EBV antibody titer evaluation method and device, medium and storage equipment

Abstract

The invention discloses an EBV antibody titer evaluation method and device, a medium and storage equipment. According to the application, the cell images of the EBV antigen under different preset exposure times are acquired, the deep learning model is utilized to predict and mask the positive cell areas of the images, and then the mathematical model is utilized to convert the mask images to adapt to the second preset exposure condition; and finally, the EBV antibody titer estimation value of each cell image is calculated based on the statistical model and the brightness distribution of the positive cell region. According to the application, the efficiency and accuracy of titer evaluation are improved, so that the problem that the titer of the EBV antibody cannot be accurately and efficiently evaluated in the prior art is solved.

Description

Technical Field

[0001] The present invention relates to the field of evaluation of EBV antibody titers, and particularly to a method, device, medium and storage device for evaluating EBV antibody titers. Background Art

[0002] The screening of NPC mainly relies on three techniques: serological detection, EBV DNA detection, and imaging examination. Serological detection is particularly common in large-scale screening due to its simple operation and low cost. Indirect immunofluorescence assay (IFA) is one type of serological detection, which plays an important role in the early diagnosis of NPC by detecting EBV antibodies EA-IgA and VCA-IgA in serum. These antibodies are key indicators for NPC screening, with high sensitivity and specificity, and their titers are usually related to the risk of NPC.

[0003] The IFA technique includes two steps: slide preparation and interpretation. First, Raji or B958 cells are cultured to express EBV antigens and fixed on a glass slide. Then, the serum sample to be tested is incubated with these cells to allow the antibodies therein to bind to the EBV antigens. A fluorescently labeled secondary antibody is further used to bind to the antibodies, and finally, the fluorescence intensity and pattern of the cells are observed through a fluorescence microscope. Professional personnel classify the samples into different titer grades based on these characteristics.

[0004] However, the IFA technique has some limitations. It relies on the experience and skills of professional personnel, and the interpretation process is highly subjective, with possible differences in interpretation among different evaluators. In addition, the slide preparation and interpretation processes of IFA are both cumbersome and time-consuming, not suitable for quickly processing a large number of samples, which limits its application in large-scale screening. The IFA technique also requires the preparation of multiple sample dilutions to determine the final titer result, which not only increases the workload but also may lead to variability in the results. In terms of quantification, the IFA technique usually can only provide discrete titer grades and cannot provide continuous quantitative results, which limits its application in accurately evaluating the EBV infection status and NPC risk. Although there are automatic estimation algorithms that can reduce the subjectivity of negative / positive discrimination, these algorithms also require the preparation of samples at multiple dilutions and cannot effectively reduce the cost of sample preparation. These problems result in the inability of the prior art to accurately and efficiently evaluate EBV antibody titers. Summary of the Invention

[0005] The present invention provides a method, device, medium and storage device for evaluating EBV antibody titers to solve the problem that the prior art cannot accurately and efficiently evaluate EBV antibody titers.

[0006] In a first aspect, the present application provides a method for evaluating EBV antibody titers, including:

[0007] Obtain each cell image of EBV antigen under the first different preset exposure times;

[0008] Input each of the cell images into a pre-trained deep learning model for prediction, so that the deep learning model outputs each first positive cell region mask image;

[0009] Process each of the first positive cell region mask images according to a mathematical model to obtain each second positive cell region mask image under the second preset exposure condition;

[0010] Calculate the EBV antibody titer estimated value of each cell image according to a preset statistical model and the brightness distribution of each positive cell region in each of the second positive cell region mask images.

[0011] This application first ensures the comprehensiveness of data by obtaining EBV antigen cell images under different exposure times. This application uses a deep learning model to analyze these images. This model can automatically identify and predict positive cell regions, which not only improves efficiency but also reduces human error. Then, this application applies a mathematical model to standardize the brightness values of positive cell regions, eliminating the influence brought by exposure time differences, so that the brightness values are more accurate and comparable. Finally, by statistically analyzing the brightness distribution of positive cell regions, the estimated value of EBV antibody titer is calculated. This application improves the efficiency and accuracy of titer evaluation to solve the problem that the prior art cannot accurately and efficiently evaluate the EBV antibody titer.

[0012] As a preferred embodiment of the first aspect, the obtaining each cell image of EBV antigen under the first different preset exposure times is specifically:

[0013] Obtain an initial cell image dataset; wherein, the initial cell image dataset is obtained by a camera shooting under the first different preset exposure times;

[0014] Input the initial cell image dataset into a pre-trained UNet network, so that the UNet network outputs each cell image of EBV antigen.

[0015] In this preferred embodiment, the present application captures cell images of EBV antigens at different preset exposure times, enabling comprehensive capture of the details of EBV antigen expression. Different exposure times can reveal antigen-antibody reactions of different intensities, providing rich feature information for the deep learning model. By inputting these initial cell image datasets into a pre-trained UNet network, its advanced feature extraction ability can be utilized to automatically identify and distinguish EBV antigen-positive cell regions. This automated image recognition process not only improves the accuracy and efficiency of detecting positive cell regions but also reduces errors caused by human factors, providing a reliable basis for subsequent quantitative analysis. Therefore, the method of the present application has significant advantages in improving the accuracy and consistency of EBV antibody titer assessment and has important application value for enhancing the efficiency and accuracy of disease screening.

[0016] As a preferred embodiment of the first aspect, the processing of each first positive cell region mask image according to the mathematical model to obtain each second positive cell region mask image under a preset exposure condition specifically includes:

[0017] According to the gain function and the first formula, convert each pixel value of each first positive cell region mask image at different first preset exposure times into each pixel value at a second preset exposure time to obtain each second positive cell region mask image under the second preset exposure condition.

[0018] The conversion of each pixel value of each first positive cell region mask image at different first preset exposure times into each pixel value at a second preset exposure time according to the gain function and the first formula specifically includes:

[0019] Preprocess each first positive cell region mask image according to the gain function to obtain each pixel gray value after eliminating the gamma effect;

[0020] Among them, the formula of the gain function is:

[0021] V′ = G -1 (V) = V γ

[0022] In the formula, V is the pixel gray value; G is the gain function, defined as G(x) = x 1 / γ , V′ is the pixel gray value after eliminating the gamma effect;

[0023] According to the first formula and each pixel gray value, obtain each pixel value at the second preset exposure time.

[0024] The obtaining of each pixel value at the second preset exposure time according to the first formula and each pixel gray value specifically includes:

[0025] According to the first formula, convert the first preset exposure time of each pixel grayscale value to obtain each pixel value under the second preset exposure time;

[0026] The first formula is:

[0027]

[0028] In the formula, T 0 is the second preset exposure time, is the pixel grayscale value after eliminating the gamma effect under the first preset exposure time T i , and is the pixel grayscale value after eliminating the gamma effect under the second preset exposure time T 0 , and α is a preset threshold.

[0029] In this preferred embodiment, the present application processes the positive cell region mask images under different exposure times through a mathematical model, converts each pixel value into a pixel value under a unified preset exposure condition. The present application can eliminate the image brightness difference caused by different exposure times, thereby obtaining a more consistent and comparable positive cell region mask image. By using a gain function and a specific conversion formula, the pixel grayscale values under different exposure conditions can be standardized to a common exposure level, ensuring the fairness and accuracy of image comparison. This method improves the reliability of image analysis because it reduces the variability introduced by changes in exposure conditions, making subsequent statistical analysis and titer estimation more accurate, thereby improving the overall accuracy and reliability of EBV antibody titer assessment.

[0030] As a preferred embodiment of the first aspect, calculating the EBV antibody titer estimation value of each cell image according to the preset statistical model and the brightness distribution of each positive cell region in each second positive cell region mask image specifically includes:

[0031] According to the second positive cell region mask image, obtain the brightness values of each pixel point in the positive cell region of each second positive cell region mask image, and calculate the average values of each brightness value to obtain each average brightness value;

[0032] According to each average brightness value and the preset statistical model, calculate the EBV antibody titer of each cell image.

[0033] In this preferred embodiment, the present application obtains the pixel brightness values of the positive cell regions in each positive cell region mask image and calculates the average of these brightness values, thereby obtaining a quantitative index representing the brightness level of the positive cell regions in each cell image, namely the average brightness value. The brightness of the positive cell regions is related to the amount of EBV antibody present. Therefore, by analyzing the average brightness of these regions, the amount of EBV antibody can be indirectly evaluated. Subsequently, using a preset statistical model, these average brightness values are converted into estimated values of the EBV antibody titer. This method based on image brightness analysis can provide a more objective and accurate titer estimate compared to traditional subjective judgment-based evaluation methods, because it reduces the interference of human factors and improves the repeatability and accuracy of the evaluation through the application of a statistical model. Therefore, the method of the present application has significant advantages in improving the accuracy and efficiency of EBV antibody titer evaluation, which has important clinical significance for the early diagnosis and treatment monitoring of diseases.

[0034] As a preferred embodiment of the first aspect, after calculating the estimated values of the EBV antibody titers of the respective cell images, the method further includes:

[0035] Calibrating the estimated values of the EBV antibody titers of the respective cell images according to a preset titer standard to obtain the calibrated estimated values of the EBV antibody titers of the respective cell images.

[0036] In this preferred embodiment, after calculating the estimated values of the EBV antibody titer, the method of the present application further includes the step of calibrating these estimated values according to a preset titer standard. This process is crucial because it ensures the accuracy and reliability of the evaluation results. The estimated values of the EBV antibody titer obtained from different individuals or under different conditions may vary, and these variations may affect the accuracy of the evaluation. By comparing and calibrating with the preset titer standard, these differences can be eliminated, making the final titer estimate more in line with the actual situation. In this way, the method of the present application can provide a more accurate and reliable EBV antibody titer evaluation.

[0037] In a second aspect, the present application provides an apparatus for evaluating the EBV antibody titer. The apparatus for evaluating the EBV antibody titer includes an acquisition module, an input / output module, a processing module, and a calculation module;

[0038] The acquisition module is used to acquire each cell image of the EBV antigen under different preset exposure times;

[0039] The input / output module is used to input the respective cell images into a pre-trained deep learning model for prediction, so that the deep learning model outputs each first positive cell region mask image;

[0040] The processing module is used to process each of the first positive cell region mask images according to a mathematical model to obtain each second positive cell region mask image under a second preset exposure condition;

[0041] The calculation module is used to calculate the EBV antibody titer estimation value of each cell image according to a preset statistical model and the brightness distributions of the positive cell regions in each of the second positive cell region mask images.

[0042] This device uses four modules to work in division of labor and coordination to better evaluate the antibody titer. This application first ensures the comprehensiveness of data by obtaining EBV antigen cell images under different exposure times. This application analyzes these images using a deep learning model, which can automatically identify and predict positive cell regions, improving efficiency and reducing human error. Then, this application applies a mathematical model to standardize the brightness values of the positive cell regions, eliminating the influence of exposure time differences, making the brightness values more accurate and comparable. Finally, the EBV antibody titer estimation value is calculated by statistically analyzing the brightness distribution of the positive cell regions. This application improves the efficiency and accuracy of titer evaluation to solve the problem that the prior art cannot accurately and efficiently evaluate the EBV antibody titer.

[0043] As a preferred embodiment of the second aspect, the obtaining of each cell image of the EBV antigen under the first different preset exposure times is specifically as follows:

[0044] Obtain an initial cell image dataset; wherein, the initial cell image dataset is obtained by the camera shooting under the first different preset exposure times;

[0045] Input the initial cell image dataset into a pre-trained UNet network so that the UNet network outputs each cell image of the EBV antigen.

[0046] In this preferred embodiment, this application can comprehensively capture the details of EBV antigen expression by obtaining cell images of the EBV antigen under different preset exposure times. Different exposure times can show antigen-antibody reactions of different intensities, providing rich feature information for the deep learning model. Inputting these initial cell image datasets into the pre-trained UNet network can utilize its advanced feature extraction ability to automatically identify and distinguish EBV antigen-positive cell regions. This automated image recognition process not only improves the accuracy and efficiency of positive cell region detection, but also reduces errors caused by human factors, providing a reliable basis for subsequent quantitative analysis. Therefore, the method of this application has significant advantages in improving the accuracy and consistency of EBV antibody titer evaluation and has important application value for improving the efficiency and accuracy of disease screening.

[0047] As a preferred embodiment of the second aspect, processing the respective first positive cell region mask images according to the mathematical model to obtain respective second positive cell region mask images under a preset exposure condition specifically includes:

[0048] Converting respective pixel values of the respective first positive cell region mask images at different first preset exposure times into respective pixel values at a second preset exposure time according to a gain function and a first formula, to obtain respective second positive cell region mask images under the second preset exposure condition.

[0049] The converting respective pixel values of the respective first positive cell region mask images at different first preset exposure times into respective pixel values at a second preset exposure time according to a gain function and a first formula specifically includes:

[0050] Preprocessing the respective first positive cell region mask images according to the gain function to obtain respective pixel gray values after eliminating the gamma effect;

[0051] Wherein, the formula of the gain function is:

[0052] V′ = G -1 (V) = V γ

[0053] In the formula, V is the pixel gray value; G is the gain function, defined as G(x) = x 1 / γ , and V′ is the pixel gray value after eliminating the gamma effect;

[0054] Obtaining respective pixel values at the second preset exposure time according to the first formula and the respective pixel gray values.

[0055] The obtaining respective pixel values at the second preset exposure time according to the first formula and the respective pixel gray values specifically includes:

[0056] Converting the first preset exposure time of the respective pixel gray values according to the first formula to obtain respective pixel values at the second preset exposure time;

[0057] The first formula is:

[0058]

[0059] In the formula, T 0 is the second preset exposure time, is the pixel gray value after eliminating the gamma effect at the first preset exposure time T i under, is the second preset exposure time T 0The pixel gray value after eliminating the gamma effect, where α is a preset threshold.

[0060] In this preferred embodiment, the present application processes the positive cell region mask images at different exposure times through a mathematical model, converts each pixel value into a pixel value under a unified preset exposure condition. The present application can eliminate the image brightness difference caused by different exposure times, thereby obtaining a more consistent and comparable positive cell region mask image. By using a gain function and a specific conversion formula, the pixel gray values under different exposure conditions can be normalized to a common exposure level, ensuring the fairness and accuracy of image comparison. This method improves the reliability of image analysis because it reduces the variability introduced by changes in exposure conditions, making subsequent statistical analysis and titer estimation more accurate, thereby improving the overall accuracy and reliability of EBV antibody titer assessment.

[0061] As a preferred embodiment of the second aspect, calculating the EBV antibody titer estimation value of each cell image according to the preset statistical model and the respective brightness distributions of the positive cell regions in each of the second positive cell region mask images specifically includes:

[0062] According to the second positive cell region mask image, obtain the respective brightness values of each pixel point in the positive cell region of each second positive cell region mask image, and calculate the respective averages of the respective brightness values to obtain respective average brightness values;

[0063] Calculate the EBV antibody titer of each cell image according to the respective average brightness values and the preset statistical model.

[0064] In this preferred embodiment, the present application can obtain a quantitative index representing the brightness level of the positive cell region of each cell image, that is, the average brightness value, by obtaining the pixel point brightness values of the positive cell regions in each positive cell region mask image and calculating the average of these brightness values. The brightness of the positive cell region is related to the amount of EBV antibody present. Therefore, by analyzing the average brightness of these regions, the amount of EBV antibody can be indirectly evaluated. Then, using the preset statistical model, these average brightness values are converted into estimated values of the EBV antibody titer. This method based on image brightness analysis can provide a more objective and accurate titer estimate compared to traditional subjective judgment-based evaluation methods because it reduces the interference of human factors and improves the repeatability and accuracy of the evaluation through the application of a statistical model. Therefore, the method of the present application has significant advantages in improving the accuracy and efficiency of EBV antibody titer assessment, which has important clinical significance for the early diagnosis and treatment monitoring of diseases.

[0065] As a preferred embodiment of the second aspect, after calculating the EBV antibody titer estimation values of the respective cell images, the method further includes:

[0066] Calibrating the EBV antibody titer estimation values of the respective cell images according to a preset titer standard to obtain the calibrated EBV antibody titer estimation values of the respective cell images.

[0067] In this preferred embodiment, after calculating the EBV antibody titer estimation values, the method of the present application further includes the step of calibrating these estimation values according to a preset titer standard. This process is crucial because it ensures the accuracy and reliability of the evaluation results. The EBV antibody titer estimation values obtained under different individuals or different conditions may vary, and these variations may affect the accuracy of the evaluation. By comparing and calibrating with the preset titer standard, these differences can be eliminated, making the final titer estimation values more in line with the actual situation. In this way, the method of the present application can provide a more accurate and reliable EBV antibody titer evaluation. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 : A flowchart of an embodiment of the method for evaluating the EBV antibody titer provided by the present application;

[0069] Figure 2 : A structural diagram of an embodiment of the device for evaluating the EBV antibody titer provided by the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0070] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying 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 of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0071] Embodiment 1

[0072] Please refer to Figure 1 , a method for evaluating the EBV antibody titer provided by an embodiment of the present invention.

[0073] In this embodiment, the process of the method for evaluating the EBV antibody titer in the present application is described in detail through steps S01 - S04.

[0074] S01: Obtain respective cell images of the EBV antigen at different preset exposure times.

[0075] As a preferred embodiment of Embodiment 1, the obtaining of each cell image of EBV antigen under different preset exposure times specifically includes:

[0076] Obtain an initial cell image dataset; wherein, the initial cell image dataset is obtained by the camera shooting under different preset exposure times.

[0077] Input the initial cell image dataset into a pre-trained UNet network so that the UNet network outputs the cell images of each EBV antigen.

[0078] Furthermore, the production process of the initial cell dataset specifically includes:

[0079] (1) Use the Raji cell line (CCL-86) purchased from the American Type Culture Collection (ATCC) and culture it in RPMI-1640 medium containing 10% fetal bovine serum (FBS). Place the cells in suspension culture at 37°C and 5% CO2.

[0080] The cells are induced with 20 ng / ml of Phorbol 12-myristate 13-acetate (TPA) and 3.3 mg / ml of sodium butyrate (SB) for 2 days at 37°C and 5% CO2.

[0081] Use the trypan blue exclusion method and an automatic cell counter (Biorad TC 10 TM ) to determine the cell viability and quantity.

[0082] After washing the cells in 1× phosphate buffer (PBS), drop them onto a teflon-coated glass slide at a density of approximately 4000 cells per well.

[0083] Fix the glass slide with acetone for 10 minutes, dry it and store it at -20°C for later use.

[0084] (2) Use the Raji cell line (CCL-86) purchased from the American Type Culture Collection (ATCC) and culture it in RPMI-1640 medium containing 10% fetal bovine serum (FBS).

[0085] Place the cells in suspension culture at 37°C and 5% CO2.

[0086] Use the trypan blue exclusion method and an automatic cell counter (Biorad TC 10 TM ) to determine the cell viability and quantity.

[0087] After washing the cells in 1× phosphate buffered saline (PBS), they were seeded onto teflon-coated slides at a density of approximately 4000 cells per well.

[0088] The slides were fixed with acetone for 10 minutes, dried, and stored at -20 °C until use.

[0089] Furthermore, the cells were subjected to an IFA assay and the IFA titers were evaluated as follows:

[0090] PBS-Tween buffer: Dissolve 1 packet of phosphate (10.2 g) in 1 liter of distilled water, add 2 ml of Tween 20 and mix well.

[0091] Reference titer standards were prepared by pooling randomly selected serum samples for each discrete titer (EA negative, 1:10, 20, 40, 80, 160, and 320, and VCA additionally 640 and 1280).

[0092] The reference and test serum samples were diluted in PBS-Tween buffer (EA negative, 1:10, 20, 40, 80, 160, and 320, and VCA additionally 640 and 1280), and applied to slides containing Raji cells and B958 cells, and incubated in a humid chamber for 30 minutes.

[0093] The slides were rinsed with PBS-Tween buffer in a beaker for 1 second, then immediately immersed in a wash cup containing PBS-Tween buffer and soaked for at least 5 minutes. If possible, a rotary shaker can be used for oscillation.

[0094] Add 25 μl of fluorescein isothiocyanate (FITC)-conjugated rabbit anti-human IgA secondary antibody to the reaction area of a clean sample addition plate. After adding all the fluorescent secondary antibodies, start incubating in a humid chamber for 30 minutes. It is recommended to use a continuous pipettor. The FITC-labeled anti-human IgG needs to be mixed well before use. To save time, the fluorescent secondary antibody can be added to the reaction area of another sample addition plate while the first incubation is in progress, repeating the washing steps, and then applying a glycerol-fixed coverslip.

[0095] The evaluator observed the cells labeled with EBV antigens under a microscope.

[0096] The evaluator independently evaluated the samples according to the reference titer standards, visually matching the patient samples with a series of discrete reference titer standards (EA negative, 1:10, 20, 40, 80, 160, and 320, and VCA additionally 640 and 1280) to obtain the titer of the serum sample.

[0097] Furthermore, a titer sample dataset is constructed. For each titer, 10 samples are collected. Each sample is prepared into multiple sparsity samples according to the above sample preparation process, and the titer is read by an evaluator as the true titer value to form a reference titer standard data. And for the 1:20 dilution samples, multiple-exposure image acquisition is performed according to the following process. In actual use, only the 1:20 dilution samples need to be prepared and image acquisition is performed to evaluate the titer.

[0098] As a preferred embodiment of Example 1, the construction of the titer sample dataset is specifically as follows:

[0099] Use single-sample dilution (1:20) to infer continuous EBV (Epstein-Barr virus) titers.

[0100] Use a EUROstar IIIPlus microscope and a Basler's Ace sensor camera (aCA3088-57uc) to obtain images of IFA-treated wells, magnified 20 times.

[0101] Use different exposure times (API settings provided by the camera are 64000, 128000, …, 2048000, 4096000) to obtain images.

[0102] As a preferred embodiment of Example 1, the inputting of the initial cell image dataset into a pre-trained UNet network to enable the UNet network to output cell images of each EBV antigen is specifically as follows:

[0103] Perform quality control on the collected images. Use the average value of multiple pictures as input data to train a deep learning image prediction model to predict and exclude images with poor quality, such as inaccurate focus, sample movement, or incomplete shooting.

[0104] Construction of the image quality control dataset: Collect a dataset of cell images that have been manually quality-controlled. Each sample contains the average of multiple different exposure images and a manually labeled quality control label ("pass" or "fail").

[0105] Image preprocessing: Add necessary data augmentation during training, such as rotation, scaling, color enhancement, etc.

[0106] Deep learning model training: Select a suitable convolutional neural network architecture, such as the ResNet series, as the basic model, and adjust its output layer to adapt to the binary classification task.

[0107] Model performance evaluation: Use the cross-entropy loss function and multiple evaluation metrics, such as accuracy, F1 score, precision, recall, and ROC-AUC, to comprehensively evaluate the model performance.

[0108] Model inference: Use the trained deep learning model to infer new image data and predict whether the image meets the quality standards.

[0109] Result verification: Manually verify the inference results of the model to ensure the accuracy of classification and make appropriate adjustments as needed.

[0110] After completing the above operations, a titer sample data set can be obtained. Each sample in it contains multiple exposure images collected from a 1:20 dilution sample, as well as the titer value evaluated from multiple dilutions.

[0111] In this preferred embodiment, the present application can comprehensively capture the details of EBV antigen expression by obtaining cell images of EBV antigen at different preset exposure times. Different exposure times can show antigen-antibody reactions of different intensities, providing rich feature information for the deep learning model. Inputting these initial cell image data sets into the pre-trained UNet network can utilize its advanced feature extraction ability to automatically identify and distinguish EBV antigen-positive cell regions. This automated image recognition process not only improves the accuracy and efficiency of positive cell region detection, but also reduces errors caused by human factors, providing a reliable basis for subsequent quantitative analysis. Therefore, the method of the present application has significant advantages in improving the accuracy and consistency of EBV antibody titer evaluation and has important application value for enhancing the efficiency and accuracy of disease screening.

[0112] S02: Input the respective cell images into a pre-trained deep learning model for prediction, so that the deep learning model outputs respective first positive cell region mask images.

[0113] As a preferred embodiment of Embodiment 1, the inputting the respective cell images into a pre-trained deep learning model for prediction, so that the deep learning model outputs respective first positive cell region mask images is specifically:

[0114] Since different titers are mainly formed by the dense expression of the secondary antibody in antigen cells, cell regions with high fluorescence intensity appear during observation. This region is called the positive cell region. To accurately identify and predict the positive cell region, the following process is adopted:

[0115] (1) Positive cell region annotation:

[0116] An assessor uses professional marking software (such as QuPath) to outline the positive cell region in the image of a selected exposure value (such as 1024000) in the titer sample data set to obtain a positive cell region mask.

[0117] The multiple-exposure photos of the titer samples, combined with the corresponding positive cell region masks, constitute the positive cell region dataset. The dataset is randomly divided into a training set and a validation set at a ratio of 8:2.

[0118] Strict quality control is carried out on the positive cell regions outlined by doctors, including cross-checking, annotation consistency assessment, and correction of potential annotation errors.

[0119] (2) Training of the UNet semantic segmentation model:

[0120] Build a UNet semantic segmentation model. The input of the model is a stack image of the multiple-exposure photos of a titer sample, and the training target is the positive cell region mask outlined by the assessor. The model can output the predicted positive cell region mask.

[0121] During training, data augmentation techniques are adopted to improve the generalization ability of the model.

[0122] Through the validation set, use indicators such as ROC-AUC and DICE coefficient to verify the model performance.

[0123] S03: According to the mathematical model, process each of the first positive cell region mask images to obtain each of the second positive cell region mask images under the second preset exposure condition.

[0124] As a preferred embodiment of Embodiment 1, the process of processing each of the first positive cell region mask images according to the mathematical model to obtain each of the second positive cell region mask images under the second preset exposure condition is specifically as follows:

[0125] Since the intensity of the titer is related to the binding strength of the secondary antibody and the antibody, the higher the binding strength, the higher the fluorescence intensity. Therefore, this application plans to only use samples with a dilution ratio of 1:20 and estimate the fluorescence intensity of the samples to complete the titer prediction.

[0126] During the camera imaging process, the relationship between the pixel gray value V, the actual brightness B of the measured object, and the exposure time T can be expressed as:

[0127]

[0128] Among them: V is the pixel gray value; G is the gain function, defined as G(x)=x 1 / γ ; α is the contrast coefficient of the sensor, determined by the photosensitive material; m is a constant, determined by the photosensitive material; τ is the lens transmittance, determined by the lens characteristics; F is the camera aperture number; B is the actual brightness of the measured object; T is the exposure time.

[0129] Generally, α, m, and τ are all intrinsic parameters of the camera and are not easily measured. Under the current multiple-exposure setting, the camera aperture number F is a fixed value, and the exposure time T is twice as long as the previous one for each exposure.

[0130] Define the time value of multiple exposures as T 0 , T 1 , …, T n , and each exposure time is twice as long as the previous one, that is, T i = 2T i-1 . To eliminate the influence of the exposure time, the pixel values under different exposure times can be converted into pixel values under the standard exposure time T 0 . First, eliminate the influence of the gain function G:

[0131] V′ = G -1 (V) = V γ ;

[0132] where: V is the pixel gray value; G is the gain function, defined as G(x) = x 1 / γ , and V′ is the pixel gray value after eliminating the gamma effect. In the samples we collected, γ = 0.2669 was fitted.

[0133] For multiple images with different exposure times T i , the pixel gray values can be converted through the following formula:

[0134]

[0135] where, T 0 is the starting exposure time, used as the standard exposure time, is the pixel gray value after eliminating the gamma effect under the exposure time T i , is the pixel gray value after eliminating the gamma effect under the standard exposure time T 0 . By converting the pixel gray values under different exposure times to the standard exposure time, the influence of different exposure times can be eliminated, and thus a more realistic brightness value can be obtained.

[0136] Here, T 0 can be set to 64000 ms, the gamma-restored pixel values of the remaining exposure photos are used, and the least squares method is used for fitting with the above formula to obtain α. In the samples collected, α = 0.2925 was fitted. Then, the exposure amounts of the photos with different exposure times of this titer sample can be restored to

[0137] Different exposure time parameters represent different observations of the same sample. By averaging the photos with different exposure times to restore the brightness estimate value at the exposure time, the accuracy of the brightness value estimate can be improved. The specific steps are as follows:

[0138] Convert the grayscale value of each image to obtain the pixel grayscale value corresponding to the standard exposure time.

[0139] Since the response curve of CMOS shows a strong non-linear relationship at too high and too low brightness, the pixel value differs greatly from the true brightness. Through multiple exposures, pixel values less than and greater than (considered invalid pixels) can be excluded to ensure that the data is within the linear response range of CMOS. 10 can be taken.

[0140] Average the valid pixels in the same area to obtain

[0141] Through the above steps, the image noise level and measurement error can be effectively reduced, avoiding inaccurate brightness measurement caused by overexposure or underexposure, so as to obtain more accurate gamma recovery pixel values.

[0142] In this preferred embodiment, the present application processes the positive cell region mask images at different exposure times through a mathematical model, and converts each pixel value into a pixel value under a unified preset exposure condition. The present application can eliminate the image brightness difference caused by different exposure times, so as to obtain a more consistent and comparable positive cell region mask image. Using a gain function and a specific conversion formula, the pixel grayscale values under different exposure conditions can be standardized to a common exposure level, ensuring the fairness and accuracy of image comparison. This method improves the reliability of image analysis because it reduces the variability introduced by changes in exposure conditions, making subsequent statistical analysis and titer estimation more accurate, thereby improving the overall accuracy and reliability of EBV antibody titer assessment.

[0143] S04: According to the preset statistical model and the respective brightness distributions of the positive cell regions in each of the second positive cell region mask images, calculate the EBV antibody titer estimate values of each of the cell images.

[0144] As a preferred embodiment of Embodiment 1, the calculating the EBV antibody titer estimate values of each of the cell images according to the preset statistical model and the respective brightness distributions of the positive cell regions in each of the second positive cell region mask images is specifically:

[0145] Let S 1:tSerum samples with a titer of t. Generally, the notations of titers are 1:10, 1:20, …, 1:320. For the convenience of notation here, the corresponding titer value t is actually 1 / 10 of the actual value, that is, t = 1, 2, …, 32. Let the fluorescence intensity of positive cells in the sample with a titer of t be Assume the fluorescence intensity is proportional to the titer t, that is where is the fluorescence intensity of positive cells when the reference titer is 1, and 2 β is used as the scaling factor. Substituting it into the camera imaging formula, we have:

[0146]

[0147] where: is the average pixel gray value after eliminating the gamma effect of the sample with a titer of t under the standard exposure time T 0 ; is the fluorescence intensity of positive cells in the sample with a titer value of t; is the fluorescence intensity of positive cells when the reference titer is 1; T 0 is the standard exposure time; α is the contrast coefficient of the sensor, which is determined by the photosensitive material; m is a constant, which is determined by the photosensitive material; τ is the lens transmittance, which is determined by the lens characteristics; F is the camera aperture number; β is the scaling coefficient between the fluorescence intensity and the titer.

[0148] Similarly, for the sample with a titer of t = 1, we have:

[0149]

[0150] Subtracting the two equations, we can get:

[0151]

[0152] After simple transformation, we can get:

[0153]

[0154] In this formula, can be obtained by calculating the average pixel gray value after eliminating the gamma effect of the positive cell region of all samples with a titer of t, and is the corresponding value of the sample with a titer of t = 1. Fitting again can obtain β. In the collected EA samples, β = 0.3989 is fitted; in the VCA samples, β = 0.3919 is fitted. For the new titer samples, we can let After equalizing the average pixel gray value after eliminating the gamma effect of the positive cell region and substituting it into the above formula, the titer t is obtained. It should be noted that t here is a shorthand expression of the titer. That is, if t = 2, the actual final titer should be expressed as 1:20.

[0155] In this preferred embodiment, the present application can obtain a quantitative index representing the brightness level of the positive cell region in each cell image, that is, the average brightness value, by obtaining the pixel point brightness values of the positive cell regions in the mask images of each positive cell region and calculating the average of these brightness values. The brightness of the positive cell region is related to the amount of EBV antibody. Therefore, by analyzing the average brightness of these regions, the amount of EBV antibody can be indirectly evaluated. Then, using a preset statistical model, these average brightness values are converted into an estimated value of the EBV antibody titer. This method based on image brightness analysis can provide a more objective and accurate titer estimate compared to the traditional subjective judgment-based evaluation method, because it reduces the interference of human factors and improves the repeatability and accuracy of the evaluation through the application of a statistical model. Therefore, the method of the present application has significant advantages in improving the accuracy and efficiency of EBV antibody titer evaluation, which has important clinical significance for the early diagnosis and treatment monitoring of diseases.

[0156] The present application first ensures the comprehensiveness of the data by obtaining EBV antigen cell images at different exposure times. The present application uses a deep learning model to analyze these images, which can automatically identify and predict the positive cell regions, improving the efficiency and reducing human errors. Then, the present application applies a mathematical model to standardize the brightness values of the positive cell regions, eliminating the influence caused by exposure time differences, so that the brightness values are more accurate and comparable. Finally, by statistically analyzing the brightness distribution of the positive cell regions, an estimated value of the EBV antibody titer is calculated. The present application improves the efficiency and accuracy of titer evaluation to solve the problem that the prior art cannot accurately and efficiently evaluate the EBV antibody titer.

[0157] Embodiment 2

[0158] Please refer to Figure 2 , which is an evaluation device for EBV antibody titer provided by an embodiment of the present application.

[0159] In this embodiment, the evaluation device for EBV antibody titer includes an acquisition module 10, an input / output module 20, a processing module 30, and a calculation module 40.

[0160] The acquisition module 10 is used to acquire each cell image of the EBV antigen at different preset exposure times.

[0161] As a preferred embodiment of the second embodiment, the obtaining of each cell image of the EBV antigen at different preset exposure times specifically includes:

[0162] Obtain an initial cell image dataset; wherein, the initial cell image dataset is obtained by the camera shooting at different preset exposure times.

[0163] Input the initial cell image dataset into a pre-trained UNet network so that the UNet network outputs the cell images of each EBV antigen.

[0164] Furthermore, the production process of the initial cell dataset specifically includes:

[0165] (1) Use the Raji cell line (CCL-86) purchased from the American Type Culture Collection (ATCC) and culture it in RPMI-1640 medium containing 10% fetal bovine serum (FBS). Place the cells in suspension culture at 37°C and 5% CO2.

[0166] Induce the cells with 20 ng / ml of Phorbol 12-myristate 13-acetate (TPA) and 3.3 mg / ml of sodium butyrate (SB) for 2 days at 37°C and 5% CO2.

[0167] Use the trypan blue exclusion method and an automatic cell counter (Biorad TC 10 TM ) to determine the cell viability and quantity.

[0168] After washing the cells in 1× phosphate buffered saline (PBS), drop them onto a teflon-coated slide at a density of approximately 4000 cells per well.

[0169] Fix the slide with acetone for 10 minutes, dry it, and store it at -20°C for later use.

[0170] (2) Use the Raji cell line (CCL-86) purchased from the American Type Culture Collection (ATCC) and culture it in RPMI-1640 medium containing 10% fetal bovine serum (FBS).

[0171] Place the cells in suspension culture at 37°C and 5% CO2.

[0172] Use the trypan blue exclusion method and an automatic cell counter (Biorad TC 10 TM ) to determine the cell viability and quantity.

[0173] After washing the cells in 1× phosphate buffered saline (PBS), they were seeded onto teflon-coated slides at a density of approximately 4000 cells per well.

[0174] The slides were fixed with acetone for 10 minutes, dried and stored at -20 °C until use.

[0175] Furthermore, the cells were subjected to an IFA assay and the IFA titers were evaluated as follows:

[0176] PBS-Tween buffer: Dissolve 1 packet of phosphate (10.2 g) in 1 liter of distilled water, add 2 ml of Tween 20 and mix well.

[0177] Reference titer standards were prepared by pooling randomly selected serum samples for each discrete titer (EA negative, 1:10, 20, 40, 80, 160 and 320, and VCA with additional 640 and 1280).

[0178] The reference and test serum samples were diluted in PBS-Tween buffer (EA negative, 1:10, 20, 40, 80, 160 and 320, and VCA with additional 640 and 1280), and applied to slides containing Raji cells and B958 cells, and incubated in a humid chamber for 30 minutes.

[0179] The slides were rinsed with PBS-Tween buffer in a beaker for 1 second, then immediately immersed in a wash cup containing PBS-Tween buffer and soaked for at least 5 minutes. A rotary shaker can be used for oscillation if available.

[0180] Add 25 μl of fluorescein isothiocyanate (FITC)-conjugated rabbit anti-human IgA secondary antibody to the reaction area of a clean sample addition plate. After adding all the fluorescent secondary antibodies, start incubating in a humid chamber for 30 minutes. A continuous pipettor is recommended. The FITC-labeled anti-human IgG needs to be mixed well before use. To save time, the fluorescent secondary antibody can be added to the reaction area of another sample addition plate while the first incubation is in progress, repeating the washing steps, and then a cover slip is fixed with glycerol.

[0181] The evaluator observed the cells labeled with EBV antigens under a microscope.

[0182] The evaluator independently evaluated the samples according to the reference titer standards, visually matching the patient samples with a series of discrete reference titer standards (EA negative, 1:10, 20, 40, 80, 160 and 320, and VCA with additional 640 and 1280) to obtain the titer of the serum sample.

[0183] Furthermore, a titer sample dataset is constructed. For each titer, 10 samples are collected. Each sample is prepared into multiple sparsity samples according to the above sample preparation process, and the titer is read by an assessor as the true titer value to form a reference titer standard data. And for the 1:20 dilution samples, multiple-exposure image acquisition is performed according to the following process. In actual use, only the 1:20 dilution samples need to be prepared and image acquisition is performed to evaluate the titer.

[0184] As a preferred embodiment of Example 2, the construction of the titer sample dataset is specifically as follows:

[0185] Use single-sample dilution (1:20) to infer continuous EBV (Epstein-Barr virus) titers.

[0186] Use a EUROstar IIIPlus microscope and a Basler's Ace sensor camera (aCA3088-57uc) to obtain images of IFA-treated wells, magnified 20 times.

[0187] Use different exposure times (API settings provided by the camera are 64000, 128000, …, 2048000, 4096000) to obtain images.

[0188] As a preferred embodiment of Example 2, the inputting of the initial cell image dataset into a pre-trained UNet network to enable the UNet network to output cell images of each EBV antigen is specifically as follows:

[0189] Perform quality control on the collected images. Use the average value of multiple pictures as input data to train a deep learning image prediction model to predict and exclude images with poor quality, such as inaccurate focus, sample movement, or incomplete shooting.

[0190] Construction of the image quality control dataset: Collect a cell image dataset that has been manually quality-controlled. Each sample contains the average image of multiple different exposure images and a manually labeled quality control label ("pass" or "fail").

[0191] Image preprocessing: Add necessary data augmentation during training, such as rotation, scaling, color enhancement, etc.

[0192] Deep learning model training: Select a suitable convolutional neural network architecture, such as the ResNet series, as the basic model, and adjust its output layer to adapt to the binary classification task.

[0193] Model performance evaluation: Use the cross-entropy loss function and multiple evaluation metrics, such as accuracy, F1 score, precision, recall, and ROC-AUC, to comprehensively evaluate the model performance.

[0194] Model inference: Use the trained deep learning model to perform inference on new image data and predict whether the image meets the quality standards.

[0195] Result verification: Manually verify the inference results of the model to ensure the accuracy of classification and make appropriate adjustments as needed.

[0196] After completing the above operations, a titer sample data set can be obtained. Each sample in it contains multiple-exposure images collected from a 1:20 dilution sample, as well as the titer value evaluated from multiple dilutions.

[0197] In this preferred embodiment, the present application can comprehensively capture the details of EBV antigen expression by acquiring cell images of EBV antigen at different preset exposure times. Different exposure times can show antigen-antibody reactions of different intensities, providing rich feature information for the deep learning model. Inputting these initial cell image data sets into the pre-trained UNet network can utilize its advanced feature extraction ability to automatically identify and distinguish EBV antigen-positive cell regions. This automated image recognition process not only improves the accuracy and efficiency of positive cell region detection, but also reduces errors caused by human factors, providing a reliable basis for subsequent quantitative analysis. Therefore, the method of the present application has significant advantages in improving the accuracy and consistency of EBV antibody titer evaluation and has important application value for enhancing the efficiency and accuracy of disease screening.

[0198] The input-output module 20 is used to input the respective cell images into a pre-trained deep learning model for prediction, so that the deep learning model outputs respective first positive cell region mask images.

[0199] As a preferred embodiment of the second embodiment, the inputting the respective cell images into a pre-trained deep learning model for prediction, so that the deep learning model outputs respective first positive cell region mask images is specifically as follows:

[0200] Since different titers are mainly formed by the dense expression of the secondary antibody in antigen cells, cell regions with high fluorescence intensity appear during observation. This region is called the positive cell region. To accurately identify and predict the positive cell region, the following process is adopted:

[0201] (1) Positive cell region annotation:

[0202] An evaluator uses professional marking software (such as QuPath) to outline the positive cell region in the image of a selected exposure value (such as 1024000) in the titer sample data set to obtain a positive cell region mask.

[0203] The multiple-exposure photos of the titer samples, combined with the corresponding positive cell region masks, constitute the positive cell region dataset. The dataset is randomly divided into a training set and a validation set at a ratio of 8:2.

[0204] Strict quality control is carried out on the positive cell regions outlined by doctors, including cross-checking, annotation consistency assessment, and correction of potential annotation errors.

[0205] (2) Training of the UNet semantic segmentation model:

[0206] Build a UNet semantic segmentation model. The input of the model is a stack image of multiple-exposure photos of a titer sample, and the training target is the positive cell region mask outlined by the assessor. The model can output the predicted positive cell region mask.

[0207] During training, data augmentation techniques are adopted to improve the generalization ability of the model.

[0208] Through the validation set, use indicators such as ROC-AUC and DICE coefficient to verify the model performance.

[0209] The processing module 30 is used to process each of the first positive cell region mask images according to a mathematical model to obtain each second positive cell region mask image under a second preset exposure condition.

[0210] As a preferred embodiment of the second embodiment, the processing of each of the first positive cell region mask images according to a mathematical model to obtain each second positive cell region mask image under a second preset exposure condition is specifically as follows:

[0211] Since the intensity of the titer is related to the binding intensity of the secondary antibody and the antibody, the higher the binding intensity, the higher the fluorescence intensity. Therefore, this application plans to only use samples with a 1:20 dilution and estimate the fluorescence intensity of the samples to complete titer prediction.

[0212] During the camera imaging process, the relationship between the pixel gray value V, the actual brightness B of the measured object, and the exposure time T can be expressed as:

[0213]

[0214] where: V is the pixel gray value; G is the gain function, defined as G(x)=x 1 / γ ; α is the contrast coefficient of the sensor, determined by the photosensitive material; m is a constant, determined by the photosensitive material; τ is the lens transmittance, determined by the lens characteristics; F is the camera aperture number; B is the actual brightness of the measured object; T is the exposure time.

[0215] Generally, α, m, and τ are all intrinsic parameters of the camera and are not easily measured. Under the current multiple-exposure setting, the camera aperture number F is a fixed value, and the exposure time T is twice as long as the previous one for each exposure.

[0216] Define the time value of multiple exposures as T 0 , T 1 , …, T n , where each exposure time is twice as long as the previous one, i.e., T i = 2T i-1 . To eliminate the influence of exposure time, the pixel values under different exposure times can be converted into pixel values under the standard exposure time T 0 . First, eliminate the influence of the gain function G:

[0217] V′ = G -1 (V) = V γ ;

[0218] where: V is the pixel gray value; G is the gain function, defined as G(x) = x 1 / γ , and V′ is the pixel gray value after eliminating the gamma effect. In the samples we collected, γ = 0.2669 was fitted.

[0219] For multiple images with different exposure times T i , the pixel gray values can be converted through the following formula:

[0220]

[0221] where, T 0 is the starting exposure time, used as the standard exposure time, is the pixel gray value after eliminating the gamma effect under the exposure time T i , and is the pixel gray value after eliminating the gamma effect under the standard exposure time T 0 . By converting the pixel gray values under different exposure times to the standard exposure time, the influence of different exposure times can be eliminated, and thus a more realistic brightness value can be obtained.

[0222] Here, let T 0 = 64000 ms, restore the gamma pixel values of the remaining exposed photos, and use the above formula for least-squares fitting to obtain α. In the samples collected, α = 0.2925 was fitted. Then, the exposure amounts of the photos with different exposure times of this titer sample can be restored to

[0223] Different exposure time parameters represent different observations of the same sample. By averaging the photos with different exposure times to restore the brightness estimate value at the exposure time, the accuracy of the brightness value estimate can be improved. The specific steps are as follows:

[0224] Convert the grayscale value of each image to obtain the pixel grayscale value corresponding to the standard exposure time.

[0225] Since the response curve of CMOS shows a strong non-linear relationship at too high and too low brightness, the pixel value differs greatly from the true brightness. Through multiple exposures, pixel values less than and greater than (considered invalid pixels) can be excluded to ensure that the data is within the linear response range of CMOS. 10 can be taken.

[0226] Average the valid pixels in the same area to obtain

[0227] Through the above steps, the image noise level and measurement error can be effectively reduced, and the inaccurate brightness measurement caused by overexposure or underexposure can be avoided, so as to obtain more accurate gamma recovery pixel values.

[0228] In this preferred embodiment, the present application processes the positive cell region mask images at different exposure times through a mathematical model, converts each pixel value into a pixel value under a unified preset exposure condition. The present application can eliminate the image brightness difference caused by different exposure times, so as to obtain a more consistent and comparable positive cell region mask image. Using a gain function and a specific conversion formula, the pixel grayscale values under different exposure conditions can be standardized to a common exposure level, ensuring the fairness and accuracy of image comparison. This method improves the reliability of image analysis because it reduces the variability introduced by changes in exposure conditions, making subsequent statistical analysis and titer estimation more accurate, thereby improving the overall accuracy and reliability of EBV antibody titer assessment.

[0229] The calculation module 40 is used to calculate the EBV antibody titer estimate value of each cell image according to a preset statistical model and the respective brightness distributions of the positive cell regions in the respective second positive cell region mask images.

[0230] As a preferred embodiment of the second embodiment, the calculating the EBV antibody titer estimate value of each cell image according to a preset statistical model and the respective brightness distributions of the positive cell regions in the respective second positive cell region mask images is specifically:

[0231] Let S 1:tSerum samples with a titer of t. Under normal circumstances, the notations of titers are 1:10, 1:20, …, 1:320. Here, for the convenience of notation, the corresponding titer value t is actually 1 / 10 of the actual value, that is, t = 1, 2, …, 32. Let the fluorescence intensity of positive cells in the sample with a titer of t be Assume the fluorescence intensity is proportional to the titer t, that is where is the fluorescence intensity of positive cells when the reference titer is 1, and 2 β is used as the scaling coefficient. Substituting into the camera imaging formula, we have:

[0232]

[0233] where: is the average pixel gray value after eliminating the gamma effect of the sample with a titer of t under the standard exposure time T 0 ; is the fluorescence intensity of positive cells in the sample with a titer value of t; is the fluorescence intensity of positive cells when the reference titer is 1; T 0 is the standard exposure time; α is the contrast coefficient of the sensor, which is determined by the photosensitive material; m is a constant, which is determined by the photosensitive material; τ is the lens transmittance, which is determined by the lens characteristics; F is the camera aperture number; β is the scaling coefficient between the fluorescence intensity and the titer.

[0234] Similarly, for the sample with a titer of t = 1, we have:

[0235]

[0236] Subtracting the two equations, we can get:

[0237]

[0238] After simple transformation, we can get:

[0239]

[0240] In this formula, can be obtained by calculating the average pixel gray value after eliminating the gamma effect of the positive cell region of all samples with a titer of t, and After obtaining the average pixel gray value after eliminating the gamma effect of the positive cell area and substituting it into the above formula, the titer t is obtained. It should be noted that t here is a shorthand expression of the titer. That is, if t = 2, the actual final titer should be expressed as 1:20.

[0241] In this preferred embodiment, the present application can obtain a quantitative index representing the brightness level of the positive cell area in each positive cell area mask image by acquiring the pixel point brightness values of the positive cell areas in these images and calculating the average of these brightness values, that is, the average brightness value. The brightness of the positive cell area is related to the amount of EBV antibody present. Therefore, by analyzing the average brightness of these areas, the amount of EBV antibody can be indirectly evaluated. Then, using a preset statistical model, these average brightness values are converted into an estimated value of the EBV antibody titer. This method based on image brightness analysis can provide a more objective and accurate titer estimate compared to the traditional subjective judgment-based evaluation method, because it reduces the interference of human factors and improves the repeatability and accuracy of the evaluation through the application of the statistical model. Therefore, the method of the present application has significant advantages in improving the accuracy and efficiency of EBV antibody titer evaluation, which has important clinical significance for the early diagnosis and treatment monitoring of diseases.

[0242] This device can better evaluate the antibody titer by using four modules to work in division and coordination. The present application first ensures the comprehensiveness of the data by acquiring EBV antigen cell images under different exposure times. The present application uses a deep learning model to analyze these images. This model can automatically identify and predict the positive cell area, which not only improves the efficiency but also reduces human error. Then, the present application applies a mathematical model to standardize the brightness values of the positive cell area, eliminating the influence brought by the exposure time difference, so that the brightness values are more accurate and comparable. Finally, by statistically analyzing the brightness distribution of the positive cell area, an estimated value of the EBV antibody titer is calculated. The present application improves the efficiency and accuracy of titer evaluation to solve the problem that the prior art cannot accurately and efficiently evaluate the EBV antibody titer.

[0243] Embodiment 3:

[0244] The embodiment of the present application provides a computer-readable storage medium, and the computer-readable storage medium includes a stored computer program, wherein when the computer program runs, it controls the device where the computer-readable storage medium is located to execute the method for evaluating the titer of EBV antibody described above;

[0245] Among them, for the method for evaluating the titer of EBV antibody, when it is implemented in the form of a software functional unit and used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, to implement all or part of the processes in the above-mentioned embodiment methods of the present invention, it can also be completed by instructing relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, the steps of the above-mentioned various method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in the form of source code, object code, executable file or some intermediate form, etc. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disc, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electrical carrier signal, telecommunication signal, and software distribution medium, etc.

[0246] Embodiment 4

[0247] This application provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the method for evaluating the titer of EBV antibody according to any one of the embodiments as described in Embodiment 1.

[0248] The above-mentioned specific embodiments have further elaborated on the purpose, technical solutions, and beneficial effects of the present invention. It should be understood that the above-mentioned are only specific embodiments of the present invention and are not used to limit the protection scope of the present invention. It is particularly pointed out that for those skilled in the art, any modifications, equivalent replacements, 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. A method for evaluating EBV antibody titer, characterized in that: include: Acquire individual cell images of EBV antigens at first different preset exposure times; Inputting each cell image into a pre-trained deep learning model for prediction, so that the deep learning model outputs each first positive cell region mask image; According to the mathematical model, each of the first positive cell region mask images is processed to obtain each of the second positive cell region mask images under a second preset exposure condition; The EBV antibody titer estimation value of each cell image is calculated based on the preset statistical model and each brightness distribution of the positive cell region in each second positive cell region mask image.

2. The method for evaluating EBV antibody titer according to claim 1, characterized in that: The acquiring of each cell image of the EBV antigen under the first different preset exposure time is specifically: Acquire an initial cell image data set; wherein the initial cell image data set is obtained by photographing with a camera at a first different preset exposure time; The initial cell image data set is input into a pre-trained UNet network so that the UNet network outputs cell images of various EBV antigens.

3. The method for evaluating EBV antibody titer according to claim 1, characterized in that: The first positive cell region mask images are processed according to the mathematical model to obtain the second positive cell region mask images under the preset exposure conditions, specifically: According to the gain function and the first formula, each pixel value of each first positive cell area mask image under different first preset exposure times is converted into each pixel value under a second preset exposure time to obtain each second positive cell area mask image under the second preset exposure condition.

4. The method for evaluating EBV antibody titer according to claim 3, characterized in that: According to the gain function and the first formula, the pixel values ​​of the first positive cell region mask images at different first preset exposure times are converted into pixel values ​​at the second preset exposure time, specifically: Preprocessing each of the first positive cell region mask images according to the gain function to obtain each pixel grayscale value after eliminating the gamma effect; Wherein, the formula of the gain function is: V′=G -1 (V)=V γ Where V is the pixel grayscale value; G is the gain function, defined as G(x) = x 1 / γ , V′ is the pixel gray value after eliminating the gamma effect; According to the first formula and the grayscale values ​​of each pixel, the values ​​of each pixel under the second preset exposure time are obtained.

5. The method for evaluating EBV antibody titer according to claim 4, characterized in that: The pixel values ​​under the second preset exposure time are obtained according to the first formula and the grayscale values ​​of the pixels, specifically: According to the first formula, the first preset exposure time of each pixel gray value is converted to obtain each pixel value under the second preset exposure time; The first formula is: Where T0 is the second preset exposure time, is the first preset exposure time T i The pixel gray value after eliminating the gamma effect is is the pixel grayscale value after eliminating the gamma effect at the second preset exposure time T0, and α is the preset threshold.

6. The method for evaluating EBV antibody titer according to claim 1, characterized in that: The EBV antibody titer estimation value of each cell image is calculated based on the preset statistical model and each brightness distribution of the positive cell area in each second positive cell area mask image, specifically: According to the second positive cell region mask image, obtaining each brightness value of each pixel point in the positive cell region in each of the second positive cell region mask images, and calculating each average value of the each brightness value to obtain each average brightness value; The EBV antibody titer of each cell image is calculated based on the average brightness values ​​and a preset statistical model.

7. The method for evaluating EBV antibody titer according to claim 1, characterized in that: After the EBV antibody titer estimation value of each cell image is obtained by calculation, the method further includes: According to a preset titer standard, the estimated EBV antibody titer value of each cell image is calibrated to obtain a calibrated estimated EBV antibody titer value of each cell image.

8. An EBV antibody titer assessment device, characterized in that: It includes an acquisition module, an input and output module, a processing module and a calculation module; The acquisition module is used to acquire each cell image of the EBV antigen under a first different preset exposure time; The input-output module is used to input each cell image into a pre-trained deep learning model for prediction, so that the deep learning model outputs each first positive cell region mask image; The processing module is used to process the first positive cell region mask images according to the mathematical model to obtain the second positive cell region mask images under the second preset exposure condition; The calculation module is used to calculate the estimated value of the EBV antibody titer of each cell image according to a preset statistical model and each brightness distribution of the positive cell area in each second positive cell area mask image.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located is controlled to execute the method for evaluating the EBV antibody titer according to any one of claims 1 to 7.

10. A terminal device, characterized in that: The method comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, the method for evaluating the EBV antibody titer according to any one of claims 1 to 7 is implemented.

Images