Individual health risk dynamic assessment system based on multi-modal biological feature fusion

Abstract

The invention discloses an individual health risk dynamic assessment system based on multi-modal biological feature fusion, and relates to the technical field of medical health information processing. The method comprises the following steps: constructing a high / low-dimension data combination rule; dynamically adjusting the feature fusion weight according to the clinical feature importance weight and the medical efficiency value calculated by the detection time interval; a cost difference value formula is introduced to quantify the economical efficiency of combined evaluation and independent evaluation; calculating a health risk comprehensive coefficient of the monitoring scene by combining the health data credibility score and the index type / amplitude / duration weighted risk value; the risk level is rechecked through expert assessment or a secondary model, so that the assessment accuracy is ensured; updating the individual credibility score according to the evaluation result; and for the result exceeding the confidence interval, classifying and optimizing a feature fusion algorithm, a model parameter or a data combination strategy. The system realizes accurate, economic and real-time individual health risk dynamic assessment, and is suitable for scenes of chronic disease management, health intervention and the like.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of medical and health information processing, and specifically relates to an individual health risk dynamic assessment system based on multi-modal biological feature fusion. BACKGROUND

[0002] With the popularization of health management concepts and the development of medical technology, individual health risk assessment has become a key link in disease prevention and health intervention. At present, health risk assessment technology relies on single type of biological feature data (such as physiological indicators or biochemical test results), which is difficult to comprehensively reflect the complexity of individual health status. At the same time, the existing assessment methods have significant shortcomings in data processing, risk quantification and dynamic optimization, which restrict the assessment accuracy and practicality. Specifically, the existing technology has the following problems: Firstly, the multi-modal biological feature data fusion is insufficient, and the physiological, biochemical, and image data of multiple dimensions lack standardized matching mechanism. The data dimension division is fuzzy, and the rationality of high and low dimensional data combination is difficult to guarantee, resulting in one-sided or distorted assessment of basic data; Secondly, the risk assessment is mostly static analysis, without combining the detection time sequence characteristics and historical data correlation analysis, and the feature fusion weight is fixed, which cannot be dynamically adjusted according to the detection interval and medical efficiency, affecting the timeliness and accuracy of risk quantification; Thirdly, there is a lack of scientific cost-benefit analysis, and the assessment method selection depends on experience, which is easy to cause waste of medical resources or insufficient assessment; Fourthly, the risk scene verification mechanism is imperfect, only the risk level is output by a single model, without expert review or secondary verification process, the high-risk monitoring scene has high misjudgment and omission rate, and the data quality warning mechanism is missing, which is difficult to identify the interference of low credibility data on the assessment result; Therefore, an individual health risk dynamic assessment technology capable of realizing efficient fusion of multi-modal biological features, dynamic quantification of risk, accurate verification of scene, optimization of cost-benefit and forming a closed loop feedback is urgently needed to solve the above problems existing in the prior art and improve the scientificity and practicality of health risk management. The present application proposes an individual health risk dynamic assessment system based on multi-modal biological feature fusion. SUMMARY

[0003] In order to overcome the shortcomings and deficiencies of the prior art, the present application adopts the following technical solutions: The individual health risk dynamic assessment system based on multi-modal biological feature fusion comprises a health information acquisition module: collecting individual health related information, and monitoring time, data credibility score and index risk coefficient of health index abnormal event; Health data matching and risk assessment module: for the standardization processing and preliminary risk assessment of health data, the characteristic information dimension is determined by constructing a dimensional relationship model, the high and low dimensional data are divided according to the threshold value, and the double matching rules are used, including health indicators and basic information combined data; the detection time sequence is screened, and the historical abnormal data are synchronously acquired; the medical cost benefit analysis is used to determine the evaluation method, and the characteristic fusion weight is optimized combined with the detection time and efficiency improvement value; Health risk comprehensive coefficient evaluation module: evaluate the health risk comprehensive coefficient of the monitoring scene, and mark the scene as a high-risk monitoring scene to be verified when the coefficient exceeds the preset threshold value; Monitoring scene review verification and early warning module: review and verify the high-risk monitoring scene, confirm the risk level by calling multi-modal data, expert evaluation or secondary model prediction, and increase the sample size for re-verification when inconsistent; update the individual credibility score after verification, and issue a data quality warning containing missing items, abnormal indicators and retest suggestions if the score is lower than the standard; Health risk assessment report generation module: generate double version reports, including patient version and doctor version; Result verification and feedback optimization module: verify whether the evaluation result is within a reasonable confidence interval, analyze the reason and optimize it when it is out of the interval; for the high-risk monitoring scene associated population, send a reminder at the preset monitoring interval combined with their health intervention plan.

[0004] The system workflow is as follows: Step one, obtain individual health information, including biological characteristic categories, characteristic data volume, health indicator clinical reference interval, detection time, individual basic health information and health indicator abnormal events; Step two, determine the characteristic information dimension according to the biological characteristic data volume of the individual, and standardize the matching of health data combined with the individual health indicator clinical reference interval, individual basic health information and detection time, and integrate the historical data for risk assessment; Step three, based on the health data credibility score of the individual, the indicator risk coefficient and the corresponding data of the historical individual, evaluate the health risk comprehensive coefficient of the monitoring scene; Step four, review and verify the monitoring scene to determine whether it is a high-risk monitoring scene; Step five, generate a health risk assessment report; Step six, collect the health risk assessment results output by the system after the evaluation is completed, verify whether the evaluation result is within a reasonable confidence interval; analyze the reason and feedback optimization when it is out of the interval; at the same time, send a reminder to the high-risk monitoring scene associated population.

[0005] Preferably, in step one, the biometric characteristics categories include physiological characteristics, biochemical characteristics, and imaging characteristics; the individual's basic health information includes age, gender, medical history, lifestyle habits, and genetic history; and the monitoring time, health data credibility score, and indicator risk coefficient when the individual has abnormal health indicator events are simultaneously obtained.

[0006] Preferably, step 2 includes the following sub-steps: S21. Obtain the individual's biometric characteristics and data volume, refer to the characteristic information dimensions of historical electronic health record data, determine the information dimensions of the individual's biometric characteristics, and perform combined matching on the multi-dimensional health data; S22. Screen the detection time series of the health data of the matching combination, eliminate the combination where the detection time of the high-dimensional data is earlier than that of the low-dimensional data, ensure that the detection time of the low-dimensional data takes precedence over the high-dimensional data, and take the detection time difference ≤ the set number of days as the priority condition; synchronously obtain from the health database the abnormal events of the health indicators of historical individuals in the same monitoring scenario, and the corresponding health data credibility score and indicator risk coefficient.

[0007] Preferably, S21 includes the following sub-steps: S211. Dimensional relationship model construction: Extract the biometric categories, data volume, and feature information dimensions from historical health records, classify them by biometric category, establish a mapping model between biometric data volume and feature information dimensions, and calculate the feature information dimensions of the current individual health data; S212, Data Dimension Classification: Preset clinical feature dimension thresholds to divide health data into high-dimensional data and low-dimensional data; S213. Multi-condition matching rules: set dual matching conditions. The health indicator matching condition is that the low-dimensional data detection value must be within the high-dimensional data reference range; the individual basic health information matching condition is to give priority to matching individual data of age ±5 years, same sex, and similar basic disease type and severity; combine the clinical information processing capabilities of a single assessment to perform combined matching of high- and low-dimensional data.

[0008] Preferably, S22 further includes the following sub-steps: S221. Medical cost-benefit analysis: Calculate the medical costs of individual and combined evaluations of the data to be screened, and determine whether to use combined or individual evaluation by calculating the cost difference; S222. Dynamic weight optimization model: Calculate feature fusion weights by combining detection time interval and medical efficiency improvement value.

[0009] Preferably, in step three, if the comprehensive health risk coefficient is greater than a preset threshold, the monitoring scenario is determined to be a high-risk monitoring scenario to be verified.

[0010] Preferably, in step 4, reviewing and verifying the monitoring scenario includes the following steps: Retrieve multimodal biometric data of individuals in the monitoring scenario, including physiological indicators, behavioral data, and medical records; The scenario risk level is reviewed through expert evaluation or secondary model prediction. If the two evaluation results are consistent, the risk level is confirmed; if not, the sample size is increased, the comprehensive health risk coefficient is recalculated, and then verified again.

[0011] Preferably, in step six, analyzing the cause and providing feedback for optimization includes the following steps: When the evaluation result exceeds the reasonable confidence interval due to unreasonable feature fusion, optimize the feature fusion algorithm of the federated learning framework; When the cause is due to inappropriate model parameters, adjust the hyperparameters of the disease-specific assessment model; When it is caused by unreasonable data combination, re-verify the high- and low-dimensional data division thresholds and adjust the data combination strategy based on the contribution of clinical characteristics.

[0012] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are: 1. This invention achieves precise risk quantification through multimodal data fusion and dynamic assessment. By integrating multimodal biometric data, including physiological, biochemical, and imaging data, and combining it with individual basic health information to build a dynamic feature fusion model, this approach addresses the one-sided nature of traditional single-modality assessments. Furthermore, data relevance is ensured through time-difference screening, and the feature fusion ratio is dynamically adjusted based on the importance weights of clinical features, achieving precise quantification of risk factors.

[0013] 2. The present invention improves resource utilization efficiency through medical cost-effectiveness optimization. The economic benefits of combined evaluation and individual evaluation are quantified through the medical cost-effectiveness analysis formula. The system can automatically select the optimal evaluation mode (for example, when the amortized cost of high-dimensional data is lower than that of individual evaluation, combined evaluation is used) to avoid unnecessary waste of resources.

[0014] 3. This invention ensures timely risk intervention through closed-loop feedback and intervention linkage. Through a review and verification module (consistency check between expert assessment and secondary model prediction) and a health data credibility score update mechanism, the system automatically corrects assessment biases (such as errors caused by irrational data combination) and sends customized reminders to people associated with high-risk monitoring scenarios. Furthermore, when assessment results fall outside a reasonable confidence interval, the system can optimize feature fusion algorithms, model hyperparameters, or data combination strategies to continuously improve model accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0016] Figure 1 The module diagram of the individual health risk dynamic assessment system based on multimodal biometric fusion of the present invention is shown; Figure 2 Shown is a workflow diagram of the system of the present invention; Figure 3 A flow chart of health data matching and risk assessment according to the present invention is shown. DETAILED DESCRIPTION

[0017] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0018] In addition, the described features, structures or characteristics can be combined in any suitable manner in one or more example embodiments. In the following description, many specific details are provided to provide a full understanding of the example embodiments of the present disclosure. However, those skilled in the art will appreciate that the technical solutions of the present disclosure can be practiced while omitting one or more of the specific details, or other methods, components, steps, etc. can be adopted. In other cases, well-known structures, methods, implementations or operations are not shown or described in detail to avoid obscuring various aspects of the present disclosure.

[0019] Example 1: See Figure 1 As shown, the individual health risk dynamic assessment system based on multimodal biometric fusion of this embodiment includes Health information acquisition module: collects individual health-related information, including biometrics (physiological, biochemical, imaging, and other categories and data volume), clinical reference ranges for health indicators, detection time, individual basic health information (age, medical history, etc.), as well as monitoring time of abnormal events in health indicators, data credibility score and indicator risk coefficient, and supports real-time or periodic acquisition of abnormal event data.

[0020] The Health Data Matching and Risk Assessment Module is responsible for the standardized processing of health data and preliminary risk assessment. This module determines the dimension of feature information by constructing a dimensional relationship model, divides high- and low-dimensional data according to thresholds, and combines data based on a dual matching rule (health indicators and basic information). It screens test time series to ensure rationality and simultaneously acquires historical abnormal data. It also conducts a medical cost-benefit analysis to determine the assessment method, optimizing feature fusion weights based on test time and efficiency gains.

[0021] Health risk comprehensive coefficient assessment module: Based on the credibility score of individual and historical individual health data and the indicator risk coefficient (weighted calculation based on indicator type, fluctuation amplitude, and duration), the comprehensive health risk coefficient of the monitoring scenario is evaluated. When the coefficient exceeds the preset threshold, the scenario will be marked as a high-risk monitoring scenario to be verified.

[0022] Monitoring scenario review, verification and early warning module: Review and verify high-risk monitoring scenarios by retrieving multimodal data, confirming the risk level through expert evaluation or secondary model prediction, and increasing the sample size for recalculation and verification in case of inconsistency; update the individual credibility score after verification, and issue a data quality warning containing missing items, abnormal indicators and re-testing suggestions if the score is lower than the standard.

[0023] Health risk assessment report generation module: Federated learning is used to integrate multi-source data, and disease-specific models are used to output probabilities. After calibration, calculation of confidence intervals and classification, a dual-version report is generated. The patient version contains risk levels and recommendations, and the doctor version contains detailed probabilities, intervals, levels, warning indicators and guideline basis.

[0024] Result verification and feedback optimization module: Verify whether the assessment results are within a reasonable confidence interval. If they exceed the interval, analyze the reasons (problems with feature fusion, model parameters, data combination, or temporary interference) and make targeted optimizations. For people associated with high-risk monitoring scenarios, combined with their health intervention plans, reminders containing risk levels, intervention recommendations, etc. are issued at preset monitoring intervals.

[0025] The beneficial effects of this embodiment include: The system achieves accurate matching of health data and risk quantification through multimodal biometric fusion and dynamic assessment. Verification and verification enhance assessment reliability, and dual-version reports meet diverse needs. Combining cost-effectiveness-optimized assessment methods with real-time early warning and feedback optimization mechanisms ensures timely risk intervention, improving the efficiency and accuracy of individual health risk management.

[0026] Example 2: See Figure 2 As shown, the individual health risk dynamic assessment system based on multimodal biometric fusion in this embodiment has the following workflow: Step 1. Obtain individual health information, including biometric categories, feature data volume, clinical reference range of health indicators, detection time, individual basic health information and abnormal health indicator events; among them, biometric categories include physiological characteristics (heart rate, blood pressure, etc.), biochemical characteristics (blood routine, blood lipids, etc.), and imaging characteristics (ultrasound, CT images, etc.); individual basic health information includes age, gender, past medical history, living habits, genetic history, etc.; simultaneously obtain the monitoring time when the individual has abnormal health indicator events, health data credibility score and indicator risk coefficient.

[0027] Obtaining abnormal events of individual health indicators includes the following steps: obtaining individual health monitoring data in real time, and when an individual indicator is abnormal, obtaining abnormal events of individual health indicators in real time; or regularly obtaining individual health monitoring reports, and obtaining abnormal events of individual health indicators based on the health monitoring reports.

[0028] Step 2: Determine the characteristic information dimension based on the amount of individual biometric data, and standardize and match the health data based on the individual's clinical reference range of health indicators, individual basic health information, and testing time. At the same time, integrate historical data for risk assessment: See Figure 3 As shown, the specific process of step 2 is as follows: S21. Obtain the individual's biometric category and data volume, refer to the characteristic information dimension of historical electronic health record data, determine the information dimension of the individual's biometrics, and combine and match the multi-dimensional health data.

[0029] Purpose: By unifying information dimensions, it ensures that health data from different individuals are comparable when evaluating them. For example, high-dimensional data (such as CT images) must be matched with low-dimensional data (such as blood pressure) in time series to avoid evaluation bias caused by differences in data dimensions.

[0030] Specifically including: S211, dimensional relationship model construction: extract the biometric categories, data volume and characteristic information dimensions in the historical health records, classify them according to the biometric categories, including physiological characteristics, biochemical characteristics, and imaging characteristics, establish a mapping model between the biometric data volume and characteristic information dimensions, and calculate the characteristic information dimensions of the current individual health data; S212, data dimension classification: preset clinical characteristic dimension thresholds, and divide health data into high-dimensional data (multi-parameter imaging omics data) and low-dimensional data (single blood routine indicator); S213, multi-condition matching rules: set double matching conditions, and the health indicator matching condition is that the low-dimensional data detection value must be within the reference range of the high-dimensional data; the individual basic health information matching condition is to give priority to matching individual data of age ±5 years, same sex, and similar basic disease type and severity (using ICD-10 coding matching); combine high and low dimensional data for matching in combination with the clinical information processing capabilities of a single assessment; S22. Screen the detection time series of the health data of the matching combination, eliminate the combination where the detection time of the high-dimensional data is earlier than that of the low-dimensional data, ensure that the detection time of the low-dimensional data takes precedence over the high-dimensional data, and take the detection time difference ≤ 90 days as the priority condition; synchronously obtain from the health database the abnormal events of the health indicators of historical individuals in the same monitoring scenario, and the corresponding health data credibility score and indicator risk coefficient.

[0031] It also includes medical cost-benefit analysis and dynamic weight optimization: S221. Medical cost-benefit analysis: Calculate the medical costs of individual and combined assessments of the data to be screened, and divide the assessed costs into three categories: direct medical costs, indirect costs, and hidden costs. The calculation formula is: ;in, is the cost difference (economic efficiency of combined evaluation relative to separate evaluation, unit: yuan / case); 、 are the number of high-dimensional and low-dimensional data (examples), respectively; 、 Separately evaluate the cost for high-dimensional and low-dimensional data (yuan / case); Fixed cost of system operation and maintenance for the combined evaluation system (yuan / time); 、 Amortized cost of high-dimensional and low-dimensional data for combined evaluation (unit cost reduction due to equipment / personnel reuse); 、 、 is the scenario cost weight ( ); V is the clinical value coefficient, which is used to balance economic costs and medical necessity, and avoid determining data combination strategies based solely on cost. For high-clinical-value scenarios (such as imaging genomics data for early cancer screening), V is set to 1.2–1.5 to increase the weight of high-dimensional data in cost-effectiveness analysis (even if the direct cost is high, the clinical value of reducing missed diagnoses can offset the economic cost). For basic screening scenarios (such as routine physical examinations and blood tests), V is set to 0.8–1.0, with economic cost as the primary decision-making basis. The transportation cost of high-dimensional testing (transportation fees for off-site imaging examinations), time cost (calculated by local average daily wage × testing time), and repeated testing costs (secondary testing costs due to data mismatch); Used to quantify the potential costs caused by assessment bias, including the incremental costs of subsequent treatment caused by missed diagnosis / misdiagnosis (such as additional costs of late-stage treatment due to missed diagnosis of tumors in the early stages) and the costs of false positive interventions evaluated solely with low-dimensional data (such as unnecessary medication / examination costs caused by erroneous abnormality warnings).

[0032] when <−500 yuan / case (significant cost savings): Forced to use high-low dimensional data combination evaluation, preferentially matching 90-day time series data.

[0033] When −500 ≤ <200 yuan / case (marginal cost balance): Combined with the clinical value coefficient V, if V≥1.2, use combination evaluation, otherwise use low-dimensional data alone.

[0034] When ≥200 yuan / case (significant incremental cost): Use low-dimensional data alone for evaluation, and simultaneously mark the potential needs of high-dimensional data (such as triggering supplementary evaluation when subsequent abnormal progression occurs).

[0035] Effect: High-dimensional data (such as multi-parameter imaging) usually has high acquisition cost and complex analysis, while low-dimensional data (such as blood routine) has low cost and high efficiency. Through information dimension division, it can be determined whether to use combination evaluation (mixing high and low dimensional data) or separate evaluation, and optimize the allocation of medical resources.

[0036] S222, dynamic weight optimization model: Combine detection time interval and medical efficiency improvement value to calculate feature fusion weight, formula is: ; Wherein, is the i-th feature fusion weight (after normalization); is the clinical feature importance weight; is the medical efficiency improvement value, the calculation formula is: ; Wherein, is the adjustment coefficient; is the detection time interval (days).

[0037] Effect: Information dimension and detection time interval jointly affect the feature fusion weight. For example, if the detection time of high-dimensional data is too early, it may reduce its weight in risk assessment.

[0038] Step three, based on the health data credibility score of the individual, the index risk coefficient and the corresponding data of the historical individual, the health risk comprehensive coefficient of the monitoring scene is evaluated, and the calculation formula is as follows: ; Wherein, is the health risk comprehensive coefficient (range 0~1, the larger the value, the higher the risk); is the health data credibility score of the i-th individual in the monitoring scene; is the index risk coefficient of the i-th individual in the monitoring scene; is the number of historical individuals (in the same monitoring scene). If the health risk comprehensive coefficient is greater than the preset threshold, the monitoring scene is determined to be a high-risk monitoring scene to be verified.

[0039] Monitoring scenarios include routine inspection scenarios, abnormal event scenarios, and chronic disease management scenarios.

[0040] Routine checkup scenario: Individuals submit health reports regularly, and the system assesses risks based on historical data.

[0041] Abnormal event scenario: If an individual shows abnormal indicators (such as a sudden increase in blood pressure), the system automatically triggers the high-risk monitoring scenario verification process.

[0042] Chronic disease management scenario: Diabetic patients continuously monitor their blood sugar, and the system calculates the risk factor based on the fluctuation amplitude and duration.

[0043] The calculation of the indicator risk coefficient includes the following steps: obtaining the indicator type, abnormal fluctuation amplitude and duration corresponding to the abnormal event of the individual health indicator; setting the indicator type weight, fluctuation amplitude weight and duration weight, and calculating the indicator risk coefficient based on the weighted sum: ;in, is the risk value of the indicator type (e.g., physiological indicator tphysiological = 1.0, biochemical indicator tbiochemical = 1.5, imaging indicator timaging = 2.0); The abnormal fluctuation range risk value (if it exceeds the reference interval by 20%, , more than 50% ); is the risk value of duration (e.g., if it lasts for 3 days, , lasting 7 days ); 、 、 are the corresponding weights respectively.

[0044] Step 4: Review and verify the monitoring scenario to determine whether it is a high-risk monitoring scenario; after verification, update the individual score based on the health data credibility score calculation formula, which is as follows: ;in, Score the credibility of health data (ranging from 0 to 1, with higher values ​​indicating higher credibility); is the total number of high-risk monitoring scenarios to be verified under the monitoring scenario; The number of monitoring scenarios that have been identified as high-risk monitoring scenarios; The continuous monitoring duration of individual health data (days); 、 is the weight.

[0045] The review and verification of the monitoring scenario includes the following steps: retrieving the multimodal biometric data of individuals in the monitoring scenario, including physiological indicators, behavioral data and medical records; reviewing the scenario risk level through expert evaluation or secondary model prediction. If the two evaluation results are consistent, the risk level is confirmed; if not, the sample size is increased, the comprehensive health risk coefficient is recalculated, and then verified again.

[0046] It also includes a data quality warning step: if the individual's health data credibility score is lower than the preset standard, a data quality warning reminder will be issued to the individual or his health manager. The warning content includes missing data items, abnormal fluctuation indicators and re-testing suggestions.

[0047] Step 5: Generate a health risk assessment report. The process includes: feature fusion (federated learning framework integrates multi-source data), model construction (disease-specific assessment model outputs raw probabilities), and result generation (calibrating probability values ​​→ calculating confidence intervals → assigning risk levels). Generate a dual-version report: a patient version (risk level + recommendations) and a doctor version (probability value / CI / level / warning indicator / guideline basis). Step 6. After the assessment is completed, collect the health risk assessment results output by the system and verify whether the assessment results are within a reasonable confidence interval; if they are outside the interval, analyze the reasons (direct reasons: unreasonable feature fusion, inappropriate model parameters, unreasonable data combination; occasional reasons: temporary interference in data collection) and provide feedback for optimization; at the same time, issue reminders to people associated with high-risk monitoring scenarios.

[0048] Issuing reminders to people associated with high-risk monitoring scenarios includes the following steps: obtaining the health intervention plan of individuals in the population associated with the high-risk monitoring scenario to determine whether the high-risk monitoring scenario is within the individual's health intervention focus range; if the high-risk monitoring scenario is within the individual's health intervention focus range, then when the interval between the individual's next monitoring time and the indicator associated with the high-risk monitoring scenario is a preset time, a health risk reminder is issued to the individual, and the reminder content includes the risk level, associated indicators, intervention recommendations and review time.

[0049] Cause analysis and feedback optimization include the following steps: when the evaluation result exceeds the reasonable confidence interval due to unreasonable feature fusion, optimize the feature fusion algorithm of the federated learning framework; when it is caused by inappropriate model parameters, adjust the hyperparameters of the disease-specific evaluation model (based on the 10-fold cross-validation results); when it is caused by unreasonable data combination, re-verify the high- and low-dimensional data division thresholds, and adjust the data combination strategy based on the contribution of clinical features.

[0050] The beneficial effects of this embodiment include: accurate matching of health data and risk quantification through multimodal biometric fusion and dynamic assessment. Verification and verification enhance reliability, and dual-version reports accommodate diverse needs. Combining cost-effectiveness-optimized assessment methods with real-time early warning and feedback optimization mechanisms ensures timely risk intervention, improving the efficiency and accuracy of individual health risk management.

[0051] The weighting coefficients of the present invention are used to measure the degree of influence of different factors or variables on a particular outcome or decision. A weighting coefficient is defined as a numerical value assigned to each factor when comparing and evaluating multiple factors to reflect its importance or priority. These weighting coefficients can be determined based on specific circumstances and needs, and are typically developed and confirmed by professionals or relevant stakeholders. By properly setting weighting coefficients, programs or systems can be helped to make more accurate decisions or predictions.

[0052] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solution and inventive concept of the present invention, should be covered by the scope of protection of the present invention.

[0053] The preferred embodiments of the present invention disclosed above are intended only to help illustrate the present invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the present invention to specific embodiments. Obviously, many modifications and variations are possible based on the contents of this specification. These embodiments are selected and described in detail in this specification to better explain the principles and practical applications of the present invention, thereby enabling those skilled in the art to better understand and utilize the present invention. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. An individual health risk dynamic assessment system based on multimodal biometric fusion, characterized by: The system includes Health information acquisition module: collects individual health-related information, as well as the monitoring time of abnormal health indicator events, data credibility score and indicator risk coefficient; Health Data Matching and Risk Assessment Module: This module is used for standardized processing of health data and preliminary risk assessment. It determines the dimension of feature information by building a dimensional relationship model, divides high- and low-dimensional data according to thresholds, and combines health indicators and basic information according to dual matching rules. It also screens detection time series and simultaneously obtains historical abnormal data. It also conducts medical cost-benefit analysis to determine the evaluation method and optimizes feature fusion weights based on detection time and efficiency improvement values. Health risk comprehensive coefficient assessment module: evaluates the comprehensive health risk coefficient of the monitoring scenario. When the coefficient exceeds the preset threshold, the scenario is marked as a high-risk monitoring scenario to be verified; Monitoring scenario review, verification, and early warning module: Review and verify high-risk monitoring scenarios by retrieving multimodal data and confirming the risk level through expert evaluation or secondary model prediction. In case of inconsistency, the sample size is increased and re-verified. After verification, the individual credibility score is updated. If the score is below the standard, a data quality warning containing missing items, abnormal indicators, and re-testing suggestions is issued. Health risk assessment report generation module: Generate dual-version reports, including patient version and doctor version; Result verification and feedback optimization module: Verify whether the assessment results are within a reasonable confidence interval, analyze the reasons if they exceed the interval, and make targeted optimizations; for people associated with high-risk monitoring scenarios, combined with their health intervention plans, reminders are issued at preset monitoring intervals.

2. The individual health risk dynamic assessment system based on multimodal biometric fusion according to claim 1 is characterized in that: The system workflow is as follows: Step 1: Obtain individual health information, including biometric type, feature data volume, clinical reference range of health indicators, testing time, individual basic health information and abnormal health indicator events; Step 2: Determine the dimension of characteristic information based on the amount of individual biometric data, and standardize and match the health data by combining the clinical reference range of individual health indicators, individual basic health information, and testing time, while integrating historical data for risk assessment; Step 3: Based on the individual's health data credibility score, indicator risk coefficient and historical individual corresponding data, evaluate the comprehensive health risk coefficient of the monitoring scenario; Step 4: Review and verify the monitoring scenario to determine whether it is a high-risk monitoring scenario; Step 5: Generate a health risk assessment report; Step 6. After the assessment is completed, collect the health risk assessment results output by the system and verify whether the assessment results are within a reasonable confidence interval; if they exceed the interval, analyze the reasons and provide feedback for optimization; at the same time, issue reminders to people associated with high-risk monitoring scenarios.

3. The individual health risk dynamic assessment system based on multimodal biometric fusion according to claim 2 is characterized in that: In the step 1, the biometric characteristics include physiological characteristics, biochemical characteristics, and imaging characteristics; Individual basic health information includes age, gender, medical history, lifestyle habits, and genetic history; the monitoring time when an individual has abnormal health indicator events, the health data credibility score, and the indicator risk coefficient are obtained simultaneously.

4. The individual health risk dynamic assessment system based on multimodal biometric fusion according to claim 2 is characterized in that: The second step includes the following sub-steps: S21. Obtain the individual's biometric characteristics and data volume, refer to the characteristic information dimensions of historical electronic health record data, determine the information dimensions of the individual's biometric characteristics, and perform combined matching on the multi-dimensional health data; S22. Screen the detection time series of the matched health data, and eliminate combinations where the detection time of high-dimensional data is earlier than that of low-dimensional data, to ensure that the detection time of low-dimensional data takes precedence over that of high-dimensional data, and set the detection time difference ≤ the set number of days as the priority condition; Synchronously obtain historical abnormal health indicator events of individuals in the same monitoring scenario from the health database, as well as the corresponding health data credibility score and indicator risk coefficient.

5. The individual health risk dynamic assessment system based on multimodal biometric fusion according to claim 2 is characterized in that: The S21 includes the following sub-steps: S211. Dimensional relationship model construction: Extract the biometric categories, data volume, and feature information dimensions from historical health records, classify them by biometric category, establish a mapping model between biometric data volume and feature information dimensions, and calculate the feature information dimensions of the current individual health data; S212, Data Dimension Classification: Preset clinical feature dimension thresholds to divide health data into high-dimensional data and low-dimensional data; S213, multi-condition matching rule: set dual matching conditions, the health indicator matching condition is that the low-dimensional data detection value must be within the high-dimensional data reference range; The matching conditions for individual basic health information are to give priority to matching individual data of age ±5 years, same sex, and similar underlying disease types and severity; and to combine high- and low-dimensional data with the clinical information processing capabilities of a single assessment.

6. The individual health risk dynamic assessment system based on multimodal biometric fusion according to claim 2 is characterized in that: The S22 further includes the following sub-steps: S221. Medical cost-benefit analysis: Calculate the medical costs of individual and combined evaluations of the data to be screened, and determine whether to use combined or individual evaluation by calculating the cost difference; S222. Dynamic weight optimization model: Calculate feature fusion weights by combining detection time interval and medical efficiency improvement value.

7. The individual health risk dynamic assessment system based on multimodal biometric fusion according to claim 2 is characterized in that: In step three, if the comprehensive health risk coefficient is greater than a preset threshold, the monitoring scenario is determined to be a high-risk monitoring scenario to be verified.

8. The individual health risk dynamic assessment system based on multimodal biometric fusion according to claim 2 is characterized in that: In step 4, reviewing and verifying the monitoring scenario includes the following steps: Retrieve multimodal biometric data of individuals in the monitoring scenario, including physiological indicators, behavioral data, and medical records; The scenario risk level is reviewed through expert evaluation or secondary model prediction. If the two evaluation results are consistent, the risk level is confirmed; if not, the sample size is increased, the comprehensive health risk coefficient is recalculated, and then verified again.

9. The individual health risk dynamic assessment system based on multimodal biometric fusion according to claim 2 is characterized in that: In step 6, analyzing the causes and providing feedback for optimization include the following steps: When the evaluation result exceeds the reasonable confidence interval due to unreasonable feature fusion, optimize the feature fusion algorithm of the federated learning framework; When the cause is due to inappropriate model parameters, adjust the hyperparameters of the disease-specific assessment model; When it is caused by unreasonable data combination, re-verify the high- and low-dimensional data division thresholds and adjust the data combination strategy based on the contribution of clinical characteristics.

Images