Polycystic ovarian syndrome detection method based on multi-dimensional biomarker analysis

Abstract

The invention relates to the technical field of medical care, in particular to a polycystic ovarian syndrome detection method based on multi-dimensional biomarker analysis, which comprises the following steps: S1, patient data acquisition: acquiring multi-modal health data of a patient; s2, integrating and standardizing the biomarker data: integrating the health data of different dimensions; s3, performing pathogenesis inference based on a hormone and metabolism coupling model: performing coupling analysis on the hormone level and metabolism data of the patient, and revealing a mutual influence mechanism of the hormone level and the metabolism data in the polycystic ovarian syndrome; s4, risk score calculation: calculating the risk score of the polycystic ovarian syndrome of the patient; and S5, diagnosis feedback and dynamic optimization: dynamically adjusting the detection scheme. According to the method, the accuracy and the flexibility of the diagnosis and treatment scheme are improved, all-around data support is provided for doctors, and personalized and dynamic long-term health management and optimization of the treatment scheme are assisted.

Description

Technical Field

[0001] The present invention relates to the field of medical care technology, and in particular to a polycystic ovary syndrome detection method based on multidimensional biomarker analysis. Background Art

[0002] Polycystic ovary syndrome (PCOS) is a common endocrine disorder, mainly manifested by abnormal menstruation, excessive androgen and polycystic ovary lesions, which seriously affect women's reproductive health. In recent years, with the rapid development of biomedical technology, more and more biomarkers have been used to assist in the diagnosis of PCOS, such as hormone levels, metabolic indicators, immune markers and genetic testing. However, due to the complex etiology of PCOS, which involves multiple systems such as endocrine, metabolic, and immune systems, a single-dimensional detection method is difficult to fully reflect the complex etiology of the disease. How to effectively integrate multidimensional biomarkers and deeply analyze their association with PCOS has become a key issue in diagnosis and treatment.

[0003] In the existing technology, the diagnosis method of PCOS mainly relies on the patient's hormone level detection and ultrasound image analysis, lacking comprehensive analysis at the metabolic, immune and genetic levels. This single-dimensional analysis method easily ignores the important role of other systems in PCOS, resulting in the accuracy of the diagnostic results being affected. In addition, the existing technology usually lacks effective integration and standardized processing of multimodal data, and cannot fully utilize the patient's multidimensional health data. Due to the complex coupling relationship between biomarkers, the existing diagnostic technology has limitations in capturing these complex interactions, and cannot provide doctors with a comprehensive analysis of the etiology mechanism, which also limits the formulation of personalized treatment plans.

[0004] In view of the shortcomings of the existing technology, the present invention proposes a polycystic ovary syndrome detection method based on multidimensional biomarker analysis, which can improve the diagnostic accuracy and personalized treatment effect of PCOS, provide doctors with more accurate clinical decision-making basis, and promote the advancement of PCOS diagnostic technology. Summary of the invention

[0005] The present invention provides a method for detecting polycystic ovary syndrome based on multidimensional biomarker analysis.

[0006] The method for detecting polycystic ovary syndrome based on multidimensional biomarker analysis comprises the following steps:

[0007] S1, patient data collection: collect multimodal health data of patients, including imaging data, genetic test data, hormone levels, metabolic indicators, immune markers and biomarkers in physiological fluids (such as saliva, sweat, etc.);

[0008] S2, biomarker data integration and standardization: standardize the collected multimodal health data and integrate health data of different dimensions through the biomarker database;

[0009] S3, Etiology inference based on the hormone and metabolism coupling model: Couple the patient's hormone levels with metabolic data, establish a hormone and metabolic function coupling model, reveal the mutual influence mechanism between the two in PCOS, analyze the impact of hormone disorders on metabolic disorders and the impact of metabolic disorders on endocrine function imbalance, including:

[0010] S31, hormone and metabolic data processing: extract hormone levels and metabolic indicators from patients' health data, and pre-process hormone levels and metabolic indicators, including removing noise and outliers;

[0011] S32, construction of hormone and metabolic coupling model: based on pre-treatment hormone levels and metabolic indicators, a coupling model of hormone and metabolic function was constructed;

[0012] S33, Etiological mechanism analysis: Based on the constructed coupling model of hormones and metabolic functions, the etiological mechanism analysis was conducted to reveal the effects of hormone disorders on metabolic disorders and the effects of metabolic disorders on endocrine function imbalance;

[0013] S4, Risk score calculation: Calculate the patient's PCOS risk score based on the integrated health data and the results of etiology inference;

[0014] S5, diagnostic feedback and dynamic optimization: The calculated risk score and the results of etiology inference are fed back to clinicians, and the testing plan is dynamically adjusted according to the patient's biomarker changes, etiology inference and clinical feedback. The testing plan includes a health monitoring plan, treatment plan recommendations and follow-up plan.

[0015] Optionally, the patient data collection in S1 includes:

[0016] S11, imaging data collection: using medical imaging technology (ultrasound, MRI), collect imaging data related to PCOS;

[0017] S12, genetic testing data collection: Collect the patient's genetic testing data, including genetic information on gene mutations, gene expression levels, and polymorphisms;

[0018] S13, hormone level testing: measuring the patient's hormone levels, including follicle-stimulating hormone, luteinizing hormone, and androgen, through blood or body fluid samples;

[0019] S14, metabolic index collection: monitoring the patient's metabolic status, including blood glucose, insulin sensitivity, and lipid metabolism;

[0020] S15, immune marker collection: Collect markers related to immune function through blood or body fluids, including cytokines and immune cell counts;

[0021] S16, physiological fluid data collection: Collect physiological fluid samples (such as saliva, sweat, etc.) from patients and extract metabolites and hormones as supplementary health data.

[0022] Optionally, the biomarker data integration and standardization in S2 includes:

[0023] S21, data standardization: Use data standardization methods to normalize health data from different sources and dimensions;

[0024] S22, Data integration and collaborative analysis: Standardized health data are compared and contrasted through the biomarker database, and health data of different dimensions are integrated to form a synergistic data set.

[0025] Optionally, the data integration and collaborative analysis in S22 includes:

[0026] S221, control and comparison of standardized data: control and compare the standardized health data through the biomarker database, and the comparison process matches and verifies the patient's health data with the known biomarker reference range based on the reference values ​​in the biomarker database;

[0027] S222, principal component analysis dimensionality reduction processing: based on the compared health data, the principal component analysis (PCA) algorithm is used to perform dimensionality reduction processing on the health data;

[0028] S223, Correlation analysis and data integration: Use the Pearson correlation coefficient to measure the correlation between different health data dimensions. Based on the calculated Pearson correlation coefficient, identify the strong or weak correlation between different health data dimensions, and integrate the strongly correlated health data.

[0029] Optionally, the hormone and metabolic data processing in S31 includes:

[0030] S311, noise removal: hormone levels and metabolic indicators were smoothed using the moving average method;

[0031] S312, Outlier Detection and Removal: Detecting outliers in hormone and metabolic data using the triple standard deviation method.

[0032] Optionally, the coupling model of hormone and metabolic function in S32 adopts a graph convolutional network (GCN) model, and the graph convolutional network (GCN) model includes:

[0033] S321, construct weighted adjacency matrix: Construct weighted adjacency matrix A w , used to represent the coupling relationship between hormone levels and metabolic indicators;

[0034] S322, dynamic update mechanism of convolutional layer: Hormone levels and metabolic indicators change dynamically over time and with treatment plans. A dynamic update mechanism is introduced to allow the adjacency matrix to be dynamically adjusted as the data changes. That is, after each layer of convolution calculation, the weight of the adjacency matrix is ​​recalculated;

[0035] S323, Differentiated processing of multimodal data: In order to process hormone levels, metabolic indicators, and genetic testing data, multiple subgraphs are constructed for different data sources. Each subgraph uses a separate adjacency matrix and performs convolution operations based on the characteristics of health data of different modalities.

[0036] S324, Loss function optimization: Optimize classification or regression tasks by adding correlation loss.

[0037] Optionally, the etiology mechanism analysis in S33 includes:

[0038] S331, Analyze the effect of hormones on metabolism: Through multi-layer convolution operations, calculate the effect of hormone nodes on metabolic nodes in each layer. In each layer of the model, hormone features are convolved to affect metabolic features, thereby revealing how hormone levels affect metabolic disorders;

[0039] S332, Analyzing metabolic feedback on hormone function: Capturing the feedback of metabolic dysregulation on hormone function through reverse pathways;

[0040] S333, Nonlinear correlation analysis of causal mechanisms: using Pearson correlation coefficient r HM Calculate the nonlinear correlation between hormones and metabolic indicators and quantify the coupling relationship between hormone H and metabolism M.

[0041] Optionally, the risk score calculation in S4 includes:

[0042] S41, feature weighted scoring: weighting the integrated health data;

[0043] S42, combination of etiology inference results: The etiology inference results based on the coupling model of hormone and metabolic function, combined with the relevant score of the cause S disease Calculate the final PCOS risk score S total ;

[0044] S43, Risk level classification: According to the calculated final PCOS risk score S total, combined with clinical experience, it is divided into different risk levels, including low risk, medium risk and high risk.

[0045] Optionally, the cause-related score S disease The quantification includes:

[0046] Hormone and metabolic coupling feature extraction: Use the graph convolutional network (GCN) model to perform coupling analysis on hormone and metabolic data and extract etiology features;

[0047] Etiology inference score: the final output of the etiology feature matrix After multiple layers of convolution, the cause-related score S is calculated by weighted summation of the generated cause feature matrix. disease .

[0048] Optionally, the diagnostic feedback and dynamic optimization in S5 include:

[0049] S51, feedback result provision: the patient's calculated PCOS risk score S total And the results of etiology inference disease Feedback to clinicians;

[0050] S52, Dynamic adjustment of testing plan: According to the patient's biomarker changes (such as hormone levels, dynamic changes in metabolic indicators), etiology inference results and clinical feedback information, the testing plan is adjusted in real time, including:

[0051] Health monitoring plan: Adjust the monitoring frequency of health data according to the patient's risk score, focusing on monitoring affected hormone levels and metabolic indicators;

[0052] Treatment plan recommendations: Based on the feedback from etiology inference, doctors are advised to adjust the patient's personalized treatment plan, including drug dosage adjustment or replacement of treatment plan;

[0053] Follow-up plan: Provide personalized follow-up recommendations based on the patient's changing risk.

[0054] Beneficial effects of the present invention:

[0055] The present invention, through non-invasive data collection and multimodal health data integration, can comprehensively collect and analyze patients' imaging data, genetic test data, hormone levels, metabolic indicators, immune markers, etc., ensuring multi-dimensional and comprehensive capture of patients' health status. This method not only improves the integrity of the data, but also through standardized processing and collaborative analysis, enables health information from different data sources to be effectively integrated, thereby improving the comparability of the data and the accuracy of the analysis.

[0056] The present invention analyzes the complex interactive relationship between hormones and metabolism through a graph convolutional network model, reveals the pathological mechanism between hormone disorders and metabolic disorders, and by introducing a dynamic update mechanism and multimodal data processing, can capture the dynamic changes of hormones and metabolism during the treatment process, providing more accurate etiology analysis results, and effectively improving the diagnostic accuracy of polycystic ovary syndrome and the scientific nature of risk prediction.

[0057] The present invention realizes personalized diagnosis and treatment management of polycystic ovary syndrome through risk score calculation, diagnostic feedback and dynamic optimization functions. The risk score comprehensively considers the patient's health data and etiology analysis results, and combines with the dynamic adjustment mechanism to optimize the health monitoring plan, treatment plan and follow-up plan in real time according to the patient's biomarker changes and clinical feedback. This method greatly improves the accuracy and flexibility of the diagnosis and treatment plan, provides doctors with all-round data support, and helps personalized and dynamic long-term health management and optimization of treatment plans. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only for the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0059] Figure 1 A schematic diagram of a detection method flow chart of an embodiment of the present invention;

[0060] Figure 2 This is a schematic diagram of etiology inference based on the hormone and metabolism coupling model according to an embodiment of the present invention. DETAILED DESCRIPTION

[0061] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. At the same time, it is explained here that in order to make the embodiments more detailed, the following embodiments are the best and preferred embodiments, and those skilled in the art may also adopt other alternatives to implement some known technologies; and the accompanying drawings are only for more specific description of the embodiments, and are not intended to specifically limit the present invention.

[0062] It should be noted that the references to "one embodiment", "an embodiment", "an exemplary embodiment", "some embodiments" and the like in the specification indicate that the embodiments described may include specific features, structures or characteristics, but not every embodiment may include the specific features, structures or characteristics. In addition, when a specific feature, structure or characteristic is described in conjunction with an embodiment, it should be within the knowledge of a person skilled in the art to implement such feature, structure or characteristic in conjunction with other embodiments (whether or not explicitly described).

[0063] In general, a term can be understood, at least in part, from its use in context. For example, depending, at least in part, on the context, the term "one or more" as used herein can be used to describe any feature, structure, or characteristic in the singular sense, or can be used to describe a combination of features, structures, or characteristics in the plural sense. Additionally, the term "based on" can be understood as not necessarily intended to convey an exclusive set of factors, but can instead, depending, at least in part, on the context, allow for the presence of other factors that are not necessarily explicitly described.

[0064] like Figure 1-Figure 2 As shown, the detection method of polycystic ovary syndrome based on multidimensional biomarker analysis includes the following steps:

[0065] S1, patient data collection: collect multimodal health data of patients, including imaging data, genetic test data, hormone levels, metabolic indicators, immune markers and biomarkers in physiological fluids (such as saliva, sweat, etc.);

[0066] S2, biomarker data integration and standardization: standardize the collected multimodal health data and integrate health data of different dimensions through the biomarker database;

[0067] S3, Etiology inference based on the hormone and metabolism coupling model: Couple the patient's hormone levels with metabolic data, establish a hormone and metabolic function coupling model, reveal the mutual influence mechanism between the two in PCOS, analyze the impact of hormone disorders on metabolic disorders and the impact of metabolic disorders on endocrine function imbalance, including:

[0068] S31, hormone and metabolic data processing: extract hormone levels and metabolic indicators from patients' health data, and pre-process hormone levels and metabolic indicators, including removing noise and outliers;

[0069] S32, construction of hormone and metabolic coupling model: based on pre-treatment hormone levels and metabolic indicators, a coupling model of hormone and metabolic function was constructed;

[0070] S33, Etiological mechanism analysis: Based on the constructed coupling model of hormones and metabolic functions, the etiological mechanism analysis was conducted to reveal the effects of hormone disorders on metabolic disorders and the effects of metabolic disorders on endocrine function imbalance;

[0071] S4, Risk score calculation: Calculate the patient's PCOS risk score based on the integrated health data and the results of etiology inference;

[0072] S5, diagnostic feedback and dynamic optimization: Feedback the calculated risk score and etiology inference results to clinicians, and dynamically adjust the testing plan based on the patient's biomarker changes, etiology inference and clinical feedback. The testing plan includes a health monitoring plan, treatment plan recommendations and follow-up plan;

[0073] Through the above content, based on the comprehensive analysis of multidimensional biomarkers, combined with non-invasive data collection, precise data integration and standardized processing, and etiology inference of the hormone and metabolic function coupling model, it can not only comprehensively capture the patient's health status, but also deeply reveal the potential etiology mechanism of polycystic ovary syndrome, provide personalized risk assessment and dynamic optimization solutions, and provide more accurate, flexible and efficient support for doctors' diagnosis and treatment decisions, thereby improving the comprehensiveness of detection and clinical application value.

[0074] Patient data collection in S1 includes:

[0075] S11, imaging data collection: using medical imaging technology (ultrasound, MRI), collect imaging data related to PCOS;

[0076] S12, genetic testing data collection: Collect the patient's genetic testing data, including genetic information on gene mutations, gene expression levels, and polymorphisms;

[0077] S13, hormone level testing: measuring the patient's hormone levels, including follicle-stimulating hormone, luteinizing hormone, and androgen, through blood or body fluid samples;

[0078] S14, metabolic index collection: monitoring the patient's metabolic status, including blood glucose, insulin sensitivity, and lipid metabolism;

[0079] S15, immune marker collection: Collect markers related to immune function through blood or body fluids, including cytokines and immune cell counts;

[0080] S16, physiological fluid data collection: collect physiological fluid samples (such as saliva, sweat, etc.) from patients and extract metabolites and hormones therein as supplementary health data;

[0081] Through the above content, it can simultaneously cover imaging, genes, hormones, metabolism, immunity and biomarkers in physiological fluids, and comprehensively reflect the patient's health status from multiple dimensions. It not only improves the diagnostic accuracy of polycystic ovary syndrome, but also helps doctors to deeply analyze the causes from multiple angles such as genetics, endocrinology, metabolism and immunity, providing a scientific basis for the formulation of personalized treatment plans, ensuring the comprehensiveness and accuracy of diagnosis and treatment.

[0082] Biomarker data integration and normalization in S2 include:

[0083] S21, Data standardization: Use data standardization methods to normalize health data from different sources and dimensions so that all types of data can be effectively analyzed and compared within the same range;

[0084] S22, Data integration and collaborative analysis: The standardized health data are compared and contrasted through the biomarker database to ensure the accuracy of the biological significance of each data point, and the health data of different dimensions are integrated to form a synergistic data set;

[0085] Through the above content, data from different sources are unified into a comparable range, so that genetic data, hormone levels, metabolic indicators, etc. can be compared and analyzed under the same framework. Through collaborative analysis, these multidimensional data can be effectively integrated to generate a comprehensive data set with biological significance, which provides a more comprehensive basis for accurate diagnosis and personalized treatment of diseases, and helps to improve the efficiency and accuracy of clinical decision-making.

[0086] Data integration and collaborative analysis in S22 include:

[0087] S221, Control and comparison of standardized data: The standardized health data is compared and compared with the biomarker database to ensure the biological significance of each data point is accurate. The comparison process is based on the reference values ​​in the biomarker database, matching and verifying the patient's health data with the known biomarker reference range to ensure the biological accuracy of the data;

[0088] S222, principal component analysis dimensionality reduction processing: Based on the compared health data, the principal component analysis (PCA) algorithm is used to reduce the dimensionality of the health data to reduce redundant information and retain key features, which is expressed as:

[0089] Z = XW;

[0090] Among them, Z is the data matrix after dimension reduction, X is the original health data matrix, and W is the principal component matrix;

[0091] S223, Correlation analysis and data integration: Use the Pearson correlation coefficient to measure the correlation between different health data dimensions. According to the calculated Pearson correlation coefficient, identify the strong or weak correlation between different health data dimensions, and integrate the strongly correlated health data. The Pearson correlation coefficient is expressed as:

[0092]

[0093] Among them, r xy is the correlation between two variables x and y, x i and iThey are data points in health data of different dimensions. and are the means of variables x and y respectively;

[0094] Through the above content, the correlation between health data of different dimensions can be accurately measured, data redundancy can be effectively reduced, and the extraction and integration of key health information can be ensured. Multimodal data (such as hormone levels, metabolic indicators, genetic data, etc.) can be synergistically integrated to provide doctors with a more comprehensive and simplified diagnostic basis, thereby optimizing the accuracy and efficiency of clinical decision-making. At the same time, the efficiency of data processing is improved through algorithms such as dimensionality reduction, making personalized diagnosis and etiology analysis more efficient and with a global perspective.

[0095] Hormone and metabolic data processing in S31 includes:

[0096] S311, noise removal: hormone levels and metabolic indicators are smoothed using the moving average method to reduce random fluctuations in the sampled data, expressed as:

[0097]

[0098] Among them, y t is the smoothed value, x t-i is the data value at the i-th moment before time point t, and n is the window size;

[0099] S312, Outlier Detection and Removal: Detect outliers in hormone and metabolic data using the triple standard deviation method, expressed as:

[0100]

[0101] Where z is the standardized value, x is the original data point in the hormone or metabolic data, μ is the mean of the data set, and σ is the standard deviation;

[0102] Through the above content, the noise in the sampling process is effectively reduced, and the outliers are accurately identified and removed through the triple standard deviation method, ensuring the stability and accuracy of the data, which helps to improve the reliability of subsequent analysis, making the cause inference and diagnosis results more accurate, and providing a high-quality data foundation for the formulation of personalized treatment plans.

[0103] The coupling model of hormone and metabolic function in S32 adopts the graph convolutional network (GCN) model, which includes:

[0104] S321, construct weighted adjacency matrix: Construct weighted adjacency matrix A w , which is used to represent the coupling relationship between hormone levels and metabolic indicators, is expressed as:

[0105] A w =αij A;

[0106] Among them, A w is the weighted adjacency matrix, α ij is the weight between hormone i and metabolic index j, A is the standard adjacency matrix, representing the interconnections between hormone and metabolic variables;

[0107] S322, dynamic update mechanism of convolution layer: Hormone levels and metabolic indicators change dynamically with time and treatment plans. The dynamic update mechanism is introduced to allow the adjacency matrix to be dynamically adjusted as the data changes. That is, after each layer of convolution calculation, the weight of the adjacency matrix is ​​recalculated, expressed as:

[0108]

[0109] Among them, H (l+1) is the output of the l+1th layer, D w is a weighted degree matrix, the value of which is the sum of the row or column elements of the adjacency matrix. (l) is the node feature matrix of the lth layer, W (l) is the weight matrix of the lth layer, σ is the ReLU activation function;

[0110] S323, Differentiated processing of multimodal data: In order to process hormone levels, metabolic indicators, and genetic testing data, multiple subgraphs are constructed for different data sources. Each subgraph uses a separate adjacency matrix and performs convolution operations based on the characteristics of health data of different modalities, expressed as:

[0111]

[0112] Among them, H final is the final output after the integration of multimodal features, M represents the number of multimodal data (such as hormones, metabolism, genes, etc.), is the output of the mth mode in the l+1th layer, β m is the weight of each mode, which is used to adjust the influence of different modes on the final result;

[0113] S324, loss function optimization: By adding correlation loss, the classification or regression task is optimized to ensure that the correlation between different health data dimensions is effectively captured and utilized by the model, expressed as:

[0114]

[0115] in, is the total loss function, is the task-related loss (such as classification loss or regression loss), is the correlation loss, which measures the correlation difference between hormones and metabolic indicators, and λ is the balance term, which controls the weights of the two loss terms;

[0116] Task-related losses Choose according to the task type, including:

[0117] Classification task: If the task is classification, use the cross entropy loss function, expressed as:

[0118]

[0119] Where N is the number of samples, y i is the actual label (such as hormone function classification), is the probability distribution predicted by the model;

[0120] Regression task: If the task is regression (such as predicting metabolic indicators), use the mean square error (MSE), expressed as:

[0121]

[0122] Among them, y i is the actual value, is the predicted value of the model;

[0123] Correlation loss It is used to measure the correlation between two variables (hormones and metabolic indicators) to ensure that the model can capture the relationship between these variables. The negative value of the Pearson correlation coefficient is used to define the correlation loss, which is expressed as:

[0124]

[0125]

[0126] Among them, r xy is the Pearson correlation coefficient, x i and i They are hormone and metabolic data, and are the means of x and y respectively;

[0127] Through the above content, the introduction of weighted adjacency matrix, dynamic update mechanism and multimodal data processing can more accurately capture the complex interactive relationship between hormone levels and metabolic indicators. At the same time, the multi-objective optimization mechanism of task loss, correlation loss and smoothing loss is combined, so that the model can not only effectively complete the classification or regression tasks, but also maintain the stability and accuracy of the model during dynamic adjustment, making the model more suitable for processing the coupled analysis of multidimensional health data, providing strong support for personalized diagnosis and precision treatment, and significantly improving the diagnostic efficiency and scientificity of the present invention.

[0128] The etiological mechanism analysis in S33 includes:

[0129] S331, Analyze the impact of hormones on metabolism: Through multi-layer convolution operations, calculate the impact of hormone nodes on metabolic nodes in each layer. In each layer of the model, the characteristics of hormones are convolved to affect the characteristics of metabolism, thereby revealing how hormone levels affect metabolic disorders. For example, abnormal androgen levels may affect insulin sensitivity or blood sugar levels. The deeper the convolution layer, the more complex the impact captured, which can be expressed as:

[0130]

[0131] in, is the feature matrix of the metabolic node in the l+1th layer, which represents the features of the metabolic node after being affected by the hormone node. is the feature matrix of the hormone node in the lth layer, which represents the features of the hormone node in the convolutional layer l. wH→M It is a directed weighted adjacency matrix representing hormone nodes to metabolic nodes, which is specifically used to represent the effect of hormones on metabolism. wH is the weighted degree matrix of hormone nodes, used for normalization of the adjacency matrix, It is a weight matrix representing the effect of hormones on metabolism, which is specifically used to control the intensity of this effect in convolutional layer l;

[0132] S332, Analysis of metabolic feedback on hormonal function: The feedback of metabolic disorders on hormonal function is captured through the reverse path. For example, insulin resistance may in turn affect the level of follicle-stimulating hormone (FSH) or luteinizing hormone (LH). Through the reverse convolution operation, the model can capture the feedback effect of this metabolic disorder on hormonal regulation, which is expressed as:

[0133]

[0134] in, is the feature matrix of the hormone node in the l+1th layer, which represents the characteristics of the hormone node after being reflected by the metabolic node. is the feature matrix of the metabolic node in the lth layer, which represents the features of the metabolic node in the convolutional layer l. wM→H It represents the directed weighted adjacency matrix of metabolic nodes to hormone nodes, which is specifically used to represent the feedback of metabolism to hormones. wM is the weighted degree matrix of metabolic nodes, used to normalize the adjacency matrix, It is the weight matrix representing the metabolic feedback to the hormone, which is used in the convolutional layer l to control the intensity of this feedback effect;

[0135] S333, Nonlinear correlation analysis of causal mechanisms: using Pearson correlation coefficient r HMThe nonlinear correlation between hormones and metabolic indicators was calculated to quantify the coupling relationship between hormone H and metabolism M, which is expressed as:

[0136]

[0137] Among them, r HM is the Pearson correlation coefficient between hormone level H and metabolic index M, H i is the data point of the ith hormone level, is the mean of hormone level data, M i is the data point of the ith metabolic index, is the mean value of metabolic index data;

[0138] Through the above content, not only can the bidirectional effects of hormones and metabolism be reflected through nonlinear convolution operations, but the coupling strength between the two can also be quantitatively evaluated through correlation analysis, providing reliable model support for in-depth understanding of disease mechanisms and improving the diagnostic accuracy of multidimensional biomarkers and the effectiveness of personalized treatment.

[0139] The risk score calculation in S4 includes:

[0140] S41, feature weighted scoring: weighted processing of the integrated health data, expressed as:

[0141]

[0142] Among them, S feature represents the feature weighted score, w i is the weight of the i-th feature, reflecting the contribution of this feature to the risk of PCOS, f i is the i-th eigenvalue (hormone level, metabolic index);

[0143] S42, combination of etiology inference results: The etiology inference results based on the coupling model of hormone and metabolic function, combined with the relevant score of the cause S disease Calculate the final PCOS risk score S total , expressed as:

[0144] S total =α·S feature +β·S disease ;

[0145] Among them, S total represents the final PCOS risk score, S disease It represents the characteristic score based on etiology analysis, α and β are adjustment coefficients, which are used to balance the weight of characteristic score and etiology inference result;

[0146] S43, Risk level classification: According to the calculated final PCOS risk score S total , combined with clinical experience, divided into different risk levels, including low risk, medium risk and high risk;

[0147] Through the above content, the patient's PCOS risk can be assessed more accurately, the graph convolutional network model can be used to extract the coupling characteristics of hormone and metabolic function, and the complex interaction between hormone disorders and metabolic disorders can be quantified. Then, a weighted algorithm is used to reasonably allocate the impact of each feature on the risk score. By integrating the etiology inference score with health data, the final risk score not only improves the accuracy and personalization of the prediction, but also provides a more reliable basis for clinicians, which helps to optimize diagnosis and treatment plans and conduct early intervention.

[0148] Cause-related score S disease The quantification includes:

[0149] Hormone and metabolic coupling feature extraction: The graph convolutional network (GCN) model is used to perform coupling analysis on hormone and metabolic data and extract etiological features, which are expressed as:

[0150]

[0151] in, is the etiology feature matrix after l+1 layers of graph convolution, A w is a weighted adjacency matrix representing the interaction between hormone levels and metabolic indicators, D w is the degree matrix, used to normalize the adjacency matrix A w , H (l) is the input feature matrix in layer l, including hormone and metabolic data, W (l) is the weight matrix of the lth layer, σ is the ReLU activation function, which is used to introduce nonlinear effects;

[0152] Etiology inference score: the final output of the etiology feature matrix After multiple layers of convolution, the cause-related score S is calculated by weighted summation of the generated cause feature matrix. disease , expressed as:

[0153]

[0154] in, is the value of the i-th cause-related feature after convolution in the L-layer (final layer), w disease,i is the weight of the etiology-related feature, reflecting its contribution to the risk of PCOS, and n is the number of etiology features;

[0155] Through the above content, the complex coupling relationship between hormones and metabolic functions can be accurately captured and evaluated, thereby effectively revealing the etiology of PCOS. The score is based on the graph convolutional network model, which extracts etiology features through multi-layer convolution and performs weighted summation on them, ensuring that the contribution of each feature in the final etiology inference is reasonably reflected, thereby improving the accuracy of risk assessment and the scientific nature of diagnosis.

[0156] Diagnostic feedback and dynamic optimization in S5 include:

[0157] S51, feedback result provision: the patient's calculated PCOS risk score S total And the results of etiology inference disease Feedback to clinicians to help them make accurate diagnostic decisions;

[0158] S52, Dynamic adjustment of testing plan: According to the patient's biomarker changes (such as hormone levels, dynamic changes in metabolic indicators), etiology inference results and clinical feedback information, the testing plan is adjusted in real time, including:

[0159] Health monitoring plan: Adjust the monitoring frequency of health data according to the patient's risk score, focusing on monitoring affected hormone levels and metabolic indicators;

[0160] Treatment plan recommendations: Based on the feedback from etiology inference, doctors are advised to adjust the patient's personalized treatment plan, including drug dosage adjustment or replacement of treatment plan;

[0161] Follow-up plan: Provide personalized follow-up recommendations based on the patient's risk changes to ensure dynamic tracking and evaluation of the patient during treatment;

[0162] Through the above content, the calculated risk scores and etiology inference results can be promptly fed back to clinicians, ensuring that doctors can make decisions based on the latest data information, dynamically adjust the patient's testing plan, and make real-time optimization based on changes in biomarkers and clinical feedback, including health monitoring plans, treatment plan recommendations and follow-up arrangements. This feedback mechanism improves the accuracy of diagnosis and treatment, helps doctors provide patients with personalized and dynamic medical services, effectively reduces treatment risks and improves patients' long-term health management effects.

[0163] The present invention covers any substitution, modification, equivalent method and scheme made on the essence and scope of the present invention. In order to make the public have a thorough understanding of the present invention, specific details are described in detail in the following preferred embodiments of the present invention, but those skilled in the art can fully understand the present invention without the description of these details. In addition, in order to avoid unnecessary confusion about the essence of the present invention, well-known methods, processes, procedures, components and circuits are not described in detail.

[0164] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for detecting polycystic ovary syndrome based on multidimensional biomarker analysis, characterized in that: The following steps are involved: S1, patient data collection: collect multimodal health data of patients, including imaging data, genetic testing data, hormone levels, metabolic indicators, immune markers, and biomarkers in physiological fluids; S2, biomarker data integration and standardization: standardize the collected multimodal health data and integrate health data of different dimensions through the biomarker database; S3, Etiology inference based on the hormone and metabolism coupling model: Couple the patient's hormone levels with metabolic data, establish a hormone and metabolic function coupling model, reveal the mutual influence mechanism between the two in PCOS, analyze the impact of hormone disorders on metabolic disorders and the impact of metabolic disorders on endocrine function imbalance, including: S31, hormone and metabolic data processing: extract hormone levels and metabolic indicators from patients' health data, and pre-process hormone levels and metabolic indicators, including removing noise and outliers; S32, construction of hormone and metabolic coupling model: based on pre-treatment hormone levels and metabolic indicators, a coupling model of hormone and metabolic function was constructed; S33, Etiological mechanism analysis: Based on the constructed coupling model of hormones and metabolic functions, the etiological mechanism analysis was conducted to reveal the effects of hormone disorders on metabolic disorders and the effects of metabolic disorders on endocrine function imbalance; S4, Risk score calculation: Calculate the patient's PCOS risk score based on the integrated health data and the results of etiology inference; S5, diagnostic feedback and dynamic optimization: The calculated risk score and the results of etiology inference are fed back to clinicians, and the testing plan is dynamically adjusted according to the patient's biomarker changes, etiology inference and clinical feedback. The testing plan includes a health monitoring plan, treatment plan recommendations and follow-up plan.

2. The method for detecting polycystic ovary syndrome based on multidimensional biomarker analysis according to claim 1, characterized in that: The patient data collection in S1 includes: S11, imaging data collection: using medical imaging technology to collect imaging data related to PCOS; S12, genetic testing data collection: Collect the patient's genetic testing data, including genetic information on gene mutations, gene expression levels, and polymorphisms; S13, hormone level testing: measuring the patient's hormone levels, including follicle-stimulating hormone, luteinizing hormone, and androgen, through blood or body fluid samples; S14, metabolic index collection: monitoring the patient's metabolic status, including blood glucose, insulin sensitivity, and lipid metabolism; S15, immune marker collection: Collect markers related to immune function through blood or body fluids, including cytokines and immune cell counts; S16, physiological fluid data collection: Collect physiological fluid samples from patients and extract metabolites and hormones as supplementary health data.

3. The method for detecting polycystic ovary syndrome based on multidimensional biomarker analysis according to claim 1, characterized in that: The biomarker data integration and standardization in S2 includes: S21, data standardization: Use data standardization methods to normalize health data from different sources and dimensions; S22, Data integration and collaborative analysis: Standardized health data are compared and contrasted through the biomarker database, and health data of different dimensions are integrated to form a synergistic data set.

4. The method for detecting polycystic ovary syndrome based on multidimensional biomarker analysis according to claim 3, characterized in that: The data integration and collaborative analysis in S22 include: S221, control and comparison of standardized data: control and compare the standardized health data through the biomarker database, and the comparison process matches and verifies the patient's health data with the known biomarker reference range based on the reference values ​​in the biomarker database; S222, principal component analysis dimensionality reduction processing: based on the compared health data, the principal component analysis algorithm is used to perform dimensionality reduction processing on the health data; S223, Correlation analysis and data integration: Use the Pearson correlation coefficient to measure the correlation between different health data dimensions. Based on the calculated Pearson correlation coefficient, identify the strong or weak correlation between different health data dimensions, and integrate the strongly correlated health data.

5. The method for detecting polycystic ovary syndrome based on multidimensional biomarker analysis according to claim 1, characterized in that: The hormone and metabolic data processing in S31 includes: S311, noise removal: hormone levels and metabolic indicators were smoothed using the moving average method; S312, Outlier Detection and Removal: Detecting outliers in hormone and metabolic data using the triple standard deviation method.

6. The method for detecting polycystic ovary syndrome based on multidimensional biomarker analysis according to claim 5, characterized in that: The coupling model of hormones and metabolic functions in S32 adopts a graph convolutional network model, and the graph convolutional network model includes: S321, construct weighted adjacency matrix: Construct weighted adjacency matrix A w , used to represent the coupling relationship between hormone levels and metabolic indicators; S322, dynamic update mechanism of convolutional layer: Hormone levels and metabolic indicators change dynamically over time and with treatment plans. A dynamic update mechanism is introduced to allow the adjacency matrix to be dynamically adjusted as the data changes. That is, after each layer of convolution calculation, the weight of the adjacency matrix is ​​recalculated; S323, Differentiated processing of multimodal data: In order to process hormone levels, metabolic indicators, and genetic testing data, multiple subgraphs are constructed for different data sources. Each subgraph uses a separate adjacency matrix and performs convolution operations based on the characteristics of health data of different modalities. S324, Loss function optimization: Optimize classification or regression tasks by adding correlation loss.

7. The method for detecting polycystic ovary syndrome based on multidimensional biomarker analysis according to claim 6, characterized in that: The etiology analysis in S33 includes: S331, Analyze the effect of hormones on metabolism: Through multi-layer convolution operations, calculate the effect of hormone nodes on metabolic nodes in each layer. In each layer of the model, hormone features are convolved to affect metabolic features, thereby revealing how hormone levels affect metabolic disorders; S332, Analyzing metabolic feedback on hormone function: Capturing the feedback of metabolic dysregulation on hormone function through reverse pathways; S333, Nonlinear correlation analysis of causal mechanisms: using Pearson correlation coefficient r HM Calculate the nonlinear correlation between hormones and metabolic indicators and quantify the coupling relationship between hormone H and metabolism M.

8. The method for detecting polycystic ovary syndrome based on multidimensional biomarker analysis according to claim 7, characterized in that: The risk score calculation in S4 includes: S41, feature weighted scoring: weighting the integrated health data; S42, combination of etiology inference results: The etiology inference results based on the coupling model of hormone and metabolic function, combined with the relevant score of the cause S disease Calculate the final PCOS risk score S total ; S43, Risk level classification: According to the calculated final PCOS risk score S total , combined with clinical experience, it is divided into different risk levels, including low risk, medium risk and high risk.

9. The method for detecting polycystic ovary syndrome based on multidimensional biomarker analysis according to claim 8, characterized in that: The cause's associated score S disease The quantification includes: Hormone and metabolic coupling feature extraction: Use graph convolutional network models to conduct coupling analysis on hormone and metabolic data and extract etiology features; Etiology inference score: the final output of the etiology feature matrix After multiple layers of convolution, the cause-related score S is calculated by weighted summation of the generated cause feature matrix. disease .

10. The method for detecting polycystic ovary syndrome based on multidimensional biomarker analysis according to claim 9, characterized in that: The diagnostic feedback and dynamic optimization in S5 include: S51, feedback result provision: the patient's calculated PCOS risk score S total And the results of etiology inference disease Feedback to clinicians; S52, Dynamically adjust the testing plan: According to the patient's biomarker changes, etiology inference results and clinical feedback information, the testing plan is adjusted in real time, including: Health monitoring plan: Adjust the monitoring frequency of health data according to the patient's risk score, focusing on monitoring affected hormone levels and metabolic indicators; Treatment plan recommendations: Based on the feedback from etiology inference, doctors are advised to adjust the patient's personalized treatment plan, including drug dosage adjustment or replacement of treatment plan; Follow-up plan: Provide personalized follow-up recommendations based on the patient's changing risk.

Images