Medical data analysis method and system for postmenopausal osteoporosis patient

Abstract

The invention relates to the technical field of medical data processing, in particular to a postmenopausal osteoporosis patient medical data analysis method and system, and the method comprises the following steps: extracting a bone mineral density measurement value and a bone metabolism index concentration based on patient bone mineral density data, bone metabolism index data and fracture event records; according to the method, through combination of covariance matrix calculation, feature value and feature vector extraction and revealing of deep relevance between data, fine description of bone mineral density and metabolism index changes is achieved, state probability values, fracture records and drug use data are subjected to normalization weighting processing, and an optimized posterior state distribution matrix is generated; the accuracy and comprehensiveness of data fusion are improved, a key dynamic feature set is extracted through screening and accumulation operation, the pertinence and efficiency of analysis and normalization and time interval prediction of a trend curve are optimized, a quantitative result of the bone mineral density change trend and the fracture risk is formed, and high-precision support is provided for illness state assessment and personalized analysis.

Description

Technical Field

[0001] The present invention relates to the technical field of medical data processing, and in particular to a medical data analysis method and system for postmenopausal osteoporosis patients. Background Art

[0002] The field of medical data processing technology refers to the technical system for the collection, storage, analysis, processing and application of medical and health-related data. This technical field covers electronic medical record analysis, medical image analysis, personalized medical recommendations, health data monitoring, disease prediction and diagnosis, etc., using modern information technology and data science methods to improve the efficiency and quality of medical services. In this field, data accuracy, security and privacy protection are particularly important, and related technologies usually include artificial intelligence algorithms, big data processing, cloud computing and Internet of Things technologies. While promoting the digital transformation of the medical industry, these technologies also provide support for precision medicine and health analysis.

[0003] Among them, the medical data analysis method for postmenopausal osteoporosis patients refers to the process of comprehensively collecting, storing and analyzing the health information of postmenopausal osteoporosis patients through medical data processing technology, thereby providing scientific support for the analysis of osteoporosis of patients. Its main uses include the formulation of personalized treatment plans, real-time monitoring of disease progression, evaluation of treatment effects, and support for auxiliary medical decision-making, thereby helping to improve the quality of life of patients and the efficiency of disease analysis.

[0004] Existing technologies focus on a single data dimension, ignore the dynamic interaction between multidimensional data, and lack time series correlation modeling. In the analysis of postmenopausal osteoporosis, only bone density changes are focused on, and bone metabolism indicators and fracture events are not effectively integrated, resulting in limitations in disease analysis. Multi-source data fusion fails to fully characterize the state dynamics, lacks accurate trend prediction and risk assessment, resulting in one-sided disease assessment and insufficient targeted treatment plans. Quantitative support for trend changes is limited, making it difficult to meet the needs of precision medicine. These deficiencies reduce the application value of data analysis in personalized treatment and real-time decision-making, affecting the timeliness and scientific nature of medical services. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a medical data analysis method and system for postmenopausal osteoporosis patients.

[0006] In order to achieve the above object, the present invention adopts the following technical solution: a medical data analysis method for postmenopausal osteoporosis patients, comprising the following steps: S1: Based on the patient's bone density data, bone metabolism index data and fracture event records, extract the bone density measurement value and bone metabolism index concentration, calculate the change between time points, count the number of fracture events, and generate a dynamic feature time series matrix; S2: Based on the dynamic feature time series matrix, extract the bone density change and the bone metabolism index change value, calculate the covariance matrix between the change values, extract the eigenvalue and eigenvector of the covariance, establish a transfer parameter matrix, and generate a hidden state conversion result; S3: Based on the hidden state conversion result, the state probability value in the transfer parameter is calculated, the fracture event record and the drug use data are combined, the state probability is superimposed, normalized and weighted, and an optimized posterior state distribution matrix is ​​generated; S4: Based on the optimized posterior state distribution matrix, extract the state probability values ​​of the bone density change rate and the bone metabolism index change rate, screen and accumulate the probability values, calculate the cumulative value of the influencing features, and generate a key dynamic feature set; S5: Based on the key dynamic feature set, extract the bone density change trend and fracture event prediction data, calculate the cumulative value of the prediction probability and time interval, normalize and generate a trend curve, and generate a feature evolution prediction value.

[0007] The dynamic feature time series matrix includes bone density measurement values, bone metabolism index concentrations, bone density changes, bone metabolism index changes, and the number of fracture events. The hidden state conversion results include bone density change eigenvalues, bone metabolism index change value eigenvalues, covariance eigenvalues, covariance eigenvectors, and transfer parameter matrices. The optimized posterior state distribution matrix includes state probability values, fracture event records, drug use data, normalized probability values, and weighted calculation values. The key dynamic feature set includes bone density change rate, bone metabolism index change rate, state probability value, cumulative calculation results, and impact feature cumulative values. The feature evolution prediction value includes bone density change trend, fracture event prediction data, prediction probability, time interval cumulative value, and trend curve.

[0008] As a further solution of the present invention, the step of acquiring the dynamic feature time series matrix is ​​specifically as follows: S111: extracting each bone density measurement value and bone metabolism index concentration from the patient's data record, calculating the change between two consecutive measurements, and obtaining the bone density change and bone metabolism change; S112: Record and count the number of fractures of each patient, and combine the bone density change and bone metabolism change to form time series characteristic data; S113: Integrate the time series feature data using the formula: ; The dynamic characteristic time series matrix of each patient is calculated; in, Representing patients In the The dynamic characteristic matrix value of the measurement, Representative The patient in The change in bone density measured each time, Representative The patient in The changes in bone metabolism indicators measured by Representative Number of fractures per patient.

[0009] As a further solution of the present invention, the step of obtaining the hidden state conversion result is specifically: S211: extracting the bone density change and the bone metabolism index change value from the dynamic feature time series matrix, calculating the covariance matrix based on the change values, and performing covariance component calculation on the bone density change and the bone metabolism index change to generate a covariance matrix; S212: performing eigendecomposition on the covariance matrix to extract eigenvalues ​​and eigenvectors of the covariance matrix, and simultaneously calculating a plurality of eigenvalues ​​and corresponding eigenvectors based on distribution characteristics of the elements of the covariance matrix to obtain an eigenvalue matrix and an eigenvector matrix; S213: Based on the eigenvalue matrix and the eigenvector matrix, the formula is adopted: ; Combine the weight relationship between the elements of the matrix to calculate and generate the transfer parameter matrix; in, Indicates eigenvalues, Represents the eigenvector of the covariance matrix The row corresponding to eigenvector components, Represents the eigenvector of the covariance matrix The row corresponding to eigenvector components, is the total number of eigenvalues, represents the first Line Elements of a column; S214: Based on the transfer parameter matrix, analyze the probability transfer relationship of the elements in the matrix, calculate the conversion mode between hidden states, and generate a hidden state conversion result.

[0010] As a further solution of the present invention, the step of obtaining the optimized posterior state distribution matrix is ​​specifically as follows: S311: extracting a transfer parameter matrix from the hidden state conversion result, calculating the probability value of each state, traversing the transfer probability elements of each row in the matrix, accumulating the probability values ​​of the states corresponding to the fracture event records, and generating a superimposed state probability matrix; S312: Based on the state probability matrix, combined with the drug action intensity and frequency in the drug use data, the drug action intensity is used as a weight component and the frequency is used as an adjustment factor, and each state probability is weighted to generate a weighted state probability matrix; S313: Based on the weighted state probability matrix, the probability value is normalized using the formula: ; Calculate the normalized state probability value and generate the optimized posterior state distribution matrix; in, is the drug effect intensity weight parameter, To adjust the factor weight parameters, Tuning parameters for normalized benchmarks, is the weighted state probability matrix Line The probability value of the column, is the adjustment factor corresponding to the frequency of drug use, is the total number of states, Normalized state probability value.

[0011] As a further solution of the present invention, the step of acquiring the key dynamic feature set is specifically: S411: extracting state probability values ​​of bone density change rate and bone metabolism index change rate from the optimized posterior state distribution matrix, filtering the state probability values ​​according to the screening conditions, removing probability values ​​below the threshold, retaining the corresponding change rates, and generating a screened state probability value set; S412: Based on the screened state probability value set, group and accumulate the bone density change rate and the bone metabolism index change rate according to the weight of the state probability value, calculate the accumulated change rate weighted value based on the state probability value, and generate a cumulative state probability change rate matrix; S413: Based on the cumulative state probability change rate matrix, the cumulative value of each dynamic feature is calculated using the formula: ; Generate a set of key dynamic features; in, Indicates The cumulative value of dynamic features, is the cumulative state probability change rate matrix The probability value of a state, For the Under the state The rate of change of a dynamic feature, is the deviation value between the dynamic feature and its mean value in the state, is the mean of the feature, is the total number of states.

[0012] As a further solution of the present invention, the step of obtaining the feature evolution prediction value is specifically as follows: S511: extracting bone density change trend and fracture event prediction data from the key dynamic feature set, calculating state probability values ​​and time intervals corresponding to multiple time points based on the extracted trend data, and establishing a mapping matrix between state probability and time interval; S512: Based on the state probability and time interval mapping matrix, the state probability values ​​in the matrix are accumulated, and the time intervals are weightedly accumulated to generate a cumulative prediction probability time interval matrix; S513: Based on the cumulative prediction probability time interval matrix, calculate the normalized trend curve at each time point using the formula: ; generating feature evolution predictions; in, Indicates The characteristic evolution value at a time point, For the The dynamic features are The state probability value of the time interval is For the time interval, For the The dynamic features are The rate of change at a time point, For the The characteristic adjustment factor for the time interval, is the total number of time intervals.

[0013] A medical data analysis system for postmenopausal osteoporosis patients, the medical data analysis system for postmenopausal osteoporosis patients is used to execute the above-mentioned medical data analysis method for postmenopausal osteoporosis patients, the system comprising: The bone density dynamic analysis module extracts bone density measurement values ​​and bone metabolism index concentrations based on bone density data, bone metabolism index data and fracture event records, calculates the bone density change and bone metabolism index change values ​​between time points, counts the number of fracture events, and organizes them into a dynamic feature time series matrix; The hidden state feature extraction module extracts the bone density change and the bone metabolism index change value based on the dynamic feature time series matrix, calculates the covariance matrix between the change values, extracts the eigenvalue and eigenvector of the covariance, establishes a transfer parameter matrix, and obtains the hidden state conversion parameters; The state probability distribution calculation module calculates the state probability value in the transfer parameter based on the hidden state transition parameter, combines the fracture event record and the drug use data, superimposes the fracture event record and the drug use data with the state probability value, and forms a posterior state distribution matrix after normalization and weighted calculation; The characteristic evolution prediction module extracts the state probability values ​​of the bone density change rate and the bone metabolism index change rate based on the posterior state distribution matrix, screens and accumulates the probability values, calculates the cumulative value of the bone density change trend and the fracture event prediction data, normalizes them to form a trend curve, and obtains the trend evolution characteristic value.

[0014] Compared with the prior art, the advantages and positive effects of the present invention are: In the present invention, by extracting bone density data, bone metabolism index concentration and fracture event records, calculating the change amount and event statistics, and generating a dynamic feature time series matrix, a quantitative basis is provided for data analysis. Combined with the covariance matrix calculation, eigenvalues ​​and eigenvectors are extracted to reveal the deep correlation between the data and achieve a detailed description of the changes in bone density and metabolic indicators. The state probability values, fracture records, and drug use data are normalized and weighted to generate an optimized posterior state distribution matrix to improve the accuracy and comprehensiveness of data fusion. The key dynamic feature set is extracted through screening and cumulative operations to optimize the pertinence and efficiency of the analysis. The normalization of the trend curve and the time interval prediction form the quantitative results of the bone density change trend and the fracture risk, providing high-precision support for disease assessment and personalized analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a schematic diagram of the workflow of the present invention; Figure 2 The following is a flow chart of the steps for obtaining the dynamic feature timing matrix of the present invention; Figure 3 Flow chart of the steps for obtaining the hidden state conversion result of the present invention; Figure 4 A flow chart of the steps for obtaining the optimized posterior state distribution matrix of the present invention; Figure 5 A flowchart of the steps for obtaining a key dynamic feature set of the present invention; Figure 6 The figure is a flow chart of the steps for obtaining the feature evolution prediction value of the present invention. DETAILED DESCRIPTION

[0016] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0017] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, "multiple" means two or more, unless otherwise clearly and specifically defined.

[0018] See also Figure 1 The present invention provides a technical solution: a method for analyzing medical data of postmenopausal osteoporosis patients, comprising the following steps: S1: Based on the patient's bone density data, bone metabolism index data and fracture event records, extract the bone density measurement value and bone metabolism index concentration, calculate the change between time points, count the number of fracture events, and generate a dynamic feature time series matrix; S2: Based on the dynamic feature time series matrix, extract the change in bone density and bone metabolism index, calculate the covariance matrix between the change values, extract the eigenvalues ​​and eigenvectors of the covariance, establish the transfer parameter matrix, and generate the hidden state conversion results; S3: Based on the hidden state conversion results, the state probability value in the transfer parameter is calculated, the fracture event record and drug use data are combined, the state probability is superimposed, normalized and weighted, and the optimized posterior state distribution matrix is ​​generated; S4: Based on the optimized posterior state distribution matrix, the state probability values ​​of the bone density change rate and the bone metabolism index change rate are extracted, the probability values ​​are screened and accumulated, the cumulative values ​​of the influencing features are calculated, and a set of key dynamic features is generated; S5: Based on the key dynamic feature set, extract the bone density change trend and fracture event prediction data, calculate the cumulative value of the prediction probability and time interval, normalize and generate the trend curve, and generate the feature evolution prediction value.

[0019] The dynamic feature time series matrix includes bone density measurement values, bone metabolism index concentrations, bone density changes, bone metabolism index changes, and the number of fracture events. The hidden state conversion results include bone density change eigenvalues, bone metabolism index change value eigenvalues, covariance eigenvalues, covariance eigenvectors, and transfer parameter matrix. The optimized posterior state distribution matrix includes state probability values, fracture event records, drug usage data, normalized probability values, and weighted calculation values. The key dynamic feature set includes bone density change rate, bone metabolism index change rate, state probability value, cumulative calculation results, and cumulative values ​​of influencing features. The feature evolution prediction value includes bone density change trend, fracture event prediction data, prediction probability, time interval cumulative value, and trend curve.

[0020] See also Figure 2 , the specific steps for obtaining the dynamic feature timing matrix are: S111: extracting each bone density measurement value and bone metabolism index concentration from the patient's data record, calculating the change between two consecutive measurements, and obtaining the bone density change and bone metabolism change; Automatic data acquisition is achieved through a specially designed data collection system. The system automatically records data each time a patient visits. After the data is recorded, the data processing system screens and processes abnormal values ​​to ensure data quality. The processed data is stored in the database for subsequent analysis. The standard difference calculation method is used to calculate the difference. The difference between each measured data and the previous one is the change in that time. This processing method ensures the real-time and accuracy of the data.

[0021] S112: Record and count the number of fractures of each patient, and combine the change in bone density and the change in bone metabolism to form time series characteristic data; First, the patient's fracture data is extracted through the medical record system, and then the number of fractures is counted using statistical software. The software can handle complex medical data and ensure the accuracy of the statistical results. At the same time, the counted number of fractures is merged with the data on changes in bone density and bone metabolism to form a complete patient data record. These records will be used for further data analysis and health prediction. The entire process ensures the integrity and reliability of the data.

[0022] S113: Integrate the time series feature data using the formula: ; The dynamic characteristic time series matrix of each patient is calculated; in, Representing patients In the The dynamic characteristic matrix value of the measurement, Representative The patient in The change in bone density measured each time, Representative The patient in The changes in bone metabolism indicators measured by Representative Number of fractures per patient.

[0023] formula: ; The benefit of the formula is that, by introducing square, square root and logarithmic calculations, the formula can more carefully reflect the nonlinear relationship between changes in bone density and bone metabolism indicators and the impact of the logarithmic transformation of the number of fractures on the results. This method improves the model's sensitivity to data fluctuations and helps to more accurately predict the patient's bone health status.

[0024] Detailed explanation of the formula and the process of formula calculation and derivation: Set specific data: , derivation process: ; calculate: ; calculate: ; calculate: ; The results show that for a given change in bone density, bone metabolism, and number of fractures, the calculated value of the dynamic characteristic time series matrix is ​​1.1854. This value will be used in further analysis and health prediction models to assess the patient's bone health status and provide a quantitative health indicator.

[0025] See also Figure 3 , the steps to obtain the hidden state conversion results are as follows: S211: extracting the bone density change and the bone metabolism index change value from the dynamic feature time series matrix, calculating the covariance matrix based on the change value, and performing covariance component calculation on the bone density change and the bone metabolism index change to generate a covariance matrix; Based on these change values, the covariance matrix between the two is calculated and the extracted change values ​​are called. This process involves statistical analysis of continuous data sets to determine the relationship between variables. The covariance between each pair of variables is obtained by calculating the average of the products of each pair of variables minus the product of the variable means. The obtained covariance matrix provides the necessary basic data for the next step of feature decomposition. The covariance matrix serves as the basic data for subsequent calculations.

[0026] S212: performing eigendecomposition on the covariance matrix to extract eigenvalues ​​and eigenvectors of the covariance matrix, and calculating a plurality of eigenvalues ​​and corresponding eigenvectors based on the distribution characteristics of the elements of the covariance matrix to obtain an eigenvalue matrix and an eigenvector matrix; This step involves using mathematical tools to process the covariance matrix to identify the main directions of the data. Through eigendecomposition, an orthogonal basis can be found that maximizes the variance of the data on this basis vector. Each eigenvector is associated with an eigenvalue, which represents the magnitude of the variance in that direction, resulting in an eigenvalue matrix and an eigenvector matrix.

[0027] S213: Based on the eigenvalue matrix and eigenvector matrix, the formula is: ; Combine the weight relationship between the elements of the matrix to calculate and generate the transfer parameter matrix; in, Indicates eigenvalues, Represents the eigenvector of the covariance matrix The row corresponding to eigenvector components, Represents the eigenvector of the covariance matrix The row corresponding to eigenvector components, is the total number of eigenvalues, represents the first Line Elements of a column; formula: ; The benefit of the formula is that it adjusts the differences and similarities between eigenvectors through weights, so that the model can be optimized according to the size of the eigenvalues ​​and the direction differences of the eigenvectors, which helps to more accurately describe the state changes of complex systems in mathematical modeling.

[0028] Detailed explanation of the formula and the process of formula calculation and derivation: Assume a specific example, the three eigenvalues ​​of the covariance matrix are 2, 0.5, and 0.1, and the corresponding eigenvectors are [1, 0, 0], [0, 1, 0], and [0, 0, 1]. Calculate ,Right now: ; This result shows that the transition probability between the first and second rows is 2.6, reflecting the transition strength and probability between the two states.

[0029] S214: Based on the transfer parameter matrix, analyze the probability transfer relationship of the elements in the matrix, calculate the conversion mode between hidden states, and generate the hidden state conversion result.

[0030] Call the transfer parameter matrix, analyze the probability transfer relationship of the elements in the matrix, and calculate the transition pattern between hidden states. This process includes parsing each matrix element to infer possible state transitions. Each element in the transfer parameter matrix represents the probability of going from one state to another. The size of the probability value can be used to determine the possibility of a state transferring to another state. This is crucial for predicting how the system transitions from the current state to the future state and generating hidden state transition results.

[0031] See also Figure 4 , the steps to obtain the optimized posterior state distribution matrix are as follows: S311: extracting a transfer parameter matrix from the hidden state conversion result, calculating the probability value of each state, traversing the transfer probability elements of each row in the matrix, accumulating the probability values ​​of the states corresponding to the fracture event records, and generating a superimposed state probability matrix; When calculating the state probability value, we must first extract the transfer parameter matrix from the hidden state conversion results. This process involves the precise reading and processing of the matrix data. By traversing each row and column of the matrix, the probability values ​​of the corresponding states are accumulated. On this basis, the probability values ​​of the corresponding states of each event in the fracture event record are superimposed. This operation needs to consider the actual frequency of fracture events and their corresponding state influences. Each fracture event needs to obtain its time point and occurrence state through database query, and then these data are matched and accumulated with the corresponding states in the transfer parameter matrix to generate a superimposed state probability matrix that integrates the fracture event data.

[0032] S312: Based on the state probability matrix, combined with the drug action intensity and frequency in the drug use data, the drug action intensity is used as a weight component and the frequency is used as an adjustment factor, and each state probability is weighted to generate a weighted state probability matrix; When combining the superimposed state probability matrix with drug usage data, the drug's intensity of action and frequency of use become key weight factors. The integration of drug data needs to be extracted from the medical record database, including drug name, dosage, administration frequency, etc. The mechanism of action of each drug and the corresponding pharmacological action intensity need to be quantitatively analyzed according to the drug instructions or medical research reports. The impact of these data on the state probability is reflected through weighted calculation, specifically by multiplying the drug's intensity of action as a weight by the probability value of each state, and considering the frequency of drug use as an adjustment factor, thereby generating a weighted state probability matrix.

[0033] S313: Based on the weighted state probability matrix, the probability value is normalized using the formula: ; Calculate the normalized state probability value and generate the optimized posterior state distribution matrix; in, is the drug effect intensity weight parameter, To adjust the factor weight parameters, Tuning parameters for normalized benchmarks, is the weighted state probability matrix Line The probability value of the column, is the adjustment factor corresponding to the frequency of drug use, is the total number of states, Normalized state probability value.

[0034] formula: ; The benefit of the formula is that by introducing the intensity and frequency of drug use as adjustment factors, it can more accurately reflect the impact of drug treatment on the patient's state transition and make the state distribution more consistent with the actual situation.

[0035] Detailed explanation of the formula and the process of formula calculation and derivation: Set a specific value. , , , assuming that for a specific state, the frequency of drug use , the original state probability , put it into the formula: ; Assumptions And the original probability of all states is the same as the drug frequency, and the calculation is: ; The results show that the optimized posterior state distribution matrix provides a more accurate state probability value for each state after adjustment for drug effects, reflecting the actual effect of drug treatment.

[0036] See also Figure 5 , the steps for obtaining the key dynamic feature set are as follows: S411: extracting state probability values ​​of bone density change rate and bone metabolism index change rate from the optimized posterior state distribution matrix, filtering the state probability values ​​according to the screening conditions, removing probability values ​​below the threshold, retaining the corresponding change rates, and generating a screened state probability value set; First, it is necessary to determine the threshold of the state probability value, which is based on the statistical analysis of previous data, such as calculating the mean or median of all state probabilities to set the threshold. For each state probability value, if it is greater than this threshold, it will be retained, otherwise it will be eliminated. The retained state probability value will be used for subsequent weighted and cumulative calculations. The key to this step is to optimize the quality and availability of the data through threshold screening to ensure the accuracy and representativeness of the data in subsequent steps.

[0037] S412: based on the screened state probability value set, grouping and accumulating the bone density change rate and the bone metabolism index change rate according to the weight of the state probability value, calculating the accumulated change rate weighted value based on the state probability value, and generating a cumulative state probability change rate matrix; The calculation of the cumulative state probability change rate matrix includes associating the screened state probability values ​​with the corresponding change rates. This requires multiplying each retained state probability value with its corresponding bone density change rate and bone metabolism index change rate, and then accumulating all the product results to obtain the weighted sum of each change rate. This calculation not only relies on accurate data input, but also needs to ensure the calculation accuracy during the multiplication and accumulation process to reflect the comprehensive influence of each change rate under different states. This method provides a basis for further data processing and feature extraction through quantitative analysis.

[0038] S413: Based on the cumulative state probability change rate matrix, calculate the cumulative value of each dynamic feature using the formula: ; Generate a set of key dynamic features; in, Indicates The cumulative value of dynamic features, is the cumulative state probability change rate matrix The probability value of a state, For the Under the state The rate of change of a dynamic feature, is the deviation value between the dynamic feature and its mean value in the state, is the mean of the feature, is the total number of states.

[0039] formula: ; The usefulness of the formula is that it not only takes into account the probability of each state , and through the rate of change of the combined state and the deviation of the rate of change from its mean , which enhances the sensitivity to data variability and enables the formula to more carefully capture the dynamic changes of each feature under different states.

[0040] Detailed explanation of the formula and the process of formula calculation and derivation: The specific parameters are set as follows: Assume there are three states, namely , state probability value The change rates are 0.2, 0.5, and 0.3 respectively. They are 0.1, -0.2, 0.15 respectively, and the rate of change deviation They are 0.05, 0.1, -0.05 respectively, and the mean Set to 0.1. The calculation process is: ; ; ; ; The results show that considering the combined effect of probability weighting and feature change rate, the cumulative value of the key dynamic feature set is 0.162, which reflects the overall dynamic change trend of the feature under a given state probability and change rate.

[0041] See also Figure 6 , the specific steps for obtaining the feature evolution prediction value are: S511: extracting bone density change trend and fracture event prediction data from the key dynamic feature set, calculating state probability values ​​and time intervals corresponding to multiple time points based on the extracted trend data, and establishing a mapping matrix between state probability and time interval; The operation of extracting bone density change trend and fracture event prediction data from the key dynamic feature set begins with the collection and preprocessing of dynamic data. The program automatically screens the patient monitoring data of the past six months, and then uses the data cleaning algorithm to remove outliers and missing values ​​to ensure the integrity and accuracy of the data. The data after preliminary screening is subjected to preliminary trend analysis through advanced statistical models, and the time series analysis method is used to estimate the change trend of bone density, providing a basis for subsequent in-depth analysis. Through this series of operations, a mapping matrix containing time-stamped state probabilities and time intervals is established.

[0042] S512: Based on the state probability and time interval mapping matrix, the state probability values ​​in the matrix are accumulated, and the time intervals are weighted accumulated to generate a cumulative prediction probability time interval matrix; Calling the state probability and time interval mapping matrix, the specific operation involves importing the data in the mapping matrix into the statistical software, using the built-in cumulative function of the software to accumulate the state probability item by item, and applying the weighted accumulation algorithm to the time interval data to ensure that the weight of the time interval properly reflects its impact on the prediction model. This calculation process involves multiple data conversion and algorithm adjustment steps to ensure the accuracy of the data and the effectiveness of the operation. The cumulative prediction probability time interval matrix generated by this process provides key numerical input for the next trend analysis.

[0043] S513: Based on the cumulative prediction probability time interval matrix, calculate the normalized trend curve at each time point using the formula: ; generating feature evolution predictions; in, Indicates The characteristic evolution value at a time point, For the The dynamic features are The state probability value of the time interval is For the time interval, For the The dynamic features are The rate of change at a time point, For the The characteristic adjustment factor for the time interval, is the total number of time intervals.

[0044] formula: ; The benefit of the formula is that by combining the state probability, time interval and change rate adjustment, the trend prediction is made more accurate and can reflect the complex relationship between the bone density change trend and the predicted data in real time.

[0045] Detailed explanation of the formula and the process of formula calculation: Assumptions , sky, The rate of change, , the specific calculation process is as follows: ; ; ; This result shows that the calculated The value is 0.12, which means that after considering the time factor and the dynamic change rate, the predicted time point The bone density change trend indicator can be directly used to assess the risk of future fractures.

[0046] A medical data analysis system for postmenopausal osteoporosis patients, the medical data analysis system for postmenopausal osteoporosis patients is used to execute the above-mentioned medical data analysis method for postmenopausal osteoporosis patients, and the system includes: The bone density dynamic analysis module extracts bone density measurement values ​​and bone metabolism index concentrations based on bone density data, bone metabolism index data and fracture event records, calculates the bone density change and bone metabolism index change values ​​between time points, counts the number of fracture events, and organizes them into a dynamic feature time series matrix; The hidden state feature extraction module extracts the bone density change and bone metabolism index change value based on the dynamic feature time series matrix, calculates the covariance matrix between the change values, extracts the eigenvalue and eigenvector of the covariance, establishes the transfer parameter matrix, and obtains the hidden state conversion parameters; The state probability distribution calculation module calculates the state probability value in the transfer parameter based on the hidden state transition parameter, combines the fracture event record and the drug use data, superimposes the fracture event record and the drug use data with the state probability value, and forms the posterior state distribution matrix after normalization and weighted calculation; The characteristic evolution prediction module extracts the state probability values ​​of the bone density change rate and the bone metabolism index change rate based on the posterior state distribution matrix, screens and accumulates the probability values, calculates the cumulative values ​​of the bone density change trend and fracture event prediction data, normalizes them to form a trend curve, and obtains the trend evolution characteristic value.

[0047] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A method for analyzing medical data of postmenopausal osteoporosis patients, characterized in that: The following steps are involved: Based on the patient's bone density data, bone metabolism index data and fracture event records, the bone density measurement values ​​and bone metabolism index concentrations are extracted, the changes between time points are calculated, the number of fracture events is counted, and a dynamic feature time series matrix is ​​generated; Based on the dynamic feature time series matrix, the bone density change and the bone metabolism index change value are extracted, the covariance matrix between the change values ​​is calculated, the eigenvalue and eigenvector of the covariance are extracted, the transfer parameter matrix is ​​established, and the hidden state conversion result is generated; Based on the hidden state conversion result, the state probability value in the transfer parameter is calculated, the fracture event record and the drug use data are combined, the state probability is superimposed, normalized and weighted, and an optimized posterior state distribution matrix is ​​generated; Based on the optimized posterior state distribution matrix, the state probability values ​​of the bone density change rate and the bone metabolism index change rate are extracted, the probability values ​​are screened and accumulated, the cumulative values ​​of the influencing features are calculated, and a key dynamic feature set is generated; Based on the key dynamic feature set, the bone density change trend and fracture event prediction data are extracted, the cumulative value of the prediction probability and time interval is calculated, the trend curve is normalized to generate the feature evolution prediction value.

2. The medical data analysis method for postmenopausal osteoporosis patients according to claim 1, characterized in that: The dynamic feature time series matrix includes bone density measurement values, bone metabolism index concentrations, bone density changes, bone metabolism index changes, and the number of fracture events. The hidden state conversion results include bone density change eigenvalues, bone metabolism index change value eigenvalues, covariance eigenvalues, covariance eigenvectors, and transfer parameter matrices. The optimized posterior state distribution matrix includes state probability values, fracture event records, drug use data, normalized probability values, and weighted calculation values. The key dynamic feature set includes bone density change rate, bone metabolism index change rate, state probability value, cumulative calculation results, and impact feature cumulative values. The feature evolution prediction value includes bone density change trend, fracture event prediction data, prediction probability, time interval cumulative value, and trend curve.

3. The medical data analysis method for postmenopausal osteoporosis patients according to claim 2, characterized in that: The steps of obtaining the dynamic feature time series matrix are specifically as follows: Extract each bone density measurement value and bone metabolism index concentration from the patient's data record, calculate the change between two consecutive measurements, and obtain the bone density change and bone metabolism change; Record and count the number of fractures of each patient, and combine the bone density change and bone metabolism change to form time series characteristic data; Integrate the time series feature data and use the formula: ； The dynamic characteristic time series matrix of each patient is calculated; in, Representing patients In the The dynamic characteristic matrix value of the measurement, Representative The patient in The change in bone density measured each time, Representative The patient in The changes in bone metabolism indicators measured by Representative Number of fractures per patient.

4. The medical data analysis method for postmenopausal osteoporosis patients according to claim 3, characterized in that: The steps for obtaining the hidden state conversion result are specifically as follows: Extracting the bone density change and the bone metabolism index change value from the dynamic feature time series matrix, calculating the covariance matrix based on the change value, and performing covariance component calculation on the bone density change and the bone metabolism index change to generate a covariance matrix; Performing eigendecomposition on the covariance matrix to extract eigenvalues ​​and eigenvectors of the covariance matrix, and simultaneously calculating a plurality of eigenvalues ​​and corresponding eigenvectors based on the distribution characteristics of the elements of the covariance matrix to obtain an eigenvalue matrix and an eigenvector matrix; Based on the eigenvalue matrix and eigenvector matrix, the formula is adopted: ； Combine the weight relationship between the elements of the matrix to calculate and generate the transfer parameter matrix; in, Indicates eigenvalues, Represents the eigenvector of the covariance matrix The row corresponding to eigenvector components, Represents the eigenvector of the covariance matrix The row corresponding to eigenvector components, is the total number of eigenvalues, represents the first Line Elements of a column; Based on the transfer parameter matrix, the probability transfer relationship of the elements in the matrix is ​​analyzed, the conversion mode between the hidden states is calculated, and the hidden state conversion result is generated.

5. The medical data analysis method for postmenopausal osteoporosis patients according to claim 4, characterized in that: The steps for obtaining the optimized posterior state distribution matrix are specifically as follows: Extracting a transfer parameter matrix from the hidden state conversion result, calculating the probability value of each state, traversing the transfer probability elements of each row in the matrix, accumulating the probability values ​​of the states corresponding to the fracture event records, and generating a superimposed state probability matrix; Based on the state probability matrix, combined with the drug action intensity and frequency in the drug use data, the drug action intensity is used as a weight component and the frequency is used as an adjustment factor, and each state probability is weighted to generate a weighted state probability matrix; Based on the weighted state probability matrix, the probability value is normalized using the formula: ； Calculate the normalized state probability value and generate the optimized posterior state distribution matrix; in, is the drug effect intensity weight parameter, To adjust the factor weight parameters, Tuning parameters for normalized benchmarks, is the weighted state probability matrix Line The probability value of the column, is the adjustment factor corresponding to the frequency of drug use, is the total number of states, Normalized state probability value.

6. The method for analyzing medical data of postmenopausal osteoporosis patients according to claim 5, characterized in that: The steps of obtaining the key dynamic feature set are specifically as follows: Extracting state probability values ​​of bone density change rate and bone metabolism index change rate from the optimized posterior state distribution matrix, filtering the state probability values ​​according to the screening conditions, removing probability values ​​below the threshold, retaining the corresponding change rates, and generating a screened state probability value set; Based on the screened state probability value set, the bone density change rate and the bone metabolism index change rate are grouped and accumulated according to the weights of the state probability values, and the accumulated change rate weighted values ​​are calculated based on the state probability values ​​to generate a cumulative state probability change rate matrix; Based on the cumulative state probability change rate matrix, the cumulative value of each dynamic feature is calculated using the formula: ； Generate a set of key dynamic features; in, Indicates The cumulative value of dynamic features, is the cumulative state probability change rate matrix The probability value of a state, For the Under the state The rate of change of a dynamic feature, is the deviation value between the dynamic feature and its mean value in the state, is the mean of the feature, is the total number of states.

7. The medical data analysis method for postmenopausal osteoporosis patients according to claim 6, characterized in that: The steps for obtaining the feature evolution prediction value are specifically as follows: Extracting bone density change trend and fracture event prediction data from the key dynamic feature set, calculating state probability values ​​and time intervals corresponding to multiple time points based on the extracted trend data, and establishing a mapping matrix between state probability and time interval; Based on the state probability and time interval mapping matrix, the state probability values ​​in the matrix are accumulated, and the time intervals are weightedly accumulated to generate a cumulative prediction probability time interval matrix; Based on the cumulative prediction probability time interval matrix, the normalized trend curve at each time point is calculated using the formula: ； generating feature evolution predictions; in, Indicates The characteristic evolution value at a time point, For the The dynamic features are The state probability value of the time interval is For the time interval, For the The dynamic features are The rate of change at a time point, For the The characteristic adjustment factor for the time interval, is the total number of time intervals.

8. A medical data analysis system for postmenopausal osteoporosis patients, characterized in that: According to the method for analyzing medical data of postmenopausal osteoporosis patients according to any one of claims 1 to 7, the system comprises: The bone density dynamic analysis module extracts bone density measurement values ​​and bone metabolism index concentrations based on bone density data, bone metabolism index data and fracture event records, calculates the bone density change and bone metabolism index change values ​​between time points, counts the number of fracture events, and organizes them into a dynamic feature time series matrix; The hidden state feature extraction module extracts the bone density change and the bone metabolism index change value based on the dynamic feature time series matrix, calculates the covariance matrix between the change values, extracts the eigenvalue and eigenvector of the covariance, establishes a transfer parameter matrix, and obtains the hidden state conversion parameters; The state probability distribution calculation module calculates the state probability value in the transfer parameter based on the hidden state transition parameter, combines the fracture event record and the drug use data, superimposes the fracture event record and the drug use data with the state probability value, and forms a posterior state distribution matrix after normalization and weighted calculation; The characteristic evolution prediction module extracts the state probability values ​​of the bone density change rate and the bone metabolism index change rate based on the posterior state distribution matrix, screens and accumulates the probability values, calculates the cumulative value of the bone density change trend and the fracture event prediction data, normalizes them to form a trend curve, and obtains the trend evolution characteristic value.

Images