Immunodetection method based on holographic imaging mode conversion and deep learning

Abstract

The invention discloses an immunodetection method based on holographic imaging mode conversion and deep learning, and belongs to the field of immunoassay detection. The deep learning holographic mode conversion system constructed based on the non-paired data set reduces the influence of diffraction interference on target quantification to the greatest extent, so that holographic super-resolution imaging is realized, and challenges in paired data set generation are overcome; besides, a functional glass slide is designed, and rapid and sensitive holographic imaging detection is realized in combination with a click reaction signal amplification strategy; the functionalized glass slide is combined with the portable holographic imaging equipment, so that the detection efficiency of the system is greatly improved while accurate signal amplification is carried out, the application scene is expanded, and antibiotic residues can be rapidly and sensitively detected. The invention opens up a new research direction for high-sensitivity on-site convenient detection of trace target objects in portable equipment and complex samples.

Description

Technical Field

[0001] The present invention relates to an immune detection method based on holographic imaging modal conversion and deep learning, and belongs to the field of immune analysis detection. Background Art

[0002] Detection strategies based on immunological methods, such as enzyme-linked immunosorbent assay (ELISA), lateral flow immunoassay, and microsphere-mediated immunobiosensors, are widely used in clinical diagnosis, food safety, and environmental monitoring. Although these basic immunoassay methods have the advantages of high specificity and ease of use, they have limited sensitivity and accuracy in detecting trace targets in complex real samples. Currently, microscopic bright-field optical immunobiosensors based on immunomicrospheres have attracted widespread interest due to their high sensitivity. This immunobiosensor can achieve high sensitivity and high accuracy in a short time by capturing microsphere signals through microscopic imaging through changes in the number of microspheres mediated by target concentration. However, traditional microsphere counting detection analysis using ordinary optical microscopes is limited by small imaging field of view and poor portability. The imaging field of view decreases with increasing magnification, which hinders the collection of various sample signals and easily leads to signal loss. In addition, sampling errors are easily amplified under a small imaging field of view. Therefore, achieving a larger imaging field of view at the same magnification is crucial for the field of microsphere imaging-based immunosensing.

[0003] Lensless holographic imaging technology has become a widely-watched technology due to its low cost, magnification comparable to that of traditional microscopes, large imaging field of view and no loss. Lensless holographic technology transcends the optical limitations imposed by lenses. It can directly place the specimen on the imaging photosensor to capture the image without using a lens, and a clear object image reconstruction can be achieved using the corresponding image reconstruction algorithm.

[0004] The use of coherent light sources in lensless holographic imaging technology produces interference and diffraction patterns, which appear as concentric fringes. These diffraction interferences greatly reduce the image quality and hinder the identification of analytes. In order to restore the focusing information of the holographic image, a holographic reconstruction algorithm is required to perform a modal transformation (from the holographic state to the microscopic state) to achieve super-resolution imaging. Currently, reconstruction algorithms are divided into physical and optically driven reconstruction methods and data-driven reconstruction methods based on deep learning. Physical and optically driven reconstruction methods rely on analysis or iterative calculation of hardware parameters (such as axial distance and wavelength), so they are generally suitable for static objects. However, these methods require a lot of measurements and adjustments, which are time-consuming and complex to operate.

[0005] In contrast, conventional deep learning data-driven algorithms rely on aligned and labeled image datasets for training, and are heavily dependent on the accuracy and stability of the aligned datasets during the learning process. However, in fields such as real-time diagnosis and biomedical imaging, the background features of most imaging environments are complex, and the calibration process is also very complicated, making it very difficult to construct large paired datasets with the same magnification and the same position under different modal conditions. Therefore, the practical application of deep learning modal conversion based on paired datasets is still quite challenging. Unpaired dataset algorithms do not require special calibration, facilitate data collection, and have no fixed quantity requirements. Unpaired dataset algorithms can effectively reduce pairing bias and enhance the model's resistance to external factors. Therefore, realizing unpaired dataset modal conversion is crucial for the practical application of holographic imaging. Summary of the invention

[0006] In order to solve the above problems, the present invention provides an immune detection method based on holographic imaging modality conversion and deep learning, and the technical solution is as follows:

[0007] Step 1: Mix the target object to be detected, the pre-prepared CuO2@SiO2 complex and the nanomagnetic particle conjugate to perform a competitive immune reaction; the surface of the CuO2@SiO2 complex is modified with a detection antibody that recognizes the target object to be detected, and the surface of the nanomagnetic particle conjugate is modified with a complete antigen of the target object to be detected;

[0008] Step 2: magnetic separation, dissolving the remaining CuO2@SiO2 composite after reaction in hydrochloric acid and sodium ascorbate to release copper ions;

[0009] Step 3: Copper ions act as catalysts to rapidly catalyze the azide-alkyne cycloaddition reaction at room temperature, so that an azide-alkyne cycloaddition reaction occurs between the alkyne-modified polystyrene microspheres and the azide ultrathin functionalized glass slide, i.e., a click reaction;

[0010] Step 4: performing holographic imaging on the polystyrene microspheres obtained in step 3 and fixed on the functionalized glass slide through the click reaction;

[0011] Step 5: construct an unpaired modal conversion data set using the holographic imaging image of the isogradient diluted polystyrene microspheres and the microscopic imaging image of the isogradient diluted polystyrene microspheres, and train them using a deep learning modal conversion model, and then use the trained deep learning modal conversion model to perform modal conversion on the holographic imaging image obtained in the experiment of step 4 to obtain a microscopic modal image;

[0012] Step 6: Input the microscopic modal image obtained after the modal conversion and the microscopic imaging image of the isogradient diluted polystyrene microspheres into a deep learning target detection network for training, and then use the trained deep learning target detection network to identify and count the microscopic modal images that have undergone modal conversion in the experiment of step 4;

[0013] Step 7: Using the linear relationship between the number of the polystyrene microspheres and the concentration of the target to be detected, the concentration information of the target to be detected is obtained.

[0014] Optionally, the modality conversion model uses a cyclic generative adversarial network CycleGAN as the model backbone network.

[0015] Optionally, the deep learning target detection network uses YOLOV7 as the backbone network.

[0016] Optionally, the copper ions are monovalent copper ions generated by a reduction reaction of sodium ascorbate.

[0017] Optionally, the click reaction is a reaction in which azide and alkyne react to form a covalent bond under the catalysis of monovalent copper ions.

[0018] Optionally, the polystyrene microspheres are 6 μm polystyrene microspheres with alkyne modified on the surface.

[0019] Optionally, the method further comprises processing the holographic image using a non-local mean filtering algorithm.

[0020] Optionally, the nanomagnetic particle conjugate is: a 150 nm nanomagnetic particle with a complete antigen of the target object to be detected modified on the surface.

[0021] Optionally, the nanomagnetic particle conjugate is added to the reaction system at a concentration of 20 μg / mL.

[0022] Optionally, the CuO2@SiO2 complex is added to the reaction system at a concentration of 20 μg / mL.

[0023] Optionally, the surface of the functionalized glass slide is functionalized and modified with azide.

[0024] Optionally, the copper ions are monovalent copper ions generated by a reduction reaction of sodium ascorbate.

[0025] Optionally, the target object is a small molecule antibiotic, including at least one of chloramphenicol, neomycin, and clarithromycin.

[0026] Optionally, the maximum amount of hydrochloric acid added to the reaction system is 8 mM.

[0027] Optionally, the polystyrene microsphere-azide complex is added to the reaction system at a concentration of 50 μg / mL.

[0028] Optionally, the complete antigen-alkyne complex is added to the reaction system at a concentration of 4 mg / mL.

[0029] Optionally, the functionalized slides are pre-made and stored at 4°C.

[0030] Optionally, the nanomagnetic particle conjugate is incubated in the reaction system at 20-40° C. for at least 10 min, or at 37° C. for 30 min.

[0031] The beneficial effects of the present invention are:

[0032] The present invention combines a holographic unpaired data set super-resolution modal conversion system with a signal amplification strategy based on click chemistry, and can rapidly and sensitively detect the target object to be detected.

[0033] The present invention utilizes portable holographic imaging technology to replace the traditional bulky microscope as a signal readout device. Holographic imaging technology has the advantages of wide field of view, lossless magnification, low cost and good portability. It realizes holographic super-resolution reconstruction through the unpaired data set modal conversion algorithm, solves the diffraction interference caused by holographic imaging, overcomes background interference to a great extent, and ensures the sensitivity of detection.

[0034] The present invention combines a competitive immune response with a signal amplification system based on a click reaction, relies on the high specificity of the competitive immune response and the precipitation of a large amount of copper ions encapsulated in the CuO2@SiO2 nanomaterial, and realizes efficient and accurate signal transmission based on the efficient catalytic properties of the click reaction.

[0035] The present invention designs a functionalized glass slide and can be combined with a portable holographic imaging device, which greatly improves the detection efficiency and application scenarios of the system while accurately amplifying the signal.

[0036] The detection system of the present invention has high sensitivity, strong specificity, simple operation, high detection efficiency and broad application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0038] Figure 1The flowchart of the unpaired dataset modal transformation immunoassay method based on holographic imaging of the present invention is shown in Figure 1. (a) is a workflow diagram of antibiotic detection of unpaired dataset modal transformation immunosensor based on holographic imaging; (b) is an architecture diagram of the lensless holographic imaging hardware and software system.

[0039] Figure 2 The figure is a preparation principle and flow chart of the ultrathin functionalized glass slide of the present invention.

[0040] Figure 3 This is a flow chart of the unpaired data set modality transformation algorithm system of the present invention.

[0041] Figure 4 The figures are the training results of the unpaired dataset modality conversion algorithm of the present invention, where (a) is the cycle consistency loss curve; (b) is the object detection loss curve; (c) is the mAP@0.5, (d) is the bounding box loss curve; and (e) is the Precision-Recall curve.

[0042] Figure 5 It is a comparison diagram of the original hologram image, the filtered image and the modal transformation image at different concentration gradients in an embodiment of the present invention.

[0043] Figure 6 This is a comparison chart of the consistency between manual counting and the deep learning target detection counting algorithm in an embodiment of the present invention.

[0044] Figure 7 This figure shows the effect of the modal conversion model in an embodiment of the present invention on processing the superimposed diffraction interference phenomenon. The left side shows the effect of the modal transformation and deep learning object detection of the modal conversion model when dealing with the superimposed diffraction of two and three polystyrene microspheres. The right side shows the transformation effect of the modal conversion model when dealing with the superimposed diffraction interference of five polystyrene microspheres.

[0045] Figure 8 It is the gradient signal response standard curve and linear range of the unpaired data set modal transformation immunosensor based on holographic imaging for detecting chloramphenicol (CAP) in an embodiment of the present invention.

[0046] Fig. 9 This is the recovery rate of spiked CAP at different concentrations determined in the examples of the present invention.

[0047] Fig.10 This is a graph showing the experimental results of the specific response of the unpaired data set modal transformation immune sensor to different targets based on holographic imaging in an embodiment of the present invention.

[0048] Fig.11This is a comparison chart of the unpaired data set modal transformation immunoassay method based on holographic imaging and ELISA detection of CAP in real samples (urine, aquaculture wastewater and fish) using heat map distribution in an embodiment of the present invention. DETAILED DESCRIPTION

[0049] In order to make the objectives, technical solutions and advantages of the present invention more clear, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0050] Example 1

[0051] This embodiment provides an immune detection method based on holographic imaging modality conversion and deep learning, see Figure 1 , this embodiment is used to detect the concentration of chloramphenicol in the sample to be tested.

[0052] First, the sources of the main reagents used in this example are introduced:

[0053] Carboxyl functionalized polystyrene microspheres (PSM-COOH, 6 μm) were purchased from Bangs Laboratories, Inc (USA).

[0054] Carboxylated magnetic nanoparticles (MNPs 150 -COOH) was purchased from Ocean Nano-Tech (USA).

[0055] Azide (Azide-PEG4-NH2) and alkyne (Alkyze-PEG4-NH2) were purchased from Ruixi Biotechnology Co., Ltd. (Xi'an).

[0056] 1-Ethyl-3-(3-dimethylaminopropyl)-carbodiimide hydrochloride (EDC), N-hydroxysulfosuccinimide sodium salt (Sulfo-NHS), phosphate buffered saline (PBS) and 2-(N-morpholino)ethanesulfonic acid hydrate (MES) were purchased from Aladdin Company (Shanghai).

[0057] Bovine serum albumin (BSA) was purchased from Amresco (USA).

[0058] CuCl2·2H2O and polyvinylpyrrolidone (PVP) were purchased from Sinopharm Chemical Reagent Company (Shanghai).

[0059] Tetraethyl orthosilicate (TEOS), 3-aminopropyltriethoxysilane (APTES), and sodium ascorbate were purchased from Shanghai Aladdin Biochemical Technology Co., Ltd. (Shanghai).

[0060] Chloramphenicol (CAP), neomycin and clarithromycin were purchased from Sigma Aldrich (USA).

[0061] Chloramphenicol antibody (CAP-Ab) and BSA-chloramphenicol antigen (BSA-CAP-Ag) were purchased from Sangon Biotech Co., Ltd. (Shanghai).

[0062] The experimental water was deionized by a water purification system (Millipore, USA). All chemicals were of analytical grade and no further purification was required before use.

[0063] The preparation method of the relevant reagents used in this embodiment is:

[0064] PBS buffer (10 mM, pH = 7.4): Take 8.00 g NaCl, 0.20 g KCl, 0.20 g KH2PO4 and 2.90 g Na2HPO4·12H2O in a 1000 mL volumetric flask, make up to volume and shake well.

[0065] MES buffer (0.1M, pH=6.0): Dissolve 21.325 g of MES in deionized water to make up to 1000 mL to obtain solution A; dissolve 4 g of NaOH in deionized water to make up to 1000 mL to obtain solution B; mix 1000 mL of solution A and 400 mL of solution B and shake well.

[0066] PBST, MEST solution: Add 0.5 mL Tween-20 to 1000 mL of prepared PBS or MES buffer and shake well.

[0067] The immunoassay method of this embodiment mainly includes the following contents:

[0068] (1) Preparation of CuO2@SiO2 nanomaterials at room temperature

[0069] Dissolve CuCl2·2H2O (17 mg) and polyvinylpyrrolidone PVP (1 g) in 10 mL ultrapure water, add NH3·H2O (10 mL, 1 mol / L) and H2O2 (200 μL, 10 mol / L) dropwise, and stir for 30 minutes. Add tetraethyl orthosilicate (TEOS, 100 μL), stir the mixture for 2 hours, centrifuge (10000 r / min, 5 minutes) to collect the crude product, rinse with ultrapure water three times, and freeze-dry overnight. The obtained CuO2@SiO2 product is stored at 4°C in the dark for later use.

[0070] (2) Preparation of functionalized glass slides at room temperature

[0071] First, bovine serum albumin-azide (BSA-Azide) was coupled in advance. BSA (2 mg / mL, 5 mL) was coupled with Azide-PEG4-NH2 (10 mg / mL, 16 μL), dissolved in PBS (10 mM, pH = 7.4), incubated for 2 hours, and then the reaction was terminated with Tris-HCl (50 mM, pH = 8, 200 μL).

[0072] Secondly, functionalized glass slides were prepared. The ultrathin glass slides were soaked in 0.1M NaOH for 1 hour and rinsed three times with ultrapure water after treatment to activate the silanol groups on the surface of the ultrathin glass slides. The activated glass slides were then immersed in an anhydrous ethanol solution containing APTES (1%, v / v) for 12 hours to introduce amino groups into the activated glass slides. Subsequently, the glass slides were washed three times with ultrapure water and dried in a 37°C oven. The dried glass slides were incubated with GA aqueous solution (2.5%, v / v) for 2 hours, and the glass slides were rinsed three times with ultrapure water after the incubation. Finally, the pre-coupled BSA-Azide solution (4 mg / mL) was added to the surface of the glass slide and incubated for 2 hours. After completion, the glass slides were rinsed three times with ultrapure water, and the treated functionalized glass slides were immersed in a PBS solution at 4°C for storage.

[0073] (3) Preparation of complexes required for competitive immune response

[0074] Synthesis of CuO2@SiO2-CAP_Ab. CuO2@SiO2 (2 mg) was soaked in an anhydrous ethanol solution containing APTES (1%, 500 uL, v / v) for 12 hours to introduce amino groups on the surface of the material. Subsequently, it was washed three times with ultrapure water and centrifuged (10,000 rpm, 5 minutes). The dried CuO2@SiO2 was added to GA solution (500 uL, 2.5%, v / v) and incubated for 2 hours, and then washed three times with ultrapure water by centrifugation. Finally, 20 μg CAP-Ab was dropped and incubated for 2 hours, and then washed three times with ultrapure water by centrifugation. The unbound sites were blocked with BSA (1%, 500 uL) for 30 minutes, then washed three times with PBST by centrifugation and stored at 4°C.

[0075] Synthesize magnetic nanoparticles-chloramphenicol complete antigen (MNP-CAP_BSA). 150-COOH (1 mg), washed twice with MEST, and the supernatant was removed after magnetic separation. EDC (5 mg / mL, 50 μL) and NHS (5 mg / mL, 25 μL) were added respectively, activated at room temperature for 30 minutes, and washed twice with MEST after activating the carboxyl group. CAP_BSA (100 μg) was added to the carboxyl-activated MNPs and coupled at 37°C for 3 hours. The supernatant was removed after magnetic separation. Finally, BSA solution (1%, 1000 μL) was added to the above system to block unbound sites and reacted at 37°C for 30 minutes. After rinsing with PBST 4 times, MNP-CAP_BSA was resuspended with PBST (1000 μL pH=7.4, 0.5% BSA) and stored at 4°C.

[0076] (4) Chloramphenicol detection

[0077] MNP-CAP_BSA (20μg / mL, 100μL), CAP (100μL), and CuO2@SiO2-CAP_Ab (20μg / mL, 100μL) were mixed, and then a competitive immune reaction was carried out at 37°C for 25 minutes. The magnetic separation separated the complex of magnetic particles and CuO2@SiO2 immune binding, and then the supernatant was taken. In the competitive immune reaction, there was a quantitative relationship between the CAP concentration and the content of the remaining CuO2@SiO2 complex. Hydrochloric acid (8mM, 100μL) was added to the supernatant to release copper ions from the CuO2@SiO2-CAP_Ab after the magnetic separation reaction. The content of copper ions and the content of CuO2@SiO2 complex were directly quantitatively related, and an indirect quantitative relationship between the copper ion content and the CAP concentration was constructed. Then sodium ascorbate (SA) (5mM, 10μL) was added to the Cu 2+ Reduction to Cu + Finally, the coupled Alkyze-PS 6μm (50 μg / mL, 100 μL) and the converted Cu + When added to the functionalized glass slide, the content of copper ions is quantitatively related to the polystyrene microspheres involved in the click reaction, thus constructing a reaction system of CAP concentration and polystyrene microsphere signal method. After 10 minutes of blowing and washing reaction, rinse the glass slide with a slow stream of water, and place the cleaned ultra-thin glass slide into the self-developed holographic imaging system for imaging.

[0078] (5) Construction of modal conversion unpaired dataset and deep learning target detection dataset

[0079] The unpaired modal transfer dataset consists of a microscope dataset and a holographic imaging dataset. The microscope dataset is constructed by gradient dilution of 6μm polystyrene microspheres, imaging them at 40x magnification with an ordinary optical microscope, using a random sampling method, and then cropping the microscope imaging images into 512*512 pixel images, and performing data augmentation for model training. The microscope dataset contains 1,200 images in total.

[0080] Holographic imaging data set: Polystyrene microspheres with the same concentration gradient as mentioned above were imaged directly on a self-developed lensless holographic imaging device, and the holographic images were subjected to non-local mean filtering and cropping. At the same time, a random sampling method was used to obtain a data set of 400 holographic images, thereby establishing an unpaired data set for modal transformation.

[0081] In order to ensure the stability and accuracy of microsphere counting quantification, the target detection dataset consists of images after modal transformation and microscopic imaging datasets. The microscopic modal images generated after modal transformation and the microscopic imaging images of equal gradient dilution were annotated, and 2824 target detection datasets were obtained by combining the data enhancement algorithm, in which the ratio of training set to validation set was 8:2. The target detection dataset was used to train the polystyrene microsphere recognition and counting model.

[0082] (6) Unpaired dataset modality conversion algorithm and deep learning target detection algorithm construction

[0083] The modality conversion model uses CycleGAN as the backbone model network for the transformation from the holographic domain to the microscopic image domain. Different from the traditional supervised modality generative adversarial network (GAN) which has only one generator and one discriminator and requires one-to-one mapping, the modality conversion model achieves image transformation between two different domains (holographic image domain and microscopic image domain) by learning the mapping relationship between the two different domains, and maintains the consistency of the transformation through inverse mapping. This model does not rely on paired datasets, but uses cycle consistency loss to achieve image transformation of unpaired datasets.

[0084] The modal conversion model consists of two mapping functions: generator G: X (holographic image domain) → Y (microscopic image domain), generator F: Y (microscopic image domain) → X (holographic image domain), and discriminator D X , D Y , adversarial losses L1 and L2 are used to control the differences in generated images within the domain, and cycle consistency loss Loss CycleUsed to control the difference between domains. The model discriminator is a 70×70 PatchGAN network, the generator is a Resnet_9blocks network, and the Unaligned dataset mode is used. Due to hardware device limitations, Batch_size is set to 1, the number of training epochs is set to 200, and the number of epochs with linear decay of the learning rate is 100. During the training process, the initial learning rate of Adam is set to 0.0002, after which the learning rate decays linearly, the momentum term of Adam is set to 0.5, the training set image is set to 800×800, and grayscale images are used for training.

[0085] The target count quantification model uses YOLOV7 as the backbone model network, and performs modal transformation on the image to quantify the polystyrene microsphere count. The model has a multi-scale detection strategy and good generalization ability. The model first extracts specific feature information through the forward propagation network (Conv), then fuses feature information of different scales through the extended efficient layer aggregation network (E-ELAN), and then further extracts features and optimizes the network structure through the spatial pyramid pooling-cross-stage partial network (SPPCSP) to maintain network performance. The prediction box is generated by the detection head and the internal parameters are adjusted through the back-propagation algorithm to improve the model performance. Finally, the non-maximum suppression algorithm (NMS) is used to process overlapping predictions to retain the best prediction results. The model pre-training weight is yolov7-e6, the training image size is 640×640, the initial learning rate is 0.01, Optimizer = SGD, Optimizer weight decay is 0.0005, SGD is 0.937, Epoch is 300, Batch_size is 32, and multi-scale training is adopted, in which MP and Upsample are used for downsampling and upsampling respectively, and CBS is used for image convolution processing. In order to adapt to the statistics of the number of polystyrene microspheres under the large field of view of holographic imaging, the maximum number of detection frames retained by max-det, that is, the maximum number of object detections, is set to 10,000.

[0086] This embodiment is based on Python 3.8, Pytorch 3.0.1+cu11.8, Numpy, Opencv, Pandas, Visdom, piq and other algorithm packages. The hardware used for model training is: GPU NVIDIA GeForce RTX 3090Ti, CPU i9-12900K.

[0087] Example 2: Detecting the content of chloramphenicol (CAP) in aquatic product samples.

[0088] The aquatic product samples were purchased randomly from supermarkets.

[0089] Pretreatment: Use a grinder to grind the aquatic product samples into minced meat. Add CAP solution of unknown concentration to the minced meat randomly, mix well and let it stand overnight at 4°C. Ultrasonic extraction of CAP in minced meat was performed with a mixture of ethyl acetate / acetonitrile (1:1, v / v), repeated twice, 30 minutes each time. The supernatant was dried with nitrogen and resuspended in PBS solution for subsequent detection.

[0090] The immunoassay method described in Example 1 was used to detect the chloramphenicol (CAP) content in the pretreated aquatic product samples. The test results are as follows: Fig.11 As shown, among the 10 aquatic product samples, CAP was detected in three samples, and the same results were detected in parallel samples, which was the same as the gold standard method ELISA method in detecting CAP sample size and the samples were consistent.

[0091] Example 3: Detection of chloramphenicol (CAP) content in aquaculture wastewater samples

[0092] Aquaculture wastewater samples were randomly collected from a certain breeding area.

[0093] Pretreatment: Use a 0.22 μm organic filter membrane to filter out impurities, and randomly add CAP solution of unknown concentration for subsequent testing.

[0094] The immunoassay method described in Example 1 was used to detect the chloramphenicol (CAP) content in the pretreated aquaculture wastewater sample. The test results are as follows: Fig.11 As shown, among the 10 aquaculture wastewater samples, CAP was detected in three samples, and the same results were detected in parallel samples, which was the same as the gold standard method ELISA method in detecting CAP sample size and the samples were consistent.

[0095] Example 4: Detection of chloramphenicol (CAP) content in urine

[0096] Urine samples were collected from 10 healthy subjects.

[0097] Pretreatment: CAP solution of unknown concentration was randomly added to urine samples for subsequent testing.

[0098] The chloramphenicol (CAP) content of the pretreated urine sample was detected by the immunoassay method described in Example 1. The test results are as follows: Fig.11 As shown in the figure, among the 10 urine samples, CAP was detected in three samples, and the same results were detected in parallel samples, while the gold standard method ELISA method did not detect a low concentration of CAP, indicating that the detection accuracy of this method is higher.

[0099] Some steps in the embodiments of the present invention may be implemented using software, and the corresponding software program may be stored in a readable storage medium, such as a CD or a hard disk.

[0100] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. An immunoassay method, characterized in that: The method comprises: Step 1: Mix the target object to be detected, the pre-prepared CuO2@SiO2 complex and the nanomagnetic particle conjugate to perform a competitive immune reaction; the surface of the CuO2@SiO2 complex is modified with a detection antibody that recognizes the target object to be detected, and the surface of the nanomagnetic particle conjugate is modified with a complete antigen of the target object to be detected; Step 2: magnetic separation, dissolving the remaining CuO2@SiO2 composite after reaction in hydrochloric acid and sodium ascorbate to release copper ions; Step 3: Copper ions act as catalysts to rapidly catalyze the azide-alkyne cycloaddition reaction at room temperature, so that an azide-alkyne cycloaddition reaction occurs between the alkyne-modified polystyrene microspheres and the azide ultrathin functionalized glass slide, i.e., a click reaction; Step 4: performing holographic imaging on the polystyrene microspheres obtained in step 3 and fixed on the functionalized glass slide through the click reaction; Step 5: construct an unpaired modal conversion data set using the holographic imaging image of the isogradient diluted polystyrene microspheres and the microscopic imaging image of the isogradient diluted polystyrene microspheres, and train them using a deep learning modal conversion model, and then use the trained deep learning modal conversion model to perform modal conversion on the holographic imaging image obtained in the experiment of step 4 to obtain a microscopic modal image; Step 6: Input the microscopic modal image obtained after the modal conversion and the microscopic imaging image of the isogradient diluted polystyrene microspheres into a deep learning target detection network for training, and then use the trained deep learning target detection network to identify and count the microscopic modal images that have undergone modal conversion in the experiment of step 4; Step 7: Using the linear relationship between the number of the polystyrene microspheres and the concentration of the target to be detected, the concentration information of the target to be detected is obtained.

2. The immunoassay method according to claim 1, characterized in that: The deep learning modality transfer model uses the cycle generative adversarial network CycleGAN as the backbone network.

3. The immunoassay method according to claim 1, characterized in that: The deep learning target detection network uses YOLOV7 as the backbone network.

4. The immunoassay method according to claim 1, characterized in that: The copper ions are monovalent copper ions generated by the reduction reaction of sodium ascorbate.

5. The immunoassay method according to claim 1, characterized in that: The click reaction is that azide and alkyne react to form a covalent bond under the catalysis of monovalent copper ions.

6. The immunoassay method according to claim 1, characterized in that: The polystyrene microspheres are 6 μm polystyrene microspheres with alkynes modified on the surface.

7. The immunoassay method according to claim 1, characterized in that: The method also includes processing the holographic image using a non-local means filtering algorithm.

8. The immunoassay method according to claim 1, characterized in that: The nanomagnetic particle conjugate is: a 150nm nanomagnetic particle with the complete antigen of the target object to be detected modified on the surface.

9. The immunoassay method according to claim 1, characterized in that: The added concentration of the nanomagnetic particle conjugate in the reaction system is 20 μg / mL.

10. The immunoassay method according to claim 1, characterized in that: The CuO2@SiO2 complex is added to the reaction system at a concentration of 20 μg / mL.

Images