Nail fold blood microcirculation image data analysis method and system

Abstract

The invention relates to the technical field of image data analysis, and particularly discloses a nail fold blood microcirculation image data analysis method and system, and the method comprises the steps: collecting nail fold blood microcirculation original image data, and carrying out the preprocessing of the nail fold blood microcirculation original image data, a deep learning-based U-Net convolutional neural network model is adopted to carry out nail fold blood vessel segmentation, a particle image velocity measurement algorithm is utilized to calculate the blood flow velocity of nail fold blood, the blood flow velocity of nail fold blood is corrected, and a nail fold blood vessel-blood flow-periappendage combined model and a nail fold blood microcirculation analysis influence model are constructed. And updating the nail fold blood conjoint analysis factor, and outputting a nail fold blood microcirculation image data classification result by using a support vector machine model. The problems that traditional nail fold blood microcirculation image data analysis is low in accuracy and efficiency, data analysis interference factors are not comprehensively considered, and analysis accuracy is insufficient are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of image data analysis, and specifically to a method and system for analyzing nailfold blood microcirculation image data. Background Art

[0002] The microvessels in the nailfold area are richly distributed. With the development of optical imaging technology, high-resolution and high-definition nailfold microcirculation image acquisition devices have emerged, which can obtain clearer and richer microcirculation image information. At the same time, computer technology and digital image processing algorithms have been continuously advancing, including image denoising, segmentation, feature extraction, etc., providing technical means for in-depth analysis of nailfold blood microcirculation images. Through the analysis of nailfold blood microcirculation image data, it is helpful to explore the change mechanism of microcirculation and provide a theoretical basis for new treatment targets and intervention strategies.

[0003] Nowadays, there are still some deficiencies in the research on the analysis of nailfold blood microcirculation image data. Specifically, the traditional method of relying on the naked eye to observe nailfold microcirculation images has low accuracy and efficiency, and cannot meet the needs of accurate analysis of nailfold blood microcirculation image data. The interference factors in the analysis of nailfold blood microcirculation image data are not comprehensively considered, the analysis accuracy is insufficient, and there is still room for improvement in the accuracy and reliability of the results. Summary of the Invention

[0004] In view of the deficiencies of the prior art, the present invention provides a method and system for analyzing nailfold blood microcirculation image data, which can effectively solve the problems involved in the above background art.

[0005] To achieve the above objectives, the present invention is realized through the following technical solutions: In the first aspect of the present invention, a method for analyzing nailfold blood microcirculation image data is provided, including the following steps: using a microcirculation imaging device to collect original nailfold blood microcirculation image data and preprocess the original nailfold blood microcirculation image data; adopting a U-Net convolutional neural network model based on deep learning for nailfold blood vessel segmentation; in the segmented nailfold blood vessel image, using the particle image velocimetry algorithm to calculate the nailfold blood flow velocity and correct the nailfold blood flow velocity; collecting nailfold blood vessel morphological feature data, nailfold blood flow feature data, and nailfold blood loop perimeter environment feature data, constructing a nailfold blood vessel-blood flow-loop perimeter joint model, and outputting a nailfold blood joint analysis factor; obtaining nailfold blood microcirculation analysis feature data, constructing a nailfold blood microcirculation analysis influence model, and outputting a nailfold blood microcirculation analysis influence factor; based on the nailfold blood microcirculation analysis influence factor, correcting the nailfold blood joint analysis factor to obtain an updated nailfold blood joint analysis factor; based on the updated nailfold blood joint analysis factor, using a support vector machine model to output a classification result of nailfold blood microcirculation image data.

[0006] As a further method, a microcirculation imaging device is used to collect the original image data of nailfold blood microcirculation, and the original image data of nailfold blood microcirculation is preprocessed. The specific analysis process is as follows: Use the microcirculation imaging device to collect the original image data of nailfold blood microcirculation; Use the non-local means denoising algorithm to denoise the original image data of nailfold blood microcirculation:

[0007]

[0008] where NLM(p) is the value of pixel p after denoising, C(p) is the normalization constant, Ω is the set of all pixels in the original image data of nailfold blood microcirculation, w(p, q) is the weight between pixels p and q, and I(q) is the gray value of the original image at pixel q;

[0009]

[0010] where i traverses each position offset in the neighborhood window centered on p and q, is the weighted Euclidean distance with standard deviation σ, h is the parameter controlling the denoising intensity, I(p + i) is the gray value of the original image at position p + i within the neighborhood window centered on pixel p, and I(q + i) is the gray value of the original image at position q + i within the neighborhood window centered on pixel q;

[0011] An adaptive gray correction method based on histogram equalization is adopted to adjust the gray histogram of the image to a uniform distribution and enhance the contrast of the image: For a gray image, the pixel value J(x, y) after its histogram equalization is calculated by the following formula:

[0012]

[0013] where L is the total number of gray levels, M * N is the size of the image, n a is the frequency of occurrence of gray level a, round is the rounding function, and I(x, y) is the gray value of the pixel at coordinates (x, y) in the image; The preprocessed image data of nailfold blood microcirculation is obtained.

[0014] As a further method, a U-Net convolutional neural network model based on deep learning is used for nailfold blood vessel segmentation. The specific analysis process is as follows: Collect a large number of labeled nailfold blood microcirculation images as the training set, train the U-Net convolutional neural network model based on deep learning, and use the cross-entropy loss function to measure the difference between the prediction result and the true label:

[0015]

[0016] Among them, b is the sample label, b = 1, 2, 3,..., N1, where N1 is the number of samples, c is the class label, c = 1, 2, 3,..., C, and C is the number of classes, y bc is the true label that sample b belongs to class c, and p bc is the probability that the model predicts sample b belongs to class c;

[0017] The model parameters are continuously adjusted through the backpropagation algorithm to minimize the cross-entropy loss function; the preprocessed nailfold blood microcirculation image data is input into the U-Net convolutional neural network model based on deep learning. The U-Net convolutional neural network model based on deep learning outputs the probability that each pixel belongs to the nailfold blood vessels. Based on the nailfold blood vessel probability division rules stored in the database, the segmented nailfold blood vessel image is obtained;

[0018] The nailfold blood vessel probability division rules are as follows:

[0019]

[0020] Among them, r is the probability that the pixel output by the U-Net convolutional neural network model based on deep learning belongs to the nailfold blood vessels, and r0 is the nailfold blood vessel probability division threshold stored in the database.

[0021] As a further method, in the segmented nailfold blood vessel image, the particle image velocimetry algorithm is used to calculate the nailfold blood flow velocity and correct the nailfold blood flow velocity. The specific analysis process is as follows: In the segmented nailfold blood vessel image, based on the window division scheme stored in the database, two adjacent frames of nailfold blood vessel images I1 and I2 are divided into multiple small windows, and the displacement of pixels in each window is calculated to estimate the blood flow velocity:

[0022]

[0023] Among them, I1(x, y) and I2(x, y) are the pixel gray values at coordinates (x, y) in two adjacent frames of nailfold blood vessel images, is the gray mean value of I1 within the window, is the gray mean value of I2 within the window, C(Δx, Δy) is the normalized cross-correlation function, and (Δx, Δy) is the displacement of pixels in the image;

[0024] The (Δx, Δy) that maximizes the normalized cross-correlation function C(Δx, Δy) is denoted as the actual displacement (Δx1, Δy1); the blood flow velocity v measured can be calculated from the actual displacement (Δx1, Δy1) and the frame interval time Δt:

[0025]

[0026] The angle between the blood vessel and the image plane is θ, and the blood flow velocity of the nail fold is corrected to obtain the corrected blood flow velocity v of the nail fold. real :

[0027]

[0028] As a further method, a nail fold blood vessel-blood flow-periloop combined model is constructed to output a nail fold blood combined analysis factor. The specific analysis process is as follows: According to the nail fold blood vessel segmentation image and the nail fold blood flow velocity, nail fold blood vessel morphological feature data, nail fold blood flow feature data, and nail fold blood periloop environmental feature data are collected. Among them: The nail fold blood vessel morphological feature data specifically includes the input branch diameter div of the nail fold blood vessel, the output branch diameter dov of the nail fold blood vessel, and the nail fold blood vessel length Length; the nail fold blood flow feature data specifically includes the nail fold blood flow velocity v real , the density drbca of the red blood cell aggregation area in the nail fold blood, and the number Number of white blood cells in the nail fold blood w ; the nail fold blood periloop environmental feature data specifically includes the size Size of the subpapillary venous plexus plexus , the bleeding area Area around the blood vessel; a nail fold blood vessel-blood flow-periloop combined model is constructed, and based on the nail fold blood vessel morphological feature data, the nail fold blood flow feature data, and the nail fold blood periloop environmental feature data, a nail fold blood combined analysis factor is output. The nail fold blood combined analysis factor is used as the analysis basis for obtaining the updated nail fold blood combined analysis factor.

[0029] As a further method, for the nail fold blood vessel-blood flow-periloop combined model, the specific analysis process is as follows:

[0030]

[0031] In the formula, combine is the nail fold blood combined analysis factor, vessel is the nail fold blood vessel analysis factor, flow is the nail fold blood flow analysis factor, periloop is the nail fold blood periloop environmental analysis factor, A1 is the set weight factor of vessel, A2 is the set weight factor of flow, A3 is the set weight factor of periloop, and e is the natural constant.

[0032] As a further method, a model for analyzing the influencing factors of nailfold blood microcirculation is constructed to output the influencing factors of nailfold blood microcirculation analysis. The specific analysis process is as follows: Obtain the characteristic data of nailfold blood microcirculation analysis, which specifically includes the environmental temperature tem of nailfold blood microcirculation detection, the resolution f bw of the imaging device for nailfold blood microcirculation detection, and the magnification factor bn of the imaging device for nailfold blood microcirculation detection; Construct a model for analyzing the influencing factors of nailfold blood microcirculation, and based on the characteristic data of nailfold blood microcirculation analysis, output the influencing factors of nailfold blood microcirculation analysis. The influencing factors of nailfold blood microcirculation analysis serve as the analysis basis for obtaining the updated combined analysis factors of nailfold blood.

[0033] The model for analyzing the influencing factors of nailfold blood microcirculation, the specific analysis process is as follows:

[0034]

[0035] In the formula, Inf is the influencing factor of nailfold blood microcirculation analysis, and e is the natural constant.

[0036] As a further method, based on the influencing factors of nailfold blood microcirculation analysis, the combined analysis factors of nailfold blood are corrected to obtain the updated combined analysis factors of nailfold blood. The specific analysis process is as follows:

[0037]

[0038] In the formula, is the updated combined analysis factor of nailfold blood, combine is the combined analysis factor of nailfold blood, S1 is the weight factor of combine set, and S2 is the weight factor of Inf set.

[0039] As a further method, based on the updated combined analysis factors of nailfold blood, using the support vector machine model, the classification result of nailfold blood microcirculation image data is output. The specific analysis process is as follows: Obtain the nailfold blood sample data with known states stored in the database, and divide it into a training set and a test set according to a set ratio; Use the training set to train the support vector machine model by minimizing the objective function while satisfying the constraint conditions and 0 ≤ α i ≤ C; where, α i is the Lagrange multiplier corresponding to the i-th nailfold blood sample, α j is the Lagrange multiplier corresponding to the j-th nailfold blood sample, y i is the i-th nailfold blood sample, y j is the j-th nailfold blood sample, i and j are the numbers of nailfold blood samples, n is the total number of nailfold blood samples, C is the penalty parameter, K(x i , x j ) is the radial basis kernel function. σ is the width parameter of the radial basis kernel function, and x i is the feature vector of the i-th nailfold blood sample, and x j is the feature vector of the j-th nailfold blood sample; the trained support vector machine model is input into the preprocessed updated nailfold blood combined analysis factor, and the state of the nailfold blood is classified through the decision function, and the classification result of the nailfold blood microcirculation image data is output.

[0040] The second aspect of the present invention provides a nailfold blood microcirculation image data analysis system, including an original image data preprocessing module, a blood vessel segmentation module, a blood flow velocity analysis and correction module, a combined analysis factor output module, an analysis influencing factor output module, a combined analysis factor update module, and a classification result output module, wherein: the original image data preprocessing module is used to collect the original nailfold blood microcirculation image data by using a microcirculation imaging device and preprocess the original nailfold blood microcirculation image data; the blood vessel segmentation module is used to perform nailfold blood vessel segmentation by using a U-Net convolutional neural network model based on deep learning; the blood flow velocity analysis and correction module is used to calculate the nailfold blood flow velocity by using the particle image velocimetry algorithm in the segmented nailfold blood vessel image and correct the nailfold blood flow velocity; the combined analysis factor output module is used to collect the nailfold blood vessel morphological feature data, the nailfold blood flow feature data, and the nailfold blood loop perimeter environment feature data, construct a nailfold blood vessel-blood flow-loop perimeter combined model, and output the nailfold blood combined analysis factor; the analysis influencing factor output module is used to obtain the nailfold blood microcirculation analysis feature data, construct a nailfold blood microcirculation analysis influencing model, and output the nailfold blood microcirculation analysis influencing factor; the combined analysis factor update module is used to correct the nailfold blood combined analysis factor based on the nailfold blood microcirculation analysis influencing factor to obtain the updated nailfold blood combined analysis factor; the classification result output module is used to output the classification result of the nailfold blood microcirculation image data based on the updated nailfold blood combined analysis factor by using a support vector machine model.

[0041] Compared with the prior art, the embodiments of the present invention have at least the following advantages or beneficial effects:

[0042] (1) The present invention provides a method and system for analyzing nailfold blood microcirculation image data. The U-Net convolutional neural network model can accurately segment nailfold blood vessels. Compared with traditional methods, it can better identify complex vascular structures, providing an accurate basis for subsequent blood flow velocity calculation and feature data acquisition, and reducing diagnostic errors caused by inaccurate vascular segmentation. Feature data such as the morphology, blood flow, and peritrabecular environment of nailfold blood vessels are collected, a joint model is constructed and a joint analysis factor is output, and the joint analysis factor is corrected by considering the influencing factors of microcirculation analysis. This analysis method that synthesizes multi-dimensional information can more comprehensively reflect the true state of nailfold blood microcirculation.

[0043] (2) The present invention provides a tool for studying the interrelationships of various factors in microcirculation by constructing a nailfold blood vessel-blood flow-peritrabecular joint model and a nailfold blood microcirculation analysis influence model. By continuously optimizing and validating these models, the understanding of the mechanism of nailfold blood microcirculation can be further improved. A variety of advanced technologies and algorithms such as deep learning, particle image velocimetry algorithm, and support vector machine are comprehensively used to improve the accuracy and reliability of data processing. The process of gradual correction can reduce the accumulation of errors and improve the accuracy of the final classification result. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] The present invention will be further described with reference to the accompanying drawings. However, the embodiments in the drawings do not constitute any limitation to the present invention. For those of ordinary skill in the art, other drawings can be obtained according to the following drawings without creative efforts.

[0045] Figure 1 It is a schematic flowchart of the method steps of the present invention.

[0046] Figure 2 It is a schematic diagram of the connection of system modules of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0047] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0048] Referring to Figure 1 As shown, the first aspect of the present invention provides a method for analyzing nailfold blood microcirculation image data, including: using a microcirculation imaging device to collect original nailfold blood microcirculation image data and preprocessing the original nailfold blood microcirculation image data.

[0049] The specific analysis process is as follows: Use a microcirculation imaging device to collect the original image data of nailfold blood microcirculation; Use the non-local means denoising algorithm to denoise the original image data of nailfold blood microcirculation:

[0050]

[0051] Among them, NLM(p) is the value of pixel p after denoising, C(p) is the normalization constant, Ω is the set of all pixels in the original image data of nailfold blood microcirculation, w(p,q) is the weight between pixels p and q, and I(q) is the gray value of the original image at pixel q;

[0052]

[0053] Among them, i traverses each position offset in the neighborhood window centered on p and q, is the weighted Euclidean distance with standard deviation σ, h is the parameter controlling the denoising intensity, I(p + i) is the gray value of the original image at the position p + i within the neighborhood window centered on pixel p, and I(q + i) is the gray value of the original image at the position q + i within the neighborhood window centered on pixel q;

[0054] Adopt an adaptive gray correction method based on histogram equalization to adjust the gray histogram of the image to a uniform distribution and enhance the contrast of the image: For a gray image, the pixel value J(x,y) after its histogram equalization is calculated by the following formula:

[0055]

[0056] Among them, L is the total number of gray levels, M*N is the size of the image, n a is the frequency of occurrence of gray level a, round is the rounding function, and I(x,y) is the gray value of the pixel at the coordinate (x,y) in the image; Obtain the preprocessed image data of nailfold blood microcirculation.

[0057] The non-local means denoising algorithm utilizes the self-similarity of the image, can effectively remove the noise in the original image, and retain key structural information such as blood vessels. This reduces the interference of noise on subsequent analysis and improves the image clarity. The adaptive gray correction by histogram equalization improves the gray deviation caused by uneven light sources, enhances the image contrast, makes the blood vessels more distinct from the surrounding tissues, and is conducive to subsequent segmentation and feature extraction.

[0058] High-quality preprocessed images provide a reliable basis for subsequent vascular segmentation, blood flow velocity calculation, etc. Denoising and gray correction reduce data errors, enabling the U-Net model based on deep learning to more accurately identify blood vessels, and the particle image velocimetry algorithm to more accurately calculate blood flow velocity, ultimately improving the accuracy of the entire analysis process. The preprocessed data is more standardized and clear, facilitating model construction and further analysis.

[0059] The U-Net convolutional neural network model based on deep learning is used for nailfold blood vessel segmentation.

[0060] The specific analysis process is as follows: A large number of labeled nailfold blood microcirculation images are collected as the training set, and the U-Net convolutional neural network model based on deep learning is trained. The cross-entropy loss function is used to measure the difference between the prediction result and the true label:

[0061]

[0062] Among them, b is the sample label, b = 1, 2, 3,..., N1, where N1 is the number of samples, c is the class label, c = 1, 2, 3,..., C, and C is the number of classes (in this implementation, C = 2, namely blood vessels and non-blood vessels), and y bc is the true label that sample b belongs to class c (in this implementation, the true label that sample b belongs to class c is 0 or 1), and p bc is the probability that the model predicts that sample b belongs to class c;

[0063] The model parameters are continuously adjusted through the backpropagation algorithm to minimize the cross-entropy loss function; the preprocessed nailfold blood microcirculation image data is input into the U-Net convolutional neural network model based on deep learning. The U-Net convolutional neural network model based on deep learning outputs the probability that each pixel belongs to the nailfold blood vessel. Based on the nailfold blood vessel probability division rule stored in the database, the segmented nailfold blood vessel image is obtained;

[0064] The nailfold blood vessel probability division rule is:

[0065]

[0066] Among them, r is the probability that the pixel output by the U-Net convolutional neural network model based on deep learning belongs to the nailfold blood vessel, and r0 is the nailfold blood vessel probability division threshold stored in the database.

[0067] The U-Net model has a unique structure that combines a contracting path and an expanding path, which can effectively extract image features and restore the resolution. For the complex structure of nailfold blood vessels, it can achieve high-precision segmentation. Through training with a large number of labeled images, the cross-entropy loss function is used to measure the difference and backpropagation is used to adjust the parameters, enabling the model to accurately learn the vascular features and output the probability that each pixel belongs to a blood vessel, thus obtaining a precise segmented image and reducing the error and subjectivity of manual segmentation.

[0068] There is no need for manual pixel-by-pixel segmentation. Inputting the preprocessed image data into the model can automatically output the segmentation result, greatly saving labor and time costs, improving work efficiency, being suitable for processing a large amount of nailfold blood microcirculation image data, and meeting the requirements for rapid data processing in clinical and research settings.

[0069] As long as there is sufficient diverse and accurately labeled training data, this model can adapt to nailfold blood microcirculation images of different individuals and under different imaging conditions. It can also adjust the model parameters or combine with other technologies for further optimization according to new research or clinical needs to improve the segmentation effect and expand the application scope.

[0070] In the segmented nailfold blood vessel image, the particle image velocimetry algorithm is used to calculate the nailfold blood flow velocity and correct the nailfold blood flow velocity.

[0071] The specific analysis process is as follows: In the segmented nailfold blood vessel image, based on the window division scheme stored in the database, two adjacent frames of nailfold blood vessel images I1 and I2 are divided into multiple small windows, and the displacement of pixels within each window is calculated to estimate the blood flow velocity:

[0072]

[0073] Among them, I1(x,y) and I2(x,y) are the pixel gray values at coordinates (x,y) in two adjacent frames of nailfold blood vessel images, is the gray mean value of I1 within the window, is the gray mean value of I2 within the window, C(Δx,Δy) is the normalized cross-correlation function, and (Δx,Δy) is the displacement of pixels in the image;

[0074] The (Δx,Δy) that maximizes the normalized cross-correlation function C(Δx,Δy) is denoted as the actual displacement (Δx1,Δy1); the blood flow velocity v can be calculated from the actual displacement (Δx1,Δy1) and the frame interval time Δt:

[0075]

[0076] The angle between the blood vessel and the image plane is θ, and the nailfold blood flow velocity is corrected to obtain the corrected nailfold blood flow velocity v real:

[0077]

[0078] By dividing the adjacent-frame nailfold blood vessel images into small windows and calculating the pixel displacement within the windows to estimate the blood flow velocity, the flow condition of the blood in the blood vessels can be carefully captured. The application of the normalized cross-correlation function can accurately find the actual displacement of the pixels, thereby obtaining a relatively accurate measurement value, providing reliable blood flow velocity data for analyzing the nailfold blood microcirculation condition.

[0079] Considering the influence of the angle between the blood vessel and the image plane on the blood flow velocity measurement and performing correction, it makes up for the measurement error caused by the imaging angle. It makes the measurement result closer to the true blood flow velocity, enhances the reliability and scientific nature of the data, and helps to more accurately evaluate the microcirculation function.

[0080] It simulates the real flow state of the blood in the blood vessels, analyzes the displacement at the pixel level, and conforms to the physiological characteristics of the nailfold blood microcirculation. It is of great significance for in-depth study of the nailfold blood microcirculation mechanism.

[0081] Collect the nailfold blood vessel morphological feature data, nailfold blood flow feature data, and nailfold blood perianastomotic environment feature data, construct a nailfold blood vessel-blood flow-perianastomotic joint model, and output the nailfold blood joint analysis factor.

[0082] The specific analysis process is as follows: According to the nailfold blood vessel segmentation image and the nailfold blood flow velocity, collect the nailfold blood vessel morphological feature data, nailfold blood flow feature data, and nailfold blood perianastomotic environment feature data, where: The nailfold blood vessel morphological feature data specifically includes the input branch diameter div of the nailfold blood vessel, the output branch diameter dov of the nailfold blood vessel, and the nailfold blood vessel length Length; the nailfold blood flow feature data specifically includes the nailfold blood flow velocity v real , the density drbca of the red blood cell aggregation area in the nailfold blood, and the number Number of white blood cells in the nailfold blood w ; the nailfold blood perianastomotic environment feature data specifically includes the size Size plexus of the subpapillary venous plexus and the bleeding area Area around the blood vessel; construct a nailfold blood vessel-blood flow-perianastomotic joint model, and based on the nailfold blood vessel morphological feature data, nailfold blood flow feature data, and nailfold blood perianastomotic environment feature data, output the nailfold blood joint analysis factor, and the nailfold blood joint analysis factor is used as the analysis basis for obtaining the updated nailfold blood joint analysis factor.

[0083] The nailfold blood vessel-blood flow-perianastomotic joint model, the specific analysis process is as follows:

[0084]

[0085] In the formula, combine is the combined analysis factor of nailfold blood, vessel is the vascular analysis factor of nailfold blood, flow is the blood flow analysis factor of nailfold blood, periloop is the periloop environment analysis factor of nailfold blood, A1 is the set weight factor of vessel, A2 is the set weight factor of flow, A3 is the set weight factor of periloop, and e is the natural constant.

[0086] Collect multi-dimensional data covering vascular morphology, blood flow characteristics, and periloop environment, breaking through the limitations of single-index analysis. For example, it considers both the diameter and length of blood vessels, as well as blood flow velocity, blood cell aggregation, and the size and bleeding area of the periloop venous plexus, etc., and can comprehensively and objectively reflect the true state of nailfold blood microcirculation as a whole.

[0087] Integrate various characteristic data into a combined analysis factor through a specific formula, enabling the quantification of the microcirculation state. Compared with qualitative descriptions, quantitative indicators are more accurate and objective. The analysis factors in the combined model are combined through weight factors, reflecting the mutual relationship between vascular, blood flow, and periloop environment factors. This is more in line with the actual situation of the nailfold blood microcirculation system and can more accurately evaluate the impact of the synergistic effect of various factors on microcirculation.

[0088] Obtain the characteristic data of nailfold blood microcirculation analysis, construct an influence model for nailfold blood microcirculation analysis, and output the influence factor of nailfold blood microcirculation analysis.

[0089] The specific analysis process is as follows: Obtain the characteristic data of nailfold blood microcirculation analysis. The characteristic data of nailfold blood microcirculation analysis specifically includes the temperature tem of the nailfold blood microcirculation detection environment, the resolution f bw of the nailfold blood microcirculation detection imaging device, and the magnification bn of the nailfold blood microcirculation detection imaging device; construct an influence model for nailfold blood microcirculation analysis, and based on the characteristic data of nailfold blood microcirculation analysis, output the influence factor of nailfold blood microcirculation analysis. The influence factor of nailfold blood microcirculation analysis is used as the analysis basis for obtaining the updated combined analysis factor of nailfold blood.

[0090] The influence model for nailfold blood microcirculation analysis, the specific analysis process is as follows:

[0091]

[0092] In the formula, Inf is the influence factor of nailfold blood microcirculation analysis.

[0093] Incorporate external factors such as the detection environment temperature, imaging device resolution, and magnification into the analysis, recognizing that these factors will affect the nailfold blood microcirculation image data and analysis results, avoiding analysis deviations caused by ignoring these factors, and making the analysis more comprehensive and objective.

[0094] Quantify external factors into the influencing factors of nailfold blood microcirculation analysis through a specific formula, providing specific numerical references for evaluating the influence degree of external factors on microcirculation analysis, and serving as the basis for updating the combined analysis factors of nailfold blood. This influencing factor can correct the combined analysis factors, reduce the errors caused by external factors, make the final analysis results closer to the true nailfold blood microcirculation state, and thus improve the accuracy and reliability of nailfold blood microcirculation image data analysis.

[0095] Based on the influencing factors of nailfold blood microcirculation analysis, correct the combined analysis factors of nailfold blood to obtain updated combined analysis factors of nailfold blood.

[0096] The specific analysis process is as follows:

[0097]

[0098] In the formula, is the updated combined analysis factor of nailfold blood, combine is the combined analysis factor of nailfold blood, S1 is the weight factor of the set combine, and S2 is the weight factor of the set Inf.

[0099] The influencing factors of nailfold blood microcirculation analysis incorporate external factors such as the detection environment temperature and imaging device parameters. By correcting the combined analysis factors through it, the interference and deviation brought by these external factors to the analysis can be eliminated or reduced, making the updated combined analysis factors more accurately reflect the true state of nailfold blood microcirculation.

[0100] Different detection environments and equipment conditions will cause data differences. This correction mechanism enables the analysis results to adapt to diverse detection scenarios. Regardless of the data collection environment, through the adjustment of the influencing factors, the combined analysis factors can be made comparable and stable, enhancing the applicability and generality of the entire analysis method.

[0101] Combining external influencing factors with the inherent characteristics of microcirculation improves the analysis system of nailfold blood microcirculation. From simply analyzing the internal characteristics of microcirculation to considering external influencing factors, a more comprehensive and systematic analysis process is formed, which helps to deeply explore relevant information about nailfold blood microcirculation.

[0102] Based on the updated combined analysis factors of nailfold blood, use the support vector machine model to output the classification results of nailfold blood microcirculation image data.

[0103] The specific analysis process is as follows: Obtain the nailfold blood sample data with known states stored in the database, divide it into a training set and a test set according to a set ratio; use the training set to train the support vector machine model, by minimizing the objective function while satisfying the constraint conditions and 0 ≤ αi ≤C; where α i is the Lagrange multiplier corresponding to the i-th nailfold blood sample, α j is the Lagrange multiplier corresponding to the jth nailfold blood sample, y i is the i-th nailfold blood sample, y j is the jth nailfold blood sample, i and j are the nailfold blood sample numbers, n is the total number of nailfold blood samples, C is the penalty parameter, K(x i ,x j ) is the radial basis kernel function, σ is the width parameter of the radial basis kernel function, x i is the feature vector of the i-th nailfold blood sample, x j is the characteristic vector of the jth nailfold blood sample; the trained support vector machine model is input into the preprocessed updated nailfold blood joint analysis factor, the state of the nailfold blood is classified through the decision function, and the classification result of the nailfold blood microcirculation image data is output.

[0104] The support vector machine model can effectively handle linear and nonlinear classification problems by finding the optimal classification hyperplane in high-dimensional space. By training with nailfold blood sample data of known states, the mapping relationship between sample characteristics and states can be learned. The updated nailfold blood joint analysis factor comprehensively considers multiple factors and provides the model with more representative and accurate data, so that the model can more accurately classify nailfold blood microcirculation imaging data and improve the accuracy of diagnosis.

[0105] During the training process, the support vector machine model can effectively avoid overfitting by minimizing the objective function and satisfying the constraints, making the model have good generalization ability. This means that the model not only performs well on the training data, but also can give reliable classification results for new and unseen nail fold blood sample data, and is suitable for judging the nail fold blood microcirculation status of different individuals.

[0106] This method makes full use of the updated nailfold blood joint analysis factors obtained in the previous processing, which integrates information such as nailfold blood vessels, blood flow, periarthritis environment and external influencing factors. The support vector machine model learns and classifies based on these rich data, which can dig out the information hidden behind the data and analyze the state of nailfold blood microcirculation more comprehensively.

[0107] Reference Figure 2 As shown, the second aspect of the present invention provides a nailfold blood microcirculation image data analysis system, including an original image data preprocessing module, a blood vessel segmentation module, a blood flow velocity analysis and correction module, a joint analysis factor output module, an analysis influence factor output module, a joint analysis factor update module and a classification result output module.

[0108] The original image data preprocessing module is used to collect the original image data of nailfold blood microcirculation using a microcirculation imaging device and preprocess the original image data of nailfold blood microcirculation.

[0109] The blood vessel segmentation module is used to perform nailfold blood vessel segmentation using a U-Net convolutional neural network model based on deep learning.

[0110] The blood flow velocity analysis and correction module is used to calculate the blood flow velocity of nailfold blood using the particle image velocimetry algorithm in the segmented nailfold blood vessel image and correct the blood flow velocity of nailfold blood.

[0111] The combined analysis factor output module is used to collect the nailfold blood vessel morphological feature data, nailfold blood flow feature data, and nailfold blood loop perimeter environment feature data, construct a nailfold blood vessel-blood flow-loop perimeter combined model, and output the nailfold blood combined analysis factor.

[0112] The analysis influencing factor output module is used to obtain the nailfold blood microcirculation analysis feature data, construct a nailfold blood microcirculation analysis influencing model, and output the nailfold blood microcirculation analysis influencing factor.

[0113] The combined analysis factor update module is used to correct the nailfold blood combined analysis factor based on the nailfold blood microcirculation analysis influencing factor to obtain an updated nailfold blood combined analysis factor.

[0114] The classification result output module is used to output the classification result of the nailfold blood microcirculation image data based on the updated nailfold blood combined analysis factor using a support vector machine model.

[0115] The above content is only an example and explanation of the structure of the present invention. Those skilled in the art of this technology can make various modifications, supplements, or use similar methods to replace the specific embodiments described, as long as they do not deviate from the structure of the invention or exceed the scope defined by this claims, they should fall within the protection scope of the present invention.

Claims

1. A method for analyzing nailfold blood microcirculation image data, characterized in that, It includes the following steps: Using a microcirculation imaging device, collect the original image data of nailfold blood microcirculation and preprocess the original image data of nailfold blood microcirculation; Adopt a U-Net convolutional neural network model based on deep learning for nailfold blood vessel segmentation; In the segmented nailfold blood vessel image, use the particle image velocimetry algorithm to calculate the nailfold blood flow velocity and correct the nailfold blood flow velocity; Collect the nailfold blood vessel morphological feature data, nailfold blood flow feature data, and nailfold blood loop perimeter environment feature data, construct a nailfold blood vessel-blood flow-loop perimeter joint model, and output the nailfold blood joint analysis factor; Obtain the nailfold blood microcirculation analysis feature data, construct a nailfold blood microcirculation analysis influence model, and output the nailfold blood microcirculation analysis influence factor; Based on the nailfold blood microcirculation analysis influence factor, correct the nailfold blood joint analysis factor to obtain an updated nailfold blood joint analysis factor; Based on the updated nailfold blood joint analysis factor, use a support vector machine model to output the classification result of the nailfold blood microcirculation image data.

2. The method for analyzing nailfold blood microcirculation image data according to claim 1, wherein: The specific analysis process of using a microcirculation imaging device to collect the original image data of nailfold blood microcirculation and preprocess the original image data of nailfold blood microcirculation is as follows: Use a microcirculation imaging device to collect the original image data of nailfold blood microcirculation; Use the non-local means denoising algorithm to denoise the original image data of nailfold blood microcirculation: Where NLM(p) is the value of pixel p after denoising, C(p) is the normalization constant, Ω is the set of all pixels in the original image data of nailfold blood microcirculation, w(p,q) is the weight between pixel p and q, and I(q) is the original image gray value at pixel q; where \(i\) traverses each position offset in the neighborhood window centered at \(p\) and \(q\). is the weighted Euclidean distance with standard deviation \(\sigma\), \(h\) is a parameter controlling the denoising intensity, \(I(p + i)\) is the gray value of the original image at position \(p + i\) within the neighborhood window centered at pixel \(p\), and \(I(q + i)\) is the gray value of the original image at position \(q + i\) within the neighborhood window centered at pixel \(q\). Adopt an adaptive gray correction method based on histogram equalization to adjust the gray histogram of the image to a uniform distribution and enhance the contrast of the image: For a grayscale image, the pixel value J(x,y) after its histogram equalization is calculated by the following formula: where L is the total number of gray levels, M*N is the size of the image, n a is the frequency of occurrence of gray level a, round is the rounding function, and I(x, y) is the pixel gray value at the coordinate (x, y) in the image; Obtain the preprocessed nailfold blood microcirculation image data.

3. A method for analyzing nailfold blood microcirculation image data according to claim 1, characterized in that: The specific analysis process of adopting a U-Net convolutional neural network model based on deep learning for nailfold blood vessel segmentation is as follows: Collect a large number of labeled nailfold blood microcirculation images as the training set, train the U-Net convolutional neural network model based on deep learning, and use the cross-entropy loss function to measure the difference between the prediction result and the true label: Among them, b is the sample label, b = 1, 2, 3,..., N1, where N1 is the number of samples, c is the class label, c = 1, 2, 3,..., C, and C is the number of classes, y bc is the true label that sample b belongs to class c, and p bc is the probability that the model predicts that sample b belongs to class c; Continuously adjust the model parameters through the backpropagation algorithm to minimize the cross-entropy loss function; Input the preprocessed nailfold blood microcirculation image data into the U-Net convolutional neural network model based on deep learning. The U-Net convolutional neural network model based on deep learning outputs the probability that each pixel belongs to the nailfold blood vessel. Based on the nailfold blood vessel probability division rule stored in the database, obtain the segmented nailfold blood vessel image; The nailfold blood vessel probability division rule is: Where r is the probability that the pixel output by the U-Net convolutional neural network model based on deep learning belongs to the nailfold blood vessel, and r0 is the nailfold blood vessel probability division threshold stored in the database.

4. A method for analyzing nailfold blood microcirculation image data according to claim 1, characterized in that: In the segmented nail fold blood vessel image, the particle image velocimetry algorithm is used to calculate the blood flow velocity of the nail fold blood, and the blood flow velocity of the nail fold blood is corrected. The specific analysis process is as follows: In the segmented nail fold blood vessel image, based on the window division scheme stored in the database, two adjacent frames of nail fold blood vessel images I1 and I2 are divided into multiple small windows, and the displacement of pixels in each window is calculated to estimate the blood flow velocity: Wherein, I1(x, y) and I2(x, y) are the pixel gray values at the coordinate (x, y) in two adjacent frames of fingernail fold blood vessel images, is the average gray value of I1 within the window, is the average gray value of I2 within the window, C(Δx, Δy) is the normalized cross-correlation function, and (Δx, Δy) is the displacement of the pixel in the image; The (Δx, Δy) that maximizes the normalized cross-correlation function C(Δx, Δy) is denoted as the actual displacement (Δx1, Δy1); Blood flow velocity v measured It can be calculated from the actual displacement (Δx1, Δy1) and the frame interval time Δt: The angle between the blood vessel and the image plane is θ, and the blood flow velocity of the nail fold is corrected to obtain the corrected blood flow velocity v of the nail fold real :

5. A method for analyzing nailfold blood microcirculation image data according to claim 1, characterized in that: The construction of the nail fold blood vessel - blood flow - periloop joint model outputs the nail fold blood joint analysis factor. The specific analysis process is as follows: According to the nail fold blood vessel segmentation image and the nail fold blood flow velocity, the nail fold blood vessel morphological feature data, the nail fold blood flow feature data, and the nail fold blood periloop environmental feature data are collected, where: The nail fold blood vessel morphological feature data specifically includes the diameter div of the input branch of the nail fold blood vessel, the diameter dov of the output branch of the nail fold blood vessel, and the length Length of the nail fold blood vessel; The specific data of the blood flow characteristics of the nail fold include the blood flow velocity v of the nail fold blood real , the density drbca of the red blood cell aggregation area in the nail fold blood, and the number Number of white blood cells in the nail fold blood w ; The specific data on the characteristics of the perivascular environment of the nail fold blood loop include the size of the subpapillary venous plexus Size plexus , the area of perivascular hemorrhage Area; Construct a nail fold blood vessel - blood flow - periloop joint model. Based on the nail fold blood vessel morphological feature data, the nail fold blood flow feature data, and the nail fold blood periloop environmental feature data, output the nail fold blood joint analysis factor. The nail fold blood joint analysis factor is used as the analysis basis for obtaining the updated nail fold blood joint analysis factor.

6. A method for analyzing nailfold blood microcirculation image data according to claim 5, characterized in that: The nail fold blood vessel - blood flow - periloop joint model, the specific analysis process is as follows: In the formula, combine is the nail fold blood joint analysis factor, vessel is the nail fold blood vessel analysis factor, flow is the nail fold blood flow analysis factor, periloop is the nail fold blood periloop environmental analysis factor, A1 is the set weight factor of vessel, A2 is the set weight factor of flow, A3 is the set weight factor of periloop, and e is the natural constant.

7. A method for analyzing nailfold blood microcirculation image data according to claim 1, characterized in that: The construction of the nail fold blood microcirculation analysis influence model outputs the nail fold blood microcirculation analysis influence factor. The specific analysis process is as follows: Obtain the nail fold blood microcirculation analysis feature data. The nail fold blood microcirculation analysis feature data specifically includes the temperature tem of the nail fold blood microcirculation detection environment, the resolution f bw of the nail fold blood microcirculation detection imaging device, and the magnification bn of the nail fold blood microcirculation detection imaging device; Construct a nail fold blood microcirculation analysis influence model. Based on the nail fold blood microcirculation analysis feature data, output the nail fold blood microcirculation analysis influence factor. The nail fold blood microcirculation analysis influence factor is used as the analysis basis for obtaining the updated nail fold blood joint analysis factor; The nail fold blood microcirculation analysis influence model, the specific analysis process is as follows: In the formula, Inf is the nail fold blood microcirculation analysis influence factor, and e is the natural constant.

8. A method for analyzing nailfold blood microcirculation image data according to claim 7, characterized in that: Based on the nail fold blood microcirculation analysis influence factor, the nail fold blood joint analysis factor is corrected to obtain the updated nail fold blood joint analysis factor. The specific analysis process is as follows: Wherein, is the updated combined nailfold blood analysis factor, combine is the combined nailfold blood analysis factor, S1 is the set weight factor of combine, and S2 is the set weight factor of Inf.

9. A method for analyzing nailfold blood microcirculation image data according to claim 1, characterized in that: Based on the updated nail fold blood joint analysis factor, using the support vector machine model, output the classification result of the nail fold blood microcirculation image data. The specific analysis process is as follows: Obtain the data of nailfold blood samples with known states stored in the database, and divide them into a training set and a test set according to a set ratio; Train the support vector machine model using the training set by minimizing the objective function while satisfying the constraint conditions and 0 ≤ α i ≤ C; where, α i is the Lagrange multiplier corresponding to the i-th nailfold blood sample, α j is the Lagrange multiplier corresponding to the j-th nailfold blood sample, y i is the i-th nailfold blood sample, y j is the j-th nailfold blood sample, i and j are the nailfold blood sample numbers, n is the total number of nailfold blood samples, C is the penalty parameter, K(x i , x j ) is the radial basis kernel function, σ is the width parameter of the radial basis kernel function, x i is the feature vector of the i-th nailfold blood sample, x j is the feature vector of the j-th nailfold blood sample; Input the trained support vector machine model into the preprocessed updated nailfold blood joint analysis factors, classify the states of nailfold blood through the decision function, and output the classification results of nailfold blood microcirculation image data.

10. A nail fold blood microcirculation image data analysis system, which is applied to the nail fold blood microcirculation image data analysis method according to any one of claims 1-9, and is characterized in that, It includes an original image data preprocessing module, a blood vessel segmentation module, a blood flow velocity analysis and correction module, a joint analysis factor output module, an analysis influencing factor output module, a joint analysis factor update module, and a classification result output module, where: The original image data preprocessing module is used to collect the original image data of nailfold blood microcirculation using a microcirculation imaging device and preprocess the original image data of nailfold blood microcirculation; The blood vessel segmentation module is used to perform nailfold blood vessel segmentation using a U-Net convolutional neural network model based on deep learning; The blood flow velocity analysis and correction module is used to calculate the nailfold blood flow velocity using the particle image velocimetry algorithm in the segmented nailfold blood vessel image and correct the nailfold blood flow velocity; The joint analysis factor output module is used to collect the nailfold blood vessel morphological feature data, nailfold blood flow feature data, and nailfold blood loop perimeter environment feature data, construct a nailfold blood vessel-blood flow-loop perimeter joint model, and output the nailfold blood joint analysis factors; The analysis influencing factor output module is used to obtain the nailfold blood microcirculation analysis feature data, construct a nailfold blood microcirculation analysis influencing model, and output the nailfold blood microcirculation analysis influencing factors; The joint analysis factor update module is used to correct the nailfold blood joint analysis factors based on the nailfold blood microcirculation analysis influencing factors to obtain updated nailfold blood joint analysis factors; The classification result output module is used to output the classification results of nailfold blood microcirculation image data based on the updated nailfold blood joint analysis factors using a support vector machine model.

Images