A highly sensitive handheld multiplex fluorescence immunoassay analyzer

Abstract

The present invention discloses a highly sensitive handheld multiplex fluorescence immunoassay analyzer, comprising: a reagent card, a detection module, a data processing module, a calibration verification module, a quantitative calculation module, and an output and display module. By performing compressive sensing sparse sampling, signal reconstruction, calibration, and adversarial verification on the fluorescent markers on the reagent card, and optimizing according to the feedback of the adversarial verification results, the targeted enhancement of the fluorescent signals of ultra-low concentration target substances is achieved. This process not only effectively reduces background noise, but also improves the detection sensitivity of the handheld fluorescence immunoassay analyzer through multiple technical means, and has a lower detection limit. Compared with the prior art, the present invention solves the problems of insufficient detection sensitivity and accuracy caused by weak fluorescent signals when the handheld fluorescence immunoassay analyzer detects low-concentration samples.

Description

Technical Field

[0001] The present invention relates to the technical field of medical devices, and particularly to a highly sensitive handheld multiplex fluorescence immunoassay analyzer. Background Art

[0002] Fluorescence immunoassay is a commonly used biological detection technology, which is widely applied in fields such as medical diagnosis and food safety. Existing fluorescence immunoassay analyzers usually use LEDs as excitation light sources to detect the concentration of target substances by exciting specific fluorescent markers. However, traditional LED light sources adopt a constant-on mode during the excitation period, which limits the current of the LED lamp and cannot provide sufficient excitation energy. Especially in the detection of low-concentration samples, the fluorescence signal is weak, affecting the sensitivity and accuracy of the detection.

[0003] In order to improve the detection sensitivity, some studies have tried to improve the excitation mode of the light source, such as increasing the power of the LED or using a laser light source. However, these methods have problems such as high cost and complex operation, and it is difficult to be widely applied in portable devices.

[0004] Despite various improvement measures, in portable fluorescence immunoassay analyzers, how to improve the detection sensitivity while ensuring the portability of the device is still an urgent problem to be solved. In the prior art, the excitation energy of the constant-on LED light source is limited, resulting in a weak fluorescence signal for low-concentration samples, seriously affecting the accuracy and reliability of the detection results.

[0005] Therefore, it is necessary to provide a highly sensitive handheld multiplex fluorescence immunoassay analyzer to solve the above technical problems. Summary of the Invention

[0006] The present invention overcomes the deficiencies of the prior art and provides a highly sensitive handheld multiplex fluorescence immunoassay analyzer.

[0007] To achieve the above object, the technical solution adopted by the present invention is: a highly sensitive handheld multiplex fluorescence immunoassay analyzer, comprising:

[0008] A reagent card,

[0009] A detection module, which obtains the compressed measurement signal in the reagent card through sparse sampling and image fusion;

[0010] A data processing module, which is used to perform signal reconstruction processing on the compressed measurement signal to obtain a first enhanced signal;

[0011] A calibration and verification module, which is used to calibrate the first enhanced signal and verify the authenticity of the first enhanced signal through an adversarial verification algorithm, and feedback processing to obtain a second enhanced signal;

[0012] A quantitative calculation module for calculating the concentration of the target substance corresponding to the second enhanced signal;

[0013] An output and display module for outputting and displaying the concentration of the target substance through a visualization interface.

[0014] In a preferred embodiment of the present invention, the detection module includes:

[0015] A card slot for inserting the reagent card to a specified position;

[0016] An excitation light source assembly, including a light source, a pulse modulation circuit, and a spectroscopic system;

[0017] An optical acquisition assembly, including a CMOS image sensor, for receiving and converting the fluorescence signal into an electrical signal;

[0018] Wherein, the pulse modulation circuit controls the light source to excite the fluorescent substance with a pulse having a pulse width < 10 ns.

[0019] In a preferred embodiment of the present invention, the sparse sampling includes the following steps:

[0020] Step 1: Construct a sparse representation dictionary D based on the quantum dot fluorescence lifetime feature library;

[0021] Step 2: Obtain the peak intensity V of the fluorescence signal through pre-scanning p , calculate the sampling rate, and control the CMOS image sensor to perform pixel-level random sampling, with the sampling window synchronized with the quantum dot fluorescence lifetime window;

[0022] Step 3: Synchronize the pulsed light source with the global shutter of the CMOS image sensor, and start imaging within the quantum dot fluorescence lifetime window on the reagent card;

[0023] Continuously collect 5 - 10 frames of images, with the exposure time of each frame between 10 - 100 μs.

[0024] In a preferred embodiment of the present invention, the image fusion is specifically:

[0025] The N frames of fluorescence images obtained through sparse sampling , with the first frame I1 as the reference image, for each frame I k , calculate its displacement vector field relative to I1 , by solving: ;

[0026] Wherein, is the pixel offset, constrained within ±5 pixels; is the coordinate position of the pixel in the image;

[0027] Calculate the registration weight using an improved non-local homogeneous algorithm and perform multi-frame weighted fusion to generate a fused image I fusion : ;

[0028] where the weight w k is proportional to the square of the signal-to-noise ratio: , preferentially retaining the information of high-quality frames.

[0029] In a preferred embodiment of the present invention, the signal reconstruction is based on a sparse representation dictionary D, and the compressed measurement signal output by the detection module is iteratively optimized multiple times using the Bregman algorithm;

[0030] The iterative reconstruction process includes:

[0031] Minimize the objective function to obtain a first enhanced signal S, and the objective function is:

[0032] ;

[0033] where Ψ is a single-substance specific dictionary, M is a random sampling mask matrix, Y is the compressed measurement signal, and λ is a weight coefficient.

[0034] In a preferred embodiment of the present invention, the reagent card includes: an observation area placed in the middle and reference areas on both sides;

[0035] The reference area is equally divided into multiple sub-areas, and each sub-area contains a standard substance with a known concentration;

[0036] While the detection module collects the signal of the observation area, it synchronously collects the fluorescence signal of the reference area and records the fluorescence intensity value of each sub-area.

[0037] In a preferred embodiment of the present invention, the calibration and verification module includes: an error factor calculation unit and a signal calibration unit;

[0038] The error factor calculation unit calculates the error factor under the current detection conditions through the difference between the reference area signal and the theoretical value, and the error factor includes:

[0039] Light source fluctuation compensation: ;

[0040] Sensor drift compensation: ;

[0041] The signal calibration unit calibrates the first enhanced signal output by the data processing module through the error factor: .

[0042] In a preferred embodiment of the present invention, the calibration verification module further includes: an adversarial verification unit, which detects anomalies or interferences in the calibrated first enhanced signal through an adversarial verification algorithm to ensure the authenticity of the measurement result;

[0043] The specific detection process includes:

[0044] Construct an adversarial network architecture, including:

[0045] A generator, configured to generate simulated anomaly signals for training the discriminator;

[0046] A discriminator, configured to distinguish between real signals and generated anomaly signals and output a confidence score;

[0047] Input the calibrated first enhanced signal into the discriminator to output a confidence score D score ;

[0048] D score < θ, then the signal is determined to be abnormal;

[0049] If D score > θ, then it passes the verification.

[0050] In a preferred embodiment of the present invention, according to the output result of the adversarial verification of the calibration verification module, adjust the objective function of iterative reconstruction, and re - output the second enhanced signal. The adjusted objective function:

[0051] ;

[0052] Wherein, By penalizing signals with a large difference from the real distribution, guide the reconstruction process away from false features, and μ is the weight coefficient of the adversarial loss.

[0053] In a preferred embodiment of the present invention, the quantitative calculation module calculates the concentration of the target substance corresponding to the second enhanced signal according to the existing standard model of the target substance concentration.

[0054] The present invention solves the defects existing in the background technology and has the following beneficial effects:

[0055] (1) The present invention provides a highly sensitive handheld multiplex fluorescence immunoassay analyzer. By performing compressive sensing sparse sampling, signal reconstruction, calibration, and adversarial verification on the fluorescence markers on the reagent card, and optimizing based on the results of the adversarial verification, it realizes the targeted enhancement of the fluorescence signal of ultra-low concentration target substances. This process not only effectively reduces background noise but also improves the detection sensitivity of the handheld fluorescence immunoassay analyzer through multiple technical means, with a lower detection limit. Compared with the prior art, the present invention solves the problem of insufficient detection sensitivity and accuracy of the handheld fluorescence immunoassay analyzer when detecting low-concentration samples due to weak fluorescence signals.

[0056] (2) The present invention can effectively reduce random noise and improve the signal-to-noise ratio by sparse sampling and fusing multiple frames of images. Specifically, continuously collecting 5 - 10 frames of images and aligning and fusing these images through the non-local means (NLM) algorithm can significantly reduce noise interference. This method not only improves the image quality but also enhances the detection ability of low-concentration samples without increasing the hardware complexity.

[0057] (3) Based on the sparse representation dictionary, the present invention uses the Bregman algorithm to perform multiple iterative optimizations on the compressed measurement signals output by the detection module, and recovers a high-resolution first enhanced signal from a small number of compressed measurement signals. Especially in low-concentration detection scenarios (such as the TSH detection at 0.01 ng / mL), this method significantly reduces the signal distortion caused by insufficient sampling. By optimizing the selection of the sparse representation dictionary and the setting of iterative parameters, the present invention can greatly reduce the computational complexity and processing time while ensuring high precision.

[0058] (4) The present invention corrects the light source fluctuation and sensor drift in real time through the reference area on the edge of the reagent card, and uses the adversarial verification of the generative adversarial network (GAN) to ensure the authenticity of the first enhanced signal. When the statistical difference between the reconstructed signal and the true distribution is significant, the adversarial loss function will dynamically increase the weight of the penalty term through the gradient backpropagation mechanism, forcing the iterative optimization process to adjust the core parameters. This feedback mechanism can effectively suppress the false features introduced by algorithm biases such as undersampling error or excessive noise suppression during the reconstruction process by strengthening the constraint on the distribution consistency, thereby improving the fitting accuracy of the reconstructed signal and the physical true value. It is particularly important in low-concentration detection and can significantly reduce false positive or false negative results caused by algorithm errors.

[0059] (5) By precisely controlling the excitation conditions of the light source (such as pulse width, frequency, and peak current) and combining with the characteristics of quantum dot fluorescent markers, the present invention maximizes the generation efficiency of fluorescence signals. Secondly, by using sparse sampling technology and Bregman iterative algorithm, high-quality original signals are recovered from a small number of compressed measurement signals, avoiding the high cost and complexity brought by traditional full sampling. In addition, abnormal signals are identified and removed through the adversarial verification algorithm, ensuring the authenticity and reliability of the final detection results. By integrating these technical means, the present invention exhibits excellent performance in low-concentration detection, can significantly reduce the detection limit, and meet the requirements of high-sensitivity detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments recorded in the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0061] Figure 1 It is a schematic diagram of the handheld immunoassay analyzer body of the preferred embodiment of the present invention;

[0062] Figure 2 It is a block diagram of the components of the handheld immunoassay analyzer body of the preferred embodiment of the present invention;

[0063] Figure 3 It is the optical structure composition of the preferred embodiment of the present invention.

[0064] In the figure: 1. Immunoassay analyzer body; 101. Light source; 102. Collimating mirror; 103. Dichroic mirror; 104. First condenser lens; 105. CMOS image sensor; 106. Diaphragm; 107. Second condenser lens; 108. Filter. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0065] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the protection scope of the present invention.

[0066] Many specific details are set forth in the following description in order to provide a thorough understanding of the present invention, but the present invention can also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited by the specific embodiments disclosed below.

[0067] In the description of the present application, it should be understood that the orientation or positional relationship indicated by the terms "center", "longitudinal", "transverse", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc. is based on the orientation or positional relationship shown in the drawings, and is only for the convenience of describing the present application and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as limiting the protection scope of the present application. In addition, the terms "first", "second", etc. are only used for descriptive purposes and cannot be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Therefore, the features defined with "first", "second", etc. may explicitly or implicitly include one or more of such features. In the description of the present invention, unless otherwise specified, the meaning of "a plurality" is two or more.

[0068] In the description of the present application, it should be noted that unless otherwise clearly specified and limited, the terms "mounted", "connected", and "coupled" should be understood in a broad sense. For example, it may be a fixed connection, a detachable connection, or an integral connection; it may be a mechanical connection or an electrical connection; it may be directly connected or indirectly connected through an intermediate medium, and it may be the communication inside two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present application can be understood through specific circumstances.

[0069] As Figure 1 shown, the present invention provides a highly sensitive handheld multiplex fluorescence immunoassay analyzer. The handheld immunoassay analyzer includes a reagent card and an immunoassay analyzer body 1 for obtaining fluorescence detection information in the reagent card and performing data processing to obtain the concentration of the target substance.

[0070] The reagent card uses quantum dots as fluorescent markers. After the sample to be tested is mixed with the fluorescently labeled antibody prefabricated on the conjugate pad, an immune reaction occurs to form a complex of the substance to be tested - fluorescently labeled antibody. This complex is captured and enriched by the antibody coated on the NC membrane through the chromatography process, while the remaining components are separated and removed.

[0071] Specifically, a number of drainage channels are designed on the reagent card, and the drainage channels are connected to the uniformly arranged detection channels. The serum sample is dropped into the sample application area of the reagent card, and the sample flows along the drainage channels by capillary action and mixes and reacts with the fluorescently labeled antibody in the detection channels. Above the detection channels, an observation area is arranged in the middle and reference areas are arranged on both sides.

[0072] It should be noted that on the reagent card, reference areas are set on both sides of the observation area. The reference areas are equally divided into multiple sub-areas, and each sub-area contains standard substances with known concentrations (such as 0.1, 1, 10 ng / mL of TSH). The same fluorescent label is used in the reference area as in the observation area, and the light source 101 and the sensor respond consistently.

[0073] As Figure 2 shown, the immunological analyzer body 1 includes: a detection module, a data processing module, a calibration and verification module, a quantitative calculation module, and an output and display module.

[0074] Among them, the detection module is used to obtain the compressed measurement signal in the reagent card through the excitation light source 101. The detection module includes: a card slot, an excitation light source 101 assembly, and an optical acquisition assembly. Specifically:

[0075] The card slot is set on the immunological analyzer body 1 for inserting the reagent card. After insertion, the Observation area on the reagent card is aligned with the optical acquisition assembly. Further, an adjustable snap-type fixture is set in the card slot to fix the reagent card through an elastic pressing piece.

[0076] As Figure 3 shown, the excitation light source assembly includes: a light source 101, a pulse modulation circuit, and a beam splitting system. Among them, the light source 101 is configured as a multi-wavelength LED array, integrating multi-wavelength light sources 101 such as a 365nm ultraviolet LED (exciting quantum dots) and a 650nm near-infrared LED (exciting fluorescent microspheres), supporting the requirements of multiplex detection. The light source 101 irradiates the fluorescent substance through the excitation light to excite a fluorescent signal, thereby completing the detection, and the wavelength of the light source 101 needs to be determined according to the fluorescent substance. The pulse modulation circuit is configured as a nanosecond-level pulse drive chip, preferably a PWM controller, to control the pulsed excitation of the light source 101, and the pulse width is controlled within <10ns. The beam splitting system includes: a collimating mirror 102, a dichroic mirror 103 arranged on the path of the light source 101, and a first condenser lens 104 arranged perpendicular to the dichroic mirror 103; among them, the dichroic mirror 103 is installed at a 45° angle, and the condenser lens is directly above the observation area of the reagent card.

[0077] The optical acquisition assembly includes: a CMOS image sensor 105, a diaphragm 106, a second condenser lens 107, and a filter 108 arranged in sequence from top to bottom; the CMOS image sensor 105, the diaphragm 106, the second condenser lens 107, and the filter 108 are all arranged directly above the dichroic mirror 103, and the center points of the CMOS image sensor 105, the diaphragm 106, the second condenser lens 107, the filter 108, the dichroic mirror 103, and the first condenser lens 104 are on a vertical straight line. Among them, the CMOS image sensor 105 is used to receive and convert the optical signal, that is, to convert the fluorescent signal into an electrical signal.

[0078] In the optical design of the excitation light source component and the optical acquisition component, the reflection of the dichroic mirror 103 can change the path of the excitation light, so that the excitation light irradiates the reagent card, and the excited fluorescence can directly reach the optical acquisition component through the dichroic mirror 103 to complete the reception of the compressed measurement signal.

[0079] Furthermore, the sparse sampling specifically includes the following steps:

[0080] Step 1: Construct a sparse representation dictionary D based on the quantum dot fluorescence lifetime feature library. The dictionary includes multiple decay basis functions , where is the fluorescence lifetime related to the quantum dot particle size, and is the fluorescence oscillation frequency.

[0081] The basis functions in the sparse representation dictionary D need to match the fluorescence characteristics of the target substance. For example, when detecting TSH, select the decay basis function with the closest fluorescence lifetime (such as 50 ns) and oscillation frequency (such as 2 MHz) to the quantum dot bound to the TSH-labeled antibody to ensure the accuracy of the subsequent reconstructed signal.

[0082] Step 2: Obtain the peak intensity V of the fluorescence signal through pre-scanning p , and dynamically select the construction mode of the observation matrix Φ according to the intensity interval of V p :

[0083] ;

[0084] Among them, is the Fourier observation matrix that retains the low-frequency components, generated by intercepting the low-frequency coefficients in the frequency domain; is the Bernoulli random matrix that captures transient characteristics, and the elements take values of ±1 and follow independent and identical distributions; the symbol ⊕ represents the concatenation operation of matrices, that is, the hybrid sampling mode covering the entire frequency band.

[0085] For high signal intensity (Vp > 1V), use the Fourier low-frequency observation matrix. Under strong signals, the main components of the fluorescence decay curve are concentrated in the low-frequency band (such as the exponential decay basis related to the quantum dot lifetime). Through Fourier low-frequency sampling, the core energy of the signal can be efficiently captured, redundant data can be reduced, and high-frequency noise can be suppressed.

[0086] For medium signal intensity (0.1V < V p ≤ 1V), use the hybrid observation matrix. Medium-intensity signals need to balance the low-frequency main body and high-frequency details (such as transient oscillations caused by quantum dot aggregation). The Fourier component retains the main decay trend, and the Bernoulli random matrix captures sparse high-frequency features through random projection to balance the data volume and information integrity.

[0087] For weak signal intensity (Vp (≤0.1V), a full - band observation matrix is adopted. At extremely low concentrations, the high - frequency components of weak signals (such as single - molecule fluorescence blinking of quantum dots) may carry key information. The cascaded observation matrix covers the full band, and by increasing the sampling rate, the loss of high - frequency features is avoided to ensure the detection limit.

[0088] The FPGA hardware is used to control the CMOS image sensor 105 for pixel - level random sampling. The sampling window is synchronized with the quantum dot fluorescence lifetime window to generate a compressed measurement signal , where x is the original fluorescence decay signal, ε is the measurement noise and satisfies .

[0089] Step 3: Pulse light source 101 is synchronized with the global shutter of CMOS image sensor 105, and imaging is started within the quantum dot fluorescence lifetime window on the reagent card. During imaging, 5 - 10 frames of images are continuously acquired, with the exposure time of each frame between 10 - 100 μs and the frame interval of 2 ms to match the chromatography flow velocity.

[0090] Next, the N frames of fluorescence images collected are matched and fused. Specifically:

[0091] For the N frames of fluorescence images collected continuously , with the first frame I1 as the reference image, for each frame I k (k = 1 to N), calculate its displacement vector field relative to I1 , by solving:

[0092] , where is the pixel offset, constrained within ±5 pixels; is the coordinate position of the pixel in the image;

[0093] By minimizing the pixel difference between the reference frame I1 and the k - th frame I k , find the optimal displacement vector Δ to align the two frames.

[0094] The improved non - local homogeneous algorithm is used to calculate the registration weight , and multi - frame weighted fusion is performed to generate a fused image I fusion : ;

[0095] where the weight w k is proportional to the square of the signal - to - noise ratio: , preferentially retaining the information of high - quality frames, and multi - frame fusion improves the signal - to - noise ratio through weighted averaging to obtain a compressed measurement signal.

[0096] The data processing module is used to perform signal reconstruction processing on the compressed measurement signal to obtain a first enhanced signal.

[0097] Signal reconstruction is based on the sparse representation dictionary D, and the Bregman algorithm is used to iteratively optimize the compressed measurement signal output by the detection module multiple times. In the initial stage, the signal is assumed to be in a zero-value state, and the signal estimate value is gradually adjusted until the change between two consecutive estimate values is less than 0.01% to achieve signal reconstruction and gradually approach the original signal. Specifically, this unit can recover high-resolution signals from the compressed measurement signals output by the detection module, especially under low-concentration conditions, significantly reducing signal distortion caused by insufficient sampling.

[0098] It should be noted that the iterative reconstruction process includes:

[0099] Minimize the objective function to obtain the first enhanced signal S, and the objective function is ; where Ψ is a single-substance dedicated dictionary (such as the specific basis function combination corresponding to TSH), M is a random sampling mask matrix, and Y is the compressed measurement signal. For example, the quantum dots conjugated with TSH-labeled antibodies have specific fluorescence lifetimes (such as 50 ns) and oscillation frequencies (such as 2 MHz), and these characteristics are encoded in the dictionary Ψ. The L1 norm encourages the solution S to be as sparse as possible under the dictionary Ψ, that is, as few basis functions as possible are involved in signal reconstruction, thereby removing redundant information. And the parameter λ is a weight coefficient used to balance the relationship between sparsity and data fidelity.

[0100] Use Bregman iteration to solve the minimized objective function.

[0101] Initialize the signal S （0） = 0, that is, start constructing the signal from 0; at the same time, initialize the Bregman variable B (0) = 0, and this variable is used to store the cumulative error in each iteration.

[0102] Each iteration includes:

[0103] Update the signal estimate S (K+1) : ;

[0104] In each iteration, the current signal estimate S and the cumulative error B (k) are added as the new input. Through the L1 norm constraint, the new signal estimate S (k+1) is encouraged to be as sparse as possible under the dictionary Ψ. At the same time, through the L2 norm constraint, it is ensured that the new signal estimate S (k+1) is as close as possible to the compressed measurement signal Y.

[0105] Update the Bregman variable B (k+1) : ;

[0106] The Bregman variable B (k+1)Records the cumulative error in each iteration. After each iteration, B is updated (k+1) by adding the difference between the new and old signal estimates to the cumulative error of the previous iteration, i.e., . This helps to gradually correct the bias in the signal estimate.

[0107] The iteration process continues until the following stopping condition is met: ;

[0108] where is a preset small threshold (e.g., 0.01%). When the change in signal estimate between two consecutive iterations is less than this threshold, the algorithm is considered to have converged, and the final signal reconstruction result S, i.e., the first enhanced signal, is obtained.

[0109] Through the above iteration process, the Bregman algorithm can recover the high-resolution first enhanced signal S from a small number of compressed measurement signals Y. Especially in low-concentration detection scenarios (such as TSH detection at 0.01 ng / mL), it significantly reduces signal distortion caused by insufficient sampling. Due to the sparsity constraint of the L1 norm, the Bregman algorithm can effectively suppress noise, especially high-frequency noise that does not conform to the characteristics of the dictionary Ψ. At the same time, the data fidelity constraint of the L2 norm ensures the consistency between the reconstructed signal and the actual observed signal, avoiding the loss of effective signals caused by over-smoothing. In summary, the Bregman iterative reconstruction algorithm effectively recovers the original high-resolution signal from compressed measurement signals by combining sparse representation and data fidelity constraints.

[0110] It should be noted that while the detection module collects the signals in the observation area, the fluorescence signals in the reference area are synchronously collected, and the fluorescence intensity values of each sub-region are recorded.

[0111] The calibration and verification module calibrates the first enhanced signal based on the reference area signal, including compensation for light source fluctuations and sensor drift, and verifies the authenticity of the first enhanced signal through an adversarial verification algorithm, and the second enhanced signal is obtained through feedback processing.

[0112] Based on the signals of standard substances with known concentrations, a calibration curve of concentration-fluorescence intensity is established.

[0113] This calibration and verification module includes: an error factor calculation unit, a signal calibration unit, and an adversarial verification unit.

[0114] Among them, the error factor calculation unit calculates the error factor under the current detection conditions through the difference between the reference area signal and the theoretical value. This error factor includes:

[0115] Light source fluctuation compensation: ;

[0116] If the light source intensity decays over time, light source fluctuation compensation is used to adjust the signal in the observation area;

[0117] Sensor drift compensation: ;

[0118] By comparing the repeated measurement values in the reference area, the change in sensor sensitivity is corrected.

[0119] The signal calibration unit calibrates the first enhanced signal output by the data processing module through an error factor:

[0120] 。

[0121] The adversarial verification unit detects anomalies or interferences in the calibrated first enhanced signal through an adversarial verification algorithm to ensure the authenticity of the measurement results. The specific detection process includes:

[0122] The calibrated first enhanced signal and the real signal in the reference area are respectively input into the adversarial network architecture.

[0123] The adversarial network architecture includes:

[0124] The generator is configured to generate simulated anomaly signals (such as noise interference, concentration tampering) for training the discriminator. The input of the generator is the real signal in the reference area + random noise, and the output is the synthetic anomaly signal (such as signal intensity mutation, lifespan anomaly).

[0125] The discriminator is configured to distinguish between real signals and generated anomaly signals and output a confidence score (0 - 1). The input of the discriminator is the real signal in the reference area and the synthetic anomaly signal output by the generator. The discriminator determines whether the signal is real (0 for anomaly, 1 for real).

[0126] Based on the above - constructed adversarial network architecture, pre - training and adversarial training are carried out in sequence. In the pre - training stage, the discriminator is trained using the known concentration signals in the reference area so that it learns the distribution characteristics of real signals (such as fluorescence lifespan, oscillation frequency, intensity distribution). In the adversarial training stage, the generator continuously generates more realistic anomaly signals, and the discriminator continuously optimizes its classification ability until the Nash equilibrium is reached (the generator cannot interfere with the discriminator).

[0127] The calibrated first enhanced signal is input into the discriminator, and a confidence score D is output score 。If D score <θ, θ = 0.8, then it is determined that the signal is abnormal, that is, there is interference or measurement error; if D score >θ, then it passes the verification and the signal is credible.

[0128] Furthermore, according to the adversarial verification result, adjust the hyperparameters or optimization objectives of the signal reconstruction unit for iterative reconstruction, including:

[0129] The false feature is a noise misjudgment. Increase λ to enhance data fidelity;

[0130] The false feature is signal distortion. Decrease λ to allow more basis functions to participate in the reconstruction.

[0131] Specifically, adjust the objective function of the iterative reconstruction according to the output result of the adversarial verification, and re-output the second enhanced signal. The adjusted objective function is: , where By punishing the signal with a large difference from the true distribution, guide the reconstruction process away from false features, and μ is the weight coefficient of the adversarial loss, which needs to be adjusted according to experiments.

[0132] By implementing an adversarial verification mechanism on the first enhanced signal, when the statistical difference between the reconstructed signal and the true distribution (defined by the reference standard substance in the reference area) is significant, the adversarial loss function will dynamically increase the weight of the penalty term through the gradient backpropagation mechanism, forcing the iterative optimization process to adjust the core parameters. This feedback mechanism can effectively suppress the false features (such as pseudo-lifetime signals or oscillation frequency distortions) introduced by algorithmic biases such as undersampling errors or excessive noise suppression in the reconstruction process by strengthening the constraint on distribution consistency, thereby improving the fitting accuracy of the reconstructed signal and the physical true value. It is particularly important in low-concentration detection and can significantly reduce false positive or false negative results caused by algorithm errors.

[0133] A quantitative calculation module is used to calculate the concentration of the target substance corresponding to the second enhanced signal. Specifically: Calculate the concentration of the target substance corresponding to the second enhanced signal according to the existing standard model of the target substance concentration.

[0134] An output and display module is used to output and display the concentration of the target substance through a visualization interface. Specifically: The output result (concentration of the target substance and auxiliary information) of the quantitative calculation module is displayed in a user-friendly form in real time, and data recording and export are supported.

[0135] The output and display module further includes: A visualization interface provided on the immunoassay analyzer body 1 for displaying the concentration of the target substance and auxiliary information.

[0136] The present invention provides a highly sensitive handheld multiplex fluorescence immunoassay analyzer. By performing compressive sensing sparse sampling, signal reconstruction, calibration, and adversarial verification on the fluorescent markers on the reagent card, and optimizing based on the feedback of the adversarial verification results, it realizes the targeted enhancement of the fluorescent signals of ultra-low concentration target substances. This process not only effectively reduces background noise but also improves the detection sensitivity of the handheld fluorescence immunoassay analyzer through multiple technical means, with a lower detection limit. Compared with the prior art, the present invention solves the problem of insufficient detection sensitivity and accuracy of the handheld fluorescence immunoassay analyzer due to weak fluorescent signals when detecting low-concentration samples.

[0137] To verify the performance of this handheld multiplex fluorescence immunoassay analyzer in low-concentration detection, a gradient concentration test was conducted on thyroid-stimulating hormone (TSH). TSH serum samples with a concentration range of 0.001 - 1 ng / mL were selected, and each group was repeatedly detected 20 times and compared with a traditional fluorescence immunoassay analyzer. The experimental results show that:

[0138] Higher sensitivity: At an ultra-low concentration of 0.01 ng / mL, the signal-to-noise ratio (SNR) of the present invention reaches 4.2 ± 0.3, which is significantly higher than 1.8 ± 0.5 of the traditional instrument, indicating that the compressive sensing sparse sampling and Bregman iterative reconstruction technology effectively enhance weak signals;

[0139] Lower detection limit: By dynamically compensating for light source fluctuations and sensor drift through the calibration verification module, and combining the rejection of abnormal signals by the adversarial verification unit (the rejection rate is 100% when Dscore < 0.8), the LOD of the present invention is as low as 0.008 ng / mL (blank signal mean + 3σ), which is reduced by one order of magnitude compared with the traditional instrument (LOD = 0.1 ng / mL);

[0140] And in a complex matrix containing 1 mg / mL human serum albumin (HSA), the detection recovery rate of 0.01 ng / mL TSH is 97.5% - 102.3%, which confirms the suppression effect of the multi-frame fusion algorithm and the adversarial network on false features.

[0141] In summary, the present invention realizes a breakthrough in the sensitivity and reliability of the handheld fluorescence immunoassay analyzer in ultra-low concentration detection through the pulse excitation - sparse sampling - adversarial verification closed-loop technology chain.

[0142] Based on the ideal embodiments of the present invention as inspiration, through the above description, relevant personnel can completely make various changes and modifications without departing from the technical idea of this invention. The technical scope of this invention is not limited to the content in the specification, and the technical scope must be determined according to the scope of the claims.

Claims

1. A highly sensitive handheld multi-detection fluorescent immunoassay analyzer, characterized in that: include: Reagent Card, A detection module, which obtains the compressed measurement signal in the reagent card through sparse sampling and image fusion; A data processing module, performing signal reconstruction processing on the compressed measurement signal based on a Bregman algorithm to obtain a first enhanced signal; A calibration verification module, used to calibrate the first enhanced signal, verify the authenticity of the first enhanced signal through an adversarial verification algorithm, and obtain a second enhanced signal through feedback processing; A quantitative calculation module, used to calculate the concentration of the target substance corresponding to the second enhanced signal; The output and display module is used to output and display the concentration of the target substance through a visual interface.

2. A highly sensitive handheld multi-detection fluorescent immunoassay analyzer according to claim 1, characterized in that: The detection module comprises: A card slot, used to insert the reagent card into a designated position; An excitation light source assembly, including a light source, a pulse modulation circuit and a light splitting system; An optical collection component, including a CMOS image sensor, for receiving and converting the fluorescence signal into an electrical signal; The pulse modulation circuit controls the light source to excite the fluorescent substance with a pulse with a pulse width of less than 10 ns.

3. A highly sensitive handheld multi-detection fluorescent immunoassay analyzer according to claim 2, characterized in that: The sparse sampling comprises the following steps: Step 1: construct a sparse representation dictionary D based on the quantum dot fluorescence lifetime feature library; Step 2: Obtain the peak intensity V of the fluorescence signal through pre-scanning p , calculating the sampling rate, and controlling the CMOS image sensor to perform pixel-level random sampling, with the sampling window being synchronized with the quantum dot fluorescence lifetime window; Step 3: Using a pulse light source to synchronize with the global shutter of the CMOS image sensor to start imaging within the fluorescence lifetime window of the quantum dots on the reagent card; 5-10 frames of images were collected continuously, and the exposure time of each frame was between 10-100 μs.

4. A highly sensitive handheld multi-detection fluorescent immunoassay analyzer according to claim 1, characterized in that: The image fusion is specifically as follows: N frames of fluorescence images acquired by sparse sampling , taking the first frame I1 as the reference image, for each frame I k , calculate its displacement vector field relative to I1 , by solving: ; in, is the pixel offset, constrained within ±5 pixels; is the coordinate position of the pixel in the image; The registration weights are calculated using the improved non-local means algorithm. , and perform multi-frame weighted fusion to generate a fused image I fusion : ; The weight w k Proportional to the square of the signal-to-noise ratio: , giving priority to preserving the information of high-quality frames.

5. A high-sensitivity handheld multi-detection fluorescent immunoassay analyzer according to claim 3, characterized in that: The signal reconstruction is based on the sparse representation dictionary D, and the Bregman algorithm is used to perform multiple iterations of optimization on the compressed measurement signal output by the detection module; The iterative refactoring process includes: The first enhanced signal S is obtained by minimizing the objective function, which is: ； Among them, Ψ is a single-substance-specific dictionary, M is a random sampling mask matrix, Y is the compressed measurement signal, and λ is a weight coefficient.

6. A highly sensitive handheld multi-detection fluorescent immunoassay analyzer according to claim 1, characterized in that: The reagent card comprises: an observation area in the middle and reference areas on both sides; The reference area is equally divided into a plurality of sub-areas, each of which contains a standard substance of known concentration; The detection module collects the signal of the observation area and the fluorescence signal of the reference area at the same time, and records the fluorescence intensity value of each sub-area.

7. A high-sensitivity handheld multi-detection fluorescent immunoassay analyzer according to claim 6, characterized in that: The calibration verification module includes: an error factor calculation unit and a signal calibration unit; The error factor calculation unit calculates the error factor under the current detection conditions by the difference between the reference zone signal and the theoretical value. The error factor includes: Light source fluctuation compensation: ; Sensor drift compensation: ; A signal calibration unit is used to calibrate the first enhanced signal output by the data processing module using an error factor: .

8. A high-sensitivity handheld multi-detection fluorescent immunoassay analyzer according to claim 1, characterized in that: The calibration verification module further includes: a confrontation verification unit, which detects anomalies or interferences in the calibrated first enhanced signal through a confrontation verification algorithm to ensure the authenticity of the measurement result; The specific testing process includes: Build an adversarial network architecture, including: A generator configured to generate a simulated abnormal signal for training the discriminator; A discriminator, configured to distinguish between real signals and generated anomaly signals and output a confidence score; The calibrated first enhanced signal is input into the discriminator to output the confidence score D score ; D score <θ, then the signal is judged to be abnormal; If D score >θ, the verification is passed.

9. A high-sensitivity handheld multi-detection fluorescent immunoassay analyzer according to claim 5, characterized in that: The objective function of iterative reconstruction is adjusted according to the output result of the calibration verification module against the verification, and the second enhanced signal is re-output. The adjusted objective function is: ； in, By penalizing signals that are far different from the true distribution, the reconstruction process is guided away from false features, and μ is the weight coefficient of the adversarial loss.

10. A high-sensitivity handheld multi-detection fluorescent immunoassay analyzer according to claim 1, characterized in that: The quantitative calculation module calculates the target substance concentration corresponding to the second enhanced signal according to an existing target substance concentration standard model.

Images