Broiler blood immunoglobulin detection model construction method and system

Abstract

A method and system for constructing a detection model for broiler blood immunoglobulins, relating to the field of bioinformatics, alleviates the problems of high pollution and high cost in existing broiler blood immunoglobulin detection technologies. The method for constructing a broiler blood immunoglobulin detection model, based on broiler blood immunoglobulin indicators, effectively constructs a comprehensive evaluation index of broiler blood immunoglobulins through comprehensive analysis, and combines this index with the visible-near-infrared reflectance spectrum of chicken feces to establish a comprehensive broiler blood immunoglobulin indicator detection model based on constructed spectral indices. This invention is applicable to the field of rapid, non-destructive detection of poultry immune status, as well as in the biopharmaceutical industry for evaluating the immunomodulatory effects of functional feed additives, analyzing big data on vaccine immune responses, and screening diagnostic biomarkers for poultry diseases through data processing.

Description

Technical Field

[0001] This invention relates to the field of bioinformatics technology, specifically to the field of blood immunoglobulin detection technology. Background Technology

[0002] Immunoglobulins are important indicators of humoral immunity in animals, with IgG, IgA, and IgM being the three most prevalent immunoglobulins in poultry. IgG specifically recognizes and binds to antigens such as bacteria, viruses, and parasites that invade the body, triggering an immune response to eliminate these pathogens. It also neutralizes toxins and enhances the phagocytic activity of immune cells, thereby improving the body's resistance to infection. IgA primarily functions as an immune defense in mucosal sites such as the respiratory and digestive tracts, effectively preventing pathogens from entering the body through the mucosa and maintaining local immune balance. IgM is the first-responding immunoglobulin; it attaches to pathogenic microorganisms and blocks their receptors, playing a crucial role in clearing and neutralizing pathogens and regulating immunity. Therefore, changes in the levels of IgG, IgA, and IgM during broiler rearing directly reflect the flock's health level, immune function status, and responsiveness to disease and environmental stress, making them key physiological and biochemical indicators for broiler health assessment and precision farming management.

[0003] Different immunoglobulins exhibit significant differences in their immune response stages, sites of action, and functional mechanisms. In actual production, using a single immunoglobulin indicator to evaluate the immune status of broiler flocks often fails to comprehensively reflect the overall immune function of the organism. By comprehensively evaluating IgG, IgA, and IgM, a comprehensive immunoglobulin evaluation index can be constructed. This not only reduces the evaluation bias caused by fluctuations in single indicators but also more comprehensively reflects the overall level of immune status in broiler flocks, thereby improving the stability and reliability of health assessment. Therefore, establishing a comprehensive index integrating multiple immunoglobulins is of great significance in poultry health monitoring.

[0004] Currently, immunoglobulin detection mainly relies on enzyme-linked immunosorbent assay (ELISA). Although these methods have high detection accuracy, they have long detection cycles and are invasive sampling methods that can easily cause stress to chickens. They are not suitable for rapid detection in the breeding field and are not conducive to large-scale, continuous monitoring.

[0005] Chicken manure, as a digestive and metabolic product, contains incompletely digested nutrients and microorganisms, providing a direct reflection of the chicken's digestive metabolism and health status. It contains a wealth of information, and its composition is significantly influenced by the chicken's immune status, intestinal function, and overall health. Related research indicates that changes in the color and morphology of chicken manure are closely related to the digestive function and health status of the flock, serving as a preliminary reference indicator for assessing flock health. Therefore, analyzing the changing characteristics of relevant substances in chicken manure holds promise for achieving non-invasive detection of the flock's immune status and immunoglobulin levels.

[0006] Visible-near-infrared spectroscopy is characterized by its speed, non-destructive nature, and high information content, enabling it to reflect relevant information about organic matter, nitrogen-containing compounds, and other chemical components in samples. However, chicken manure spectral data is easily affected by moisture, organic matter, and background noise, exhibiting high information overlap, high variable dimensionality, and potentially non-linear and weak correlations with immune indicators. Spectral indices, through combined calculations of specific bands, can highlight spectral features highly correlated with target indicators, reduce background interference, and improve the robustness and interpretability of the model. Compared to full-spectrum modeling methods, spectral index models have advantages such as fewer parameters, lower computational cost, and ease of engineering implementation, making them more suitable for rapid on-site detection and embedded device applications.

[0007] In the prior art, Chinese patent document CN120913662A discloses "a method, system and detection method for constructing a model for detecting crude protein content in chicken manure". This technical solution is based on the visible protein content of chicken manure. Near-infrared reflectance spectroscopy and crude protein content were used to screen for optimal wavelength positions and corresponding three spectral indices. A dataset was constructed based on these three spectral indices and crude protein content, and the best fitting equation, corresponding preprocessing method, and wavelength position were selected as the model for detecting crude protein content in chicken manure. This technical solution uses fecal spectral information as input to directly detect the crude protein content of the feces itself. This type of detection principle is based on the target indicator having a clear, stable, and dominant information contribution in the spectrum, establishing a reliable correspondence, belonging to a direct and explicit indicator-spectral correspondence. However, the prediction of blood immunoglobulins IgG, IgA, and IgM belongs to the problem of "cross-media physiological state characterization," that is, indirectly reflecting the level of immunoglobulins in the body's blood through fecal spectral information. This type of detection is not based on the target indicator directly dominating the spectral response, but relies on weak correlation information in the fecal spectrum related to the body's immune status. This correlation information is easily interfered with by factors unrelated to the immune status, belonging to an indirect and implicit state-spectral mapping relationship, leading to insufficient detection accuracy and further increasing detection complexity.

[0008] In summary, existing broiler blood immunoglobulin detection technologies suffer from several problems, including significant contamination during the detection process, high costs, poor model stability, insufficient accuracy and high complexity when spectral analysis technology is directly applied to broiler blood immunoglobulin detection, and difficulty in meeting the real-time health monitoring needs of modern large-scale farming. Summary of the Invention

[0009] This invention alleviates the problems of existing broiler blood immunoglobulin detection technologies, such as high contamination during the detection process, high cost, poor model stability, insufficient detection accuracy and high complexity when directly applying spectral analysis technology to broiler blood immunoglobulin detection, and difficulty in meeting the real-time health monitoring needs of modern large-scale farming. This invention provides the following solution: Option 1: A method for constructing a broiler blood immunoglobulin detection model, the method comprising the following steps: Step A1: Obtain a sample set, wherein the sample set includes visible chicken manure. Near-infrared reflectance spectra and corresponding blood immunoglobulin levels; Step A2, using a baseline offset correction preprocessing method to preprocess the visible... The near-infrared reflectance spectrum is preprocessed to obtain the corresponding baseline shift correction preprocessed spectrum; Step A3: A comprehensive evaluation of several blood immunoglobulin indicators is performed to obtain a comprehensive blood immunoglobulin index. Step A4, based on the visible The positions of all wavelengths in the near-infrared reflectance spectrum and the baseline offset correction preprocessed spectrum are used to obtain the corresponding three spectral indices. The three spectral indices mentioned include the ratio spectral index RI, the difference spectral index DI, and the normalized spectral index NDVI. Step A5: Evaluate the correlation between the three spectral indices and their corresponding comprehensive blood immunoglobulin indicators, and select the visible... The optimal wavelength position for near-infrared reflectance spectra and baseline shift correction preprocessed spectra; Step A6: Based on the three spectral indices corresponding to the optimal wavelength position, construct data sets DA1, DA2, or DA3 corresponding to the visible-near-infrared reflectance spectrum and the comprehensive blood immunoglobulin index, as well as data sets DB1, DB2, and DB3 corresponding to the baseline offset correction preprocessed spectrum. Multiple fitting equations were obtained by fitting each of the above data sets using various fitting methods. Step A7: Combine the data group DB1, the data group DB2 and the data group DB3 into a data group DB, and fit the data group DB with various machine learning methods to obtain multiple detection models; Step A8: The multiple fitting equations and multiple detection models are evaluated using the coefficient of determination and root mean square error, and the best fitting equation or detection model, along with the corresponding preprocessing method and wavelength position, is selected as the blood immunoglobulin detection model.

[0010] Furthermore, in one embodiment of the present invention, the blood immunoglobulin indicators include blood immunoglobulin indicators IgG, IgA, and IgM.

[0011] Furthermore, in one embodiment of the present invention, the comprehensive evaluation in step A3 includes the following steps: Step A31, for the aforementioned blood immunoglobulin indicators Perform Bartlett's test of sphericity and KMO test to obtain the Bartlett's test of sphericity results. KMO test results ; Step A32, through

[0012] Obtaining the allocation coefficient ; Step A33, through

[0013] Obtain the index weights of the blood immunoglobulin index. ,in Principal component coefficients, The weights are determined by the entropy weighting method. Step A34, through

[0014] Obtain comprehensive blood immunoglobulin index ,in, To standardize blood immunoglobulin levels.

[0015] Furthermore, in one embodiment of the present invention, the principal component coefficients are obtained through the following steps: Step A321, for the aforementioned several blood immunoglobulin indicators Principal component loadings and eigenvalues ​​were obtained using principal component analysis. Step A322: Divide the principal component loading value by the corresponding principal component eigenvalue and take the square root to obtain the principal component coefficients.

[0016] Furthermore, in one embodiment of the present invention, the various fitting methods described in step A6 include sigmoid equation, linear fitting, quadratic fitting, cubic fitting, logarithmic fitting, and exponential fitting.

[0017] Furthermore, in one embodiment of the present invention, the various machine learning methods described in step A7 include Particle Swarm Optimization Support Vector Machine (PSO-SVR), Backpropagation Neural Network (BP-NN), and Extreme Learning Machine (ELM) model.

[0018] Option 2: A system for constructing a broiler blood immunoglobulin detection model, comprising the following modules: Module 1 is used to obtain a sample set, which includes visible chicken manure. Near-infrared reflectance spectra and corresponding blood immunoglobulin levels; Module 2 is used to perform baseline offset correction preprocessing on the visible... The near-infrared reflectance spectrum is preprocessed to obtain the corresponding baseline shift correction preprocessed spectrum; Module 3 is used to comprehensively evaluate the aforementioned blood immunoglobulin indicators to obtain a comprehensive blood immunoglobulin index. Module 4, used for respectively based on the visible The positions of all wavelengths in the near-infrared reflectance spectrum and the baseline offset correction preprocessed spectrum are used to obtain the corresponding three spectral indices. The three spectral indices mentioned include the ratio spectral index RI, the difference spectral index DI, and the normalized spectral index NDVI. Module 5 is used to evaluate the correlation between the three spectral indices and their corresponding comprehensive blood immunoglobulin indicators, and to screen and obtain the visible... The optimal wavelength position for near-infrared reflectance spectra and baseline shift correction preprocessed spectra; Module 6 is used to construct, based on the three spectral indices corresponding to the optimal wavelength position and the comprehensive index of blood immunoglobulins, to obtain data sets DA1, DA2 or DA3 corresponding to the visible-near infrared reflectance spectrum, and data sets DB1, DB2 and DB3 corresponding to the baseline offset correction preprocessed spectrum. Multiple fitting equations were obtained by fitting each of the above data sets using various fitting methods. Module 7 combines data group DB1, data group DB2 and data group DB3 into data group DB, and fits data group DB with various machine learning methods to obtain multiple detection models; Module 8 is used to evaluate the various fitting equations and detection models using the coefficient of determination and root mean square error, and to select the best fitting equation or detection model, along with the corresponding preprocessing method and wavelength position, as the blood immunoglobulin detection model.

[0019] Option 3: An electronic device includes a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; Memory, used to store computer programs; When a processor executes a program stored in memory, it implements the method described in Scheme 1.

[0020] Option 4: A computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the method described in Option 1.

[0021] The present invention provides a method and system for constructing a detection model for broiler blood immunoglobulins, which effectively alleviates the problems of existing broiler blood immunoglobulin detection technologies, such as high contamination during the detection process, high cost, poor model stability, insufficient detection accuracy and high detection complexity when spectral analysis technology is directly applied to broiler blood immunoglobulin detection, and difficulty in meeting the real-time health monitoring needs of modern large-scale farming. Specific beneficial effects include: 1. The method for constructing a broiler blood immunoglobulin detection model according to this invention takes broiler blood immunoglobulin indicators (IgG, IgA, and IgM) as the research object and visible-near-infrared reflectance spectroscopy as the research method. Through a comprehensive evaluation method, it effectively constructs a comprehensive evaluation index for broiler blood immunoglobulins and combines it with the visible-near-infrared reflectance spectrum of chicken feces to establish a comprehensive broiler blood immunoglobulin detection model based on constructed spectral indices. This model effectively solves the problems of high pollution and high cost in existing broiler blood immunoglobulin detection processes, providing technical support for broiler health monitoring and precision farming. This invention reduces evaluation bias caused by fluctuations in single indicators by constructing a comprehensive evaluation index for blood immunoglobulins, and improves the stability of the overall immune status characterization. Furthermore, by combining chicken feces spectra to establish a predictive model, it achieves non-destructive and rapid detection of broiler blood immune function, thereby reducing the operational complexity and time cost of traditional blood testing and better meeting the needs of modern large-scale farming for real-time health monitoring. It effectively solves the problems of insufficient detection accuracy, high detection complexity, and difficulty in meeting the real-time health monitoring requirements of modern large-scale farming in broiler blood immunoglobulin detection.

[0022] 2. The method for constructing a broiler blood immunoglobulin detection model described in this invention selects broiler blood immunoglobulin indicators IgG, IgA, and IgM for comprehensive evaluation. These three immunoglobulins reflect the immune status of broiler blood from different aspects and exhibit synergistic changes during the regulation of broiler blood. If a single indicator or any number of broiler blood immunoglobulin indicators are directly used for evaluation, it is easily affected by individual differences and short-term fluctuations in broilers, leading to inaccurate and unstable evaluation results.

[0023] 3. The method for constructing a broiler blood immunoglobulin detection model according to the present invention is based on the synergistic changes of IgG, IgA, and IgM in broiler blood during humoral immunity. Through an improved comprehensive evaluation method, information is fused from these broiler blood immunoglobulin indicators to construct a comprehensive evaluation index that can characterize the overall humoral immunity level. This transforms the originally interrelated multi-indicator variables into potential variables that can characterize the overall humoral immune function status. It reduces the impact of single-indicator fluctuations and multicollinearity on modeling stability, improves the overall characterization ability of the target variable, and achieves the integration and stable expression of multi-indicator immune data from broiler blood. This reduces the impact of fluctuations in broiler blood immunoglobulin indicators, eliminates multicollinearity interference, and improves the prediction accuracy and engineering application stability of subsequent models, thereby achieving efficient and accurate evaluation of the broiler immune status.

[0024] 4. The method for constructing a broiler blood immunoglobulin detection model described in this invention addresses the fact that the relationship between spectral indices and in vivo immunoglobulin levels is not a simple linear one, but rather influenced by multiple factors such as the body's immune status. Therefore, the relationship typically exhibits non-linear characteristics. Based on an improved comprehensive evaluation, this invention utilizes visible-near-infrared reflectance spectral data from chicken feces and employs a baseline shift correction preprocessing method to eliminate baseline drift interference in the spectrum, highlighting effective spectral information relevant to the detection target. Based on visible... Near-infrared reflectance spectroscopy and baseline shift correction preprocessing spectral wavelength positions yielded three corresponding spectral indices, which were then fitted to the comprehensive blood immunoglobulin index to obtain multiple fitting equations. Data sets DB1, DB2, and DB3 were combined into data set DB, and various machine learning methods were used to fit data set DB to obtain multiple detection models. By comparing the various fitting equations, the mapping and expression ability between spectral indices and the comprehensive broiler blood immunoglobulin index was enhanced. By modeling using multiple machine learning methods, the different machine learning algorithms were fully utilized in processing visible light. Near-infrared reflectance spectroscopy and broiler blood immunoglobulin comprehensive index have different processing effects and generalization abilities when dealing with nonlinear and small sample data.

[0025] Therefore, this invention, by combining fitting equations and machine learning methods, can characterize the response pattern between spectral indices and comprehensive immune indicators under different mathematical mapping spaces, enhance the mapping expression ability between spectral indices and comprehensive immunoglobulin indicators in broiler blood, improve the reliability and applicability of cross-media detection of "chicken feces spectroscopy-blood comprehensive immune indicators", and thus screen out expression models with stronger correlation, higher goodness of fit and better stability. This enables indirect and non-destructive prediction of blood immune function, thereby avoiding the problems of complex and long cycle of traditional blood sampling and testing, improving the real-time performance of immune health monitoring under large-scale farming conditions, and meeting the real-time health monitoring requirements of modern large-scale farming.

[0026] The method described in this invention provides a new technical approach for the rapid and non-destructive detection of poultry immune status. It is applicable to the field of rapid and non-destructive detection of poultry immune status, as well as to the fields of evaluating the immunomodulatory effects of functional feed additives, analyzing big data on vaccine immune responses, and screening diagnostic biomarkers for poultry diseases in the biopharmaceutical industry through data processing. Attached Figure Description

[0027] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 This is a flowchart of the method for constructing a broiler blood immunoglobulin detection model as described in Implementation Method 1; Figure 2 This is a diagram of the visible-near-infrared reflectance spectroscopy acquisition device described in Embodiment 1; Figure 3 This is a graph showing the correlation analysis results as described in Implementation Method Seven; Figure 4 This is the principal component analysis result diagram described in Implementation Method Seven; Figure 5 This is the principal component loading matrix result diagram described in Implementation Method Seven; Figure 6 This is a diagram showing the optimal spectral index wavelength combination position results as described in Implementation Method Seven; Figure 7 This is a graph showing the fitting equation results described in Implementation Method Seven; Figure 8 The following are the fitting equation verification results as described in Implementation Method 7: (a) is the fitting equation verification result for data group DA1; (b) is the fitting equation verification result for data group DA2; (c) is the fitting equation verification result for data group DA3; (d) is the fitting equation verification result for data group DB1; (e) is the fitting equation verification result for data group DB2; and (f) is the fitting equation verification result for data group DB3.

[0028] Figure 9 This is a result diagram of the machine learning method described in Implementation Method Seven.

[0029] Figure label: 1. Computer; 2. White reference plate; 3. Sample; 4. Blackboard; 5. Reflection probe bracket; 6. Light source; 7. Spectrometer; 8. Fiber optic cable. Detailed Implementation

[0030] Various embodiments of the present invention will now be clearly and completely described with reference to the accompanying drawings. The embodiments described with reference to the drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0031] Implementation Method 1: The method for constructing a broiler blood immunoglobulin detection model described in this implementation method, such as... Figure 1 As shown, the method includes the following steps: Step A1: Obtain a sample set, wherein the sample set includes visible chicken manure. Near-infrared reflectance spectra and corresponding blood immunoglobulin levels; Step A2, using a baseline offset correction preprocessing method to preprocess the visible... The near-infrared reflectance spectrum is preprocessed to obtain the corresponding baseline shift correction preprocessed spectrum; Step A3: A comprehensive evaluation of several blood immunoglobulin indicators is performed to obtain a comprehensive blood immunoglobulin index. Step A4, based on the visible The positions of all wavelengths in the near-infrared reflectance spectrum and the baseline offset correction preprocessed spectrum are used to obtain the corresponding three spectral indices. The three spectral indices mentioned include the ratio spectral index RI, the difference spectral index DI, and the normalized spectral index NDVI. Step A5: Evaluate the correlation between the three spectral indices and their corresponding comprehensive blood immunoglobulin indicators, and select the visible... The optimal wavelength position for near-infrared reflectance spectra and baseline shift correction preprocessed spectra; Step A6: Based on the three spectral indices corresponding to the optimal wavelength position, construct data sets DA1, DA2, or DA3 corresponding to the visible-near-infrared reflectance spectrum and the comprehensive blood immunoglobulin index, as well as data sets DB1, DB2, and DB3 corresponding to the baseline offset correction preprocessed spectrum. Multiple fitting equations were obtained by fitting each of the above data sets using various fitting methods. Step A7: Combine the data group DB1, the data group DB2 and the data group DB3 into a data group DB, and fit the data group DB with various machine learning methods to obtain multiple detection models; Step A8: The multiple fitting equations and multiple detection models are evaluated using the coefficient of determination and root mean square error, and the best fitting equation or detection model, along with the corresponding preprocessing method and wavelength position, is selected as the blood immunoglobulin detection model.

[0032] In this embodiment, the ratio spectral index RI among the three spectral indices is obtained through...

[0033] get; Difference Spectral Index (DI)

[0034] get; Normalized Spectral Index (NDVI) via

[0035] Obtain, among which, , Let be the spectral reflectance at any spectral position.

[0036] In this embodiment, step A4 preferably takes place within the wavelength range of 400nm to 900nm, based on the visible light spectrum. All wavelength positions in the near-infrared reflectance spectrum and baseline shift correction preprocessed spectrum were used to obtain the corresponding three spectral indices. The wavelength range of 400nm to 900nm was selected based on a comprehensive consideration of the characteristics of blood immunoglobulin indicators and the overall distribution characteristics of chicken feces spectral information. Changes in the body's immune status can affect intestinal function, which in turn affects the spectral response of chicken feces. These effects are manifested to varying degrees in both the visible and near-infrared regions. Selecting the 400–900nm band can fully utilize the complementary information of visible and near-infrared light.

[0037] In this embodiment, in step A6, data group DA1 corresponds to the ratio RI spectral index of the original spectrum; data group DA2 corresponds to the difference DI spectral index of the original spectrum; data group DA3 corresponds to the normalized NDVI spectral index of the original spectrum; data group DB1 corresponds to the ratio RI spectral index of the baseline offset correction preprocessed spectrum; data group DB2 corresponds to the difference DI spectral index of the baseline offset correction preprocessed spectrum; and data group DB3 corresponds to the normalized NDVI spectral index of the baseline offset correction preprocessed spectrum.

[0038] In this embodiment, the determination coefficient mentioned in step A8 for:

[0039] Root mean square error for:

[0040] in, For the sample size, For the first The actual value of each sample For the first The predicted value for each sample, This is the average of the actual values ​​for all samples.

[0041] In this embodiment, the visible-near-infrared reflectance spectra of the sample set in step S1 are obtained as follows: Step S11: Collect chicken manure and pre-treat the collected chicken manure to obtain pre-treated chicken manure; Step S12: Collect the visible-near infrared reflectance spectrum of the pretreated chicken manure.

[0042] The collection of chicken manure in step S11 includes the following steps: Step S111, Feeding Experiment To accurately detect comprehensive immunoglobulin indicators in broiler blood using visible-near-infrared reflectance spectroscopy of chicken feces, this invention adds different concentrations of *Lactobacillus casei* to the daily drinking water of broilers, resulting in variations in the collected fecal and blood immunological indicators. This is primarily because probiotics can influence the digestion and absorption of nutrients by regulating the body's immune function, improving the intestinal flora structure, and maintaining the integrity of intestinal structure and function. Related studies have shown that adding *Lactobacillus casei* to the diet or drinking water can promote broiler growth performance, improve immunity and antioxidant levels, and influence the intestinal microbial community structure, thereby improving breeding efficiency. Therefore, based on the above research, this invention sets five *Lactobacillus casei* concentration levels: 0 CFU / mL, 1.2 × 10⁻⁶, and 1.2 × 10⁻⁶. 6 CFU / mL, 2.0×10 6 CFU / mL, 1.0×10 7 CFU / mL and 2.0×10 7 The CFU / mL range covers low, medium, and high doses to ensure comparability of experimental data between treatment groups.

[0043] Step S112, collect chicken manure When the broilers were raised to 42 days old, chicken manure samples were collected from each treatment group, and then sealed in plastic bags for storage within each group.

[0044] The pretreatment of chicken manure in step S11 includes: 10 mL of 10% H2SO4 was added to each 100 g chicken manure sample for nitrogen fixation treatment. After 4 days, the samples from each treatment group were thoroughly mixed and dried at 65℃ for 72 h, followed by rehydration for 24 h. After the samples were dried, the chicken manure samples were crushed and passed through a 40-mesh sieve to ensure sample homogeneity, and then sealed and stored for later use.

[0045] The process of collecting the visible-near-infrared reflectance spectrum of the pretreated chicken manure in step S12 is as follows: When acquiring visible-near-infrared reflectance spectra, the following methods are used: Figure 2 The apparatus shown connects the optical fiber of the reflective probe to the spectrometer and the light source, respectively. The reflective probe is then fixed in place using a reflective probe holder, ensuring a 90° angle between the probe and the sample surface. Finally, the light source is preheated for 8 minutes, white plate calibration is performed, and sample spectrum acquisition begins. During the experiment, white plate calibration is repeated every 30 minutes to ensure the stability and reliability of the spectral data. For data acquisition, 50 g of dried chicken manure samples from each treatment group are used as a test sample and evenly spread in a 90 mm diameter petri dish. Eight reflectance spectra are acquired for each sample, resulting in a total of 120 visible-near-infrared reflectance spectra of chicken manure.

[0046] In this embodiment, the method for obtaining the blood immunoglobulin index in the sample set in step S1 is as follows: When broilers reach 42 days of age, blood is collected via the jugular vein. The collected blood is placed in a vacuum container and allowed to coagulate at room temperature. The coagulated blood sample is centrifuged at 3000 rpm for 10 minutes to separate the serum, which is then aliquoted into multiple sterile centrifuge tubes. These tubes are stored at -20°C for subsequent determination of serum immunoglobulins. The levels of chicken immunoglobulins A (IgA), G (IgG), and M (IgM) in the samples are determined using a double-antigen sandwich method. The absorbance (OD value) is measured at 450 nm using an ELISA reader, and the concentrations of immunoglobulins A (IgA), G (IgG), and M (IgM) in the samples are calculated using a standard curve.

[0047] In this embodiment, the processing of spectral data is achieved using existing spectral data processing software tools. For example, AvaSoft 8 spectral acquisition software is used to export the acquired spectral data, Excel 2010 is used to input and analyze the exported basic data, The Unscrambler X 10.4 is used to preprocess the acquired spectral data, Matlab 2023b software is used to train and validate the model, and SPSS 23 is used to construct quantitative equations.

[0048] The method for constructing a broiler blood immunoglobulin detection model described in this embodiment employs a baseline shift correction preprocessing method. The baseline refers to the spectral drift caused by instrument vibration, fluorescence interference, or noise. The presence of a baseline leads to an increase in the relative intensity of some spectral data and masks low-intensity signals, easily resulting in the loss of characteristic variables in the analytical data. Baseline correction achieves pure spectral information estimation by fitting a background baseline, effectively improving the information of low-intensity signal variables. Existing technologies use MSC and SNV preprocessing methods to eliminate physical scattering effects caused by sample particle size, inhomogeneity, and optical path differences. However, considering the problem that weak signals such as immunoglobulins in chicken feces spectra are easily interfered with by instrument fluctuations, this embodiment uses baseline shift correction preprocessing. The advantage of this method is that changes in instrument status, optical path, and environmental conditions are unavoidable, and these factors can interfere with spectral signals. Signals related to broiler blood immunoglobulins are usually weak and easily masked. Baseline shift correction eliminates these effects without altering the shape of the original spectrum, ensuring that the spectral curve reflects the characteristics of the sample itself. It reduces background interference caused by instrument noise and other factors, effectively reducing spectral variations caused by non-sample information, and providing a spectral data foundation with a higher signal-to-noise ratio and a more stable baseline for subsequent modeling.

[0049] Given the complex surface structure of chicken manure samples, an overall baseline shift is easily generated in the original visible-near infrared reflectance spectrum, which affects the extraction of effective spectral information. This invention adopts a baseline shift correction preprocessing method to eliminate the baseline shift of the spectrum, reduce background interference, and improve modeling accuracy and detection stability while maintaining the spectral characteristics.

[0050] In this embodiment, three spectral indices are used, including the ratio spectral index (RI), the difference spectral index (DI), and the normalized spectral index (NDVI). The ratio, difference, and normalized spectral indices are widely used in near-infrared spectral analysis to reduce non-chemical background interference and amplify target features, thereby enhancing the spectral information's responsiveness to changes in the comprehensive immunoglobulin index of broiler blood and providing stable input for subsequent feature screening and modeling.

[0051] In this embodiment, the fitting equation is used to model the relationship between a single optimal spectral index and the comprehensive blood immunoglobulin index, which can directly reflect the quantitative relationship between the spectral index and the comprehensive blood immunoglobulin index, and is suitable for rapid screening and preliminary analysis; while the machine learning method uses the three optimal spectral indices and the comprehensive blood immunoglobulin index under the extracted baseline offset correction preprocessing to establish a model, which can fully integrate the information of multiple spectral indices and capture the spectral response of nonlinearity and multi-factor coupling.

[0052] Implementation Method 2: This implementation method further defines the method for constructing a broiler blood immunoglobulin detection model described in Implementation Method 1. In this implementation method, the blood immunoglobulin indicators include blood immunoglobulin indicators IgG, IgA, and IgM.

[0053] This embodiment further defines Embodiment 1, providing examples of several blood immunoglobulin indicators. Since a single indicator cannot comprehensively reflect the impact of *Lactobacillus casei* on broiler immune proteins, this embodiment comprehensively considers IgG, IgA, and IgM to construct a comprehensive immune protein indicator to fully assess the impact of *Lactobacillus casei* on broiler immune proteins. In existing technologies, poultry immune status assessment typically uses single immunoglobulin indicators (such as IgG or IgA) or immune organ indices as evaluation parameters. These methods often fail to comprehensively reflect the overall functional state of the humoral immune system and easily overlook the synergistic effects and dynamic changes between different immunoglobulins. This embodiment comprehensively considers three key immunoglobulins—IgG, IgA, and IgM—and constructs a comprehensive immune protein indicator through an improved integrated evaluation method. This not only transforms the originally interrelated multi-indicator variables into potential variables that can characterize the overall humoral immune function status but also provides a more comprehensive and systematic assessment of the overall regulatory effect of *Lactobacillus casei* on broiler immune proteins. Compared with traditional single-indicator evaluation, the comprehensive index of this invention can reflect the overall response of the immune system and reduce the impact of fluctuations in a single indicator on the evaluation results, thus providing a scientific and effective technical means for rapid and non-destructive evaluation of the role of probiotics in the immune function of broilers.

[0054] This implementation method further detected and analyzed IgG, IgA, and IgM in the blood of broiler chickens by adding different concentrations of Lactobacillus casei to drinking water, further verifying that different concentrations of Lactobacillus casei have different effects on different immune protein indicators. Through comprehensive evaluation of changes in multiple indicators, the overall influence of different doses of Lactobacillus casei on humoral immunity levels can be clarified, thus providing experimental evidence for constructing comprehensive immune evaluation indicators and improving the accuracy and stability of evaluating the immunomodulatory effects of functional feed additives. This provides a scientific basis for optimizing feed additive usage programs and achieving rapid, non-destructive monitoring of immune status, and offers reliable technical support for evaluating the immunomodulatory effects of functional feed additives.

[0055] Implementation Method 3: This implementation method further defines the method for constructing a broiler blood immunoglobulin detection model as described in Implementation Method 1 or 2. In this implementation method, the comprehensive evaluation in step A3 includes the following steps: Step A31, for several blood immunoglobulin indicators Perform Bartlett's test of sphericity and KMO test to obtain the Bartlett's test of sphericity results. KMO test results ; Step A32, through

[0056] Obtaining the allocation coefficient ; Step A33, through

[0057] Obtain the index weights of the blood immunoglobulin index. ,in Principal component coefficients, The weights are determined by the entropy weighting method. Step A34, through

[0058] Obtain comprehensive blood immunoglobulin index ,in, To standardize blood immunoglobulin levels.

[0059] This embodiment further defines step A3 and explains the comprehensive evaluation. This embodiment determines the weighting method based on the results of Bartlett's sphericity test and KMO test. The weighting method includes principal component analysis, entropy weighting, and a combination of the two.

[0060] Principal Component Analysis (PCA) is a multivariate statistical method based on dimensionality reduction. Through linear transformation, it converts multiple correlated variables into a few uncorrelated composite indicators. This preserves the main information of the original data while eliminating correlations between indicators, achieving objective weighting. Entropy weighting is a classic algorithm for calculating weighted indicators. It is used to determine the dispersion of an indicator. It judges the objective weight by the magnitude of the indicator's variability; that is, the smaller the information entropy, the greater the dispersion, the greater the information content, and the greater the weight for the comprehensive evaluation. A combination of both methods considers both the collaborative information between indicators and the variability characteristics of each indicator itself.

[0061] In this embodiment, blood immunoglobulin levels reflect the body's immune status. However, there are certain correlations and information redundancy among the various indicators, making it difficult to comprehensively characterize immune function through individual analysis. Principal component analysis (PCA) transforms the original indicators into independent principal components, achieving information compression and comprehensive characterization. The entropy weighting method assigns weights to each indicator based on its dispersion and information entropy in the sample, determining the impact of each indicator on the overall evaluation. This combined approach considers both the synergistic relationship between indicators and the information differences of individual indicators, enabling the final comprehensive index to systematically integrate IgG, IgA, and IgM data, providing a reliable data foundation for subsequent rapid and non-destructive immune status assessment based on chicken fecal spectra.

[0062] If the correlation coefficients between variables are very small, or if the eigenvalues ​​of the correlation coefficients are evenly distributed, then it is unnecessary to use principal component analysis for a comprehensive evaluation of this group of observations. Entropy weighting, on the other hand, is suitable for handling problems with weak correlations between indicators. It calculates weights using objective data, avoiding interference from subjective factors and addressing the issue of insufficient adaptability of single weighting methods to data structures.

[0063] This implementation method uses a weighting interval selection method based on the results of the KMO and Bartlett's test of sphericity. The KMO test (Kaiser-Meyer-Olkin Test) is a method used to measure the correlation between variables. The KMO test is typically used to check the suitability of a data sample to determine whether it is suitable for factor analysis or other multivariate data analysis. Bartlett's test of sphericity is used to test the independence between data points; this test is generally performed before factor analysis to determine whether the variables are suitable for factor analysis.

[0064] When KMO < 0.5 or P > 0.05, the correlation between variables is weak and the preconditions for principal component analysis are not met. Therefore, the entropy weight method is used to determine the weights.

[0065] When 0.5 ≤ KMO ≤ 0.7 and p ≤ 0.05, a potential but unstable correlation structure exists between variables. Therefore, by interval normalizing KMO and p values ​​and constructing a function using a product method, α increases synchronously with the structural strength and significance, thus achieving a transition between principal component weighting and entropy weighting, avoiding the instability caused by simply using either method. The product method is used because both indicators need to function simultaneously; if one indicator is poor, the product result will decrease, thus avoiding unreasonably increasing the principal component weight based solely on a high single indicator. When the KMO test value is between 0.5 and 0.7, a comprehensive weight is constructed by combining the two methods to consider both variable structural information and information content differences, thereby improving the robustness and scientific rigor of the comprehensive evaluation results.

[0066] When KMO > 0.7 and P < 0.05, it indicates a strong correlation between variables, and principal component analysis is used to determine the weights.

[0067] Implementation Method Four: This implementation method further defines the method for constructing a broiler blood immunoglobulin detection model described in Implementation Method Three. In this implementation method, the principal component coefficients are obtained through the following steps: Step A321, for the aforementioned several blood immunoglobulin indicators Principal component loadings and eigenvalues ​​were obtained using principal component analysis. Step A322: Divide the principal component loading value by the corresponding principal component eigenvalue and take the square root to obtain the principal component coefficients.

[0068] In this embodiment, the principal component coefficients are used as weights for calculating the principal component scores.

[0069] In this embodiment, principal component analysis is used to construct a comprehensive index of immune proteins to remove redundancy, improve evaluation reliability, and reduce subjective weighting.

[0070] Implementation Method 5: This implementation method further defines the method for constructing a broiler blood immunoglobulin detection model as described in Implementation Method 3. In this implementation method, the various fitting methods mentioned in step A6 include sigmoid equation, linear fitting, quadratic fitting, cubic fitting, logarithmic fitting, and exponential fitting.

[0071] This implementation further defines step A6, providing examples of various fitting methods. The above relationships are chosen for fitting primarily because the S-shaped equation describes the growth process of variables with factors, characterized by a non-linear growth trend of "slow at first, rapid in the middle, and stable at the end." It describes the non-linear change trend of the immunoglobulin comprehensive index with the increase of the spectral index, exhibiting a "slow at first, rapid in the middle, and stable at the end." Linear fitting is suitable for data with stable and monotonic relationships between variables, offering advantages such as simple model form, clear parameter meanings, and easy interpretation of results. It can be used to characterize the stable and monotonic effects of spectral index changes on the immunoglobulin comprehensive index in broiler blood. Quadratic and cubic polynomial fitting can effectively characterize the non-linear change features and potential inflection points in the data, making it more suitable for describing complex trends and reflecting the enhancement, inhibition, or inflection point changes of the immunoglobulin comprehensive index caused by spectral index changes. Logarithmic fitting is suitable for relationships where the rate of change gradually slows down with the increase of the independent variable, and is applicable to situations where the increase of the immunoglobulin comprehensive index gradually slows down with the increase of the spectral index. Exponential fitting is suitable for characterizing rapid growth or decay processes, and is applicable to scenarios where immunoglobulins are highly sensitive to changes in the spectral index and exhibit accelerated growth or decay trends. By comparing and analyzing the fitting effects of different models, selecting the model with the best fit and reasonable structure can help avoid excessive model complexity while ensuring prediction accuracy, thereby improving the stability, reliability and interpretability of the model.

[0072] By comparing the goodness of fit of different function models and selecting the optimal model, we can not only improve the detection accuracy of spectral indices for comprehensive immunoglobulin indicators, but also improve the stability and interpretability of the model, thereby better serving the quantitative detection and evaluation of blood immunoglobulin indicators.

[0073] Implementation Method Six: This implementation method further defines the method for constructing a broiler blood immunoglobulin detection model as described in Implementation Method Three. In this implementation method, the various machine learning methods mentioned in step A7 include Particle Swarm Optimization Support Vector Machine (PSO-SVR), Backpropagation Neural Network (BP-NN), and Extreme Learning Machine (ELM) model.

[0074] This implementation further defines step A7, explaining various machine learning methods. The main reason is that PSO-SVR can robustly fit nonlinear relationships under small sample and high-dimensional spectral conditions, exhibiting strong noise resistance. SVR, or Support Vector Regression, is based on statistical learning theory and the principle of structural risk minimization. It uses a kernel function to map input samples from the original feature space to a higher-dimensional feature space, transforming the original nonlinear regression problem into a dual quadratic programming problem in a high-dimensional feature space. It features small sample size, nonlinearity, and high-dimensional pattern recognition. The predictive performance of SVR mainly depends on the selection of two key parameters: the penalty parameter c and the kernel parameter g. If the penalty factor c is too large, it will cause overfitting, reducing the generalization performance of the prediction model; if this value is too small, it will increase the model's tolerance for error, easily leading to underfitting. If the kernel parameter g is too large, the influence between support vectors will be too strong, reducing the algorithm's accuracy; if this value is too small, the connection between support vectors will be loose, resulting in poor model generalization performance. Particle Swarm Optimization (PSO) is an intelligent algorithm inspired by the search strategies employed in the defense and predation behaviors of biological communities such as flocks of birds and schools of fish. PSO is easy to implement, converges quickly, and requires few parameters, making it a highly efficient search method. Compared to other optimization algorithms, it exhibits stronger randomness within biological populations and requires fewer control variables. Backpropagation neural networks (BPNNs) can capture complex nonlinear relationships and are a type of multilayer feedforward neural network commonly used in supervised learning. By simulating the information transmission mechanism of biological neurons, they continuously adjust network weights and biases through error backpropagation to minimize prediction errors. Through learning and training on sample data, BPNNs can construct a mapping model between input and output without needing to understand the underlying physical mechanisms. BPNNs possess powerful nonlinear mapping capabilities, effectively extracting and fitting nonlinear relationships between complex variables.

[0075] Extreme Learning Machines (ELM) are suitable for rapid modeling due to their fast training speed and strong generalization ability. Unlike traditional neural network training methods, ELM avoids frequent adjustments to the input weights and hidden element biases during training, and can obtain a unique optimal solution. Its purpose is to reduce the complexity of training by using random weights assigned to the hidden layers of the neural network. ELM overcomes the shortcomings of traditional neural network methods, such as long training periods and sensitivity to learning rates, with its rapid learning ability, outstanding generalization ability, and convenient parameter setting.

[0076] Implementation Method Seven: The method for constructing the broiler blood immunoglobulin detection model in this implementation method is based on the method for constructing the broiler blood immunoglobulin detection model described in Implementation Method One, combined with several blood immunoglobulin indicators optimized in Implementation Method Two, the comprehensive evaluation optimized in Implementation Method Three, the principal component analysis weights optimized in Implementation Method Four, the various fitting methods optimized in Implementation Method Five, and the various machine learning methods optimized in Implementation Method Six.

[0077] In this embodiment, a correlation analysis was performed on three indicators: blood immunoglobulin IgG, blood immunoglobulin IgA, and blood immunoglobulin IgM. The results are as follows: Figure 3 As shown in the figure, significant positive correlations were found among IgG, IgA, and IgM. The correlation coefficients between IgG and both IgA and IgM were 0.77; the correlation coefficient between IgA and IgM was 0.97, indicating a high degree of correlation among the indicators.

[0078] Correlation analysis showed a significant correlation between IgG, IgA, and IgM, indicating a strong relationship among the indicators (p < 0.05, KMO test value 0.70006). Therefore, principal component analysis (PCA) was suitable for data analysis. PCA was performed on IgG, IgA, and IgM, and the results are as follows: Figure 4 As shown. Based on the principle that the eigenvalue is ≥1 and the cumulative variance contribution rate is ≥85%, the first principal component was extracted, with a cumulative variance contribution rate of 89.3457%, indicating that the extracted first principal component contains 89.3457% of the original data information.

[0079] Figure 5 The principal component loading matrices for each experimental indicator show that the loading coefficients of IgG, IgA, and IgM on principal component 1 are 0.8908, 0.9707, and 0.9719, respectively, indicating that these three indicators have large positive coefficient values. When the first principal component is large, the values ​​of these three indicators are also large. The communality (common factor variance) of the three is 0.7935, 0.9423, and 0.9446, respectively, indicating that principal component 1 has a high information retention rate. Therefore, principal component 1 can serve as a core factor comprehensively reflecting changes in immunoglobulins in broiler blood, providing a basis for constructing a comprehensive immunoglobulin index.

[0080] Divide the principal component loadings of each test index by the square root of the corresponding principal component eigenvalues ​​to obtain the principal component coefficients for each test index. Use these principal component coefficients as weights to establish the formula for calculating the principal component score.

[0081] In the formula, F represents the principal component score, and X1, X2, and X3 represent the standardized values ​​of IgG, IgA, and IgM, respectively.

[0082] In this embodiment, since only Principal Component 1 is extracted for the construction of the comprehensive index, its variance contribution rate is 89.35%. Therefore, the comprehensive score is equivalent to the score of Principal Component 1, and no further weighting is required. The comprehensive immune protein index is constructed as follows:

[0083] In this embodiment, to clarify the correlation between the ratio, difference, and normalized spectral index and the comprehensive immunoglobulin index, this study used the correlation matrix method to perform correlation analysis between the three spectral indices and the comprehensive immunoglobulin index, respectively. The wavelength positions i and j where the maximum correlation coefficient was located were taken as the optimal wavelength combination. The results are as follows: Figure 6 As shown.

[0084] To clarify the quantitative relationship between the optimal spectral index and the comprehensive immunoglobulin index under the original spectra and B pretreatment, this study selected 80% of the sample data, using the optimal spectral index as the independent variable and the comprehensive immunoglobulin index as the dependent variable. Linear, quadratic, cubic, logarithmic, and exponential equations were fitted, and the results are as follows: Figure 7 As shown.

[0085] Among them, the quantitative equations established using the DI spectral index for both the original spectrum and the B preprocessed spectrum showed the best results, with R² of 0.861 and F-value of 286.96 for both. The established equations were all... In this invention, the F-value is used to evaluate the goodness of different quantitative equations. The larger the F-value, the more obvious the fitting effect, thus providing a basis for screening the optimal quantitative equation.

[0086] To evaluate the quantitative equation for the comprehensive immunoglobulin index constructed using the spectral index, this study selected the remaining 20% ​​of the samples for equation validation. The predicted values ​​of the comprehensive immunoglobulin index were calculated using the fitted quantitative equations and compared with the measured values. The results are as follows: Figure 8 As shown in the figure. The results show that the DI spectral index R2 is the highest for both the original spectrum and the B preprocessed spectrum, with R2 of 0.8454 and RMSE of 0.8245 for both in the test set.

[0087] To further explore the predictive performance of the optimal spectral index, the optimal spectral index was used as input and the comprehensive immunoglobulin index as output. The training and test sets were randomly divided in a 4:1 ratio. PSO-SVR, BP-NN, and ELM models were established, and the results are as follows: Figure 9 As shown in the figure, the ELM model performed best, with Rp2 and RMSEp values ​​of 0.81 and 0.5913 respectively on the test set. This demonstrates that visible-near-infrared reflectance spectroscopy can achieve rapid detection of comprehensive immunoglobulin indicators in broiler blood.

Claims

1. A method for constructing a broiler chicken blood immunoglobulin detection model, characterized in that, The method includes the following steps: Step A1: Obtain a sample set, wherein the sample set includes visible chicken manure. Near-infrared reflectance spectra and corresponding blood immunoglobulin levels; Step A2, using a baseline offset correction preprocessing method to preprocess the visible... The near-infrared reflectance spectrum is preprocessed to obtain the corresponding baseline shift correction preprocessed spectrum; Step A3: A comprehensive evaluation of several blood immunoglobulin indicators is performed to obtain a comprehensive blood immunoglobulin index. Step A4, based on the visible The positions of all wavelengths in the near-infrared reflectance spectrum and the baseline offset correction preprocessed spectrum are used to obtain the corresponding three spectral indices. The three spectral indices mentioned include the ratio spectral index RI, the difference spectral index DI, and the normalized spectral index NDVI. Step A5: Evaluate the correlation between the three spectral indices and their corresponding comprehensive blood immunoglobulin indicators, and select the visible... The optimal wavelength position for near-infrared reflectance spectra and baseline shift correction preprocessed spectra; Step A6: Based on the three spectral indices corresponding to the optimal wavelength position, construct data sets DA1, DA2, or DA3 corresponding to the visible-near-infrared reflectance spectrum and the comprehensive blood immunoglobulin index, as well as data sets DB1, DB2, and DB3 corresponding to the baseline offset correction preprocessed spectrum. Multiple fitting equations were obtained by fitting each of the above data sets using various fitting methods. Step A7: Combine the data group DB1, the data group DB2 and the data group DB3 into a data group DB, and fit the data group DB with various machine learning methods to obtain multiple detection models; Step A8: The multiple fitting equations and multiple detection models are evaluated using the coefficient of determination and root mean square error, and the best fitting equation or detection model, along with the corresponding preprocessing method and wavelength position, is selected as the blood immunoglobulin detection model.

2. The method for constructing a broiler blood immunoglobulin detection model according to claim 1, characterized in that, The blood immunoglobulin indicators include blood immunoglobulin IgG, blood immunoglobulin IgA, and blood immunoglobulin IgM.

3. The method for constructing a broiler blood immunoglobulin detection model according to claim 1 or 2, characterized in that, The comprehensive evaluation described in step A3 includes the following steps: Step A31, for several blood immunoglobulin indicators Perform Bartlett's test of sphericity and KMO test to obtain the Bartlett's test of sphericity results. KMO test results ; Step A32, through Obtaining the allocation coefficient ; Step A33, through Obtain the index weights of the blood immunoglobulin index. ,in Principal component coefficients, The weights are determined by the entropy weighting method. Step A34, through Obtain comprehensive blood immunoglobulin index ,in, To standardize blood immunoglobulin levels.

4. The method for constructing a broiler blood immunoglobulin detection model according to claim 3, characterized in that, The principal component coefficients are obtained through the following steps: Step A321, for the aforementioned several blood immunoglobulin indicators Principal component loadings and eigenvalues ​​were obtained using principal component analysis. Step A322: Divide the principal component loading value by the corresponding principal component eigenvalue and take the square root to obtain the principal component coefficients.

5. The method for constructing a broiler blood immunoglobulin detection model according to claim 3, characterized in that, The various fitting methods described in step A6 include S-shaped equations, linear fitting, quadratic fitting, cubic fitting, logarithmic fitting, and exponential fitting.

6. The method for constructing a broiler blood immunoglobulin detection model according to claim 3, characterized in that, The various machine learning methods described in step A7 include Particle Swarm Optimization Support Vector Machine (PSO-SVR), Backpropagation Neural Network (BP-NN), and Extreme Learning Machine (ELM) model.

7. A system for constructing a broiler chicken blood immunoglobulin detection model, characterized in that, Includes the following modules: Module 1 is used to obtain a sample set, which includes visible chicken manure. Near-infrared reflectance spectra and corresponding blood immunoglobulin levels; Module 2 is used to perform baseline offset correction preprocessing on the visible... The near-infrared reflectance spectrum is preprocessed to obtain the corresponding baseline shift correction preprocessed spectrum; Module 3 is used to comprehensively evaluate several blood immunoglobulin indicators and obtain a comprehensive blood immunoglobulin index. Module 4, used for respectively based on the visible The positions of all wavelengths in the near-infrared reflectance spectrum and the baseline offset correction preprocessed spectrum are used to obtain the corresponding three spectral indices. The three spectral indices mentioned include the ratio spectral index RI, the difference spectral index DI, and the normalized spectral index NDVI. Module 5 is used to evaluate the correlation between the three spectral indices and their corresponding comprehensive blood immunoglobulin indicators, and to screen and obtain the visible... The optimal wavelength position for near-infrared reflectance spectra and baseline shift correction preprocessed spectra; Module 6 is used to construct, based on the three spectral indices corresponding to the optimal wavelength position and the comprehensive index of blood immunoglobulins, to obtain data sets DA1, DA2 or DA3 corresponding to the visible-near infrared reflectance spectrum, and data sets DB1, DB2 and DB3 corresponding to the baseline offset correction preprocessed spectrum. Multiple fitting equations were obtained by fitting each of the above data sets using various fitting methods. Module 7 combines data group DB1, data group DB2 and data group DB3 into data group DB, and fits data group DB with various machine learning methods to obtain multiple detection models; Module 8 is used to evaluate the various fitting equations and detection models using the coefficient of determination and root mean square error, and to select the best fitting equation or detection model, along with the corresponding preprocessing method and wavelength position, as the blood immunoglobulin detection model.

8. An electronic device, characterized in that, It includes a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; Memory, used to store computer programs; A processor, when executing a program stored in memory, implements the method of any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the method of any one of claims 1-6.

Images