A cross-age health status assessment and anti-aging decision support method, system, device and medium based on a deterministic safety kernel and probabilistic reasoning

Abstract

The application discloses a kind of based on certainty safety kernel and probability inference's cross-age health state evaluation and anti-aging decision support method, system, equipment and medium.The method comprises the following steps: receiving multi-source physiological data, calculating biological age, and generating high-dimensional state vector containing biological age, intrinsic ability index;Calculate the robust distance index between the high-dimensional state vector and the health baseline distribution;Determine the current regime state of the individual based on the robust distance index;Based on the current regime state of the individual, generate candidate intervention strategies;Each candidate intervention strategy in the candidate intervention strategy list is reviewed through a multi-security review mechanism, and at least one target intervention strategy is generated.Low-cost, high-precision aging assessment and multi-security barrier decision support based on conventional data are realized.

Description

Technical Field

[0001] This invention relates to the interdisciplinary application technology of computational medicine, artificial intelligence data processing, and structural health monitoring (SHM), and particularly to a method, system, device, and medium for cross-age health status assessment and anti-aging decision support based on a deterministic safety kernel and probabilistic reasoning. Background Technology

[0002] With the increasing aging of the global population, quantifying aging and implementing precise interventions has become a medical challenge. Current technologies suffer from the following main shortcomings: 1. High testing costs: Current mainstream biological age assessments rely on DNA methylation sequencing (such as HorvathClock), which is expensive and difficult to popularize on a large scale in primary healthcare.

[0003] 2. Lack of theoretical models: Traditional medicine lacks an engineered description of the human body’s “systemic decline” and it is difficult to quantify the overall physiological load under the coexistence of multiple diseases.

[0004] 3. Algorithm black box risk: Purely data-driven AI models are prone to "illusion" and lack the ability to refuse to answer questions in the unknown domain (out-of-distribution), making it difficult to meet the stringent safety requirements for medical device registration (such as NMPA / FDA). Summary of the Invention

[0005] This invention provides a method, system, device, and medium for cross-age health status assessment and anti-aging decision support based on a deterministic safety kernel and probabilistic reasoning, in order to solve the problem of nonlinear status assessment and safe intervention of the human body under the influence of aging and multidimensional physiological load.

[0006] Firstly, a method for cross-age health status assessment and anti-aging decision support based on a deterministic security kernel and probabilistic reasoning is provided, including: It receives multi-source physiological data to calculate biological age and generates a high-dimensional state vector containing biological age and intrinsic ability indicators. Calculate the robust distance index between the high-dimensional state vector and the healthy baseline distribution; Based on the robust distance index, determine the individual's current physical condition; Based on the individual's current physical condition, candidate intervention strategies are generated; Based on the high-dimensional state vector and the candidate intervention strategy list, calculate the cognitive uncertainty score and the random uncertainty score of each candidate intervention strategy in the candidate intervention strategy list; Based on the high-dimensional state vector, the cognitive uncertainty score and the accidental uncertainty score corresponding to each candidate intervention strategy, a multi-layer security review mechanism is used to review each candidate intervention strategy in the candidate intervention strategy list. If the review is passed, at least one target intervention strategy is generated.

[0007] Optionally, the step of receiving multi-source physiological data to calculate biological age and generating a high-dimensional state vector containing biological age and intrinsic ability indicators includes: It receives physiological data from multiple sources and outputs biological age through a pre-trained deep neural network model. The biological age is integrated with pre-obtained intrinsic ability indicators to construct a high-dimensional state vector.

[0008] Optionally, calculating a robust distance metric between the high-dimensional state vector and the healthy baseline distribution includes: Based on robust statistical methods, a target data subset is determined from the health dataset; Based on the mean vector and covariance matrix of the target data subset, a healthy baseline distribution is constructed; Calculate the robust distance index between the high-dimensional state vector and the healthy baseline distribution.

[0009] Optionally, determining an individual's current physical condition based on the robust distance index includes: Based on the aforementioned health dataset, high-risk thresholds and recovery thresholds are adaptively determined. If the robustness distance index is higher than the high-risk threshold and the duration reaches the first preset residence time length, the individual's state is determined to be a weakened constitution.

[0010] If the robustness distance index is lower than the recovery threshold and the duration reaches the second preset residence time length, the individual's status is determined to be healthy.

[0011] Optionally, generating candidate intervention strategies based on the individual's current physical condition includes: Retrieve target intervention objects from the evidence database that correspond to the individual's current physical condition; The current recommendation weight of each target intervention object is calculated by comprehensively weighting the evidence level, quality score and timeliness weight corresponding to each target intervention object, wherein the timeliness weight is calculated by the half-life decay function; Based on the recommendation weights, the target intervention objects are sorted from largest to smallest, and a preset number of first target intervention objects are selected from the sorted list; Based on each of the first target intervention subjects, a list of candidate intervention strategies is generated according to clinical rules.

[0012] Optionally, the multiple security review mechanisms include: a red line threshold determination mechanism, an uncertainty rejection mechanism, a human-in-the-loop intervention mechanism, and a forced blocking mechanism.

[0013] Secondly, a cross-age health status assessment and anti-aging decision support system based on a deterministic security kernel and probabilistic reasoning is provided, characterized in that the system is used to execute the cross-age health status assessment and anti-aging decision support method based on a deterministic security kernel and probabilistic reasoning as described in any one of the embodiments of the present invention, including: The first generation module receives multi-source physiological data, calculates biological age, and generates a high-dimensional state vector containing biological age and intrinsic ability indicators. The first calculation module is used to calculate the robust distance index between the high-dimensional state vector and the health baseline distribution; The determination module is used to determine the current physical condition of an individual based on the robust distance index. The second generation module is used to generate candidate intervention strategies based on the individual's current physical condition. The second calculation module is used to calculate the cognitive uncertainty score and the random uncertainty score of each candidate intervention strategy in the candidate intervention strategy list based on the high-dimensional state vector and the candidate intervention strategy list. The third generation module is used to review each of the candidate intervention strategies in the candidate intervention strategy list based on the high-dimensional state vector, the cognitive uncertainty score and the accidental uncertainty score corresponding to each of the candidate intervention strategies, through a multi-security review mechanism. If the review is passed, at least one target intervention strategy is generated.

[0014] Optionally, the first generation module is specifically used for: It receives physiological data from multiple sources and outputs biological age through a pre-trained deep neural network model. The biological age is integrated with pre-obtained intrinsic ability indicators to construct a high-dimensional state vector.

[0015] Thirdly, an electronic device is provided, the electronic device comprising: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, which enables the at least one processor to execute the cross-age health status assessment and anti-aging decision support method based on deterministic security kernel and probabilistic reasoning as described in any embodiment of the present invention.

[0016] Fourthly, a computer-readable storage medium is provided, the computer-readable storage medium storing computer instructions, the computer instructions being used to cause a processor to execute and implement the cross-age health status assessment and anti-aging decision support method based on deterministic security kernel and probabilistic reasoning as described in any embodiment of the present invention.

[0017] The technical solution of this invention calculates biological age and intrinsic ability scores from low-cost data using a pre-trained model, achieving low-cost, high-precision quantification of aging. It employs MCD to calculate robust distances, resisting outlier interference and accurately quantifying heterostable loads. A Schmitt trigger model is used to determine physical condition, avoiding misjudgments due to short-term fluctuations by setting thresholds and residence time, ensuring stable and reliable status. Intervention strategies are generated from an evidence base, integrating evidence levels and timeliness weights to guarantee the scientific validity and timeliness of recommendations. A deep ensemble model calculates uncertainty scores to quantify recommendation reliability. Finally, a multi-layered security review kernel integrates red-line judgment and rejection mechanisms to build a security barrier, preventing unsafe outputs and ensuring the inherent security and compliance of the system.

[0018] It should be understood that the content described in this section is not intended to identify key or important features of the embodiments of the present invention, nor is it intended to limit the scope of the present invention. Other features of the present invention will become readily apparent from the following description. Furthermore, during operation, the system of the present invention dynamically interacts with each processing stage through multi-round feedback, state linkage, and collaborative control mechanisms. Its execution order and dependencies are adaptively adjusted according to the input state and internal control strategy, avoiding the formation of fixed or predictable processing paths, thereby preventing the reconstruction of the overall decision-making logic through a linear process.

[0019] In this invention, the mapping relationship and intermediate parameters between each processing stage can be implemented in a non-public manner, including but not limited to encrypted parameters, nonlinear transformation and distributed storage, thereby preventing reverse derivation of the internal logic of the system through external observation. Attached Figure Description

[0020] 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 some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 The flowchart of the cross-age health status assessment and anti-aging decision support method based on deterministic security kernel and probabilistic reasoning is provided in Embodiment 1 of the present invention. Figure 2This is a framework diagram of a cross-age health status assessment and anti-aging decision support system based on deterministic security kernel and probabilistic reasoning, provided in Embodiment 3 of the present invention. Figure 3 A schematic diagram of an electronic device that can be used to implement embodiments of the present invention is shown. Detailed Implementation

[0022] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0023] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0024] Application Overview: Existing health assessment and intervention recommendation methods have significant limitations. At the state assessment level, while biological age detection based on omics technologies such as DNA methylation is accurate, it is costly and difficult to widely implement in primary healthcare and health management. Assessments based on routine physical examination indicators lack robust theoretical models for quantifying overall physiological load and aging dynamics. At the decision support level, existing clinical decision support systems or health algorithms often struggle to handle the nonlinearity and time-varying nature of individual physiological states, as well as the complex interactions between multiple indicators. Their recommendations lack quantification of their own decision-making uncertainties and, more importantly, lack a mandatory review framework to ensure the highest safety priority. This limits their reliability and compliance in medical settings.

[0025] This invention introduces an engineering paradigm of "human structural health monitoring," using the Transformer architecture to model routine clinical data, enabling high-precision prediction of biological age; this provides a possibility for low-cost, universally accessible aging assessment. Meanwhile, structural health monitoring (SHM) technology in civil engineering is already very mature in the safety assessment of large bridges and tunnels, and its theories of "damage accumulation" and "modal analysis" have a high degree of mathematical isomorphism with those of human aging assessment.

[0026] Furthermore, this invention constructs an evaluation and decision-making system that integrates robust statistics, deep learning, and deterministic safety rules. Specifically, the system utilizes a pre-trained deep neural network to mine complex relationships between indicators from low-cost routine physical examination data, calculates the biological age reflecting the true degree of aging, and integrates it with intrinsic ability scores into a high-dimensional state vector. Further, robust statistical methods are used to quantify the distance of this state vector from the baseline of healthy individuals, as an anisostatic load, and a dual-threshold model with hysteresis characteristics is used to stably determine whether an individual is in a healthy or frail state. Based on the state generation intervention strategy, the system calculates the cognitive uncertainty (model knowledge blind spot) and accidental uncertainty (inherent noise) of each strategy through a deep ensemble model, and finally makes a decision through a multi-layered safety review kernel including red-line thresholds, uncertainty rejection, human-in-the-loop, and mandatory circuit breaker mechanisms. It enables high-precision status assessment using low-cost data, and forms an inseparable technical closed loop through a three-layer architecture of "probabilistic reasoning, deterministic adjudication, and evidence maintenance." The three must work together to ensure the scientific nature, stability, and inherent security of the system in status determination and intervention recommendations, providing a complete technical closed loop for inclusive and reliable digital health interventions. Example 1

[0027] Figure 1 This invention provides a flowchart of a cross-age health status assessment and anti-aging decision support method based on a deterministic security kernel and probabilistic reasoning, as described in Embodiment 1 of the present invention. This embodiment is applicable to nonlinear state assessment and safety intervention of the human body under aging and multidimensional physiological loads. This method can be executed by a cross-age health status assessment and anti-aging decision support system. Figure 1 As shown, the method includes: S110: Receive multi-source physiological data, calculate biological age, and generate a high-dimensional state vector containing biological age and intrinsic ability indicators. S120. Calculate the robust distance index between the high-dimensional state vector and the healthy baseline distribution; S130. Based on the robust distance index, determine the individual's current physical condition; S140. Based on the individual's current physical condition, generate candidate intervention strategies; S150. Based on the high-dimensional state vector and the candidate intervention strategy list, calculate the cognitive uncertainty score and the random uncertainty score of each candidate intervention strategy in the candidate intervention strategy list; S160. Based on the high-dimensional state vector, the cognitive uncertainty score and the accidental uncertainty score corresponding to each candidate intervention strategy, the candidate intervention strategies in the candidate intervention strategy list are reviewed through a multi-level security review mechanism. If the review is passed, at least one target intervention strategy is generated.

[0028] The robust estimation, candidate intervention strategy generation, and multiple security reviews are executed collaboratively through a coupled multi-layer decision-making closed-loop structure. The outputs of each layer serve as inputs for subsequent processing and participate in overall constraints, thus forming an indivisible decision-making system.

[0029] In this embodiment, a pre-trained model calculates biological age and intrinsic ability scores from low-cost data, achieving low-cost, high-precision quantification of aging. MCD (Mean Distance Difference) is used to calculate robust distances, resisting outlier interference and accurately quantifying heterostable loads. A Schmitt trigger model is used to determine physical condition, and thresholds and residence times are set to avoid misjudgments due to short-term fluctuations, ensuring stable and reliable status. Intervention strategies are generated from an evidence base, integrating evidence levels and timeliness weights to ensure the scientific validity and timeliness of recommendations. A deep ensemble model calculates uncertainty scores to quantify recommendation reliability. Finally, a multi-layered security review kernel integrates red-line judgment and rejection mechanisms to build a security barrier, preventing unsafe outputs and ensuring the inherent security and compliance of the system. Example 2

[0030] The technical solution in this embodiment is a further refinement based on the above embodiments.

[0031] In step S110, multi-source physiological data is received to calculate biological age, and a high-dimensional state vector containing biological age and intrinsic ability indicators is generated, including: It receives physiological data from multiple sources and outputs biological age through a pre-trained deep neural network model. The biological age is integrated with pre-obtained intrinsic ability indicators to construct a high-dimensional state vector.

[0032] The biological age is inferred through a pre-trained deep neural network model based on routine clinical data, and the model architecture is not limited to a specific network type; (receiving multi-source physiological and routine clinical data, constructing a high-dimensional normalized state vector containing biological age and multi-dimensional physiological indicators) Multi-source physiological data refers to diverse data reflecting the physiological state of the human body, obtained from different channels and devices. Specifically, it can include: routine clinical data from medical institutions, wearable device data, and imaging data. Routine clinical data refers to relatively low-cost laboratory indicators widely used in routine physical examinations or clinical tests, such as complete blood counts, urinalysis, and biochemical indicators (e.g., hemoglobin, albumin, creatinine, cholesterol, blood glucose, etc.). Wearable device data refers to data such as heart rate and activity levels continuously monitored by devices like smartwatches and fitness trackers. Imaging data refers to medical imaging data such as X-rays, CT scans, and MRI scans. Pre-trained deep neural network models refer to deep learning models based on the Transformer architecture that have been pre-trained on large-scale health datasets, enabling them to extract features from routine clinical data and predict biological age.

[0033] Biological age refers to the true degree of aging at the cellular level, reflecting an individual's physiological functions. A high-dimensional normalized state vector refers to a standardized mathematical vector containing multiple health dimensions, used to comprehensively characterize an individual's health status. Its core components include at least: biological age and intrinsic capacity score. The Intrinsic Capacity (IC) score is a comprehensive assessment framework proposed by the World Health Organization (WHO) in 2015 to measure the comprehensive functional reserve of older adults at both physiological and psychological levels. Its core objective is to shift from a "disease-centered" to a "function-centered" assessment of healthy aging. Specifically, multi-source physiological data can be automatically retrieved from multiple data sources via API interfaces. For example, continuous monitoring data from wearable devices can be automatically synchronized and retrieved through cloud API interfaces provided by equipment manufacturers, as well as data from interconnected medical information systems (routine clinical data, imaging data). Additionally, data upload interfaces, such as front-end pages or mobile applications, allow users or medical personnel to manually upload physical examination report files containing complete blood counts and biochemical indicators (e.g., 30-50 routine biochemical indicators such as hemoglobin, albumin, creatinine, cholesterol, and blood glucose). These reports can be PDFs, images, or structured data files. After uploading, the uploaded files can be parsed to extract key indicators, such as hemoglobin (reflecting hematopoiesis and oxygen carrying), albumin (reflecting nutrition and liver synthesis), creatinine (reflecting kidney function), cholesterol, and blood glucose (reflecting metabolic function), each corresponding to different key physiological systems in the human body. Changes in the levels of these key indicators directly or indirectly reflect the functional state of the system, and their combined patterns can effectively indicate the cumulative degree of aging or "damage" across multiple systems.

[0034] The system receives multi-source physiological data input from the user. This data is constructed into a time-stamped sequence; for example, physical examination data of the same individual in different years (T1, T2, T3...) are organized chronologically, with each indicator linked to its measurement time point, forming a multi-dimensional time series. Using a pre-trained deep neural network model and an attention mechanism, the system automatically analyzes and captures the complex, non-linear interactions between the indicators, as well as the patterns of change of each indicator over time. Combining the interactions between indicators, the system comprehensively assesses the individual's aging status and outputs their biological age.

[0035] The deep neural network model can include, but is not limited to, attention-based neural network structures or other model structures capable of modeling complex nonlinear relationships. It should be understood that the specific architecture of the model does not constitute a limitation of this invention. Pre-training strategies can employ self-supervised pre-training on a large amount of unlabeled or weakly labeled biomedical data (such as publicly available multi-omics databases). Common methods include: masked prediction, where a portion of the input indicators is randomly masked, allowing the model to predict the masked indicator values ​​based on the context, thereby learning the co-occurrence and dependencies between indicators; and contrastive learning, which involves constructing similar and dissimilar sample pairs and training the model to make the representations of similar samples more similar in the vector space. The purpose of pre-training is to enable the model to extract features from routine clinical data and predict biological age, providing high-quality initialization parameters for predicting biological age based on extracted features, effectively preventing overfitting and improving generalization ability.

[0036] Multi-source physiological data is fed into the model as an input sequence, and a multi-head attention mechanism is used to capture nonlinear interactions. Each "attention head" can be viewed as an independent feature detector. Different heads adaptively focus on different types of relationships between indicators. For example, one head might specifically learn synergistic or antagonistic relationships between metabolic indicators (such as blood glucose and blood lipids). Another head might focus on the association between inflammatory markers and organ function indicators. Through parallel computation and subsequent fusion of multiple heads, the model can characterize complex, high-order nonlinear interaction networks far exceeding simple addition or multiplication. Furthermore, the multi-source physiological data is constructed as a time-labeled sequence for input. The attention mechanism can dynamically measure the importance of measurements at different time points, identifying patterns such as "a certain indicator is rapidly deteriorating in the recent period" or "remaining stable in the long term," thereby capturing the dynamic trajectory of aging.

[0037] After processing through multiple Transformer blocks, the information from all indicators is fused into a comprehensive, high-dimensional individual state representation vector. This vector contains the mined relationships between indicators and temporal patterns. Finally, a regression output layer (typically a fully connected neural network) is connected to map this comprehensive representation onto a specific biological age value. The training objective of the model is to minimize the error between the predicted biological age and the "true" biological age (or aging state) defined based on clinical outcomes or aging phenotypes.

[0038] The intrinsic ability score can be pre-assessed by professional healthcare personnel using a complete scale (such as SPPB, MMSE, GDS-15, etc.), or users can obtain it themselves through simplified testing combined with the Health Engine IC assessment tool for preliminary judgment. This intrinsic ability score is used as a key input dimension and integrated with the calculated biological age and other physiological indicators (e.g., hemoglobin, creatinine, etc.) to construct a high-dimensional state vector for subsequent analysis.

[0039] In this embodiment, a pre-trained deep neural network (such as Transformer) is used to extract features from low-cost, easily accessible multi-source physiological data and calculate biological age. Combined with intrinsic ability (IC) scores, a high-dimensional state vector is constructed, which realizes low-cost and high-precision quantification of aging degree and greatly improves the accessibility of health assessment.

[0040] In step S120, a robust distance index between the high-dimensional state vector and the healthy baseline distribution is calculated; Based on robust statistical methods, a target data subset is determined from the health dataset; Based on the mean vector and covariance matrix of the target data subset, a healthy baseline distribution is constructed; Calculate the robust distance index between the high-dimensional state vector and the healthy baseline distribution.

[0041] The healthy baseline distribution refers to a high-dimensional multivariate statistical distribution fitted from physiological data of a large-scale healthy population using statistical methods. It can be figuratively called a "healthy structural reference surface" or "healthy envelope," similar to the stress-strain distribution model of a bridge under normal conditions in engineering. It serves as the "gold standard" or reference system for assessing individual health. Robust distance can be calculated using multidimensional correlation measures, including but not limited to Mahalanobis distance or other statistical distance measures. It is a statistical index that measures the distance from a point to a distribution, considering the correlation between various dimensions. The robust distance index can be quantified as a numerical measure of "heterostatic load" or "cumulative damage." The robust statistical methods include, but are not limited to, the minimum covariance determinant (MCD) method or other equivalent robust estimation methods. Their goal is to find a subset from a dataset containing "outliers" (abnormal values) such that the determinant of the covariance matrix of that subset is minimized. The mean vector and covariance matrix calculated for this subset are insensitive to outliers, thus representing a "robust" estimate. A health dataset can refer to a large-scale, screened dataset of "normalized state vectors" of healthy individuals (e.g., individuals clinically assessed as having no major chronic diseases and in good functional status). It should be understood that the robust statistical methods described above are merely exemplary implementations, and this invention is not limited to any specific statistical method or distance calculation method.

[0042] Specifically, a large-scale, screened dataset of "normalized state vectors" is collected from a healthy population (e.g., individuals clinically assessed as having no major chronic diseases and in good functional condition). Each vector represents a healthy individual's multidimensional physiological state. This "normalized state vector" dataset is then fed into a robust statistical method, such as the minimum covariance determinant estimation algorithm. The core of this algorithm is to find a subset of data points that minimizes the determinant of the sample covariance matrix—the target data subset, for example, a subset containing 75% of the data points. Data points within this subset are considered the core of the data, representing the most "typical" health pattern, while unselected subsets are considered outliers. Based on this found "most compact" and "most typical" target data subset, its mean vector and covariance matrix are calculated. These two results constitute the "robust mean vector." "and robust covariance matrix" This led to a conclusion based on... Centered on, with A robust multivariate Gaussian distribution model describing the shape, known as the "healthy baseline distribution," will serve as the benchmark for comparing the states of all subsequent individuals.

[0043] Use the robust mean vector obtained during the training phase. and robust covariance matrix Calculate the Mahalanobis distance D from the high-dimensional state vector x to the reference surface. M The formula is as follows: ; Among them, D M represents the calculated robust Mahalanobis distance, i.e., the quantified heterostable load. x represents the high-dimensional normalized state vector of the individual to be evaluated. This represents the robust mean vector of the healthy baseline distribution estimated using the minimum covariance determinant (MCD) method. The inverse of the robust covariance matrix representing the healthy baseline distribution is used to "whiten" the data, eliminating correlations between different dimensions and standardizing variance. (x- This indicates the degree to which an individual's state deviates from the center of health.

[0044] The formula above calculates D M The robust distance index is a scalar value that can be directly interpreted as a quantified value of heterostable loads. (D) M The larger the value, the greater the deviation of the high-dimensional normalized state vector x of the individual being evaluated from the "healthy baseline distribution" after considering the overall variation patterns of the healthy population. The system will continuously monitor changes in this value; if D... M If the damage continues to increase over time, it indicates that the individual's "cumulative physiological damage" is increasing, meaning that the system is deviating from homeostasis and developing towards a weakened state.

[0045] In this embodiment, a robust statistical method is used to establish a healthy baseline distribution and calculate the robust distance (Maharanobis distance) between the individual's state vector and the baseline. This method can resist the interference of outliers in the data and accurately quantify the degree to which the individual's overall physiological system deviates from the healthy homeostasis (i.e., the abnormal homeostasis load), providing a reliable and robust quantitative indicator for state determination.

[0046] In step S130, based on the robust distance index, the individual's current physical condition is determined, including: Based on the aforementioned health dataset, high-risk thresholds and recovery thresholds are adaptively determined. If the robustness distance index is higher than the high-risk threshold and the duration reaches the first preset residence time length, the individual's state is determined to be a weakened constitution.

[0047] If the robustness distance index is lower than the recovery threshold and the duration reaches the second preset residence time length, the individual's status is determined to be healthy.

[0048] The dual-threshold Schmitt trigger model can refer to a state-determining logic model with hysteresis characteristics, which sets two different thresholds, one for high risk and one for low risk. A recovery threshold The system uses dwell time conditions to simulate a judgment process with "inertia," ensuring stable state transitions and preventing repeated and frequent state reversals near the critical point caused by short-term fluctuations or noise in the input signal. Physical condition refers to the category of health status determined by the system for an individual, specifically two types: Fit (healthy / robust) and Frail (weak).

[0049] High risk threshold ( This can refer to the trigger boundary when the state changes from Fit to Frail. When An individual is considered to have entered a weakened state when their risk level remains consistently above the high-risk threshold and the required residence time is met. Recovery threshold ( This can refer to the trigger boundary for the state to switch from Frail back to Fit. When An individual is considered to have recovered to a healthy state when their condition remains below the recovery threshold and the required residence time is met. > The interval between the two is called the "hysteresis interval" or "dead zone," which is a buffer zone for maintaining a stable state. The first preset dwell time length can refer to the time required to trigger a switch from Fit to Frail. continuously higher than The shortest duration. Used to confirm the persistence of deviations from a healthy state trend and filter out temporary physiological fluctuations. The second preset dwell time length can refer to the minimum duration required to trigger a switch from Frail back to Fit. continuously below The shortest duration. Used to confirm the stability and reliability of the recovery trend, preventing premature diagnosis of recovery.

[0050] Specifically, the system continuously calculates and monitors robust distance indicators that represent an individual's "abnormal steady-state load" or "cumulative damage". . This is a quantified value of the degree to which the state vector deviates from the healthy baseline, and also the direct input signal of the state determination model (i.e., the double-threshold Schmitt trigger model with hysteresis characteristics). The Schmitt trigger model is used, and a double-boundary hysteresis mechanism is applied for determination. Internally, the system maintains a current system state variable (initially Fit). The determination logic is as follows: If the current physical condition is Fit (healthy): when Persistently above the high-risk threshold The system state will only be switched to Frail when the duration of the "above" state reaches the first preset dwell time length.

[0051] like It was only temporarily higher than The temperature dropped again, and since the "duration" condition was not met, the systemic state would not switch. By setting a first preset residence time length, misjudgments were avoided due to temporary drifts in physiological indicators caused by acute infections, short-term sleep deprivation, etc.

[0052] If the current state is Frail (weak): when Persistently below the recovery threshold Furthermore, the individual's physical condition will only be switched back to Fit if the duration of the "below" state reaches the second preset dwell time. By setting the second preset dwell time, it is ensured that the individual needs to undergo a stable and continuous recovery period before being re-identified as healthy, preventing the physical condition from fluctuating near the critical point.

[0053] High-risk and recovery thresholds can be adaptively determined based on a health dataset. The adaptive determination process specifically involves: based on the "health baseline distribution" determined in the above steps (composed of a robust mean vector)... and robust covariance matrix (Definition) Calculates the robust distance metric for each healthy individual in the dataset. Analyze historical data of a large-scale healthy population (i.e., a health dataset) for all healthy individuals. The distribution. Initialize based on distribution characteristics (such as percentiles, mean, and standard deviation). and .For example, It can be set for healthy people Near the 95th percentile of the distribution, The threshold can be set near the 75th percentile. Furthermore, during system operation, new, labeled data can be continuously used to optimize and fine-tune the two thresholds through optimization algorithms (such as grid search, Bayesian optimization, etc.) to achieve the best balance between sensitivity and specificity. It should be understood that the interval division method is not limited to a specific quantile or fixed threshold. The determination of the threshold is not limited to specific values, quantiles, or statistical methods; it can be obtained through any form of adaptive function, learning model, or strategy configuration, and does not constitute a limitation of the present invention.

[0054] In this embodiment, a dual-threshold Schmitt trigger model with hysteresis characteristics is adopted to determine the individual's physical condition (healthy / weak) based on the robust distance index. By setting the conditions of "high-risk threshold", "recovery threshold" and "staying time", the frequent switching or misjudgment of the system state caused by short-term normal fluctuations of physiological indicators (such as temporary discomfort) is effectively avoided, making the state classification results more stable and reliable.

[0055] In step S140, based on the individual's current physical condition, candidate intervention strategies are generated, including: Retrieve target intervention objects from the evidence database that correspond to the individual's current physical condition; The current recommendation weight of each target intervention object is calculated by comprehensively weighting the evidence level, quality score and timeliness weight corresponding to each target intervention object, wherein the timeliness weight is calculated by the half-life decay function; Based on the recommendation weights, the target intervention objects are sorted from largest to smallest, and a preset number of first target intervention objects are selected from the sorted list; Based on each of the first target intervention subjects, a list of candidate intervention strategies is generated according to clinical rules.

[0056] The evidence base refers to a structured medical knowledge database that stores systematically organized interventions (drugs, exercise, nutrition, etc.) and their corresponding medical research evidence. Each record includes metadata such as the intervention target, effectiveness conclusion, evidence level, quality score, and publication date. The target intervention target refers to a specific intervention entity retrieved from the evidence base that matches the user's current physical condition (e.g., "Frail"). For example, "150 minutes of moderate-intensity aerobic exercise per week" or "daily supplementation with 800 IU of vitamin D" could be an intervention target. The evidence level refers to a scale that measures the source of medical evidence and the scientific rigor of the research design. It typically follows a standard system (such as the Oxford Centre for Evidence-Based Medicine grading), with randomized controlled trials (RCTs) having a higher level than observational studies, which in turn have a higher level than expert opinions. The quality score refers to a score that further evaluates the methodological quality of an individual study within the same evidence level. For example, a double-blind, large-sample, low-dropout RCT would have a higher score than a methodologically flawed RCT.

[0057] Timeliness weight can refer to a weighting factor dynamically calculated based on the evidence's release date, used to quantify the "freshness" of the evidence. It is calculated using a half-life decay function, where t is the time since the evidence's release; the more recent the release, the higher the timeliness weight. Recommendation weight can refer to the final comprehensive score calculated for each target intervention object. It is derived by weighting the evidence level, quality score, and timeliness weight according to a preset ratio. The half-life decay function can refer to a mathematical model, the formula of which is... This system dynamically adjusts the recommendation weights based on the time of evidence release, simulating the updating and iteration of medical knowledge. The ranking list can refer to a sequential list formed by arranging all target intervention subjects in descending order of their recommendation weights. The preset quantity can refer to a system-configured constant N, such as 6, used to select the top N intervention subjects from the ranking list as the core candidate set, controlling the complexity and focus of the strategy. The primary target intervention subjects can refer to the top N intervention subjects selected from the ranking list according to the preset quantity. They are the "raw materials" for generating the final strategy. Clinical rules can refer to a set of coded medical practice guidelines, including the synergistic effects between interventions (e.g., exercise and nutritional supplementation can be combined), contraindications and conflicts (e.g., two drugs cannot be taken together), and the order of application. The candidate intervention strategy list can refer to a list containing one or more complete intervention protocols. Each protocol is composed of several primary target intervention subjects combined according to clinical rules, and may include preferred strategies, alternative strategies, etc.

[0058] Specifically, the systemic condition is determined, such as "Frail". Using this condition as a keyword, the evidence database is searched for all intervention records where the "Indications" or "Recommended Population" fields contain "Frail" or "Frail". The retrieved intervention records are the current target intervention subjects. For example, evidence records such as "Comprehensive Resistance Training", "Protein Energy Supplementation", and "Vitamin D Supplementation" may be retrieved.

[0059] For each piece of evidence corresponding to a target intervention object, obtain its publication time (publication date). Calculate the interval t (usually in years) from the publication time to the current time. Assign an initial timeliness weight W to this evidence. init (Usually 1 or a base score) and the interval t are substituted into the half-life decay function: ; Where W(t) represents the timeliness weight of the evidence at the current moment. init represents the initial weight of the evidence. is the decay coefficient, λ, which determines the rate at which the weight decays over time. t is the time since the evidence was published. The larger t is, the smaller W(t) is.

[0060] Extract the evidence level score, quality score, and timeliness weight calculated in the previous step for each target intervention subject from the evidence database. Three preset weight coefficients α, β, and γ represent the degree of importance attached to evidence level, quality, and timeliness, respectively (α+β+γ=1).

[0061] Calculate the recommendation weight using the following formula: ; Where R is the current final priority score of the target intervention subject, which integrates scientific validity, quality, and timeliness. α, β, and γ are pre-set weighting coefficients that measure the relative importance of evidence level, quality score, and timeliness weight in the comprehensive evaluation, typically satisfying α + β + γ = 1. S g S represents the level of evidence score. q The weighting is represented by the quality score. W(t) represents the time-dependent weight calculated from the half-life decay function. The above weighting calculation method is an exemplary implementation. In practical applications, nonlinear mapping, piecewise functions, or black-box models can also be used for weight fusion, which does not constitute a limitation of the present invention.

[0062] Sort all target intervention objects in descending order of their recommendation weight R. This forms a sorted list, with the intervention object ranked first having the highest recommendation weight. Read the preset number N (e.g., N=5) from the configuration. Starting from the top of the sorted list, select the first N intervention objects sequentially. These N selected objects constitute the primary target intervention objects, representing the core set of measures with the highest recommendation value.

[0063] The selected N primary intervention subjects are input into the clinical rule engine. The engine processes these subjects based on coded knowledge, including conflict detection and elimination, synergistic combination, ranking and stratification, and generating alternative solutions. Conflict detection and elimination checks for mutually exclusive interventions among the N primary intervention subjects, such as contraindications to drug interactions. If such interactions exist, the intervention with the lower weight is removed, or a warning is generated. Synergistic combination identifies combinations that can synergistically enhance each other, such as resistance training and protein supplementation, and prioritizes packaging them into the same sub-strategy. Ranking and stratification determine the order of interventions based on clinical practice, such as correcting nutrition before starting exercise. The engine typically assembles more than one strategy option; for example, the "preferred strategy" can include combinations of the first three primary intervention subjects, while the "alternative strategy" can include combinations of the fourth to sixth primary intervention subjects for users with contraindications.

[0064] Ultimately, a list of candidate intervention strategies is generated. This list typically contains multiple complete and actionable intervention options for physicians or users to choose from. Each option is composed of several primary target intervention subjects rationally arranged according to clinical rules, and includes corresponding explanations and precautions.

[0065] In this embodiment, intervention measures matching the current physical condition are retrieved from the evidence database, and a recommendation weight is calculated by combining the evidence level, quality score, and timeliness weight with half-life decay, thereby generating a list of candidate intervention strategies. This ensures the scientific nature and timeliness of the intervention recommendations and allows for dynamic adjustment based on the latest medical evidence.

[0066] In step S150, based on the high-dimensional state vector and the candidate intervention strategy list, the cognitive uncertainty score and the random uncertainty score of each candidate intervention strategy in the candidate intervention strategy list are calculated.

[0067] Deep ensemble models can refer to machine learning models used to quantify prediction uncertainty. These models consist of multiple structurally identical but independently trained deep learning sub-models. Their core function is to measure the reliability of the model's predictions by analyzing the differences in the prediction results of each sub-model. Cognitive uncertainty score is a quantified value representing the degree of uncertainty in prediction results caused by the deep ensemble model's own insufficient knowledge, such as encountering user states and policy combinations not fully covered in the training data. A higher score indicates that the model is more "uncertain" or "unaware." Random uncertainty score is a quantified value representing the inherent, unavoidable uncertainty in predictions caused by random noise in the data, such as measurement errors and daily fluctuations in individual physiology.

[0068] Specifically, a high-dimensional state vector representing the user's state is paired with a candidate intervention strategy to be evaluated, forming a complete "user state, intervention strategy" decision instance. This paired instance is then simultaneously input into each sub-model of the deep ensemble model. Each sub-model works independently, outputting its prediction of "whether the candidate intervention strategy is applicable to the individual and what its effect is," typically a probability value. Cognitive uncertainty can be measured by analyzing the differences (variance) between the predictions of these sub-models. For example, there are M sub-models, each outputting a probability value. Cognitive uncertainty score is often measured by calculating the variance of probability values: If all sub-models give highly consistent predictions (small variance), it indicates that the model is "very confident" in applying the intervention strategy to the individual's decision-making scenario, and the cognitive uncertainty is low. If the sub-models differ greatly (large variance), for example, giving vastly different prediction results, it indicates that the model "knows very little" about applying the intervention strategy to the individual's decision-making scenario or has encountered a situation not covered by training data, and the cognitive uncertainty is high.

[0069] In a deep ensemble model, each sub-model can output a probability distribution (such as a Gaussian distribution), where the mean represents the prediction performance and the variance represents the random uncertainty perceived by that sub-model. The final random uncertainty can be the average of the variances of all sub-model outputs.

[0070] Based on the above calculations, the cognitive uncertainty score can be represented as the variance among the prediction results of each sub-model. The random uncertainty score can be represented as the average of the prediction variances of each sub-model. Furthermore, the calculation method for the uncertainty score is not limited to variance calculation; it can also be achieved through probability distribution modeling, Bayesian inference, or other equivalent methods. This invention does not limit the specific methods used.

[0071] Iterate through each strategy in the candidate intervention strategy list and calculate the corresponding pairwise uncertainty scores for each of them.

[0072] In this embodiment, a cognitive uncertainty score (model knowledge blind spot) and a random uncertainty score (inherent noise in the data) are calculated for each candidate intervention strategy using a deep ensemble model. This quantitatively evaluates the reliability of each recommendation decision and provides a key basis for subsequent security review.

[0073] In step S160, based on the high-dimensional state vector, the cognitive uncertainty score and the accidental uncertainty score corresponding to each of the candidate intervention strategies, a multi-layer security review mechanism is used to review each of the candidate intervention strategies in the candidate intervention strategy list. If the review is passed, at least one intervention strategy is generated.

[0074] The multi-layered security review mechanism refers to a set of multiple parallel-operating hard security rules integrated by a deterministic security kernel module; it is a comprehensive security checkpoint used to make the final decision on preliminary decisions. The multi-layered security review mechanism includes: a red-line threshold determination mechanism, an uncertainty rejection mechanism, a human-in-the-loop intervention mechanism, and a forced blocking mechanism. The red-line threshold determination mechanism refers to setting an absolute prohibition safety boundary (red line) for key physiological indicators (such as heart rate, blood potassium concentration, etc.). Once a state vector value in the user's current high-dimensional state vector touches the red line, all candidate intervention strategies will be rejected; this can be set as the second priority rule. The uncertainty rejection mechanism refers to setting a maximum tolerance threshold for cognitive uncertainty scores. When the cognitive uncertainty score of a candidate intervention strategy exceeds this maximum tolerance threshold, a forced rejection is triggered, prohibiting the recommendation of that candidate intervention strategy; this can be set as the third priority rule. The human-in-the-loop intervention mechanism refers to a rule that forcibly suspends automated decision-making and transfers control to a human physician for final judgment when facing moderate-risk, ambiguous situations, such as when the uncertainty score is in the critical zone or the intervention strategy involves significant adjustments. For example, if cognitive uncertainty is low but accidental uncertainty is high, it indicates that the model "knows" the situation, but the predicted outcome inherently has a large range of fluctuations. The review mechanism may not directly reject the answer, but it may trigger a "human-in-the-loop intervention," transferring the final decision to a human doctor; this could be set as the fourth priority rule. A mandatory blocking mechanism can refer to a highest-priority safety rule used to respond to extreme danger signals, such as receiving an emergency circuit breaker order from a regulatory agency or the system detecting a fatal error. Once triggered, it will unconditionally terminate the entire decision-making process. These safety mechanisms are centrally scheduled through a unified adjudication interface, and after performing parallel constraint judgments on candidate strategies, a unified adjudication result is formed.

[0075] Specifically, the list of candidate intervention strategies to be reviewed, along with the cognitive uncertainty score and accidental uncertainty score corresponding to each strategy, and a high-dimensional state vector representing the user's current state, are input into a multi-layered security review mechanism. This multi-layered security review mechanism is centrally scheduled through a unified decision-making interface, performing parallel constraint judgments on the candidate intervention strategies and outputting a unified decision based on priority rules. Once activated, the security review mechanism's internal red-line threshold judgment mechanism, uncertainty rejection mechanism, human-in-the-loop intervention mechanism, and forced blocking mechanism are initiated in parallel, performing independent and synchronous security filtering on each strategy in the list. For each candidate intervention strategy in the list, the review mechanism performs the following parallel judgments: A mandatory blocking mechanism checks for the presence of high-priority blocking signals, such as external circuit breaker commands or fatal system errors. If any are detected, the entire review process immediately and unconditionally terminates all intervention strategies. External circuit breaker commands can refer to official safety announcements issued by authoritative regulatory agencies such as the National Medical Products Administration (NMPA), requiring the immediate cessation of use or high-level attention to a particular drug, medical device, or health intervention. For example, an emergency safety announcement: a notice issued by the NMPA for a specific medical device or drug, requiring immediate modification of instructions for use, suspension of sales, or recall. Safety information updates: regulatory agencies, based on the latest research or adverse event monitoring, clearly indicate that a certain treatment method is contraindicated in a specific population. Fatal system errors can refer to internal, serious technical failures that prevent the system from operating reliably. These may be detected as runtime errors in internal algorithms / models, such as failure to load the core prediction model or an undetectable anomaly in the inference process leading to a crash. Critical data pipeline failures, such as interruptions in real-time physiological data streams or permanent loss of connection to the hospital laboratory information system (LIS), resulting in the system losing core input. Severe breaches of data consistency or integrity, such as the detection of accidental tampering with critical database tables or checksum anomalies in core calculation results. Or, there may be warnings about the health of hardware and infrastructure, such as server critical hardware (e.g., CPU, memory) failure, or the container platform detecting service unavailability.

[0076] The red line threshold determination mechanism reads a high-dimensional state vector representing the user's real-time status and compares predefined key indicators (e.g., heart rate and serum potassium concentration) in this vector with preset "red line" thresholds. Key indicators refer to physiological parameters that have clear consensus in clinical medicine, and whose values ​​exceeding a certain limit directly constitute an imminent life-threatening danger or extremely high health risk. These can be indicators with recognized clinical "critical value" ranges, such as serum potassium concentration, blood glucose (extremely high or low), blood oxygen saturation, and heart rate. They can also be indicators that directly reflect core vital signs or organ function, such as blood pressure, respiratory rate, and creatinine / estimated glomerular filtration rate (eGFR) reflecting kidney function. The preset "red line" thresholds are derived from critical value ranges published by authoritative domestic and international medical associations (such as the Chinese Medical Association and the American Heart Association) in clinical guidelines, expert consensus, or textbooks. For example, the "red line" for serum potassium concentration might be directly set as the clinically recognized lower limit (e.g., <2.8 mmol / L) and upper limit (e.g., >6.0 mmol / L) for reporting critical values. If any key indicator reaches or exceeds the red line, all candidate intervention strategies will be rejected outright.

[0077] The uncertainty-based rejection mechanism reads the cognitive uncertainty score of candidate strategies and determines whether it exceeds (or is greater than) the rejection threshold. If so, the mechanism outputs a "rejection status," such as a DATA_VOID signal, prohibits the generation of any intervention suggestions, and prompts for manual evaluation.

[0078] Human-in-the-Loop Intervention Mechanism: A low safety threshold is set for the cognitive uncertainty score, which is less than the refusal threshold. For cases where the safety threshold ≤ cognitive uncertainty score ≤ refusal threshold, and the random uncertainty is extremely high (e.g., greater than a pre-set acceptable threshold), the human-in-the-loop mechanism is triggered. The strategy will not be automatically generated or rejected; instead, it will be suspended, triggering a "human-in-the-loop" state, awaiting physician intervention and decision. This [safety threshold, refusal threshold] is the "critical zone." And / or, if the candidate intervention strategy involves significant adjustments, such as recommendations involving prescription drug changes, surgery, or invasive procedures, a human-in-the-loop state is mandatory, awaiting physician intervention and decision. Examples include, according to a pre-set "significant adjustments" rule: changes in the type of intervention, i.e., introducing a new intervention category from none (e.g., the user currently has no medication intervention, and the new strategy introduces prescription drugs for the first time); or a significant jump in intervention intensity, e.g., an exercise recommendation changing from "walking" to "high-intensity interval training"; nutritional supplementation changing from dietary adjustments to requiring tube feeding or parenteral nutrition. Alternatively, the candidate intervention strategy may include any invasive examinations or treatment recommendations. Or, considering the user's current "physical condition" (e.g., already in a frail state), any new intervention may be considered a "major change" by the system because frail individuals have poor physiological reserves and low tolerance for change. Or, the strategy may involve multiple adjustments simultaneously (e.g., simultaneous changes to medication, exercise, and diet), and these adjustments need to be implemented in a coordinated manner, potentially being judged as a major, complex clinical protocol change. Or, by maintaining an internal list, certain interventions (e.g., the use of glucocorticoids, anticoagulants, chemotherapy drugs, etc.) may be marked as "high-risk" or "major." Any strategy containing such measures will be automatically judged as involving a major change. If one or more of the above rules are triggered, it will be marked as "involving a major change," potentially triggering the "human-in-the-loop intervention mechanism," requiring the physician to conduct a thorough review and confirmation of this major change.

[0079] A strategy is considered "approved" only if it does not trigger the "mandatory blocking," "red line judgment," or "uncertainty refusal" mechanisms, nor the "human-in-the-loop" mechanism, or if it is approved by a physician after triggering such a mechanism. All "approved" strategies are generated as formal target intervention strategies (i.e., at least one). Target intervention strategies are given structured descriptions and natural language explanations, such as, "It is recommended to increase sun exposure to supplement vitamin D because serum vitamin D levels are low and there is strong evidence of a link between low serum vitamin D levels and improved sarcopenia and fall risk." These strategies are recorded in the audit log and delivered to the physician as final recommendation support, but do not constitute a diagnosis.

[0080] Furthermore, if more than one target intervention strategy is generated, the strategies are arranged in descending order of cognitive uncertainty score before being delivered to the doctor. For example, the arranged strategies can be displayed on a visualization interface, allowing the doctor to view them via touch or mouse.

[0081] In addition, the entire decision-making process, from raw data, status determination, candidate strategies, uncertainty assessment results to final recommendations, will be recorded in an immutable log. This complies with the stringent traceability requirements of medical device regulations (such as NMPA / IEC 62304). In case of problems, the entire decision-making chain can be reviewed to clarify responsibilities.

[0082] In this embodiment, a multi-layered security review kernel, integrating mechanisms such as red-line threshold determination, uncertainty rejection, human-in-the-loop intervention, and forced blocking, makes a final decision on candidate strategies. Its beneficial effect is that it constructs a security barrier with the highest priority, which can proactively identify high-risk situations, reject suggestions that exceed the model's cognitive range, and transfer them to human judgment when necessary. This fundamentally prevents the output of unsafe suggestions and ensures the inherent safety and compliance of the system in medical applications.

[0083] Furthermore, the target intervention strategy of this invention is only used as decision support information output and does not constitute a diagnostic or treatment instruction. Example 3

[0084] Figure 2 This is a framework diagram of a cross-age health status assessment and anti-aging decision support system based on a deterministic security kernel and probabilistic reasoning, provided in Embodiment 3 of the present invention. Figure 2 As shown, the system includes: The first generation module 210 is used to receive multi-source physiological data, calculate biological age, and generate a high-dimensional state vector containing biological age and intrinsic ability indicators. The first calculation module 220 is used to calculate the robust distance index between the high-dimensional state vector and the healthy baseline distribution; The determination module 230 is used to determine the current physical condition of an individual based on the robust distance index. The second generation module 240 is used to generate candidate intervention strategies based on the individual's current physical condition; The second calculation module 250 is used to calculate the cognitive uncertainty score and the random uncertainty score of each candidate intervention strategy in the candidate intervention strategy list based on the high-dimensional state vector and the candidate intervention strategy list. The third generation module 260 is used to review each of the candidate intervention strategies in the candidate intervention strategy list based on the high-dimensional state vector, the cognitive uncertainty score and the accidental uncertainty score corresponding to each of the candidate intervention strategies, through a multi-security review mechanism. If the review is passed, at least one target intervention strategy is generated.

[0085] Optionally, the first generation module 210 is specifically used for: It receives physiological data from multiple sources and outputs biological age through a pre-trained deep neural network model. The biological age is integrated with pre-obtained intrinsic ability indicators to construct a high-dimensional state vector.

[0086] Optionally, the first computing module 220 is specifically used for: Based on robust statistical methods, a target data subset is determined from the health dataset; Based on the mean vector and covariance matrix of the target data subset, a healthy baseline distribution is constructed; Calculate the robust distance index between the high-dimensional state vector and the healthy baseline distribution.

[0087] Optionally, the determination module 230 is specifically used for: Based on the aforementioned health dataset, high-risk thresholds and recovery thresholds are adaptively determined. If the robustness distance index is higher than the high-risk threshold and the duration reaches the first preset residence time length, the individual's state is determined to be a weakened constitution.

[0088] If the robustness distance index is lower than the recovery threshold and the duration reaches the second preset residence time length, the individual's status is determined to be healthy.

[0089] Optionally, the second generation module 240 is specifically used for: Retrieve target intervention objects from the evidence database that correspond to the individual's current physical condition; The current recommendation weight of each target intervention object is calculated by comprehensively weighting the evidence level, quality score and timeliness weight corresponding to each target intervention object, wherein the timeliness weight is calculated by the half-life decay function; Based on the recommendation weights, the target intervention objects are sorted from largest to smallest, and a preset number of first target intervention objects are selected from the sorted list; Based on each of the first target intervention subjects, a list of candidate intervention strategies is generated according to clinical rules.

[0090] Optionally, the multiple security review mechanisms include: a red line threshold determination mechanism, an uncertainty rejection mechanism, a human-in-the-loop intervention mechanism, and a forced blocking mechanism.

[0091] The cross-age health status assessment and anti-aging decision support system based on deterministic security kernel and probabilistic reasoning provided in the embodiments of the present invention can execute the cross-age health status assessment and anti-aging decision support method based on deterministic security kernel and probabilistic reasoning provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method. Example 4

[0092] Figure 3 A schematic diagram of an electronic device that can be used to implement embodiments of the present invention is shown. The electronic device 10 is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (e.g., helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0093] like Figure 3 As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12, a random access memory (RAM) 13, etc., which is communicatively connected to the at least one processor 11. The memory stores a computer program that can be executed by the at least one processor 11, and the computer program is executed by the at least one processor 11 to enable the at least one processor 11 to perform the method provided by the present invention.

[0094] The processor 11 can perform various appropriate actions and processes based on a computer program stored in the read-only memory (ROM) 12 or a computer program loaded from the storage unit 18 into the random access memory (RAM) 13. The RAM 13 can also store various programs and data required for the operation of the electronic device 10. The processor 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0095] Multiple components in electronic device 10 are connected to I / O interface 15, including: input unit 16, such as keyboard, mouse, etc.; output unit 17, such as various types of displays, speakers, etc.; storage unit 18, such as disk, optical disk, etc.; and communication unit 19, such as network card, modem, wireless transceiver, etc. Communication unit 19 allows electronic device 10 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0096] Processor 11 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, digital signal processors (DSPs), and any suitable processor, controller, microcontroller, etc. Processor 11 performs the various methods and processes described above, such as a cross-age health status assessment and anti-aging decision support method based on a deterministic secure kernel and probabilistic reasoning.

[0097] In some embodiments, a method for cross-age health status assessment and anti-aging decision support based on a deterministic security kernel and probabilistic reasoning can be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed on electronic device 10 via ROM 12 and / or communication unit 19. When the computer program is loaded into RAM 13 and executed by processor 11, one or more steps of the method for cross-age health status assessment and anti-aging decision support based on a deterministic security kernel and probabilistic reasoning described above can be performed. Alternatively, in other embodiments, processor 11 can be configured by any other suitable means (e.g., by means of firmware) to perform a method for cross-age health status assessment and anti-aging decision support based on a deterministic security kernel and probabilistic reasoning.

[0098] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard parts (ASSPs), systems-on-chip (SoCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0099] Computer programs used to implement the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0100] In the context of this invention, a computer-readable storage medium stores computer instructions that, when executed by a processor, implement the cross-age health status assessment and anti-aging decision support method based on a deterministic security kernel and probabilistic reasoning provided by this invention. The computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. The computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. Alternatively, the computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0101] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device for displaying information to the user (e.g., a cathode ray tube (CRT) or liquid crystal display (LCD) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0102] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or middleware components (e.g., application servers), or frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0103] A computing system can include clients and servers. Clients and servers are generally geographically separated and typically interact via communication networks. The client-server relationship is established by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system. It addresses the shortcomings of traditional physical hosts and Virtual Private Server (VPS) services, such as high management difficulty and weak business scalability.

[0104] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.

[0105] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A cross-age health status assessment and anti-aging decision support method based on deterministic security kernel and probabilistic reasoning, characterized in that, include: Receive multi-source physiological data to calculate biological age and generate a high-dimensional state vector containing biological age and intrinsic ability indicators; Calculate the robust distance index between the high-dimensional state vector and the healthy baseline distribution; Based on the robust distance index, determine the individual's current physical condition; Based on the individual's current physical condition, candidate intervention strategies are generated; Based on the high-dimensional state vector and the candidate intervention strategy list, calculate the cognitive uncertainty score and the random uncertainty score of each candidate intervention strategy in the candidate intervention strategy list; Based on the high-dimensional state vector, the cognitive uncertainty score and the accidental uncertainty score corresponding to each candidate intervention strategy, a multi-layer security review mechanism is used to review each candidate intervention strategy in the candidate intervention strategy list. If the review is passed, at least one target intervention strategy is generated.

2. The method of claim 1, wherein, The process of receiving multi-source physiological data to calculate biological age and generating a high-dimensional state vector containing biological age and intrinsic ability indicators includes: It receives physiological data from multiple sources and outputs biological age through a pre-trained deep neural network model. The biological age is integrated with pre-obtained intrinsic ability indicators to construct a high-dimensional state vector.

3. The method of claim 1, wherein, Calculating the robust distance index between the high-dimensional state vector and the healthy baseline distribution includes: Based on robust statistical methods, a target data subset is determined from the health dataset; Based on the mean vector and covariance matrix of the target data subset, a healthy baseline distribution is constructed; Calculate the robust distance index between the high-dimensional state vector and the healthy baseline distribution.

4. The method of claim 1, wherein, The determination of an individual's current physical condition based on the robust distance index includes: Based on the aforementioned health dataset, high-risk thresholds and recovery thresholds are adaptively determined. If the robustness distance index is higher than the high-risk threshold and the duration reaches the first preset residence time length, then the individual's state is determined to be a weakened constitution. If the robustness distance index is lower than the recovery threshold and the duration reaches the second preset residence time length, the individual's status is determined to be healthy.

5. The method according to claim 1, characterized in that, The generation of candidate intervention strategies based on the individual's current physical condition includes: Retrieve target intervention objects from the evidence database that correspond to the individual's current physical condition; The current recommendation weight of each target intervention object is calculated by comprehensively weighting the evidence level, quality score and timeliness weight corresponding to each target intervention object, wherein the timeliness weight is calculated by the half-life decay function; Based on the recommendation weights, the target intervention objects are sorted from largest to smallest, and a preset number of first target intervention objects are selected from the sorted list; Based on each of the first target intervention subjects, a list of candidate intervention strategies is generated according to clinical rules.

6. The method according to claim 1, characterized in that, The multiple security review mechanisms include: a red line threshold determination mechanism, an uncertainty rejection mechanism, a human-in-the-loop intervention mechanism, and a forced blocking mechanism.

7. A cross-age health status assessment and anti-aging decision support system based on a deterministic security kernel and probabilistic reasoning, characterized in that, The system is used to execute the cross-age health status assessment and anti-aging decision support method based on deterministic security kernel and probabilistic reasoning as described in any one of claims 1-6, including: The first generation module receives multi-source physiological data, calculates biological age, and generates a high-dimensional state vector containing biological age and intrinsic ability indicators. The first calculation module is used to calculate the robust distance index between the high-dimensional state vector and the healthy baseline distribution; The determination module is used to determine the current physical condition of an individual based on the robust distance index. The second generation module is used to generate candidate intervention strategies based on the individual's current physical condition. The second calculation module is used to calculate the cognitive uncertainty score and the random uncertainty score of each candidate intervention strategy in the candidate intervention strategy list based on the high-dimensional state vector and the candidate intervention strategy list. The third generation module is used to review each of the candidate intervention strategies in the candidate intervention strategy list based on the high-dimensional state vector, the cognitive uncertainty score and the accidental uncertainty score corresponding to each of the candidate intervention strategies, through a multi-security review mechanism. If the review is passed, at least one target intervention strategy is generated.

8. The system according to claim 7, characterized in that, The first generation module is specifically used for: It receives physiological data from multiple sources and outputs biological age through a pre-trained deep neural network model. The biological age is integrated with pre-obtained intrinsic ability indicators to construct a high-dimensional state vector.

9. An electronic device, characterized in that, include: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to execute the cross-age health status assessment and anti-aging decision support method based on deterministic security kernel and probabilistic reasoning as described in any one of claims 1-6.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that cause a processor to execute the cross-age health status assessment and anti-aging decision support method based on a deterministic secure kernel and probabilistic reasoning as described in any one of claims 1-6.

Images