired connections with intelligent switching to upload the calibrated antibody concentration and equipment operating status; through the feedback of regional quality control statistics and clinical correlation information from the LIS system, it sends back as self-calibration input to achieve real-time interactive closed loop of detection data.

[0014] Furthermore, the process of generating a standardized test report is as follows: The system receives regional quality control statistics and local multi-module data to form a data pool. Through horizontal comparison, vertical traceability and multimodal fusion calculation of regional quality control collaborative algorithms, calibration coefficients are calculated and parameters are adjusted. Finally, a standardized test report with a three-layer structure of results, quality control and traceability is output and the whole process is traceable.

[0015] (III) Beneficial Effects This invention provides a fully automated serological detection system for non-Treponema pallidum, which has the following advantages: (1) This invention combines chemiluminescent immunoassay with AI multimodal interference correction technology to improve detection sensitivity, which is superior to traditional RPR and TRUST methods. It can accurately capture early syphilis infection, such as low-concentration antibodies before the appearance of the chancre in primary syphilis, so as to achieve early detection and early treatment. At the same time, the AI ​​deep residual network can automatically identify and remove sample interference such as hemoglobin absorbance interference in hemolysis and scattering characteristics interference in lipemia. Even in hemolyzed and lipemic samples, it can output stable and reliable detection results, which solves the problem of false negatives and false positives caused by sample quality such as lipemia and hemolysis in traditional methods. Especially in prenatal check-ups for pregnant women and screening of high-risk groups, it can effectively avoid missed diagnosis and misdiagnosis, reduce the incidence of congenital syphilis, and provide early intervention basis for the prevention of late complications such as neurosyphilis and cardiovascular syphilis, thereby reducing the irreversible damage of syphilis to individual health from the source.

[0016] (2) This invention adopts a fully automated process of magnetic bead processing, signal acquisition, AI correction and data interaction in a closed loop, which improves detection efficiency and can meet the high-efficiency needs of large-scale population screening. Through dual-link networking and real-time interaction with the LIS system, combined with regional quality control collaborative algorithms, it automatically compares the detection data of similar samples in the region to achieve self-calibration, promotes the homogenization of regional test results, and provides standardized and traceable syphilis epidemic data for public health departments, helping to accurately formulate prevention and control strategies. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the system of the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0019] Please see Figure 1 This embodiment provides a fully automated non-treponemal serological testing system, which includes: The sample and magnetic bead processing module mixes samples with magnetic beads and uses dynamic magnetic flux control to process the magnetic beads and construct a reaction system. The process of mixing the sample with the magnetic beads is as follows: To address the specific requirements for detecting non-Treponema pallidum antibodies, a chemiluminescent immunoassay was developed, innovating from two dimensions: antigen labeling and reaction kinetics. Acridinium ester was selected as the luminescent label, which, after binding to non-Treponema pallidum antigens, undergoes an oxidation reaction with hydrogen peroxide under alkaline conditions, releasing high-energy photons. This reaction features high quantum yield, ensuring high detection sensitivity.

[0020] In the antigen immobilization stage, when immobilizing non-Treponema pallidum antigens on the surface of superparamagnetic beads, gold nanoparticles are introduced as spacer arms to reduce non-specific binding by increasing the steric hindrance of the antigen. The sample and magnetic bead processing module introduces the sample flow and magnetic bead flow into a microfluidic bifurcation channel at a volume ratio of 10:1. A micropillar array is placed within the channel, creating a laminar flow while increasing turbulence to promote full collision between the sample and the magnetic beads. By precisely controlling the modification density of the antigen on the magnetic bead surface, the spatial conformation of the antigen is made closer to its native state, thereby significantly improving the specificity of binding with the target antibody. Experimental verification shows that when the modification density is controlled within a specific range, the antigen-antibody binding specificity is optimal, effectively reducing detection errors caused by cross-reactions. During this process, based on a dynamic magnetic flux regulation strategy, the magnetic beads are in a medium-intensity magnetic flux environment during the mixing stage, ensuring uniform dispersion of the magnetic beads and improving mixing efficiency.

[0021] The process of constructing the reaction system is as follows: Magnetic bead processing is achieved through dynamic magnetic flux regulation, including bead capture, washing and separation, and bead release. Precise control of magnetic flux is realized through dynamic current adjustment of a variable air gap electromagnet.

[0022] High-intensity magnetic flux is used in the magnetic bead capture stage, and the coil current is adjusted by a PID controller to ensure that the magnetic field gradient is greater than 10. 5 T / m causes magnetic beads containing non-treponemal antigen and antibody complexes to rapidly settle and accumulate in the direction of the magnetic field. By using a magnetic bead sedimentation kinetic model, combined with parameters such as magnetic bead radius, magnetization intensity and liquid viscosity, it is ensured that the magnetic beads are captured in a short time to form aggregates of magnetic beads and complexes to be washed.

[0023] During the washing and separation stage, a medium-intensity magnetic flux is used, and a pulsed washing liquid is introduced with a sinusoidal flow rate to generate periodic shear force. This ensures that the magnetic flux intensity can maintain the stability of the magnetic bead aggregates while effectively removing non-specific binders on the surface of the magnetic beads through fluid shear force. After three washing cycles, the residual rate of non-specific binders is <0.1%. During the magnetic bead release phase, the magnetic flux is adjusted to a low intensity, so that the magnetic beads, antigen and antibody complex are uniformly dispersed in the reaction system. At this time, the magnetic beads are in a free suspension state, providing a uniform reaction environment for the addition of luminescent substrate and signal detection in the subsequent chemiluminescence reaction, and finally constructing a chemiluminescence reaction system that can be used for photon signal acquisition.

[0024] The luminescence signal acquisition module acquires the chemiluminescence photon signal in the reaction system, combines it with a wavelet threshold algorithm to suppress noise, and converts the noise-suppressed photon signal into a quantifiable original signal. The specific process for collecting chemiluminescent photon signals in the reaction system is as follows: Using a single-photon avalanche diode (SPAD) array as a photon detector, it has high sensitivity and can accurately capture a single chemiluminescent photon. In the reaction system, after the acrid ester and hydrogen peroxide are oxidized and release photons under alkaline conditions during the chemiluminescent reaction, the photons are dispersed in all directions. The SPAD array receives the photons released in the reaction system from all directions through a large area of ​​photosensitive region.

[0025] Utilizing time-correlated single-photon counting technology, with a sampling frequency of ≥100MHz and extremely high time resolution, such as... The system accurately records the arrival time of each photon, forming a sequence of photon arrival times. Based on the characteristic that the number of chemiluminescent photons follows a Poisson distribution, the system achieves quantitative acquisition of photon signals by statistically analyzing the arrival times of a large number of photons.

[0026] The acquired photon counting signal is decomposed into wavelet coefficients at different scales. A soft threshold function is used to process the wavelet coefficients at different scales: wavelet coefficients smaller than the set threshold are set to zero, and wavelet coefficients larger than the threshold are shrunk and adjusted. After processing, the wavelet coefficients are reconstructed to effectively suppress background noise and obtain a quantifiable original photon signal with a signal-to-noise ratio improved by more than 20dB.

[0027] The process of converting the signal into a quantifiable raw signal is as follows: A calibration curve was established in advance using non-treponemal antibody standards of known concentrations. A four-parameter logistic model was used, and the noise-suppressed photon signal was substituted into the model to calculate the corresponding original signal.

[0028] The AI ​​multimodal interference correction module pre-acquires multi-dimensional interference features, including hemoglobin absorbance and lipid scattering characteristics, and combines them with the original signal to remove multi-dimensional interference features through a deep residual network. The antibody concentration is calculated based on the original signal after interference removal and corrected by kinetic fit. The process of pre-acquiring multi-dimensional interference features is as follows: The absorbance of hemoglobin is obtained through a built-in spectrophotometric detection module. Before or simultaneously with the sample entering the reaction system, a specific wavelength in the visible light region is detected. This wavelength is the characteristic absorption peak of hemoglobin, which can accurately capture the absorption characteristics of hemoglobin. Based on the absorption law of substances to light, the absorbance of hemoglobin at this characteristic wavelength will change regularly with its own concentration. The higher the concentration, the greater the absorbance value, which indirectly reflects the degree of hemolysis of the sample. Finally, characteristic data of hemoglobin absorbance that can characterize the degree of hemolysis are obtained.

[0029] The acquisition of lipemia scattering characteristics utilizes a polarized light scattering detection module and a turbidity detection unit. Through the collaborative work of multiple modules, multi-dimensional scattering characteristics related to lipemia are collected simultaneously. In the scattered light intensity and polarization degree detection stage, a specific laser source is used, whose wavelength will not interfere with the subsequent chemiluminescence signal. After the sample is irradiated by the laser, a detector is set at a specific scattering angle position to collect two types of scattered light intensities: parallel to the incident light polarization direction and perpendicular to the incident light polarization direction. By calculating the ratio of the difference between these two light intensities to their sum, the polarization degree data is obtained. This polarization degree can directly reflect the scattering characteristics of lipoprotein particles in lipemia. In the lipemia degree quantification index detection stage, the triglyceride concentration in the sample can be quickly detected by turbidimetry or enzymatic methods. Alternatively, the turbidity of the sample can be measured by a turbidimeter. Based on the light transmission attenuation effect, the higher the sample turbidity, the higher the concentration of lipemia particles. This turbidity serves as a quantitative reference for lipemia degree and, together with the polarization degree data, constitutes the lipemia scattering characteristics.

[0030] The process of removing interference from hemoglobin absorbance and lipid scattering characteristics in the original signal is as follows: S301: The pre-collected multi-dimensional feature data is used as the input signal of the deep residual network. The features include hemoglobin absorbance and lipid scattering characteristics, as well as photon signals after noise suppression. All input features are preprocessed to eliminate the influence of dimensional differences between different features by unifying the feature scale, so that the features of each dimension remain consistent within the numerical range.

[0031] S302: The deep residual network consists of multiple layers of residual blocks connected in series. Each residual block contains core components such as convolutional layers and activation layers, splitting the input signal into two parts: the real effective signal and the interference residual signal, capturing the interference residual signal. During network operation, each residual block receives the output features of the previous layer and extracts higher-level interference feature representations through convolution operations and activation functions. The unique structure of the residual block allows the signal to be directly passed through the shortcut path, allowing the network to focus on learning the nonlinear relationship between the interference residual and the input features, rather than learning the complete signal from scratch. This avoids the gradient vanishing problem in deep network training and allows the network to explore the complex correlation between hemoglobin absorbance, lipid scattering characteristics and interference signals more deeply, ensuring the comprehensive capture of interference features.

[0032] For example, to address the spectral interference caused by hemoglobin absorbance, the network learns the correspondence between the absorbance and background interference in the photon signal; to address the scattering characteristics of lipemia, the network explores the intrinsic relationship between scattering polarization, triglyceride concentration or turbidity and scattering interference signals, thereby achieving separate identification and joint learning of the two types of interference.

[0033] S303: The deep residual network employs a specially designed composite learning objective that balances the accuracy of interference removal and the thoroughness of interference suppression. This objective measures the deviation between the corrected signal and the true effective signal, ensuring the corrected result is as close as possible to the signal corresponding to the actual antibody concentration of the sample. Simultaneously, it emphasizes the effectiveness of interference suppression, driving the deep residual network to remove interference components to the maximum extent possible. Through this dual-objective constraint, the network does not simply pursue minimizing signal deviation during training. Instead, under the premise of controllable deviation, it prioritizes the complete removal of interference components, avoiding overfitting to the effective signal while leaving residual interference or over-removing interference while losing the effective signal. This achieves dual optimization of correction accuracy and interference suppression rate.

[0034] S304: During the deep residual network training phase, a large-scale dataset containing samples with varying degrees of interference was used for model training. The dataset included samples with different hemoglobin concentrations corresponding to different degrees of hemolysis, different triglyceride concentrations, and different degrees of scattering polarization corresponding to different degrees of lipemia. Each sample was paired with a known concentration of non-Treponema pallidum antibody standard to ensure the network could learn the signal patterns under different interference intensities and modes. During training, the weight parameters and bias terms of each layer of the network were continuously adjusted using the backpropagation algorithm. Based on the error value calculated according to the composite learning objective, the learning effect of the network was gradually optimized. After multiple rounds of iterative training, the network developed a stable interference identification and removal capability. It could accurately extract the corresponding interference residual signals for different combinations and degrees of hemolysis and lipemia interference, ensuring the model had good generalization ability.

[0035] The process for calculating antibody concentration is as follows: When a well-trained deep residual network processes a new test sample, it receives pre-processed multi-dimensional feature data and calculates the corresponding interference residual signal for the sample using optimized network parameters and structure. This residual signal accurately includes spectral interference caused by hemoglobin absorbance and scattering interference caused by lipemia scattering characteristics. The interference residual signal output by the network is subtracted from the noise-suppressed original photon signal to completely remove the interference components. The output signal is the effective signal after removing hemolysis and lipemia interference, which can truly reflect the concentration level of non-treponemal antibodies in the sample.

[0036] The process of correction based on the dynamic fit is as follows: By incorporating a kinetic model of antigen-antibody binding reaction, based on the secondary reaction characteristics of non-treponemal antigens and antibodies, the influence of key factors such as reaction temperature, reaction time, and reagent concentration on the binding reaction process was considered. Through numerous standard experiments, basic parameters were calibrated to determine the ideal reaction kinetic curve characteristics under different detection conditions, such as a standard reaction temperature of 37℃ and a reaction equilibrium time of 10 minutes. Key indicators, including the reaction initiation rate, the time node to reach equilibrium, and the signal stability during the equilibrium stage, were established to form a standard kinetic template library.

[0037] During the chemiluminescence reaction, the luminescence signal of the reaction system is continuously and frequently acquired through the photon detection module, forming a real-time signal sequence that changes over time. The acquisition frequency is matched with the reaction process, and the acquisition density is appropriately increased in the early stage of the reaction and the stage of rapid signal change to ensure that the signal change characteristics of each stage, such as the reaction start-up, rapid rise and tendency to equilibrium, can be fully captured.

[0038] The acquired real-time signal sequence is input into the system's data processing unit, where it is matched with the corresponding model in the standard kinetic template library using a preset fitting algorithm. During the fitting process, the algorithm automatically adjusts key parameters of the model, such as the reaction rate coefficient and equilibrium constant, to optimize the overlap between the standard model curve and the real-time signal sequence. The core evaluation index of the fit is the degree of agreement between the real-time signal sequence and the fitted model curve. A higher degree of agreement indicates a stronger consistency between the actual reaction process and the theoretical kinetic laws, and a higher reliability of the signal data. Conversely, a lower degree of agreement indicates that there may be abnormal deviations in the actual reaction, requiring further correction.

[0039] The system presets a goodness-of-fit threshold, which is determined through validation with a large number of clinical samples to ensure effective screening of normal and abnormal reactions. The real-time calculated goodness-of-fit is compared with this threshold. If the goodness-of-fit reaches or exceeds the threshold, it indicates that the actual reaction process conforms to the theoretical kinetics of antigen-antibody binding, the reaction system is stable, and there are no abnormalities in temperature, reagent activity, or sample status. The real-time signal sequence accurately reflects the antibody concentration in the sample, requiring no additional correction, and the system proceeds directly to the subsequent quantitative calculation stage. If the goodness-of-fit is below the threshold, the system automatically initiates a deviation analysis program. This program combines synchronously monitored auxiliary parameters during the reaction process, such as real-time temperature of the reaction chamber, reagent addition, and sample processing time, to determine the source of the deviation. Common causes of deviation include fluctuations in reaction temperature, insufficient reaction time, slight attenuation of reagent activity, and the presence of trace amounts of unknown interfering substances in the sample. The system will accurately locate the core factors causing the poor goodness-of-fit based on a preset deviation type library.

[0040] Based on the deviation analysis results, the system automatically implements targeted parameter adjustments. If the deviation originates from temperature fluctuations, the signal intensity conversion coefficient is adjusted according to the temperature deviation amplitude to offset the effect of temperature on the reaction rate. If the deviation originates from insufficient reaction time, the reaction monitoring duration is appropriately extended to collect more signal data approaching equilibrium. If the deviation originates from changes in reagent activity, the correction parameters corresponding to the reagent batch are called to adjust the reaction rate correlation coefficient in the model. After the parameter adjustments are completed, the corrected real-time signal sequence is re-fitted with the standard kinetic model to determine whether the fit meets the acceptable threshold. If it still does not meet the threshold, the system will iteratively adjust the parameters based on the new deviation situation until the fit meets the requirements, ensuring that the final signal sequence used for quantitative calculation fully conforms to the kinetic laws of antigen-antibody binding.

[0041] After the fit meets the standard, the system verifies the validity of the corrected signal sequence by checking the stability of the signal during the equilibrium phase, ensuring that there are no abnormal fluctuations or abrupt changes, and ensuring that no new deviations are introduced during the correction process. After the verification is passed, the signal sequence is output as valid data for subsequent quantitative calculation of antibody concentration, ultimately ensuring that the test results can truly reflect the actual level of non-treponemal antibodies in the sample, and improving the accuracy and reliability of the test results.

[0042] The adaptive vector iteration module interacts with the LIS system in real time with the corrected antibody concentration through a dual-link network, performs self-calibration using a regional quality control collaborative algorithm, and outputs a standardized test report. The specific process of real-time interaction with the LIS system via dual-link networking is as follows: A dual-link redundancy design using 5G and wired networks is employed to construct the data transmission channel. The 5G link is prioritized for real-time data transmission, leveraging its high bandwidth and low latency characteristics to ensure that key data such as corrected antibody concentration, interference correction residuals, and kinetic fit are sent to the LIS system within milliseconds. The wired link serves as a backup channel, automatically switching to the wired network when the 5G signal fluctuates or is interrupted, ensuring uninterrupted data transmission. Through a built-in network status monitoring module, the bandwidth, latency, packet loss rate, and other indicators of the dual links are monitored in real time. When the performance of one link falls below a threshold, link switching is automatically triggered, ensuring continuous data interaction without manual intervention. The corrected antibody concentration and overall quality data are packaged and uploaded to the LIS system synchronously via dual links. The uploaded content includes: the corrected non-treponemal antibody concentration values, units, and interpretation of clinical reference intervals; the interference residual map and kinetic fit score of the AI ​​multimodal interference correction module; the dynamic magnetic flux control log of the magnetic bead processing module; and the wavelet threshold noise reduction parameters of the luminescence signal acquisition module; as well as the real-time operating parameters of each module, such as magnetic bead separation time and photon acquisition frequency, providing a basis for equipment quality control management of the LIS system.

[0043] After receiving the data, the LIS system simultaneously provides two types of key information: regional quality control statistics, which include the distribution, mean, and standard deviation of antibody concentrations for similar samples within the same region during the same period, as well as the consistency deviation rate of each laboratory; and clinically relevant information, which includes the patient's basic information, historical test records, and preliminary clinical diagnosis for the sample, and recommendations for retesting based on medical history. Once the dual links have completed the entire process of uplink data transmission and downlink information feedback, a closed loop of data interaction is formed: Regional quality control statistics show that the deviation between the system's test results and the regional mean is within the allowable range, such as <3%, and there are no special prompts in the clinically relevant information. In this case, the interaction is deemed valid, and the system directly proceeds to the standardized report output stage. If the deviation exceeds the threshold or there is clinical uncertainty, the module automatically triggers the regional quality control collaboration algorithm. Combining the returned regional data and the traceability information of each module within the system, the antibody concentration is self-calibrated to ensure the consistency of the results with the regional testing level. Finally, a self-calibrated standardized test report is output, realizing a closed loop of data interaction, quality control collaboration, and result calibration.

[0044] Output standardized test reports: By receiving regional quality control statistics from the LIS system, including the distribution, mean, standard deviation, and consistency deviation rate of antibody concentrations of similar samples in the same period within the region; and integrating the historical operating parameters and quality control records of the system's magnetic bead processing, signal acquisition, and AI correction modules, a multi-dimensional data pool of regional and local data is formed.

[0045] The regional quality control collaborative algorithm performs horizontal comparison: comparing the current sample's corrected antibody concentration with the regional mean to calculate the deviation coefficient. If the deviation exceeds the clinically permissible error threshold, such as 3%, a calibration warning is triggered. Vertical traceability: retrieving the system's quality control sample testing data for the past 7 days and the operation logs of each module, analyzing the stability of magnetic flux regulation, photon acquisition noise level, etc., to generate the equipment drift coefficient. Multimodal fusion: weighted fusion of the horizontal deviation coefficient, vertical drift coefficient, AI interference residual, and kinetic fit to calculate the final calibration coefficient, clarifying the direction and magnitude of equipment parameter adjustments.

[0046] Based on the calibration coefficient, the dynamic magnetic flux control parameters of the magnetic beads are automatically adjusted, such as the current intensity during the capture phase, the gain of the luminescence signal acquisition, and the weights of the AI ​​model, to correct the antibody concentration. After correction, the results are compared again with the regional quality control model. If the deviation still does not meet the standard, the parameters are iteratively adjusted until the deviation is within the allowable range.

[0047] The standardized test report output has a three-layer structure: the results layer clearly marks the self-calibrated antibody concentration, clinical reference range, and interpretation suggestions, such as an elevated concentration indicating an active infection phase; the quality control layer intuitively presents the reliability of the results by embedding regional quality control deviation rate, interference correction residual map, and kinetic fitting curve; and the traceability layer achieves full-process traceability by associating flux control logs processed by magnetic beads, noise suppression parameters of signal acquisition, and AI-corrected feature extraction records.

[0048] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution.

[0049] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0050] The above description is merely a specific embodiment of this application, but the scope of protection of this application 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 this application should be included within the scope of protection of this application.

Claims

1. A fully automated non-Treponema pallidum serological testing system, characterized in that: The system includes: The sample and magnetic bead processing module mixes samples with magnetic beads and uses dynamic magnetic flux control to process the magnetic beads and construct a reaction system. The luminescence signal acquisition module acquires the chemiluminescence photon signal in the reaction system, combines it with a wavelet threshold algorithm to suppress noise, and converts the noise-suppressed photon signal into a quantifiable original signal. The AI ​​multimodal interference correction module pre-acquires multi-dimensional interference features, including hemoglobin absorbance and lipid scattering characteristics. It removes multi-dimensional interference features from the original signal through a deep residual network, calculates the antibody concentration based on the original signal after interference removal, and corrects it through kinetic fit. The adaptive vector iteration module interacts with the LIS system in real time with the corrected antibody concentration through a dual-link network, performs self-calibration using a regional quality control collaborative algorithm, and outputs a standardized test report.

2. The fully automated non-treponemal serological testing system according to claim 1, characterized in that: The process of mixing the sample with the magnetic beads is as follows: The sample and magnetic beads are mixed at a volume ratio of 10:1 to form a sample flow and a magnetic bead flow. The sample flow and magnetic bead flow are introduced into a microfluidic bifurcation channel. A micropillar array is set in the microfluidic bifurcation channel to make the sample flow and magnetic bead flow form a laminar flow state, so that they can collide fully.

3. The fully automated non-treponemal serological testing system according to claim 2, characterized in that: The process of constructing the reaction system is as follows: The magnetic flux is adjusted and controlled by the dynamic current of a variable air gap electromagnet to capture, wash and separate magnetic beads and release them. The magnetic bead capture method uses high-intensity magnetic flux to rapidly deposit and enrich magnetic beads that bind antigen-antibody complexes to non-Treponema pallidum in the direction of the magnetic field. The washing and separation process switches to a medium-intensity magnetic flux while simultaneously introducing pulsed washing liquid. The release of magnetic beads is regulated by using a low-intensity magnetic flux to ensure that the magnetic beads, antigen and antibody complex are uniformly dispersed in the reaction system.

4. The fully automated non-treponemal serological testing system according to claim 1, characterized in that: The process of collecting chemiluminescent photon signals in the reaction system is as follows: A single-photon avalanche diode array is used to capture individual chemiluminescent photons. The arrival time of each photon is recorded using time-correlated single-photon counting to form a photon signal with the arrival time of the photon. A wavelet thresholding algorithm is then used to suppress noise.

5. The fully automated non-treponemal serological testing system according to claim 4, characterized in that: The process of converting the signal into a quantifiable raw signal is as follows: A calibration curve was established in advance using non-treponemal antibody standards of known concentrations. A four-parameter logistic model was used, and the noise-suppressed photon signal was substituted into the model to calculate the corresponding original signal.

6. The fully automated non-treponemal serological testing system according to claim 1, characterized in that: The process of removing multi-dimensional interference features using deep residual networks is as follows: S301: Collect multi-dimensional feature data as input signals for deep residual networks. Features include hemoglobin absorbance and lipid scattering characteristics, as well as photon signals after noise suppression. S302: The deep residual network is composed of multiple residual blocks connected in series. Each residual block contains core components such as convolutional layers and activation layers. It splits the input signal into two parts: the real effective signal and the interference residual signal, and captures the interference residual signal. S303: The deep residual network adopts a composite learning objective to correct the interfering residual signal to obtain the corrected signal. By measuring the deviation between the corrected signal and the true effective signal, the correction result is made to converge to the photon signal corresponding to the actual antibody concentration of the sample; at the same time, the deep residual network removes the interfering components in the interfering residual signal. S304: During the deep residual network training phase, a large-scale dataset containing samples with different levels of interference is used for model training. The dataset includes samples with different hemoglobin concentrations corresponding to different degrees of hemolysis and different degrees of scattering polarization corresponding to different degrees of lipemia. Each sample is paired with a non-treponemal antibody standard of known concentration to learn the signal patterns under different interference intensities and interference modes.

7. The fully automated non-treponemal serological testing system according to claim 6, characterized in that: The process of calculating antibody concentration based on the original signal after removing interference is as follows: When the trained deep residual network processes a new sample, it receives preprocessed multi-dimensional feature data and calculates the corresponding interference residual signal for the sample. This residual signal includes spectral interference from hemoglobin absorbance and scattering interference from lipid scattering characteristics. The original photon signal after noise suppression is then processed to remove the interference residual signal output by the deep residual network, and the antibody concentration is output.

8. The fully automated non-Treponema pallidum serological testing system according to claim 7, characterized in that: The process of correction based on the dynamic fit is as follows: A standard kinetic model based on the secondary reaction of non-treponemal antigen and antibody was established and the basic parameters were calibrated. Photon signals of the chemiluminescent reaction were collected, fitted with the standard kinetic model, and the goodness-of-fit threshold was determined. When the goodness-of-fit did not meet the standard, the deviation was analyzed and the parameters were adjusted for a second fitting. After the goodness-of-fit was met, the validity of the corrected photon signal was confirmed and the correction was completed.

9. The fully automated non-treponemal serological testing system according to claim 1, characterized in that: The process of real-time interaction with the LIS system via dual-link networking is as follows: It adopts a dual-link redundancy architecture of 5G and wired connections with intelligent switching to upload the calibrated antibody concentration and equipment operating status; through the feedback of regional quality control statistics and clinical correlation information from the LIS system, it sends back as self-calibration input to achieve real-time interactive closed loop of detection data.

10. The fully automated non-Treponema pallidum serological testing system according to claim 1, characterized in that: The process of generating a standardized test report is as follows: The system receives regional quality control statistics and local multi-module data to form a data pool. Through horizontal comparison, vertical traceability and multimodal fusion calculation of regional quality control collaborative algorithms, calibration coefficients are calculated and parameters are adjusted. Finally, a standardized test report with a three-layer structure of results, quality control and traceability is output and the whole process is traceable.

Images