Disease marker structure evolution characteristic change point determination method

Abstract

A disease marker structure evolution characteristic change point determination method belongs to the field of disease markers, and comprises the steps of obtaining and preprocessing time sequence structure characteristic data of a disease marker, constructing a characteristic transformation image and calculating characteristic intensity distribution, establishing a structure characteristic fitting model and calculating a time evolution coefficient and an intensity evolution coefficient, determining a structural feature evolution trajectory and calculating a correlation index; establishing a structural feature piecewise function and identifying a feature mutation point; calculating a structural feature contribution value and generating a feature evolution matrix; calculating an evolution stability index and determining a structural feature change point; hierarchical clustering is carried out, main structural feature change points and secondary structural feature change points are determined, a feature weight distribution diagram is constructed, a change point time sequence table is generated, a time sequence corresponding relation is established, and a structural feature change point determination result is output; and fine analysis and accurate description of structural evolution characteristics of complex disease markers are realized.

Description

Technical Field

[0001] The present invention belongs to the technical field of disease markers, and more specifically, relates to a method for determining change points in the structural evolution characteristics of disease markers. Background Art

[0002] The analysis of the structural evolution characteristics of disease markers is an important research topic in the field of biomedicine. By accurately capturing the key change points in the temporal evolution of the marker structure, it can provide crucial basis for the early diagnosis, prognosis judgment, and treatment plan formulation of diseases. Currently, the commonly used methods for identifying change points in the structural evolution characteristics of disease markers mainly include technical means such as statistical analysis, spectral analysis, and pattern recognition.

[0003] The method based on statistical analysis usually first segments the time series data, and then calculates the statistical characteristics of each segment of data such as mean, variance, etc. When these statistics exceed the preset threshold, that time point is considered as a change point. This method is simple to implement, but it is difficult to cope with the non-stationarity and sudden changes in the marker structure evolution process. For example, in the diagnosis of some diseases, the change of marker concentration over time often shows non-linear and non-continuous characteristics, making it difficult for a single statistical threshold to accurately capture the change point.

[0004] The method based on spectral analysis is to transform the time series data into the frequency domain space, and identify the change points by analyzing spectral characteristics such as energy distribution, frequency components, etc. This method can provide multi-scale analysis capabilities, but it has limitations in dealing with non-linear evolution processes. Since the disease marker structure usually shows complex non-linear dynamic characteristics, a single spectral analysis method is difficult to comprehensively describe its evolution law.

[0005] The method based on pattern recognition identifies the change points by constructing feature templates and using pattern matching. The specific approach is to pre-define several typical change patterns, and then calculate the similarity between the measured data and these templates to determine the change points. This method highly depends on prior knowledge and is difficult to adapt to the actual complex and changeable situations. For example, in practical applications, there are often various complex change patterns in the disease marker structure evolution process, and it is difficult to exhaust and construct a complete template library in advance.

[0006] Generally speaking, the existing methods for identifying change points in the structural evolution characteristics of disease markers have problems in that they cannot effectively handle non-stationary and non-linear evolution characteristics, resulting in difficulties in comprehensively capturing the multi-dimensional feature changes in the evolution process. Summary of the Invention

[0007] In view of this, the present invention provides a method for determining the change points of the structural evolution characteristics of disease markers, which can solve the problem that the existing methods for identifying the change points of the structural evolution characteristics of disease markers cannot effectively process non-stationary and non-linear evolution characteristics, resulting in difficulty in comprehensively capturing the multi-dimensional characteristic changes in the evolution process. The method includes the following operation steps:

[0008] Obtain the time-series structural characteristic data of disease markers and perform preprocessing, construct a feature transformation image and calculate the feature density distribution, establish a structural feature fitting model and calculate the time evolution coefficient and the intensity evolution coefficient, determine the structural feature evolution trajectory and calculate the correlation index, establish a structural feature piecewise function and identify the feature mutation points, calculate the structural feature contribution value and generate a feature evolution matrix, calculate the evolution stability index and determine the change points of the structural features, perform hierarchical clustering and determine the main change points and secondary change points of the structural features, construct a feature weight distribution map and generate a change point time-series table, establish a time-series correspondence relationship and output the result of determining the change points of the structural features.

[0009] Among them, to obtain the time-series structural characteristic data of disease markers, specifically, a biomarker mass spectrometry system is used to obtain marker sample data, record the mass spectrometry map information of the markers, collect the marker structure information, and perform time-series integration on the mass spectrometry map information and the marker structure information to form the time-series structural characteristic data of disease markers.

[0010] Among them, to construct the feature transformation image of the time-series structural characteristic data of the disease markers, specifically, a time-intensity coordinate system is established for the preprocessed time-series data, the data points are made continuous, and a time-series grayscale image is generated; to calculate the feature density distribution in the feature transformation image, specifically, the local peak points in the image are identified, the number distribution of the peak points is statistically analyzed, the distances between the peak points are calculated, and a peak point distribution function is formed.

[0011] Among them, to establish a structural feature fitting model based on the feature density distribution, specifically, the characteristic points of the distribution curve are selected, the fitting coefficients are determined, the fitting parameters are adjusted, and a fitting curve is obtained; to calculate the time evolution coefficient and the intensity evolution coefficient of the structural feature fitting model, specifically, the first-order change rate is calculated in the time dimension as the time evolution coefficient, and the second-order change rate is calculated in the intensity dimension as the intensity evolution coefficient.

[0012] Among them, to determine the structural feature evolution trajectory according to the time evolution coefficient and the intensity evolution coefficient, specifically, the time evolution coefficient and the intensity evolution coefficient are synthesized in direction to obtain the evolution direction and form the structural feature evolution trajectory; to calculate the correlation index of the structural feature evolution trajectory, specifically, the similarity between the trajectory points is calculated, the local correlation degree is statistically analyzed, and the overall correlation index is obtained.

[0013] Among them, the correlation index is numerically segmented to establish a structural feature piecewise function, and the characteristic mutation points of the structural feature piecewise function are identified; the structural feature contribution values at the characteristic mutation points are calculated, and the structural feature contribution values are used to characterize the evolution intensity of the structural features at the characteristic mutation points.

[0014] Among them, establishing the characteristic mapping relationship between the structural feature contribution value and the structural feature evolution trajectory specifically involves constructing a characteristic transformation matrix to map the structural feature contribution value to the evolution feature space; generating a characteristic evolution matrix, specifically arranging the characteristic mapping data in time sequence to construct a two-dimensional time sequence characteristic matrix.

[0015] Among them, calculating the evolution stability index based on the characteristic evolution matrix specifically involves analyzing the time sequence correlation of matrix elements, calculating the characteristic fluctuation degree, and evaluating the stability of the evolution process; determining the structural feature change points according to the evolution stability index, specifically setting a stability threshold, identifying the moment of index change, and determining the key points of feature change.

[0016] Among them, hierarchical clustering is performed on the structural feature change points to determine the main structural feature change points and the secondary structural feature change points; the characteristic weight values of the main structural feature change points are calculated, and a characteristic weight distribution diagram is constructed.

[0017] Among them, the key structural feature change points are identified according to the characteristic weight distribution diagram to generate a time sequence table of change points; the time sequence correspondence between the time sequence table of change points and the structural evolution process of disease markers is established, and the determination result of the structural feature change points is output.

[0018] The effects of the present invention are as follows:

[0019] The present invention proposes a method for determining the change points of the structural evolution characteristics of disease markers based on data driving, which can effectively solve the above technical problems. Through a series of innovative technical means such as constructing a two-dimensional transformation image, establishing a structural feature fitting model, determining the evolution trajectory, and analyzing the evolution stability, the method realizes the refined analysis and quantitative description of the marker structure evolution characteristics. Compared with traditional statistical analysis, spectrum analysis, and pattern recognition methods, the method of the present invention has the following advantages:

[0020] 1. It has strong adaptability. This method does not require a large number of empirical parameters to be set artificially, and can automatically adjust the analysis strategy according to data characteristics, improving the reliability of the analysis results. Traditional methods often require a lot of manual experience to determine key parameters such as thresholds and templates, and it is difficult to adapt to complex and changeable actual situations.

[0021] 2. It has good anti-interference ability. The method of the present invention effectively reduces the influence of noise on the results through a multi-level feature extraction and analysis mechanism. In contrast, the existing methods are vulnerable to noise interference because they rely too much on a single statistic or spectral feature.

[0022] 3. It has clear physical interpretability. Each step of this method has clear physical and biological meanings, improving the interpretability of the analysis results. In contrast, traditional pattern recognition methods lack in-depth understanding of physical mechanisms because they rely too much on empirical knowledge.

[0023] 4. It realizes hierarchical recognition of change points. The method of the present invention provides more detailed information reference for the in-depth study of the structural evolution characteristics of disease markers by dividing major change points and minor change points. Such a hierarchical recognition mechanism is difficult to achieve in existing methods.

[0024] Generally speaking, the method of the present invention makes full use of the inherent characteristics of the data itself, overcomes the limitations of the existing technologies, and provides strong support for the refined analysis and accurate characterization of the structural evolution characteristics of disease markers. It solves the problem that the existing methods for identifying change points in the structural evolution characteristics of disease markers cannot effectively process non-stationary and non-linear evolution characteristics, resulting in difficulty in comprehensively capturing multi-dimensional feature changes in the evolution process. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 is a flowchart of the method provided by the present invention;

[0026] Figure 2 is the transformed image of the marker time-series structural feature in Example 2;

[0027] Figure 3 is the structural feature evolution trajectory diagram in Example 2;

[0028] Figure 4 is the stability index change diagram in Example 2. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0029] To make the objectives, technical solutions and advantages of the embodiments of the present invention clearer, 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.

[0030] As Figure 1 shown, the present invention includes the following operation steps:

[0031] S01. Obtain the disease marker time-series structural feature data, and preprocess the disease marker time-series structural feature data to eliminate noise interference;

[0032] S02. Construct a feature transformation image of the time-series structural feature data of the disease biomarker, and calculate the feature density distribution in the feature transformation image;

[0033] S03. Establish a structural feature fitting model based on the feature density distribution, and calculate the time evolution coefficient and intensity evolution coefficient of the structural feature fitting model;

[0034] S04. Determine the structural feature evolution trajectory according to the time evolution coefficient and the intensity evolution coefficient, and calculate the correlation degree index of the structural feature evolution trajectory;

[0035] S05. Perform numerical segmentation on the correlation degree index, establish a structural feature piecewise function, and identify the feature mutation points of the structural feature piecewise function;

[0036] S06. Calculate the structural feature contribution value of the feature mutation point, and the structural feature contribution value is used to characterize the evolution intensity of the structural feature at the feature mutation point;

[0037] S07. Establish a feature mapping relationship between the structural feature contribution value and the structural feature evolution trajectory, and generate a feature evolution matrix;

[0038] S08. Calculate the evolution stability index based on the feature evolution matrix, and determine the structural feature change point according to the evolution stability index;

[0039] S09. Perform hierarchical clustering on the structural feature change points to determine the main structural feature change points and secondary structural feature change points;

[0040] S10. Calculate the feature weight value of the main structural feature change point, and construct a feature weight distribution map;

[0041] S11. Identify the key structural feature change points according to the feature weight distribution map, and generate a change point time series table;

[0042] S12. Establish a time series correspondence relationship between the change point time series table and the structural evolution process of the disease biomarker, and output the determination result of the structural feature change point.

[0043] Among them, to obtain the time-series structural feature data of the disease biomarker, specifically, a biomass spectrometry system is used to obtain biomarker sample data, record the mass spectrometry map information of the biomarker, collect the biomarker structure information, and perform time-series integration on the mass spectrometry map information and the biomarker structure information to form the time-series structural feature data of the disease biomarker.

[0044] To preprocess the time-series structural feature data of the disease biomarker to eliminate noise interference, specifically, the original data is decomposed at multiple scales, the high-frequency components are removed, the random fluctuations are eliminated, and data normalization processing is performed.

[0045] Construct a feature transformation image of the time-series structural feature data of the disease biomarker. Specifically, establish a time-intensity coordinate system for the preprocessed time-series data, perform continuous processing on the data points, and generate a time-series grayscale image.

[0046] Calculate the feature density distribution in the feature transformation image. Specifically, identify the local peak points in the image, count the distribution of the number of peak points, calculate the distance between peak points, and form a peak point distribution function.

[0047] Establish a structural feature fitting model based on the feature density distribution. Specifically, select the feature point of the distribution curve, determine the fitting coefficient, adjust the fitting parameter, and obtain the fitting curve.

[0048] Calculate the time evolution coefficient and intensity evolution coefficient of the structural feature fitting model. Specifically, calculate the first-order change rate in the time dimension as the time evolution coefficient, and calculate the second-order change rate in the intensity dimension as the intensity evolution coefficient.

[0049] Determine the structural feature evolution trajectory according to the time evolution coefficient and the intensity evolution coefficient. Specifically, perform direction synthesis on the time evolution coefficient and the intensity evolution coefficient to obtain the evolution direction and form the structural feature evolution trajectory.

[0050] Calculate the correlation degree index of the structural feature evolution trajectory. Specifically, calculate the similarity between trajectory points, count the local correlation degree, and obtain the overall correlation degree index.

[0051] Perform numerical segmentation on the correlation degree index. Specifically, set segmentation values according to the distribution of the correlation degree index, divide the correlation degree index into intervals, and analyze the interval data.

[0052] Establish a structural feature piecewise function. Specifically, perform function fitting on the interval data, perform smoothing processing at the interval connection, and construct a piecewise composite function.

[0053] Identify the feature mutation points of the structural feature piecewise function. Specifically, calculate the derivative of the piecewise function, detect the position of the derivative jump, and determine the position of the feature mutation points.

[0054] Calculate the structural feature contribution value of the feature mutation points. Specifically, analyze the change of feature quantities before and after the feature mutation points, calculate the local gradient, and evaluate the impact of the feature mutation points on the structural evolution.

[0055] Establish the feature mapping relationship between the structural feature contribution value and the structural feature evolution trajectory. Specifically, construct a feature transformation matrix and map the structural feature contribution value to the evolution feature space.

[0056] Generate a feature evolution matrix. Specifically, arrange the feature mapping data in time series and construct a two-dimensional time series feature matrix.

[0057] Calculate the evolutionary stability index based on the feature evolution matrix. Specifically, analyze the temporal correlation of matrix elements, calculate the degree of feature fluctuation, and evaluate the stability of the evolutionary process.

[0058] Determine the structural feature change points according to the evolutionary stability index. Specifically, set a stability threshold, identify the moment of index change, and determine the key points of feature change.

[0059] Perform hierarchical clustering on the structural feature change points to determine the main structural feature change points and the secondary structural feature change points. Specifically, group the change points, calculate the difference between groups, and divide the levels of change points.

[0060] Calculate the feature weight values of the main structural feature change points and construct a feature weight distribution map. Specifically, analyze the central tendency of the weights, select the change points with high weights, and generate a distribution image.

[0061] Identify the key structural feature change points according to the feature weight distribution map and generate a change point time series table. Specifically, analyze the weight distribution, select the change points with higher feature weight values, and generate a list in time series.

[0062] Establish the temporal correspondence relationship between the change point time series table and the structural evolution process of disease markers, and output the determination result of the structural feature change points. Specifically, match the change point time series with the structural evolution stage, establish a temporal relationship, and form an analysis result.

[0063] The following describes the specific implementation manners of the above steps in detail.

[0064] Step S01: Obtain the time series structural feature data of disease markers, and preprocess the time series structural feature data of disease markers to eliminate noise interference. Specifically, in this step, first use a biomass spectrometry system to obtain marker sample data, record the mass spectrometry map information of the markers, collect the structural information of the markers, and perform temporal integration on the mass spectrometry map information and the marker structural information to form the time series structural feature data of disease markers. Then perform multi-scale decomposition on the original data, remove high-frequency components, eliminate random fluctuations, and perform data normalization processing to eliminate noise interference and improve the signal-to-noise ratio of the data.

[0065] Step S02: Construct a feature transformation image of the time series structural feature data of the disease markers, and calculate the feature density distribution in the feature transformation image. Specifically, in this step, first establish a time-intensity coordinate system for the preprocessed time series data, perform continuous processing on the data points to generate a time series grayscale image. Then identify the local peak points in the image, count the distribution of the number of peak points, calculate the distance between peak points, and form a peak point distribution function, thereby obtaining the feature density distribution. The calculation formula for the feature density distribution is: where D(x, y) is the feature density distribution function; P ij is the peak point intensity; d ij is the distance between peak points; (x i , y i ) is the coordinate of the i-th peak point; α, β are attenuation coefficients; n, m are the number of peak points. The purpose of this step is to extract the key information of the temporal structure features of disease markers.

[0066] Step S03, establish a structural feature fitting model based on the feature density distribution, and calculate the time evolution coefficient and intensity evolution coefficient of the structural feature fitting model. Specifically, this step first selects the characteristic points of the distribution curve, determines the fitting coefficients, adjusts the fitting parameters, and obtains the structural feature fitting model where φ k (t) is the basis function, ψ k (t) is the orthogonal basis function, a k , b k are the fitting coefficients, λ is the smoothing parameter, η(t) is the random error term, and K is the number of basis functions. Then calculate the first-order change rate in the time dimension as the time evolution coefficient Calculate the second-order change rate in the intensity dimension as the intensity evolution coefficient where γ is the time scale parameter, μ is the intensity scale parameter, and ξ is the intensity correction term. The purpose of this step is to establish a mathematical model describing the evolution process of the marker structural features.

[0067] Step S04, determine the structural feature evolution trajectory according to the time evolution coefficient and the intensity evolution coefficient, and calculate the correlation degree index of the structural feature evolution trajectory. Specifically, this step first synthesizes the time evolution coefficient and the intensity evolution coefficient in the direction to obtain the evolution direction and form the structural feature evolution trajectory where is the unit vector in the time direction, is the unit vector in the intensity direction, and ω(t) is the trajectory fluctuation term. Then calculate the similarity between trajectory points, statistically analyze the local correlation degree, and obtain the overall correlation degree index where ρ k is the correlation weight, τ is the time decay constant, σ k is the standard deviation, and N is the number of reference points. The purpose of this step is to characterize the temporal evolution process and its internal connection of the marker structural features.

[0068] Step S05: Numerically segment the correlation index, establish a structural feature piecewise function, and identify the characteristic mutation points of the structural feature piecewise function. Specifically, in this step, the segmentation values are first set according to the distribution of the correlation index, the correlation index is divided into intervals, and the interval data is analyzed. Then, the function fitting is performed on the segmented interval data, and smoothing processing is carried out at the interval connection points to construct a piecewise composite function where f i (t) is the interval fitting function, q i (t) is the transition function, u(t) is the unit step function, t i is the segmentation point, ζ(t) is the segmentation error term, and I is the number of segments. Then, calculate the derivative of the piecewise function, detect the derivative jump position, and determine the position of the characteristic mutation point where η is the sensitivity parameter and λ is the weight of the higher-order derivative. The purpose of this step is to identify the key turning points in the evolution process of the marker structure characteristics

[0069] Step S06: Calculate the structural feature contribution value of the characteristic mutation point. The structural feature contribution value is used to characterize the evolution intensity of the structural feature at the characteristic mutation point. Specifically, this step analyzes the changes in the characteristic quantities before and after the characteristic mutation point, calculates the local gradient, evaluates the impact of the characteristic mutation point on the structural evolution, and obtains the structural feature contribution value where V(p) is the contribution value at the characteristic mutation point p, ΔF(p) is the function value mutation amount, is the gradient value, δ is the scale parameter, h j (p) is the local characteristic function, w j is the weight coefficient, and J is the number of characteristic functions. The purpose of this step is to quantify the importance of the marker structure characteristics at the mutation point

[0070] Step S07: Establish the characteristic mapping relationship between the structural feature contribution value and the structural feature evolution trajectory, and generate a characteristic evolution matrix. Specifically, in this step, first construct a characteristic transformation matrix where V i is the characteristic contribution value, x i , y j are the characteristic space coordinates, σ q is the space scale parameter, and γ ij is the transformation correction term. Then, arrange the characteristic mapping data in time sequence to construct a two-dimensional time sequence characteristic matrix, that is, the characteristic evolution matrix. The purpose of this step is to establish the mapping relationship between the marker structure characteristics and the evolution trajectory, providing a data basis for subsequent analysis

[0071] Step S08: Calculate the evolutionary stability index based on the feature evolution matrix, and determine the structural feature change points according to the evolutionary stability index. Specifically, this step first analyzes the temporal correlation of the matrix elements φ(t1,t2), where M k (t) is the time series of the matrix elements, φ(t1,t2) is the correlation correction term, and K is the feature dimension. Then calculate the feature fluctuation degree where M(t) is the matrix element, T is the time window length, and α v is the acceleration weight. Finally, calculate the evolutionary stability index according to the temporal correlation and the feature fluctuation degree where N is the total number of matrix elements, θ is the time scale parameter, and κ is the diffusion coefficient. Set the stability threshold, identify the moment of index change, and determine the structural feature change points. The purpose of this step is to identify the key stages in the evolution process of the marker's structural features through a data-driven method

[0072] Step S09: Perform hierarchical clustering on the structural feature change points to determine the main structural feature change points and the secondary structural feature change points. Specifically, this step first groups the change points and calculates the between-group difference where is the k-th dimensional feature of the i-th point, w k is the feature weight, W(p) is the change point weight, and β is the weight difference coefficient. Then divide the levels of the change points according to the grouping results to determine the main change points and the secondary change points. The purpose of this step is to conduct a more detailed analysis and description of the evolution process of the marker's structural features

[0073] Step S10: Calculate the feature weight values of the main structural feature change points and construct a feature weight distribution map. Specifically, this step analyzes the central tendency of the weights of the main change points, selects the change points with higher feature weight values, and calculates the weight values where g l (p) is the local weight function, c l is the clustering center, α l is the weight coefficient, σ is the spatial scale parameter, β is the contribution value weight, and L is the number of clusters. Then generate a feature weight distribution map to visually display the feature importance of the key change points. The purpose of this step is to identify the key feature change points in the marker's structural evolution

[0074] Step S11: Identify the key structural feature change points based on the feature weight distribution diagram and generate a change point time sequence table. Specifically, this step analyzes the feature weight distribution diagram, selects the change points with higher feature weight values, and generates a change point time sequence table according to the time sequence. The purpose of this step is to present the key structural feature change point information in an intuitive form.

[0075] Step S12: Establish the time sequence correspondence relationship between the change point time sequence table and the structural evolution process of the disease biomarker, and output the determination result of the structural feature change point. Specifically, this step matches the change point time sequence with the biomarker structure evolution stage to establish a time sequence relationship. where s k is the reference time point, τ k is the time scale, f k (t, s) is the local correspondence function, α k is the correspondence weight, η L is the correspondence error term. Finally, an analysis result is formed and the determination information of the biomarker structural feature change point is output. The purpose of this step is to associate the analysis result with the actual structural evolution process of the disease biomarker to provide support for clinical applications.

[0076] The following details the specific calculation process involved in the present invention.

[0077] 1. The calculation of the feature density distribution is expressed as follows:

[0078]

[0079] In the formula, D(x, y) is the feature density distribution function; P ij is the peak point intensity; d ij is the distance between peak points; (x i , y i ) is the coordinate of the i-th peak point; α, β are attenuation coefficients; n, m are the number of peak points.

[0080] 2. The structural feature fitting model is expressed as follows:

[0081]

[0082] In the formula, F(t) is the fitting function; φ k (t) is the basis function; ψ k (t) is the orthogonal basis function; a k , b k are the fitting coefficients; λ is the smoothing parameter; ∶(t) is the random error term; K is the number of basis functions.

[0083] 3. The calculation of the time evolution coefficient is expressed as follows:

[0084]

[0085] In the formula, C t is the time evolution coefficient; γ is the time scale parameter; F(t) is the fitting function.

[0086] 4. The calculation of the intensity evolution coefficient is expressed as follows:

[0087]

[0088] In the formula, C i is the intensity evolution coefficient; μ is the intensity scale parameter; ξ is the intensity correction term.

[0089] 5. The calculation of the structural feature evolution trajectory is expressed as follows:

[0090]

[0091] In the formula, T(t) is the evolution trajectory function; is the unit vector in the time direction; is the unit vector in the intensity direction; ω(t) is the trajectory fluctuation term.

[0092] 6. The calculation of the correlation degree index is expressed as follows:

[0093]

[0094] In the formula, R(t) is the correlation degree index; ρ k is the correlation weight; τ is the time decay constant; σ k is the standard deviation; N is the number of reference points.

[0095] 7. The calculation of the contribution value of the structural feature is expressed as follows:

[0096]

[0097] In the formula, V(p) is the contribution value at the characteristic mutation point p; ΔF(p) is the mutation amount of the function value; is the gradient value; δ is the scale parameter; h j (p) is the local characteristic function; w j is the weight coefficient; J is the number of characteristic functions.

[0098] 8. The characteristic mapping relationship matrix is expressed as follows:

[0099]

[0100] In the formula, m ij is the mapping coefficient; n is the characteristic dimension.

[0101] 9. The calculation of the evolution stability index is expressed as follows:

[0102]

[0103] In the formula, S(t) is the evolution stability index; M k is the k-th element of the characteristic evolution matrix; θ is the time scale parameter; κ is the diffusion coefficient; N is the total number of matrix elements.

[0104] 10. The calculation of the feature weight is expressed as follows:

[0105]

[0106] In the formula, W(p) is the weight value of the change point p; g l (p) is the local weight function; c l is the clustering center; α l is the weight coefficient; σ is the spatial scale parameter; β is the contribution value weight; L is the number of clusters.

[0107] 11. The calculation of the numerical segmentation threshold is expressed as follows:

[0108]

[0109] In the formula, H(r) is the segmentation threshold function; R(t) is the correlation degree index; σ h is the smoothing parameter; ε h is the threshold correction term; r is the reference time point.

[0110] 12. The structural feature segmentation function is expressed as follows:

[0111]

[0112] In the formula, G(t) is the segmentation function; f i (t) is the interval fitting function; q i (t) is the transition function; u(t) is the unit step function; t i is the segmentation point; ζ(t) is the segmentation error term; I is the number of segments.

[0113] 13. The feature mutation point recognition function is expressed as follows:

[0114]

[0115] In the formula, P(t) is the mutation point recognition function; η is the sensitivity parameter; λ is the high-order derivative weight.

[0116] 14. The calculation of the feature transformation matrix is expressed as follows:

[0117]

[0118] In the formula, Q is the feature transformation matrix; q ij is the transformation coefficient; Vi is the characteristic contribution value; x i , y j are the characteristic space coordinates; σ q is the spatial scale parameter; γ ij is the conversion correction term.

[0119] 15. The calculation of temporal correlation is expressed as follows:

[0120]

[0121] In the formula, C(t1, t2) is the temporal correlation function; M k (t) is the time series of matrix elements; φ(t1, t2) is the correlation correction term; K is the characteristic dimension.

[0122] 16. The calculation of the degree of feature fluctuation is expressed as follows:

[0123]

[0124] In the formula, F v (t) is the fluctuation intensity function; M(t) is the matrix element; T is the time window length; α v is the acceleration weight.

[0125] 17. The calculation of the hierarchical clustering distance is expressed as follows:

[0126]

[0127] In the formula, D(p1, p2) is the distance between change points; is the k-th dimensional feature of the i-th point; w k is the feature weight; W(p) is the change point weight; β is the weight difference coefficient.

[0128] 18. The temporal correspondence function is expressed as follows:

[0129]

[0130] In the formula, L(t, s) is the temporal correspondence function; s k is the reference time point; τ k is the time scale; f k (t, s) is the local correspondence function; α k is the correspondence weight; η L is the correspondence error term.

[0131] These formulas are constructed considering the following principles:

[0132] 1. The feature density distribution adopts the Gaussian kernel function, considering the spatial attenuation effect;

[0133] 2. The structural feature fitting model introduces orthogonal basis functions, improving the fitting accuracy and stability;

[0134] 3. The calculation of the evolution coefficient considers the high-order derivative terms, enhancing the sensitivity to mutation features;

[0135] 4. The correlation degree index adopts an exponential decay form, reflecting the natural decay characteristic of time correlation;

[0136] 5. The calculation of the feature contribution value combines local gradient information and global feature functions;

[0137] 6. The time series correlation analysis adopts a normalized form, eliminating the influence of dimension;

[0138] 7. The hierarchical clustering distance combines the feature space distance and weight difference, improving the classification accuracy;

[0139] 8. The time series correspondence function considers multi-scale time features, enhancing the robustness of the correspondence.

[0140] The following is a detailed description of the establishment and derivation process of the feature density distribution function and the structural feature fitting model:

[0141] Derivation process of the feature density distribution function D(x, y):

[0142] The first step is to establish an initial model based on the point source diffusion theory:

[0143]

[0144] The second step is to introduce the Gaussian kernel function considering the spatial decay effect:

[0145]

[0146] The third step is to combine the initial model and the decay function to obtain the final expression:

[0147]

[0148] Parameter acquisition method: P ij The peak intensity is directly measured by a mass spectrometry instrument; d ij Calculated by the Euclidean distance: α, β are optimized by the cross-validation method, and the general value range is 0.01 - 0.1; (x i , y i ) are the peak point coordinates in the mass spectrometry diagram and are directly obtained by the instrument.

[0149] Derivation process of the structural feature fitting model F(t):

[0150] Step 1: Construct a basic polynomial fitting model:

[0151] Step 2: Introduce orthogonal basis functions to improve stability:

[0152] Step 3: Add smoothing terms and error terms to obtain the final model:

[0153] Parameter acquisition method: φ k (t) selects the Legendre polynomial system; ψ k (t) selects the Chebyshev polynomial system; a k , b k is solved by the least squares method; λ is determined by the L-curve method, and the general value range is 0.001 - 0.1; η(t) is estimated by residual analysis and follows the normal distribution N(0, σ 2 ).

[0154] Among them, the time evolution coefficient C t The establishment process first considers the first-order change rate in the time dimension to characterize the instantaneous evolution speed of the system, and further introduces the second-order derivative term to depict the acceleration characteristics of the evolution. The first-order and second-order effects are balanced by the weight coefficient γ, and finally the expression of the time evolution coefficient is obtained Among them, γ is determined by the optimization method, and the general value range is 0.1 - 1.0. This evolution coefficient can simultaneously reflect the evolution speed and acceleration characteristics of the system, providing more comprehensive dynamic information;

[0155] Among them, the intensity evolution coefficient C i The establishment process considers the change characteristics in the intensity dimension and uses the second-order derivative as the main term to capture the curvature characteristics of the intensity change, introduces the first-order derivative term to describe the trend of the intensity change, adds the correction term ξ to compensate for the system error, and obtains the expression of the intensity evolution coefficient Among them, μ is determined by numerical experiments, and the range is 0.2 - 2.0. ξ is obtained through system calibration. This coefficient effectively characterizes the complex characteristics of the intensity change;

[0156] Among them, the establishment process of the structural feature evolution trajectory T(t) is to decompose the contributions of the time evolution coefficient C t and the intensity evolution coefficient C i in their respective directions through the unit vectors and adopts the integral form to accumulate the evolution effect, considers the system perturbation and introduces the fluctuation term ω(t), and finally obtains the evolution trajectory function Among them, ω(t) is described by a Brownian motion model, and this trajectory function comprehensively depicts the dynamic evolution process of the system;

[0157] Among them, the establishment process of the correlation degree index R(t) is to select the key reference points in the system evolution process, and calculate the distance ||T(t) - T(t k )|| between the current state and the reference state, and introduce an exponential term considering the time decay effect Through the standard deviation σ k Perform normalization processing, introduce the weight coefficient ρ k Balance the importance of different reference points to obtain the expression of the correlation degree index Among them, τ is determined by autocorrelation analysis, and ρ k is calculated by the entropy weight method, and this index effectively quantifies the correlation strength of the system state;

[0158] Among them, the establishment process of the structural feature contribution value V(p) is to analyze the function value jump ΔF(p) at the mutation point, and combine the local gradient information to evaluate the mutation intensity, amplify the gradient effect through the exponential function introduce the local feature function h j (p) to describe the characteristic attributes of the mutation point, and use the weight coefficient w j to balance the importance of each feature to obtain the contribution value expression Among them, δ is determined by cross-validation, and h j (p) is constructed based on wavelet analysis, and this contribution value accurately reflects the importance degree of the mutation point;

[0159] Among them, the establishment process of the feature mapping relationship matrix M is to construct an n-dimensional feature space, calculate the mapping coefficient m between different dimensions in the feature space ij , and consider the interaction and transformation relationship between features to form a complete mapping matrix Among them, the mapping coefficient m ij is determined by the principal component analysis and canonical correlation analysis methods, and this matrix realizes the complete mapping of the feature space;

[0160] Among them, the establishment process of the evolution stability index S(t) is to analyze the time derivative k of the feature evolution matrix M to characterize the change rate, introduce a time decay term consider the historical influence, evaluate the spatial stability through the Laplace operator and use the coefficient κ to balance the contribution of the change rate and stability to obtain the stability index expression Among them, θ is determined through time series analysis, and κ is obtained through numerical optimization. This index comprehensively evaluates the evolutionary stability of the system;

[0161] Among them, the process of establishing the feature weight W(p) is to describe the importance of the change point in different clusters through the local weight function g l (p), and an exponential decay term is introduced considering the spatial distance effect Combined with the influence of the feature contribution value V(p), the weight coefficients α l and β are used to balance each contribution, and the weight calculation expression is obtained Among them, g l (p) is constructed through kernel density estimation, and c l is determined through cluster analysis. This weight calculation method effectively quantifies the importance of the change point.

[0162] Among them, the process of establishing the numerical piecewise threshold function H(r) is to first consider the distribution characteristics of the correlation degree index R(t), and through the Gaussian kernel function smooth the correlation degree to eliminate the influence of local fluctuations, and use the Fourier transform form to achieve frequency domain analysis, and introduce the threshold correction term ε h to compensate for the system error, and the threshold function expression is obtained Among them, σ h is optimized and determined through cross-validation, and ε h is estimated through error analysis. This function can adaptively determine the optimal threshold for data segmentation;

[0163] Among them, the process of establishing the structural feature piecewise function G(t) is to establish a local fitting function f i (t) for each segmentation interval, design a transition function q i (t) to ensure the continuity at the segmentation point, use the unit step function u(t) to achieve segmentation switching, and introduce the segmentation error term ζ(t) to compensate for the fitting error, and the piecewise function expression is obtained Among them, f i (t) is obtained by polynomial fitting, q i (t) is constructed through Hermite interpolation, and ζ(t) is determined through residual analysis. This function realizes the piecewise fitting and smooth transition of the data;

[0164] Among them, the process of establishing the feature mutation point recognition function P(t) is to calculate the second derivative of the piecewise function to characterize the curvature change, amplify the first derivative effect through the exponential function to highlight the change rate, introduce the third derivative term to evaluate the acceleration change, and use the weight coefficient λ to balance the contributions of each order derivative, and the recognition function expression is obtained Among them, η is determined through sensitivity analysis, and λ is solved through an optimization algorithm. This function can effectively identify the mutation characteristics of the system;

[0165] Among them, the process of establishing the feature transformation matrix Q is to construct the matrix element q ij , combined with the feature contribution value V i and the spatial distance effect introduce the conversion correction term γ ij , to form the calculation expression of the conversion coefficient Organize all conversion coefficients into a matrix form Among them, σ q is determined by minimizing the conversion error, and γ ij is obtained through system calibration. This matrix realizes the effective transformation of the feature space;

[0166] Among them, the process of establishing the time series correlation function C(t1,t2) is to adopt the inner product form of the matrix elements calculate the correlation, and through normalization eliminate the influence of dimension, introduce the relevant correction term φ(t1,t2) to compensate for the nonlinear effect, and obtain the correlation function expression Among them, φ(t1,t2) is determined by nonlinear regression. This function accurately describes the correlation characteristics of time series data;

[0167] Among them, the feature fluctuation degree function F v (t) is established by calculating the root mean square velocity within the time window to characterize the fluctuation intensity, introduce the second derivative term evaluate the influence of acceleration, and through the weight coefficient α v balance the contributions of velocity and acceleration, and obtain the fluctuation function expression Among them, T is determined through time series analysis, and α v is obtained through numerical optimization. This function comprehensively evaluates the fluctuation characteristics of the system;

[0168] Among them, the process of establishing the hierarchical clustering distance function D(p1,p2) is to calculate the weighted Euclidean distance in the feature space Combined with the weight difference term β·|W(p1)-W(p2)|, consider the importance difference of the change points, and obtain the distance function expression Among them, w k is determined through feature importance analysis, and β is obtained through clustering effect optimization. This function provides a reasonable distance metric for hierarchical clustering;

[0169] Among them, the process of establishing the time series correspondence function L(t,s) is to design the local correspondence function f k (t,s) to describe the mapping relationship between time points, through the exponential decay term Considering the time-scale effect, introduce the corresponding weight α k Balance the contributions of different scales and add the corresponding error term η L Compensate for the mapping error to obtain the corresponding function expression f k (t, s) + η L , where τ k Determined by multi-scale analysis, f k (t, s) is constructed by dynamic time warping, and this function realizes the exact correspondence of time series data.

[0170] The traditional methods for determining the change points of the structural evolution characteristics of disease markers mainly include the following technical routes:

[0171] One is the method based on statistical analysis. This method calculates statistical features such as the mean and variance of the marker structure data and sets fixed statistical thresholds to judge the change points. The specific approach is to first segment the time series data, calculate the statistics of each segment, and when the statistics exceed the preset threshold, mark this time point as a change point. This method is simple to implement but difficult to handle the non-stationarity and sudden changes of the data.

[0172] Another is the method based on spectral analysis. This method transforms the time series data into the frequency domain space and identifies the change points by analyzing the spectral features. The specific approach is to use Fourier transform or wavelet transform to obtain the frequency composition of the data and locate the change points according to the change of spectral energy. This method can provide multi-scale analysis ability but has limitations in dealing with non-linear evolution processes.

[0173] There is also a method based on pattern recognition. This method constructs feature templates and uses pattern matching to identify the change points. The specific approach is to pre-define several typical change patterns and determine the change points by calculating the similarity between the data to be measured and the templates. This method relies on prior knowledge and is difficult to adapt to complex and changeable actual situations.

[0174] In contrast, the technical effects of the method of the present invention are reflected in the following aspects:

[0175] In the feature extraction step, the present invention adopts the method of combining two-dimensional transformed images with density distribution analysis, which can not only capture the instantaneous features of the marker structure but also reflect the overall characteristics in its evolution process. By establishing a polynomial fitting model, the accurate characterization of complex non-linear evolution processes is realized. This method overcomes the problem of insufficient processing ability of traditional statistical analysis methods for non-stationary data.

[0176] In the change point recognition stage, the present invention introduces a dual evaluation mechanism of time evolution coefficient and intensity evolution coefficient. Through the construction of the structural feature evolution trajectory, the multi-dimensional characterization of change features is realized. This method breaks through the limitation of traditional spectrum analysis methods that only focus on single-dimensional features.

[0177] In the result evaluation stage, the present invention establishes a mapping relationship between the feature contribution value and the evolution feature, uses the feature evolution matrix to describe the overall evolution characteristics of the system, and conducts quantitative evaluation through the evolution stability index. This method overcomes the defect of traditional pattern recognition methods that overly rely on empirical knowledge.

[0178] In practical applications, the method of the present invention exhibits the following advantages: First, the method has strong adaptability and does not require a large number of empirical parameters to be set artificially. It can automatically adjust the analysis strategy according to the data characteristics. Second, the method has good anti-interference ability. Through the multi-level feature extraction and analysis mechanism, the influence of noise on the results is effectively reduced. Third, the calculation process of the method has a clear physical meaning, and each step can trace its biological interpretation, improving the reliability and interpretability of the results.

[0179] In addition, the method of the present invention also realizes the hierarchical recognition of change points. Through the division of major change points and minor change points, more detailed information is provided for the in-depth study of the structural evolution characteristics of disease markers. This hierarchical recognition mechanism is difficult to achieve in traditional methods.

[0180] Generally speaking, through innovative theoretical design and algorithm implementation, the method of the present invention is significantly superior to traditional methods in terms of data processing ability, feature extraction accuracy, reliability of change point recognition, etc., providing more powerful technical support for the research on the structural evolution characteristics of disease markers.

[0181] Next, a specific Embodiment 1 of the present invention is provided. The specific implementation manners of each step in this Embodiment 1 are described in detail as follows: Step S01, obtain the time-series structural feature data of disease markers, and preprocess the time-series structural feature data of disease markers to eliminate noise interference. This step first uses a bio-mass spectrometry system to obtain marker sample data, record the mass spectrometry map information of the markers, collect the marker structure information, and perform time-series integration on the mass spectrometry map information and the marker structure information to form the time-series structural feature data of disease markers. Then, perform multi-scale decomposition on the original data, remove the high-frequency components, eliminate random fluctuations, and perform data normalization processing to eliminate noise interference and improve the signal-to-noise ratio of the data. The purpose of this step is to obtain and preprocess the original structural evolution feature data of disease markers, laying a foundation for subsequent feature analysis.

[0182] Step S02: Construct a feature transformation image of the time-series structural feature data of the disease marker, and calculate the feature density distribution in the feature transformation image. In this step, a time-intensity coordinate system is first established for the preprocessed time-series data, the data points are made continuous to generate a time-series grayscale image. Then, the local peak points in the image are identified, the number distribution of the peak points is statistically analyzed, the distances between the peak points are calculated, and a peak point distribution function is formed, thereby obtaining a feature density distribution function. where D(x, y) is the feature density distribution function, P ij is the intensity of the peak point, d ij is the distance between the peak points, (x i , y i ) is the coordinate of the i-th peak point, α, β are attenuation coefficients, and n, m are the numbers of peak points. The purpose of this step is to extract the key information of the time-series structural features of the disease marker and lay a foundation for subsequent model construction and analysis.

[0183] Step S03: Based on the feature density distribution, establish a structural feature fitting model, and calculate the time evolution coefficient and intensity evolution coefficient of the structural feature fitting model. In this step, first, the characteristic points of the distribution curve are selected, the fitting coefficients are determined, and the fitting parameters are adjusted to obtain a structural feature fitting model. where φ k (t) is the basis function, ψ k (t) is the orthogonal basis function, a k , b k are the fitting coefficients, λ is the smoothing parameter, η(t) is the random error term, and K is the number of basis functions. Then, the first-order change rate is calculated in the time dimension as the time evolution coefficient. where γ is the time scale parameter; the second-order change rate is calculated in the intensity dimension as the intensity evolution coefficient. where μ is the intensity scale parameter and ξ is the intensity correction term. The purpose of this step is to establish a mathematical model describing the evolution process of the marker structural features and provide a theoretical basis for subsequent feature analysis.

[0184] Step S04: Determine the structural feature evolution trajectory according to the time evolution coefficient and the intensity evolution coefficient, and calculate the correlation degree index of the structural feature evolution trajectory. In this step, first, the time evolution coefficient and the intensity evolution coefficient are combined in direction to obtain an evolution direction, forming a structural feature evolution trajectory. where is the unit vector in the time direction, is the unit vector in the intensity direction, and ω(t) is the trajectory fluctuation term. Then, the similarity between the trajectory points is calculated, the local correlation degree is statistically analyzed, and the overall correlation degree index is obtained. where ρ k is the correlation weight, τ is the time decay constant, σk where σ is the standard deviation and N is the number of reference points. The purpose of this step is to characterize the temporal evolution process and internal relationships of the marker structure features, providing basic data for subsequent change point identification.

[0185] Step S05: Numerically segment the correlation degree index, establish a piecewise function of the structure features, and identify the characteristic mutation points of the piecewise function of the structure features. This step first sets the segmentation values according to the distribution of the correlation degree index, divides the correlation degree index into intervals, and analyzes the interval data. Then, function fitting is performed on the segmented interval data, and smoothing processing is carried out at the interval connection points to construct a piecewise composite function where f i (t) is the interval fitting function, q i (t) is the transition function, u(t) is the unit step function, t i is the segmentation point, ζ(t) is the segmentation error term, and I is the number of segments. Then, the derivative of the piecewise function is calculated, the position of the derivative jump is detected, and the position of the characteristic mutation point is determined where η is the sensitivity parameter and λ is the weight of the high-order derivative. The purpose of this step is to identify the key turning points in the evolution process of the marker structure features, providing a basis for subsequent feature contribution analysis.

[0186] Step S06: Calculate the contribution value of the structure features at the characteristic mutation points, where the contribution value of the structure features is used to characterize the evolution intensity of the structure features at the characteristic mutation points. This step analyzes the changes in the feature quantities before and after the characteristic mutation points, calculates the local gradient, evaluates the impact of the characteristic mutation points on the structure evolution, and obtains the contribution value of the structure features where V(p) is the contribution value at the characteristic mutation point p, ΔF(p) is the sudden change amount of the function value, is the gradient value, δ is the scale parameter, h j (p) is the local feature function, w j is the weight coefficient, and J is the number of feature functions. The purpose of this step is to quantify the importance of the marker structure features at the mutation points, providing a basis for subsequent feature screening.

[0187] Step S07: Establish the feature mapping relationship between the contribution value of the structure features and the evolution trajectory of the structure features, and generate a feature evolution matrix. This step first constructs a feature transformation matrix where V i is the feature contribution value, x i , y j are the feature space coordinates, σ q is the space scale parameter, γ ijFor conversion correction terms. Then arrange the feature mapping data in time series to construct a two-dimensional time series feature matrix, that is, the feature evolution matrix. The purpose of this step is to establish the mapping relationship between the marker structure features and the evolution trajectory, and provide a data basis for subsequent evolution stability analysis.

[0188] Step S08, calculate the evolution stability index based on the feature evolution matrix, and determine the structural feature change points according to the evolution stability index. This step first analyzes the time series correlation of the matrix elements where M k (t) is the time series of matrix elements, φ(t1,t2) is the correlation correction term, and K is the feature dimension. Then calculate the feature fluctuation degree where M(t) is the matrix element, T is the time window length, and α v is the acceleration weight. Finally, calculate the evolution stability index according to the time series correlation and the feature fluctuation degree where N is the total number of matrix elements, θ is the time scale parameter, and κ is the diffusion coefficient. Set the stability threshold, identify the moment of index change, and determine the structural feature change points. The purpose of this step is to identify the key stages in the evolution process of the marker structure features through a data-driven method, and provide a basis for subsequent classification of change points.

[0189] Step S09, perform hierarchical clustering on the structural feature change points to determine the main structural feature change points and the secondary structural feature change points. This step first groups the change points and calculates the inter-group difference where is the k-dimensional feature of the i-th point, w k is the feature weight, W(p) is the change point weight, and β is the weight difference coefficient. Then divide the levels of the change points according to the grouping results, and determine the main change points and the secondary change points. The purpose of this step is to conduct a more detailed analysis and description of the evolution process of the marker structure features, and lay a foundation for subsequent extraction of key features.

[0190] Step S10, calculate the feature weight values of the main structural feature change points, and construct a feature weight distribution map. This step analyzes the concentration trend of the weights of the main change points, selects the change points with higher feature weight values, and calculates the weight values where g l (p) is the local weight function, c l is the clustering center, and α lis the weight coefficient, σ is the spatial scale parameter, β is the contribution value weight, and L is the number of clusters. Then, a feature weight distribution map is generated to visually display the feature importance of key change points. The purpose of this step is to identify the key feature change points in the structural evolution of the biomarker and provide support for the subsequent result output.

[0191] Step S11: Identify key structural feature change points based on the feature weight distribution map and generate a change point time series table. This step analyzes the feature weight distribution map, selects the change points with higher feature weight values, and generates a change point time series table according to the time sequence. The purpose of this step is to present the information of key structural feature change points in an intuitive form and provide data support for clinical applications.

[0192] Step S12: Establish a time sequence correspondence relationship between the change point time series table and the structural evolution process of the disease biomarker, and output the determination result of the structural feature change point. This step matches the change point time sequence with the biomarker structural evolution stage to establish a time sequence relationship. where s k is the reference time point, τ k is the time scale, f k (t, s) is the local correspondence function, α k is the correspondence weight, and η L is the correspondence error term. Finally, an analysis result is formed and the determination information of the biomarker structural feature change point is output. The purpose of this step is to associate the analysis result with the actual structural evolution process of the disease biomarker and provide support for clinical applications.

[0193] The following provides Example 2 of a specific application scenario of the present invention:

[0194] A biomedical research institution hopes to use the disease biomarker structural evolution feature analysis technology to provide a basis for the early diagnosis of a certain rare disease. The institution extracts a key biomarker from clinical samples and systematically tests its time sequence structural features using a bio-mass spectrometry system. The following is the specific process of the institution using the method of the present invention for biomarker structural evolution feature analysis.

[0195] First, the researchers obtained the time sequence structural feature data of the biomarker. Specifically, they continuously collected samples of 20 healthy subjects and 30 patients with this disease using a bio-mass spectrometry system, recorded the mass spectrometry map information and structural information of the biomarker, and integrated these data according to the time sequence to form a biomarker time sequence structural feature database. Table 1 shows the sample situation of some biomarker structural feature data.

[0196] Table 1 Sample of biomarker structural feature data

[0197]

[0198] Among them, represents the structural feature value of the j-th type of sample (1 for healthy and 2 for patient) at the i-th time point.

[0199] For the obtained original data, the researchers first performed preprocessing with the aim of eliminating noise interference. The specific approach is as follows:

[0200] 1) Perform multi-scale decomposition on the original data and decompose it into components of different frequency bands through wavelet transform.

[0201] 2) Suppress the high-frequency components to remove the noise influence caused by random fluctuations.

[0202] 3) Standardize the preprocessed data to eliminate the influence of dimension.

[0203] After the above preprocessing, the researchers obtained clean time-series structural feature data of the biomarker, laying a foundation for subsequent feature analysis.

[0204] Next, the researchers constructed a two-dimensional transformed image of the time-series structural features of the biomarker. The specific approach is as follows:

[0205] 1) Establish a time-intensity coordinate system for the preprocessed time-series data, with the x-axis representing time and the y-axis representing feature intensity.

[0206] 2) Continuously process the data points to generate a grayscale image. Figure 1 The transformed image of the time-series structural features of this biomarker is given.

[0207] 3) Analyze this image and find that there are a large number of local peak points, indicating that there are multiple dense regions in the time evolution process of the biomarker structure.

[0208] 4) Statistically analyze the quantity distribution and spatial distance of these peak points, and calculate the feature density distribution function D(x, y). According to the formula:

[0209]

[0210] where P ij is the intensity of the i-th peak point, d ij is the distance between the i-th and j-th peak points, (x i , y i ) is the coordinate of the i-th peak point, α, β are attenuation coefficients, and n, m are the number of peak points. The researchers calculated D(x, y) = 0.214.

[0211] Such as Figure 2The figure shows the changing trends of the temporal structural characteristics of healthy samples and patient samples. The horizontal axis represents time t (unit: minute), and the vertical axis represents the feature intensity F(t) (unit: μm^-1). The blue solid line represents the changing trend of healthy samples, and the red dashed line represents the changing trend of patient samples. It can be seen from the figure that the two groups of samples show different changing patterns during the time evolution process.

[0212] Next, the researchers established a structural feature fitting model based on the feature density distribution. The specific steps are as follows:

[0213] 1) Select Figure 1 several feature-dense regions in k as key feature points, such as (t1, F1), (t2, F2),..., (t k ), F

[0214] 2) Use a polynomial function for fitting to obtain the structural feature fitting model F(t):

[0215]

[0216] where φ k (t) is the basis function, ψ k (t) is the orthogonal basis function, a k , b k are the fitting coefficients, λ is the smoothing parameter, η(t) is the random error term, and K is the number of basis functions. By adjusting the fitting parameters, the researchers finally obtained t(t) = 0.372t 2 + 0.108t + 0.894.

[0217] 3) Calculate the first-order change rate in the time dimension as the time evolution coefficient C t :

[0218]

[0219] where γ is the time scale parameter, and the researchers took γ = 0.2.

[0220] 4) Calculate the second-order change rate in the intensity dimension as the intensity evolution coefficient C i :

[0221]

[0222] where μ is the intensity scale parameter, and the researchers took μ = 0.3, ξ = 0.05.

[0223] Through the above steps, the researchers obtained the time evolution coefficient and intensity evolution coefficient of the biomarker structural characteristics, laying a foundation for the subsequent construction of the evolution trajectory.

[0224] Based on the time evolution coefficient and the intensity evolution coefficient, the researchers began to determine the evolution trajectory of the marker structural characteristics. The specific approach is as follows:

[0225] 1) Synthesize the directions of the time evolution coefficient and the intensity evolution coefficient to obtain the evolution direction vector

[0226] 2) Integrate and calculate the evolution trajectory function T(t) according to the formula:

[0227]

[0228] where ω(t) is the trajectory fluctuation term, and the researchers take ω(t) = 0.01t.

[0229] As Figure 3 shown, it presents the evolution trajectory of the marker structural characteristics in the space coordinate system. The horizontal axis and the vertical axis represent the space coordinates x and y (unit: μm) respectively. The blue curve represents the evolution trajectory, and the red scatter points mark the positions of two key turning points (t = 3.6 and t = 7.2). It can be clearly seen from the figure that significant changes in the structural characteristics occurred at these two time points.

[0230] 3) Calculate the similarity between points on the evolution trajectory, statistically analyze the local correlation degree, and obtain the overall correlation degree index R(t):

[0231]

[0232] where ρ k = 2.14 is the correlation weight, τ = 0.5 is the time decay constant, σ k = 0.3 is the standard deviation, and N = 20 is the number of reference points.

[0233] After the above calculations, the researchers obtained the evolution trajectory of the marker structural characteristics and its correlation degree index. Next, they carried out a numerical piecewise analysis on the correlation degree index to identify the key turning points in the marker structure evolution process.

[0234] 1) According to the distribution characteristics of the correlation degree index, set the piecewise threshold H(r) = 1.78. Divide the correlation degree index into a high - correlation interval and a low - correlation interval.

[0235] 2) For the high - correlation interval, use a smooth polynomial function f i (t) for fitting; for the low - correlation interval, use a transition function q i (t) for connection.

[0236] 3) Construct the piecewise composite function G(t):

[0237]

[0238] Among them, u(t) is the unit step function, and t i is the segmentation point, ζ(t) is the segmentation error term, and I = 3 is the number of segments.

[0239] 4) Calculate the second derivative and the third derivative of the piecewise function G(t), and it is found that obvious jumps occur at t = 3.6 and t = 7.2, which are identified as the key turning points of the marker structure characteristics.

[0240] Through the above steps, the researchers successfully identified two key change points in the marker structure evolution process. To further quantify the importance of these change points, they calculated the contribution values of the structural characteristics of these points:

[0241]

[0242] Among them, h1(p) and h2(p) are local characteristic functions, and δ = 0.5 is the scale parameter. It can be seen that the change at the moment of 3.6 has a greater impact on the marker structure evolution.

[0243] The researchers further established the mapping relationship between the contribution value of the structural characteristics and the evolution trajectory, and constructed the characteristic evolution matrix M:

[0244]

[0245] Among them, are the coordinates of the change point and the evolution trajectory in the feature space respectively.

[0246] Based on the characteristic evolution matrix, the researchers calculated the stability index S(t) of the evolution process:

[0247]

[0248] Among them, M k is the k-th element of the matrix, θ = 0.3 is the time scale parameter, κ = 0.1 is the diffusion coefficient, and N = 80 is the total number of matrix elements.

[0249] As Figure 4 shown, it shows the change of the stability index S(t) of the marker structure characteristics over time. The horizontal axis represents time t (unit: minute), and the vertical axis represents the stability index S(t). The red dashed line marks the main change point (t = 3.6), and the green dashed line marks the secondary change point (t = 7.2). It can be observed from the figure that the stability index decreases significantly at these two time points.

[0250] The researchers found that S(t) decreased significantly near t = 3.6 and t = 7.2, indicating that the structural characteristics of the biomarker changed greatly at these two time points. Combining with the feature contribution values calculated previously, they determined that these two time points were the key change points in the structural evolution of the biomarker.

[0251] Finally, the researchers conducted further analysis on these two main change points. First, they calculated the feature weight values of these two change points:

[0252]

[0253] where g1(p), g2(p) are local weight functions, c1, c2 are cluster centers, and σ = 0.2 is the spatial scale parameter. It can be seen that the change point at time 3.6 has a higher feature weight.

[0254] According to the feature weight distribution, the researchers identified that the change point at time 3.6 was the key turning point in the structural evolution of the biomarker. They organized this result in the form of a time sequence table as follows:

[0255] Table 2 Time Sequence Table of Change Points in Biomarker Structural Characteristics

[0256] Change point time Feature weight value Degree of change t＝3.6 2.41 Main change point t＝7.2 2.07 Minor change point

[0257] Finally, the researchers conducted a corresponding analysis of the time sequence table of change points and the process of biomarker structural evolution, forming a complete analysis result. Specifically, as shown in Table 3.

[0258] Table 3 Analysis Results of Biomarker Structural Evolution

[0259] Time period Structural feature change situation t<3.6 The structure of the marker changes slowly and tends to be stable as a whole 3.6≤t<7.2 The structure of the marker changes violently and a major turning point appears t≥7.2 The structure of the marker enters a new evolutionary stage and minor changes occur

[0260] In summary, the method of the present invention successfully identified two key change points in the structural evolution of the biomarker, where the change at time 3.6 was the main turning point, having a greater impact on the overall evolution process. This result provides an important basis for subsequent disease diagnosis.

[0261] Technical Principle: The core principle of the method of the present invention is to construct a mathematical model that can comprehensively describe the structural evolution characteristics of disease biomarkers, and identify key change points based on this model. Specifically, this method is achieved through the following steps:

[0262] First, the instantaneous features of the biomarker structure are extracted by means of two-dimensional transformed images, and the feature density distribution is calculated. This method can not only capture the local features of the biomarker structure but also reflect the law of its overall evolution. Compared with traditional one-dimensional time series data analysis, two-dimensional transformation can provide more abundant information.

[0263] Secondly, a structural feature fitting model is established, and the time evolution coefficient and the intensity evolution coefficient are calculated simultaneously. These two coefficients respectively characterize the change trends of the marker structure in the time and intensity dimensions, providing a basis for constructing the subsequent evolution trajectory. Compared with single statistics or spectral features, this multi-dimensional description ability is more comprehensive.

[0264] Then, based on the time evolution coefficient and the intensity evolution coefficient, a structural feature evolution trajectory is constructed, and the correlation degree index of the trajectory is calculated. The trajectory describes the overall trend of the marker structure in the spatio-temporal evolution process, and the correlation degree index reflects the consistency of the evolution process. This lays an important foundation for the subsequent identification of change points.

[0265] Next, by numerically segmenting the correlation degree index, a structural feature piecewise function is established, and the characteristic mutation points of the function are identified. The mutation points represent the key turning points in the evolution process of the marker structure and are the direct basis for change point identification. This piecewise analysis can better capture the key features in the non-linear evolution process.

[0266] Furthermore, the contribution value of the structural features of the characteristic mutation points is calculated, and the mapping relationship between the contribution value of the features and the evolution trajectory is established. This step quantifies the influence degree of the mutation points on the overall evolution and provides a basis for the subsequent change point evaluation. Compared with the traditional pattern matching method, this feature contribution-based method is more in line with the physical mechanism.

[0267] Finally, based on the feature evolution matrix, the stability of the evolution process is analyzed, and the key change points are determined in combination with the feature weight distribution. The stability index can reflect the smoothness of the marker structure evolution, while the feature weight distribution reveals the importance of the key feature changes. Through the comprehensive analysis of these two dimensions, the key stages of the marker structure evolution can be accurately identified.

[0268] It should be noted that the detailed explanations of the variables involved in the present invention are shown in Table 4 below.

[0269] Table 4 Variable Explanation Table

[0270]

[0271]

[0272] The above is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, and all should be covered by the protection scope of the present invention.

Claims

1. A method for determining the change points of the structural evolution characteristics of a disease biomarker, characterized in that Including obtaining time-series structural feature data of disease markers and preprocessing them, constructing feature transformation images and calculating feature density distributions, establishing structural feature fitting models and calculating time evolution coefficients and intensity evolution coefficients, determining structural feature evolution trajectories and calculating correlation degree indices, establishing structural feature piecewise functions and identifying feature mutation points, calculating structural feature contribution values and generating feature evolution matrices, calculating evolution stability indices and determining structural feature change points, performing hierarchical clustering and determining main and secondary structural feature change points, constructing feature weight distribution maps and generating change point time series tables, establishing time series correspondence relationships and outputting the results of determining structural feature change points.

2. The method for determining the change points of the structural evolution characteristics of a disease biomarker according to claim 1, wherein, Obtaining time-series structural feature data of disease markers, specifically, using a biomass spectrometry system to obtain marker sample data, recording marker mass spectrometry map information, collecting marker structural information, and performing time series integration on the mass spectrometry map information and the marker structural information to form time-series structural feature data of disease markers.

3. The method for determining the change point of the structural evolution characteristics of the disease biomarker according to claim 1, wherein Constructing a feature transformation image of the time-series structural feature data of the disease marker, specifically, establishing a time-intensity coordinate system for the preprocessed time series data, performing continuous processing on the data points to generate a time series grayscale image; calculating the feature density distribution in the feature transformation image, specifically, identifying local peak points in the image, statistically analyzing the number distribution of peak points, calculating the distances between peak points, and forming a peak point distribution function.

4. The method for determining the change points of the structural evolution characteristics of a disease biomarker according to claim 1, wherein Establishing a structural feature fitting model based on the feature density distribution, specifically, selecting distribution curve feature points, determining fitting coefficients, adjusting fitting parameters to obtain a fitting curve; calculating the time evolution coefficient and intensity evolution coefficient of the structural feature fitting model, specifically, calculating the first-order change rate in the time dimension as the time evolution coefficient and the second-order change rate in the intensity dimension as the intensity evolution coefficient.

5. The method for determining the change points of the structural evolution characteristics of a disease biomarker according to claim 1, wherein Determining the structural feature evolution trajectory according to the time evolution coefficient and the intensity evolution coefficient, specifically, performing direction synthesis on the time evolution coefficient and the intensity evolution coefficient to obtain an evolution direction and form a structural feature evolution trajectory; calculating the correlation degree index of the structural feature evolution trajectory, specifically, calculating the similarity between trajectory points, statistically analyzing the local correlation degree, and obtaining the overall correlation degree index.

6. The method for determining the change points of the structural evolution characteristics of a disease biomarker according to claim 1, wherein Performing numerical segmentation on the correlation degree index, establishing a structural feature piecewise function, and identifying the feature mutation points of the structural feature piecewise function; calculating the structural feature contribution value of the feature mutation point, and the structural feature contribution value is used to characterize the evolution intensity of the structural feature at the feature mutation point.

7. The method for determining the change points of the structural evolution characteristics of a disease biomarker according to claim 1, wherein Establishing a feature mapping relationship between the structural feature contribution value and the structural feature evolution trajectory, specifically, constructing a feature transformation matrix to map the structural feature contribution value to the evolution feature space; generating a feature evolution matrix, specifically, arranging the feature mapping data in time series and constructing a two-dimensional time series feature matrix.

8. The method for determining the change points of the structural evolution characteristics of a disease biomarker according to claim 1, wherein Calculating an evolution stability index based on the feature evolution matrix, specifically, analyzing the time series correlation of matrix elements, calculating the feature fluctuation degree, and evaluating the stability of the evolution process; determining the structural feature change point according to the evolution stability index, specifically, setting a stability threshold, identifying the moment of index change, and determining the key feature change points.

9. The method for determining the change points of the structural evolution characteristics of a disease biomarker according to claim 1, wherein Perform hierarchical clustering on the change points of the structural features to determine the main change points of the structural features and the secondary change points of the structural features; calculate the feature weight values of the main change points of the structural features and construct a feature weight distribution map.

10. The method for determining the change point of the structural evolution characteristics of a disease biomarker according to claim 1, wherein Identify the key change points of the structural features according to the feature weight distribution map, generate a change point time series table; establish the time series correspondence between the change point time series table and the structural evolution process of the disease biomarker, and output the determination result of the change points of the structural features.

Images