A Multimodal Longevity Comprehensive Health Index Calculation Method and System Based on Adaptive Confidence Fusion

Abstract

This invention provides a method and system for calculating a multimodal longevity comprehensive health index based on adaptive confidence fusion. The method acquires individual multimodal health data, eliminates dimensional differences in multimodal health data through z-score standardization, and employs a three-factor confidence model to handle data quality differences and time-related decay. It calculates age acceleration score, vitality score, and risk inverse score separately, and finally generates a comprehensive longevity health index using an additive fusion model. This invention incorporates PET-MR metabolic imaging age into a unified longevity scoring framework; designs adaptive stratification thresholds for four age groups; introduces a three-factor confidence mechanism to handle data of different quality and time-related aspects; and uses an additive fusion model to ensure horizontal comparability of scores in scenarios with missing data. This invention can be applied to scenarios such as physical examination centers, health management institutions, insurance company underwriting and pricing, and aging intervention research, providing individuals with a scientific, interpretable, robust, and highly clinically applicable quantitative assessment of longevity health.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence health data analysis technology, and in particular to a method for calculating a multimodal longevity comprehensive health index based on adaptive confidence fusion. Background Technology

[0002] With the increasing prevalence of population aging, longevity health assessment has become an important research direction in preventive medicine and health management. Biological age, as a core indicator for measuring an individual's degree of aging, more accurately reflects an individual's healthy lifespan potential than chronological age. Traditional health assessments typically rely on single indicators, such as blood glucose, blood lipids, or BMI, which are insufficient to comprehensively reflect an individual's health status and degree of aging. In recent years, research teams both domestically and internationally have developed various biological age assessment systems and commercial products based on different data modalities, but all have significant limitations. For example... Figure 2 As shown, the existing technology has the following common defects: (1) Lack of multimodal integration: Existing systems have not effectively integrated PET-MR metabolic images with epigenetic molecular markers (DNA methylation) and blood biochemical phenotypes under a unified scoring framework, and the complementary advantages of the three types of modal information have not been utilized.

[0003] (2) Data missing vulnerability: Existing algorithms mostly require full data input. When the examinee lacks high-cost DNA methylation or imaging detection data, the scoring system often fails or the output results are incomparable.

[0004] (3) Lack of a unified comprehensive score: All systems output age values ​​or risk percentages, without a comparable comprehensive score of 0-100 that covers biological aging acceleration, vitality status, and risk of multiple diseases, and cannot support individual ranking across data completeness.

[0005] (4) Lack of functional health dimensions: Existing research focuses on molecular clocks and does not include functional dimensions that can be directly intervened in clinical practice, such as muscle reserve, cardiorespiratory fitness, and mental health, which are not practical enough.

[0006] (5) Confidence and timeliness are missing: The impact of differences in the quality of data sources (such as clinical testing in tertiary hospitals vs. self-testing by home devices) and the time decay of data on the accuracy of the score is not considered.

[0007] (6) Fixed weights cannot be adapted: All system weights are fixed inside the model and cannot be dynamically adjusted according to different population standards, age characteristics or clinical scenarios, resulting in high research iteration costs.

[0008] (7) Lack of transparency in the chain of evidence: Commercial system algorithms are closed, making it impossible to trace the original data source and calculation process of a single scoring dimension, which does not meet the compliance requirements of medical institutions. Summary of the Invention

[0009] The purpose of this invention is to provide a multimodal longevity comprehensive health index calculation method based on adaptive confidence fusion, in order to solve the technical problems in the existing technology such as lack of multimodal integration, incomparability of scores in scenarios with missing data, lack of functional health dimensions, lack of consideration of confidence and data timeliness, and fixed and unconfigurable weights, so as to provide individuals with a scientific, interpretable and traceable longevity health assessment.

[0010] To achieve the above-mentioned objectives, the first aspect of this invention provides a method for calculating a multimodal longevity comprehensive health index based on adaptive confidence fusion, the method comprising: S101. Obtain multimodal health data of the target individual. Multimodal health data includes one or more health indicators from blood biochemistry test data, DNA methylation test data, PET-MR metabolic imaging data, and lifestyle questionnaire data. S102. Standardize the health indicators in the multimodal health data to obtain standardized health indicators; S103. Calculate the confidence factor for each standardized health indicator, wherein the confidence factor comprehensively reflects the quality of the data source, the reliability of the algorithm, and the data decay over time; S104. Based on standardized health indicators, calculate the scores for the three modules, including the accelerated aging score, vitality score, and risk inverse score. S105. Input the age acceleration score, vitality score, risk inverse score and confidence factor into the additive fusion model to calculate the comprehensive longevity and health index. S106. Output the comprehensive longevity and health index, the scoring results of health indicators in each dimension, the data source, multimodal health data, and the calculation process.

[0011] Furthermore, in step S102, the z-score standardization method is used to process each health indicator in the multimodal health data, and the processing results are mapped to 0-100 points through piecewise linear interpolation or Sigmoid function mapping to obtain standardized health indicators.

[0012] Furthermore, the confidence factor is calculated as follows:

[0013] In the formula, This represents the confidence factor. Indicates the data quality coefficient. Represents the algorithm's reliability coefficient. Indicates the time decay coefficient. Indicates the time span of the data.

[0014] Furthermore, the accelerated age score is based on a comprehensive assessment of three sub-dimensions: blood test phenotypic age difference, DNA methylation age difference, and PET metabolic age difference. The formula for calculating the accelerated age score is as follows:

[0015]

[0016] In the formula, Indicates the aging acceleration index, Indicates the first Weights of each sub-dimension Indicates the first Biological age calculated from individual dimensions Indicates calendar age. Indicates the first in the reference population Standard deviation of the difference between body temperature and age Indicates that age accelerates the scoring. This represents the sensitivity coefficient.

[0017] Furthermore, differentiated age difference judgment thresholds are set for different age groups, and different age difference judgment threshold systems are used for blood biochemical test data, DNA methylation test data, and PET-MR metabolic imaging data. This includes classifying age difference values ​​into corresponding levels based on the target individual's age group and corresponding health data modality, generating age difference judgment results, and using these results to determine the age acceleration sub-dimension mapping score, generate health assessment interpretation messages, determine key focus dimensions, and form a traceable evidence chain.

[0018] Furthermore, the vitality score is obtained by weighted calculation based on ten dimensions: metabolic health, inflammation level, muscle reserve, cardiorespiratory fitness, bone strength, organ function, vision level, mental health, environmental conditions, and genetic advantages.

[0019] Furthermore, in step S104, the risk of tumor, cardiovascular disease, diabetes, Alzheimer's disease, and metabolic syndrome is considered, and a probability coupling model is used to calculate the overall incidence probability. Based on the overall incidence probability, a risk inverse score is calculated.

[0020] Furthermore, the calculation formula for the additive fusion model is:

[0021] In the formula, This represents the comprehensive longevity and health index. Indicates the baseline score. Indicates the first The scores for each module, Indicates the first Preset weights for each module, Indicates the first Confidence factor for each module.

[0022] Furthermore, in step S106, based on the comprehensive longevity health index and the scores of health indicators in each dimension, personalized intervention suggestions are automatically generated, prioritizing behavioral, nutritional, or medical intervention plans for the interventionable dimensions with the lowest scores.

[0023] A second aspect of the present invention provides a multimodal longevity comprehensive health index calculation system based on adaptive confidence fusion, the system comprising: The data acquisition module is used to acquire multimodal health data of the target individual. The multimodal health data includes one or more health indicators from blood biochemistry test data, DNA methylation test data, PET-MR metabolic imaging data, and lifestyle questionnaire data. The standardization processing module is used to standardize various health indicators in multimodal health data to obtain standardized health indicators. The confidence calculation module is used to calculate the confidence factor of each standardized health indicator. The confidence factor comprehensively reflects the quality of the data source, the reliability of the algorithm, and the data decay over time. The scoring calculation module is used to calculate scores for three modules based on standardized health indicators, including the accelerated age score, vitality score, and risk inverse score. The fusion calculation module is used to input the age acceleration score, vitality score, risk inverse score and confidence factor into the additive fusion model to calculate the comprehensive longevity and health index; The results output module is used to output the comprehensive longevity and health index, the scores of health indicators in each dimension, the data source, multimodal health data, and the calculation process.

[0024] Compared with the prior art, the beneficial effects of the present invention are: (1) Multimodal integration: For the first time, a unified framework integration of blood biochemical phenotypic age, DNA methylation epigenetic age and PET-MR metabolic imaging age is achieved, with complementary information and more comprehensive and accurate assessment.

[0025] (2) Robustness of missing data: The comprehensive longevity and health index is calculated by using an additive fusion model. The scores are robust and comparable in the case of missing data, which solves the core problem of score failure caused by missing data in the existing technology. It can provide support for horizontal ranking of subjects with different packages / different data completeness.

[0026] (3) Adaptive confidence mechanism: The confidence model automatically processes different data quality and data timeliness. High-quality and latest data contribute the most, while the contribution of old data naturally decays over time, and the score is closer to the current real health status of the examinee.

[0027] (4) Scientific nature of age stratification: The adaptive age stratification threshold fully reflects the age dependence of aging, and the assessment of different age groups is more targeted, avoiding the systematic low score problem of the elderly group caused by the "one-size-fits-all" threshold.

[0028] (5) Clinical interpretability: This invention provides the original data source and calculation path for each scoring dimension, meets the compliance requirements of medical institutions, and enhances the trust of examinees and the persuasiveness of intervention guidance.

[0029] (6) Wide range of applications: This invention can be applied to a variety of scenarios such as comprehensive evaluation of high-end physical examination centers, long-term tracking of health management institutions, commercial insurance underwriting and pricing, evaluation of the effects of aging intervention, and actuarial science of life insurance, and has strong commercial value. Attached Figure Description

[0030] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only preferred embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0031] Figure 1 This is a schematic diagram of the overall process of a multimodal longevity comprehensive health index calculation method based on adaptive confidence fusion provided by the present invention.

[0032] Figure 2 This is a schematic diagram illustrating the main limitations of existing biological age assessment systems / products.

[0033] Figure 3 This is a schematic diagram of the threshold system for judging blood test phenotypic age difference / DNA methylation age difference provided by the present invention.

[0034] Figure 4 This is a schematic diagram of the PET metabolic age difference judgment threshold system provided by the present invention.

[0035] Figure 5 This is a schematic diagram of the functional health dimensions provided by the present invention.

[0036] Figure 6 This is a schematic diagram of the five major disease risk dimensions provided by the present invention.

[0037] Figure 7 This is a schematic diagram of the data missing handling mechanism provided by the present invention.

[0038] Figure 8 This is a schematic diagram of blood testing indicators provided in Embodiment 1 of the present invention.

[0039] Figure 9 This is a schematic diagram of biological age data provided in Embodiment 1 of the present invention.

[0040] Figure 10 This is a schematic diagram of the disease risk assessment results provided in Embodiment 1 of the present invention.

[0041] Figure 11 This is a schematic diagram of the confidence calculation results provided in Embodiment 1 of the present invention.

[0042] Figure 12 This is a schematic diagram of blood test results provided in Embodiment 2 of the present invention.

[0043] Figure 13 This is a schematic diagram of health trends provided in Embodiment 3 of the present invention. Detailed Implementation

[0044] The principles and features of the present invention are described below with reference to the accompanying drawings. The listed embodiments are only used to explain the present invention and are not intended to limit the scope of the present invention.

[0045] Reference Figure 1 This embodiment provides a method for calculating a multimodal longevity comprehensive health index based on adaptive confidence fusion. The method includes the following steps: S101. Obtain multimodal health data of the target individual. Multimodal health data includes one or more health indicators from blood biochemistry test data, DNA methylation test data, PET-MR metabolic imaging data, and lifestyle questionnaire data.

[0046] S102. Standardize the health indicators in the multimodal health data to obtain standardized health indicators.

[0047] S103. Calculate the confidence factor for each standardized health indicator. The confidence factor comprehensively reflects the quality of the data source, the reliability of the algorithm, and the data decay over time.

[0048] S104. Based on standardized health indicators, calculate the scores for the three modules, including the accelerated age score, vitality score, and risk inverse score.

[0049] S105. Input the age acceleration score, vitality score, risk inverse score, and confidence factor into the additive fusion model to calculate the comprehensive longevity and health index.

[0050] S106. Output the comprehensive longevity and health index, the scoring results of health indicators in each dimension, the data source, multimodal health data, and the calculation process.

[0051] As a preferred implementation, in step S102, the z-score standardization method is used to process health indicators with different dimensions to achieve cross-modal comparability. The corresponding expression is:

[0052] In the above formula, This represents the z-value calculated using the z-score method. This represents unprocessed health indicators in multimodal health data. and These represent the mean and standard deviation of health indicators for a reference population of the same age and sex, respectively. The reference population data are from large-scale domestic and international epidemiological cohort study databases.

[0053] Next, the health indicators are converted into standardized health indicators ranging from 0 to 100. This implementation provides two methods. By default, piecewise linear interpolation is used, employing a predefined piecewise linear interpolation function to map the z-values ​​or age differences of each health indicator to a score of 0-100, as shown below: 1) Excellent: Significantly better than peers, with a score of 90-100; 2) Good: Better than peers, score 75-89; 3) Normal: Comparable to the same age group, score 60-74; 4) Slightly high (slightly low): Slightly deviates from the normal range, with a score of 45-59; 5) Poor: Significantly deviates from the normal range, with a score of 25-44; Poor: Significantly deviates from the normal range, score 0-24.

[0054] The alternative method is the Sigmoid function mapping method, and the corresponding calculation formula is shown below:

[0055] In the formula, This represents the calculated mapping value. The sensitivity coefficient is taken here. It can be adjusted via a YAML file. Compared to piecewise linear interpolation, the Sigmoid function mapping method provides a smoother score transition and is suitable for continuous indicators.

[0056] As another preferred implementation, in step S103, this embodiment introduces a three-factor confidence model to calculate the confidence factor for each dimension of the standardized health index. The calculation formula is as follows:

[0057] In the formula, This represents the confidence factor. Indicates the data quality coefficient. Represents the algorithm's reliability coefficient. Indicates the time decay coefficient. Indicates the time span of the data.

[0058] Data quality coefficient This reflects the reliability of the testing equipment or institution. For example, it can be determined based on the source of the multimodal health data. The possible values ​​for are shown below: 1) Clinical testing at a top-tier hospital: q=1.0, the highest quality gold standard; 2) Secondary hospitals or formal medical examination institutions: q=0.9, high quality; 3) Home medical device self-test: q=0.7, medium quality; 4) Questionnaire self-reported data: q=0.6, relatively subjective.

[0059] Algorithm reliability coefficient This reflects the degree of scientific validation of the algorithm used to calculate biological age. For example, the algorithm used to calculate biological age is compared with... The correspondence is as follows: 1) Directly measured value: 0.95≤r≤1.0, direct observation has the highest reliability; 2) DNA methylation inferred value: 0.85≤r≤0.92, depending on the accuracy of the detection platform and the algorithm version; 3) Image analysis value (PET-MR): 0.88≤r≤0.95, determined according to equipment resolution and analysis model; 4) Gene inference / prediction value: 0.60≤r≤0.80, with limited prediction accuracy.

[0060] Time decay coefficient Used to control the decay rate of historical data confidence over time (default) Data time span is used to measure data timeliness, representing the time interval from the date of data detection to the current evaluation date. For example, The value is determined according to the following rules: 1) Current test (≤1 month): , ; 2) 6 months ago: , ; 3) 1 year ago: , ; 4) 2 years ago: , ; 5) 3 years or more: , .

[0061] In this embodiment, the accelerated aging score is calculated based on three sub-dimensions: blood test phenotypic age difference, DNA methylation age difference, and PET metabolic age difference, to comprehensively assess the aging rate of the target individual.

[0062] The data source for the Phenoage Difference (PhenoAge Diff) is blood biochemistry (9 items) test data. It is calculated based on the Phenoage algorithm removed by Levine et al. (2018) by inputting albumin, creatinine, glucose, C-reactive protein, lymphocyte percentage, mean corpuscular volume, red blood cell distribution width, alkaline phosphatase, white blood cell count and calendar age.

[0063] The data source for the DNA methylation age difference (Epigenetic Age Diff) is whole blood or saliva DNA methylation detection. It supports calculation using various epigenetic clock algorithms such as GrimAge, DunedinPACE, and PhenoAge methylated version. The algorithm can be specified through a configuration file.

[0064] The data source for PET Metabolic Age Diff is PET-MR image analysis, obtained through multi-organ FDG uptake analysis. It is used to assess the deviation of glucose metabolic activity from that of a peer reference population and outputs the metabolic age difference.

[0065] As another preferred implementation, the age difference across the three sub-dimensions is standardized and weighted by introducing the Age Acceleration Index (AAI), calculated as follows:

[0066] In the formula, Indicates the aging acceleration index; Indicates the first Weights of each sub-dimension; Indicates the first Biological age calculated from individual dimensions; Indicates calendar age; Indicates the first in the reference population The standard deviation of the temperature-age difference is used to eliminate differences in the dimensions of different clocks.

[0067] Then, based on The age acceleration score is calculated using an exponential decay function and is expressed as:

[0068] Sensitivity coefficient here The default setting is 0.5. When That is, when the biological age is less than the calendar age. Approaching 100; when That is, when the biological age is greater than the calendar age, the score decreases, and the final score is limited to the range of [0, 100].

[0069] As a further preferred implementation, differentiated age difference judgment thresholds are set for different age groups to reflect the biological principle that "tolerance for accelerated aging increases with age." Different age difference judgment threshold systems are used for blood biochemistry test data, DNA methylation test data, and PET-MR metabolic imaging data.

[0070] Specifically, for the blood test phenotypic age difference / DNA methylation age difference, the applicable age difference judgment threshold system is as follows: Figure 3 As shown. For the metabolic age difference in PET, the applicable judgment threshold system is as follows: Figure 4 As shown.

[0071] The judgment results of the age difference judgment threshold system serve as important intermediate results for age acceleration score calculation and health assessment interpretation. Specifically, after obtaining the blood test phenotype age difference, DNA methylation age difference, or PET metabolic age difference, the system first determines the age group to which the target individual belongs based on their calendar age. Then, based on the corresponding data modality, it calls the corresponding age difference judgment threshold table to classify the age difference value into levels such as excellent, good, normal, slightly high, poor, or very poor, and generates age difference status labels, sub-dimension level results, and corresponding sub-dimension mapping scores.

[0072] The judgment results of the age difference judgment threshold system are used for at least one of the following processes: First, to determine the segmented mapping score of the age acceleration sub-dimension, serving as one of the inputs for calculating the age acceleration score; Second, to explain the causes of the age acceleration score, identifying the main sources of age acceleration in blood biochemistry, DNA methylation, or PET-MR metabolic imaging; Third, to generate stratified prompts in health assessment reports, enabling examinees to know the degree of deviation of their biological age from the reference population of the same age group; Fourth, to determine the priority of personalized intervention recommendations. When any age difference judgment result is at a high, low, or poor level, the system marks the corresponding modality as a key focus dimension and prioritizes lifestyle, nutrition, exercise, follow-up examination, or medical assessment recommendations related to this modality in subsequent health management recommendations; Fifth, to form a traceable chain of evidence, recording the original age difference value, the age group to which it belongs, the applicable threshold range, the judgment level, the sub-dimension score, and its contribution to the comprehensive longevity health index.

[0073] Therefore, the age difference judgment threshold system is not only used to statically stratify age difference values, but also further participates in scoring calculation, result interpretation, risk warning, intervention suggestion generation and evidence chain tracing, thereby improving the clinical interpretability and health management operability of the comprehensive longevity health index.

[0074] In this implementation, all thresholds are managed in the form of TAML configuration files and can be dynamically updated based on new population study data.

[0075] The vitality score encompasses ten clinically measurable dimensions of functional health, comprehensively assessing an individual's current physical "reserve capacity" and vitality level. These functional health dimensions include metabolic health, inflammation levels, muscle reserve, cardiorespiratory fitness, bone strength, organ function, vision, mental health, environmental factors, and genetic predispositions, specifically as follows: Figure 5 As shown.

[0076] Vitality Score The calculation formula is as follows:

[0077] In the formula, This represents the weight of the i-th functional health dimension. This represents the confidence factor for the i-th functional health dimension. This represents the mapping score for the i-th functional health dimension.

[0078] For the reverse risk score, a comprehensive assessment is made based on cancer risk, cardiovascular risk, diabetes risk, Alzheimer's disease risk, and metabolic syndrome risk (e.g., ...). Figure 6 As shown in the figure, a probability coupling model is used to calculate the comprehensive incidence probability, and a risk inverse score is calculated based on the comprehensive incidence probability.

[0079] Using a probabilistic coupling model to calculate the overall incidence probability can avoid overestimation or underestimation caused by simple weighted summation. Overall Incidence Probability The calculation formula is as follows:

[0080] in, For the first Weighted incidence probability across disease dimensions .

[0081] Based on this, risk reverse scoring The calculation formula is as follows:

[0082] As can be seen from the above formula, the lower the overall incidence rate, the higher the risk inverse score.

[0083] In this embodiment, the comprehensive longevity and health index is calculated using an additive fusion model of "baseline score + increase / decrease score", and the corresponding expression is as follows:

[0084] In the formula, This represents a comprehensive longevity and health index. This represents the baseline score, which is preset to 60 points and represents the default health level estimate when data is completely missing. Indicates the first The scores for each module, This includes the age acceleration score, vitality score, and risk inverse score. Indicates the first The preset weights for each module are, for example, the weight of the age acceleration score is 0.35, the weight of the vitality score is 0.40, and the weight of the risk reverse score is 0.25. Indicates the first Confidence factor for each module.

[0085] When all three modules have complete data and the confidence factor is 1, the above formula degenerates into a standard weighted average, as shown below:

[0086] When data is missing from the three modules, different processing strategies are adopted according to the specific circumstances of the missing data. The specific processing mechanisms are as follows: Figure 7 As shown.

[0087] The processing mechanism ensures that examinees with different levels of data completeness are on the same scale, allowing for direct horizontal comparison and ranking, and preventing abnormally high or low scores due to missing data.

[0088] In this invention, all weights, thresholds, and parameters are managed through an external YAML configuration file. After the configuration file is updated, the system automatically loads the new parameters during the next calculation, eliminating the need to recompile or republish the algorithm package. This allows the R&D and operations teams to maintain the configuration independently. The YAML hot-update configuration mechanism decouples business parameters from algorithm logic, enabling the operations and research teams to independently iterate on weights and thresholds to adapt to new population research findings without needing to republish the algorithm package.

[0089] Simultaneously, this invention generates a complete chain of evidence report for each assessment, including data sources, test values, and calculation processes. This chain of evidence report allows for the retrospective analysis of historical comprehensive longevity and health index assessments. The complete chain of evidence traceability mechanism provides the original data source and calculation path for each scoring dimension, thereby meeting the compliance requirements of medical institutions and enhancing the trust of examinees and the persuasiveness of intervention guidance.

[0090] As another preferred implementation method, after obtaining the comprehensive longevity and health index, personalized intervention suggestions are automatically generated based on the comprehensive longevity and health index and the scores of health indicators in each dimension, prioritizing specific behavioral, nutritional or medical intervention plans for the interventionable dimensions with the lowest scores.

[0091] Based on the foregoing method embodiments, another embodiment of the present invention provides a multimodal longevity comprehensive health index calculation system based on adaptive confidence fusion, the system comprising: The data acquisition module is used to acquire multimodal health data of the target individual. The multimodal health data includes one or more health indicators from blood biochemistry test data, DNA methylation test data, PET-MR metabolic imaging data, and lifestyle questionnaire data. The standardization processing module is used to standardize various health indicators in multimodal health data to obtain standardized health indicators. The confidence calculation module is used to calculate the confidence factor of each standardized health indicator. The confidence factor comprehensively reflects the quality of the data source, the reliability of the algorithm, and the data decay over time. The scoring calculation module is used to calculate scores for three modules based on standardized health indicators, including the accelerated age score, vitality score, and risk inverse score. The fusion calculation module is used to input the age acceleration score, vitality score, risk inverse score and confidence factor into the additive fusion model to calculate the comprehensive longevity and health index; The results output module is used to output the comprehensive longevity and health index, the scores of health indicators in each dimension, the data source, multimodal health data, and the calculation process.

[0092] In a preferred embodiment, the fusion calculation module supports time-decay fusion of historical data, calculates a comprehensive index by weighted fusion of historical detection data and current detection data, and outputs the rate of change of health trend.

[0093] The system is used to execute the methods described in the foregoing method embodiments. Its working principle and technical effects can be completely referred to the content described in the foregoing method embodiments, and will not be repeated here.

[0094] The technical solution of the present invention is illustrated below through three specific embodiments corresponding to different scenarios: Example 1: Comprehensive Health Assessment Based on Multimodal Data (Health Checkup Center Scenario) Basic information of the examinee: Age: 45 years old; Gender: Female; Testing date: 2026-03-01; Blood test indicators such as Figure 8As shown.

[0095] Biological age data such as Figure 9 As shown.

[0096] Calculation of the aging acceleration index (assuming) , ): ; The score is limited to 95 points.

[0097] Vitality score (simplified example, only 5 blood test dimensions): The score is approximately 70 points.

[0098] Disease risk assessment results such as Figure 10 As shown.

[0099] P = 1 (1 0.08)(1 0.06)(1 0.05), approximately 0.178; The score was approximately 82.

[0100] Confidence calculation as follows Figure 11 As shown.

[0101] The corresponding comprehensive longevity and health index is:

[0102] Output result: AgeScore: 95; VitalityScore: 70; RiskReverseScore: 82; Overall Longevity and Health Index: 81; Health level: Good (Good Longevity Status).

[0103] Example 2: Blood testing only (DNA and imaging data missing) A 50-year-old male subject provided only blood test data; DNA methylation and PET-MR data were missing. Blood test results are as follows: Figure 12 As shown: Based on the missing data, the scores for the three modules are calculated separately according to the processing mechanism: AgeScore: DNA Methylation Dimension PET-MR dimensions AgeScore LI The contribution of the increase or decrease in points is 0; VitalityScore ≈ 54 (based on existing blood indicators); RiskReverseScore: Based on limited data evaluation, RiskReverseScore≈70.

[0104] The additive fusion model calculates the comprehensive longevity and health index:

[0105] It can be seen that the comprehensive longevity and health index of the subjects with missing data converges to the baseline score of 60, without producing abnormally high or low scores due to the smaller denominator, and can be meaningfully compared with the subjects with full data (81 points).

[0106] Example 3: Long-term health tracking and monitoring A subject (initially 60 years old) underwent comprehensive assessments three times consecutively. The system used a time decay mechanism to fuse historical data and output trends, such as... Figure 13 As shown.

[0107] System output: Health trends continue to improve; it is recommended to maintain the current intervention plan.

[0108] The formula for fusing historical data with time decay is expressed as follows: .

[0109] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for calculating a multimodal longevity comprehensive health index based on adaptive confidence fusion, characterized in that, The method includes: S101. Obtain multimodal health data of the target individual. Multimodal health data includes one or more health indicators from blood biochemistry test data, DNA methylation test data, PET-MR metabolic imaging data, and lifestyle questionnaire data. S102. Standardize the health indicators in the multimodal health data to obtain standardized health indicators; S103. Calculate the confidence factor for each standardized health indicator, wherein the confidence factor comprehensively reflects the quality of the data source, the reliability of the algorithm, and the data decay over time; S104. Based on standardized health indicators, calculate the scores for the three modules, including the accelerated aging score, vitality score, and risk inverse score. S105. Input the age acceleration score, vitality score, risk inverse score and confidence factor into the additive fusion model to calculate the comprehensive longevity and health index. S106. Output the comprehensive longevity and health index, the scoring results of health indicators in each dimension, the data source, multimodal health data, and the calculation process.

2. The method for calculating a multimodal longevity comprehensive health index based on adaptive confidence fusion according to claim 1, characterized in that, In step S102, the z-score standardization method is used to process each health indicator in the multimodal health data. The processing results are mapped to 0-100 points through piecewise linear interpolation or Sigmoid function mapping to obtain standardized health indicators.

3. The method for calculating a multimodal longevity comprehensive health index based on adaptive confidence fusion according to claim 1, characterized in that, The confidence factor is calculated as follows: In the formula, This represents the confidence factor. Indicates the data quality coefficient. Represents the algorithm's reliability coefficient. Indicates the time decay coefficient. Indicates the time span of the data.

4. The method for calculating a multimodal longevity comprehensive health index based on adaptive confidence fusion according to claim 1, characterized in that, The accelerated age score is a comprehensive assessment based on three sub-dimensions: blood test phenotypic age difference, DNA methylation age difference, and PET metabolic age difference. The formula for calculating the accelerated age score is as follows: In the formula, Indicates the aging acceleration index, Indicates the first Weights of each sub-dimension Indicates the first Biological age calculated from individual dimensions Indicates calendar age. Indicates the first in the reference population Standard deviation of the difference between body temperature and age Indicates that age accelerates the scoring. This represents the sensitivity coefficient.

5. A method for calculating a multimodal longevity comprehensive health index based on adaptive confidence fusion according to claim 1 or 4, characterized in that, Different age difference judgment thresholds are set for different age groups, and different age difference judgment threshold systems are used for blood biochemistry test data, DNA methylation test data, and PET-MR metabolic imaging data. This includes classifying age difference values ​​into corresponding levels according to the age group and corresponding health data modality of the target individual, generating age difference judgment results, and using the age difference judgment results to determine the age acceleration sub-dimension mapping score, generate health assessment interpretation messages, determine key focus dimensions, and form a traceable evidence chain.

6. The method for calculating a multimodal longevity comprehensive health index based on adaptive confidence fusion according to claim 1, characterized in that, The vitality score is calculated by weighting ten dimensions: metabolic health, inflammation level, muscle reserve, cardiorespiratory fitness, bone strength, organ function, vision level, mental health, environmental conditions, and genetic advantages.

7. The method for calculating a multimodal longevity comprehensive health index based on adaptive confidence fusion according to claim 1, characterized in that, In step S104, the risk of cancer, cardiovascular disease, diabetes, Alzheimer's disease, and metabolic syndrome is considered, and a probabilistic coupling model is used to calculate the overall incidence probability. Based on the overall incidence probability, a risk inverse score is calculated.

8. The method for calculating a multimodal longevity comprehensive health index based on adaptive confidence fusion according to claim 1, characterized in that, The calculation formula for the additive fusion model is: In the formula, This represents the comprehensive longevity and health index. Indicates the baseline score. Indicates the first The scores for each module, Indicates the first Preset weights for each module, Indicates the first Confidence factor for each module.

9. The method for calculating a multimodal longevity comprehensive health index based on adaptive confidence fusion according to claim 1, characterized in that, In step S106, based on the comprehensive longevity health index and the scores of health indicators in each dimension, personalized intervention suggestions are automatically generated, prioritizing behavioral, nutritional, or medical intervention plans for the interventionable dimensions with the lowest scores.

10. A multimodal longevity comprehensive health index calculation system based on adaptive confidence fusion, characterized in that, The system includes: The data acquisition module is used to acquire multimodal health data of the target individual. The multimodal health data includes one or more health indicators from blood biochemistry test data, DNA methylation test data, PET-MR metabolic imaging data, and lifestyle questionnaire data. The standardization processing module is used to standardize various health indicators in multimodal health data to obtain standardized health indicators. The confidence calculation module is used to calculate the confidence factor of each standardized health indicator. The confidence factor comprehensively reflects the quality of the data source, the reliability of the algorithm, and the data decay over time. The scoring calculation module is used to calculate scores for three modules based on standardized health indicators, including the accelerated age score, vitality score, and risk inverse score. The fusion calculation module is used to input the age acceleration score, vitality score, risk inverse score and confidence factor into the additive fusion model to calculate the comprehensive longevity and health index; The results output module is used to output the comprehensive longevity and health index, the scores of health indicators in each dimension, the data source, multimodal health data, and the calculation process.

Images