Collagen activity analysis method based on artificial intelligence

Abstract

The invention relates to the field of biomedical image processing, in particular to a collagen activity analysis method based on artificial intelligence. The method comprises the following steps: collecting and preprocessing image data to obtain an initial feature value; performing nonlinear mapping on the initial characteristic value to obtain a characteristic value; introducing a recursive attention mechanism, and carrying out layer-by-layer weighting and nonlinear combination on the feature value to obtain a final enhanced feature value; and introducing a self-supervised correction mechanism, correcting the final enhanced characteristic value, and calculating to obtain a collagen activity score value. The problems that a traditional algorithm has high dependence on large-scale annotation data, and consequently an analysis result is unstable are solved; a traditional static feature extraction method is difficult to effectively capture microcosmic dynamic changes of collagen along with time, and the sensitivity of analysis is reduced; an existing scoring strategy mostly adopts a linear processing method and is easily interfered by extreme data, so that the scoring result is distorted, and the judgment on the activity level of collagen is influenced.

Description

Technical Field

[0001] The present invention relates to the field of biomedical image processing, and in particular to a collagen activity analysis method based on artificial intelligence. Background Art

[0002] Collagen is an important component of tissues such as skin, cartilage, and joints. The concentration and activity level of collagen directly affect the health of tissues. At present, collagen analysis mainly relies on microscopic imaging technology, such as confocal microscopes and second harmonic microscopy systems, which can capture high-resolution images of tissue samples. In order to extract useful information from the images, convolutional neural networks (CNNs) or other deep learning models are traditionally used for feature extraction. However, the deep learning models usually require a large amount of labeled data to achieve good performance, and lack the ability to capture time-related changes, which limits the application of collagen in microstructure and dynamic analysis.

[0003] In summary, the current collagen activity analysis methods have shortcomings in the following aspects: First, convolutional neural networks (CNN) and other traditional algorithms are highly dependent on large-scale labeled data. When the sample data volume is small, the generalization ability and adaptability of the model will significantly decrease, resulting in unstable analysis results; second, the traditional static feature extraction method can only capture the characteristic information of the sample at a certain moment, and it is difficult to effectively capture the microscopic dynamics of collagen over time, which will directly affect the accurate assessment of the activity state of collagen and reduce the sensitivity of the analysis; in addition, the existing scoring strategies mostly use linear processing methods, which lack robustness to noise and outliers, and are easily disturbed by extreme data, resulting in distorted scoring results and affecting the judgment of the activity level of collagen. Therefore, it is urgent to improve it through a new algorithm and processing technology. Summary of the invention

[0004] The present invention provides a collagen activity analysis method based on artificial intelligence to solve the problems that convolutional neural networks (CNN) and other traditional algorithms are highly dependent on large-scale labeled data. When the amount of sample data is small, the generalization ability and adaptability of the model will be significantly reduced, resulting in unstable analysis results; traditional static feature extraction methods can only capture the characteristic information of the sample at a certain moment, and it is difficult to effectively capture the microscopic dynamics of collagen changing over time, which will directly affect the accurate evaluation of the activity state of collagen and reduce the sensitivity of the analysis; existing scoring strategies mostly use linear processing methods, which lack robustness to noise and outliers and are easily disturbed by extreme data, resulting in distorted scoring results and affecting the judgment of the activity level of collagen.

[0005] The present invention provides a method for analyzing collagen activity based on artificial intelligence, which specifically includes the following technical solutions:

[0006] A collagen activity analysis method based on artificial intelligence, comprising the following steps:

[0007] S1: Collect image data and perform preprocessing to obtain initial eigenvalues; perform nonlinear mapping on the initial eigenvalues ​​to obtain eigenvalues; introduce a recursive attention mechanism to perform layer-by-layer weighting and nonlinear combination on the eigenvalues ​​to obtain the final enhanced eigenvalues;

[0008] S2: A self-supervised correction mechanism is introduced to correct the final enhanced eigenvalue to obtain a corrected eigenvalue; based on the corrected eigenvalue, a nonlinear error weighted average scoring strategy is adopted to calculate the collagen activity score.

[0009] Preferably, the S1 specifically includes:

[0010] The specific implementation process of the preprocessing includes: converting the image data into a single-channel grayscale image through grayscale processing; performing denoising processing on the obtained single-channel grayscale image to obtain denoised image data; dividing the denoised image data into image blocks, and treating the image blocks as feature points, performing feature extraction operations on the image blocks, and calculating the average intensity value of the image blocks as the initial feature value.

[0011] Preferably, the S1 specifically includes:

[0012] Adopting the adaptive nonlinear mapping algorithm, the adaptive weight is dynamically adjusted according to the time step, and the initial eigenvalue is nonlinearly mapped to obtain the eigenvalue; the specific formula is as follows:

[0013]

[0014] Among them, f i (t) is the feature value of the i-th feature point at time step t; i is the index variable of the feature point; n is the total number of feature points; k is the index variable of the feature point in the accumulation process; is the adaptive weight, which represents the adaptive weight of the k-th feature point in the nonlinear mapping at time step t; z i is the initial eigenvalue of the i-th feature point at time step t; β k is the scaling factor of the kth feature point; k is the time adjustment coefficient of the kth feature point.

[0015] Preferably, the S1 specifically includes:

[0016] The feature value is used as the input of the recursive attention mechanism, and the initial enhancement operation is used to perform a preliminary weighted combination on each feature point to obtain the initial value of the enhanced feature value.

[0017] Preferably, the S1 specifically includes:

[0018] Based on the initial value of the enhanced eigenvalue, recursive weighted calculation is performed layer by layer. The calculation of each layer is advanced and adjusted based on the result of the previous layer, and the enhanced eigenvalue is output. When the set total number of recursive layers is reached, the enhanced eigenvalue output by the last layer is used as the final enhanced eigenvalue. The recursive weighted formula is as follows:

[0019]

[0020] Among them, F i (l) (t) is the enhanced feature value output at layer l of the i-th feature point at time step t; l is the layer index in the recursive attention mechanism; is the weight coefficient of the i-th feature point in the l-1th layer recursive weighting process at time step t; is the weight coefficient of the k-th feature point in the recursive weighting process of the l-1 layer at time step t; F i (l-1) (t) is the enhanced feature value output at the l-1 layer at the i-th feature point at time step t; is the enhanced feature value outputted at the l-1 layer at the k-th feature point at time step t; ξ is the smoothing coefficient.

[0021] Preferably, the S2 specifically includes:

[0022] In the process of implementing the self-supervised correction mechanism, the internal reference value is obtained based on the final enhanced eigenvalue; based on the final enhanced eigenvalue and the internal reference value, the sine function and cosine function are introduced to calculate the deviation between the final enhanced eigenvalue and the overall collagen activity level of the current biological tissue sample, and the error of the feature point is obtained; the error calculation formula is as follows:

[0023]

[0024] Among them, ∈ i is the error between the final enhanced feature value of the i-th feature point and the internal reference value; L is the total number of recursive layers in the recursive attention mechanism; F i (L) (t) is the final enhanced feature value of the i-th feature point at time step t; is the final enhanced feature value of the kth feature point at time step t; y is the internal reference value; is the sensitivity coefficient; k is the feature point index variable in the accumulation process; n is the total number of feature points.

[0025] Preferably, the S2 specifically includes:

[0026] Based on the error of the feature points, a correction weight coefficient is introduced to perform nonlinear correction on the final enhanced eigenvalue to obtain the corrected eigenvalue.

[0027] Preferably, the S2 specifically includes:

[0028] In the scoring strategy of nonlinear error weighted average, the error of the feature point is used as the weight, and the corrected feature values ​​of all feature points are nonlinearly weighted averaged to calculate the collagen activity score.

[0029] The beneficial effects of the technical solution of the present invention are:

[0030] 1. The present invention dynamically adjusts the initial eigenvalue through an adaptive nonlinear mapping algorithm and captures the microscopic changes of the characteristic points according to the time step; the adaptive nonlinear mapping algorithm can flexibly respond to the fluctuations of collagen activity in different time periods, making the collagen activity analysis method more sensitive to changes in activity levels, improving the accuracy and real-time performance of the analysis, and providing effective support for tissue health monitoring and repair evaluation.

[0031] 2. The present invention introduces a recursive attention mechanism, which strengthens the complex correlation between feature points and accumulates important information layer by layer through layer-by-layer weighting and nonlinear combination; the recursive attention mechanism not only improves the detection accuracy of the collagen activity analysis method for the collagen activity level, but also enhances the ability to capture subtle feature changes, making the analysis results more comprehensive and reliable.

[0032] 3. The present invention introduces a self-supervised correction mechanism to perform nonlinear correction on the final enhanced feature value, uses the internal benchmark value as a reference, avoids relying on real labeled data, and enhances the adaptability of the collagen activity analysis method; at the same time, adopts a nonlinear error weighted average scoring strategy to make the contribution of abnormal feature points in the overall score more reasonable, ensures that the analysis results are more accurate, and can effectively deal with the interference of noise and extreme values. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 This is a flow chart of the collagen activity analysis method based on artificial intelligence described in the present invention. DETAILED DESCRIPTION

[0034] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined purpose of the invention, the technical scheme in the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiment of the present invention. Obviously, the described embodiment is only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0035] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0036] The specific scheme of the collagen activity analysis method based on artificial intelligence provided by the present invention is described in detail below with reference to the accompanying drawings.

[0037] See attached Figure 1 , which shows a flow chart of a collagen activity analysis method based on artificial intelligence provided by an embodiment of the present invention, the method comprising the following steps:

[0038] S1: Collect image data and perform preprocessing to obtain initial eigenvalues; perform nonlinear mapping on the initial eigenvalues ​​to obtain eigenvalues; introduce a recursive attention mechanism to perform layer-by-layer weighting and nonlinear combination on the eigenvalues ​​to obtain the final enhanced eigenvalues;

[0039] First, image data is collected. The image data required for collagen activity analysis comes from a biological microscopic imaging device, such as a confocal microscope or a second harmonic microscopic imaging system. The biological microscopic imaging device can capture high-resolution images of biological tissue samples, such as skin, joints, cartilage, tendons, and other biological tissue samples rich in collagen;

[0040] Then, the image data is preprocessed to obtain the initial eigenvalue; the preprocessing includes: grayscale processing, denoising processing, image segmentation and feature extraction. Specifically, the image data is a color RGB image, and the color RGB image is converted into a single-channel grayscale image through grayscale processing; then a Gaussian filter is used to perform denoising on the obtained single-channel grayscale image to eliminate random noise and obtain denoised image data; the denoised image data is divided into multiple image blocks of fixed size, and each image block is regarded as a feature point, that is, each image block represents a feature point of a local area, and a feature extraction operation is performed on the image block. The size of the image block is set according to the specific implementation scenario, which will not be elaborated here. The average intensity value of the image block is calculated as the initial eigenvalue, which represents the collagen signal intensity of the local area represented by the image block. The collagen signal intensity reflects the concentration and distribution state of collagen in biological tissue samples. The stronger the collagen signal, the higher the collagen content, the more complete the structure, and the stronger the collagen activity level;

[0041] Furthermore, the initial eigenvalues ​​are nonlinearly mapped to obtain eigenvalues. The nonlinear mapping uses an adaptive nonlinear mapping algorithm to dynamically adjust the adaptive weights according to the time step, increase the accuracy of feature expression, and improve the sensitivity and reliability of collagen activity analysis. The specific formula is as follows:

[0042]

[0043] Among them, f i (t) is the feature value of the i-th feature point at time step t; i is the index variable of the feature point; t is the time step, which is used to control the dynamic adjustment in the mapping process. The time step is set according to the specific implementation scenario and will not be described here; n is the total number of feature points, that is, the total number of image blocks, which comes from the total number of image blocks obtained after the image data is divided according to the preset image block size in the preprocessing process; k is the feature point index variable in the accumulation process; is the adaptive weight, which represents the adaptive weight of the kth feature point in the nonlinear mapping at time step t, and is used to dynamically adjust the contribution of the feature point in the mapping process to ensure that the influence of different feature points can be flexibly adjusted at different time steps; i is the initial eigenvalue of the i-th feature point at time step t; β k is the scaling factor, which represents the scaling factor of the kth feature point, used to adjust the amplitude of the initial eigenvalue during the mapping process, and is obtained through experimental tuning; k is the time adjustment coefficient of the kth feature point, used to control the adaptive weight The rate of change is obtained through experimental tuning;

[0044] Finally, a recursive attention mechanism is introduced to perform weighted and nonlinear combination of the eigenvalues ​​layer by layer to obtain the final enhanced eigenvalues. The recursive attention mechanism dynamically adjusts the weights of the feature points through multi-layer recursive calculations, and gradually strengthens the feature expression through nonlinear combinations of each layer, outputs enhanced eigenvalues, and the enhanced eigenvalues ​​output by the last layer of the recursive attention mechanism are the final enhanced eigenvalues. Compared with the traditional attention mechanism, the recursive attention mechanism can not only capture the complex relationship between features, but also accumulate information as the number of layers increases, reflect the local and overall changes in collagen activity, and improve the sensitivity and robustness to subtle features. The calculation process of the recursive attention mechanism is as follows:

[0045] The feature value is used as the input of the recursive attention mechanism, and the initial enhancement operation is used to perform a preliminary weighted combination on each feature point to obtain the initial value of the enhanced feature value, laying the foundation for subsequent recursive weighting. The specific formula of the initial enhancement operation is:

[0046]

[0047] Among them, F i (0) (t) is the initial value of the enhanced eigenvalue of the i-th feature point at time step t; f i (t) is the feature value of the i-th feature point at time step t; n is the number of feature points; k is the feature point index variable in the accumulation process; γ kis the dynamic weighting coefficient of the kth eigenvalue, which is used to control the importance of different eigenvalues ​​in the combination process. It is obtained by normalizing all eigenvalues ​​and dividing the normalized kth eigenvalue by the sum of all normalized eigenvalues. cos(.) is the cosine function, which is used to capture the periodic relationship between eigenvalues. f k (t) is the eigenvalue of the kth feature point at time step t;

[0048] The initial enhancement operation captures the periodic relationship between feature points and outputs the initial value of the enhanced eigenvalue. The initial value of the enhanced eigenvalue is used as the starting point of the recursive calculation. On the basis of the initial value of the enhanced eigenvalue, recursive weighted calculation is performed layer by layer. The calculation of each layer is promoted and adjusted based on the result of the previous layer. The recursive weighted formula is as follows:

[0049]

[0050]

[0051] Among them, F i (l) (t) is the enhanced feature value output at layer l of the i-th feature point at time step t; l is the layer index in the recursive attention mechanism, l∈{1, 2, ..., L}, where L is the total number of recursive layers in the recursive attention mechanism, which is set according to the specific implementation scenario; is the weight coefficient, which represents the weight coefficient of the i-th feature point in the l-1th layer recursive weighting process at time step t; is the weight coefficient, which represents the weight coefficient of the k-th feature point in the recursive weighting process of the l-1 layer at time step t; F i (l-1) (t) is the enhanced feature value output at the l-1 layer at the i-th feature point at time step t, F i (0) (t) is the initial value of the enhanced eigenvalue of the i-th feature point at time step t; is the enhanced feature value output at the l-1 layer of the k-th feature point at time step t; ξ is a smoothing coefficient, which is used to adjust the sensitivity of the weight distribution to control the importance difference of the enhanced feature value in the recursive weighting process, and is obtained through experimental tuning;

[0052] The recursive process continues for L layers. When the set total number of recursive layers L is reached, the enhanced feature value output by the last layer (i.e., the Lth layer) is the output result of the feature point in the recursive attention mechanism, i.e., the final enhanced feature value.

[0053] S2: A self-supervised correction mechanism is introduced to correct the final enhanced eigenvalue to obtain a corrected eigenvalue; based on the corrected eigenvalue, a nonlinear error weighted average scoring strategy is adopted to calculate the collagen activity score.

[0054] The recursive attention mechanism can capture subtle changes in collagen activity levels by weighting the influence of dynamic enhancement features layer by layer, but it will also introduce noise or extreme values. Therefore, it is necessary to introduce a self-supervised correction mechanism to correct the final enhanced feature value to obtain a corrected feature value. The self-supervised correction mechanism uses an internal reference value to correct the final enhanced feature value to ensure the accuracy of feature expression, and the corrected feature value is used for subsequent calculation of collagen activity scores. The internal reference value is the mean of all final enhanced feature values ​​obtained by calculation, and the internal reference value is used as an adaptive reference standard to reflect the overall collagen activity level of the current biological tissue sample. The specific implementation process of the self-supervised correction mechanism is as follows:

[0055] Firstly, the error of the feature point is calculated using the difference between the final enhanced feature value and the internal reference value. The error of the feature point reflects the degree of deviation between the final enhanced feature value and the internal reference value, and is used to measure the consistency and abnormality of the feature point with the overall collagen activity level.

[0056] The error calculation formula is as follows:

[0057]

[0058] Among them, ∈ i is the error between the final enhanced feature value of the i-th feature point and the internal reference value, which is used to measure the degree to which the final enhanced feature value deviates from the overall collagen activity level of the current biological tissue sample; F i (L) (t) is the final enhanced feature value of the i-th feature point at time step t; is the final enhanced eigenvalue of the kth feature point at time step t; y is the internal reference value, which represents the mean of the final enhanced eigenvalues ​​of all feature points and represents the overall collagen activity level of the current biological tissue sample; is the sensitivity coefficient, which is used to control the response of the sine function and cosine function to the change of the final enhanced eigenvalue, and is obtained through experimental tuning; k is the feature point index variable in the accumulation process, which is used in the error calculation formula to traverse the final enhanced eigenvalues ​​of all feature points for normalization calculation;

[0059] The error calculation formula calculates the deviation between the final enhanced feature value of the feature point and the overall collagen activity level of the current biological tissue sample, enhances the sensitivity to feature abnormalities through sine and cosine functions, and uses a normalized denominator to avoid the influence of individual abnormal features;

[0060] Furthermore, based on the error of the feature points, each final enhanced feature value is corrected to obtain a corrected feature value. The correction formula is as follows:

[0061]

[0062] in, is the corrected eigenvalue of the ith feature point at time step t; η i is the correction weight coefficient of the i-th feature point, which is used to adjust the contribution of the feature point in the error correction process and is obtained through experimental tuning;

[0063] The self-supervised correction mechanism performs nonlinear correction on the final enhanced eigenvalue, making it closer to the internal benchmark value, thereby improving the stability and accuracy of the collagen analysis results.

[0064] Based on the corrected eigenvalues, a scoring strategy of nonlinear error weighted average is adopted, and the error of the feature point is used as the weight to reflect the importance of the feature point. The feature point with a larger error has a higher weight in the collagen activity score, ensuring that the collagen activity analysis pays more attention to the feature points that deviate from the internal benchmark value and improves the sensitivity of the analysis. The collagen activity score is calculated by taking a nonlinear weighted average of the corrected eigenvalues ​​of all feature points, so as to accurately reflect the true collagen activity level and improve the robustness and reliability of the collagen activity analysis.

[0065] The scoring formula is as follows:

[0066]

[0067] Among them, Score is the collagen activity score; ∈ k is the error of the kth feature point.

[0068] The collagen activity score reflects the concentration, distribution and functional status of collagen in biological tissue samples by quantifying and modifying characteristic values. The higher the collagen activity score, the more complete the collagen structure, the higher the density and the stronger the activity. A lower collagen activity score indicates fiber breakage, degradation or weakened activity, reflecting the degree of degradation or damage of biological tissues. The collagen activity score is used to judge the health status of biological tissues, formulate repair or treatment plans, and track the efficacy and progress of tissue recovery.

[0069] In summary, an artificial intelligence-based collagen activity analysis method was completed.

[0070] The order of the embodiments of the invention is for description only and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0071] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

[0072] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that the technical solutions described in the above embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention.

Claims

1. A collagen activity analysis method based on artificial intelligence, characterized in that: The following steps are involved: S1: collect image data and preprocess them to obtain initial eigenvalues; Perform nonlinear mapping on the initial eigenvalue to obtain the eigenvalue; Introducing a recursive attention mechanism to perform layer-by-layer weighting and nonlinear combination of eigenvalues ​​to obtain the final enhanced eigenvalues; S2: Introduce a self-supervised correction mechanism to correct the final enhanced eigenvalue to obtain the corrected eigenvalue; Based on the modified eigenvalue, the collagen activity score was calculated using a nonlinear error weighted average scoring strategy.

2. The method for analyzing collagen activity based on artificial intelligence according to claim 1, characterized in that: The S1 specifically includes: The specific implementation process of the preprocessing includes: converting the image data into a single-channel grayscale image through grayscale processing; performing denoising processing on the obtained single-channel grayscale image to obtain denoised image data; dividing the denoised image data into image blocks, and treating the image blocks as feature points, performing feature extraction operations on the image blocks, and calculating the average intensity value of the image blocks as the initial feature value.

3. The method for analyzing collagen activity based on artificial intelligence according to claim 2, characterized in that: The S1 specifically includes: Adopting the adaptive nonlinear mapping algorithm, the adaptive weight is dynamically adjusted according to the time step, and the initial eigenvalue is nonlinearly mapped to obtain the eigenvalue; the specific formula is as follows: Among them, f i (t) is the feature value of the i-th feature point at time step t; i is the index variable of the feature point; n is the total number of feature points; k is the index variable of the feature point in the accumulation process; is the adaptive weight, which represents the adaptive weight of the k-th feature point in the nonlinear mapping at time step t; z i is the initial eigenvalue of the i-th feature point at time step t; β k is the scaling factor of the kth feature point; k is the time adjustment coefficient of the kth feature point.

4. The method for analyzing collagen activity based on artificial intelligence according to claim 3, characterized in that: The S1 specifically includes: The feature value is used as the input of the recursive attention mechanism, and the initial enhancement operation is used to perform a preliminary weighted combination on each feature point to obtain the initial value of the enhanced feature value.

5. The method for analyzing collagen activity based on artificial intelligence according to claim 4, characterized in that: The S1 specifically includes: Based on the initial value of the enhanced eigenvalue, recursive weighted calculation is performed layer by layer. The calculation of each layer is advanced and adjusted based on the result of the previous layer, and the enhanced eigenvalue is output. When the set total number of recursive layers is reached, the enhanced eigenvalue output by the last layer is used as the final enhanced eigenvalue. The recursive weighted formula is as follows: Among them, F i (l) (t) is the enhanced feature value output at layer l of the i-th feature point at time step t; l is the layer index in the recursive attention mechanism; is the weight coefficient of the i-th feature point in the l-1th layer recursive weighting process at time step t; is the weight coefficient of the k-th feature point in the recursive weighting process of the l-1 layer at time step t; F i (l-1) (t) is the enhanced feature value output at the l-1 layer at the i-th feature point at time step t; is the enhanced feature value outputted at the l-1 layer at the k-th feature point at time step t; ξ is the smoothing coefficient.

6. The method for analyzing collagen activity based on artificial intelligence according to claim 1, characterized in that: The S2 specifically includes: In the process of implementing the self-supervised correction mechanism, the internal reference value is obtained based on the final enhanced eigenvalue; based on the final enhanced eigenvalue and the internal reference value, the sine function and cosine function are introduced to calculate the deviation between the final enhanced eigenvalue and the overall collagen activity level of the current biological tissue sample, and the error of the feature point is obtained; the error calculation formula is as follows: Among them, ∈ i is the error between the final enhanced feature value of the i-th feature point and the internal reference value; L is the total number of recursive layers in the recursive attention mechanism; F i (L) (t) is the final enhanced feature value of the i-th feature point at time step t; is the final enhanced feature value of the kth feature point at time step t; y is the internal reference value; is the sensitivity coefficient; k is the feature point index variable in the accumulation process; n is the total number of feature points.

7. The method for analyzing collagen activity based on artificial intelligence according to claim 6, characterized in that: The S2 specifically includes: Based on the error of the feature points, a correction weight coefficient is introduced to perform nonlinear correction on the final enhanced eigenvalue to obtain the corrected eigenvalue.

8. The method for analyzing collagen activity based on artificial intelligence according to claim 7, characterized in that: The S2 specifically includes: In the scoring strategy of nonlinear error weighted average, the error of the feature point is used as the weight, and the corrected feature values ​​of all feature points are nonlinearly weighted averaged to calculate the collagen activity score.

Images