Immunoquantitative detection ai model training and testing method

Abstract

An AI model training and detection method for quantitative immunoassay involves adding detection microparticles to a sample to form a microscopic sample. The sample contains the component to be detected, and the surface of the detection microparticles has antibodies or antigens. A microscopic image of the sample is obtained by photographing the microscopic sample. The microscopic image is labeled with the concentration of the component to be detected, resulting in a labeled microscopic image. An AI model is trained using the labeled microscopic image to obtain an AI model feature dataset or a trained AI model. The AI ​​model feature dataset, combined with an AI computational model, can identify the concentration of the component to be detected in the sample; or the trained AI model can identify the concentration of the component to be detected in the sample. The AI ​​model feature dataset obtained by the above method, combined with the AI ​​computational model or the trained AI model, identifies the concentration of the component to be detected from the microscopic image.

Description

Technical Field

[0001] This application belongs to the field of in vitro diagnostic immunology technology, and specifically relates to a quantitative immunoassay and AI training method based on the detection of microparticle aggregation in microscopic images. Background Technology

[0002] In vitro diagnostic immunology is a method for diagnosing diseases by detecting specific immune markers in bodily fluids. These immune markers can be antibodies, antigens, immunoglobulins, etc. Detecting their presence or changes in levels can help doctors diagnose diseases, monitor treatment effectiveness, or predict disease progression.

[0003] There are several different technical approaches for the quantitative and qualitative analysis of trace amounts of immunomarkers. Established techniques include enzyme-linked immunosorbent assay (ELISA), radioimmunoassay, immunoturbidimetric assay, immunofluorescence assay, and chemiluminescent immunoassay.

[0004] Chinese patent application number "CN201410197209.1" entitled "A Method for Detecting Biological Macromolecules or Microorganisms" proposes a method for detecting biological macromolecules or microorganisms. This patent integrates immunomagnetic enrichment with visual detection, offering a simple operation. However, this patent reveals that "the degree of aggregation of the aggregates is positively correlated with the concentration of biological macromolecules or microorganisms," and that "the concentration of biological macromolecules or microorganisms in the sample to be tested is determined based on the degree of aggregation of the aggregates, specifically by acquiring a photograph of the aggregates and quantitatively determining the concentration of biological macromolecules or microorganisms in the sample to be tested by analyzing the grayscale values ​​of the photographs."

[0005] In existing technologies, establishing fitting formulas based on the aggregation degree and morphology of aggregated particles requires a large amount of experimental data and complex mathematical modeling. Extracting features to establish the mathematical relationship between the image and the component to be detected requires extensive experimentation and modeling analysis, and the industry has yet to find a satisfactory solution. Therefore, technologies for obtaining the concentration of the analyte based on image analysis of aggregated particles are difficult to commercialize. Even fitting formulas obtained through extensive experimental data often fail to achieve high accuracy.

[0006] Technical Name: Antibody: A protective protein produced by the body in response to antigen stimulation. (Immunoglobulins are not limited to antibodies.) It is a large, Y-shaped protein secreted by plasma cells (effector B cells) and used by the immune system to identify and neutralize foreign substances such as bacteria and viruses. It is found only in the blood and other bodily fluids of vertebrates, and on the cell membrane surface of their B cells. Antibodies recognize a unique characteristic of a specific foreign substance; this foreign target is called an antigen. Summary of the Invention

[0007] This application, validated through extensive experimentation, demonstrates that AI models can be directly trained using images with calibrated concentrations, enabling the AI ​​to directly identify the concentration of target substances in test samples. It avoids the complex derivation of fitting formulas, replacing mathematical modeling with AI computing power. Extensive experimental verification shows that this method achieves the required accuracy for medical testing. It has broad prospects for industrial application and represents a significant breakthrough in the field of immunoassay detection technology.

[0008] The method for solving the above-mentioned technical problems in this application is an AI model training method for quantitative immunoassay detection, used to obtain a trained AI model or an AI model feature dataset; adding detection microparticles to a sample to form a microscopic sample, the sample including the component to be detected, the detection microparticles having antibodies or antigens on their surface; taking microscopic images of the microscopic sample; labeling the microscopic image with the concentration of the component to be detected to obtain a labeled microscopic image; training an AI model with the labeled microscopic image to obtain an AI model feature dataset or a trained AI model; the AI ​​model feature dataset, in conjunction with an AI computational model, can identify the concentration of the component to be detected in the sample; or the trained AI model can identify the concentration of the component to be detected in the sample.

[0009] The microscopic images are multiple microscopic images taken of the same sample. After each microscopic image is labeled, an AI model is trained.

[0010] The microscopic sample is placed into a detection chamber at the same or different heights, and a microscopic image of the sample in the detection chamber is obtained by taking a picture.

[0011] The microscopic examination sample is an immune sample of known concentration with added detection particles; the concentrations of different concentrations of immune samples in a set of training samples are set according to a step concentration.

[0012] Immunization samples of different concentrations include training sample group A, which includes samples with a 1x unit concentration and samples with a 2x unit concentration.

[0013] Immunization samples of different concentrations include training sample group B, which includes samples with a 1x unit concentration and samples with a 10x unit concentration.

[0014] The immune samples of different concentrations include training sample group C, which includes two or more training sample subgroups, including training sample group C1 and training sample group C2. Training sample group C1 includes multiple training samples of different concentrations in magnitude K1. Training sample group C2 includes multiple training samples of different concentrations in magnitude K2, where magnitude K2 is greater than magnitude K1.

[0015] In the training method for an AI model for quantitative immunoassay, the sample may be serum, and the detected components may include any one or more of proteins, lipids, polysaccharides, small molecule hormones, bacteria, and fungi. Alternatively, the sample may be a diluted whole blood solution, and the detected components may include any one or more of proteins, lipids, polysaccharides, small molecule hormones, and cells. Or, the sample may be urine, and the detected components may include any one or more of proteins, lipids, polysaccharides, small molecule hormones, cells, and bacteria. Finally, the sample may be feces, and the detected components may include any one or more of proteins, protozoa, eggs, cells, and microorganisms.

[0016] An immunoassay quantitative detection method includes: adding detection microparticles to a sample to form a microscopic sample, the sample containing the component to be detected, and the detection microparticles having antibodies or antigens on their surface; taking a microscopic image of the microscopic sample; using an AI model feature dataset obtained by the above method in conjunction with an AI computational model to identify the concentration of the component to be detected from the microscopic image; or using a trained AI model obtained by the above method to identify the concentration of the component to be detected from the microscopic image.

[0017] The microscopic images include microscopic image A1 and microscopic image A2; the concentration W1 of the component to be detected is obtained by identifying microscopic image A1; the concentration W2 of the component to be detected is obtained by identifying microscopic image A1; the concentration of the component to be detected is equal to the average value of the concentrations W1 and W2 of the component to be detected.

[0018] In the immunoassay method, the sample may be serum, and the detected components may include any one or more of proteins, lipids, polysaccharides, small molecule hormones, bacteria, and fungi.

[0019] The sample may be a whole blood dilution, and the detected components may include any one or more of proteins, lipids, polysaccharides, small molecule hormones, and cells.

[0020] The sample may be urine, and the detected components may include any one or more of the following: proteins, lipids, polysaccharides, small molecule hormones, cells, and bacteria.

[0021] The sample may be feces, and the detected components may include any one or more of the following: protein, protozoa, eggs, cells, and microorganisms.

[0022] A detection device or computing processing device is used to run all or part of the above-described methods; the memory of the detection device or computing processing device includes all or part of the data of all or part of the above-described methods.

[0023] A data storage device for storing all or part of the program code for performing the above methods; or storing all or part of the data for performing the above methods.

[0024] The technical effects of the above-mentioned solution include: Extensive experimental verification has shown that AI models can be directly trained using images with calibrated concentrations, enabling the AI ​​to directly identify the concentration of target substances in test samples. This avoids the complex derivation process of fitting formulas, replacing mathematical modeling with AI computing power. Extensive experimental verification has demonstrated that this method achieves the accuracy required for medical testing. It has broad prospects for industrial application and represents a significant breakthrough in the field of immunoassay detection technology.

[0025] The technical effects of the above-mentioned technical solution include: the AI ​​model feature dataset, combined with the AI ​​calculation model or the trained AI model, can identify the concentration of the component to be detected corresponding to the image and directly establish the correlation between the image and the concentration of the component to be detected; it greatly simplifies the intermediate image processing, recognition and intermediate calculation process; and greatly improves the detection efficiency.

[0026] The technical advantages of the above-mentioned technical solution include: compared with traditional optical absorbance detection, it greatly reduces costs, eliminates the need for a sophisticated photoelectric detection system, and only requires microscopic imaging to complete the immunoassay, which is a major breakthrough in this field.

[0027] The technical effects of the above-mentioned technical solution include: training the AI ​​model with multiple microscopic images of the same sample, that is, using multiple images of samples with the same concentration to repeatedly train the AI, so that the AI ​​can extract the relationship between concentration and key graphic elements, which can improve the efficiency and accuracy of recognition.

[0028] The technical advantages of the above-mentioned solution include: Existing quantitative analysis techniques require calculating the sample volume based on the image area and the height of the sample-containing cavity, and the accuracy of the sample volume significantly impacts the accuracy of the detection parameters. In this application, microscopic images of samples examined in detection cavities at different heights are obtained and used for AI training. This enables the AI ​​to recognize the concentration of the sample at different cavity heights, reducing the precision requirements of the sample-containing device and improving the accuracy and robustness of the detection.

[0029] The technical effects of the above-mentioned technical solution include: adding detection particles to immune samples of known concentration; and generating sufficient characteristic images by setting the concentrations of multiple groups of immune samples of different concentrations according to a stepped concentration.

[0030] The technical effects of the above-mentioned technical solution include: training sample group A includes samples with 1x unit concentration and samples with 2x unit concentration, and the sample concentration intervals are set relatively close, which can be used in occasions where high concentration resolution is required.

[0031] The technical effects of the above-mentioned technical solution include: training sample group B includes samples with 1x unit concentration and samples with 10x unit concentration, and the sample concentration intervals are set relatively close, which can be used in occasions with a large concentration measurement range.

[0032] The technical effects of the above-mentioned technical solution include: training sample group C1 and training sample group C2; training sample group C1 includes multiple training samples of different concentrations in order of magnitude K1; training sample group C2 includes multiple training samples of different concentrations in order of magnitude K2, where order of magnitude K2 is greater than order of magnitude K1; training the two groups of samples simultaneously can improve both the range and accuracy of concentration measurement.

[0033] The technical effects of the above-mentioned technical solution include: the concentration of the component to be detected can be obtained directly based on the microscopic image of the sample and the training data set or model with concentration labels, which is an efficient method for detecting the concentration of the component to be detected and opens up a completely new path.

[0034] The technical effects of the above technical solution include: the concentration of the component to be detected is equal to the average value of the concentrations W1 and W2 of the component to be detected, and the average calculation reduces the measurement error caused by the photographic deviation.

[0035] The technical advantages of the above-mentioned technical solution include: it can handle a variety of samples, has good adaptability, and can be used for almost all humor-related immune tests. Attached Figure Description

[0036] Figure 1 This is a flowchart illustrating the AI ​​training method for quantitative immunoassay. Figure 1 ;

[0037] Figure 2 This is a flowchart illustrating the AI ​​training method for quantitative immunoassay. Figure 2 ;

[0038] Figure 3 This is a flowchart illustrating the AI ​​training method for quantitative immunoassay. Figure 3 ;

[0039] Figure 4 This is a flowchart illustrating the quantitative immunoassay method.

[0040] Figure 5 These are microscopic images obtained from an immunoassay sample at a 1x concentration.

[0041] Figure 6 These are microscopic images obtained from immunoassay samples at twice the concentration.

[0042] Figure 7 These are microscopic images obtained from immunoassay samples at a 3-fold concentration.

[0043] Figure 8 These are microscopic images obtained from immunized samples at a 4-fold concentration;

[0044] Figure 9 This is a schematic diagram of training sample group A used for AI training;

[0045] Figure 10 This is a schematic diagram of training sample group B used for AI training;

[0046] Figure 11 This is a schematic diagram of multiple training sample groups used in AI training. Detailed Implementation

[0047] The contents of this application will be further described in detail below with reference to the accompanying drawings. It should be noted that the following description is of preferred embodiments of the present invention and does not constitute any limitation on the present invention. The description of the preferred embodiments of the present invention is merely an explanation of the general principles of the invention. The designations "first," "second," "A," and "B" used in this invention are for ease of explanation only and do not represent a temporal or spatial order. The combinations of letters and numbers "TA," "TB," and "H" used in this invention are for ease of explanation only, and their specific meanings are determined by the specific terms they represent.

[0048] In existing technologies, most immune substance detection is based on optical absorbance detection, which requires the setup of a sophisticated photoelectric detection system, resulting in high design, maintenance, and usage costs.

[0049] Image recognition technology based on aggregated particles is used to detect antigens and antibodies by using the aggregation phenomenon of micron-level detection particles to quantitatively measure the content of target substances at the molecular level.

[0050] After years of research and significant investment of human and material resources, the applicant has developed a detection method that directly uses AI to identify images of aggregated microparticles. Based on the number and area of ​​individual detection particles and aggregated microparticles identified by microscopic photography, the concentration of the target analyte is calculated using empirical formulas. This method is also used in the applicant's self-developed formed element analyzer. However, because it requires the identification of individual detection particles and aggregated microparticles, it is computationally intensive.

[0051] The method in this application involves adding detection particles to a sample to form a microscopic sample. The sample includes the component to be detected, and the surface of the detection particles has antibodies or antigens. It is also based on the changes in the microscopic image caused by the antigen-antibody binding characteristics in the sample. However, instead of using AI to identify individual detection particles and particle aggregates or clusters in various localities, the AI ​​is trained by directly identifying the concentration of the target substance. This allows the AI ​​to be trained to directly obtain the content of the component to be detected from the image. The AI ​​computing power is used to avoid the calculation of the fitting formula, enabling the industrial application of the technology of obtaining the concentration of the target substance based on the image analysis of clustered particles. This is a major breakthrough in the field of immunoassay technology.

[0052] The solution presented in this application, which enables immunoassay as long as microscopic imaging is achieved, represents a significant breakthrough in the field. Compared to methods that rely on identifying individual detectable microparticles and aggregated microparticles through microscopic imaging and then establishing fitting formulas, this approach directly uses the concentration of the component to be detected to label the microscopic image, fully leveraging the advantages of AI.

[0053] like Figure 1 An AI model training method for quantitative immunoassay is disclosed, comprising: obtaining a trained AI model or an AI model feature dataset; adding detection microparticles to a sample to form a microscopic sample, the sample containing the component to be detected, the detection microparticles having antibodies or antigens on their surface; capturing a microscopic image of the microscopic sample; labeling the microscopic image with the concentration of the component to be detected to obtain a labeled microscopic image; training the AI ​​model with the labeled microscopic image to obtain an AI model feature dataset or a trained AI model; the AI ​​model feature dataset, in conjunction with an AI computational model, is capable of identifying the concentration of the component to be detected in the sample; or the trained AI model is capable of identifying the concentration of the component to be detected in the sample.

[0054] In some embodiments, the microscopic images are multiple microscopic images taken of the same sample, and each microscopic image is labeled before AI model training. Figure 5 These are microscopic images obtained from an immunoassay sample at a 1x concentration. Figure 6 These are microscopic images obtained from immunoassay samples at twice the concentration. Figure 7 These are microscopic images obtained from immunoassay samples at a 3-fold concentration. Figure 8 These are microscopic images obtained from immune samples at 4 times the concentration; both can be used for AI model training.

[0055] Typically, the microscopic sample is placed into a detection chamber of the same height, and a microscopic image of the sample inside the chamber is obtained. Alternatively, as... Figure 3 The microscopic sample is placed into a detection chamber at different heights, and microscopic images of the sample in the detection chamber are obtained by taking pictures.

[0056] like Figure 2 The microscopic examination sample is an immune sample of known concentration with added detection particles; the concentrations of a set of immune samples of different concentrations are set according to a step concentration.

[0057] like Figure 9 The immune samples of different concentrations include training sample group A, which includes samples with a concentration of 1 unit, 2 units, 3 units, and 4 units.

[0058] like Figure 10 The immune samples of different concentrations include training sample group B, which includes samples with a concentration of 1x, 10x, 100x, and 1000x.

[0059] like Figure 11 The immune samples of different concentrations include training sample group C, which includes two or more training sample subgroups, namely training sample group C1, training sample group C2, training sample group C3, and training sample group C4.

[0060] The training sample group C1 includes multiple training samples of different concentrations within the order of magnitude K1, namely, samples with a 1-unit concentration and samples with a 5-unit concentration. There can be more than one different concentration.

[0061] The training sample group C2 includes multiple training samples of different concentrations within the order of magnitude K2, namely samples with a concentration 10 times and samples with a concentration 50 times. The order of magnitude K2 is greater than the order of magnitude K1.

[0062] The training sample group C3 includes multiple training samples of different concentrations within the order of magnitude K3, namely samples with a concentration of 100 times and samples with a concentration of 500 times. The order of magnitude K3 is greater than that of order of magnitude K2.

[0063] The training sample group C4 includes multiple training samples of different concentrations within the order of magnitude K4, namely samples with a concentration of 1000 times and samples with a concentration of 5000 times. The order of magnitude K4 is greater than that of order of magnitude K3.

[0064] like Figure 4 An immunoassay quantitative detection method includes: adding detection microparticles to a sample to form a microscopic sample, the sample containing the component to be detected, and the detection microparticles having antibodies or antigens on their surface; taking a microscopic image of the microscopic sample; using an AI model feature dataset obtained by the above method in conjunction with an AI computational model to identify the concentration of the component to be detected in the microscopic image; or using a trained AI model obtained by the above method to identify the concentration of the component to be detected in the microscopic image.

[0065] The microscopic images include microscopic image A1 and microscopic image A2; the concentration W1 of the component to be detected is obtained by identifying microscopic image A1; the concentration W2 of the component to be detected is obtained by identifying microscopic image A1; the concentration of the component to be detected is equal to the average value of the concentrations W1 and W2 of the component to be detected.

[0066] In some embodiments, the sample is serum, and the detected components include any one or more of proteins, lipids, polysaccharides, small molecule hormones, bacteria, and fungi.

[0067] In some embodiments, the sample is a whole blood dilution, and the detected components include any one or more of proteins, lipids, polysaccharides, small molecule hormones, and cells.

[0068] In some embodiments, the sample is urine, and the detected components include any one or more of proteins, lipids, polysaccharides, small molecule hormones, cells, and bacteria.

[0069] In some embodiments, the sample is feces, and the detected components include any one or more of proteins, protozoa, eggs, cells, and microorganisms.

[0070] A detection device or computing processing device is used to run all or part of the above-described methods; the memory of the detection device or computing processing device includes all or part of the data of all or part of the above-described methods.

[0071] A data storage device for storing all or part of the program code for performing the above methods; or for storing all or part of the data for performing the above methods.

[0072] While the present invention has been described and illustrated with reference to preferred embodiments and several alternatives, the invention is not limited to the specific descriptions herein. Other alternatives or equivalent components may also be used to practice the invention.

Claims

1. A method for training an AI model for quantitative immunoassay, wherein the quantitative immunoassay involves adding detection microparticles to a sample to form a microscopic sample, the sample containing the component to be detected, and the detection microparticles having antibodies or antigens on their surface; the method is based on the changes in microscopic images caused by the antigen-antibody binding characteristics in the sample, and the microscopic images include detection microparticles and microparticle aggregates; Image recognition technology based on aggregated particles is used to detect antigens and antibodies by using the aggregation phenomenon of micron-level detection particles to quantitatively measure the content of target substances at the molecular level. Its features are: Used to obtain a trained AI model or a dataset of AI model features; A microscopic sample is formed by adding detection microparticles to a sample. The sample contains the component to be detected, and the detection microparticles have antibodies or antigens on their surface. The microscopic sample is an immune sample with the detection microparticles added at a known concentration. The microscopic sample is placed into a detection chamber at the same or different heights, and a microscopic image of the sample in the detection chamber is obtained by photographing it; the microscopic image is labeled with the concentration of the component to be detected to obtain a labeled microscopic image. The AI ​​model is trained using the labeled microscopic images to obtain an AI model feature dataset or a trained AI model. The AI ​​model feature dataset, combined with the AI ​​computing model, can identify the concentration of the component to be detected in the sample. Or the trained AI model can identify the concentration of the component to be detected in the sample to be detected; Instead of using AI to identify individual particles and particle aggregates or clusters in various localities, the AI ​​is trained to directly identify the concentration of the target substance. The concentrations of different concentrations of immune samples in a set of training samples were set according to a stepwise concentration pattern; Includes one or more of the following features: Feature TB10: The sample is serum, and the detected components include any one or more of proteins, lipids, polysaccharides, small molecule hormones, bacteria, and fungi; Feature TB20: The sample is a whole blood dilution, and the detected components include any one or more of proteins, lipids, polysaccharides, small molecule hormones, and cells; Feature TB30: The sample is urine, and the detected components include any one or more of the following: proteins, lipids, polysaccharides, small molecule hormones, cells, and bacteria; Feature TB40: The sample is feces, and the detected components include any one or more of the following: protein, protozoa, eggs, cells, and microorganisms.

2. The method for training an AI model for quantitative immunoassay according to claim 1, characterized in that, The microscopic images are multiple microscopic images taken of the same sample. After each microscopic image is labeled, an AI model is trained.

3. The method for training an AI model for quantitative immunoassay according to claim 1, characterized in that, Immunization samples of different concentrations include training sample group A, which includes samples with a 1x unit concentration and samples with a 2x unit concentration.

4. The method for training an AI model for quantitative immunoassay according to claim 1, characterized in that, Immunization samples of different concentrations include training sample group B, which includes samples with a 1x unit concentration and samples with a 10x unit concentration.

5. The method for training an AI model for quantitative immunoassay according to claim 1, characterized in that, The immune samples of different concentrations include training sample group C, which includes two or more training sample subgroups, including training sample group C1 and training sample group C2. The training sample group C1 includes multiple training samples of different concentrations within the order of magnitude K1; The training sample group C2 includes multiple training samples of different concentrations in the order of magnitude K2, where the order of magnitude K2 is greater than that of magnitude K1.

6. An immunoassay method, comprising: The sample is prepared by adding detection particles to form a microscopic sample. The sample contains the components to be detected, and the detection particles have antibodies or antigens on their surface. Microscopic images of samples were obtained by photographing and examining them under a microscope. Using the AI ​​model feature dataset obtained by the method described in any one of claims 1 to 5, and in conjunction with the AI ​​computing model, the concentration of the component to be detected is obtained by recognizing a microscopic image. Alternatively, the trained AI model obtained by the method described in any one of claims 1 to 5 can be used to identify the concentration of the component to be detected in a microscopic image.

7. The immunoassay method according to claim 6, characterized in that, The microscopic images include microscopic image A1 and microscopic image A2; The concentration W1 of the component to be detected is obtained by identifying the microscopic image A1; the concentration W2 of the component to be detected is obtained by identifying the microscopic image A1. The concentration of the component to be detected is equal to the average of the concentrations W1 and W2 of the component to be detected.

8. A detection apparatus for operating the method according to any one of claims 1 to 7; The detection device includes data from the methods described above.

9. A data storage device for storing program code for performing the method according to any one of claims 1 to 7; or storing data according to the method according to any one of claims 1 to 7.

Images