Health assessment method based on multi-dimensional physical examination data clustering analysis

Abstract

The application discloses a kind of multi-dimension physical examination data clustering health assessment methods, it is related to medical health data analysis technical field.The steps include: acquisition and pre-processing physical examination / wearable data (missing value processing, unit uniform, standardization);Determine each index weight based on entropy-stability fusion strategy;Around health baseline M k Construct direction-sensitive weighted distance measure;Improved fuzzy C mean iterative solution cluster center and membership, get cluster level deviation and individual risk score, and generate structured report.This method is more sensitive to the same abnormality of key indicators while ensuring robustness, improves clustering separation and risk interpretability, and is suitable for physical examination center, hospital information system and wearable data intelligent health assessment.

Description

Technical Field

[0001] This invention relates to the field of medical and health data analysis technology, specifically to a health assessment method based on multi-dimensional physical examination data clustering analysis. Background Technology

[0002] Current health assessments primarily rely on whether individual indicators exceed normal ranges or on simple weighted summation. For example, doctors typically assess an individual's health status based on whether key indicators such as blood pressure and blood sugar are abnormal. However, relying solely on thresholds for a few indicators may not comprehensively reflect an individual's overall health. When multiple physiological and biochemical indicators are simultaneously near critical values, traditional methods often fail to detect potential health risks in a timely manner. Furthermore, different indicators have varying impacts on health, and existing assessment methods lack dynamic measurement of the importance of each indicator, potentially leading to important abnormalities being buried under data noise. In conclusion, there is an urgent need for a technological solution that can fully utilize the internal correlations of multi-dimensional health examination data, differentiate the importance of different indicators, and automatically identify health risk patterns to more comprehensively and accurately assess an individual's health status. Summary of the Invention

[0003] Technical Objective: To address the problems of incomplete single-indicator assessment and difficulty in timely detection of risks associated with multiple indicator combinations in existing technologies, this invention discloses a health assessment method based on multi-dimensional physical examination data clustering analysis. Through an improved clustering analysis algorithm, unsupervised pattern recognition is performed on multiple physical examination indicator data of the examinee, automatically classifying individuals with similar health characteristics, and assessing the health risk of each group based on the clustering results.

[0004] Technical solution: To achieve the above technical objectives, the present invention adopts the following technical solution:

[0005] A health assessment method based on multi-dimensional physical examination data clustering analysis, characterized by the following steps:

[0006] S1. Obtain multi-dimensional physical examination data of the examinee, which includes multiple different categories of health indicator values;

[0007] S2. Perform data cleaning and standardization on the physical examination data to obtain complete and consistent standardized health data;

[0008] S3. Calculate the weight coefficient of each health indicator based on the variation of each health indicator value in standardized health data. The weight coefficient reflects the relative importance of each indicator to health differences.

[0009] S4. Based on standardized health data and weight coefficients, a clustering algorithm is used to divide the examined objects into several clusters with similar health statuses, and each cluster corresponds to a health status category.

[0010] S5. Analyze the central characteristics of each cluster, assess the health status level or health risk corresponding to the cluster, and output the corresponding health assessment results based on the cluster to which the tested object belongs.

[0011] Preferably, the weighting coefficients of each health indicator are calculated using the information entropy method, specifically including: normalizing the value of each health indicator to a percentage form, and calculating the information entropy of each health indicator. and according to Determine the weight of each health indicator, among which Furthermore, the sum of the weights of all health indicators is 1.

[0012] Preferably, step S4 employs a fuzzy C-means clustering algorithm, including the following processes: calculating the distance between the tested object and the center of each cluster based on a weighted distance metric, and iteratively updating the membership degree of the tested object to each cluster. The weighted distance metric is calculated by summing the center vectors of each cluster until the cluster centers converge; where the weighted distance metric is the distance between the examined object i and cluster j. , Let be the standardized value of the i-th subject on the k-th health indicator. Let the value of cluster j be the center of health index k. Let be the weighting coefficient of the k-th health indicator, and p be the total number of health indicators. The direction-sensitive penalty coefficient, , This is a deviation-indicator function. , This is the health reference baseline or target value for health indicator k.

[0013] Preferably, in the fuzzy C-means clustering algorithm, the membership degree of the examined object to the j-th cluster is... Calculate using the following formula:

[0014] ,

[0015] in For the examined object i and the cluster The weighted distance between them, where c is the number of clusters and m is the fuzziness factor; the center vector of each cluster is updated according to the following formula:

[0016] ,

[0017] in For clusters The central value of the k-th health indicator, where n is the total number of subjects examined.

[0018] Preferably, step S5 includes: calculating a comprehensive health deviation or risk score for each cluster, and dividing the clusters into different health level categories according to a preset threshold; obtaining the membership distribution of each subject to each cluster, classifying the subject to the corresponding health level, and outputting a health assessment report, wherein the health assessment report includes the overall health status level of the subject and improvement suggestions for abnormal indicators.

[0019] A health assessment system for dimensional physical examination data cluster analysis, used to perform a health assessment method for dimensional physical examination data cluster analysis as described above, includes: a data acquisition module, a data processing module, a cluster analysis module, and an assessment output module;

[0020] The data acquisition module is used to receive original physical examination records containing the unique identifier and timestamp of the examinee from physical examination equipment, hospital information system or wearable terminal, perform field integrity verification, unit of measurement uniformity and source marking on the records, and output a structured record stream with examinee identifier, timestamp, indicator code, value and unit;

[0021] The data processing module is used to perform missing value imputation, outlier suppression, unit conversion and standardization on the record stream, generate a standardized data matrix and supporting statistical parameters for clustering, and generate a parameter set for distance measurement and risk assessment.

[0022] The clustering analysis module is used to perform fuzzy clustering iteration on standardized data based on index weights and weighted distance metrics, solve the cluster centers and the membership degree of the examined objects to each cluster until the convergence condition is met, and output the cluster center vector and membership degree distribution.

[0023] The assessment output module is used to calculate cluster-level health deviation based on cluster centers and preset health references, and map it to cluster-level risk scores. It calculates individual risk scores and health levels based on the membership of the examined subjects, and generates structured health assessment results that include overall level, dominant deviation indicators and intervention recommendations, and outputs them externally.

[0024] Beneficial Effects: The health assessment method based on multi-dimensional physical examination data clustering analysis provided by this invention has the following beneficial effects:

[0025] This invention, by integrating data clustering and health risk assessment techniques, fully uncovers the implicit correlation patterns in routine physical examination data. Compared to existing assessment methods that rely solely on single indicators or linear scoring, this method can automatically identify anomalous combinations of related indicators, enabling the identification and classification of early health risks. In particular, by employing entropy weighting and balanced clustering, it avoids situations where certain indicators dominate distance calculations and mask other risk factors, resulting in more objective and comprehensive assessment results. Furthermore, this invention introduces the concept of partial membership using fuzzy clustering, allowing each examinee to simultaneously belong to multiple health categories with certain weights, thus providing accurate assessments even for individuals on the verge of sub-health. This method improves the accuracy of health assessments while exhibiting good robustness and adaptability, meeting the assessment needs of personalized medicine and group health management. Attached Figure Description

[0026] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below.

[0027] Figure 1 This is a flowchart of the health assessment method of the present invention;

[0028] Figure 2 This is a structural block diagram of the health assessment system of the present invention;

[0029] Figure 3 A schematic diagram illustrating the integration of distance measurement and direction sensitivity;

[0030] Figure 4 This is a flowchart of the fuzzy C-means iteration process. Detailed Implementation

[0031] The present invention will now be described more clearly and completely by way of a preferred embodiment in conjunction with the accompanying drawings, but this does not limit the invention to the scope of the described embodiment.

[0032] like Figure 1 As shown, a health assessment method based on multi-dimensional physical examination data clustering analysis includes the following steps:

[0033] S1. Obtain multi-dimensional physical examination data of the examinee, which includes multiple different categories of health indicator values;

[0034] S2. Perform data cleaning and standardization on the physical examination data to obtain complete and consistent standardized health data;

[0035] S3. Calculate the weight coefficient of each health indicator based on the variation of each health indicator value in standardized health data. The weight coefficient reflects the relative importance of each indicator to health differences.

[0036] S4. Based on standardized health data and weight coefficients, a clustering algorithm is used to divide the examined objects into several clusters with similar health statuses, and each cluster corresponds to a health status category.

[0037] S5. Analyze the central characteristics of each cluster, assess the health status level or health risk corresponding to the cluster, and output the corresponding health assessment results based on the cluster to which the tested object belongs.

[0038] like Figure 2 As shown, a health assessment system based on dimensional physical examination data clustering analysis is used to perform a health assessment method based on dimensional physical examination data clustering analysis as described above, including: a data acquisition module, a data processing module, a clustering analysis module, and an assessment output module;

[0039] The data acquisition module is used to receive original physical examination records containing the unique identifier and timestamp of the examinee from physical examination equipment, hospital information system or wearable terminal, perform field integrity verification, unit of measurement uniformity and source marking on the records, and output a structured record stream with examinee identifier, timestamp, indicator code, value and unit;

[0040] The data processing module is used to perform missing value imputation, outlier suppression, unit conversion and standardization on the record stream, generate a standardized data matrix and supporting statistical parameters for clustering, and generate a parameter set for distance measurement and risk assessment.

[0041] The clustering analysis module is used to perform fuzzy clustering iteration on standardized data based on index weights and weighted distance metrics, solve the cluster centers and the membership degree of the examined objects to each cluster until the convergence condition is met, and output the cluster center vector and membership degree distribution.

[0042] The assessment output module is used to calculate cluster-level health deviation based on cluster centers and preset health references, and map it to cluster-level risk scores. It calculates individual risk scores and health levels based on the membership of the examined subjects, and generates structured health assessment results that include overall level, dominant deviation indicators and intervention recommendations, and outputs them externally.

[0043] Example

[0044] In this embodiment, the health assessment method includes the following steps:

[0045] Step S1: Data Collection. Acquire multi-dimensional physical examination data from multiple subjects. This data includes basic physiological indicators, clinical laboratory indicators, and lifestyle-related indicators. For example, physiological indicators may include height, weight, blood pressure, and heart rate; clinical laboratory indicators may include fasting blood glucose, glycated hemoglobin, lipid profiles, and liver and kidney function indicators; lifestyle and health behavior data may include daily exercise volume, sleep duration, and frequency of smoking and drinking. This data is collected through hospital examination equipment, wearable devices, or questionnaires, and then aggregated to form the original health dataset for each subject. Preferably, a timestamp and subject ID are added to each data point during data collection to facilitate subsequent longitudinal tracking analysis of individuals and cross-group comparisons.

[0046] Step S2: Data Preprocessing. The collected raw physical examination data is preprocessed to obtain a complete, standardized dataset suitable for analysis. First, missing data is imputed, for example, by using mean imputation or interpolation based on similar individuals to fill in missing indicator values. Second, outliers and noisy data are detected and processed. If necessary, the 3σ principle or density-based algorithms can be used to identify extreme outliers, marking them as anomalous or replacing them with reasonable values ​​to prevent extreme values ​​from interfering with subsequent analysis. Then, the data from each dimension undergoes a unified format conversion and dimensional standardization. Preferably, [min-max] normalization or Z-score standardization methods are used to convert different indicators to a comparable numerical range, for example, for numerical indicators. Calculate standardized values or ,in and denoted as the mean and standard deviation of the k-th indicator in the sample data, respectively. This preprocessing ensures that all physical examination indicator data are clustered on the same scale and reduces the impact of data quality issues.

[0047] Step S3: Improved Cluster Analysis. Cluster analysis is performed on the preprocessed multi-dimensional physical examination data to uncover potential health category patterns among the examinees. Unlike traditional clustering, this invention introduces indicator weights, adaptive distance metrics, and fuzzy classification mechanisms during the clustering process, making the clustering results more aligned with health assessment needs. Specifically, it includes the following sub-steps:

[0048] S31. Calculation of Indicator Weights: Based on the contribution of each physical examination indicator to the differentiation of health status, calculate the weight coefficient of each indicator. This embodiment preferably uses the information entropy method (entropy weight method) to calculate the weights. First, the distribution of the k-th indicator among all n tested objects is calculated using the preprocessed data: Let... Let k be the standardized value of the k-th indicator for the i-th subject. Then calculate its weight under that indicator. :

[0049]

[0050] Where n is the total number of subjects examined. Next, the information entropy value of the k-th indicator is calculated. :

[0051]

[0052] in Entropy A smaller value indicates a higher degree of information aggregation and dispersion, a greater degree of discrimination, and a greater contribution to overall health differences. Therefore, the utility value of the k-th indicator is defined as follows: And calculate the index weights accordingly. :

[0053]

[0054] Where p is the total number of indicators, and has Through the entropy weight calculation described above, each indicator will obtain a weight value between 0 and 1. This indicator is used to characterize its importance relative to other indicators. For example, if an indicator shows large and significant variations in value within the sample (such as obvious abnormalities in a portion of the population), then its importance is considered significant. Smaller A larger weight indicates that the indicator is more crucial for classifying health categories; conversely, a smaller weight indicates a lower weight. It should be noted that, in addition to the entropy weight method, this invention can also adjust the weights of each indicator based on the experience of medical experts or statistical methods. For example, higher weights can be assigned to indicators related to the risk of specific diseases. The choice of weight calculation method does not affect the implementation effect of this invention.

[0055] S32. Distance metric definition: Based on the index weights obtained above, a multidimensional weighted distance from a sample to the cluster center is defined for cluster analysis. Traditional clustering typically uses Euclidean distance to calculate differences between individuals, but does not consider the differences in importance of different indicators. This invention improves the distance formula by introducing weighting coefficients, making key health indicators contribute more to the distance and improving the sensitivity of clustering to differences in health status. Preferably, this embodiment uses weighted Euclidean distance as the similarity measurement function, which calculates the distance between the i-th examined object and the j-th cluster in the p-dimensional index space. as follows:

[0056]

[0057] in Let be the standardized value of the i-th subject on the k-th health indicator. Let the value of cluster j be the center of health index k. Let be the weighting coefficient of the k-th health indicator, and p be the total number of health indicators. The direction-sensitive penalty coefficient, , This is a deviation-indicator function. , This serves as a health reference baseline or target value for health indicator k. Of course, those skilled in the art can choose other weighted distance metrics as needed, such as weighted Manhattan distance, Mahalanobis distance, etc., as long as they meet the principle of highlighting the differences in important indicators, they all fall within the scope of protection of this invention.

[0058] S33. Clustering Algorithm Execution: An improved fuzzy clustering algorithm is used to cluster the data of all subjects, automatically grouping the samples into several categories with similar health states. Preferably, this embodiment uses the fuzzy C-means (FCM) clustering algorithm with index weights. Compared with traditional hard clustering (such as K-means), fuzzy clustering allows a sample to belong to multiple clusters in the form of membership degrees, which is very suitable for handling marginal states in health assessment. For example, a subject may have both mild dyslipidemia and mild hypertension, and their health state is between "normal" and "pre-metabolic syndrome." Fuzzy clustering can more accurately reflect this situation. Figure 4 As shown, the clustering process is as follows: First, determine the number of clusters c (the optimal number of clusters can be determined based on prior knowledge or by using indicators such as the silhouette coefficient); then, randomly initialize c cluster center vectors v1, v2, ..., v c The sample membership and cluster centers are then iteratively updated until convergence. The update formula for the fuzzy C-means is as follows:

[0059] Membership update: For the i-th examined object and the j-th cluster, calculate its membership degree u. ij (Values ​​range from 0 to 1) are:

[0060]

[0061] in The examined object i and the cluster defined in step S32 The weighted distance between them, where c is the number of clusters, and m > 1 is the fuzziness factor (usually taken as 2); the above formula ensures that the closer the inspected object is to the center of a cluster, the higher its membership degree u to that cluster. ij The larger the membership degree, the lower the membership degree; conversely, if the distance is large, the membership degree approaches 0. Fuzzy membership degree can quantify the degree to which each subject belongs to each health category.

[0062] Cluster center update: Calculate new cluster centers As the average of the weighted membership degrees of each indicator:

[0063]

[0064] in For clusters The cluster center is the central value of the k-th health indicator, where n is the total number of subjects examined. The above formula indicates that the cluster center is the fuzzy membership degree u of all samples within that cluster. ij The weighted average is calculated. The fuzzy factor m is usually the same value as the membership formula, for example, m=2 to increase the smoothness of the clustering results. The membership degree u is calculated iteratively. ij and cluster center v j This continues until the cluster center change in adjacent iterations is less than a preset threshold (e.g., ...). , When the threshold value is set to 0, the clustering process is considered to have converged. Finally, c clusters are obtained, along with the membership degree of each examined object to each cluster.

[0065] After the clustering algorithm is executed, each cluster represents a group of people with similar health characteristics. Through the improved clustering analysis process described above, this invention can fully consider the importance of different indicators and allow for overlapping health states, thereby obtaining more reasonable clustering results. For example, in one embodiment, analyzing physical examination data of middle-aged people may yield clusters corresponding to: "stable health cluster" (all indicators are normal and balanced), "high blood pressure cluster" (significantly high blood pressure accompanied by mild overweight), "abnormal blood sugar and lipid cluster" (high blood sugar and cholesterol), etc. These clustering results provide a foundation for subsequent health assessments.

[0066] Figure 3 This diagram visually illustrates the weighted and orientation-sensitive sample-cluster center distance metric and its structure used in the clustering stage of this invention. In the figure, the object under examination, i, is represented by "sample i," and the cluster center is represented by "cluster center." "This indicates that among them" For the first The center vector of each cluster in the normalized feature space; the distance from the sample to the cluster center is denoted as... This is subsequently used for membership calculation and iterative convergence. To highlight the difference from traditional Euclidean distance, the figure also shows component weighting and the relative healthy baseline. Direction determination.

[0067] Figure 3 (a) The distance from sample i to the cluster center is represented by two bounding boxes and a middle arrow. The synthesized distance vector, where the arrow represents the geometric direction from the sample point to the cluster center. The corresponding metric is:

[0068]

[0069] This figure is intended to illustrate that when moving away from the healthy baseline in the same direction, the contribution of distance to this dimension is amplified, thereby enhancing sensitivity to anomalies in the same direction.

[0070] Figure 3 (b) Decompose the difference in the two-dimensional case into two dashed line components in the horizontal and vertical directions. The dashed lines represent the differences in the example two dimensions (denoted as...). and component difference on ) and The figures are marked outside the dashed lines. and This is used to visually demonstrate why first scaling by taking the square root of the weights, and then summing the squared distances, can give key indicators a higher discriminative power: when When the weighted distance is large, the influence of the corresponding component is amplified; conversely, it is suppressed. This figure emphasizes the interpretability of the weighted distance and its adaptive representation of the importance of multiple indicators.

[0071] Figure 3 (c) Increase health baseline The reference points are marked with triangles, and arrows pointing from the baseline to the cluster center and the sample respectively indicate the direction of their deviation from the baseline. When the sample is relative to the cluster center... When the deviation is in the same direction, it is indicated in the figure. When the opposite direction or one side is at the baseline, This design makes It not only quantifies the distance from the cluster center, but also provides differentiated amplification for whether it deviates in a direction with more adverse health risks, reflecting the improvement of the embedding of health semantics and the interpretability of clustering in this invention.

[0072] Step S4: Clustering Result Analysis. Based on the characteristics of each cluster obtained from clustering, the health status of different clusters is assessed and classified. Specifically, this includes: calculating the difference between the center of each cluster and the medical reference health value to determine the health type of the cluster; and further calculating the health risk score for each cluster. Preferably, historical healthy population data or authoritative medical indicators are first selected as a reference baseline to obtain the ideal normal value or range for each indicator. For example, for a general adult, systolic blood pressure of 120 mmHg, fasting blood glucose of 5.5 mmol / L, and BMI of 22 can be taken as ideal health reference values, denoted as... Then, the weighted distance D between the center of the j-th cluster and the reference health vector M is calculated. j This is used to quantify the overall health deviation of the cluster.

[0073]

[0074] Where v jk Let M be the value of the j-th cluster center at the k-th index. kis the ideal health reference value for this indicator, R k is the tolerable normal fluctuation range of this indicator (for example, half of the difference between the medically defined upper and lower normal limits, or determined based on the population standard deviation), w k is the weight coefficient of the k-th indicator. This formula comprehensively considers the deviation degree and importance of each indicator at the cluster center, and calculates the health deviation degree D of cluster j j . D j A larger value indicates that cluster j has a more severe overall deviation from the health baseline across multiple indicators, and the health risk is higher; conversely, a smaller D j means the cluster as a whole is closer to a healthy state.

[0075] Based on the health deviation degree D of each cluster j and the deviation of specific indicators, the clusters can be named and their health risk levels can be assessed. For example, if the D of a certain cluster j is much higher than that of other clusters, and this is mainly caused by the large deviations of the two indicators of blood pressure and blood glucose from the normal values, this cluster can be defined as a "hypertension combined with hyperglycemia high-risk cluster" with a risk level of high risk; on the contrary, if the D of a certain cluster j is close to 0 and all indicators are close to normal values, it is a "health benchmark cluster" with a risk level of low risk. In addition, characteristic indicators can be extracted for each cluster: for example, the common abnormalities of most members in the cluster are used as the markers of the cluster. This kind of analysis helps to convert the clustering results into conclusions with medical significance.

[0076] In this embodiment, the clustering results are divided into three levels of health states: healthy, sub-healthy, and high-risk. By comparing the D of each cluster j with the preset thresholds, the state category of each cluster can be determined. For example, thresholds T1<T2 can be set. If D j <T1, cluster j is determined as a healthy category; if D j >T2, it is determined as a high-risk category; the values in between belong to the sub-healthy category. The thresholds can be determined according to actual population data statistics, so that the assessment results correspond to the clinical risk levels. Certainly, the criteria and quantity for dividing categories are not fixed, and the present invention can be flexibly adjusted according to application scenarios, for example, to subdivide more categories (such as mild risk, moderate risk, severe risk, etc.).

[0077] Step S5: Health assessment output. Based on cluster analysis and result assessment, a health assessment report for each tested subject is generated. In this step, the individual is associated with the health state of the cluster to which the individual belongs, so as to obtain the individual's comprehensive health evaluation. For example, in this embodiment, for each tested subject i, the membership degree u of the subject in each cluster is obtained ij (where j=1,...,c). The personalized health risk score of the tested subject i can be calculated based on the membership degree and the risk level of each cluster :

[0078]

[0079] in Risk score for a predefined cluster j (e.g., setting a healthy cluster) =0, high-risk clusters are set relatively large. The sub-health cluster has values ​​between these two. Because fuzzy clustering allows... The non-zero values ​​of and the weighted summation in the formula reflect the comprehensive risk level of an individual if they belong to multiple categories simultaneously. For example, if an individual i has a membership degree of 0.6 to the healthy cluster and 0.4 to the sub-healthy cluster, then if we define R=0 for the healthy cluster and R=50 for the sub-healthy cluster (with 100 representing the highest risk), their individual risk score would be... This falls within the low-risk range. Accordingly, we can... The range in which a person falls will be classified as "healthy (green code)," "sub-healthy (yellow code)," or "high-risk (red code)," and will be clearly marked in the report.

[0080] In its health assessment report, this invention not only provides the overall health level or risk score of the examinee, but also offers detailed multi-dimensional analytical information. For example, the report can list the health cluster category closest to the examinee, such as "You belong to the high blood pressure cluster, characterized by blood pressure higher than normal, while other indicators are basically normal." The report can also indicate the indicator with the greatest deviation from ideal health and improvement suggestions, such as "Blood pressure is about 15% higher than normal; it is recommended to control salt intake and maintain exercise." For individuals belonging to high-risk clusters, potential disease risks requiring special attention (e.g., "high risk of cardiovascular disease") can be marked, and further medical examinations can be recommended. By combining the group characteristics of clusters with the specific indicator deviations of individuals, the assessment report of this invention can provide users and medical personnel with personalized and interpretable health assessment results, unlike traditional single-indicator normal / abnormal reports, and has higher reference value.

[0081] In summary, the multi-dimensional physical examination data clustering analysis health assessment method of the present invention, through the innovative introduction of index entropy weight assignment, fuzzy clustering analysis, and health deviation calculation, achieves a comprehensive assessment of the health status of the examinee. This method demonstrates superior performance compared to existing technologies in all embodiments, capable of identifying sub-healthy individuals and potentially high-risk groups, providing a powerful tool for preventive medicine and personalized health management. It should be noted that the method steps and parameter selection of the present invention can be adjusted according to different application scenarios. For example, for chronic disease management in the elderly, emphasis can be placed on blood pressure and blood sugar indicators, with corresponding adjustments to weights and cluster numbers; for annual health assessments of corporate employees, dimensions such as psychological stress can be added. All equivalent variations based on the ideas of this invention are within the scope of protection of this invention.

Claims

1. A health assessment method based on multi-dimensional physical examination data cluster analysis, characterized in that, Includes the following steps: S1. Obtain multi-dimensional physical examination data of the examinee, which includes multiple different categories of health indicator values; S2. Perform data cleaning and standardization on the physical examination data to obtain complete and consistent standardized health data; S3. Calculate the weight coefficient of each health indicator based on the variation of each health indicator value in standardized health data. The weight coefficient reflects the relative importance of each indicator to health differences. S4. Based on standardized health data and weight coefficients, a clustering algorithm is used to divide the examined objects into several clusters with similar health statuses, and each cluster corresponds to a health status category. S5. Analyze the central characteristics of each cluster, assess the health status level or health risk corresponding to the cluster, and output the corresponding health assessment results based on the cluster to which the tested object belongs.

2. The health assessment method based on multi-dimensional physical examination data clustering analysis according to claim 1, characterized in that, The information entropy method is used to calculate the weight coefficients of each health indicator. Specifically, this involves normalizing the value of each health indicator into a percentage form and calculating the information entropy of each health indicator. and according to Determine the weight of each health indicator, among which Furthermore, the sum of the weights of all health indicators is 1.

3. The health assessment method for multi-dimensional physical examination data cluster analysis according to claim 1, characterized in that, Step S4 employs the fuzzy C-means clustering algorithm, which includes the following processes: calculating the distance between the examined object and the center of each cluster based on a weighted distance metric, and iteratively updating the membership degree of the examined object to each cluster. The weighted distance metric is calculated by summing the center vectors of each cluster until the cluster centers converge; where the weighted distance metric is the distance between the examined object i and cluster j. , Let be the standardized value of the i-th subject on the k-th health indicator. Let the value of cluster j be the center of health index k. Let be the weighting coefficient of the k-th health indicator, and p be the total number of health indicators. The direction-sensitive penalty coefficient, , This is a deviation-indicator function. , This is the health reference baseline or target value for health indicator k.

4. The health assessment method for multi-dimensional physical examination data cluster analysis according to claim 3, characterized in that, In the fuzzy C-means clustering algorithm, the membership degree of the examined object to the j-th cluster is... Calculate using the following formula: ， in For the examined object i and the cluster The weighted distance between them, where c is the number of clusters and m is the fuzziness factor; the center vector of each cluster is updated according to the following formula: ， in For clusters The central value of the k-th health indicator, where n is the total number of subjects examined.

5. The health assessment method based on multi-dimensional physical examination data clustering analysis according to claim 1, characterized in that, Step S5 includes: calculating a comprehensive health deviation or risk score for each cluster, and dividing the clusters into different health level categories according to a preset threshold; obtaining the membership distribution of each subject to each cluster, classifying the subject to the corresponding health level, and outputting a health assessment report, which includes the overall health status level of the subject and improvement suggestions for abnormal indicators.

6. A health assessment system based on dimensional physical examination data clustering analysis, characterized in that, A health assessment method for performing a dimensional physical examination data clustering analysis as described in any one of claims 1-5, comprising: a data acquisition module, a data processing module, a clustering analysis module, and an assessment output module; The data acquisition module is used to receive original physical examination records containing the unique identifier and timestamp of the examinee from physical examination equipment, hospital information system or wearable terminal, perform field integrity verification, unit of measurement uniformity and source marking on the records, and output a structured record stream with examinee identifier, timestamp, indicator code, value and unit; The data processing module is used to perform missing value imputation, outlier suppression, unit conversion and standardization on the record stream, generate a standardized data matrix and supporting statistical parameters for clustering, and generate a parameter set for distance measurement and risk assessment. The clustering analysis module is used to perform fuzzy clustering iteration on standardized data based on index weights and weighted distance metrics, solve the cluster centers and the membership degree of the examined objects to each cluster until the convergence condition is met, and output the cluster center vector and membership degree distribution. The assessment output module is used to calculate cluster-level health deviation based on cluster centers and preset health references, and map it to cluster-level risk scores. It calculates individual risk scores and health levels based on the membership of the examined subjects, and generates structured health assessment results that include overall level, dominant deviation indicators and intervention recommendations, and outputs them externally.

Images