Method, system and storage medium for recognizing abnormal serum status based on image analysis

Abstract

The present invention relates to the technical field of serum image analysis, and in particular to a serum abnormal state identification method, system and storage medium based on image analysis, comprising the following steps: using OCT technology to image the serum sample to be analyzed, and obtaining multi-level serum image data; performing hierarchical image segmentation on the collected multi-level serum sample image; establishing a mathematical model to analyze and standardize the optical and biochemical properties of the serum sample, and then extracting key features; based on the extracted key features, applying pattern recognition technology to distinguish normal and abnormal biochemical indicators, thereby identifying abnormal states in serum; and displaying the identified abnormal state in a graphical user interface, including information on the abnormal type and severity. The present invention, by combining the analysis of spectral characteristics and scattering characteristics, can comprehensively consider a variety of physical and chemical information in the serum sample, and enhance the sensitivity and recognition ability to abnormal states.

Description

Technical Field

[0001] The present invention relates to the technical field of serum image analysis, and in particular to a serum abnormal state recognition method, system and storage medium based on image analysis. Background Art

[0002] In the field of modern medical diagnosis, the analysis of serum samples is crucial for the early detection and diagnosis of a variety of diseases. Traditional serum analysis techniques mainly rely on chemical reactions and the detection of biomarkers, which, although effective, may lack sufficient sensitivity or fail to provide detailed information about biochemical changes in some cases. In addition, conventional methods usually require complex pretreatment and a long analysis process, which limits their application in emergency and rapid diagnosis.

[0003] With the development of optical imaging technology, especially optical coherence tomography (OCT) technology, new possibilities have been provided for serum sample analysis. OCT technology can provide high-resolution deep-level images, making it possible to intuitively observe the changes in the microstructure in serum samples. However, although OCT technology provides high-quality image data, how to effectively extract and utilize information from these data remains a challenge. Especially when dealing with complex biochemical components, single-level analysis often fails to reveal the multidimensional characteristics of the sample. Therefore, a new method for serum abnormality identification based on image analysis is proposed. Summary of the invention

[0004] Based on the above objectives, the present invention provides a method, system and storage medium for identifying abnormal serum status based on image analysis.

[0005] The method for identifying abnormal serum status based on image analysis comprises the following steps:

[0006] S1: Use optical coherence tomography (OCT) technology to image the serum sample to be analyzed and obtain multi-layer serum image data, where each layer represents different depth components in the serum;

[0007] S2: Perform hierarchical segmentation on the collected multi-level serum sample images and remove image noise caused by particles and impurities;

[0008] S3: Establish a mathematical model to analyze and standardize the optical and biochemical properties of serum samples, and then extract key features from each layer of the hierarchical segmented image. The key features include spectral characteristics and scattering characteristics, which are used to identify and distinguish normal and abnormal biochemical indicators;

[0009] S4: Based on the extraction of spectral characteristics and scattering characteristics, pattern recognition technology is applied to distinguish normal and abnormal biochemical indicators, thereby identifying abnormal states in serum;

[0010] S5: Display the identified abnormal status in a graphical user interface, including information on the abnormality type and severity.

[0011] Furthermore, the S1 specifically includes:

[0012] S11, sample preparation: placing the serum sample to be analyzed in a transparent or translucent container, which is transparent to the light band of OCT imaging so as to achieve effective light wave penetration and reflection;

[0013] S12, equipment setting: adjust the scanning parameters of the OCT equipment, including scanning depth, resolution, and scanning speed, to adapt to the characteristics of the serum sample, and set the wavelength and power of the light source to maximize the recognition ability of different components in the serum;

[0014] S13, imaging scan: start the OCT device to perform continuous scanning, starting from the surface layer of the serum sample and gradually going deeper to a preset depth. Each scan acquires image data in a layer of serum, reflecting the optical properties and biochemical composition of the layer;

[0015] S14, data acquisition and processing: collect image data obtained from scanning at various depth levels and perform preliminary data processing, including signal enhancement and contrast adjustment, to improve image quality and analysis accuracy.

[0016] Furthermore, the S2 receives the collected original multi-layer serum image data and makes a preliminary evaluation to determine whether the image quality and the clarity of each layer of the image meet the standards. The S2 also includes:

[0017] S21: determining image segmentation parameters based on visible optical properties in the serum sample, including reflection intensity and scattering pattern, including setting a segmentation threshold and selecting a segmentation algorithm, wherein the segmentation algorithm includes a threshold-based method, a region growing algorithm, or an edge-based segmentation technique;

[0018] S22: applying the selected image segmentation technology to segment each layer of the image into a plurality of segmentation regions, each segmentation region representing a biochemical component or structure in the serum sample, and ensuring that each segmentation region represents its corresponding deep level feature;

[0019] S23: Evaluate the accuracy and completeness of the segmentation results to check whether any areas are misclassified or missed;

[0020] S24: Organize and annotate the segmented image regions to ensure that the data of each segmented region is correctly associated with its corresponding depth level and biochemical characteristics.

[0021] Furthermore, the S21 specifically includes:

[0022] S211, optical properties analysis:

[0023] Reflection intensity analysis: A custom script was used to analyze the reflection intensity in the multi-level serum image data, calculate the grayscale value of each pixel, and evaluate the distribution of reflection intensity within the entire image;

[0024] Scattering pattern recognition: Scattering patterns indicate the characteristics of different substances in the sample. The distribution and intensity of scattered light, as well as changes at different depth levels, are analyzed to identify structural boundaries or composition differences.

[0025] S212, set segmentation threshold:

[0026] Based on the statistics of reflection intensity, the segmentation threshold is selected to distinguish different biochemical components, with high reflection intensity indicating dense tissue areas and low reflection intensity indicating tissue areas with high water content;

[0027] Histogram analysis is used to assist in determining the optimal segmentation threshold. Different peaks on the histogram represent the reflection characteristics of different materials.

[0028] S213, select segmentation algorithm:

[0029] Threshold-based segmentation: If the image reflection intensity distribution is uniform, use a global threshold method such as the Otsu algorithm to automatically determine the segmentation threshold;

[0030] Region growing algorithm: For images with complex scattering patterns, the region growing algorithm is used;

[0031] Edge-based segmentation: When the scattering pattern shows obvious edges, edge detection algorithms (such as Sobel or Canny algorithms) are used to identify and segment different structural boundaries.

[0032] Furthermore, the S3 specifically includes:

[0033] S31: Define a mathematical model to explain and simulate the changes in spectral characteristics and scattering characteristics in serum samples. The mathematical model is used to describe the effects of different chemical components on spectral absorption and scattering, and parameterize the model, where the parameters include spectral absorbance, scattering coefficient, layer depth and sample component concentration;

[0034] S32: Apply feature extraction algorithms to automatically extract and quantify spectral and scattering properties from hierarchically segmented images, including absorbance peak wavelength, peak width, and total scattering intensity;

[0035] S33: performing Z-score standardization on the extracted feature data, converting the data into a format with zero mean and unit variance, and then performing normalization processing to make the weights of different feature data consistent;

[0036] S34: Cross-validation techniques are used to test the accuracy and generalization ability of the mathematical model and feature extraction algorithm, independent test data sets are used to evaluate the predictive ability and stability of the model, and the extracted feature data are integrated into a comprehensive analysis framework for the final abnormal state identification and classification.

[0037] Furthermore, the mathematical model is established based on the Beer-Lambert law, which is expressed as: ;in, is at wavelength The absorbance at is the molar absorptivity, obtained based on experimental fit (depends on wavelength and chemical composition), is the concentration of the solute, is the light path length, i.e. the width of the sample container;

[0038] The feature extraction algorithm includes:

[0039] Spectral feature extraction: Use spectral imaging technology to obtain the spectral image of the serum sample, record the absorbance at each wavelength, identify the absorbance peak at each wavelength, the absorbance peak corresponds to the maximum absorption point of the specific chemical component, and use the local maximum search algorithm:

[0040] ,in It is compared within a small range around the wavelength.

[0041] Scattering property extraction: For scattering properties, the scattering intensity of serum samples at different wavelengths is quantified. The scattering intensity is measured by measuring the intensity of light deviating from the original propagation direction. Using Mie scattering theory, the scattering intensity is calculated as:

[0042] in, is at wavelength and scattering angle The scattering intensity under is the incident light intensity, is the Mie scattering amplitude function;

[0043] The Mie scattering amplitude function and The calculation uses numerical methods as follows:

[0044] Particle parameter setting: First set the refractive index of the particle , particle radius and wavelength of light;

[0045] Size parameter and relative refractive index: in is the size parameter, is the relative refractive index of the particle and the medium;

[0046] The scattering coefficient is obtained by solving the Mie mathematical equation and ;

[0047] The scattering amplitude function and It is expressed as:

[0048] in, and is an angular function, dependent on the scattering angle .

[0049] Furthermore, the pattern recognition technology in S4 adopts a spectral clustering algorithm, which specifically includes:

[0050] S41: Feature vector construction, combining the extracted spectral characteristics and scattering characteristics into a feature vector, standardizing or normalizing the features to ensure that the data are on the same scale;

[0051] S42: Construct a similarity matrix based on the similarity between feature vectors, and use the Gaussian kernel function to calculate the similarity between two samples:

[0052] in, and The samples are and The feature vector of contains the spectral characteristics and scattering characteristics data, is the bandwidth parameter of the kernel function, which determines the sensitivity of the function and is optimized through cross-validation method. Representation sample and The square of the Euclidean distance between them;

[0053] S43: Spectral clustering algorithm implementation, calculation of degree matrix , the degree matrix is a diagonal matrix where each element is the similarity matrix No. Sum of row elements: , calculate the Laplacian matrix based on the similarity matrix ,in, is a diagonal matrix, each element of which is a matrix No. The sum of the rows;

[0054] S44: Eigendecomposition and clustering, perform eigendecomposition of the Laplacian matrix L, extract its front The eigenvector corresponding to the smallest non-zero eigenvalue is used to describe the low-dimensional embedding of the data. Each point in the eigenvector is regarded as a point in a new low-dimensional space. Each point in the eigenvector is clustered using k-means clustering. Each cluster represents a group of samples with similar spectral and scattering characteristics.

[0055] S45: Identify abnormal states, analyze the spectral characteristics and scattering characteristics of samples in each cluster, determine which clusters represent normal states and which represent abnormal states. Abnormal clusters will show spectral or scattering characteristics that are significantly different from normal biochemical indicators. Based on the clustering results, mark the samples belonging to abnormal clusters as abnormal states.

[0056] Furthermore, the bandwidth parameter of the kernel function The selection method is to set It is the reciprocal of the average distance of all sample pairs, that is:

[0057] The serum abnormal state recognition system based on image analysis is used to implement the above-mentioned serum abnormal state recognition method based on image analysis, and includes the following modules:

[0058] Image acquisition module: An automated robotic arm is configured to place and fix the serum sample to be analyzed in a transparent or translucent container. The OCT device is integrated, including a control unit that can adjust the scanning parameters and a light source wavelength and power regulator that adapts to the characteristics of the serum sample. The module automatically and continuously scans from the surface layer of the serum sample to a preset depth. The image data of each layer is captured to obtain multi-layer serum image data.

[0059] Image processing and segmentation module: receives multi-layer serum image data, performs signal enhancement and contrast adjustment, optimizes image quality, automatically sets segmentation parameters based on reflection intensity and scattering pattern, and selects segmentation algorithm to process each layer of image;

[0060] Feature extraction and analysis module: Based on the defined mathematical model, it automatically extracts spectral characteristics and scattering characteristics from the segmented images at each level, performs standardization and regularization processing to unify the weights of different feature data, and prepares them for pattern recognition;

[0061] Pattern recognition and diagnosis module: Using spectral clustering algorithm, construct similarity matrix according to the similarity of feature vectors, perform feature decomposition and clustering, automatically distinguish normal and abnormal biochemical indicators, and identify abnormal status based on clustering results;

[0062] User Interface and Control Module: Provides an operational interface that allows medical professionals to input sample information, view real-time imaging and processing progress, adjust analysis parameters, and view final diagnostic results and reports.

[0063] A computer-readable storage medium can be read by a computer, the storage medium stores a computer program, and the computer program is executed by a processor to implement the above method.

[0064] Beneficial effects of the present invention:

[0065] The present invention can perform a more detailed analysis on each layer by segmenting the serum sample image into different layers. Because various biochemical components in serum may show different optical properties (such as refractive index, scattering rate, etc.) at different layers, stratification can help identify these properties more clearly, thereby improving the accuracy of the overall analysis. Hierarchical segmentation helps to extract more representative features from each specific layer, such as spectral characteristics and scattering characteristics of a specific layer. These features are the key to identifying abnormal biochemical indicators. If hierarchical segmentation is not performed, mixed hierarchical characteristics may cause the loss or confusion of key features. Stratification can effectively reduce the complexity of the data to be processed.

[0066] The present invention uses multi-level serum image data acquired by optical coherence tomography (OCT) technology, and each layer reflects the biochemical components at different depths in the serum, providing rich information for accurate analysis. This multi-level data acquisition and analysis helps to reveal complex biochemical changes, especially the subtle changes that may occur under pathological conditions. Various biochemical components are effectively segmented through advanced image segmentation technology, ensuring segmentation accuracy and integrity. This precise image processing technology ensures that the features extracted from the image are accurate and comprehensive, thereby improving the reliability of subsequent analysis.

[0067] The present invention, by combining the analysis of spectral characteristics and scattering characteristics, can comprehensively consider a variety of physical and chemical information in serum samples, and enhance the sensitivity and recognition ability of abnormal states, especially in the early stages of the disease, where small biochemical changes may only appear in spectral or scattering data, and this method can effectively capture these signals, and use spectral clustering algorithms to further enhance the accuracy of pattern recognition, especially when dealing with high-dimensional and complex data sets. Spectral clustering algorithms identify similarities by exploring the essential structure of data, so that even small abnormalities can be accurately classified, thus providing the possibility of early diagnosis and treatment. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only for the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0069] Figure 1 A schematic diagram of a method flow chart of an embodiment of the present invention;

[0070] Figure 2 Schematic diagram of system function modules according to an embodiment of the present invention. DETAILED DESCRIPTION

[0071] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with specific embodiments.

[0072] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0073] like Figure 1 As shown, the method for identifying abnormal serum status based on image analysis comprises the following steps:

[0074] S1: Use optical coherence tomography (OCT) technology to image the serum sample to be analyzed and obtain multi-layer serum image data, where each layer represents different depth components in the serum;

[0075] S2: Perform hierarchical segmentation on the collected multi-level serum sample images and remove image noise caused by particles and impurities;

[0076] S3: Establish a mathematical model to analyze and standardize the optical and biochemical properties of serum samples, and then extract key features from each layer of the hierarchical segmented image. The key features include spectral characteristics and scattering characteristics, which are used to identify and distinguish normal and abnormal biochemical indicators;

[0077] S4: Based on the extraction of spectral characteristics and scattering characteristics, pattern recognition technology is applied to distinguish normal and abnormal biochemical indicators, thereby identifying abnormal states in serum;

[0078] S5: Display the identified abnormal status in a graphical user interface, including information on the abnormality type and severity.

[0079] When identifying and distinguishing normal and abnormal biochemical indicators, different levels may show different abnormal characteristics. This multi-level information provides support for complex decision-making, making the diagnosis more comprehensive and reliable.

[0080] S1 specifically includes:

[0081] S11, sample preparation: placing the serum sample to be analyzed in a transparent or translucent container, which is transparent to the light band of OCT imaging so as to achieve effective light wave penetration and reflection;

[0082] S12, equipment setting: adjusting the scanning parameters of the OCT equipment, including scanning depth, resolution, and scanning speed, to suit the characteristics of the serum sample, and setting the wavelength and power of the light source to maximize the ability to identify different components in the serum;

[0083] S13, imaging scan: start the OCT device to perform continuous scanning, starting from the surface layer of the serum sample and gradually going deeper to a preset depth. Each scan acquires image data in a layer of serum, reflecting the optical properties and biochemical composition of the layer;

[0084] S14, data acquisition and processing: collect image data obtained from scanning at various depth levels and perform preliminary data processing, including signal enhancement and contrast adjustment, to improve image quality and analysis accuracy.

[0085] S2 receives the collected original multi-layer serum image data and makes a preliminary evaluation to determine whether the image quality and the clarity of each layer of the image meet the standards. S2 also includes:

[0086] S21: determining image segmentation parameters based on visible optical properties in the serum sample, including reflection intensity and scattering pattern, including setting a segmentation threshold and selecting a segmentation algorithm, wherein the segmentation algorithm includes a threshold-based method, a region growing algorithm, or an edge-based segmentation technique;

[0087] S22: applying the selected image segmentation technology to segment each layer of the image into a plurality of segmentation regions, each segmentation region representing a biochemical component or structure in the serum sample, and ensuring that each segmentation region represents its corresponding deep level feature;

[0088] S23: Evaluate the accuracy and completeness of the segmentation results to check whether any areas are misclassified or missed;

[0089] S24: Organize and annotate the segmented image regions to ensure that the data of each segmented region is correctly associated with its corresponding depth level and biochemical characteristics.

[0090] S21 specifically includes:

[0091] S211, optical properties analysis:

[0092] Reflection intensity analysis: A custom script was used to analyze the reflection intensity in the multi-level serum image data, calculate the grayscale value of each pixel, and evaluate the distribution of reflection intensity within the entire image;

[0093] Scattering pattern recognition: Scattering patterns indicate the characteristics of different substances in the sample. The distribution and intensity of scattered light, as well as changes at different depth levels, are analyzed to identify structural boundaries or composition differences.

[0094] S212, set segmentation threshold:

[0095] Based on the statistics of reflection intensity, the segmentation threshold is selected to distinguish different biochemical components, with high reflection intensity indicating dense tissue areas and low reflection intensity indicating tissue areas with high water content;

[0096] Use histogram analysis to assist in determining the optimal segmentation threshold. On the histogram, different peaks represent the reflectance characteristics of different materials. Selecting an appropriate threshold can help separate these peaks, thereby achieving more accurate regional segmentation.

[0097] S213, select segmentation algorithm:

[0098] Threshold-based segmentation: If the image reflection intensity distribution is uniform, use a global threshold method such as the Otsu algorithm to automatically determine the segmentation threshold. This method is suitable for segmenting areas with obvious differences.

[0099] Region growing algorithm: For images with complex scattering patterns, a region growing algorithm is used, which starts from a seed point and gradually grows to adjacent pixels according to the set similarity criteria (such as light intensity, texture, etc.);

[0100] Edge-based segmentation: When the scattering pattern shows obvious edges, edge detection algorithms (such as Sobel or Canny algorithms) are used to identify and segment different structural boundaries;

[0101] After determining the preliminary segmentation threshold and segmentation algorithm, different parameter combinations are tested experimentally and iteratively to find the best image segmentation configuration.

[0102] Histogram analysis helps determine the best threshold to distinguish different areas by analyzing the brightness distribution of the image. It is effective for serum sample image segmentation because the reflection intensity of different components may be different. Here are the steps to use histogram analysis to assist in determining the best threshold:

[0103] First, a histogram is constructed for the image to be analyzed. A histogram is a graphical representation of the frequency of each brightness value in the image. For grayscale images, brightness values ​​range from 0 to 255. It is expressed as: ,in, is a specific grayscale value, The gray value in the image is The number of pixels.

[0104] Observe the shape of the histogram, paying special attention to the peaks and valleys. Each peak represents a major gray level in the image, while the valley represents the dividing point between two different components.

[0105] Determine the threshold: Use Otsu's method to automatically find the optimal threshold. Otsu's method determines the threshold by minimizing the intra-class variance after image binarization or maximizing the inter-class variance. The inter-class variance is defined as: ,in, is the threshold value, and The thresholds are The ratio of pixels below and above to the total pixels, and The thresholds are The average gray value of the pixels below and above.

[0106] The steps for calculating the between-class variance are as follows:

[0107] Calculate the overall average grayscale of the image ,in is the total number of pixels in the image.

[0108] Calculating the threshold Weights below and above:

[0109] ;

[0110] Calculating the threshold Average grayscale below and above:

[0111]

[0112] Calculate the between-class variance and find the maximum The threshold value.

[0113] Apply threshold: Apply the calculated threshold to the image, perform binarization, and segment the image into regions reflecting different components.

[0114] S3 specifically includes:

[0115] S31: Define a mathematical model to explain and simulate the changes in spectral characteristics and scattering characteristics in serum samples. The mathematical model is used to describe the effects of different chemical components on spectral absorption and scattering. The model is parameterized, where the parameters include spectral absorbance, scattering coefficient, layer depth and sample component concentration. The parameters are calibrated using known optical and chemical data.

[0116] S32: Apply feature extraction algorithms to automatically extract and quantify spectral and scattering properties from hierarchically segmented images, including absorbance peak wavelength, peak width, and total scattering intensity;

[0117] S33: performing Z-score standardization on the extracted feature data, converting the data into a format with zero mean and unit variance, and then performing normalization processing to make the weights of different feature data consistent;

[0118] S34: Cross-validation techniques are used to test the accuracy and generalization ability of the mathematical model and feature extraction algorithm, independent test data sets are used to evaluate the predictive ability and stability of the model, and the extracted feature data are integrated into a comprehensive analysis framework for the final abnormal state identification and classification.

[0119] According to the specific spectral characteristics of light absorption and scattering of different components in serum samples, the mathematical model is established based on the Beer-Lambert law, which is expressed as: ;in, is at wavelength The absorbance at is the molar absorptivity, obtained based on experimental fit (depends on wavelength and chemical composition), is the concentration of the solute, is the light path length, i.e. the width of the sample container;

[0120] Feature extraction algorithms include:

[0121] Spectral feature extraction: Use spectral imaging technology to obtain the spectral image of the serum sample, record the absorbance at each wavelength, identify the absorbance peak at each wavelength, the absorbance peak corresponds to the maximum absorption point of the specific chemical component, and use the local maximum search algorithm:

[0122] ,in It is a comparison within a small range around the wavelength.

[0123] Scattering property extraction: For scattering properties, the scattering intensity of serum samples at different wavelengths is quantified. The scattering intensity is measured by measuring the intensity of light deviating from the original propagation direction. Using Mie scattering theory, the scattering intensity Calculated as:

[0124] in, is at wavelength and scattering angle The scattering intensity under is the incident light intensity, is a function of the Mie scattering amplitude, which depends on the particle's size, shape, refractive index, and dimensions relative to the wavelength of light;

[0125] Mie scattering amplitude function and The calculation uses numerical methods as follows:

[0126] Particle parameter setting: First set the refractive index of the particle , particle radius and wavelength of light;

[0127] Size parameter and relative refractive index: in is the size parameter, is the relative refractive index of the particle and the medium;

[0128] The scattering coefficient is obtained by solving the Mie mathematical equation and , coefficients and and Related, the calculation can be obtained by modified Bessel function;

[0129] The scattering amplitude function and It is expressed as:

[0130] in, and is an angular function, dependent on the scattering angle .

[0131] Using the above theory and calculations, experiments and simulations are set up to measure and verify the scattering properties in serum samples. By measuring the scattering intensity at different angles, the physical and chemical properties of the sample, such as particle size and concentration, are inferred. Specific calculations can be based on Mie theory scattering calculation software, providing accurate simulation and prediction results.

[0132] The S1 and S2 functions describe the scattering behavior of polarized light waves perpendicular and parallel to the scattering plane. This means that when light waves interact with particles, light waves with different polarization directions will exhibit different scattering characteristics.

[0133] In order to measure the absorbance of samples of known concentration at different wavelengths and fit the molar absorptivity from these data , the following steps and calculation methods can be used:

[0134] Step 1, data collection:

[0135] Sample preparation: Prepare a series of serum samples with known concentrations containing the chemical components to be analyzed at known concentrations.

[0136] Spectral measurement: using a spectrometer to measure Measure the absorbance of each sample , ensuring that experimental conditions such as temperature and light source intensity are controlled during the measurement process.

[0137] Step 2, data processing:

[0138] Absorbance recording: Record the absorbance of each sample at each wavelength and establish a data matrix. The rows of the data matrix represent different samples and the columns represent different wavelengths.

[0139] Step 3, fitting the molar absorptivity, according to the Beer-Lambert law: ;

[0140] Data Linear Regression: To find , the formula can be rewritten as: For each wavelength , using linear regression analysis, right Performing a regression, the slope will be given by In the specific calculation, it is assumed that , we can calculate :

[0141]

[0142] The slope is calculated by the least squares method:

[0143] in, is the number of samples, It is The concentration of the samples, is the corresponding absorbance.

[0144] Error analysis: Statistical analysis was performed on the regression results to calculate the goodness of fit and standard error to evaluate the accuracy and reliability of the model.

[0145] Through the above process, the molar absorption coefficients of different chemical components at different wavelengths can be accurately determined and fitted. .

[0146] The pattern recognition technology in S4 uses a spectral clustering algorithm, which includes:

[0147] S41: Feature vector construction, combining the extracted spectral characteristics and scattering characteristics into a feature vector, standardizing or normalizing the characteristics to ensure that the data are on the same scale (including peak wavelength, peak intensity and overall spectral shape, which can reflect the presence and concentration of specific chemical components in the sample; quantifying the degree of deviation of light from the original propagation direction, the scattering angle distribution, and the scattering characteristics related to particle size and distribution);

[0148] S42: Construct a similarity matrix based on the similarity between feature vectors, and use the Gaussian kernel function to calculate the similarity between two samples:

[0149] ,in, and The samples are and The feature vector of contains the spectral characteristics and scattering characteristics data, is the bandwidth parameter of the kernel function, which determines the sensitivity of the function and is optimized through cross-validation method. Representation sample and The square of the Euclidean distance between them;

[0150] S43: Spectral clustering algorithm implementation, calculation of degree matrix , the degree matrix is a diagonal matrix where each element is the similarity matrix No. Sum of row elements: , calculate the Laplacian matrix based on the similarity matrix ,in, is a diagonal matrix, each element of which is a matrix No. The sum of the rows;

[0151] S44: Eigendecomposition and clustering, perform eigendecomposition of the Laplacian matrix L, extract its front The eigenvector corresponding to the smallest non-zero eigenvalue is used to describe the low-dimensional embedding of the data. Each point in the eigenvector is regarded as a point in a new low-dimensional space. Each point in the eigenvector is clustered using k-means clustering. Each cluster represents a group of samples with similar spectral and scattering characteristics.

[0152] S45: Identify abnormal states, analyze the spectral characteristics and scattering characteristics of samples in each cluster, determine which clusters represent normal states and which represent abnormal states. Abnormal clusters will show spectral or scattering characteristics that are significantly different from normal biochemical indicators. Based on the clustering results, mark the samples belonging to abnormal clusters as abnormal states.

[0153] Kernel function bandwidth parameter The selection method is to set It is the reciprocal of the average distance of all sample pairs, that is: Indicates the average value.

[0154] like Figure 2 As shown, the serum abnormal state recognition system based on image analysis is used to implement the above-mentioned serum abnormal state recognition method based on image analysis, and includes the following modules:

[0155] Includes the following modules:

[0156] Image acquisition module: An automated robotic arm is configured to place and fix the serum sample to be analyzed in a transparent or translucent container. The OCT device is integrated, including a control unit that can adjust the scanning parameters and a light source wavelength and power regulator that adapts to the characteristics of the serum sample. The module automatically and continuously scans from the surface layer of the serum sample to a preset depth. The image data of each layer is captured to obtain multi-layer serum image data.

[0157] Image processing and segmentation module: receives multi-layer serum image data, performs signal enhancement and contrast adjustment, optimizes image quality, automatically sets segmentation parameters based on reflection intensity and scattering pattern, and selects segmentation algorithm to process each layer of image;

[0158] Feature extraction and analysis module: Based on the defined mathematical model, it automatically extracts spectral characteristics and scattering characteristics from the segmented images at each level, performs standardization and regularization processing to unify the weights of different feature data, and prepares them for pattern recognition;

[0159] Pattern recognition and diagnosis module: Using spectral clustering algorithm, construct similarity matrix according to the similarity of feature vectors, perform feature decomposition and clustering, automatically distinguish normal and abnormal biochemical indicators, and identify abnormal status based on clustering results;

[0160] User Interface and Control Module: Provides an operational interface that allows medical professionals to input sample information, view real-time imaging and processing progress, adjust analysis parameters, and view final diagnostic results and reports.

[0161] System control unit: Integrates all hardware and software operations to ensure efficient and stable operation of the system.

[0162] A computer-readable storage medium can be read by a computer, the storage medium stores a computer program, and the computer program is executed by a processor to implement the above method.

[0163] Those skilled in the art should understand that the discussion of any of the above embodiments is merely illustrative and is not intended to imply that the scope of the present invention is limited to these examples. Under the concept of the present invention, the technical features in the above embodiments or different embodiments may be combined, the steps may be implemented in any order, and there are many other variations of the different aspects of the present invention as described above, which are not provided in detail for the sake of simplicity.

[0164] The present invention is intended to cover all such substitutions, modifications and variations that fall within the broad scope of the claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for identifying abnormal serum status based on image analysis, characterized in that: The following steps are involved: S1: Use optical coherence tomography (OCT) technology to image the serum sample to be analyzed and obtain multi-layer serum image data, where each layer represents different depth components in the serum; S2: Perform hierarchical segmentation on the collected multi-level serum sample images and remove image noise caused by particles and impurities; S3: Establish a mathematical model to analyze and standardize the optical and biochemical properties of serum samples, and then extract key features from each layer of the hierarchical segmented image. The key features include spectral characteristics and scattering characteristics, which are used to identify and distinguish normal and abnormal biochemical indicators; S4: Based on the extraction of spectral characteristics and scattering characteristics, pattern recognition technology is applied to distinguish normal and abnormal biochemical indicators, thereby identifying abnormal states in serum; S5: Display the identified abnormal state in a graphical user interface, including information on the abnormality type and severity; The S2 receives the collected original multi-layer serum image data and makes a preliminary evaluation to determine whether the image quality and the clarity of each layer of the image meet the standards. The S2 also includes: S21: determining image segmentation parameters based on visible optical properties in the serum sample, including reflection intensity and scattering pattern, including setting a segmentation threshold and selecting a segmentation algorithm, wherein the segmentation algorithm includes a threshold-based method, a region growing algorithm, or an edge-based segmentation technique; S22: applying the selected image segmentation technology to segment each layer of the image into a plurality of segmentation regions, each segmentation region representing a biochemical component or structure in the serum sample, and ensuring that each segmentation region represents its corresponding deep level feature; S23: Evaluate the accuracy and completeness of the segmentation results to check whether any areas are misclassified or missed; S24: organizing and annotating the segmented image regions to ensure that the data of each segmented region is correctly associated with its corresponding depth level and biochemical characteristics; The S21 specifically includes: S211, optical properties analysis: Reflection intensity analysis: A custom script was used to analyze the reflection intensity in the multi-level serum image data, calculate the grayscale value of each pixel, and evaluate the distribution of reflection intensity within the entire image; Scattering pattern recognition: Scattering patterns indicate the characteristics of different substances in the sample. The distribution and intensity of scattered light, as well as changes at different depth levels, are analyzed to identify structural boundaries or composition differences. S212, set segmentation threshold: Based on the statistics of reflection intensity, the segmentation threshold is selected to distinguish different biochemical components, with high reflection intensity indicating dense tissue areas and low reflection intensity indicating tissue areas with high water content; Histogram analysis is used to assist in determining the optimal segmentation threshold. Different peaks on the histogram represent the reflection characteristics of different materials. S213, select segmentation algorithm: Threshold-based segmentation: If the image reflection intensity distribution is uniform, the segmentation threshold is automatically determined using the global threshold method; Region growing algorithm: For images with complex scattering patterns, the region growing algorithm is used; Edge-based segmentation: When the scattering pattern shows obvious edges, edge detection algorithms are used to identify and segment different structural boundaries; The S3 specifically includes: S31: Define a mathematical model to explain and simulate the changes in spectral characteristics and scattering characteristics in serum samples. The mathematical model is used to describe the effects of different chemical components on spectral absorption and scattering, and parameterize the model, where the parameters include spectral absorbance, scattering coefficient, layer depth and sample component concentration; S32: Apply feature extraction algorithms to automatically extract and quantify spectral and scattering properties from hierarchically segmented images, including absorbance peak wavelength, peak width, and total scattering intensity; S33: performing Z-score standardization on the extracted feature data, converting the data into a format with zero mean and unit variance, and then performing normalization processing to make the weights of different feature data consistent; S34: Use cross-validation techniques to test the accuracy and generalization ability of mathematical models and feature extraction algorithms, use independent test data sets to evaluate the predictive ability and stability of the models, and integrate the extracted feature data into a comprehensive analysis framework for the final abnormal state identification and classification; The mathematical model is established based on the Beer-Lambert law, which is expressed as: ;in, is at wavelength The absorbance at is the molar absorptivity, obtained based on experimental fitting, is the concentration of the solute, is the light path length, i.e. the width of the sample container; The feature extraction algorithm includes: Spectral feature extraction: Use spectral imaging technology to obtain the spectral image of the serum sample, record the absorbance at each wavelength, identify the absorbance peak at each wavelength, the absorbance peak corresponds to the maximum absorption point of the specific chemical component, and use the local maximum search algorithm: ,in It is a comparison within a small range of surrounding wavelengths; Scattering property extraction: For scattering properties, the scattering intensity of serum samples at different wavelengths is quantified. The scattering intensity is measured by measuring the intensity of light deviating from the original propagation direction. Using Mie scattering theory, the scattering intensity Calculated as: in, is at wavelength and scattering angle The scattering intensity under is the incident light intensity, is the Mie scattering amplitude function; The Mie scattering amplitude function and The calculation uses numerical methods as follows: Particle parameter setting: First set the refractive index of the particle , particle radius and wavelength of light; Size parameter and relative refractive index: in is the size parameter, is the relative refractive index of the particle and the medium; The scattering coefficient is obtained by solving the Mie mathematical equation and ; The scattering amplitude function and It is expressed as: in, and is an angular function, dependent on the scattering angle ; The pattern recognition technology in S4 adopts a spectral clustering algorithm, which specifically includes: S41: Feature vector construction, combining the extracted spectral characteristics and scattering characteristics into a feature vector, standardizing or normalizing the features to ensure that the data are on the same scale; S42: Construct a similarity matrix based on the similarity between feature vectors, and use the Gaussian kernel function to calculate the similarity between two samples: in, and The samples are and The feature vector of contains the spectral characteristics and scattering characteristics data, is the bandwidth parameter of the kernel function, Representation sample and The square of the Euclidean distance between them; S43: Spectral clustering algorithm implementation, calculation of degree matrix , the degree matrix is a diagonal matrix where each element is the similarity matrix No. Sum of row elements: , calculate the Laplacian matrix based on the similarity matrix ,in, is a diagonal matrix, each element of which is a matrix No. The sum of the rows; S44: Eigendecomposition and clustering, perform eigendecomposition of the Laplacian matrix L, extract its front The eigenvector corresponding to the smallest non-zero eigenvalue is used to describe the low-dimensional embedding of the data. Each point in the eigenvector is regarded as a point in a new low-dimensional space. Each point in the eigenvector is clustered using k-means clustering. Each cluster represents a group of samples with similar spectral and scattering characteristics. S45: Identify abnormal states, analyze the spectral characteristics and scattering characteristics of samples in each cluster, determine which clusters represent normal states and which represent abnormal states, and abnormal clusters will show spectral or scattering characteristics that are significantly different from normal biochemical indicators. According to the clustering results, mark the samples belonging to the abnormal clusters as abnormal states; The bandwidth parameter of the kernel function The selection method is to set It is the reciprocal of the average distance of all sample pairs, that is:

2. The method for identifying abnormal serum status based on image analysis according to claim 1, characterized in that: The S1 specifically includes: S11, sample preparation: placing the serum sample to be analyzed in a transparent or translucent container, which is transparent to the light band of OCT imaging; S12, equipment setting: adjusting the scanning parameters of the OCT equipment, including scanning depth, resolution, and scanning speed, to suit the characteristics of the serum sample, and setting the wavelength and power of the light source; S13, imaging scan: start the OCT device to perform continuous scanning, starting from the surface layer of the serum sample and gradually going deeper to a preset depth. Each scan acquires image data in a layer of serum, reflecting the optical properties and biochemical composition of the layer; S14, data acquisition and processing: collect image data obtained from scanning at various depth levels and perform preliminary data processing, including signal enhancement and contrast adjustment, to improve image quality and analysis accuracy.

3. A serum abnormal state recognition system based on image analysis, used to implement the serum abnormal state recognition method based on image analysis as described in any one of claims 1-2, characterized in that: Includes the following modules: Image acquisition module: An automated robotic arm is configured to place and fix the serum sample to be analyzed in a transparent or translucent container. The OCT device is integrated, including a control unit that can adjust the scanning parameters and a light source wavelength and power regulator that adapts to the characteristics of the serum sample. The module automatically and continuously scans from the surface layer of the serum sample to a preset depth. The image data of each layer is captured to obtain multi-layer serum image data. Image processing and segmentation module: receives multi-layer serum image data, performs signal enhancement and contrast adjustment, optimizes image quality, automatically sets segmentation parameters based on reflection intensity and scattering pattern, and selects segmentation algorithm to process each layer of image; Feature extraction and analysis module: Based on the defined mathematical model, it automatically extracts spectral characteristics and scattering characteristics from the segmented images at each level, performs standardization and regularization processing to unify the weights of different feature data, and prepares them for pattern recognition; Pattern recognition and diagnosis module: Using spectral clustering algorithm, construct similarity matrix according to the similarity of feature vectors, perform feature decomposition and clustering, automatically distinguish normal and abnormal biochemical indicators, and identify abnormal status based on clustering results; User Interface and Control Module: Provides an operational interface that allows medical professionals to input sample information, view real-time imaging and processing progress, adjust analysis parameters, and view final diagnostic results and reports.

4. A computer-readable storage medium, wherein: The storage medium stores a computer program, and the computer program is executed by a processor to implement the method according to any one of claims 1 to 2.

Images